Risk Analysis and Management in Rainfed Rice Systems

S. Pandey, B.C. Barah, R.A. Villano, and S. Pal, Editors

INTERNATIONAL RICE RESEARCH INSTITUTE

IRRI

Contents
Introduction Units of measurement Risk and rainfed rice: some conceptual and methodological issues S. Pandey Risk and rainfed rice in India: an overview C. Ramasamy and K. Uma Decomposition of income variability in rainfed areas: the case of rice in eastern India B.C. Barah Labor use and employment pattern in rainfed rice-producing states of India G.K. Chadha Crop insurance: a policy perspective P.K. Mishra The nature and causes of changes in variability of rice production in eastern India: a district-level analysis S. Pandey and S. Pal Growth and variability in agriculture revisited: district-level evidence of rice production in eastern India S. Pal, S. Pandey, and Abedullah Rainfed rice and risk-coping strategies: some microeconomic evidence from eastern Uttar Pradesh S. Pandex H.N. Singh, and R.A. Villano Risk and the value of rainfall forecast for rainfed rice in the Philippines Abedullah and S. Pandey Characterizing risk and strategies for managing risk in flood-prone rice cultivation in Assam B.C. Bhowmick, S. Pandey, R.A. Villano, and J.K. Gogoi Risk and its management in the rainfed rice ecosystem of Bihar J. Thakur Risk and its management in rainfed rice ecosystem of West Bengal N.K. Saha, S.K. Bardhan Roy, and U.S. Aich Risk and rice production in Orissa, eastern India D. Naik, S. Pandey, D. Behura, and R.A. Villano Risk and rice technology design L.J. Wade Participants 187 195 v xi 1 21 29

39 63

73

97

115 135

143 157 165 173

Introduction

Agricultural production is an inherently risky activity, especially in rainfed areas where farmers have to contend with high levels of climatic risk. Crops may suffer from drought, submergence, or both, often in the same season. Although climatic uncertainty is a major source of risk, other sources of risk such as price uncertainty can amplify the effect of climatic risk. A better understanding of different sources of risk is essential as strategies to overcome risk may vary with sources of risk. Eastern India, which comprises six states, has approximately 20 million ha of rainfed rice. High levels of climatic risk in these areas are often believed to be a major constraint to adoption of improved technologies. In the absence of efficient crop insurance schemes, farmers have developed mechanisms for selfinsurance against these risks. The efficiency costs of these strategies can be substantial, especially for poor farmers who have very few resources to fall back on in the case of crop failure. An in-depth understanding of the role of risk and risk aversion in rainfed rice systems is needed to develop technological and policy interventions that help reduce the cost of dealing with risk. Although the importance of risk has been widely recognized by researchers and policymakers, there is a dearth of quantitative information on the effect of risk on rice production systems in rainfed rice environments. In eastern India, the work has been mainly qualitative and descriptive. Analysts have often attempted to quantify risk at a higher level of aggregation (district or state) rather than at the farm level. Similarly, much of the policy research has focused on the use of price policy

and public distribution systems in improving food security. Overall, the study of risk and risk-diffusing mechanisms in rainfed rice environments has been somewhat fragmented, with researchers from various disciplines studying particular facets of risk. A 3-day workshop on “Risk Analysis and Management in Rainfed Rice Systems” was organized jointly by the International Rice Research Institute (IRRI) and the National Center for Agricultural Economics and Policy Research (NCAP), New Delhi in September 1998. The primary objective of the workshop was to take stock of the work that has been carried out by researchers from various disciplines with a view to synthesizing the findings and identifying priority themes that need further investigation. Researchers from within and outside India presented 17 papers. The present volume consists mainly of the papers presented during the workshop. However, a few new papers not presented during the workshop are also included due to their relevance to the theme of this workshop. The first paper by Pandey provides a general overview of conceptual and empirical issues in risk analysis in the context of rainfed rice farming. The paper provides discussions on the concept and measurement of risk, the effect of risk on technology adoption, and how farmers cope with risk through various self-insurance strategies. The paper also analyzes technology and risk interactions at both the farm and aggregate level. In addition, strategies for developing and disseminating risk-reducing technologies are discussed. Technologies that require a lower degree of prior commitments of inputs, that raise income by improving other

v

components of farming systems (not just that of rice), and that help stabilize area (not just yield) are seen as important for reducing risk. As such technologies will require farmers to use information on weather and other factors that condition crop performance, provision of such information is seen as an important riskreducing strategy. Similarly, dissemination of such risk-reducing technologies that tend to be somewhat information-intensive will require reform of the extension system that is designed, in most countries, for delivering simple technology packages. Ramasamy and Uma provide an overview of risk-related literature in the context of rainfed rice farming in India. While recognizing the importance of climatic risk, they emphasize risks associated with the timely supply of inputs and with prices. The review also indicates that most of the literature on the study of farm level risk in rice production in India is somewhat dated. Perhaps a rapid growth in productivity realized through the adoption of modem varieties and other related technologies in irrigated areas during the Green Revolution period diverted attention from problems facing rainfed areas. However, with the increasing attention now being paid to rainfed rice systems of eastern India, interest in issues of risk and its management has come once again to the forefront. Ramasamy and Uma identify important areas that require increased research attention to develop risk-reducing interventions. The paper by Barah shows that price variability can be an important source of revenue variability of rice farmers. Using the method of variance decomposition, Barah finds that price variability is more important than yield variability in irrigated environments, but that the opposite holds true in the rainfed environment. The implication is that the design of price policy should vary according to the environmental conditions including basic infrastructure. Of course, this poses the challenge of how to design a differential price policy in an environment where spatial economic linkages are growing stronger. Yield-stabilizing and -enhancing measures such as biotic and abiotic stresstolerant varieties and insurance against

calamities are preferred policies particularly in the poorly endowed regions. Chada discusses the role of nonfarm rural employment in income diversification and, consequently, in risk management of farm households in India. Using rural employment data for India, the author shows that expansion of rural nonfarm employment has helped farmers manage risk better. In addition, expansion of nonfarm rural industry will directly promote economic growth by better use of forward and backward linkages associated with agricultural growth. Improvements in infrastructure and agrarian reforms are seen as important policy interventions needed to stimulate the growth of nonfarm employment in rural areas. Mishra discusses the issues related to crop insurance as a policy response for stabilization of crop income. Although the current wisdom is that publicly funded crop insurance programs are financially unviable, the author finds the total social benefit of the comprehensive crop insurance scheme in India to be higher than the total cost. Thus, from the social point of view, the author shows that crop insurance programs can be desirable, even though they may not be financially viable without the subsidy. Nevertheless, the author suggests that opportunities for improving the financial performance of crop insurance schemes should be exploited as much as possible to improve their financial viability. The problems of adverse selection and moral hazard, however, continue to erode the viability of crop insurance schemes. Two papers (Pandey and Pal, and Pal et al) have focused attention on assessing the pattern of changes in productivity and variability in eastern India using district-level data. The authors report a diverse pattern of change with variability increasing in some districts, decreasing in others, but remaining more or less unchanged in most. Eastern Uttar Pradesh and West Bengal are the two states where the change in variability (defined by the CV of production) has been stabilizing. This is attributed mainly to the availability of supplemental irrigation that reduced risk and encouraged farmers to adopt

vi

yield-increasing modem varieties. Parts of Bihar and Orissa, on the other hand, experienced an increase in production variability. The coefficient of variation of district-level yield was found to be positively related to the coefficient of rainfall and negatively related to the quantity of fertilizer used. As the latter is an indicator of the extent of adoption of improved technologies, the stability consequences of the adoption of improved technologies appear to have been favorable in eastern India. The findings of these two studies support the view that productivity gains in parts of eastern India have been achieved without increasing instability. This indicates that growth and stability are not necessarily incompatible goals. The expansion of irrigation and development of rice varieties suitable to the environment of eastern India are likely to be the causal factors that have reduced instability and increased growth simultaneously. However, there is an underlying trend toward an increasing correlation in production across districts that, if unchecked, could have a destabilizing effect. The Pandey et al paper analyzes risk management strategies using panel data from two villages with contrasting risk profiles in eastern Uttar Pradesh. Diversification and maintenance of flexibility are seen as two major strategies for reducing risk. The analysis of panel data permitted the authors to document changes in cropping patterns, varieties of rice grown, methods of crop establishment, and input use over time and relate these changes to rainfall. The paper shows that area variability is an important component of variability in rice production. Most of the biological research on rice ignores area variability and focuses on yield variability. One important contribution of this paper is that it shows that the risk benefits of stabilization of rice yield in the study villages are quite small. This is mainly due to a very small share of rice in the total household income. As a result, stabilization of rice income will not necessarily translate into stabilization of total household income. As farmer income sources are already diversified away from rice, rice research can have more impact by focusing on yield improvement rather than on yield stabilization per se. However, in other areas

where income diversification opportunities are constrained by infrastructure and biophysical factors, stabilization of rice yield can result in substantial income gains. The paper by Abedullah and Pandey provides an estimate of the economic value of rainfall forecast to rainfed rice farmers in the Philippines. This is the only paper in the volume that includes data from outside India. Using a decision-theoretical approach, the authors estimate the value of three types of seasonal rainfall forecast (average, below average, and above average). The economic value of forecasts arises from farmers being able to alter crop management practices if they have access to the forecasts. To get around the problems related to forecast accuracy, the authors estimate the economic value of a perfect forecast as such estimates provide the upper limit to the value of a forecast. The estimated value of such a forecast was found to be 1% of the net returns from rice. For the rainfed rice area of the Philippines, the total value was estimated to be $6.6 million per year. Four papers (Bhowmick et al, Thakur, Saha et al, and Naik et al) analyze rice production and instability in Assam, Bihar, West Bengal, and Orissa, respectively, the major riceproducing states of eastern India. Although rice is grown in three different time periods in eastern India, the papers show that rainfed rice is grown mainly in July to November and the variability of total rice production is determined mainly by the variability during this period. The Bhowmick et al paper highlights the importance of flood risk in Assam. Overall, the variability of rice production in Assam has changed very little. The results from Bihar show that its variability of rice production and yield is the highest of all eastern Indian states. The interaction between modem varieties and complex hydrology in Bihar is probably the main reason for increased production variability in this state. Highly variable environmental conditions could also be a reason for the shrinkage of rice area in this state. The Saha et al paper on West Bengal provides a more detailed description of changes in productivity patterns in West Bengal and how farmers manage risk by adjusting crop management. vii

West Bengal is the only state with the least variation in productivity over time. The paper shows that low variability has resulted mainly from stabilization of rainy-season rice production even though the importance of summer rice has grown over time. The Naik et al paper analyzes the instability of rice production in Orissa. The paper shows that the variability in rice yield across districts is not related to the adoption of modern varieties but mainly to soil/climatic factors. In the case of Orissa, opportunities to reduce risk by manipulating crop management practices of rice seem circumscribed by hydrological factors, especially in the coastal belt. From a biological perspective, the Wade paper brings out clearly opportunities to reduce risk through a better understanding of the genotype by environment interactions. Experimental data and crop simulation are seen as important in understanding the nature of risk and its management through manipulation of varieties and crop management. The paper also provides some insights into the more downstream aspects of technology adaptation and dissemination through farmer participatory methods and involving nontraditional extension agencies such as farmer organizations and nongovernment organizations.

may not be as high as often believed. If rice contributes to only a small proportion of the total income of farm households, interventions that stabilize rice income are not necessarily effective in stabilizing total household income. Thus, yield-stabilizing technologies for a single crop are not likely to be effective in reducing risk, even for an important crop such as rice. However, in areas where income diversification is limited due to limited infrastructure or less favorable agroclimatic conditions, the economic cost associated with instability in rice production can be substantial. Thus, there is a need to delineate rainfed rice environments in terms of the current extent of income diversification and the possibilities for diversification in the future. Interventions designed mainly for stabilizing yield and incomes (such as technologies with higher yield stability and crop insurance) are likely to be less useful in environments with ample opportunities for diversification. Naturally, technologies that improve the average yield of rice are always important, irrespective of the nature of the environment. Data needs One of the major gaps in the current empirical work on risk analysis in the context of rainfed rice systems is the lack of information on the relative importance of various risk-coping mechanisms and how they change with increasing commercialization of agriculture. Very little information is available on the determinants of various strategies that farmers employ to cope with risk and their associated cost. One of the difficulties has been the lack of panel data to study the correlation between climatic fluctuations and farmers’ responses. Village-level studies such as the ones conducted by the international Crops Research Institute for the Semi-Arid Tropics (ICRISAT) can be important in bridging the information gap for rainfed rice systems also. Risk and externality Most of the current studies on risk for rainfed rice areas focus on the farm or household as the unit of analysis. Little information is available

Summary and synthesis
The papers presented during the workshop and the discussions that ensued covered many issues related to agricultural growth in eastern India and risk management. While most of the papers used the concepts and methods that were popularized when the study of interaction between agricultural risk and technology adoption was popular during the late 1970s and 1980s, the findings reported during the workshop provide new insights into conceptual and research issues. Some of these major issues are summarized below. Yield stabilization Rice production is an important economic activity in eastern India. Despite its importance, the share of rice in the total household income viii

at a higher level of aggregation such as the community, district, or state. Often, the source of increased risk in downstream areas may be the result of inappropriate land use in the upstream areas. For example, deforestation and the use of soil-eroding practices in the upper parts of a watershed can increase the flood risk in downstream areas. Such risks are better managed through the use of more sustainable land use systems in the upper slopes than by other means. In this regard, collective institutions can play an important role in the management of overall risk for the whole watershed. However, the study of interactions between collective institutions and risk management remains a relatively uncharted territory. Macroeconomic instability and risk Another area of research that deserves adequate attention is the effect of macroeconomic instability on risk in agriculture. As agriculture changes from subsistence to commercial orientation, the agricultural sector becomes more prone to macroeconomic shocks of fluctuations in foreign exchange rates and interest rates. With the anticipated trend toward globalization of trade following the WTO agreement, the macroeconomic shocks are likely to be transmitted more easily across countries. Policies and institutional mechanisms needed to stabilize food production and farmers’ income under such conditions are yet to be adequately scrutinized. Upscaling and extrapolation A methodological issue relates to the quantification of the impacts (production losses, income, and welfare) of unpredictable shocks at different geographic scales. Farm-level losses can be quantified through a sample survey and other traditional methods. However, geographically referenced spatial information on factors that affect production losses are needed to extrapolate such information to the regional and subnational levels. Such databases are now becoming increasingly available with the growing popularity of geographic information

systems (GIS). Nevertheless, methods and approaches are needed to upscale farm level effects of risk to a higher level of aggregation. Economic value of forecasts Opportunities for reducing the economic cost of risk by providing of forecast information have not been adequately addressed in the agricultural sector worldwide, except for some specific highvalue crops. In most developing countries, climatic forecasts are rarely available in a form useful to farmers for planning agricultural operations. Similarly, opportunities for reducing risk through the better use of information of forecast prices are rarely available. Policymakers and farmers alike have not adequately appreciated the importance of information acquisition and use for risk management. Perhaps the value of information is low in traditional subsistence-oriented agriculture. But with increasing commercialization, the use of forecast information can be an important strategy for reducing price and weather risk. Risk analysis of rainfed rice systems

In rainfed rice areas, the overall theme of risk analysis and management has not been adequately studied. Parallels are drawn from earlier work conducted in irrigated areas where the interaction between risk and technology adoption was widely discussed. While theoretical and conceptual advances made in such studies are relevant to rainfed environments also, empirical applications for analyzing farmers’ decisions regarding the choice of technology, their risk-coping strategies, and the overall effect of farmers’ risk management strategies on production instability at the aggregate level have been far too few. Due to the high degree of heterogeneity in rainfed environments, the domain of a specific technology is likely to be much narrower. A careful delineation of rice production environments, based on risk profile and farmers’ socioeconomic conditions, is needed to target technology development. With the availability of more powerful computer technology, many ix

powerful tools ranging from crop simulation to GIS are now in the hands of analysts. More studies that use such tools to quantify effects of risk at the production systems level are needed. Increasing the adoption of modem varieties of rice even in rainfed areas and the growth in income of farm households observed during the 1990s in eastern India indicate that farm households have been able to mitigate the effect of risk to a certain extent. S. Pandey* B.C. Barah R.A. Villano S. Pal

*Sushil Pandey and Renato Villano are agricultural economist and assistant scientist, respectively, at the International Rice Research Institute, Los Baños, Laguna, Philippines; B.C. Barah and S. Pal are principal scientist and senior scientist, respectively, at the National Centre for Agricultural Economics and Policy Research (NCAP), New Delhi, India. The editors acknowledge support and encouragement from Dr. Mahabub Hossain, head, Social Sciences Division, IRRI; Dr. D. Jha, national professor and exdirector, NCAP and Dr. Mruthyunjaya, director, NCAP. Editorial assistance provided by Dr. Bill Hardy, Ms. Teresita Rola, Ms. Millet Magsino, Ms. Erlie Putungan, Mr. Juan Lazaro IV, and Mr. George Reyes of the Communication and Publications Services, IRRI, is gratefully acknowledged.

x

Units of measurement

All data on rice and production in this report are expressed in terms of rough rice. The conversion factor used is 1 kg of rough rice = 0.66 kg of milled rice. All monetary values for studies in India are expressed in Indian rupees. In 1997, the exchange rate was 1US$ = Rs 36.31.

xi

Risk and rainfed rice: some conceptual and methodological issues
S. Pandey

The study of risk and its interaction with technology is an important topic in agricultural development. The paper provides a review and synthesis of conceptual and empirical issues in risk analysis in the context of rainfed rice farming. Various strategies employed by farmers for managing risk are discussed and implications of these strategies for designing and disseminating technologies are derived. Methodological and measurement issues that require further development are highlighted.

Rainfed rice farmers, like farmers everywhere, have to carry out production activities in an inherently uncertain environment. Production is affected by drought, flood, and pests and diseases, which occur in an unpredictable way. In addition, farmers’ income and welfare also depend on uncertainty related to economic parameters such as price and marketing. Efficient management of risk is hence the essence of rainfed agriculture. For poor farmers, risk considerations may loom large in their choices of crops and the method of production. Hence, a study of how farmers are likely to respond to technological and policy interventions in the face of risk is critical in designing these interventions.

Definition and measurement of risk
Although the word “risk” is used in all walks of life to describe the chances of some undesirable outcome, defining it precisely and unambiguously is not easy. This is reflected in the following statement: “Risk is like love; we have a good idea of what it is, but we can’t define it precisely” (Stiglitz as quoted in Roumasset et a1 1979). The Macquarie Dictionary defines risk as “exposure to the chance of injury or loss.” As “injury” or “loss” is a subjective concept with its consequence depending on the person as well as the circumstance, what is considered to be risky by

one individual may not be seen to be so by another person. Risk, hence, is subjective. It is essential to draw a clear distinction between risk and variability. The latter term merely implies that a variable of interest is not fixed but has different values. No risk is involved if the value of the variable can be known with certainty. For example, farm size may vary from farmer to farmer but can be known with certainty. Similarly, soil type within a farm can vary from paddock to paddock but can be known with a fair degree of certainty. Uncertainty about the likely values, not the variability per se, is the source of risk. A notion such as risk that is intrinsically subjective obviously cannot be measured by an “objective” indicator. Subjective probability distribution of an uncertain outcome of concern to the decision maker, hence, is considered a suitable indicator of risk. Under this definition, risk can be measured by (1) the chance of an undesirable outcome, (2) the variability of outcome (or the converse of stability), and (3) the probability distribution of outcome. The first measure implies that a situation in which the chance of an undesirable outcome is greater is “riskier.” Although intuitively appealing, the measure is problematic because it is not clear when an outcome is “unacceptable.” Statistical descriptors of the probability distribution-such the variance and the as coefficient of variation—have found common

1

use as a measure of risk. However, Rothschild and Stiglitz (1970) showed that none of the statistical descriptors adequately measure risk. They contend that it is impossible to devise a universally valid statistical descriptor of risk without simultaneously considering both the probability distribution of outcomes and the risk attitude of the decision maker. Given these difficulties in devising the adequate measure of risk, it has been argued that the ambiguous terms ‘more risky’ or ‘less risky’ should be avoided. If no satisfactory measuring scale exists, then it is not possible to consider risk as being ‘more’ or ‘less.’ What is theoretically appealing is to view a decision as being “risk efficient” or “risk inefficient.” Such decisions may lead to an increase in the mean income and/or a reduction in the dispersion of income around the mean. Risk efficiency can be best ascertained by comparing the whole probability distributions of the uncertain outcomes that correspond to different decisions.

Risk and its impact on technology adoption
The impact of risk and risk aversion on the choice of agricultural production techniques and input use has been a topic of extensive investigation (Feder et al 1985, Anderson and Hazel1 1994). Theoretical studies on farmer behavior under risk indicate that, in the absence of a perfect market for insurance, resource allocation for risk-averse farmers differs from that for risk-neutral farmers (Sandmo 1971, Anderson et al 1977). The effect of risk is considered to depend on risk perceptions and risk attitudes. Farmers may be reluctant to adopt technologies that they perceive to be riskier. Risk perception depends on the quality of the information they have and their information-processing capabilities. To the extent that farmers perceive a technology to be riskier than it actually is, activities such as onfarm research to generate more accurate information and investment in educating farmers are warranted. Assuming farmers’ perceptions of risk associated with a technology to be reasonably accurate, whether adoption occurs also depends 2

on risk attitudes. Variability of income is irrelevant to risk-neutral farmers. A technology that generates a higher level of mean income would be preferred by such farmers. However, risk-averse farmers are likely to consider simultaneously both the level of income and risk and to reject a technology that they consider too risky. Empirical evidence indicates that farmers in developing countries are generally risk-averse (Binswanger 1980, Walker and Ryan 1990). If poorer farmers are more averse to risk, rainfed rice farmers who are mostly poor are likely to be reluctant to adopt technologies that increase risk. In addition to this direct effect, risk aversion also indirectly affects technology adoption through its impact on the credit market (Binswanger and Sillers 1983). Risk-averse lenders may demand greater collateral and may charge a higher interest rate, depressing credit use by poorer farmers. Similarly, more risk-averse farmers are less likely to demand credit. To the extent that credit use is essential for adoption of technologies that require purchased inputs (such as fertilizers), risk aversion discourages technology adoption. This indirect effect of risk aversion is often considered to be more important than the direct effect (Binswanger and Sillers 1983). The study of risk basically consists of two aspects: risk analysis and risk management analysis. Risk analysis consists of the study of the nature, magnitude, and sources of risk and how technology affects these characteristics. Risk management, on the other hand, involves the use of methods that reduce risk and its impact. Even if a technology is risky, farmers may adopt it if adequate means for diffusing risk are available. Risk could be studied at the micro (or farm) level or at the macro (region or nation) level. The purpose of farm-level analysis is to study adoption decisions in the face of risk. Macroeconomic parameters are assumed to be given and farmers’ responses to risk are studied. In the case of macro analysis. the purpose is to study the implications of fluctuating production for food security at the regional or national level. Although farmers may adopt improved technologies because they are profitable, the

instability of production at the aggregate level may increase as a consequence. Appropriate technological and policy interventions are required to reduce such adverse effects on food security.

Sources of risk
Income of farmers from agricultural production can fluctuate as a result of fluctuations in yield, price of output, area planted, price of input, and input supply. Agricultural scientists are mainly concerned with yield risk as it is often a major component of risk, especially under rainfed conditions. If a farmer is ultimately interested in reducing the uncertainty of income (as income and consumption in rural societies are highly correlated), other components of risk can also be important. For example, a negative correlation between the price of rice and yield tends to stabilize the income from rice compared with a situation when these two variables are positively correlated. Hence, if the interest is in stabilizing farmers’ incomes, it is necessary to evaluate the consequence of price instability and how it affects income stability. Evaluation of technology in terms of instability of yield alone will not be adequate. The importance of price uncertainty is likely to increase as rice production systems become more commercialized. Price risk is an important component of income risk, especially in commercialized systems. When the use of purchased inputs is minimal and output is mainly for subsistence, market prices are less relevant for resource allocation by farmers. However, in commercialized systems where traded inputs are substituted for nontraded inputs and output is mainly for the market, fluctuations in prices of both inputs and outputs can have a major impact on farmer welfare. A negative correlation between price and yield is an important feature of agricultural production, which helps in stabilizing farm income. Prices tend to be high when production is low and they tend to be low when production is high. Stabilizing prices in this situation can actually raise farm income instability. Marketlevel analyses and the study of price policy are

required to help design price policies that enhance farm income stability. Variability in input supply and input prices can lead to fluctuations in output as farmers adjust input levels to prevailing conditions. Marketing infrastructure is important in determining input supply risks. Similarly, government policies on production and marketing of agricultural inputs determine input supply and price risks.

Risk at the aggregate level
Even if aggregate food production is increasing, wide fluctuations in total supply can seriously affect food security at the household level, especially that of the poor. The nature of publicsector interventions in the food market depends on the instability of aggregate production. The economic costs of maintaining a bigger stock of food grain to deal with higher instability can be substantial. In addition, the effect of instability at the national level also spills over to international markets and can cause wide swings in prices, thus affecting food security in other countries also. Analysis of the patterns of instability in food grain production is hence relevant in the context of food security. It is now widely accepted that the adoption of improved technology consisting of highyielding varieties (HYVs) and associated crop management practices has increased food production in Asia over the last 20 years. What is still debatable is the effect on variability of production. The adoption of modern varieties and improved crop management techniques can make aggregate production more variable by increasing interregional correlation. When farmers grow similar varieties and use similar management practices, adverse climatic conditions over a large area can lead to a large drop in production. Similarly, when the supply of major inputs is unreliable and/or input prices change, farmers are likely to adjust their input use in the same direction, leading to covariate movement in output. This economic response can lead to increased production instability even if the yield of modem varieties is more stable than that of traditional varieties.

3

Empirical evidence shows that production variability in the aggregate has increased with the adoption of improved varieties in India (Hazell 1982). Much of the increase in production variability in food grains has been attributed not to the adoption of improved varieties per se but to fluctuations in input supply such as irrigation (due to power outages) and fertilizers (Hazell 1982). On the other hand, Walker (1989) found the adoption of HYVs of sorghum and pearl millet to be a major factor contributing to increased production variability of these crops. Using district-level data from India, Singh and Byerlee (1990) found that variability in wheat yield, measured by the coefficient of variation, has decreased over time, mainly as a result of expansion in irrigated area. Rao (1968), Mehra (1981), and Pandey (1989) have also discussed the effect of irrigation on production variability. In a more recent study covering three major crops–rice, wheat, and maize–Naylor et al (1997) found that the variability in global output of rice and wheat, measured by the average percentage deviation from the trend, increased initially with the adoption of modem varieties but then declined subsequently. The probability of a significant shortfall below the trend also decreased between the pre- and post-Green Revolution periods for these two crops. In the case of maize, production variability was higher during 1980-94 than during 1950-64. The dominance of U.S. maize production in global output and a greater downside sensitivity of yield to climatic conditions when yields are close to the ceiling have been attributed to the increase in variability in global maize production. Although the rapid growth in yield of rice and wheat may have swamped the increase in Variability during the Green Revolution period, instability may be more pronounced in the future as yield ceilings for these crops are also approached. More evidence on the aggregate effects of technology adoption is contained in papers by Pal et al (2000) and Pandey and Pal (2000).

Coping mechanisms of farmers
As a result of their exposure to risk, farmers have developed various strategies over time to avoid the negative consequences of unpredictable variations in agricultural output. A good understanding of these strategies is needed to assess the likely responses of farmers to new technologies or policies. Uptake of technologies that complement and reinforce farmers’ coping strategies is likely to be quite rapid. On the other hand, interventions that undermine key components of risk management strategies are likely to be rejected. Farmers’ risk-coping strategies can be classified into ex ante and ex post, depending on whether they help reduce risk or reduce the impact of risk after a production shortfall has occurred. Because of the lack of efficient market-based mechanisms for diffusing risk, farmers modify their production practices to provide “self-insurance” so that the chances of negative consequences are reduced to an acceptable level. Ex ante strategies help reduce fluctuations in income and are also referred to as income-smoothing strategies. These strategies can be costly, however, in terms of forgone opportunities for income gains as farmers select safer but low-return activities. Ex ante strategies can be grouped into two categories: those that reduce risk by diversification and those that do so through greater flexibility. Diversification is simply captured in the principle of not putting “all eggs in one basket.” The risk of income shortfall is reduced by growing several crops that have negatively or weakly correlated returns. This principle is used in different types of diversification common in rural societies. Examples are spatial diversification of farms, diversification of agricultural enterprises, and diversification from farm to nonfarm activities. Maintaining flexibility is an adaptive strategy that allows farmers to switch between activities as the situation demands. Flexibility in decision making permits farmers not only to

4

reduce the chances of low incomes but also to capture income-increasing opportunities when they do arise. Examples are using split doses of fertilizers, temporally adjusting input use to crop conditions, and adjusting the area allocated to a crop depending on climatic conditions. While postponing agricultural decisions until uncertainties are reduced can help lower potential losses, such a strategy can also be costly in terms of income forgone if operations are delayed beyond the optimal biological window. Ex post strategies are designed to prevent a shortfall in consumption when family income drops below what is necessary for maintaining consumption at its normal level. They are also referred to as consumption-smoothing strategies as they help reduce fluctuations in consumption even when income is fluctuating. These include migration, consumption loans, asset liquidation, and charity. A consumption shortfall can occur despite these ex post strategies if the drop in income is substantial. Farmers who are exposed to risk use these strategies in different combinations to ensure survival. Over a long period of time, some of these strategies are incorporated into the nature of the farming system and are often not easily identifiable as risk-coping mechanisms. Others are employed only under certain risky situations and are easier to identify as responses to risk. Ex ante coping mechanisms Ex ante coping mechanisms are designed to exploit low correlation among activity returns for stabilization of total income. These operate through various types of diversification that characterize traditional agriculture. Diversification may be considered as horizontal or vertical. The former refers to scattering of agricultural fields, growing of several crops, growing of several varieties of the same crop, and engaging in different income-generating activities. The latter relates to spreading agricultural operations over time. This refers to strategies such as staggered planting, spreading input use over a period of time, planting many seeds per hill, and temporally diverse planting.

Vertical diversification is a way of maintaining flexibility to adjust agricultural operations to the evolving uncertainty. Similarly, share cropping is viewed as a way of reducing risk through sharing of risk between the landlord and the tenant. Spatial diversification of fields. Agricultural fields vary from location to location in attributes such as soil moisture retention and fertility. In rainfed areas, these soil characteristics can vary widely even across fields. Similarly, rainfall distribution can also vary among fields in different locations. These variations in soil characteristics and rainfall across locations create an opportunity for fanners to stabilize agricultural output through spatial scattering of fields. Although output from fields in one location can decrease because of poor rainfall, it can increase in fields in other locations that receive higher rainfall. Weakly or negatively correlated crop yields across fields result in these compensating movements so that total farm output is more stable than output from individual fields. Spatial scattering of fields is a way of exploiting this stabilizing effect. In addition, this strategy may also help farmers to better exploit specific niches of different microenvironments to enhance productivity enhancement. In spite of these potential gains, spatial diversification of fields can cause an efficiency loss because of the increased costs of moving inputs across and marketing outputs from widely separated fields. Whether or not farmers use spatial scattering depends on the net effect of these factors. In addition, local institutions such as the inheritance law may condition the prevalence of such a practice. In rice-growing regions of Asia, it is not uncommon to find a farm household operating several parcels of land that are either spatially scattered or differ in their location along the toposequence. While risk considerations may have played a role in determining the extent of land fragmentation, casual observation indicates that land fragmentation is driven mainly by the desire to exploit different environmental niches that are suitable for different crops. In parts of eastern India, ail parcels of land are divided among legal heirs so that everybody gets an

5

equal share of all types of environmental niches. The desire for an equitable distribution of land of different quality among heirs is often considered to be a factor constraining efforts at land consolidation. If land fragmentation is an effective way of reducing risk, one would expect to observe a greater degree of fragmentation in areas where environmental conditions are less stable. However, such a pattern may not be observed due to other counteracting factors. For example, the extent of fragmentation in the more risky Sahel region of Africa has been less than in the more favorable Sudan region (Matlon 1991). This is attributed to the differences in environmental factors in these two regions. In the Sahel, low rainfall prevents farmers from cultivating a wider range of field types. As a result, cropping is restricted to only certain field types where crop success is more assured. In Sudan zone, higher rainfall and generally better soil conditions enable farmers to use a range of field types. In this example, the lack of feasible alternatives in the highly constraining environment of the Sahel reduced the value of spatial diversification as a risk management tool. Even if the inheritance law may have a big role in determining farm size and extent of fragmentation, farmers can and do alter their land resource base through land rental markets. Field experience from eastern India indicates that tenants with a given endowment of land types prefer to rent a different land type. Renting a better quality land increases average income. It may also simultaneously achieve the objective of risk reduction. Crop diversification. As with spatial diversification, farm output can be stabilized by growing several crops with poorly or negatively correlated yields. Environmental conditions less favorable to some crops may be more favorable to others, so that compensating variations in yields of different crops would impart stability to total output. In addition to risk reduction, there are several other potential benefits of crop diversification, such as a better exploitation of environmental niches, staggering of labor demand, and meeting the demand for a range of outputs. Mixed cropping and intercropping,

which are a common feature of traditional agriculture in Asia, are a form of crop diversification that reduces output variability (Walker and Jodha 1986, Siddiq and Kundu 1993). Crop diversification, however, can also be costly in terms of income gain forgone as farm households include crops with lower but more stable yields in their cropping pattern. In addition, economies of size that may result from specialization are also lost as production is diversified. Crop diversification is a feature of traditional farming systems in Asia. The role of crop diversification in risk reduction has been analyzed extensively in the context of farming in the semiarid tropics where farmers grow a range of intercrops and mixed crops. Crop diversification has been greater in the more risky environments in the semiarid tropics of India (Walker and Jodha 1986). In the rainfed rice environments of eastern India, crop diversification is greater in areas with a less assured supply of irrigation (Pandey et al, this volume). Crop diversification in flood-prone areas in a village in eastern India declined after dikes for protection from flood were constructed (Ballabh and Pandey 1999). Although diversification may reduce instability, whether or not farmers are able to diversify land use also depends on the environmental conditions. Again taking the example from Africa, low and unstable rainfall and poor soils in the Sahel have constrained opportunities for diversification, with the milletbased cropping pattern being the dominant one. In comparison, in the relatively favorable environments of the northern Guinea zone, the cropping pattern is more diversified (Matlon 1991). In addition, the more limited cropping opportunities in the Sahel also mean that crop yields are likely to be highly correlated, thus reducing the benefits from crop diversification. In the humid environments of Asia, drainage constraints in the submergence-prone bottom land similarly limit opportunities for crop diversification during the rainy season. Varietal diversification. Growing several varieties of a crop is a form of diversification that can stabilize the total output of the crop if

6

yields of different varieties are poorly correlated. Varieties with different duration can reduce risk by avoiding period-specific risk, For example, short-duration varieties can escape terminal drought that can severely affect the yield of a longer duration variety. Similarly, varieties with different degrees of tolerance for pests and diseases also help reduce losses. Rainfed rice farmers in eastern India almost invariably grow several varieties for different reasons, including possible risk reduction. In a rainfed rice village in Orissa, more than 70% of the farmers grow two to five varieties, with 20% of the farmers growing six to eight varieties (Kshirsagar et a1 1997). Similarly, in the rainfed lowland of Lao PDR, 60% of the farmers grow four or more rice varieties (Pandey and Sanamongkhoun 1998). As with crop diversification, other advantages of varietal diversification are niche matching, staggering labor demand, and generating a range of product characteristics. These latter motives are not directly related to risk management and may condition the extent of varietal diversification practiced by farmers in a given area. Income diversification. Like crop diversification that uses weak correlation among activity returns to stabilize farm income, diversification of income from farm to nonfarm sources is another way of stabilizing income. If fluctuations in nonfarm incomes are independent of fluctuations in farm output, income diversification through one or more members of the family working in the nonfarm sector can stabilize total family income. The extent of income diversification may depend on factors such as rural education, transportation infrastructure, access to institutional credit, and availability of local resources for nonfarm activities. These factors may constrain opportunities for income diversification even when agricultural risk is high. In areas with environmental conditions conducive to a strong agricultural base, income-generating activities that take advantage of agriculture’s forward and backward linkages expand. On the other hand, income diversification in agriculturally poor areas tends to be outward-looking, with households diversifying their income

geographically (Reardon et a1 1988, 1992). Share cropping. A large volume of literature on risk and efficiency implications of share cropping exists (Newbery and Stiglitz 1979, Otsuka et a1 1992). At their very basic, share cropping arrangements that lead to sharing of input and output also lead to sharing of risk between the landlord and the tenant. However, the existence of share cropping depends on many other factors in addition to risk benefits (Otsuka et a1 1992). Temporal adjustments. Crop growth is a biological process that occurs over a period of time. The economic output is obtained upon maturity when the crop is finally harvested. The crop is exposed to various factors during the intervening period between planting and harvest. Some of these factors are known with a fair degree of certainty, wheras others are highly uncertain. These factors, together with management interventions by farmers, determine the ultimate economic value of the crop output. Uncertainties are highest at planting time as future values of uncertain events are known very imprecisely. As uncertainties are resolved with the passage of time, farmers can gain by making decisions conditional on the occurrence of uncertain events up to that time and the revised expectation about the future occurrence of uncertain events. Such a sequential decisionmaking process imparts flexibility and allows farmers to exploit favorable events for income gains while reducing potential losses. To assess the value of sequential decision making, it may be useful to divide the cropping season into early, middle, and late stages. The early stage can be considered to include preplanting and the period immediately after planting. The major decisions to be made at this stage are the crops, the variety, the timing of planting, and the method of establishment. The middle stage is considered to be the period between successful crop establishment and flowering. Major decisions here are weeding, fertilization, control of pests, and irrigation. The final stage is the period after flowering until harvest. The rainfall pattern during the early stage may determine the choice of crops. If rains are

7

low or delayed during this period, farmers may forgo rice completely and expand the area under crops that require less water. Similarly, if too much water is received, farmers may expand the area under rice at the expense of other crops. In eastern India, sown area of rice contracts in years with low and unstable early-season rainfall (Pandey et al, 2000a). If the crop fails to establish itself because of too much or too little rain, farmers may decide to replant. Farmers similarly may engage in gap filling and thinning to reduce risk (Singh et al 1995). The choice of what rice variety to grow also depends partly on the nature of rainfall during this early period. Farm-level data from eastern India indicate that, in years with late rains, farmers expand the area under short-duration varieties as a mechanism for escaping terminal drought. Expanding the proportionate area under traditional varieties and resorting more to dry seeding as opposed to transplanting are other responses exhibited by farmers in eastern India. Once the crop is successfully established, farmers may adapt the level of input they use, depending on their assessment of crop health. If the crop potential appears to be low, farmers may leave some fields unweeded and apply lower than normal quantities of fertilizer. Surplus resources may be used for other crops in the same or the following season. Farmers even replant the area with some other crops if they anticipate the rice yield being too low and if the season has not advanced too far (Singh et al 1995). During the third stage, most of the uncertainties would have been resolved and few decisions would remain to be made. If rice fails during this stage, farmers may go for “salvage” operations to obtain at least the byproduct (straw, in the case of rice). Another response observed in eastern India is to establish post rainy-season crops early in the rice field if soil moisture conditions are favorable. The temporal adjustments described above are farmers’ mechanisms for reducing losses in poorer years and increasing gains in more favorable years. Relative to committing all resources at the beginning of the cropping season or on the basis of a fixed calendar, the

average farm income will always be higher when flexible methods are adopted. However, opportunities for using flexibility may be constrained by farmers’ ability to process the necessary information about crop status and the likely future occurrence of uncertain events. In addition, in poorer and harsher environments, flexibility may be so circumscribed that it cannot be relied upon as an effective risk-coping mechanism. Ex post coping mechanisms How do farmers cope with losses that do occur despite the various risk-reducing mechanisms adopted? The shortfall in agricultural production will reduce consumption if farmers are not able to meet the deficit through some other means. Depleting food and cash savings, earning more wage income, borrowing, liquidating assets, reducing consumption, relying on charity, and permanent migration are some of the mechanisms used for coping with a production shortfall. The economic burden and the long-term productivity impacts of these mechanisms differ. If farmers are able to save during betterthan-normal years and use the savings to meet consumption deficits during drought years, they may be able to maintain their consumption level over time despite short-term fluctuations in agricultural output. Savings in agricultural societies may take various forms. They could be held in the form of food grains, cash, and jewelry. They could also be held in the form of productive assets such as bullocks, farm implements, and land. Even if own savings are not enough to meet the consumption deficit, village-level institutions may permit sharing of risk across individuals such that individual consumption fluctuates much less than individual production. Empirical evidence from several studies in developing countries indicates that consumption smoothing is a common practice among farmers (Townsend 1994, 1995). Based on data from the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), crop inventory and cash reserves play major roles in smoothing

8

consumption in the semiarid tropics of India (Lim and Townsend 1994, Paxson and Chaudhuri 1994). The importance of these two mechanisms varies by farm size, with large farmers relying more on crop inventory and small farmers relying more on currency. The use of credit was another important mechanism. Results for Thailand were also similar (Townsend 1995). . The effectiveness of these mechanisms depends on the seventy of risk such as drought and crop output in the preceding year. Problems are less severe in a year with mild drought that follows a good year, and these mechanisms may be adequate to meet the shortfall. These internal reserves, however, may be grossly inadequate when drought years are consecutive or if drought is severe. In such situations, farmers may be forced to reduce consumption, with small farmers and landless labor suffering the most. Based on farm-level data from arid and semiarid areas of India, the decline in cereal consumption in a drought year relative to a normal year varied between 12% and 22% (Jodha 1978). In addition, there were drastic cuts in the expenditure on protective food such as milk, sugar, vegetables, fruits, meat, and others. Pandey et al (2000b) made similar observations for eastern India. Such shortfalls in consumption point to the inadequacy of consumptionsmoothing mechanisms, especially among small farmers. Livestock, in addition to being useful for agricultural production, are also an important store of wealth in rural societies. They serve an important role in consumption smoothing. During drought years, livestock are sold and proceeds are used to meet a consumption shortfall. Disposal of livestock can also help reduce carrying costs, which tend to be high, especially during drought years (Kinsey et al 1998). In the Sahel zone of Africa, where poor environmental conditions constrain the efficacy of ex ante mechanisms, manipulation of livestock inventory is an important ex post mechanism (Matlon 1991). Farmers in India similarly use the livestock inventory to reduce consumption shortfall (Jodha 1978).

A problem with the use of livestock for consumption smoothing is that this coping mechanism, while helping farmers to survive during drought years, can reduce the long-term production potential. Where livestock are simply a store of wealth, this will not create a problem. Disposing of livestock in this case would be similar to withdrawing cash from the bank. In fact, disposal of small animals such as goats and sheep, which tend to be good stores of value, is generally the initial response to income shortfalls. However, livestock are also the major source of draft power needed for several farm operations such as tillage, pumping irrigation water, threshing rice, and hauling farm inputs and outputs. Faced with the prospect of a severe shortage in consumption in a severe or prolonged drought, farmers may sell productive livestock such as cattle, buffaloes, and horses. Once these productive livestock assets are depleted, it takes a long time for them to be replenished. Thus, even after the drought is over and rainfall returns to normal, it may take several years for farmers to rebuild their stock of livestock. A typical feature of the livestock depletion-replenishment cycle is that livestock are sold when their prices are falling due to excess supply during drought years (Jodha 1978). Increased demand during the replenishment phase pushes the prices up, making it more difficult for farmers to reacquire the livestock. If several drought years occur in a row, the livestock asset may be depleted so severely that several years of normal conditions would be needed for full replenishment of the livestock. The effect of drought can thus linger on for several years until productive assets are fully replaced. As the mortality of livestock is higher in drought years due to poor nutrition, the asset base can be depleted dramatically during a run of drought years. Thus, this coping mechanism could be costly in terms of future production potential forgone. The impact is likely to be greater for small farmers than for large farmers as small farmers often need a longer time to replenish the depleted stock. As with the depletion of livestock, severe droughts can lead to excessive exploitation of common property resources (CPR) that are a

9

critical component of village livelihood systems (Jodha 1986). The CPR are resources owned in common by village residents. These include community forests, pasture/waste land, ponds, river banks and river beds, and groundwater. The poorer segments of the rural population are especially dependent on CPR, even in normal times, to generate food, fiber, and income. During drought periods, these resources become even more important. For example, the reduced supply of fodder during drought years increases the reliance on forest and community grazing areas for sustaining the livestock. Similarly, additional incomes are generated by selling timber, fuelwood, and other forest products. Collection of edible forest products such as fruits, nuts, and bamboo shoots also increases as farmers attempt to meet the shortfall in production. If these CPR are depleted excessively during drought years, the productivity of agriculture and livelihood of the poor can be adversely affected for many years even after the meteorological drought ends. Short-term or permanent migration to earn income from cities or far-away places is another coping mechanism. Migration to nearby places is likely to be less effective due to covariate movements in income within a small geographic area. Prospects for earning income within the locality affected by drought are limited due to a reduction in demand for labor in the agricultural as well as nonagricultural sector. Employment in far-away places or in sectors unlikely to be affected by drought will have a stabilizing effect as such income is less covariate with income in drought-affected areas. In addition to seasonal migration during drought periods, diversification of earning with some family members working permanently in cities helps smooth consumption. A variant of this coping mechanism is the marital relationship with families in far-away places. Income transfers through this mechanism have helped farmers in the semiarid tropics of India to stabilize consumption during drought years (Rosenzweig and Stark 1989). Similarly, diversification of income from the farm to nonfarm sector is a way of exploiting the low covariance for income and consumption stabilization. For example, the proportion of

income derived from nonfarm employment outside the region has been higher in the riskier Sahel zone than in the less risky Sudan zone of Africa (Matlon 1991). Credit can potentially play an important role in smoothing consumption. Credit permits borrowing against future income potential to meet a current consumption shortfall. In a perfectly competitive market, the opportunity cost of credit is equal to the interest on savings. Hence, long-run consumption will not depend on whether savings are used or credit is taken to meet a shortfall in consumption in poor years. In reality, credit markets are imperfect, with the effective interest rate on credit being higher than the interest on savings. Risk aversion among lenders, the high transaction cost of serving a large number of small farmers, and information asymmetry between borrowers and lenders are the major reasons for capital market failure in developing countries (Binswanger and Rosenzweig 1986). As a result, the use of credit for consumption smoothing in developing countries is limited, more so among small farmers who are considered as high-risk borrowers by formal credit institutions. Despite a poorly developed formal market for credit, the available evidence on the extent of consumption smoothing indicates the presence of informal institutional arrangements for risk sharing in rural areas. These may be villagelevel rice banks, local money lenders, mutual self-help groups, interlocked credit and labor markets, and social and family networks. Income transfers (in cash or in kind) through these informal arrangements can provide very effective insurance, especially if the risk affects only a few households (Jodha 1978, Ben-Porath 1980, Platteau 1991, Fafchamps 1992, Townsend 1995). The provision of such insurance is believed to be one of the critical functions of the family as an institution (Rosenzweig 1988). Although very effective in insuring poor households against a consumption shortfall caused by life-cycle events such as death or illness in the family, these mechanisms are less effective in dealing with covariate risks that affect everybody within the community. Historical records of mass migration, starvation,

10

and death attest to the failure of these informal mechanisms when droughts are severe and widespread. These informal arrangements that characterize traditional rural societies also seem to weaken considerably in the face of commercialization and greater exposure to the outside world (Jodha 1978). Publicly sponsored relief programs are used to deal with the failure of these ex post consumption-smoothing mechanisms in the face of large covariate risk. To the extent that food insecurity is due to the lack of exchange entitlements, these relief programs are designed to transfer income to farmers in affected areas to reduce consumption deficits and prevent excessive asset depletion. The relief programs generally take the form of income transfer/ employment generation although direct food distribution may also be a component when drought is severe. Several authors (Corbett 1988, Hay 1988, Dev 1996) have discussed the strengths and weaknesses of various types of relief programs.

Methods for risk analysis
One of the most widely applied models for studying decision making under uncertainty is the expected utility model (Anderson et a1 1977). Under risky situations, decision makers are assumed to select options that maximize the expected utility of probabilistic consequences. To implement the model, it is essential to know the decision makers’ attitudes toward risk and the probability of various outcomes resulting from an action. Attitudes toward risk are captured in the utility function that transforms monetary gains and losses into utility. Risk analysis consists of combining the subjective probability of outcomes and the associated utility to identify risk-efficient decisions. Different methods are available for estimating the utility function and the implied risk aversion coefficient (Binswanger 1980, Binswanger and Sillers 1983,Antle 1983). Two popular specifications used in applied work are the utility function with constant partial risk aversion coefficient and the utility function with

constant absolute risk aversion coefficient. Estimates of both types of risk aversion coefficients have been derived for several farming systems. The required probability distributions can be derived in at least two ways: using historical data and using a predictive simulation model. Both approaches have advantages and disadvantages. While the use of historical data is based on the assumption that the future will be similar to the past, the use of a predictive simulation model requires that the model adequately mimic the real production system. The use of a simulation model to predict the consequences of changes in technical parameters is becoming more popular (Muchow and Bellamy 1991, Lansigan et a1 1997). Using stochastic weather input to drive a suitably validated simulation model, the probability distribution of yield for a specific technical intervention can be obtained. The distribution of yield can then be transformed into the distribution of economic variable (for example, profit), which is then used for economic risk analysis. Anderson and Hazell (I 994, Lansigan et a1 (1997) have discussed the advantages and disadvantages of using simulation models to identify risk-efficient technologies. Once the utility function and probability distribution of income are obtained, several approaches could be used to identify riskefficient decisions. A popular integrative approach is the use of whole-farm planning models. Variants ranging from simple stochastic budgets to discrete stochastic programming models are available (Hardaker et a1 1991). Other approaches include stochastic dominance analysis, nonoptimizing simulation models, and variants thereof.

Objective function specification
The consequences of an action must be assessed in relation to the objective function or what farmers would like to achieve from their fanning activities. While the objective function may include aspects such as quality of life, children’s education, and level of leisure, analysts often focus on economic criteria such as farm income

11

and consumption. If the objective of a farmer is to maintain a given level of consumption, the utility function should be defined in terms of consumption. However, as income and consumption are highly correlated, income is often used as a proxy for consumption. Farmers derive income not only from rice but also from several activities that include growing other crops and nonfarm employment. Income from rice is often a smaller component of their total income, especially in the more marginal environments. Even if rice production is low, farmers may be able to maintain their consumption level by obtaining additional income from other sources. Thus, it is not enough to evaluate rice technologies in terms of how much additional income they can generate from rice. Income from all sources should be considered simultaneously.

Risk benefit and rice research
What opportunities exist for rice research to reduce fluctuations in income and consumption of farmers? What is the size of the economic benefit if rice yield and production could be stabilized? Answers to these questions are critical for designing suitable technological and policy interventions to reduce &he “cost” of risk. Before proceeding further, it is essential to define what we mean by “cost” of risk and develop a device to measure it quantitatively. For this, we use the expected utility model of decision making. The model postulates that, under risky situations, decision makers evaluate decisions in terms of expected utility of income and choose the action that maximizes the expected utility. For risk-neutral decision makers, the decision that maximizes the expected utility is also the decision that maximizes the expected income gain. A riskaverse decision maker, on the other hand, would be willing to sacrifice some income to avoid taking risk. The cost of risk is the amount of income sacrificed to protect or insure against risk. Using the expected utility theory, the cost of risk can be approximated as (Pandey et al 1999). P = 0.5 R [a2 Cr2 + 2 a (1 - a) g Cr Cy]

where P is the cost of risk (or risk deduction) expressed as a proportion of mean income, R is the coefficient of relative risk aversion, Cr is the coefficient of variation (CV) of rice income, a is the share of rice income in total income, Cy is the CV of nonrice income, and g is the correlation coefficient between rice and nonrice income. The proportional risk premium measured in this equation provides an estimate of the cost of risk currently borne by farmers relative to the situation in which the variability of rice income is completely eliminated. As there will always be some variability of rice income that cannot be eliminated, the estimate obtained from this equation can be considered an upper bound value. This indicates several ways through which the economic cost of risk can be reduced: lowering the CV of income from rice, lowering the ratio of rice income to nonrice income, and reducing the correlation of rice income with nonrice income. Stabilization of rice yield through breeding and better crop management can be an important research intervention. The lowering of the share of rice to nonrice income implies crop and income diversification. The scope of technical intervention to achieve this may be somewhat limited in the case of rainfed rice, as waterlogged conditions of the fields limit other cropping options during the rainy season. Farmers have other cropping alternatives only during the postrainy season, provided moisture is nonlimiting. Development of shorter duration varieties in areas where the success of a postrainy- season crop depends on how early it is established can facilitate diversification of crop income. Other options for encouraging crop and income diversification are related to policy interventions such as the development of road, transport, and marketing infrastructure. These policy interventions can also help reduce the correlation between rice and nonrice income by broadening the income base of rural households.

Technology and yield risk
Technical research can basically be classified into two types: plant improvement and crop management. Two risk-related issues involve

12

plant improvement research. The first is the issue of the extent to which improved varieties are more or less risky than traditional varieties. Plant breeders use stability analysis and other forms of genotype x environment (G x E) analysis to assess the stability and adaptability of alternative varieties versus the traditional check. The analysis of G x E interactions has been a topic of interest among plant breeders and powerful tools and methods have been developed (Cooper and Hammer 1996). However, most of these analyses use some statistical notion of “stability” for discriminating among cultivars. Such analysis could be usefully complemented by decision analytical tools such as stochastic dominance analysis to explicitly account for farmers’ risk aversion (Binswanger and Barah 1980, Witcombe 1989). The second is the issue of the extent to which a combination of several varieties reduces risk. Ample evidence shows that farmers grow several varieties of rice simultaneously in rainfed areas (Kshirsagar and Pandey 1995, Pandey and Sanamongkhoun 1998). Although there may be several reasons for doing so, risk reduction through varietal diversification appears to be an important one (Smale et al 1994). The strategy of varietal diversification could potentially be used to reduce the overall risk, even if modem varieties are less stable than their traditional counterparts. Crop management research, on the other hand, is concerned with altering yield risk by manipulating agronomic practices. Agronomic manipulation can reduce the yield risk associated with stress conditions such as drought, flood, and pests. For example, improved nutrient management may help reduce risk by making plants more tolerant of stress such as drought as well as by helping them to recover faster when the stress is relieved (Wade et al 1999). Similarly, options may exist for reducing risk by manipulating timing, placement and quantities of inputs. The study of the quantitative effects of input management on risk remains a major field of inquiry by agricultural economists, among others. Production function specifications that permit estimation of marginal risk effects have

been developed (Just and Pope 1979, Antle 1983). Such production functions have been applied to derive the optimal allocation of several inputs under risky situations. Although attempts have been made to quantify marginal risk effects of several inputs such as fertilizers, irrigation, and pesticides in a range of environments using such a framework, the empirical analyses have often produced somewhat inconsistent results (Roumasset et al 1989, Pandey 1989). An important area of research in the context of crop management technology is the effect of uncertainty on input use in a dynamic context. Instead of committing all inputs at the beginning of the crop season, inputs are used sequentially, with farmers revising the level of input use depending on crop conditions and their expectations regarding stochastic variables such as prices and weather. Possibilities for such dynamic adjustments of input use provide flexibility for efficient risk management. In addition, reliable forecasts of stochastic variables such as weather can improve the efficiency of resource allocation by reducing the level of uncertainty (Byerlee and Anderson 1982, Abedullah and Pandey 2000).

Data needs
Farmers are concerned about risk that manifests itself in the form of unpredictable fluctuations in yield over time. To analyze risk and risk-coping mechanisms, temporal data are hence required. The generation of temporal data, however, is expensive and time-consuming. Plant breeders have partially got around this constraint in their selection program by including several testing locations to capture different environmental conditions. Fortunately, this approach has worked well in the past. However, this kind of spatial substitute for temporal data is less useful when analyzing farmers’ coping mechanisms and in studying how risk influences resource allocation over time through its effect on assets of farm households. Spatial data are not of much help in studying these dynamic elements. The only extensive panel data that have been used widely to study risk and many other aspects of

13

the village economy is the village-level study database generated by ICRISAT (Walker and Ryan 1990). A similar kind of database covering major cropping systems would certainly be very useful for studying responses to risk in other rainfed environments. As indicated in the paper by Pandey et a1 (2000a), some progress is being made in this direction.

Strategies for developing and disseminating risk-reducing technologies
What are the implications of the above discussion for developing and disseminating risk-reducing technologies? Are the strategies and institutional mechanisms likely to be different from those now in place? Space limitations preclude me from going into much in-depth discussion on this topic, which by itself, is very broad. Nevertheless, some comments on this important topic are in order. Based on the above discussion, it can be deduced that the following features of technology help reduce risk: • less input demanding; • lower degree of prior commitments of inputs; • technologies that improve flexibility of decision making; • technologies that use information about conditioning factors; • technologies that help stabilize area, not just yield, as area variability can be an important source of production variability; and • technologies that raise income by improving the productivity of other components of the farming systems (e.g., those that facilitate crop diversification and intensification). These considerations suggest the following strategies for technology development, adaptation, and dissemination for reducing risk: • Emphasis on technologies that reduce yield losses in unfavorable years rather than those that increase yield in favorable years only (downside risk). However, trade-off between yield gain and instability may be inevitable. • More emphasis on developing durable

• •

resistance to pests/diseases and abiotic stresses. Molecular techniques may have a major role to play, especially when dealing with polygenic traits. Development of a “basket” of complementary options in which each component can also stand on its own. Although productivity improvement can be high when several components are combined in the form of a package, such packages also tend to increase risk. By allowing farmers to pick and choose from a complementary basket of options, such an approach makes sequential adoption of the most profitable (and least risky) components possible. More adaptive research and decentralized regional testing for specific adaptation. Specific technologies for each region developed through adaptive research will improve the suitability of such technologies to their target domain and reduce risk. This will also stabilize production by reducing interregional correlation. Such a strategy is more relevant for heterogeneous marginal areas where risk and its effect on the poor tend to be high. technology testing under farmers’ conditions using their own criteria for evaluation; As opportunities for management intervention decrease with the progress of the cropping season, crop management technologies may be more effective to reduce risk in the early part of the growing season. Germplasm-based intervention may be more suitable for late-season risk (such as terminal drought) or for other risk for which no management interventions are yet available. As research strategies differ according to sources of risk that vary across locations, a greater understanding of the nature and magnitude of risk in representative environments is required. Such an understanding is likely to be facilitated by strengthening socioeconomic research capacity in national agricultural research systems. Limited capacity in this field is quite a severe constraint currently.

14

Regarding the approach to extension, the dominant paradigm for extension for a long time has been that based on “transfer of technology (TOT).” This was the approach followed during the Green Revolution and it has met with success mostly in favorable areas. This one-size fits-all approach can actually increase risk as it requires increased outlay on several inputs simultaneously and is also more demanding of farmers’ skills. The TOT approach based on the promotion of technology packages is less suitable for disseminating risk-reducing technologies as it does not facilitate use of information about crop condition. The extension approach needed to disseminate risk-reducing technologies should be based on providing information (as opposed to inputs) to farmers and on empowering them to use it correctly. The extension system needs to provide options rather than recommendations and train farmers to use information as the basis for decisions. It has to be able to provide advice to farmers based on conditioning factors. Conditional recommendations can be complex and may contain several “if-then’’ rules. The usual extension approach designed to provide simple messages and technological packages is not suitable for risk-reducing technologies. Although improved varieties can spread quite effectively from farmer to farmer as witnessed during the Green Revolution, such a mode of transmission is unlikely to be successful for complex information. Research in the context of integrated pest management has shown that farmers are unlikely to adopt such technologies unless they understand some basic principles involved. Thus, the role of extension would be not only to transmit recommendations but also to impart knowledge to farmers so that they can monitor crop and field conditions better and make appropriate adjustments in their crop management practices. However, extension systems in developing countries are mostly funded inadequately. Extension agents are inadequately trained and lack incentives and skills to provide knowledge about crop management options to farmers. Incentives to extension agents are oriented to a

physical target such as distribution of a certain quantity of seeds, fertilizers, and other inputs. These institutional weaknesses of extension systems have been well recognized and various alternatives have been attempted. However, the situation in most cases has changed very little. More efforts are needed to design extension systems so that they can be more effective in promoting the dissemination of risk-reducing technologies.

Crop insurance
The topic of crop insurance remains a highly debated issue with many eminent economists contributing to this debate. A brief review of this debate is contained in the paper by Mishra (2000). The major problems with crop insurance are related to the difficulties in dealing with the moral hazard, adverse selection, and the resulting high cost of publicly sponsored crop insurance programs (Hazell et al 1986). Various innovative concepts such as those based on rainfall lotteries have been developed. However, the cost of implementing such schemes remains a major issue. In the context of poor rainfed rice farmers who are often the primary targets of crop insurance programs, the price they are willing to pay for yield stabilization through crop insurance programs may be too low for the scheme to be viable without subsidy. As the analysis presented later suggests (Pandey et al, 2000a), expansion of rural nonfarm employment opportunities and crop intensification during the dry season in areas with tubewell irrigation have helped increase and stabilize farm household incomes. As a result, the proportional risk premium associated with yield stabilization of rice is substantially lower than what it is normally presumed to be (Pandey et al 1999). In addition, crop insurance schemes based on stabilizing yield will have little value in many rainfed areas where area variability is also quite high. Nevertheless, crop insurance may be of value to regions where agroclimatic conditions and limited infrastructure constrain the operation of other mechanisms for smoothing income flow over time.

15

Concluding remarks
In concluding, managing agricultural risk effectively remains an important challenge to farmers, agricultural researchers, and policymakers alike. Much progress has been made in understanding the likely effect of risk and risk aversion on technology adoption, in developing approaches and methods to quantify risk, and in analyzing its ramifications for national food security and poverty. Despite this, much remains to be done, especially as emphasis on reducing poverty by stimulating agricultural growth in rainfed areas has increased in recent decades. The high degree of temporal and spatial heterogeneity, chronic poverty, and limited linkages with the national economy are some of the major factors that have constrained agricultural growth in these regions. Although some of these rainfed areas are now showing signs of economic dynamism, effective management of agricultural risk in these areas can be an important way of inducing agricultural growth.

References
Abedullah, Pandey S. 2000. Risk and the value of rainfall forecast for rainfed rice in the Philippines. (this volume) Anderson JR, Dillon JL, Hardaker JB. 1977. Agricultural decision analysis. Iowa (USA): The Iowa State University Press. Anderson JR, Hazell PBR. 1994. Risk considerations in the design and transfer of agricultural technology. In: Anderson JR, editor. Agricultural technology: policy issues for international communities. CAB International. p 321-328. Antle JM. 1983. Incorporating risk into production analysis. Am. J. Agric. Econ. 65: 1099-1 106. Ballabh V, Pandey S. 1999. Transitions in rice production system in eastern India: evidence from two villages in Uttar Pradesh. Econ. Pol. Wkly. 34(13):All-A16. Ben-Porath Y. 1980. The F-connection: families, friends, and firms and the organization of exchange. Popul. Dev. Rev. 6: 1-30.

Binswanger HP, Barah BC. 1980. Yield risk, risk aversion and genotype selection: conceptual issues and approaches. ICRISAT Research Bulletin 3, International Crop Research Institute for the Semi-Arid Tropics. Binswanger HP, Rosenzweig MR. 1986. Behavioral and material determinants of production relations in agriculture. J. Dev. Stud. 22:503-539. Binswanger HP, Sillers DA. 1983. Risk aversion and credit constraints in farmers’ decisionmaking: a re-interpretation. J. Dev. Stud. 20(1):5-21. Binswanger HP. 1980. Attitudes toward risk: experimental measurements in rural India. Am. J. Agric. Econ. 62:395-407. Byerlee D, Anderson J. 1982. Risk utility and the value of information in farmer decision making. Rev. Market. Agric. Econ. 50(3):231-246. Cooper M, Hammer GL. 1996. Plant adaptation and crop improvement. Oxford (UK): CAB International. Corbett J. 1988. Famine and household coping strategies. World Dev. 16:1099-1112. Dev SM. 1996. Food security: PDS vs. EGS a tale of two states. Econ. Pol. Wkly. : 17521763. Fafchamps M. 1992. Cash crop production, food price volatility, and rural market integration in the Third World. Am. J. Agric. Econ. 74:1-5. Feder G, Just JE, Zilbennan D. 1985. Adoption of agricultural innovation in developing countries: a survey. Econ. Dev. Cult. Change 33:255-298. Hardaker JB, Pandey S, Patten LH. 1991. Farm planning under uncertainty. Rev. Market. Agric. Econ. 59:9-22. Hay RW. 1988. Famine incomes and employment: Has Botswana anything to teach Africa? World Dev. 16: 1113-1125. Hazell P, Bossoco LM, Arcia G. 1986. A model for evaluating farmers’ demand for insurance: application in Mexico and Panama. In: Hazell P, editor. Insurance for agricultural development: issue and experience. Baltimore, Md. (USA): John Hopkins University Press.

16

Hazell PBR. 1982. Instability in Indian food grain production. IFPRI Research Report No. 30. Washington, D.C. (USA): International Food Policy Research Institute. Jodha NS. 1978. Effectiveness of farmers’ adjustment to risk. Econ. Pol. Wkly. 13:A38-A48. Jodha NS. 1984. Common property resources and rural poor in dry regions of India. Econ. Pol. Wkly. 21(27):1169-1181. Just RE, Pope RD. 1979. Production function estimation and related risk considerations. Am. J. Agric. Econ. 61(2):249-257. Kinsey B, Burger BK, Gunning JW. 1998. Coping with drought in Zimbabwe: survey evidence on responses of rural households to risk. World Dev. 26:89-110. Kshirsagar KG, Pandey S. 1995. Farmers’ choice of rice cultivars in rainfed lowlands of eastern India: a case of Garhmadhupur village. Social Sciences Division, International Rice Research Institute, Los Baños, Laguna, Philippines. Kshirsagar KG, Pandey S, Bellon MR. 1997. Farmer perceptions, varietal characteristics and technology adoption: the case of a rainfed rice village in eastern India. Social Science Division Paper 5/97. International Rice Research Institute, Los Baños, Laguna, Philippines. Lansigan F, Pandey S, Bouman BAM. 1997. Combining crop modeling with economic risk-analysis for the evaluation of crop management strategies. Field Crops Res. 51:135-145. Lim Y, Townsend RM. 1994. Currency transaction patterns and consumption smoothing theory and measurement in ICRISAT villages. Working Paper. International Crops Research Institute for Semi-Arid Tropics, Patancheru, India. Matlon P. 1991. Farmer risk management strategies: the case of West African semiarid tropics. In: Holden D et al, editors. Risk in agriculture. Proceedings of the Tenth Agriculture Sector Symposium. Washington, D.C. (USA): The World Bank.

Mehra S. 1981. Instability in Indian agriculture in the context of the new technology. Washington, D.C. (USA): International Food Policy Research Institute. Mishra, PK. 2000. Crop insurance: a policy perspective. (this volume) Muchow RC, Bellamy JA. 1991. Climatic risk in agricultural Production: models and management for the semi-arid tropics and subtropics. Oxford (UK): CAB International. Naylor R, Falcom W, Zabaleta E. 1997. Variability and growth in grain yields, 19501994. Popul. Dev. Rev. 23:41-58. Newbery DMG, Stiglitz J. 1979. Sharecropping, risk sharing and the importance of inperfect information. In: Roumasset JA, Boussard JM, Singh I, editors. Risk, uncertainty and agricultural development. New York (USA): Agricultural Development Council. p 311340. Otsuka K, Chuma H, Hayami Y. 1992, Land and labor contracts in agrarian economies: theories and facts. J. Econ. Lit. 30:19652018. Pandey S. 1989. Irrigation and crop yield variability a review. In: Anderson JR, Hazell PBR, editors. Variability in grain yield: implication for agricultural research and policy in developing countries. Baltimore, Md. (USA): Johns Hopkins University Press. Pandey S, Sanamongkhoun M. 1998. Rainfed lowland rice in Laos: a socioeconomic benchmark study. International Rice Research Institute, Los Baños, Philippines. Pandey S, Singh HN, Villano RA. 1999. Rainfed rice and risk-coping strategies: some microeconomic evidences from eastern India. Selected paper for presentation at the Annual Meeting of the American Agricultural Economics Association, 8-11 August 1999, Nashville, Tennessee, USA. Pandey S, Sing HN, Villano RA. 2000a. Rainfed rice and risk-coping stragegies: some microeconomic evidence from eastern Uttar Pradesh. (this volume)

17

Pandey S, Behura DD, Villano RA, Naik D. 2000b. Economic cost of drought and farmers’ coping mechanisms: a study of rainfed rice systems in eastern India. Discussion Paper Series No. 39. Los Baños (Philippines): International Rice Research Institute. 35 p. Paxson C, Chaudhuri C. 1994. Consumption smoothing and income seasonality in rural India. Manuscript. Platteau J. 1991. Traditional systems of social security and hunger insurance: past achievements and modem challenges. In: Ahmad E., Dreze J, Hills J, Sen A, editors. Social security in developing countries. Oxford: Clarendon Press. Rao CHH. 1968. Fluctuation in agricultural growth. Econ. Pol. Wkly. (3):87-94. Reardon T, Delgado C, Matlon P. 1988. Coping with household-level food insecurity in drought-affected areas of Burkina Faso. World Dev. 16:1065-1079. Reardon T, Delgado C, Matlon P. 1992. Determinants and effects of income diversification among farm households in Burkina Faso. J. Dev. Stud. 28:264-296. Rosenzweig MR, Stark 0. 1989. Consumption smoothing, migration and marriage: evidence from rural India. J. Pol. Econ. 97:905-926. Rosenzweig MR. 1988. Risk, implicit contracts and the family in rural areas of low income countries. Econ. J. 98: 1148-1170. Rothschild M, Stiglitz JE. 1970. Increasing risk. I. Definition. J. Econ. Theor. 2(3):225-243. Roumasset JA, Rosegrant MW, Chakravorty UN, Anderson JR. 1989. Fertilizer and crop yield variability: a review. In: Anderson JR, Hazell PBR, editors. Variability in grain yields: implication for agricultural research and policy in developing countries. Baltimore, Md. (USA): John Hopkins University Press. Roumasset, J.A., Boussard, J-M and Singh, I. 1979. Risk, Uncertainty and Agricultural Development. Agricultural Development Council, New York.

Sandmo A. 1971. On the theory of competitive firm under price uncertainty. Am. Econ. Rev. 61(1):65-73. Siddiq EA, Kundu DK. 1993. Production strategies for rice-based cropping systems in the humid tropics. In: Buxton DR et al, editors. International Crop Science 1. Crop Science Society of America. Singh HN, Singh JN, Singh RK. 1995. Risk management by rainfed lowland rice farmers in eastern India. In: Fragile lives in fragile ecosystems. Proceedings of the International Rice Research Conference, 13-17 Feb 1995, Los Baños, Laguna, Philippines. Manila (Philippines): International Rice Research Institute. p 135- 148. Singh IJ, Byerlee D. 1990. Relative variability in wheat yields across countries and overtime. J. Agric. Econ. 41:21-32. Smale M, Just RE, Leathers HD. 1994. Land allocation in HYV adoption models: an investigation of alternative explanation. Am. J. Agric. Econ. 76:533-546. Townsend RM. 1994. Risk and insurance in village India. Econometrica 62:539-592. Townsend RM. 1995. Consumption insurance: an evaluation of risk-bearing systems in low-income economies. J. Econ. Perspect. 9:83-102. Wade LJ, McLaren CG, Quintana L, Harnpichitvitaya D, Rajatasereekul S, Sarawgi AK, Kumar A, Ahmed HU, Sarwoto, Singh AK, Rodriguez R, Siopongco J, Sarkarung S. 1999. Genotype by environment interaction across diverse rainfed lowland rice environments. Field Crops Res. 64:35-50. Walker TS, Ryan JG. 1990. Village and household economies in India’s semi-arid tropics. Baltimore, Md. (USA): Johns Hopkins University Press. Walker TS, Jodha NS. 1986. How small farm households adapt to risk. In: Hazell P, Pomareda C, Valdes A, editors. Crop insurance for agricultural development: issues and experience. Baltomore. Md. (USA): Johns Hopkins University Press: p 17-34.

18

Walker TS. 1989. High-yielding varieties and variability in sorghum and pearl millet production in India. In: Anderson JR, Hazell PBR, editors. Variability in grain yields. Baltimore, Md. (USA): Johns Hopkins University Press. Witcombe JR. 1989. Variability in the yield of pearl millet varieties and hybrids in India and Pakistan. In: Anderson JR, Hazell PBR, editors. Variability in grain yields: implications for agricultural research and policy in developing countries. Washington, D.C. (USA): International Food Policy Research Institute.

Answer:

Discussion
Question: We are looking for the reasons of instability in rice production. Many factors affect productivity at the farm level. Is it necessary that we should examine all these factors for the management of risk? While the output is determined by a large number of deterministic factors (such as soil type), stochastic factors (such as rainfall) and farmers’ management (such as input use), the purpose of risk analysis is to understand how stochastic factors interact with deterministic factors in generating a particular management response. As risk is pervasive, it can affect many decisions of farmers. For pragmatic reasons, one would attempt to tackle only those aspects that are likely to affected more significantly by risk considerations. We have to measure risk in terms of the lower half (negative deviations from trend) of the probability distribution. A subsistence farmer is concerned more about the number of years of crop failure in a time period. But in terms of profitability, more than average returns may be received during years of favorable price changes. So

Answer:

there may be value in examining variations in both the directions of the mean. In terms of measurements of risk, there are plenty of concerns regarding the use of covariances and coefficients of variation. What are your views on the suitable measures of risk? The theoretical analysis on risk shows clearly that risk can be measured only by examining the whole probability distribution along with the decision maker’s utility function. Summary statistics such as variance and coefficient of variations are merely statistical descriptors of variations around the mean. These descriptors ignore the higher moments of distribution and, as a result, they can only tell a part of the story except under very restrictive assumptions about the nature of probability distribution and utility function. Although we often talk about technologies being “risk-increasing” or “riskdecreasing”, a theoretically more satisfactory approach would to be consider them as “risk-efficient” or “risk-inefficient”.

Notes
Author’s address: International Rice Research Institute, MCPO Box 3 127, Makati City 127 1, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

Question:

19

Risk and rainfed rice in India: an overview
C. Ramasamy and K. Uma

Rainfed rice production in eastern India is inherently risky due to high climatic variability. Rice yield fluctuates widely depending on the vagaries of the monsoon. Although variability in rice output is relatively less important in India, uncertainty about prices of inputs and their timely availability is also an important source of risk. This paper provides an overview of empirical studies that have addressed the issues of risk, risk aversion, and their effects on technology adoption in the context of rainfed rice systems of India. Opportunities for reducing risk through improved technologies and policy reforms are indicated. Agricultural decision making, whether at the sector, farm, or project level. is invariably characterized by considerable risks. Agricultural production is inherently a risky activity in rainfed areas where agroclimatic conditions largely dictate the decisions of farmers. Risks may arise from many sources. but they emerge primarily from the variability inherent in natural, climatic, and biological systems. Risk is particularly a pervasive phenomenon in the agriculture of developing countries and deserves careful attention both for policy formulations and for planning and appraising projects that aim to improve the lot of people in these countries (Scandizzo et al 1984). Eastern India accounts for 67% of the total rice area and about 56% of the total rice output of the country. In spite of the region’s importance, rice yield has remained low and highly variable. Climatic risks are the major reasons for such low and fluctuating yields (Widawsky and O’Toole 1990). Both droughts and floods are endemic to major areas, at times occurring in the same field in the same year. Because of the high risk involved. farmers are uwvilling to invest in yield-increasing technologies that require a greater cash outlay for inputs such as chemical fertilizers. A considerable amount of literature has appeared on risk in agriculture (Lipton 1979, Binswanger 1980. Anderson and Hamal 1983, Martin and McLeay 1998). Many studies have dealt with farmers’ perceptions of risk and their innate aversion to risk. Some of the key areas that have been analyzed are risks associated with Green Revolution technologies, rainfall variability, farm credit. agricultural subsidies, farm diversification, irrigation, crop insurance, agricultural instability and poverty, and impact of drought and risk management. This paper reviews some of the studies related to risk and risk management in rainfed rice with special focus on eastern India.

Sources of risk
In eastern India, weather conditions in a season emerge as the dominant source of risk in agriculture. Inadequate and unevenly distributed rainfall, droughts and floods, extremes of cold and hot temperatures. cyclones, and storms result in heavy yield losses. Weather fluctuations can result in cyclical movements in output and cause an erosion of the real value of financial investment in agriculture. These factors lead to underperformance of the agricultural sector (Ramachandra 1983). Recent studies suggest that agricultural output has become more sensitive to rainfall in the post-Green Revolution period, with elasticity of output (with respect to rainfall) increasing for most crops. This may be brought about by the

21

strong complementary relationship between the use of modem varieties and inputs and the availability of moisture through either rainfall or irrigation (Rao et a1 1988). Rainfall indices for crops, constructed as a proxy variable for measuring the effect of weather on area and output of different crops at the all-India level, have consistently shown that a significant part of the fluctuation in area and output around their respective trend lines could be explained by the computed rainfall indices (Ray 1983). Data on rainfall and food production in Orissa ( 1958-59 to 1978-79) indicate that the years of lowest production are also the years with deficit rainfall up to 30-35%. Similarly, in Punjab, the lowest production is associated with a relatively large deficit in rainfall in spite of a high percentage of area being under irrigation in the stare (Singh 1981). The impact of weather risks on agricultural production in India, where 70% of gross cropped area is still rainfed, underscores the need for systematic efforts toward improving our prediction of weather movements and our estimation of losses to the economy in terms of output, employment, etc. Another potentially important source of risk is the market price of inputs and outputs. Agricultural product prices are generally regarded as stochastic. Fluctuations in these prices can be a major source of variability in farmers’ incomes. Production risk due to weather also contributes to price fluctuations indirectly-thus,weather-induced risks and price risks are linked. Risk may also arise from unpredictable changes in consumer taste and policy. There is a growing recognition that the political framework within which support prices are determined can also lead to uncertainty (MacLaren 1983). Several supply response studies in agriculture have shown that price risk influences production (Gour 1978, Ramasamy 1979). Price instability and the resulting variations in income can affect allocation of land to various crops and use of fertilizers (Singh and Walker 1984). Price instability may similarly discourage the adoption of new production technologies. The wide swings in area of some commercial crops and vegetables can be attributed to large variations in 22

their prices. However, rice in India is protected through price support programs. The administered prices being implemented for the past several decades for rice had almost stabilized prices. The variation in rice prices likely to be induced by demand factors is insulated by the operation of a public distribution system. Overall, the uncertainty in the price of rice is not an important element of risk in India. This situation may change, however, with the opening of the export market under World Trade Organization arrangements. Unless globalization of agriculture is accompanied by policy reforms to minimize the transmission of price shocks from international to domestic markets, farmers may experience a greater price uncertainty. Timely availability of critical inputs such as fertilizers and plant protection materials is a prerequisite not only for encouraging adoption of improved production technologies but also for sustaining agricultural production. The nonavailability of crucial inputs is often considered to be a factor contributing to high yield gaps in rainfed areas (Pandey and Ashok Kumar 1993). An erratic power supply to agriculture can similarly increase risk in farm production (Ray 1983). The performance of institutions to provide inputs, credit, extension, and other support services has also influenced yield variability (Nadkami and Deshpande 1982). In areas where rainfed rice is supplemented by water from irrigation tanks such as in southern India, poor management of the irrigation system can destabilize food-grain production (Kandaswamy 1988, Sasmal 1993). Tanks probably constitute the oldest means of providing irrigation water in these areas. In recent years, tanks have become an unreliable source of irrigation water supply. and farmers consequently face greater risks and uncertainties (Randhir and Krishnamoorthy 1993). In the same manner. unavailability of goodquality inputs at critical times can increase risk to farmers. It is common knowledge that many Indian farmers complain about the poor quality of fertilizers and pesticides. Similarly. Uncertainty in labor supply and the use of unskilled labor in large numbers can have a risk-

increasing effect (Sasmal 1993). Many farmers, especially those with small holdings, do not get loans linked to new methods because of higher risks perceived to be associated with new technology (Anderson and Hamal 1983). Rokaya (1979) found that 85% of small farmers and 33% of large farmers in Nuwakot District of Nepal did not use agricultural credit because of the higher risks that they associated with the new technology.

Risk and adoption of new technology
The effects of risk, risk aversion, and risk perception on technology adoption have been widely studied. Risk and uncertainty in production have been recognized as an important constraint to the rapid adoption of high-yielding varieties (HYVs) (Sasmal 1993). Innovative methods of cultivation often introduce more uncertainty than traditional ones. If these innovations are adopted, they must not only bring higher average returns than the older methods; they should also reduce the risk (Low 1974). Adoption may be constrained if farmers are risk-averse or perceive a new technology to be riskier than traditional technology. Some evidence points to production instability accentuated by factors associated with the new agricultural technology (Mehra 1981, Hazell 1982). The shift from traditional crop varieties with a diversified genetic base to HYVs with a narrow, common genetic base might have contributed to greater instability. Covarjate movements in yields resulting from instability in input supply could have also led to the increased variability of aggregate output. Similarly, the substitution of chemical fertilizers for organic manures may have aggravated instability through potentially adverse effects on soil health (Ninan and Chandrashekar 1993).

production will be less risky (Anderson and Hamal 1983, Holden et al 1991). However, farmers may err in judgment since risk perceptions depend, in part, on their knowledge and experience (Anderson and Hamal 1983). Sasmal (1993) considered, along with weather conditions, the nature of irrigation and soil conditions, and the farmer’s efficiency in explaining output variability. Ranganathan and Ramasamy (1993) reported farm-level efficiencies of rice production in semidry conditions of Tamil Nadu. They found significant differences in resource use efficiency across farms, which can largely be ascribed to variations in farmers’ efficiencies. Their decisions on adjustments in agricultural operations, sequences, timings, and input combinations are crucial to achieving higher production efficiency. Singh and Zilberman (1984) reported inefficiency in land and fertilizer allocation by Punjab farmers. The existing farm plans were far from the efficient mean-variance frontier, implying that the crop plans followed by farmers were inefficient with regard to the minimum risk portfolio. To ensure subsistence, farmers with low levels of resources allocated most of these in a conservative manner. Farmers with higher levels of resource availability, on the other hand, allocated a greater proportion of resources to activities that increase their average income.

Risk management
Farmers use both formal and informal strategies to cope with risk in agriculture, but their reliance on formal strategies has been growing (Rustagi 1988). Several studies have suggested the use of farm diversification as a risk precaution, though the prime objective may be one of profit maximization. Under a situation of risk and capital constraints, diversification stabilizes farm income at a higher level (Gupta and Tewari 1985). Intercropping often provides insurance against complete crop failure (Singh and Walker 1984). Livestock are similarly used as a buffer against risk and uncertainty. Improvements in irrigation and marketing facilities and investment in soil conservation can reduce yield and price risks (Singh 1989). 23

Farmers’ behavior and efficiency
Risk also affects decisions on varietal selection, agronomic practices, crop combinations, input management, and marketing of products. Some inputs-such as fungicides and pesticides-are overtly risk-reducing in their effects. If such inputs are used wisely and in a timely manner,

Improved storage facilities, including on-farm storage, likewise facilitate the carryover of stocks from one period to another and help smooth consumption (Sahn and Von Braun 1986). Formal strategies include programs such as crop insurance, input subsidy, output price support, credit with variable amortization schemes, income equalization schemes, and disaster assistance schemes. Crop insurance, though a risk-reducing strategy, has several limitations. Public intervention in crop insurance may require subsidies in the initial phases. The justification of a public subsidy for crop insurance comes from the proposition that, if farmers are less risk-averse, they would expand their production and this, in turn, would increase total production and add substantially to social gains (Hazell et a1 1986). Walker and Jodha (1986), however, point out that crop insurance might simply provide a costly substitute for the existing private risk-sharing arrangements.

Conclusions
Risk will continue to be an important factor in the rainfed agriculture of India. Several interventions are needed to reduce risk as well as its impact on farmers’ welfare. These interventions include both agricultural research and policy reforms. On the agricultural research front, there is a need to better understand the sources of changes in variability of yield and area in the different agroecosystems of India. Research to develop improved varieties that have increased tolerance to various biotic and abiotic stresses must be strengthened. This may be achieved by • shortening the life cycle of cereal crops and reducing their sensitivity to seasonal signals such as daylength; • introducing hardiness into crop varieties to enable them to withstand drought, cold, heat, and other climatic stresses; • incorporating genetic resistance to pests and diseases (Holden et a1 1991); • introducing “competitiveness” in varieties to enable them to survive in marginal environments; and

improving the physiological efficiency of the rice crop in a given agroclimatic condition. Similarly, cost-reducing strategies that focus on economizing and substituting for external resources must be implemented-forinstance, legumes that are important sources of nitrogen can be used instead of synthetic fertilizers, and resource (soil and water) conservation and shifting to sustainable alternatives (natural farming, watershed management, biological methods of pest control and nutrient use) can help reduce risk. Better agronomic management of crops is essential to facilitate rapid crop recovery following stress periods and to improve input use efficiency. The economics of such practices, however, needs careful evaluation. On the policy front, provision of special credit facilities to drought-prone and rainfed areas in the form of production and consumption loans may help farmers minimize the impact of risk. Development of labor-intensive technological packages for lagging regions and crops supported by land reform would strengthen farmers’ ability to manage risk. Other risk management methods that require policy support include maintaining food buffer stocks, food subsidies, relief employment, and food for work programs (Hazell 1986, Rao et a1 1988). Price stabilization schemes and financial innovations such as commodity-linked financing (as a hybrid instrument) can similarly help reduce risk. However, the likely effectiveness of such innovative mechanisms needs to be carefully evaluated in the context of eastern India.

References
Anderson JR, Hamal KB. 1983. Risk and rice technology in Nepal. Indian J. Agric. Econ. 38(2):217-222. Binswanger HP. 1980. Attitudes toward risk: experimental measurement in rural India. Am. J. Agric. Econ. 62(3):395-407. Gour SS. 1978. Impact of risk and technology on supply response of wheat in Madhya Pradesh. Agric. Situation India 38(8):303306.

24

Gupta RP, Tewari SK. 1985. Factors affecting crop diversification: an empirical analysis. Indian J. Agric. Econ. 40(3):304-309. Hazell PBR. 1982. Instability in Indian foodgrains production. Research Report No. 30, Washington, D.C. (USA): International Food Policy Research Institute. p 5-47. Hazell PBR. 1986 Summary proceedings of a Workshop on Cereal Yield Variability book. Washington, D.C. (USA): International Food Policy Research Institute. p 15-47. Hazell P, Pomareda C, Valdes A. 1986. Crop insurance for agricultural development: issues and experience. Baltimore, Md. (USA): The Johns Hopkins University Press. Holden D, Hazell P, Pritchard A. 1991, Risk in agriculture. Proceedings of the Tenth Agriculture Sector Symposium. The World Bank, Washington, D.C., USA. Book review in Indian J. Agric. Econ. 48(4):159. Kandaswamy A. 1988. Commercial crops in India. Indian J. Agric. Econ. 43(3):445-450. Lipton M. 1979. Agricultural risk, rural credit, and the inefficiency of inequality. In: Roumasset JA, Boussard JM, Singh I, editors. Risk, uncertainty and agricultural development. Southeast Asian Regional Centre for Graduate Study and Research in Agriculture, Laguna, Philippines, and Agriculture Development Council, Inc., New York, USA. Low ARC. 1974. Decision taking under uncertainty: a linear programming model of peasant farmer behaviour. J. Agric.Econ. 25(2):311-319. MacLaren D. 1983. The output response of the risk averse firm: some comparative statistics for agricultural policy. J. Agric. Econ. 34(1):45-56. Martin S, McLeay F. 1998. The diversity of farmer’s risk management strategies in a deregulated New Zealand environment. J. Agric. Econ. 49(2):218-233. Mehra S. 1981. Instability in Indian agriculture in the context of the new technology. Research Report No. 25. Washington, D.C. (USA): International Food Policy Research Institute. Nadkarni MV, Deshpande RS. 1982. Agricultural growth, instability in

productivity and rainfall: case of Karnataka. Econ. Pol. Wkly. 17( 52):127-134. Ninan KN, Chandrashekar H. 1993. Green revolution: dryland agriculture and sustainability insights from India. Econ. Pol. Wkly. 28 (12, 13):20-27. Pandey RK, Ashok Kumar. 1993. Economic study of agricultural wages, output and productivity in Orissa. Indian J. Agric. Econ. 48(3):477-481. Ramachandra V. 1983. Investment, growth, and weather fluctuations in India. Indian J. Agric. Econ. 38(3):166-179. Ramasamy C. 1979. An analysis of supply-price relationship of rice in Tamil Nadu: a micromacro approach. PhD dissertation submitted to Tamil Nadu Agricultural University, Coimbatore, India. Randhir TO, Krishnamoorthy S. 1993. Optimal crop planning under production risk in tankfed South Indian farms. Indian J. Agric. Econ. 48(4):678-685. Ranganathan C, Ramasamy C. 1993. Measuring farm-level efficiency in rice production in Tamil Nadu. Research Paper, Department of Agricultural Economics, Tamil Nadu Agricultural University, Coimbatore, India. Rao HCH, Ray SK, Subbarao K. 1988. Unstable agriculture and drought : implications for policy. New Delhi (India): Vikas Publishing House Pvt. Ltd. 192 p. Ray SK. 1983. An empirical investigation of the nature and causes for growth and instability in Indian agriculture, 1950-80. Indian J. Agric. Econ. 38(4):459-474. Rokaya CM. 1979. Impact of a small farmer’s credit programme on farm output, net income and the adoption of new methods: a Nepalase case study. MSc thesis. University of New England, Armidale, Australia. Rustagi NK. 1988. Crop insurance in India: an analysis. Delhi (India): Publishing Corporation. Sahn DE, Von Braun J. 1986. Yield variability and income, consumption, and food security. In: Hazell P, editor. Summary Proceedings of a Workshop on Cereal Yield Variability. Washington, D.C. (USA): International Food Policy Research Institute. p 87-110.

25

Sasmal J. 1993. Considerations of risk in the production of high-yielding variety paddy: a generalised stochastic formulation for production function estimation. Indian J. Agric. Econ. 48(4):694-701. Scandizzo P, Hazell P, Anderson J. 1984. Risky agricultural markets: price forecasting and the need for intervention policies. A Westview Replica Edition, Boulder, Colorado, USA. Singh C, Zilberman D. 1984. Allocation of fertilizer among crops under risk: a quadration. Indian J. Agric. Econ. 39:77-83. Singh D. 1981. Imbalances in agricultural growth. Indian J. Agric. Econ. 36(1): 1-26. Singh IJ. 1989. Agricultural instability and farm poverty in India. Indian J. Agric. Econ. 44(1):1-15. Singh RP, Walker TS. 1984. Crop failure in the semi-arid tropics of Peninsula India: implications for technology policy. Indian J. Agric. Econ. 39(1):29-39. Walker TS, Jodha NS. 1986. How small farm households adapt to risk. In: Hazell P, Pomreda C, Valdes A, editors. Crop insurance for agricultural development: issues and experience. Baltimore, Md. (USA): Johns Hopkins University Press. p 17-34. Widawsky DA, O’Toole JC. 1990. Prioritizing the rice biotechnology research agenda for eastern India. Research Report for the Rockefeller Foundation.

Discussion
Comment: There are some more risk management strategies like migration to minimize reduction in consumption in the years of complete crop failure. This strategy is also followed in rainfed rice-growing areas of eastern Madhya Pradesh. On the sources of risk, it is important to consider risk induced by mismanagement of resources and breakdown of several institutions managing the resources, e.g., deterioration of irrigation tanks in Tamil Nadu. 26

Question: Can you just give us some idea about measurable indicators of risk and risk management strategies under different risk environments? Answer: I have just touched upon the key themes, and to address these issues, I think, one has to examine the available literature critically. Although agricultural researchers give much emphasis on climatic risk, risk arising from institutional failures can be very important. This aspect of risk is not well-addressed in the current literature. Regarding the indicators of risk, I agree with the comments made by Dr. Pandey earlier. Question: Is it possible to distinguish between perception of risk versus perception of the quality of a specific technology? For example, if a farmer says “I do not want some technology”, does this mean that the technology is not good enough, or the risk associated with adoption of technology is too high? Answer: Perhaps both. It is, however, not easy to separate these effects clearly as farmer responses are generated by the interaction of both of these factors. Nonadoption may be because the expected returns are low, or because risks are just too high. It may also be because a farmer faces other constraints to adoption such as lack of capital. A modeling framework is needed to disentangle the relative importance of these individual effects. Comment: Risk analysts often restrict their attention to certain artificial boundaries of privately owned resources. Risk management, particularly in rainfeed ecosystems. should also be looked at in terms of collective management actions. For example, if a large area is floodaffected, it is important to see the economics of its collective management. As regards risk and

technology adoption, one should evaluate technology performance in terms of farmers’ goals and under farmer conditions. Comment: We should also devote some time to discuss some macro aspects of risk management like crop insurance, institutional credit in the context of natural disasters, technological interventions (breeding strategy, crop and input management), and associated costs or trade-off. Comment: The issues like crop insurance involve some institutional aspects and it is not clear who the implementing agencies are. Since 1991, economic scenario has experienced tremendous changes with the emergence of the private seed sector. We should discuss some of these institutional issues, and the role of national and state governments and private financial institutions. Comment: It is unrealistic to think that a particular technology will have uniform and wider adoption without adaptive efforts by researchers, extension workers, or farmers. It would be better if we have some research strategy for a particular environment, as we are doing for rainfed lowlands. Some technologies like pest resistance, changing crop cycle or duration are pretty cost -(trade-off) neutral. But reemployment of energy flows in plants may involve substantial cost. Comment: Our main concern here is to identify the nature of risk and its management strategy in the context of rainfed rice in eastern India. The risk mainly arises from erratic rainfall, inadequate irrigation or moisture stress, salinity and stagnation of water. We should, therefore, concentrate on minimizing these risks with the help of technological development and appropriate policy actions.

Comment: Implementation of crop insurance is a real problem in areas with low and uncertain crop yield. The problems relate to paucity of data to determine the threshold yield level, basis of cost sharing by the national and state governments and efficiency of implementing agencies. In addition, there are moral hazards like excess claim of crop failure (as observed in the case of institutional loans), and involvement of huge subsidies that cast shadows on viability of the scheme. It would be better if other policy options are explored to safeguard against risk. Comment: A successful crop insurance scheme may first be implemented in better endowed areas and subsequently to other low productive areas. Maybe one should think of some modified crop insurance approach. As regards price risk and crop failure linked to suicide by farmers, we need some more evidences. Response: High crop losses or complete crop failures reported recently are partly due to changes in the behavior of insects-pests. When farmers apply high doses of pesticides, pests migrate to other host crops, which also suffer. The problem is further aggravated by high prices and poor quality of pesticides supplied by traders who often lend money to farmers. As a result, farmers not only lose the kharif crop but also fail to ensure the subsequent crop, resulting in cumulative indebtedness.

27

Notes
Authors’ address: Tamil Nadu Agricultural University, Coimbatore, India. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 2 1 -23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

28

Decomposition of income variability in rainfed areas: the case of rice in eastern India
B.C. Barah

Panel data from 130 districts in five rainfed states in eastern India for 1981-93 are used to decompose revenue variability into its yield and price components. This period witnessed various degrees of improvement in rice productivity in eastern India. The decomposition analysis shows that, out of 91 rainfed districts, the yield component exceeded the price component in 26 districts, whereas the price component exceeded the yield component in 38 out of 39 irrigated districts. It is inferred that the production-relevant infrastructure reduces yield variability in the well-endowed region and the price component of risk becomes stronger. In the case of rainfed areas that are also less endowed with infrastructure, yield variability was found to be the dominant source of income variability. Regionally differentiated stabilization policies that address major sources of risk in each region are hence needed.

A low level of yield, high variability. and susceptibility to erratic climatic conditions are the major characteristics of agricultural production systems in rainfed areas. These systems are subjected to extreme income fluctuations from year to year. Apart from temporal variations, interregional variations are also quite substantial. Crops in these areas suffer from drought, submergence, and often both in the same season. In addition to weather uncertainty, other sources of risk such as those arising from inefficient pricing policies may also amplify the effect of production risk. A good understanding of the component sources of risk and their interactions is essential for interventions such as price stabilization, crop insurance, buffer stock policies, and infrastructure development. The objective of this paper is to provide estimates of the relative shares of price and yield components in the total variability of gross revenue of rice. Agricultural land is classified into rainfed or irrigated mainly in terms of availability of irrigation (IFPRI/NCAP 1998). Appendix 1 presents the classification of various districts in eastern India according to irrigation status. Out of 130 major rice-producing districts in eastern

India, 39 are considered irrigated as they have more than 40% of the rice area irrigated. Table 1 shows the average rice yield and its variability. Although the region is predominantly rainfed, irrigation facilities have developed in a few of the districts in recent years. As a result, the average yield in irrigated districts is higher than that in rainfed districts. Furthermore, instability measured by the coefficient of variation is lower in the irrigated districts.

Method of analysis and data
Taylor’s series expansion) of the variance of crop income is used. The method is simple because it involves only a single crop (rice). The advantage is that it allows the measurement of variability and its components across different agoclimatic environments in a comparative fashion. It decomposes total variability into a price component, a yield (productivity) component, and price-yield interactions. An area component can also be incorporated into the decomposition procedure but is not considered in this analysis, as its variation is almost negligible.

A simple method of decomposition (based on

29

Table 1. Average yield of rice and its variability in eastern India, 1966-93. Rainfed districts State kg ha-1 Bihar Orissa West Bengal Madhya Pradesh Uttar Pradesh 1,902 1,870 2,998 1,458 2,291 Yield CV 19 21 29 42 36 Irrigated area (%) 13 25 13 4 12 Irrigated districts Yield kg ha-1 2,530 3,405 3,108 2,755 2,723 cv 16 13 12 28 25 Irrigated area (%) 80 54 58 73 69

Consider the following data set consisting of district “i” and year “t,” where i = 1,2, . . . , L (district subscript) t = 1,2, . . . , T (year subscript) Q = district-level yield of rice P = farm prices The data series on Q and P are detrended appropriately to adjust for technical change as well as inflation so that their variations reflect the pure variability effect only. M is the per unit area revenue from rice, then Mt = PtQt (1)

where CV(Q) = coefficient of variation of yield component, CV(P) = coefficient of variation of price component, CV(PQ) = coefficient of variation of interaction between P and Q. The individual components are standardized and expressed in percentage form. That is, C_Q% = 100 CV(Q)/{(CV(Q) + CV(P) + CV(PQ)} and C_P% = 100 CV(P)/{(CV(Q) + CV(P) + CWQ)} (4) The measurement of risk in crop production should ideally be based on farm-level data. Farm records are, however not available for a sufficiently long period of time. District-level data were hence used for the analysis.

By suppressing the subscript t, we can write the variance of M as’ V(M) = p2 V(Q) + q2 V(P) + pqCov(PQ) + R

Results and discussion
Rice yields have increased over time for most districts. As a result, several districts have moved from a lower to a higher class of yield category (Table 2). The arrows indicate the direction of the mobility of the districts from one productivity class to another. The mobility of the district from a lower level to a higher echelon of productivity could be considered as a reward to efficient research and development efforts and other strategic policies. West Bengal has shown a noticeable mobility by improving rice productivity in all the districts generally.

(2)

Lower-case letters denote the corresponding averages. The analysis is based on gross revenue as data on input costs, by districts, are not available. For the sake of simplicity, the CV-type measure of V(M) will be used to quantify yield risk in each district. CV(M) ~ CV(Q) + CV(P) + CV(PQ) = (3)

1

For details of the approximation of the formula, see Goldberger (1970) and Burt and Finley (1964).

30

Table 3. Districts with different yield and price components. Rainfed districts No. of districts analyzed Districts with - Price components exceeding yield: P>Q Yield components exceeding price component (Q>P) Indifferent price and yield components (P 91 Irrigated districts 39 38

26

1

-

Q)

65

However, compared with Uttar Pradesh and West Bengal, productivity gains in Orissa, Madhya Pradesh, and Assam are almost negligible. The introduction of structural reforms and vigorous implementation of several strategic policies in West Bengal and Uttar Pradesh in recent years have helped enhance productivity as compared with other states in the region. Out of 91 rainfed districts, the yield component exceeds the price component in 26 districts, whereas the price component exceeds the yield component in 38 out of 39 irrigated districts (Table 3). This indicates that the price component is the dominant source of revenue variability in irrigated areas and that the yield component is the dominant source in rainfed areas. Overall, the average of the yield component for rainfed districts was 23.6%; the

corresponding value for irrigated districts was 20.2%. The results show that the importance of yield variability is higher in rainfed districts than in irrigated districts. Figures 1-4 summarize the relationship among the yield, price, and yield-price interaction components. The vertical axis represents the yield component and the horizontal axis, the price component. For districts above the 45" line from the origin (AB in the figure), the contribution of the yield component exceeds that of the price component, whereas for districts below the line, the reverse applies. For those lying on or near the line, the sum of the contributions of the yield component and price component is exactly equal to the total variability, that is the interaction term is zero. The districts lying above the line (i.e., the ones with the yield component exceeding the price component) are usually less developed and have a poorer infrastructure base. The results indicate the need for differential policy interventions for price-variability- and yield variability-dominated districts. Productionrelevant infrastructure such as irrigation has reduced yield variability in well-endowed regions, making the price-variability component stronger. The reverse is true for rainfed districts. Pricing policy incentives are more relevant in districts where yield variability is relatively low; crop insurance (Mishra, 2000) or other yield stabilization policies are relevant in districts 31

32

where yield variability dominates. The minimum support pricing (MSP) policy and the bonus price (some state governments provide an incentive price over and above the MSP announced by the central government from time to time) are more beneficial to market-oriented farmers. However, such price incentives have little relevance to subsistence farmers who are normally subsistence-oriented.

References
Burt OR, FinIey RM. 1964. Statistical analysis of identities in random variables. J. Polit. Econ. 50:734-744. Goldberger A. 1970, Criteria and constraints in multivariate regression. Wisconsin University, Social Systems Research Institute, Madison, Wisconsin, USA. IFPRI/NCAP. 1998. Sustainable rainfed agricultural research in India. Draft report of the IFPRI/NCAP project. Washington, D.C., USA:IFPRI.

Conclusions
Farmers in rainfed areas face a much higher yield risk than those elsewhere. The poor peasants in these less endowed regions are likely to face even higher income fluctuations relative to their counterparts in better endowed areas. The major components of risk vary among the districts, with yield risk dominating in rainfed areas and price risk in irrigated areas. Therefore, the price stabilization policy alone cannot benefit producers in poor areas. Technologies and policies that help increase and stabilize yield are needed for rainfed areas.

33

Discussion
Comment: We have been discussing the risk or variability in relation to technology, farming practices and to some extent, policy aspects. Looking from an anthropological point, I think there is a middle area of local institutions, which is rather missing in our discussion. For example, farmer loose control over resources because of inefficiency in irrigation system, contributing to risk felt by him. There could be development of some middle-level institutions, derived by common interest of farmers, as an option to decrease/manage risk. Your results on regional pattern of price risk are in contradiction with the general perception. In less developed areas, markets are not that developed and farmers do not get remunerative prices. But your results indicate that yield variability is a dominant source of income variability. In developed areas like Punjab where markets are well developed, farmers get higher prices in the year of lower production, compensating for any shortfall in yield. Moreover, farmers in developed areas have the capacity to hold the stocks for sale at the time of better prices. I do not see any contradiction. The results merely indicate that price factor contributes more to the income variability in irrigated areas. This is mainly due to the fact that yields are much more stable under irrigated conditions. Price instability may in fact be lower than in rainfed areas but my finding refers to the relative importance of price versus yield instability for each environment.

Question:

Answer:

34

Appendix 1. Names of rainfed and irrigated districts. District State Rainfed (RF) /irrigated RF RF RF RF RF RF RF RF RF RF RF RF Irrigated Irrigated Irrigated Irrigated Irrigated RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF Irrigated Irrigated Irrigated Irrigated Irrigated Irrigated RF RF RF RF RF RF RF Irrigated Irrigated Yield (kg ha-1) (1 966-93) % rice irrigation (1 994-95) 0 0.9 2.0 2.7 4.4 4.9 8.4 9.3 18.5 30.9 36.0 40.0 49.8 70.2 85.4 96.8 100 0 0 0 0 0 0 0 0 0 0 0 0 0.1 0.1 0.1 0.2 0.3 0.3 1.3 1.8 2.5 2.5 2.8 3.5 3.6 3.6 6.2 6.5 6.5 7.6 8.1 9.5 19.6 33.7 40.2 47.3 50.6 97.1 99.9 100.0 15.6 22.3 22.3 23.5 25.2 30.7 36.7 41.5 47.4

Dhanbad Hazaribagh Ranchi Singbhum Santhal Saran Muzafarpur Darbhanga Purnea Saharsa Chapmapran Palarnau Munghyr Bhagalpur Gaya Shahbad Patna Betul Chatarpur Chindwara Datia Dhar Gunna lndore Narsinghpur Raisen Sehore Tikamgarh Vidisha Jhabua Sagar Satna Khandwa Ratlam Shahadol Surguja Sidhi Madla Shajapur Khargone Bastar Panna Rewa Hoshangabad Raigarh Rajgarh Damoh Jabalpur Shivpuri Seoni Bilaspur Durg Raipur Balaghat Morena Gwalior Bhind Kalahandi Mayurbhanj Sundargarh Keonjar Dhenkanal Koraput Bolangir Phulbhani Balasore

Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Bihar Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Madhya Orissa Orissa Orissa Orissa Orissa Orissa Orissa Orissa Orissa

Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh Pradesh

1,355 (435) 1,253 (374) 1,082 (227) 1,342 (291) 1,512 (371) 1,311 (389) 1,111 (279) 1,209 (297) 1.156 (303) 1,209 (218) 1,453 (380) 997 (512) 1,458 (444) 1,642 (418) 1,564 (498) 1,858 (570) 1,830 (558) 1,262 (305) 973 (345) 1,000 (338) 764 (265) 748 (1 98) 864 (250) 777 (315) 1,458 (448) 914 (268) 923 (214) 844 (279) 888 (264) 571 (230) 905 (256) 611 (141) 1,220 (323) 655 (239) 955 (271) 1,047 (305) 695 (173) 915 (209) 818 (202) 686 (179) 1.236 (279) 711 (206) 736 (215) 1,480 (398) 845 (255) 1,271 (270) 932 (203) 794 (335) 1.280 (388) 1,520 (497) 1,417 (452) 1,242 (332) 1,562 (491) 1,529 (398) 2,012 (750) 3,605 (862) 2,123 (639) 1,762 (364) 1,986 (371) 1.718 (291) 1,844 (314) 1,935 (494) 2.103 (405) 2,065 (392) 2,035 (586) 1,995 (406)

35

Sambalpur Cuttack Puri Ganjam Basti Gonda Gorakhpur Bahraich Pratapaganj Jhansi Sitapur Sultanpur Hamirpur Azamgarh Chamoli Agra Balia Deoria Faizabad Budaun Shahjahanpur Furukabad Kheri Varanasi Hardoi Jaunpur Ghazipur Barbanki Mathura Etah Mainpur Allahabad Meerut Bereily Saharanpur Lucknow Muzafarnagar Moradabad Bijnore Dehradun Buiandsahr Kanpur Etawah Banda Aligarh Jalaun 24 Pargana Bankura Burdwan Coochbihar Darjeeling Hoogly Howrah Jalphaiguri Malda Midnapore Murshidabad Nadia Purulia West Dinajpur

Orissa Orissa Orissa Orissa Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh Uttar Pradesh WB WB WB WB WB WB WB WB WB WB WB WB WB WB

Irrigated Irrigated irrigated Irrigated RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF RF Irrigated Irrigated Irrigated Irrigated irrigated Irrigated Irrigated Irrigated Irrigated irrigated Irrigated Irrigated Irrigated Irrigated Irrigated Irrigated Irrigated irrigated Irrigated Irrigated Irrigated Irrigated Irrigated Irrigated irrigated RF Irrigated Irrigated RF RF RF RF RF RF RF RF Irrigated RF RF

2,394 (735) 2,255 (41 7) 2,358 (397) 2,448 (655) 1,377 (441) 1,338 (502) 1,736 (723) 1,139 (503) 1,550 (624) 735 (244) 1,298 (520) 1,538 (636) 950 (361) 1,441 (720) 1,833 (183) 1,758 (512) 1,452 (692) 1,889 (811) 1,847 (780) 1,729 (497) 2,085 (911) 1,729 (659) 1,814 (870) 1,821 (700) 1,652 (700) 1,606 (752) 1,571 (753) 1,939 (683) 1,879 (700) 1,703 (553) 1,785 (603) 1,647 (630) 2,058 (768) 2,062 (744) 2,365 (764) 1,642 (555) 1,767 (415) 2,056 (926) 2,277 (1,052) 1,915 (502) 1,865 (547) 1,905 (705) 1,974 (674) 1,189 (329) 1,736 (627) 967 (4 15) 1,998 (518) 2,239 (542) 2,477 (568) 2,709 (698) 1,497 (238) 2,542 (617) 2,053 (647) 1,485 (21 5) 1,933 (533) 2,000 (532) 2,088 (647) 2,173 (874) 1,792 (408) 1,674 (398)

48.7 50.9 53.0 82.2 0.1 0.2 0.3 0.5 2.1 3.0 7.2 9.7 10.1 11.7 12.8 15.0 21.7 22.0 23.9 27.5 27.8 43.7 45.6 47.2 48.2 49.9 53.9 55.7 56.8 60.2 61.2 67.9 69.3 69.6 71.6 72.6 72.8 74.9 75.0 75.3 76.8 79.7 90.4 93.4 100 100 24.7 59.2 64.9 0.23 0 38.5 5.8 2.1 23.4 21.0 11.5 49.7 6.1 9.5

Summary statistics of number of rainfed and irrigated districts No. of rainfed districts MP Orissa West Bengal Bihar UP (Figures in parentheses are standard deviations) 35 8 11 12 25 No. of irrigated districts 7 6 3 5 18

36

Notes
Acknowledgment: This paper benefits from the insightful comments and suggestions from several peers. I gratefully acknowledge their help. I am particularly grateful to Dr. Sushil Pandey, agricultural economist and deputy head, Social Sciences Division, IRRI, Philippines; Dr. Dayanatha Jha, former director, NCAP, New Delhi; and Dr. A. Janaiah, IRRI, Philippines. Author’s address: International Rice Research Institute, MCPO Box 3127, Makati City 1271, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

37

Labor use and employment pattern in rainfed riceproducing states of India
G.K. Chadha

This paper looks into the role of employment as a strategy for managing risk in rainfed agriculture. Specifically, it analyzes employment patterns, seasonal adjustments in work patterns, production in and outside the homestead, on-farm diversification, sectoral patterns of employment, farm and nonfarm components of employment, hired workers in the farm/ nonfarm sector, and the position of female workers. Unstable and low yield/productivity is the main problem facing most of the eastern states of India. The noncrop sector is still a tiny fraction of the total rural economy. The inherent weaknesses of the nonfarm sector, especially rural manufacturing, are also no less daunting; in some of these states, nonfarm earnings per worker are extremely low. On the employment side, much of the work force is still confined to agriculture. Employment under the unstable production regime of this sector is not only undependable but also much less remunerative. People work hard but they do not have enough earnings. The proportion of the work force engaged in nonfarm activities is low. Parttime employment is a common practice among the female workers of this region. The paper concludes with the suggestion that more responsive and supportive institutional environments are needed so that farmers can manage agricultural risk better through diversification of employment. Such a strategy will also directly contribute to the overall growth of the rural economy. Uncertainty in production and earning stands out as the most crippling element of rainfed agriculture. Farmers, especially ones with a small land area, may exercise their best wisdom in sketching out an area allocation plan, following the input-use pattern, pursuing the recommended field crop practices, observing timeliness, and so on. Yet, low and highly fluctuating rainfall, on a season-to-season or intraseason basis, often imposes upon them suboptimal production and postproduction choices (Singh 1978). For example, farmers may allocate more of their labor to those months that promise them higher on-farm earnings. In the dry months, they are forced to put more efforts into nonfarm activities in and around their village. Again. during a particular cropping season, there may be inadequate rainfall or rains may come much in excess of the normal, and farmers are suddenly obliged to make contingency production plans, bringing them much lower earnings than what they would have reaped had the original plan succeeded. Yet again, farmers may put their eggs in different baskets, in as much as they have learned that the “risk of starvation” can be averted if incomegenerating efforts are spread over diverse activities. Quite often, small and marginal farm households are seen to be engaged in a long chain of “self-employing” on-farm activities. Other times, their production and employment plans inside the homestead differ starkly from those outside when we move from one cropping season to another. In other words, they are advised to continue much more with their homestead economy during the season that promises them higher monetary returns, and branch out to nonhomestead employment avenues when it is no longer “profitable” to stay inside (Thorat 1993). In brief, different production and employment strategies are followed to ward off

39

the risk of low production, low employment, low earning, deprivation, poverty, and so on. A multipronged involvement in the labor market is thus an inescapable reality for a small farm producer in rainfed agriculture, retaining the eventuality of occasional asset liquidation or migration. In many cases, institutional guidance and support may be available to individual farmers in rainfed agriculture, yet, in the ultimate analysis, farmers' own efforts in grappling with unwanted weather-induced contingencies are what decide the level of their earnings and consumption year-round. This paper looks into the employment patterns of rural households in 10 rainfed riceproducing eastern states of India. The chosen states are Arunachal Pradesh, Assam, Bihar, Manipur, Meghalaya, Mizoram, Nagaland, Orissa, Tripura, and West Bengal. Although rainfed rice systems prevail in some other states as well (e.g., Kerala, Madhya Pradesh, Uttar Pradesh, Maharashtra, Rajasthan), these states are not included primarily because rice cultivation does not occupy a prominent place in the farming economy of some of these states (such as Rajasthan) or, more expressly, in their cereal economy (such as Madhya Pradesh and Maharashtra). The extreme preponderance of rice in the total cereal economy of our chosen states stands out as the single most weighty consideration for more risky farming in these states. Based on secondary published data, I attempt to see how farmers in the 10 eastern states use employment strategies that alternate markedly between one cropping season and the other, that show changing priorities between homestead and outside production and employment plans, and that show their inescapable dependence on nonfarm employment. All these aim to augment their limited and risk-prone income from farming. The paper is divided into six parts. Part I defines the problem; Part II describes the agrarian background of the study areas; and part III looks into work and employment patterns, seasonal adjustments in work patterns, production in and outside the homestead, on-farm diversification, sectoral patterns of employment, farm and

nonfarm components of employment, major sources of nonfarm employment, the share of rural areas in nonfarm employment, hired workers in the farm/nonfarm sector, and the position of female workers. Parts IV and V give rough clues on land and labor productivity in and outside agriculture, respectively. Finally, part VI makes a few concluding comments. Although an effort is made to obtain facts and figures relating to agriculture and to nonagricultural economies for all 10 eastern states, usable data in some cases are missing (e.g., those for Arunachal Pradesh and Mizoram). Thus, data from only eight states were used. Moreover, the many indicators used are rough approximations only. Because of time and data constraints, temporal changes could not be looked into. Statistical techniques are consciously avoided at this stage because the paper aims only to provide preliminary insights into the employment patterns in these states, emanating from and connected as they are with the unstable agricultural base of their economies. Finally, I believe that the production structures and employment patterns in these areas, influenced as they are by the local agrarian institutions, are also highly conditioned by their physical environment. Again, gender differences here are perhaps of a different kind. I cannot accommodate all these elements in a single paper. Nevertheless, I must, at least, take note of a few specific physical and agrarian features of the states chosen (see Tables 1, 2, and 3).

Agrarian background
Most of the eastern states are high or very high rainfall areas compared with the rest of India (Table 1). In particular, the number of rainy days per year puts them in an entirely different physical setting. The production environment in Nagaland, Meghalaya, or Manipur, because of the excessively long rainy season, is strikingly different from that in many other rice-producing regions with a smaller number of rainy days per year (such as Andhra Pradesh. Haryana, Punjab. Tamil Nadu). Again, in terms of irrigation development, the rice-producing eastern states are far behind. The stark contrast is reflected in

40

Table 1. Rainfall and other relevant information on the rice economy of India. Av rainfall State (mm)a Coefficient of variation (%)a Rainy days per year (no.). Rainfed rice area (% of total rice area)b 3.0 78.9 64.2 41.1 6.5 38.0 9.2 37.8 58.6 76.1 73.7 54.0 57.0 74.0 66.2 5.6 63.8 13.2 81.4 45.0 73.1 52.4 Rice area (% of total area under cereals)b 66.7 96.0 61.2 19.3 22.0 10.0 32.0 25.3 98.3 39.9 14.4 98.4 81.2 78.0 94.8 37.8 1.7 66.0 97.0 30.7 93.6 42.2 Net area irrigated (% of net sown area)b 38.5 21.1 46.7 27.6 75.6 17.3 42.6 20.3 14.9 24.4 14.9 32.8 22.4 29.4 46.4 92.3 26.4 46.4 13.1 65.8 34.8 35.3

Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Jammu-Kashmir Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Nagaland Orissa Punjabd Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India
a

879 2,207 1,272 833 722 1,723 995 1,355 2,996 1,215 1,314 2,389 2.773 2,378 1,482 625 588 1,008 2,101 985 1,645 -

19.0 8.5 12.5 32.0 28.0 16.0 16.0 15.0 12.0 16.0 25.0c 11.0 26.0 42.0e 114.0 18.5 13.0 -

49.4 104.2 60.8 36.7 31.7 65.0 55.0 67.8 126.0 57.3 58.6 122.5 123.0 132.3 72.3 31.7 28.9 53.5 99.1 45.8 75.3 -

Govt. of India, Monthly and Annual Normals of Rainfall and Rainy Days; based on records from 1901 to 1950. Govt. of India Press. p 65,68-73. Govt. of India. Statistical Abstract. India 1992, CSO, Sep 1994. p 41. cCoeffiient of variation is given for the Marathwada region of Maharashtra to highlight the seriousness of the rainfall variability problem in the state. dGovt. of Punjab. Statistical Abstract. Punjab 1996, ESO. March 1997. p 152-163. eFor the same reason, it is given for West Rajasthan (in preference to the eastern part of the state).
b

Table 2. irrigated net operated area, irrigated and nonirrigated sown area, and area under current fallow by state and kharif/rabi seasons, 1992 NSS data. State Irrigated net sown area (% of net operated area) Kharif Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Jammu-Kashmir Karnatak Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India
Source: Govt. of India (1997g).

Nonirrigated net sown area (YO of net operated area) Kharif 52.1 72.5 45.5 57.9 8.2 64.0 61.0 70.7 67.0 67.2 18.6 52.2 89.2 86.7 76.9 6.8 53.4 36.5 36.6 15.0 44.5 49.8 Rabi 23.1 53.3 29.8 10.8 9.1 66.8 50.6 44.8 65.4 42.1 28.8 34.4 82.4 62.1 28.6 6.7 18.8 22.1 28.6 15.3 8.3 25.6

Area under current fallow (% of net operated area) Kharif 3.6 2.7 0.9 2.1 2.8 0.5 0.2 8.0 1.0 6.7 4.3 0.8 0.8 4.0 0.8 0.2 15.0 4.3 5.2 8.1 1.1 5.8 Rabi 32.7 26.0 1.5 24.7 10.6 0.8 10.9 20.2 6.2 17.8 21.5 38.8 7.9 22.8 37.0 0.0 39.6 31.7 4.3 5.7 26.7 19.5

Rabi 38.1 3.6 60.3 59.5 77.0 13.2 23.1 26.1 16.8 35.7 34.7 7.0 0.5 0.0 27.3 87.7 35.5 37.5 33.5 73.0 51.8 47.2

38.5 6.2 43.9 35.4 76.2 11.7 23.6 17.5 21.3 22.5 60.1 29.9 0.4 0.6 16.4 87.2 22.8 50.8 23.3 70.0 41.6 35.7

41

Table 3. Agrarian structure in Indian states, 1992 NSS data. State Percentage share of holdings SMRa SMR + MRGb Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Jammu-Kashmir Karnatak Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India
a

Percentage share of operated area SMR SMR + MRG 17.5 16.1 29.0 8.5 16.4 41.4 24.9 9.6 53.2 6.7 6.7 37.5 38.7 32.6 22.2 6.2 5.6 28.9 52.9 24.9 39.9 15.6 SMR + MRG + SML 40.9 47.3 54.2 22.3 25.2 63.9 59.5 25.0 76.6 22.3 18.4 72.7 67.5 72.7 52.3 16.9 14.9 57.0 83.9 51.3 70.6 34.3

Percentage of operated area leased by SMR MRG SML All

SMR + MRG + SMLc 79.8 89.8 96.6 63.1 59.3 88.4 85.0 68.6 96.8 68.2 58.2 82.6 83.6 86.5 84.0 71.2 58.0 89.7 90.8 84.7 92.3 78.0

No. of parcels per holding

29.1 34.9 46.8 22.0 30.2 41.4 23.4 24.4 74.5 14.0 19.6 25.7 21.8 8.9 27.6 48.7 17.2 62.1 47.2 36.2 52.8 35.1

58.4 69.8 71.5 43.2 45.8 76.1 58.2 48.3 90.9 33.8 39.3 62.5 65.4 55.9 59.7 59.8 38.1 76.6 78.9 66.2 78.9 60.2

2.9 0.9 8.3 1.3 0.8 12.0 3.7 1.1 23.3 0.7 0.8 5.4 4.6 2.0 3.9 1.2 0.8 7.6 13.9 5.4 11.8 3.3

9.2 8.5 7.3 5.5 1.3 5.5 4.1 4.0 2.9 5.0 4.0 1.0 4.6 1.0 7.4 13.9 1.7 12.1 5.2 9.9 15.4 8.6
b

11.5 6.3 5.3 1.8 8.4 5.9 2.8 6.0 1.2 8.6 3.1 6.8 9.9 0.8 13.0 18.7 8.0 9.9 5.1 11.9 11.5 8.7

10.2 8.7 5.5 1.7 9.1 5.8 5.6 7.6 3.8 7.5 3.0 13.3 6.7 0.4 14.4 11.7 2.8 10.9 8.4 14.4 12.1 8.5

9.6 7.0 6.2 2.9 6.5 5.8 3.2 5.4 2.1 7.8 3.3 5.2 8.7 0.7 11.0 17.3 6.5 10.8 5.1 11.2 13.2 8.7

2.5 1.8 2.8 1.9 1.9 4.0 2.6 1.9 1.4 3.0 2.4 1.0 1.8 2.1 3.1 2.0 3.2 2.1 1.5 3.1 3.7 2.7

Holdings with operated area between 0.002 and 0.40 ha are defined as submarginal (SMR). are marginal (MRG). c Those with area between 1.01 and 2.00 ha are small holdings (SML). Source: Govt. of India (1997c).

Those with operated area between 0.41 and 1.00 ha

the rainfed rice-cropped area as a percentage of total area under rice and, still more so, in terms of the importance of rice in the total cereal economy. In Manipur, Tripura, Assam, Orissa, and West Bengal, rice alone accounts for nearly 95 % of the total area under cereal crops. In brief, a very high incidence of rainfed rice cultivation on the one hand, and the heavy yearround dependence on rice on the other, leave the chosen states vulnerable. Low yields are therefore a natural consequence. Later, in Table 11, data will compare the poor performance of most of the chosen states in rice yield with that of other traditional rice-growing states such as Andhra Pradesh, Tamil Nadu, and Karnataka. leaving aside startling gaps with respect to the Green Revolution areas of Punjab and Haryana. Table 2 shows greater interseasonal variation in irrigation endowment in the eight states. Against a perfectly stable irrigation base in Punjab and Haryana and a highly stable base in Andhra Pradesh, Himachal Pradesh, JammuKashmir, Kerala, and Uttar Pradesh, the highly fluctuating irrigation base is clearly discernible in Assam, Bihar, Manipur, Orissa, Tripura, and 42

West Bengal. In most Indian states, current fallow is much higher in the rabi cropping season than in kharif. The rainfed rice-producing eastern states show this trait a bit more sharply. Table 3 informs us about a few important agrarian features of the Indian states. Without doubt, there is a pronounced concentration of submarginal. marginal, and small farms in most of them. Their concentration is all the more pronounced in the eastern states. For example, such holdings make up 97% of the total holdings in Bihar. 92% in West Bengal, 90% in Assam, and 80% and above in other eastern states. The inherent weaknesses of agriculture in our chosen states, especially in the context of a mismatch between small land area and technological compulsion of the modem farming system, need to be appreciated. In other words, farming on a fairly submntial proportion of an operated area (e.g., 84% in Tripura; more than 70.0% in West Bengal, Nagaland, and Mmnipur; about twothirds in Meghalay; etc.) is "doomed" to be carried out under a "subsistence farming regime," largely untouched by technological advances and characterized by up-and-down

swings in employment and earnings of a large number of rural households.

Work patterns and employment
Production in and outside the homestead A typical feature of Indian agriculture is that most of the farmers combine production enterprises in and around their homestead with work outside the homestead. Work outside the homestead involves various activities, such as self-employment in cultivation, wage employment on others’ farms, or a mix of employment types in a wide array of nonfarm activities in and outside the village. For households with a small land area, a scattered pattern of work is a natural phenomenon since any bit of additional earning from whatever source would mitigate the acute economic distress under which they live. In as much as the production and employment uncertainty looms larger during dry months, farmers would rather concentrate their production and selfemployment efforts inside the homestead. They will then seek employment/production opportunities outside the homestead when the rainy season starts. Table 4 clearly confirms these interseasonal production and work adjustments. The inside-outside homestead division of production and employment efforts varies greatly between the rainy season and rabi season (usually dry season) in most of the states, practically all along the farm size continuum. In Punjab and Haryana, dependence on homestead production/employment is practically nil during rabi and very marginal during kharif. An exception is the tiny farm that has to use its homestead production base largely because there is little land to till. The kharif-rabi differences are rather mild in other states (e.g., Andhra Pradesh, Tamil Nadu, Uttar Pradesh) where irrigation development is of a medium order. The eastern states of Bihar and West Bengal are medium irrigation areas. In most eastern states
1

with a low irrigation base, the kharif-rabi difference between homestead and outside production is not sharp but persists consistently across all farm sizes. For example, in Orissa, while only 11% of operational holdings continue with homestead production during kharif, this percentage goes up to 26.4 during rabi. In Manipur, the kharif-rabi percentages are 32.7 and 59.1; in Meghalaya, they are 15.6 and 24.3; and in Nagaland, 6.8 and 15.0. For obvious reasons, on smaller farms, the kharif-rabi difference is particularly high. Farmers in this situation have to observe more sweeping interseasonal production/work adjustments. In any case, the greater flexibility in dividing production/employment efforts between the homestead and outside locales is an unmistakable agrarian feature and a survival strategy followed by farmers in the rainfed rice areas of India.

On-farm

diversification

Farmers in rainfed agriculture have a much more diversified on-farm pattern of production/ employment. This is particularly the case for the rabi cropping season, which generally consists of dry months in the yearly production cycle (Table 5). In Punjab and Haryana, states which experienced Green Revolution in 60’S, the whole production regime is dominated by two activities: crop and livestock production, although there are several noncrop activities such as vegetable production, horticulture, poultry, and miscellaneous enterprises. This is not surprising because the states have a fairly stable (and expanding) base of grain production and their farmers are not exposed much to seasonal yield variation and production and earning uncertainties of the kind suffered by rainfed farmers. The most profitable activity is raising livestock. The wheat-rice combination, supplemented by livestock enterprises (mostly milk and allied products), has been their most profitable on-farm choice.’ Livestock supplements crop production in many other

The recent setbacks in Punjab agriculture, brought about by excessive and untimely rains, do not come in the way of our analysis.

43

Table 4. Seasonal distribution (%) of operational holdings reporting different types of agricultural and related production by state and farm size, 1992 NSS data. Tiny farms SMR farms MRG farms Med. farms All farms Inside homestead only Outside homestead Inside homestead only Outside homestead Inside homestead only Outside homestead Outside homestead Inside homestead only Inside homestead only Outside homestead Outside homestead SML farms

State

Season

Inside homestead only

Andhra Pradesh

Assam

Bihar

Gujarat 14.2 2.5

3.8 16.3 38.1 34.5 18.7 7.0 3.7 4.0 11.9 16.5 18.1 20.4 4.8 2.9 3.8 4.4 10.4 17.9 14.4 21.5 3.8 3.1 9.1 5.2 4.0 4.8 11.9 8.8 13.9 0.8 5.5 13.5 6.1 1.0 0.9

Haryana

Himachal Pradesh

Jammu-Kashmir

Karnatak

Kerala

Madhya Pradesh

Maharashtra

Manipur

4.9 5.5 4.8 0.8 3.5 3.2 7.1 0.3 5.4 34.3 2.9 7.4

Meghalaya

Nagaland

Orissa

Punjab

4.5 0.8 1.9 10.7 2.0 12.5 0.2 1.3 2.2 4.4 0.6 1.5 32.3 57.9 4.6 6.9 1.7 2.2 3.3 7.4 8.9 0.7 0.1 0.6 7.7 3.5 10.9 1.4 2.6 4.4 4.8 0.6 6.5 22.9 50.0 7.8 8.8 2.6 1.0 2.2 9.1 9.4 3.2 5.9 4.1 8.9

4.8 5.8 5.8 4.3 16.7

Rajasthan

Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi Kharif Rabi 2.2 5.3 2.4 9.5 7.3 6.7 4.3 8.3 0.6 0.7 29.7 46.0 14.4 18.2 4.1 10.7 3.9 8.2 14.4 6.4 8.8 4.7 95.6 89.6 82.1 85.6 78.5 96.2 69.9 90.9 94.8 100.0 99.3 99.9 99.4 92.3 96.5 89.1 98.6 97.4 95.6 95.2 99.4 93.5 77.1 50.0 92.2 91.2 97.4 99.0 97.8 90.9 90.6 100.0 95.9 91.1 96.0 95.2 88.1 91.2 86.1 99.2 94.5 86.5 93.9 100.0 99.0 99.1 100.0 95.1 94.5 95.2 99.2 96.5 96.8 92.9 99.7 94.6 65.7 100.0 97.1 92.6 100.0 100.0 95.2 94.2 94.2 100.0 95.7 83.3

58.8 72.7 73.3 74.3 43.6 35.5 59.7 83.1 66.9 77.9 4.7 0.9 54.8 45.2 68.5 81.4 37.3 33.2 67.3 65.7 72.1 88.1 44.9 62.9 52.0 55.2 71.1 41.8 41.4 61.3 72.9 76.8 68.9 70.6

41.2 27.3 26.7 25.7 56.4 64.5 40.3 16.9 33.1 22.1 95.3 99.1 45.2 54.8 31.5 18.6 62.7 66.8 32.7 34.3 27.9 11.9 55.1 37.1 48.0 44.8 28.9 58.2 58.6 38.7 27.1 23.2 31.1 29.4 96.0 88. 83.5 81.9 79.6 95.2 97.1 96.1 100.0 100.0 95.5 99.2 98.1 89.3 98.0 87.5 99.8 98.7 97.8 95.6 99.4 98.5 67.7 42.1 95.4 93.1 98.3 97.8 96.7 92.6 91.1 100.0 96.8 94.1

96.2 83.7 61.9 65.5 81.3 93.0 96.3 100.0 85.8 100.0 97.5 100.0 97.8 94.7 97.6 90.5 92.7 93.3 95.7 91.7 99.4 99.3 70.3 54.0 85.6 81.8 95.9 89.3 96.1 91.8 85.6 93.6 91.2 95.3

14.6 28.4 33.6 35.4 29.4 17.8 16.7 33.9 22.0 22.3 3.5 0.6 8.8 12.8 16.9 41.3 24.5 22.3 16.4 24.0 16.1 44.6 32.7 59.1 15.6 24.3 6.8 15.0 10.8 26.4 35.8 37.2 11.3 28.2

85.4 71.6 66.4 64.6 70.6 82.2 83.3 66.1 78.0 77.7 96.5 99.4 91.2 87.2 83.1 58.7 75.5 77.7 83.6 76.0 83.9 55.4 67.3 40.9 84.4 75.7 93.2 85.0 89.2 73.6 64.2 62.8 88.7 71.8

states as well (e.g., Andhra Pradesh, Gujarat, Jammu-Kashmir, Karnataka, Madhya Pradesh, Maharashtra, Rajasthan, Tamil Nadu, and Uttar Pradesh), though at a scale much lower than that in the Green Revolution states. In most of the rainfed rice- producing eastern states, livestock is less noticeable; the situation in Assam, Manipur, Meghalaya, and Nagaland merits particular attention. Thriving livestock enterprises depend quite heavily on timely access to and adequate availability of grains. In these eastern states, the uncertainty present in grain production may deter farmers from engaging in this grain-dependent business. They do not, however, sit back helplessly. Vegetable cultivation figures quite prominently in a way (though not on the same scale) that livestock does in the Green Revolution areas. As noted earlier in Table 4, dependence on homestead production/employment is much more evident in the rainfed rice-producing eastern states, and it now becomes obvious that vegetable cultivation and poultry (and, to a limited extent, livestock) are the natural choices for the homestead economy in these areas. Thus, farmers in these states pursue a much wider range of production activities compared with their counterparts in other states, especially those with an assured means of irrigation. The tendency to operate in diverse production lines manifests itself in all states, among all farm size categories and in both kharif and rabi seasons. The degree varies across the board. The necessity for supplementation is felt most acutely by those who have very little land base to operate on, and it declines as we move up the farm size continuum. It is no wonder, therefore, that tiny farms have the most diversified on-farm production base. For example, in Assam, just about one-fourth of the tiny farms are engaged in crop production during the rabi season, while as many as 55.0% are engaged in vegetable cultivation, 6.0% in livestock, 2.4% in poultry, 2.1% in horticulture, and 8.59% in miscellaneous enterprises. In Manipur, Meghalaya. Orissa, and West Bengal, the tiny farms are more dispersed between crop and noncrop enterprises. A similar pattern prevails during the kharif cropping season.

45

Table 5. Seasonal onfarm diversification of production activities by state and main activity, 1992 NSS data. Percentage distribution of operational holdings by activity type State Andhra Pradesh Season Kharif Holding type Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Crop production 40.1 91.3 93.6 94.4 94.7 83.4 30.1 72.3 81.6 85.6 89.3 67.4 28.9 70.7 86.1 91.5 91.6 71.5 26.1 66.1 80.1 89.6 93.2 64.5 61.0 91.6 92.8 90.0 89.9 79.3 67.3 95.8 98.5 96.7 99.2 84.3 32.2 86.8 95.0 98.0 94.7 81.3 23.2 89.9 95.1 87.6 87.6 65.3 11.5 79.0 94.4 88.1 91.3 70.0 7.8 94.3 100.0 100.0 97.2 72.4 92.7 96.8 86.2 91.1 98.6 92.2 Vegetable production 8.0 1.9 0.4 0.3 0.2 2.1 5.6 0.7 2.4 2.1 0.3 2.9 37.6 16.3 5.4 3.8 0.5 13.7 54.9 30.5 14.9 9.6 5.0 27.6 6.0 0.6 1.1 1.7 1.0 2.6 4.1 1.0 0.6 1.5 0.5 2.3 1.6 2.0 Horticulture 0.5 0.7 1.2 0.7 0.5 0.7 Plantation Livestock 26.7 3.4 2.3 2.8 2.6 7.2 41.3 16.2 7.2 9.6 6.0 18.3 7.0 2.4 1.8 1.8 3.0 5.9 Poultry 13.4 Others

Rabi

Assam

Kharif

1.5 0.8 1.8 0.6 2.6 0.7 0.5 0.8 2.1 0.3 0.1 0.2 1.1 0.8 0.4 0.8 0.2 0.4 0.6 0.5 0.5 0.1 0.1 0.9 0.6 0.4 0.0 0.4 1.2 2.3 1.1 1.2 1.1

Rabi

5.2 1.6 1.9 0.7 1.8 2.1 3.1 4.2 0.7 1.5 2.1 2.4 4.1 3.0 0.6 0.4 1.9 2.0 1.8 1.2

0.9 2.9 11.3 3.7 5.4 0.4 0.1 5.2 6.9 1.5 0.7 0.2 2.2 2.4

Bihar

Kharif

Rabi

0.8 1.1 0.1 0.1 0.1 0.2 0.5 0.4

Gujarat

Kharif

0.10

0.10

Rabi

Haryana

Kharif

0.1 0.5 1.3 7.2 0.0 1.5 0.6 1.7 6.9 0.0 4.2 1.1 1.7 5.70

0.10

0.20

2.0 0.5

Rabi

0.8 1.6 16.9 6.1 5.0 6.0 7.1 10.5 18.8 2.3 0.90 1.60 0.20 9.1 57.5 10.5 3.0 1.4 4.5 15.1 74.2 0.7 3.8 9.5 11.8 30.4 86.3 14.2 4.3 7.7 6.5 27.8 86.9

6.1 1.1 0.6 0.2 0.2 1.6 8.6 2.9 1.2 0.1 0.4 3.2 15.5 2.5 5.7 2.4 5.3 6.8 8.5 1.2 3.7 0.6 4.0 13.1 0.5 0.9 0.5 0.5 6.0 8.2 0.4

0.7 1.4 0.3 1.4 0.5 0.8 1.2

0.5 1.5

1.00

3.6 6.4 0.8 1.3 0.2 0.6 1.7

0.2 1.0 0.3 2.2 1.40 0.2 0.7 3.6

Himachal

Pradesh

Kharif

0.3 1.0 1.8 0.0 1.0 0.5 0.0 0.8

2.50 0.60 3.6 2.7 7.0 6.7 0.7 4.4

0.9 0.2

25.0 1.9 0.5 4.9 1.7 0.7 2.4

1.0

46

Table 5 continued. Percentage distribution of operational holdings by activity type State Season Rabi Holding type Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL crop production 93.3 96.6 94.3 93.7 92.2 94.6 46.5 96.4 95.3 97.8 98.6 89.6 46.4 91.8 90.2 90.8 94.3 85.3 26.0 71.9 84.3 85.5 88.1 72.5 20.5 57.9 67.1 64.9 84.8 47.9 24.6 26.4 29.3 24.9 30.6 25.5 18.4 25.9 25.5 23.9 29.5 21.0 28.0 90.0 93.5 92.5 93.1 80.1 38.6 91.1 92.7 93.1 92.1 75.4 27.6 99.5 98.9 98.5 98.0 83.5 11.5 94.8 91.7 87.6 91.8 53.1 Vegetable production 1.3 0.1 0.4 0.6 3.9 0.7 6.8 1.4 0.0 0.8 0.0 1.6 9.3 1.4 Horticulture 3.4 3.3 5.3 5.7 3.9 4.2 1.90 Plantation Livestock 1.6 Poultry Others 0.3

Jammu and Kashir Kharif

Rabi

0.3

Karnataka

Kharif

1.5 5.9 2.0 0.8 1.5 1.9 4.4 8.4 2.3 2.8 3.5 3.9 0.3 2.4 3.7 2.1 0.2 2.7 9.4 1.2 1.0 0.6 0.3 1.9 7.9 3.7 1.0 1.0 2.8 4.6 1.1 0.2 1.0 4.0 2.6 2.7 0.2 0.2 2.1

Rabi

2.0 4.5 1.8 0.5 0.1 1.5 0.6 7.2 3.8 1.6 14.1 3.9 8.3 6.9 1.5 10.8 14.0 4.0 5.3 6.7 4.0 10.7 0.8 1.1 0.6 0.5 0.9 0.7 0.4 0.1

Kerala

Kharif

Rabi

Madhya Pradesh

Kharif

10.6 8.3 3.3 5.9 4.6 6.4 5.8 12.5 16.0 15.6 9.9 9.3 32.4 61.9 60.0 65.4 66.6 43.7 42.8 66.1 68.3 69.2 66.5 51.7

0.4 42.8 2.2 3.8 1.4 1.4 8.0 44.3 6.9 9.8 9.2 5.7 13.2 44.5 10.7 5.1 5.6 5.2 13.7 59.0 14.1 13.5 11.3 4.9 32.3 5.7 4.3 1.7 2.3 1.3 4.7 2.8 0.4 0.7 0.2 1.9 41.4 6.8 3.8 6.2 5.4 11.9 34.7 3.5 5.6 5.7 6.9 15.3 53.4 0.1 0.4 11.6 74.3 0.7 3.7 8.9 7.2 38.6

0.2 2.0 0.0 0.9 0.5

8.9 0.4 1.9 4.5 1.5 2.3 6.6 0.7 0.4 4.3 3.7 0.3

2.2 4.6 3.5 1.2 0.5 2.1 5.3 1.1 0.4 2.5 12.9 2.8 0.4 0.3 8.6 14.6 1.2

2.4 12.4

Rabi

0.8 1.8

3.3 0.9 0.2 0.5 11.5

9.7 8.8 1.2 0.6 0.7 0.6 2.3 16.2 0.9 5.2 2.7 1.3

Maharashtra

Kharif

0.4 0.5 0.1 0.1 0.2 0.9 0.2 0.4 0.2

Rabi

1.4 0.5 0.4 1.0 0.8 3.1 0.4 0.8

2.7 9.4 1.2 5.1

0.7 0.3

0.1

47

Table 5 continued. Percentage distribution of operational holdings by activity type State Manipur Season Kharif Holding type Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Crop production 19.6 82.7 70.9 76.4 73.5 64.0 20.9 29.3 93.7 66.4 33.9 40.4 54.4 83.9 79.7 88.5 69.2 12.3 32.3 36.8 22.5 18.6 25.0 29.0 88.5 96.2 97.4 93.4 90.8 64.0 87.2 89.7 89.8 95.8 82.3 57.2 95.4 97.5 96.8 98.2 89.3 36.9 86.7 89.9 89.9 94.2 71.3 9.3 92.70 95.80 88.0 94.7 55.9 8.6 93.1 100.0 98.7 98.4 54.6 29.2 86.5 95.8 96.3 95.1 87.3 Vegetable production 57.8 8.9 16.2 13.6 20.5 23.2 61.6 59.5 4.4 27.1 79.1 52.5 21.4 22.1 8.3 6.8 2.5 13.3 56.7 37.3 52.2 48.0 60.7 49.3 47.4 7.9 0.4 1.0 4.0 16.0 11.1 8.7 10.2 4.2 11.4 11.6 0.9 1.1 0.5 2.7 8.2 8.6 7.4 6.3 4.8 7.5 3.6 Horticulture 1.9 0.7 0.3 0.9 Plantation 1.4 1.5 0.6 0.4 Livestock 2.9 7.0 1.6 1.4 3.3 8.4 Poultry 4.8 4.3 3.5 1.7 3.3 3.6 Others 13.6 6.5 1.6 3.2 2.1 5.2 4.2 11.2 1.9 6.5 20.9 5.0 2.5 5.1 2.5 2.7 3.5 5.0 3.6 2.1 1.3 5.3

Rabi

Meghalaya

Kharif

Rabi

Nagaland

Kharif

0.6 10.3 11.8 4.0 4.7 6.4 6.5 8.6 21.0 7.1 20.8 15.1 12.6 3.7 6.6 0.9

0.3 5.7 2.1 0.2 2.4 0.5 1.8 2.1 2.0 0.4 2.1 0.8 1.3 0.9 0.5

5.6 9.5 2.9 2.1 2.1 9.2 1.8 1.4 5.3 4.9 4.5 20.1 1.0 1.7 2.7 20.0 1.6 1.6 6.3 19.6 4.6 0.7 1.8 1.4 5.5 36.9 2.0 2.7 2.2 1.0 14.3 80.8 7.4 4.2 11.3 4.8 39.0 86.8 6.9 0.6 1.o 42.8 66.3 10.1 2.1 2.6 2.7 10.8

2.4 10.2 1.6 1.o 3.8 3.7 6.1 2.1

1.9 3.5 1.4 1.1

Rabi

Orissa

Kharif

0.4 0.9 0.30 1.00

7.0

4.2 0.3

Rabi

1.3 9.5 0.3 0.9 3.4 1.9

0.9 7.6 2.4 0.8 3.2 3.8

Punjab

Kharif

0.4 0.5 0.7 0.5 0.3 0.5 0.7 0.6 0.5 0.9 0.3 1.0 0.3

Rabi

1.9 3.1

0.3 0.4

0.8 0.6

1.7

Rajasthan

Kharif

1.3 1.9 1.3 0.4 0.8 0.8

0.3

0.3

0.1

2.6 2.5 0.7 0.3 0.3 0.8

48

Table 5 continued. Percentage distribution of operational holdings by activity type State season Rabi Holding type Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL TlNY SMR MRG SML MED ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL Tiny SMR MRG SML Med. ALL TlNY SMR MRG SML MED ALL Crop production 28.3 94.8 93.8 92.5 81.8 70.6 28.2 84.1 90.6 88.6 92.8 62.0 23.4 67.2 75.1 76.0 97.7 47.5 58.1 93.2 96.0 95.2 94.9 86.1 58.6 94.4 96.8 97.1 97.2 87.7 47.1 90.9 96.2 94.5 95.6 75.3 39.3 80.1 86.0 80.0 89.1 62.8 40.5 86.9 92.1 92.6 93.7 77.4 35.5 83.6 88.5 89.1 90.0 68.6 Vegetable production 1.7 1.5 1.4 0.8 12.3 1.0 0.6 3.5 2.3 6.3 11.4 1.3 1.2 6.1 3.2 7.2 2.5 0.7 0.4 0.5 0.1 0.9 4.7 2.9 1.6 1.4 1.2 2.4 10.5 3.4 0.4 0.1 0.6 4.9 16.0 12.1 7.1 7.6 7.5 12.1 7.6 1.8 1.1 0.9 0.5 2.7 8.0 5.3 3.0 2.5 1.4 4.7 Horticulture 0.5 0.4 0.2 2.2 2.1 0.4 0.3 0.2 1.4 3.6 4.3 5.1 1.9 3.2 3.5 1.3 1.0 0.4 0.8 1.5 0.9 1.2 1.1 0.1 0.4 1.0 0.7 1.6 0.9 0.7 0.5 0.3 1.0 3.2 1.6 0.5 1.6 0.9 2.0 2.3 1.2 0.9 0.5 0.8 1.2 2.4 1.4 0.8 1.0 0.8 1.5 4.9 6.8 2.9 2.6 1.5 4.4 4.7 8.2 4.8 3.8 8.2 5.2 Plantation Livestock 70.1 3.5 5.5 6.4 16.0 27.6 34.0 2.8 2.6 4.8 2.3 16.4 40.8 12.4 11.7 11.7 14.1 26.8 35.2 4.5 3.1 3.0 3.2 11.1 31.2 0.8 1.3 0.9 0.2 7.9 18.5 0.8 0.8 2.8 2.4 8.1 22.3 3.2 4.6 8.2 1.4 12.9 30.6 4.5 2.7 3.5 3.4 11.1 37.5 3.6 3.5 4.8 5.2 16.9 Poultry Others 1.1 0.7 0.3 0.8 0.8 8.9 2.8 3.2 0.3 0.9 5.0 6.4 0.6 2.1 0.4 0.5 3.7 2.4 0.6 0. I 0.5 0.2 0.8 2.9 0.5 0.2 0.2 0.5 0.9 6.2 3.0 1.2 1.9 1.2 3.1 5.2 1.5 1.3 1.6 1.1 3.3 7.2 1.5 1.1 0.7 0.4 2.7 6.3 0.8 0.7 0.3 0.3 2.6

Tamil Nadu

Kharif

9.4 0.3 0.7 4.4 9.7 6.1 0.3 6.1 0.6

Rabi

Uttar Pradesh

Kharif

Rabi

0.7 0.3

0.2 0.8

West Bengal

Kharif

0.2 1.4 0.6 0.7 0.3 0.8 1.6 0.6 0.3 0.9 5.0 3.9 2.0 1.5 1.1 2.8 5.1 4.5 3.0 2.1 1.8 3.6

0.2 14.7 0.6

Rabi

6.2 12.5 0.9 0.2 1.0 6.2 6.8 0.2 0.1 0.2 0.1 2.1 5.0 0.7 0.4 0.2 2.2

All India

Kharit

Rabi

Source: Govt. of India (1997g).

49

The contrasts between the (wet) kharif and (dry) rabi seasons need further examination. In most of the eight states, noncrop enterprises are favored by a large number of farms, irrespective of landholding size, more in the rabi season than in the kharif season. The higher degree of onfarm production diversification in these areas during rabi signals the greater distress prevailing during the dry months of the year.2 The striking similarity of the crop-noncrop difference during the kharif and rabi seasons in the Green Revolution areas and the milder kharif-rabi deviations in other irrigated regions needs to be seen in contrast to what is happening in the rainfed rice-producing eastern states. In the rainfed states, bigger kharif-rabi deviations on the one hand, and, within either cropping season, a relatively more spread-out canvas of cropnoncrop mixes, on the other, clearly point to a “multiple involvement” strategy, primarily to mitigate the distress that would follow if total reliance was put on crop production alone. Sectoral pattern of employment To a typical rural household in India, employment is available partly (though largely) through on-farm activities and partly through nonfarm activities. Farm activities manifest themselves in work on one’s own farm as cultivators and on others’ farms as agricultural labor, This is sometimes referred to as the crop sector and often includes plantations. Activities allied to crop production and located within as agriculture-such raising of livestock, poultry, pigs, etc., and production of milk-are the main components of the noncrop sector. Generally, as indeed defined by the Indian Census Organization, hunting and trapping, game propagation, forestry and logging, and fishing are also included in the noncrop sector. On the other hand, the range of nonfarm activities is wide and encompasses those selfemployed in industry; construction; transport and communications; wholesale and retail trade; community, social, and personal services; as

well as those employed by others in these occupations as wage-paid workers (Chadha 1993). Farm and nonfarm employment Table 6 gives state employment data of rural households under the three broad categories mentioned above and based on two independent (admittedly, not completely comparable) sources of data. For the crop sector, the 199 1 population census (giving total cultivators and agricultural laborers) is the only source. For the noncrop agricultural and nonagricultural activities, the 1990 economic census data are handy. The two sets of census data are the latest available. The crop sector is still the most dominant source of employment for rural households in all parts of India. This is especially so for a few rainfed rice-producing eastern states (e.g., Bihar, Nagaland, Mizoram, Orissa). To raise the standard of well-being of the rural populace in general and of those in the eastern states in particular, productivity in the farm sector, on both a per hectare and a per worker basis, must clearly improve. The employment base of the noncrop sector is terribly low, almost negligible, in most states. The depressingly low levels for most of the eastem states need to be emphasized. In other words, in spite of people’s efforts to get into different types of noncrop production regimes, branching off from crop to noncrop agriculture is not discernible. The possibility of a mix of noncrop employment with crop-sector employment cannot be ruled out either. For example, in a separate 1993-94 NSS Employment Survey, national crop-sector employment level accounts for 70% of rural workers (Gov’t. of India 1997b). On the other hand, the four main constituents of noncrop employment (livestock, agricultural services, forestry, and fishing) account for 5.3%, 0.6%, 0.3%, and 0.4% of employment, respectively. The total of crop and noncrop employment for 1993-94 is 78.4%, which is very close to (our) 1990 economic census figure of 81.9%. In any

2

In more recent years, expansion of irrigation has facilitated crop diversification in some eastern states in the rabi season.

50

Table 6. Broad sources of employment of rural households, 1990-91 Economic Census. Percentage of rural persons engaged in State Field-crop agricultural activitiesa 81.1 72.2 70.0 88.0 76.8 72.9 72.1 79.7 56.1 88.1 82.9 76.1 78.3 83.2 83.7 80.8 73.6 82.6 77.5 70.6 84.5 70.2 80.0 Noncropb agricultural activitiesc 3.1 0.8 0.4 0.2 7.4 5.7 0.7 5.0 1.7 1.7 2.2 1.4 0.8 0.4 0.5 1.3 0.6 1.3 1.7 0.7 0.3 2.4 1.9 Nonagricultural activitiesC 15.9 27.0 29.6 11.8 15.8 21.4 27.2 15.3 42.2 10.2 14.9 22.5 20.9 16.4 15.8 17.9 25.8 16.1 20.8 28.7 15.2 27.4 18.1

Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India
a

look into the composition of such employment opportunities in each state. Table 7 fills this gap to some extent. With all its data limitations— such as two-digit classification of production sectors and interstate variations in product composition, quality, and prices-Table gives 7 the broad sources of nonagricultural employment for rural households in each state.

Major sources of nonfarm employment
Manufacturing, trade, and community, social, and personal services (CSPS) together claim a lion’s share of total nonfarm employment in each state. With some exceptions, the all-India pattern of manufacturing, CSPS, and trade, in that order, is duly observed by individual states as well. Exceptions come largely from the east. For example, manufacturing employment figures are much lower in comparison with CSPS employment in Mizoram, Meghalaya, Nagaland, Tripura, Arunachal Pradesh, Manipur, and Assam; only in West Bengal and Orissa does manufacturing hold an edge over CSPS employment. Meanwhile, in Bihar, the two employment types are nearly even. The edge of manufacturing employment over its CSPS counterpart is far more marked in most of the noneastern states. The structure of nonfarm employment is thus relatively weaker in most parts of the eastern region than in the rest of the country. The most telling weakness is discernible in Mizoram and Meghalaya. If only the composition of the CSPS sector were available for the individual states, including the eastern states, one would have clearly seen that a substantial part of the CSPS employment in the eastern states came directly through publicsector channels or indirectly through public funds, autonomous and unguided by what was going on in the commodity sectors of their local economies. Two more facts testify to the inherent weakness of the nonfarm segment in many of the eastern states. The first is the more extended size of the hotel and restaurant business. The proliferation of roadside eating and drinking enterprises, especially when the hard-core commodity-producing sectors (such as

Taken from 1991 Population Census Reports. bNoncrop activities include raising of livestock. pigs, poultry. etc., and production of milk, agricultural services, hunting, trapping, game propagation, forestry and logging, and fishing (1990 Economic Census). Field crops show the total of cultivators and agricultural laborers (1991 Population Census). cTaken from 1990 Economic Census Reports. (Figures may, therefore, be taken as a rough approximation of the relative interstate positions rather than exact estimates for each state.) Sources: Govt. of India (1991, 1996a).

case, the fact of a very limited branching off into noncrop areas remains. Moreover, nonagricultural sources of employment are and will be more important for enhancing the earnings of rural households in India, especially those in the rainfed riceproducing areas in eastern India. Understandably, the picture is highly disparate among all states and among the eastern states. For example, among the eastern states, nonagricultural employment accounts for as high as 25-30% in Assam, Tripura, West Bengal, and Arunachal Pradesh. It languishes as low as 1520% in Nagaland, Mizoram, and Orissa and at a disparately low level of 11.8% in Bihar. It needs to be clarified, however, that a final view about the impact of nonagricultural employment on overall productivity, earnings, and well-being of rural households cannot be framed before we

51

Table 7. Census.

Distribution of usually employed rural workers engaged in major nonagricultural activities, 1990 Economic

Percentage of rural workers engaged in State Manufacturing Wholesale Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India
a

Trade Retail 17.1 16.1 15.5 28.5 18.7 18.4 15.3 19.5 20.9 18.8 18.4 19.2 18.2 20.7 15.4 20.3 20.1 19.3 16.4 23.2 22.1 22.2 19.7

Hotel and restaurant

Transport and communication 1.1 2.0 1.1 0.9 2.7 2.4 2.5 1.5 1.9 1.4 2.5 3.1 3.6 1.7 0.6 1.7 1.6 2.4 1.4 3.0 1.3 3.9 1.9

Community, social, and personal services 27.5 47.9 39.3 30.4 28.9 31.8 42.5 30.4 31.7 29.9 34.5 42.1 40.0 63.0 60.8 32.2 41.6 37.5 26.7 45.2 25.6 25.1 30.3

Othersa

44.6 19.4 32.3 32.4 39.1 39.5 28.9 36.8 30.9 41.4 33.0 27.3 9.8 6.6 15.4 36.2 29.6 30.1 43.5 20.3 45.4 38.2 37.9

1.0 0.3 0.8 0.5 2.5 1.1 0.6 1.0 1.9 0.8 1.2 0.1 3.8 0.3 2.9 0.7 1.3 0.7 2.4 0.3 0.5 3.5 1.3

4.1 6.9 3.8 3.9 1.7 1.9 4.1 6.2 7.4 2.7 3.9 4.9 12.8 4.7 3.9 4.7 1.5 3.1 5.4 4.9 2.7 2.8 3.9

3.9 7.4 7.3 3.4 6.4 4.9 6.1 4.6 5.3 4.9 6.5 3.1 11.8 3.1 1.1 4.2 4.5 6.9 4.2 3.1 2.5 4.2 4.7

Includes rural persons engaged in mining + quarrying, construction. storage and warehousing and finance, insurance and real estate.

Source: Govt. of India (1996a).

manufacturing) offer little hope to expanding the labor force, is a common phenomenon of weak economies. Compared with the Green Revolution states and the industrially progressive states (such as Gujarat, Maharashtra, and Karnataka), the employment size in this sector is far higher in most of the eastern states, except in West Bengal. The second fact relates to the size of the trade sector and the relative weights of its wholesale and retail components. In general, in most eastern states, against the backdrop of an extremely narrow employment base in wholesale trade, the overarching size of the retail trade cannot escape attention. The cases of Bihar, Manipur, Mizoram, Orissa, and Tripura are obvious. The “extended canvas” of the retail trade in these areas smacks more of desperation (approximately mirrored through the overwhelming size of self-versus wage-paid employment) rather than the requirements of an expanding economy. We have more to say about the wholesale-retail trade relationship in these areas later in the paper.

Rural area’s share in nonfarm employment In terms of NSS surveys on nonfarm employment, “employing units” fall under two broad categories: those being run exclusively by members of the household themselves (called own-account enterprises: OAEs) and those employing one or more hired hands, on a fairly regular basis, in addition to family workers themselves (called establishments: ESTTs). Being exclusively family enterprises, the OAEs are expected to be small. with fewer workers employed, compared with the ESTTs (Table 8). In most states, the rural areas have a high share of the OAEs and employment therein. It does not appear that the rural areas are devoid of ESTT nonfarm enterprises in eastern and other states, the interstate variations notwithstanding. It is remarkable to see, in each state, a very close correspondence between rural area’s share in the number of enterprises and that of employment. under the OAE segment. This indicates that the

52

Table 8. Some features of nonagricultural activities/employment in the rural economy of India. 1991 Economic Census. Workers hired to total workers employed in rural area (%) Manufacturing Wholesale Retail trade Personal services e Percentage of manufacturing units employing Wholesaleretail trade worker f 10 or more workers in ESlTs 3 or more workers in OAEs 7.4 1.1 9.0 8.8 5.1 4.2 1.6 7.3 2.6 9.4 4.7 3.6 3.6 2.1 2.2 7.8 3.9 5.9 7.6 4.9 11.3 8.0 7.5 46.9 77.3 78.6 37.8 62.9 67.7 39.4 58.1 66.9 22.3 56.0 23.5 60.0 24.4 72.4 22.0 63.0 30.6 54.5 43.1 47.5 29.9 46.5 10.5 33.9 18.0 11.8 19.4 10.0 17.7 20.2 24.0 8.2 14.9 4.3 22.4 12.9 25.6 8.6 12.3 9.9 21.7 13.9 10.2 10.6 13.9 23.0 17.6 12.5 12.0 30.0 21.3 10.6 20.1 10.6 16.4 20.5 6.7 9.4 7.2 17.1 90.0 15.8 13.1 17.4 8.5 14.8 12.0 17.5 42.6 87.0 43.9 32.1 61.2 50.0 59.4 47.7 43.7 51.8 68.1 43.7 70.9 51.9 82.3 30.6 43.7 34.5 36.1 30.0 46.1 17.0 38.4 54.2 94.4 63.9 72.8 72.6 70.9 80.2 66.5 74.9 67.6 70.0 89.1 82.4 91.3 93.9 69.1 70.7 72.6 73.7 78.7 63.3 58.1 67.6

Percentage share of rural areas in nonagricultural enterprises/employment ESTTs b EMP d 70.4 78.8 73.3 69.2 43.9 46.3 81.5 58.7 71.5 61.4 50.0 53.0 66.2 36.6 46.2 82.5 42.2 56.8 57.8 72.6 52.5 71.6 60.7 53.6 73.1 66.3 49.7 37.4 35.4 74.3 50.0 58.8 50.9 32.2 58.5 63.0 48.9 56.2 68.5 34.6 52.2 44.2 66.8 35.9 44.5 44.9 46.6 63.5 63.9 38.7 28.7 34.0 58.6 37.6 51.2 32.2 21.5 43.1 44.5 26.5 36.1 49.3 28.0 36.1 37.8 52.1 34.8 30.6 34.4 ENT EMP

State

OAEsa

ENT c

Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India

69.0 79.9 73.6 68.9 44.4 48.0 82.8 59.9 71.9 61.2 51.4 51.6 64.1 34.7 47.0 80.1 45.1 58.1 57.2 71.8 52.4 71.1 60.5

16.6 59.1 20.3 53.0 7.5 16.7 27.9 19.3 10.8 23.8 15.2 130.0 4.8 80.8 5.3 29.1 16.1 27.9 6.8 75.3 43.6 6.3 13.8

a OAEs = own-account enterprises (run by members of the household without hiring any worker, on a fairly regular basis), bESi-rs = establishments (run by employing at least one hired worker. on a fairly regular basis). c 5NT = no. of enterprises. dEMP = employment. e Personal services means community, social, and personal services. f No. of persons employed in rural retail trade for every one person employed in rural wholesale trade. Sources: Govt. of India (1996a,b.d.f).

employment content of rural OAEs is not different from that of urban OAEs. As we see in Table 8, 90-95% of rural OAEs employ up to two workers only, and the position is summarily the same with urban OAEs. The rural-urban contrasts are far more striking, however, for ESTTs. The big gap between rural area’s share in the number of ESTTs and employment therein is relatively more pronounced in the eastern states of Orissa, Nagaland, Mizoram, Meghalaya, Manipur, Bihar, West Bengal, and Tripura and points to lower employment capabilities of such units in the rural areas. Assam is an exception. It enjoys a widespread canvas of ESTTs, in both rural and urban areas, in terms of employment. A fairly sizeable number of ESTTs employing 100-199 or 200499 or even 500 or more workers are located in the state, partly in the public sector and partly in the oil-extracting private sector. This makes Assam strikingly different from the rest of the eastern states (Govt. of India 1996). The relative disadvantages of the rural areas are well known. Apart from technological gaps, inadequate and irregular supply of raw materials, lack of organized marketing channels, imperfect knowledge of market conditions, inadequate availability of credit, infrastructural constraints (including power shortages and breakdowns), deficient managerial skills, and so on are the handicaps commonly reported for the wide mass of rural industries (Chadha 1992). The employment disadvantage in rural areas, although commonly observed in other states as well, is really small in many noneastern states (e.g., Gujarat, Haryana, Kerala, Punjab, Tamil Nadu, Uttar Pradesh, etc.). The pattern of hiring workers for nonfarm jobs is strikingly different between the group of eastern states and the rest of the rural economy, on the one hand, and among eastern states themselves on the other. Let us consider the CSPS segment. In many of the eastern states (such as Arunachal Pradesh, Nagaland, Mizoram, Manipur, etc.), more than 90% of the workers in CSPS enterprises are hired hands. The figures are fairly high for other eastern states as well. As pointed out earlier, the public sector plays a very important role in boosting many employment opportunities under the CSPS 54

sector. The incidence of hiring workers for manufacturing, however, follows a different pattern. In many eastern states, manufacturing enterprises are managed by family workers themselves (see, for example, the low figures for Orissa, Manipur, Mizoram, West Bengal, and Bihar, Table 8). Again, the retail trade is run largely by family hands, and, in relative terms, the eastern states do this on a much higher scale than the rest of rural India. Finally, Table 8 clearly reflects the smaller size of manufacturing enterprises, in most eastern states, whether under the OAE or ESTT umbrella. For example, in not more than 4-5% of own-account manufacturing enterprises, three or more workers are at work in Arunachal Pradesh, Mizoram, Nagaland, Manipur, Meghalaya, and Tripura. A similar handicap with respect to ESTTs is discernible for many of the eastern states.

Hired workers in the farm/nonfarm sector
Table 9 shows the extent to which farm and nonfarm activities are carried out with family and hired workers in different states. The eastem states have many contrasts, both among themselves and with the rest of rural India. For many of the eastern states (e.g., Bihar, Orissa, Tripura, and West Bengal), hired labor accounts for a very large proportion of total workers for crop production, whereas, for the same states, the noncrop sector sustains itself largely through family labor. On the other hand, in Arunachal Pradesh, Assam, Meghalaya, Mizoram, and Nagaland, crop production is sustained largely by family workers, whereas noncrop enterprises are manned quite substantially by outside labor. It is interesting to see a clear substitution between family and outside workers in one group of eastern states, and vice versa for the other, for crop or noncrop activities. In most other (noneastern) states, crop production depends substantially on outside labor, while the noncrop sector relies on family labor. This is consistent with common perceptions about the much higher amount of work and its more demanding nature (including the needed weather-induced flexibility in the supply of labor) in farm production compared with the less

Table 9. Pattern of hiring rural male and female workers in different sectors of India s rural economy: 1991 population census and 1990 economic census data. Percentage share of hired female workers to total female workers Crop sector Noncrop sector Nonagricultural activities 40.2 72.6 79.0 40.9 51.1 79.2 70.4 57.2 75.3 26.7 53.3 30.2 52.1 57.6 73.9 26.6 84.7 45.1 54.9 71.8 39.3 29.2 47.6 51.9 38.1 23.2 26.4 39.3 13.9 18.4 50.7 32.6 47.3 53.8 57.3 43.4 34.8 45.2 35.5 6.2 37.7 48.5 23.2 24.7 19.8 39.3 Crop sector Noncrop sector 30.1 3.9 9.2 12.5 39.5 5.9 5.3 53.8 10.8 28.5 13.8 17.5 14.8 15.3 13.3 15.0 5.9 12.0 28.0 2.9 7.4 7.6 32.3 9.2 18.4 18.1 8.4 8.6 6.9 5.5 41.2 15.9 6.4 6.0 2.9 18.8 11.7 24.1 8.9 12.4 3.2 14.9 5.4 3.9 4.2 14.6 Percentage share of hired female workers to total hired workers Nonagricultural activities 29.9 11.7 16.6 13.5 14.9 20.2 14.8 27.4 37.6 15.3 15.3 23.1 21.5 26.1 20.3 15.4 18.6 12.2 34.2 20.1 10.4 15.9 21.0

State Noncrop sector Nonagricultural activities 42.2 79.3 62.7 42.7 55.6 57.0 56.7 51.7 57.1 37.5 53.8 48.1 63.2 67.5 77.6 37.9 56.2 46.6 52.5 52.2 43.7 23.5 46.9 72.0 5.6 20.7 64.5 58.4 33.7 2.4 64.7 86.2 41.6 52.0 15.7 18.1 2.0 5.7 67.4 54.9 20.7 70.8 44.7 42.7 69.2 55.2

Percentage share of hired workers to total workers engaged in

Crop sector

Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Havana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India 10.8 58.5 30.5 15.1 10.8 20.5 19.1 34.3 27.1 8.9 12.1 5.8 33.8 24.3 44.2 13.3 28.7 6.8 17.7 24.8 12.5 13.8 16.1

58.6 7.3 19.9 45.97 39.9 32.3 5.0 44.8 66.8 30.6 44.3 12.4 18.8 2.7 6.3 39.1 40.4 14.0 57.4 37.7 25.9 45.9 39.6

Sources: Govt. of India (1991, 1996a).

rigid and flexible schedules set for many noncrop activities. The picture is undoubtedly different for nonfarm activities. A substantial proportion of rural households do not need financial capability, technical know-how, or market intelligence to mount nonfarm enterprises on their own. They are nonetheless ready to work as wage-paid workers in nonfarm jobs, depending on their family occupation, educational attainment, job locale and its technical requirements, and so on. Specific to the eastern states, it is interesting to see a much higher proportion of hired labor in Arunachal Pradesh, Assam, Meghalaya, Mizoram, and Nagaland. The three major eastern states (Bihar, Orissa, and West Bengal) are more similar to the noneastern states than to their adjoining sisters. This could be partly because these three states are relatively “more industrialized” and partly because of the fairly noticeable base of industrial activity in the rural areas for which hired labor has been an indispensable prerequisite. For the rural people in the eastern states as a whole, the outlets of nonfarm employment are open through wage labor rather than self-employment. For the “seven eastern sisters,” the handicap is particularly severe for manufacturing. What has been said about total of hired labor is also true for hired female workers (Table 9). The two contrasting positions within the group of eastern states need to be reiterated for all three categories of employment. In West Bengal, Bihar, and Orissa, nearly two-thirds of the female workers engaged in the crop sector have to work as agricultural wage laborers. In these states, the proportion of hired female workers doing nonfarm activities is relatively lower than in the other eastern states, signifying economic distress of some kind. Perhaps, the females here are under greater pressure to mount “selfemploying” nonfarm ventures because wagepaid agricultural employment is not sufficient to sustain their families. Nonetheless, the situation in the remaining eastern states (see particularly Arunachal Pradesh, Mizoram, and Nagaland) is relatively more reassuring. Here, the incidence of nonfarm wage employment for females is much higher (in some cases, exceeding 70%)

than in the large states of West Bengal, Bihar, and Orissa. The hired female workers do not, however, constitute a very high proportion of all hired workers. In general, hiring of females is preferred in the crop sector, whereas more males are involved in nonfarm activities. This validates the common perception of more women workers staying in agriculture and more men workers venturing out to nonagricultural activities. This applies to the eastern states and to the rest of the country, the varying state-level figures notwithstanding.

Female workers
It is evident in Table 9 that female cultivators are not common in the states of West Bengal, Bihar, and Orissa. This is also true in Assam. The other eastern states (Arunachal Pradesh, Manipur, Meghalaya, Mizoram, and Nagaland) have a higher proportion of female cultivators; the problem of agricultural development in these states would thus have an additional element of “gender infirmity,” which manifests itself in many different ways. A fairly substantial proportion of the females work as agricultural laborers in practically all the eastern states, except in Punjab and Haryana. Females likewise constitute a fairly large share of noncrop agricultural and nonagricultural employment. On the whole, female participation in all walks of rural economic life is evidently substantial. In addition, a sizeable proportion of female workers in several eastern states have to shoulder the responsibility of agricultural entrepreneurship, which makes the task of enhancing agricultural growth more difficult. If we look into the sectoral distribution of female workers themselves, a majority of them are “dumped” in agriculture; in most states, the proportion of female workers in nonfarm jobs is either very low or negligible (Table 10). West Bengal in the east, Punjab in the north, and Kerala in the south are exceptions. In the eastern states, although two-thirds or more (in some cases. more than nine-tenths) of the female workers are engaged in agriculture, they are mainly cultivators. In contrast, female

56

Table 10. Gender differences in some aspects of employment of rural workers in crop and noncrop sectors of agriculture and nonagricultural activities, 1991 Population Census and 1990 Economic Census. Percentage share of females among State Cultivators Agricultural laborers 51.9 38.1 23.2 26.4 39.3 13.9 18.4 50.7 32.6 47.3 53.8 57.3 43.4 35.8 45.2 35.5 6.2 37.7 48.5 23.2 24.7 19.8 39.1 Noncrop production workers 35.2 12.5 15.4 22.5 49.8 17.5 18.7 44.8 10.4 39.5 27.6 35.0 26.6 31.9 24.4 22.4 13.9 25.8 33.1 13.6 23.6 24.8 35.6 Nonagricultural workers 31.3 12.8 13.1 14.0 16.3 14.5 11.9 24.7 28.5 21.5 15.5 36.7 26.0 30.6 21.4 21.9 12.4 12.7 32.7 14.6 11.4 18.2 20.8 Percentage distribution of female workers as Cultivators Agricultural laborers 63.2 5.1 14.1 60.1 49.8 29.0 2.3 55.1 41.7 39.1 48.6 12.3 15.6 1.9 5.3 57.5 34.1 19.4 59.8 34.8 38.3 45.5 47.9 Agricultural workers 87.8 92.0 68.1 93.2 85.3 86.2 92.4 85.1 48.4 94.0 93.5 78.1 86.0 94.0 92.4 85.3 62.1 93.5 84.4 77.9 89.6 65.7 86.9 Percentage of hired females among total female workers Noncrop sector 9.2 18.4 18.1 8.4 8.6 6.9 5.5 41.2 15.9 6.4 6.0 2.9 18.8 11.6 24.1 8.9 12.4 3.2 14.9 5.4 3.9 4.2 14.6 Nonagricultural activities 40.2 72.5 79.0 41.0 51.1 79.2 70.4 57.2 7.3 26.7 53.3 30.2 52.1 57.6 73.9 26.6 84.7 45.1 54.9 71.8 39.3 29.2 47.6

Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Havana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal All India

28.6 50.7 22.2 12.3 18.6 13.1 38.6 22.5 10.5 29.3 39.4 43.6 45.6 47.4 50.4 10.9 3.5 23.6 26.9 17.4 11.5 7.5 20.8

24.6 86.9 54.0 33.1 35.5 57.2 90.1 30.0 6.7 54.9 44.9 65.9 70.5 92.1 87.1 27.8 28.0 74.2 24.6 43.1 51.3 20.2 39.9

Sources: Govt. of India (1991,1996a).

agricultural laborers are numerically far stronger than female cultivators in West Bengal, Bihar, and Orissa.

Land/labor productivity in agriculture
Agricultural productivity differs from state to state. We can look at productivity in terms of per hectare or per worker output. Table 11 gives interstate variations in both parameters. Three important individual crops (rice, maize, and wheat) and four crop groups (total of pulses, oilseeds, cereals, and food grains) are chosen. The productivity gaps in most eastern states are apparent. For example, rice yield is lower than the national average (indeed much lower than in &he Green Revolution states) in Arunachal Pradesh, Assam, Bihar, Meghalaya, Mizoram, Nagaland, and Orissa, and only marginally better in Tripura and West Bengal. In five out of six eastern states reporting maize cultivation, yield is behind the national average. The eastern states seem to be doing better in pulses, but, for

crop aggregates (total of cereals, food grains, and oilseeds), their performance is inferior. The relative performance of the eastern states is assessed using I-year data (though the latest available) only. Without doubt, the year-toyear up-and-down swings in yield, most ostensibly of the main rice crop, would be far more frightening. The magnitude of these yearto-year yield variations render production in these states unstable. Agriculture here is thus of a different nature. To be fair, therefore, the relative agricultural backwardness of the eastern states is hardly portrayed by I-year figures for crop yields. Much more statistical work needs to be done to capture the real problems of their agriculture, especially the instability that is built into their production base. This is simply beyond the purview of this paper. Rough estimates of per worker productivity (proxied by net domestic product originating in the crop/primary sector divided by total workers) reveal the same backwardness (Table 11). Most of these states lag behind the national

57

Table 11. Some aspects of land and labor productivity in Indian agriculture. Av yield of individual cropdcrop groups (quintals ha-1) 1995-96a Per worker productivity: Rs at constant 1980-83 pricesb Total food grains Total oilseeds Crop sector alone 1990-91 total primary sector Crop sector alone 1980-81 total primary sector Wheat Total pulses Total cereals Land-person Operated area per person (ha in 1992) ratio Cropped area per rural person (ha in 1992)

State

Rice

Maize

4.2 6.9 6.1 6.9 6.8 12.8 9.2 10.6 5.2 6.2 2263 3171 2964 1918 3369 9455 3219 -

25.5 13.3 6.8 17.4 9.9 18.3 21.5 15.9 36.3 13.4 14.6 16.1 12.8 38.8 25.0 24.5 26.1 24.9 7.7 10.7 5.4 8.2 4.2 3.7 4.7 4.7 8.0 8.0 4.6 -

11.1 20.2 22.2 36.9 14.4 14.6 6.9 -

2290 10499 2732 2072 3482 941 2 3749

-

25.0 11.5 13.5 13.7 14.5 22.2 13.5 18.6 23.8 19.7 11.0 16.9 25.1 11.6 15.4 13.2 13.8 31.3 4.2 6.8 6.0 8.7 7.8 4.6 6.4 9.1 8.9 5.2 12.0 8.0 14.7 7.2 8.6 7.8 8.5 2007 1600 281 9 1846 3686 7320 2586 3599 2696 461 3 2004 2498 2314 1611 1006 1123 2507 7540 281 8 1529 2960 2899 3113 2869 2780 561 6 2550 2866 2146 1685 2011 1164 1966 10606 4026 2002 3447 3207 3658 3291 2049 4615 2665 1956 3809 7272 3344 4250 2700 4329 2335 2715 2459 1575 1117 1260 2748 7569 2770 1517 3470 2997 338 1 3048

12.2 -

10.9 12.2 28.0 8.9

Andhra Pradesh Arunachal Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Jammu and Kashmir Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Mizoram Nagaland Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal India

33.9 20.1 18.7 20.0 18.6

13.8 15.7

21.0 12.1 13.4 15.6 12.5 27.2 16.7 15.9 15.3 19.5 11.9 10.0 25.1 11.7 15.9 12.4 13.3 35.2 9.8 26.6 20.0 20.7 20.3 17.3

17.2 5.6 12.7 14.5 10.8 23.5 16.2 10.5 12.9 12.2 10.8 8.1 25.1 11.6 15.9 12.4 10.9 38.4 8.1 21.4 19.4 18.9 21.3 15.0

2733 481 8 2634 3103 2426 1619 187628 1807 2188 10603 4076 2027 3778 3227 3947 3429

0.19 0.22 0.13 0.10 0.27 0.37 0.14 0.09 0.27 0.06 0.34 0.32 0.10 0.13 0.22 0.08 0.17 0.16 0.47 0.11 0.06 0.15 0.10 0.20

0.26 0.34 0.19 0.13 0.39 0.47 0.21 0.18 0.40 0.14 0.49 0.44 0.15 0.1 7 0.29 0.22 0.36 0.53 0.57 0.19 0.20 0.23 0.18 0.30

a

Govt. of India (1998a,b),

b

Population Census 1991 (see Table 6).

levels in terms of per worker productivity in the crop sector or the primary sector as a whole. The productivity gaps are more pronounced for Bihar, Manipur, Meghalaya, Nagaland, and Orissa. The relative backwardness of these states seems to persist (see 1990-91 vs 1980-81 figures). Incidentally, per worker productivity in the noncrop sector is higher than that in the crop sector. This seems to be a decisive advantage of these states compared with many others in the rest of the country. Note in particular the levels of the two types of productivity (Table 11) in Punjab, Haryana, Uttar Pradesh, Tamil Nadu, Andhra Pradesh, and Karnataka versus differing levels in Bihar, Manipur, Nagaland, Orissa, Tripura, and West Bengal. In other words, most of the eastern states can make up for the low labor productivity in the crop sector through diversification to noncrop activities. As seen earlier (Table 8), the noncrop sector in these states is still a tiny fraction of total agriculture, and the future policy thrust must aim at moving more people from crop production to noncrop enterprises, depending on the physical environment, local resource base, infrastructure support, and marketing network available in each state. The biggest consideration, inter alia, for embarking upon crop to noncrop diversification is the extremely low land base of most of the eastern states. The land-person ratio, particularly for operated area per person, points to this most severe constraint (Table 11).

Productivity in nonagriculturalactivities
We have noted earlier (Table 7) that manufacturing is the most dominant component of the rural nonfarm sector in India, followed by CSPS, trade, and so on. It would be useful to look into rural manufacturing to get insights into the productivity and earning levels of rural nonfarm workers. Table 12 provides some firsthand information obtained from the 1994-95 nationwide survey on unorganized manufacturing enterprises in rural India. It may be mentioned, in passing, that manufacturing in rural India is covered mostly under what is known as the “unorganized segment.” Own-

account manufacturing enterprises (OAMEs) in rural India dominate in each state (Table 12), in terms of both numerical strength and proportion of employment offered. Without doubt, such enterprises are run inside and around residential premises. The value of fixed assets per enterprise is much lower in most eastern states; the gaps between these states and Gujarat, Haryana, Punjab, etc., need to be analyzed in particular. The investment gaps are reflected in value-added gaps. The most appalling gaps are discernible in per worker salary/wages for hired and paid household workers. Such earnings are ridiculously low for Bihar, Orissa, and Tripura (eastern states) and for Jammu-Kashmir, Madhya Pradesh, and Rajasthan (noneastern states). In general, the relative weaknesses of the eastern states are quite apparent in this segment of the rural economy, just as was seen earlier for agriculture. A fact that might account for the lower per worker productivity/earnings or gross value added in the eastern states is that, in their (unorganized) manufacturing sector, part-time workers make up a much higher proportion of total workers than in most other states. For example, in Assam, Nagaland, and Orissa, more than 25% of the workers are part-time workers; in West Bengal, the percentage is 17; in Tripura, it is 12, and so on. In most of the eastern states, part-time employment in rural areas is much higher than in urban areas (Table 12). Finally, an employment feature that reveals a marked structural weakness of the rural manufacturing segment in the eastern states is the high percentage of unpaid girl workers to total unpaid workers (Table 12). The high numbers show that the OAMEs here are household/cottage enterprises undertaken as an adjunct to the main work (agriculture), more in the hope of earning some additional income for the household rather than running them as independent commercial entities. The small and ad hoc production routines including work on a part-time and “as-is-convenient” basis are natural consequences of the use of outdated technology, inferior products, and catering to local or nearby markets.

59

Table 12. Salient features of unorganized manufacturing (UM) enterprises, 1994-95 NSS data.

State Rural Urban

% Share of OAMEa among all UM enterprises

% Share of employment in OAME among all UM enterprises 83.1 84.7 90.0 60.3 69.3 83.9 85.8 62.6 51.9 90.4 71.9 99.0 89.4 80.1 96.3 72.1 88.8 62.5 88.6 78.1 86.0 507 2110 391 8167 874 2356 1612 3.4 24.4 10.3 3.3 9.4 13.6 22.7 4.1 17.7 6.5 9.4 6.5 9.8 30.6 38.3 9.5 7.2 5.9 11.9 10.7 17.0 79.8 87.9 85.0 69.2 62.1 57.9 67.5 81.6 64.0 83.7 75.3 65.0 87.4 61.1 92.2 53.2 81.4 86.8 86.5 74.4 90.8 57.7 43.1 50.7 17.0 26.5 39.7 21.7 61.4 59.5 47.6 39.3 55.7 17.1 41.4 53.0 32.6 47.9 68.8 57.8 55.4 56.4 8623 5875 7190 31928 25732 14550 11182 11891 15618 9461 15524 7773 6239 17380 4024 18659 18403 18601 3663 10323 4385 7067 8161 8856 12899 15828 6040 6661 7485 8926 9557 11548 6156 14210 13929 4553 15243 14448 10708 6880 9852 8758 1678 1022 254 1991 2019 10071 684 2302 11241 792 3400 52.9 58.2 65.0 42.4 42.6 12.5 49.8 65.9 68.9 35.4 35.1 73.8 40.8 65.0 54.0 67.4 44.4 57.6 45.0 52.3 58.2 1.4 2.4 3.7 1.4 3.1 2.3 1.8 0.8 1.3 0.9 1.9 1.9 6.1 2.3 3.9 2.4 2.5 1.7 1.4 1.6 2.8

% of HH enterprises run inside residence or as moving enterprises

Value of fixed assets (less land) per enterprise

Per enterprise estimate of gross value added in OAMES

Per worker % of part% of salary/wages time workers females for hired and engaged in among paid unorganized part-time household manufacturing workers: workers in rural units rural areas OAMES

Ruralurban ratio of the proportion of part-time workers

% of unpaid girl workers to total unpaid workers in rural unorganized units 9.3 12.1 7.2 5.3 3.7 2.3 4.5 8.2 5.1 4.4 3.0 27.0 5.1 16.3 14.8 5.4 6.3 8.8 13.1 9.8 10.7

Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Jammu and Kashmir Karnataka Kerala Madhya Pradesh Maharashtra Manipur Meghalaya Nagalang Orissa Punjab Rajasthan Tamil Nadu Tripura Uttar Pradesh West Bengal

85.6 89.0 96.0 87.8 89.8 94.6 93.0 82.6 76.9 94.6 89.2 99.7 94.5 92.0 97.8 88.5 94.4 84.8 91.7 90.0 93.0

a

OAME = own-account manufacturing enterprises. Source: Govt. of India (1997e,h).

Concluding remarks
The eastern states are mostly rural and agriculture is their mainstay. Agriculture is plagued by many problems-it unstable and is has low yield/productivity. The noncrop sector is still a tiny fraction of the total rural economy. The inherent weaknesses of the nonfarm sector, especially rural manufacturing, are no less daunting; in some of these states, per worker nonfarm earnings are extremely low. On the employment side, much of the work force is still confined to agriculture. Employment under an unstable production regime is both undependable and less remunerative. People work hard but do not have enough earnings. So, they have to look for opportunities outside agriculture, where opportunities are not that promising either. Not only is the proportion of the work force engaged in nonfarm activities low, but many people work on a part-time basis. Part-time employment is a common practice among female workers of this region. Overall, their economies suffer from many structural handicaps. The well-being of rural people in these areas is thus perpetually at stake. It is important that agriculture be put on top of the policy agenda. A two-pronged policy intervention is needed, one at the institutional level and the other at the household level. A more responsive and supportive institutional environment has to be in place so that what is technically feasible becomes financially possible for individual farmers. A people-government partnership, on an enduring basis, is a must for these areas. Once agriculture improves, the nonagricultural segment of the local economy would improve as well. After all, the intersectoral forward and backward linkages, though limited, do operate at the local level.

References
Bhalla GS, Singh Gurmail. 1998. Recent developments in Indian agriculture: a district level analysis. Centre for Regional Development, Jawaharlal Nehru University, New Delhi. Chadha GK. 1992. Adoption of improved technology in India’s cottage industries: constraints and impact. In: Rizwanul Islam, editor. Transfer, adoption and diffusion of technology for small and cottage industries. ILO-ARTEP, New Delhi. Chadha GK. 1993. Nonfarm sector in India’s rural economy: policy, performance and growth prospects, VRF Series No. 220. Tokyo: Institute of Developing Economies. Chadha GK. 1994. Employment, earnings and poverty: a study of rural India and Indonesia. Indo-Dutch Studies on Development Alternatives. New Delhi: Sage Publications. Chadha GK. 1996. The industrialization strategy and growth of rural industry in India. New Delhi: ILO-SAAT. Govt. of India. 199 1. Census of India 1991, Series-1, INDIA, Paper 3 of 1991, Registrar General (Census). Govt. of India. 1962. Monthly and annual normals of rainfall and rainy days based on records from 1901 to 1950. Govt. of India Press. Govt. of India. 1995. Economic census 1990: All-India Report. New Delhi: Ministry of Planning, C.S.O. Govt. of India. 1996a. Economic census 1990: number of nonagricultural enterprises and employment therein in states/union territories according to major activity groups. New Delhi: Ministry of Planning, C.S.O. Govt. of India. 1996b. Economic census 1990. State/Districtwise Aggregates of Principal Characteristics of Enterprises. New Delhi: Ministry of Planning, C.S.O. Govt. of India. 1997a. Dwelling in India. NSS 50th Round, July 1993-June 1994, NSSO, March.

61

Govt. of India. 1997b. Employment and unemployment in India, 1993. NSS Fiftieth Round (July 1993-June 1994), NSSO, March. Govt. of India. 1997c. Operational landholdings in India, 1991-92 salient features. NSSO, March. p A 18-A34, A46-A56. Govt. of India. 1997d. Poverty in India: methodology and incidence. Planning Commission (mimeo.), Annexure V, March. Govt. of India. 1997e. Salient features of unorganized manufacturing enterprises in India. NSS 51st Round, July 1994-June 1995, NSSO, September. Govt. of India. 1997f. Slums in India. NSS 50th Round, July 1993-3une 1994, CSO, March. Govt. of India. 1997g. Some aspects of seasonwise operation of holdings. Draft Report No. 414, NSS 48th Round, JanuaryDecember 1992, January. p A102-Al67. Govt. of India. 1997h. Unorganized manufacturing sector in India, its size, employment and some key estimates. NSS 51st Round, July 1994-June 1995, NSSO, September. Govt. of India. 1998b. Statistical abstracts of India, 1997. C.S.O., Ministry of Planning. p 68-70. Govt. of India. 1998a. Economic survey, 199798. Ministry of Finance. Thorat S. 1993. Technological change and regional differentiation. New Delhi: Khama Publishers. Singh Tapeshwar. 1978. Drought-prone areas in India. New Delhi: People’s Publishing House.

The phenomenon of less incidence of poverty among landless in comparison to marginal farmers is observed in developed as well as in backward areas, simply because of high occupational diversity of landless people. They are free from land and can move throughout the year to seek employment. The linkage between productivity growth and non-farm employment for national and regional levels is obvious, but one may not observe such linkage in village-level data. However, we must recognize that there is a limit to the generation of non-farm employment through farm productivity growth alone. Question: Risk in rice production for small farmers is very high, and therefore they diversify to non-farm activities. What would be opportunities for making an impact through research and improving productivity of rice sector? When the farm size is small, actual income from rice will be quite low even with productivityincreasing technologies. Answer: Yes, I agree with you. Given the size of these farms (small and marginal), it is rather difficult to have adequate income even with the best possible technology intervention through research. Something more substantive should be thought of.

Answer:

Discussion
Question: Are the non-farm employment opportunities independent of growth in the productivity in farm sector? The studies have shown that unless you trigger growth in farm sector, non-farm employment opportunities will not grow. This kind of link was not visible from the information shown here.

Author’s address: Jawaharlal Nehru University, New Delhi, India. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAPARRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

Notes

62

Crop insurance: a policy perspective
P. K. Mishra

In agriculture, risk can arise from unpredictable variations in yield or price or both. Crop insurance addresses the problem of yield risk that is the dominant source of risk in rainfed agriculture, This paper provides an overview of issues related to crop insurance, especially in the context of India. Despite the conventional wisdom regarding high costs of crop insurance schemes, an analysis of the Indian comprehensive crop insurance scheme shows that this scheme has a positive effect on crop inputs, output, and incomes. The gross benefit was found to be higher than the value of the government subsidy. This paper suggests alternatives for designing crop insurance schemes that could improve their financial viability.

As an institutional response to agricultural risk, crop insurance exists in several countries, both developed and developing. Agricultural risks can take the form of output or yield risk and price risk. Crop insurance addresses the problem of yield risk that predominates in rainfed agriculture. The objective of this paper is to examine certain conceptual and operational issues relating to crop insurance with a policy perspective.

Crop insurance as a public policy
Crop insurance, mainly as hail insurance, emerged in Europe, the United States, and Canada more than a century ago. Coverage of multiple risks was introduced by some private companies in the U.S. in the early part of the 20th century. Public-sector crop insurance was a later phenomenon and was introduced in countries such as Japan and the U.S. in the 1930s. In the developing countries of Asia, Africa, and Latin America, however, it was introduced much later. Crop insurance was, from the early part of the 20th century, looked upon as an instrument of public policy to address the problem of agricultural risk. The pioneering work of J.S. Ckakravarti, culminating in the publication in

1920 of his book, Agricultural Insurance: A Practical Scheme Suited to Indian Conditions, was an epoch-marking event. The first half of the 20th century witnessed pioneering studies, conceptualization, exploration, and experiments in the realm of crop insurance. Optimism and excitement have been immense regarding crop insurance as a strategy for agricultural risk management. In the 1960s and 1970s, policymakers and analysts envisaged crop insurance as an effective method to enable farmers, especially those in developing countries, to tackle risks and thereby encourage adoption of new technology. The target was small farmers who are more prone to agricultural risks and who choose more secure (though less profitable) cropping activities. These risks also pose a threat to their survival. Another consideration is the link of insurance to agricultural credit-lending institutions would only lend to small farmers if insurance is included in the deal. Thus, crop insurance was chosen as a policy intervention for both efficiency and welfare considerations. Consequently, multiple-risk crop insurance programs came into being in many countries. The sense of optimism did not last very long. Some presented “a case against crop insurance in developing countries” (Roumasset

63

1978). Others cautioned governments against introducing large crop insurance schemes on the grounds that all-risk crop insurance programs have had a disappointing performance (Hazell et al 1986). Similar views were expressed in the early 1990s (Hazell 1992, Wright and Hewitt 1993). Most of these studies argue that crop insurance programs are too costly, financially not viable, and thrive on government subsidy. They contend that such insurance programs do not result in welfare gains because the cost is much larger than the benefit. Other studies, however, do not subscribe to this view. An important issue is whether some types of crop insurance programs can and do work, even though others cannot and do not. The view that crop insurance does not work seems to be derived from inadequate analysis of a particular type of scheme (i.e., public-sector allrisk insurance, based on the individual approach). In reality, crop insurance exists in a variety of forms. Hail insurance has been in existence for more than a century. Crop insurance covering specific risks in the private sector has shown a rising trend in recent times (Gudger 1991). Further, some public-sector crop insurance programs had a good financial performance (Roberts and Dick 1991). Another important development relates to area-yield insurance that, according to some analysts, provides a promising approach to a workable crop insurance program (Dandekar 1976, Ahsan 1985, Miranda 1991, Williams et al 1993). Two decades ago, Roumasset (1978) argued that crop insurance as a public policy cannot be justified using welfare economic analysis or the results of empirical research. He said that practically no evidence showed that farmers are really risk-averse and that, even if they are, formal and informal risk-sharing institutions are already available to farmers. He further argued that the fact that private insurance companies do not insure crops indicates that the benefits of crop insurance could be much less than its costs. Some studies (Koropecky 1980, Ray 1985, Mishra 1996) strongly rebut the arguments put forward by Roumasset. We will not pursue this aspect further. Suffice it to say that farmers’ risk aversion and the need for policy intervention to enable farmers to tackle agricultural risks are 64

well established in the recent literature. The issues that are more relevant to public policy are benefits and costs and operational viability. Some studies conclude that the crop insurance experience in developed and developing countries has not been encouraging. Problems related to moral hazard and cost of administration abound. The welfare gains intended are often not achieved. It may be noted, however, that these studies do not suggest that crop insurance be given up altogether. Some prefer insurance for specific risks to all risks, some recommend area-yield insurance, and others advocate an increasing role of the private sector in providing crop insurance.

Conceptual and operational aspects
Insurance is based on the law of large numbers. It is basically a mechanism by which a large number of individuals pool risks. It also helps spread risk over space and time. However, insurability of risk depends on three factors: 1. It should be possible to quantify the probability of the event that triggers payment of indemnity. 2. There should be substantial independence in the occurrence of such events; otherwise, the problem of covariability will arise. 3. It should be feasible to attribute and evaluate the damage caused by the event. Problems arise because of asymmetric information resulting in moral hazard and adverse selection. Unlike the insured, the insurer does not possess adequate information and collection of information involves cost. The problem of moral hazard arises if the behavior of those insured can affect the occurrence of an adverse event or the amount of compensation. The problem of adverse selection arises if the insured knows his own risks, but the insurer does not. The latter will determine a premium rate on the basis of average experience. The premium rate thus determined may be higher than the fair premium for the low-income individuals, who may not participate in or opt out of the insurance. These problems are not unique to agricultural insurance; they exist in other forms of insurance also. However, these problems can

be more severe in the case of crop insurance. The cost and efforts required to gather information are much higher because of the vastness of the area and scattering of plots. It may be easier to introduce insurance for commercial crops to prosperous farmers. But the problem of insuring low-income rural households, who are more vulnerable to catastrophic income losses, has remained a challenge. To design a scheme of agricultural insurance, one has to look at the agricultural situation, socioeconomic factors, and administrative infrastructure. These aspects vary from country to country, even across various regions within a country. It is useful, however, to identify certain aspects in developing and operating agricultural insurance schemes. Many issues need to be addressed in the design of an operationally viable crop insurance program. From both conceptual and operational points of view, the following aspects are critical1 : 1. Degree of comprehensiveness: perils to be covered 2. Sector: public or private 3. Approach: individual or area 4. Participation: voluntary or compulsory 5. Target farmers: small and large, borrowers and nonborrowers 6. Coverage: all or some crops Other aspects-indemnity level, premium rate, determination of loss, reinsurance, marketing, and management-are relevant to the policy and operation of crop insurance.

Degree of comprehensiveness: perils to be covered
A fundamental issue in the design of a crop insurance scheme is whether to cover all risks or certain specified risks. The former implies yield insurance. In other words, an insured farmer is eligible to receive an indemnity if yield is below a certain guaranteed level. It is argued that, in

the case of yield insurance, it is difficult to identify losses arising out of uninsured events, such as farmer negligence or poor farming practices that are within the farmer's control. This may lead to moral hazard in the sense that the insured farmer will not take all possible steps to adopt improved farming practices to prevent or reduce damage to insured crops. In view of the above, an alternative approach envisages coverage against crop losses caused by specific perils, such as hail, windstorm, typhoon, and so on. Recent literature indicates that more emphasis is placed on such schemes. In Mauritius, windstorm was the only risk covered during the first 27 years of the scheme. In Cyprus, the risks covered are hail and drought. In the Windward Islands, the scheme offers coverage against wind damage only. The schemes being operated in Brazil, Canada, India, Japan, Philippines, Sri Lanka, and the United States are of the all-risk type. An FAO (1991) study classifies, on the basis of perils covered, crop insurance schemes into four categories: single peril, named peril, multiperil, and all-risk. The first category covers a single peril such as fire, hail, or windstorm. Namedperil schemes cover four specific perils and, in some cases, six if two or more perils are related (e.g., flood, excessive rains, and excessive humidity). Multiperil schemes are those that insure more than four perils. All-risk schemes are sometimes included in the multiperil category. According to the same FAO survey, the percentages of different types of schemes are as follows: single peril, 5%; named peril, 41%; multiperil, 52%; and all-risk, 2%. Thus, most of the schemes are in the named-peril and multiperil categories. Two aspects of crop insurance with specific perils need special mention. • A question may arise whether schemes covering specified risks provide much less economic benefit than the all-risk type. No empirical study analyzes the relative

This part is derived from a presentation made by the author as a resource person at a seminar sponsored by the Asian Productivity Organization on agricultural insurance held in Manila in July 1998.
1

65

economic benefits of schemes covering various degrees of risk. • At the operational level, it may not always be possible to attribute and measure the loss due to the insured perils. An insured crop is exposed to a variety of natural hazards and it is not easy to determine how much effect each hazard has on crop yield. Hence, one has to consider the agroclimatic situation to determine the degree of comprehensiveness or to identify the risks to be covered by a crop insurance scheme. A scheme based on named perils is feasible if the insured crops are affected by specific perils causing damage that is measurable. Even in such situations, one may start with limited risk coverage and, with experience, expand the range of perils and the crops to be covered by an insurance scheme. If a scheme envisages coverage of all risks, it is necessary to provide adequate safeguards to minimize the occurrence of moral hazard, which may not be easy.

Public or private sector
Historically, larger crop insurance schemes have been developed in the public sector. They are often multiple-risk or all-risk types. Some of these schemes are linked to agricultural credit. An insurance scheme in the public sector has the advantage of being able to access the government budget and obtain the cooperation of other public institutions and banks. The role of the government can take various forms: (I) the government bears fully or partly the cost of administration; (2) the government also shares a part of the indemnity or pays a part of the premium with a view to ensuring that farmers can afford to buy insurance; and (3) the government generally underwrites the loss over a specified level by providing capital funds (as in the U.S.), reinsurance (as in Japan and Canada), or through general budgetary resources in some countries.

There are disadvantages too, such as interference from the bureaucracy and the political executive, which could adversely affect a professional approach to the operation of the scheme. Private agricultural insurance2 has existed for many years in the form of hail insurance in Europe, the U.S., Canada, and Australia. In recent times, several private agencies have shown interest in crop insurance which covers cereals, pulses, forage crops, fruits, aquaculture, flowers, forest products, tree crops, and so on. Private-sector insurance has three characteristics: (1) coverage of specific risks, which are insurable, (2) an unsubsidized premium, and (3) voluntary insurance. Private-sector agricultural insurance is no longer confined to Europe, the U.S., Canada, and Australia. It has expanded to Latin America, Asia, and even Africa, though on a very small scale in most cases. This insurance caters mainly to specific risks of commercial and large farmers. In many developing countries, insurance institutions are not developed and are nationalized in many cases. Consequently, there is a dearth of diversified and experienced private insurers with experience in other forms of insurance. An insurance infrastructure that could introduce agricultural insurance in the private sector is lacking. For crop insurance, there is quite often a divergence between private profit and social benefit. Crop losses during natural disasters could be very large. Further, many farmers and areas could be affected during a natural catastrophe. For these reasons, it may not always be feasible to have private-sector insurance. For some types of agricultural insurance, private insurers with experience and expertise will be willing to participate. For other activities, particularly those addressing small farms and less developed agriculture, the role of the public sector will be important if there are larger

2

Agricultural insurance covers, in addition to crop insurance, insurance of forestry, sericulture, livestock,

poultry, and related activities.

66

economic benefits. An alternative is to have a parastatal body with autonomy and independence, which should operate on sound principles of insurance.

Individual or area approach
Assessment of indemnity is a critical element of any insurance scheme. Two main approaches are used to determine indemnity in crop insurance: the individual approach and the area approach. In the individual approach, loss is assessed separately for each insured farmer, for each plot or for the farm as a whole (consisting of more than one plot at different locations). In the area approach, indemnity is determined not separately for each farmer but for a group of farmers. The insured farmers are indemnified on the basis of the average loss experienced by a specified homogeneous area, which could be a district, a block, or even a village. Halcrow suggested this approach in 1949 while evaluating the U.S. Crop Insurance Program’s first decade of operation. Subsequently, Dandekar gave it a concrete shape in the Indian context (Dandekar 1976). In fact, Chakravarti (1920) advocated an area approach as early as 1920 in his drought insurance scheme for Mysore State. The area approach reduces the problem of moral hazard-action the insured affecting of indemnity-and can consequently allow the insurer to provide a larger range of risk coverage. Further, because the indemnity will be uniform for all insured farmers in an identified region, the problem of adverse selection in the sense that farmers facing below-average risk will tend to opt out of the insurance program may be less. This is because all farmers are indemnified. The administrative costs are much less because there is no need for field inspection and loss assessment with respect to individual plots of land. However, certain problems are inherent with this approach. Farmers may be less interested in buying insurance if individual farm yields are not adequately correlated with average area yield of the region. Further, it may be difficult to insure damage, which affects an area smaller than the specified area unit. Covariability can also be a problem. When the average

yield is less than the guaranteed yield, all the farmers will have to be paid indemnity simultaneously. In the context of the area approach, it is suggested that rainfall insurance could be introduced (Mishra 1995). Indemnity will depend on the average rainfall for a region. It may be easier to measure rainfall than average yield. Another advantage of rainfall insurance is that it would, unlike a single crop, reflect the income outcome of all the crops. The individual approach is more common among the existing crop insurance programs. It is more acceptable and appealing to farmers, but it involves a higher administration cost and is more prone to the problem of moral hazard. On the other hand, the area approach is easier and cheaper to operate, but it has many limitations. The choice of either the individual or area approach depends on the nature of the agricultural insurance program and the agroeconomic conditions: target farmers, farm size, crops insured, and even communication facilities. For example, the individual approach is more suitable for commercial and highyielding crops, whereas the area approach is easier to operate with small farms spread over large tracts.

Participation: voluntary or compulsory
Participation in a crop insurance scheme may be voluntary or compulsory. In the voluntary approach, participation is optional for a farmer who is eligible to be insured; examples are the schemes in Canada, the U.S., and Chile. In compulsory participation, certain categories of farmers who are eligible to be insured or who grow specified crops participate automatically. The word “compulsory” implies that there is a system of automatic insurance for a group of farmers. In Japan, crop insurance is compulsory for all farmers who are growing the insured crop over more than a minimum prescribed area. The Mauritius, Cyprus, and Windward Islands schemes are compulsory for all growers of certain crops. In India and the Philippines, crop insurance is compulsory for farmers who borrow from banks and other financial institutions. 67

The compulsory approach has two advantages. The problem of adverse selection is minimized significantly and the cost of sale of insurance is reduced. There may, however, be dissatisfaction among low-risk farmers who will have to cross-subsidize high-risk farmers. Under certain circumstances, it may be expedient to provide compulsory insurance for small farmers with a view to preventing adverse selection and high administrative cost, but it will be necessary to provide incentives to counter the moral hazard problem.

Target farmers
A report by the United Nations Conference on Trade and Development (cited in Goenka 1996) examines the role and viability of agricultural insurance in the context of the following categories of agricultural operation: 1. Traditional or subsistence 2. Semicommercial 3. Commercial 4. Specialized production systems This general classification may vary across countries. Further, there may be some overlap among the subsectors and it may be difficult to have an exclusive classification of farmers. Yet, the above categorization can help us to focus on some issues relating to the design of agricultural insurance schemes. The last category includes activities such as those found on horticultural farms, aquaculture farms, and poultry farms, and in orchards. This subsector adopts improved technology, has access to institutional financing, and has a commercial orientation. There is scope for expanding agricultural insurance here and so it is in this subsector that private insurance has already shown interest. The commercial subsector has medium and large farms, which are integrated with the market. They are commercially viable and risks are insurable. One can follow an individual approach. In this case also, there is tremendous scope for private-sector insurance. The semicommercial system consists of small- and medium-sized holdings, which are in transition from traditional to commercial

agriculture and offer opportunities for private insurers. However, there is scope for publicsector insurance to operate viably. This sector can pay a fair premium. Farmers with small holdings who usually employ family labor and produce primarily for self-consumption are in the traditional or subsistence sector. They are most vulnerable to agricultural risks and need insurance the most. However, the basic criteria of insurability may not be satisfied in the conventional sense. It is for this sector that insurance, though not financially viable, can yield substantial economic benefits in the sense that the extra value added is more than the cost of the scheme. In many developing countries, public-sector programs try to address this sector, which poses the greatest challenge. Private-sector insurance might not be feasible in this sector. Also not viable is a scheme based on the individual approach. Rainfall insurance could also become relevant in certain agroclimatic situations.

Coverage of crops
The basic objective of agricultural insurance is to stabilize farmers’ income. Logically, all crops grown by a farmer should therefore be covered by insurance. In the literature, there is a suggestion for combined crop insurance instead of a scheme based on individual crops. In practice, it is not feasible to cover all crops. Indeed, during the initial years, the scheme was limited to some crops; it expanded gradually to other crops depending on the experience and ability of the implementing agency. To maximize benefits, the selection of the crop or crops to be covered should depend on the importance of such crops to the economy and farmers. Some of the criteria are as follows: • The crop is important in providing food security to farmers. • It is necessary to increase production of the crop, keeping in view the strategy for economic development. • The crop is marketed and many farmers grow it so that risks are spread. • Time-series data on crop yields and crop losses are available.

68

As discussed earlier, crop insurance is more viable if coverage is for commercial crops. It is more difficult to insure crops that are used for self-consumption and there is no linkage with the market or with credit institutions.

Some empirical aspects
All-risk crop insurance schemes, in both in developed and developing countries, are often not financially viable and depend heavily on government subsidy. The subsidy ranges from 25% of indemnities in the U.S. to 50% and 80%, in Brazil and Mexico, respectively (Hazell et al 1986). More recent data reveal that the subsidy has grown to about 65% in the U.S. In Japan, the government has financed more than 90% of the cost of the crop insurance program (Hazell et al 1986). In the case of the Comprehensive Crop Insurance Scheme (CCIS) of India, government subsidy from 1985 to 1995 was 84% of the total cost of the scheme (Mishra 1996). On the other hand, crop insurance programs covering specific risks have been financially more viable. Mauritius has operated such a scheme successfully since 1946. The long experience with hail insurance was mentioned earlier. A question arises whether some economic benefits could justify the existence of crop insurance schemes, even for all-risk types. Some recent studies touched on this issue. The Agricultural Credit Policy Council of the Philippines evaluated the performance of the crop insurance program during the first-eight years since its introduction in 1981. The results indicate that insured farmers have higher incomes than uninsured farmers. Further, income fluctuations are lower for insured farmers than for uninsured ones (Faustino 1998). Mishra (1996) analyzed the impact of the CCIS on farm inputs, outputs, and income based on 1990-9 1 farm-level data from Gujarat, Orissa, and Tamil Nadu (150 households in Gujarat and 158 each in the other two states). Analysis of farm-level data from Gujarat indicated that an increase of Rs 806 per ha in annual net farm income—at 1990-91 prices— could be ascribed to the CCIS. The corresponding figures for Orissa and Tamil Nadu were Rs 520 and Rs 1,647, respectively. The

average cropping area insured per annum in India during 1985-95 was 8.02 million ha. Taking the lowest of the per hectare increases in income, Rs 520, and multiplying it by the area figure, one can estimate the annual benefit of the CCIS in terms of an increase in net farm income—Rs 4.17 billion for the country as a whole. This estimate needs to be viewed with some caution. It is based on farm-level data for 1990-91, thereby implying that the same level of benefits accrued over all years and in all areas. It is an overestimate, to the extent that the increase in net farm income was lower during other years, particularly in earlier years. Furthermore, the sample may not be representative of all areas of the country. Three other studies—based on data from south India—found that crop insurance had positive effects on crop inputs, output, and income. Mosley and Krishnamurthy (1995) used data for the 199 1-92 crop year from 280 farmers from two districts, Sanga Reddy and Mahboobnagar, of Andhra Pradesh. The analysis was based on comparisons of “with” and “without” situations. The finding was that insured farmers spent more, relative to the uninsured, on fertilizers, pesticides, and minor irrigation, and they obtained higher farm income. Regression analysis indicated that the CCIS is associated with an increase in net income of Rs 520 per farmer. The annual average number of insured farmers for the period studied by Mishra was 4.67 million, similar in magnitude to that studied by Mosley and Krishnamurthy. Assuming an increase of Rs 520 per farmer in net farm income, the benefit to 4.67 million farmers can be estimated at Rs 2.43 billion a year for India as a whole. The CCIS involved a total cost-indemnity and administration expenses—of Rs 12.58 billion during 1985-95, based on cumulative figures at current prices for different years. At 1990-91 prices, the cost would be Rs 13.49 billion. which implies an average annual cost of Rs 1.35 billion. Even if we take the estimate of Rs 2.43 billion as the annual benefit of the CCIS, it is significantly more than the government subsidy.

69

Santhi (1991) studied 90 farmers—covered by the CCIS—from the Agastheeswaram block of Kanyakumari District in Tamil Nadu. The analysis involved “before” and “after” comparisons, the latter represented by the 1986 and 1987 kharif seasons. The percentage of farmers who use high-yielding variety (HYV) seeds increased from 57% before the CCIS to 80% after the CCIS. The use of fertilizers N, P, and K increased in physical quantities. More labor was also used. Of course, in a “before” and “after” comparison, it is not possible to attribute all or part of this positive effect solely to crop insurance. Subramanian (1986), studied the Pilot Scheme of Crop Insurance, which operated from 1979-80 to 1984-85. Subramanian studied a sample of 90 farmers from Madurantakam taluk of Chengalpattu District in Tamil Nadu for the crop year 1984-85. Following a “with” and “without” approach with 60 insured and 30 uninsured farmers, he found that the insured farmers had more area under HYVs and used more certified seeds and more fertilizers. There was no significant difference in the use of plant protection chemicals. The insured farmers used more labor. These differences are in terms of both quantity and expenditure. The insured farmers had a significantly higher yield of rice. In the regression analysis, the crop insurance dummy variable was significant. If the benefit of the CCIS in net farm income is significantly more than the cost of the scheme, should not the entire cost be charged to the insured farmers through a higher premium? The average annual cost of the scheme during 1985-95 was Rs 1.35 billion versus a premium income of Rs 198.5 million, which is 1.7% of the sum insured, at constant 1990-91 prices. For the scheme to be self-financing, the premium should have been 11.4%; in other words, it needed to be raised to 6.7 times the present rate. Because of adverse selection, this would not be feasible, especially for a scheme based on an area approach. Here, all farmers neither face similar risk situations nor derive the same amount of benefit. A higher premium rate would lead to less demand for insurance if it were voluntary, and to resentment, on the part of farmers, if it were compulsory. 70

This does not mean that premium rates cannot be raised. The point is that the premium can be raised moderately. Nevertheless, in most situations, it would not meet the total cost of the scheme. Consequently, it is necessary to reexamine the design aspects of the scheme itself to ascertain whether there is scope to improve its financial performance and reduce its operational problems.

Concluding

remarks

This paper has attempted to analyze, with a policy perspective, certain basic issues relating to crop insurance. In the early part of the 20th century, crop insurance emerged as an important instrument of public policy to tackle the problem of agricultural risks. Later, some analysts questioned the relevance and efficacy of crop insurance—though there was never any unanimity—derived from theoretical formulations and empirical research. The scenario has changed, however, in recent times. The relevance of crop insurance as a policy instrument has gained acceptance because the problem of risk aversion by farmers is well recognized and some empirical studies show that crop insurance can yield economic and social benefits that outweigh the cost involved. This does not mean that the groundwork is made easier. Issues of practice and operation still remain. Problems of moral hazard and adverse selection arising on account of asymmetric information need to be solved. Operational issues and financial sustainability of crop insurance programs are to be addressed. This paper has analyzed some of these aspects and has suggested alternatives for designing crop insurance programs that could be more sustainable.

References
Ahsan SM. 1985. Agricultural insurance: a new policy for developing countries. Gower Aldershot, England. Chakravarti JS. 1920. Agricultural insurance: a practical scheme suited to Indian conditions. Government Press, Bangalore.

Dandekar VM. 1976. Crop insurance in India. Econ. Pol. Wkly. 11(26):A61-80. FAO (Food and Agriculture Organization). 1991. FAO crop insurance compendium 199 1. Rome. Faustino BP. 1998. The Philippine crop insurance corporation: its origin and operations. In: Agricultural insurance in Asia. Tokyo: Asian Productivity Organization. Goenka A. 1996. An overview of agricultural insurance in developing countries. Paper presented at the workshop on Agricultural Insurance in Africa, Mombassa, Kenya, 2526 November. Gudger M. 1991. Crop insurance: failure of the public sector and the rise of the private sector alternative. In: Holden D, Hazell P, Pritchard A, editors. Risk in agriculture: Proceedings of the Tenth Agriculture Sector Symposium. World Bank, Washington, D.C. Hazell PBR. 1992. The appropriate role of agricultural insurance in developing countries. J. Int. Dev. 4(6):567-581. Hazell P, Pomareda C, Valdes A, editors. 1986. Crop insurance for agricultural development: issues and experience. Baltimore, Md (USA): The Johns Hopkins University Press. Koropecky O. 1980. Risk sharing, attitudes and institutions in the rural sector: a critique of a case against crop insurance in developing countries. FCIC, Washington, D.C. Miranda MJ. 1991. Area-yield crop insurance reconsidered. Am. J. Agric. Econ. 73(2):233242. Mishra PK. 1995. Is rainfall insurance a new idea: pioneering scheme revisited. Econ. Pol. Wkly. 30(25):A84-88. Mishra PK. 1996. Agricultural risk, insurance and income: a study of the impact and design of India’s comprehensive crop insurance scheme. Vermont (USA): Ashgate Publishing Co. Mosley P, Krishnamurthy R. 1995. Can crop insurance work? The case of India. J. Dev. Stud. 31(3):428-450.

Ray PK. 1985. Economics of crop insurance with special reference to the needs and situations in developing countries. Calcutta: Central Publishing Concern. Roberts RAJ, Dick WJA, editors. 1991. Strategies for crop insurance planning. FAO Agricultural Services Bulletin 86. FAO, Rome. Roumasset JA. 1978. The case against crop insurance in developing countries. Philipp. Rev. Bus. Econ.l5(1):87-108. Santhi S. 1991. Comprehensive crop insurance scheme in Agastheeswaram Block of Kanyakumari District: an economic appraisal. MS dissertation, Tamil Nadu Agricultural University (Department of Agricultural Economics), Coimbatore. India. Subramanian P. 1986. Crop insurance for agricultural development: an economic appraisal in Madurantakam Taluk of Chengalpattu District. MS dissertation. Tamil Nadu Agricultural University (Department of Agricultural Economics), Coimbatore, India. Williams JR, Carriker GL, Barnaby A, Harper JK. 1993. Crop insurance and disaster assistance designs for wheat and grain sorghum. Am. J. Agric. Econ. 75(2):435447. Wright BD, Hewitt JA. 1993. Crop insurance for developing countries. In: Berck P, Bigman D, editors. Food security and food inventories in developing countries. Wallingford (Oxon): CAB International.

Notes
Author’s address: Gujarat Electricity Board, Baroda, Government of Gujarat, India. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems. 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

71

The nature and causes of changes in variability of rice production in eastern India: a district-level analysis
S. Pandey and S. Pal

The likely effect of the adoption of modern varieties and production practices on variability of production has been widely debated. This paper examines the nature of changes in variability of rice production in eastern India between 1969-81 and 1982-94. Aggregate data on production, yield, and area from 71 rice-producing districts of eastern India are analyzed using a variance decomposition method. The variance of yield and production was found to have increased, with increasing interdistrict correlation as a major contributing factor. Production variability, measured by the coefficient of variation, however, showed a declining trend in several states. Changes in the coefficient of variation of yield were explained mainly by the changes in the coefficient of variation of rainfall. Various factors that may have altered the variability of rice production in eastern India are discussed. In contrary to the conventional belief, modern varieties can be less risky than the traditional varieties. The coefficient of variation of production decreased in districts where rice yield increased rapidly, indicating that improvements in rice yields can be achieved without increasing instability.

It is now widely accepted that the spread of improved technology consisting of high-yielding varieties and associated crop management practices has been the main source of growth in food- grain production in Asia over the past two decades. What is still debatable is the effect on the variability of production. If growth arising from the adoption of improved technology has increased the instability of production around the trend, appropriate policies to reduce-food shortages when production dips below the trend may be needed. In addition, agricultural research priorities may need to be altered to emphasize those components of technology that have stabilizing effects so that any trade-off between growth and instability may be reduced. The possibility of increased production instability with an increase in yield was first raised in the late 1960s in the context of foodgrain production in India (Sen 1967, Rao 1968). Sen ( 1967) suggested that production variability is likely to increase if output growth is achieved through an expansion of cultivation to marginal areas. In addition, he hypothesized that production instability will tend to rise with an

increased use of purchased inputs such as fertilizers. As the price elasticity of supply increases with an increase in the share of purchased inputs, the effect on output of a given level of instability in prices will be amplified. Since yield tends to be a more unstable component of output than area, Rao (1968) suggested that production variability is likely to increase with the increasing importance of yield growth as the source of output growth. Several other researchers have associated the increase in variability of food-grain production in India with the spread of modem varieties (Mehra 1981, Barker et al 1981, Griffin 1988). Based on a variance decomposition method, Hazell (1982) found that the variance of cerealgrain production in India has increased over time. However, the increase in variance was found to be not due to the adoption of highyielding varieties (HYVs) per se but due to the instability in input supplies. In a subsequent analysis, Hazell (1984) suggested that increased interregional correlation of output due to similarity of varieties grown may have contributed to increased production variance.

73

Walker (1989) found the adoption of HYVs of sorghum and pearl millet to be a major contributing factor to increased production variability of these crops. Anderson and Hazell (1989) reported that the variability of world food-grain production, measured by the coefficient of variation (CV), increased between the 1960s and the 1970s, although the variability for some crops and some regions decreased. Using district-level data from India, Singh and Byerlee (1990) found that the CV of wheat yield decreased over time, mainly because of the expansion of irrigation. Based on this result, they hypothesized that the increase in variability reported in earlier studies might partly reflect the disequilibrium that characterizes the initial periods when new technologies are introduced. Such instability tends to disappear over time as farmers learn more about the technologies and adjust their production systems. In a more recent study, Naylor et a1 (1997) reported that the production variability of maize has increased over the past three decades. Their results also indicate that the production variability of wheat and rice declined initially but has shown a tendency to increase in the more recent period (1980-94). They attribute this tendency toward an increase in instability for rice and wheat to the near full exploitation of the potential of the Green Revolution technology and a tapering off of investment in irrigation. The diversity of results on the pattern of changes in production variability implies that the factors that govern production instability vary across crops, locations, and time periods. Hazell (1982) conducted a major study of changes in production variability of rice for India using state-level data up until 1977. The diffusion of Green Revolution technology for rice in irrigated areas such as Punjab, Haryana, and Tamil Nadu was well advanced by 1977. However, the productivity growth in eastern India where more than 17 million ha of rice are grown under rainfed conditions, started only in the early 1980s. The impact of productivity growth on the variability of rice production in eastern India in recent years has not been adequately analyzed. Eastem India is characterized by a high degree of dependence on

rice production, a high level of poverty, and a high population pressure. An improvement in food security in eastern India critically depends on the stability of rice production. The aim of this paper is to analyze the changes in production variability in eastern India during 1969-94 using both state- and district-level data.

Method and data
Eastern India comprises eastern Uttar Pradesh, eastern Madhya Pradesh, Bihar, Orissa, West Bengal, and Assam. In the first part of this paper, in which growth patterns are examined, data on rice yield, area, and production for the period 1961-97 are included. As there was a clear change in the productivity trend after 1981, data were divided into two segments: before 198 I and after 198 1. Growth rates for each period were estimated using a semilogarithmic trend. Changes in instability of rice production, area, and yield were examined by comparing changes in variance and the coefficient of variation. The second part of the analysis, which involves a decomposition of the change in variance, covers from 1969 to 1994. The choice of this period is based on the availability of district-level data. Assam was excluded from the analysis as district-level data for this state were not available at the time of writing this paper. Seventy-one districts from the remaining five states of eastern India were included in the analysis. For comparative purposes, the analysis has been conducted also for northern India (Punjab and Haryana) and southern India (Andhra Pradesh, Kamataka, and Tamil Nadu), which are now believed to be in the post-Green Revolution phase (Byerlee 1995). Broadly, eastern India represents the rainfed environment, whereas northern and southern India represent irrigated environments. Hazell’s (1982) variance decomposition analysis basically consists of decomposing the change in production variance between the two periods into 10 components (Appendix 1). The first two components capture the effect of changes in mean yield and mean area, respectively. The third and fourth components measure the effect of changes in yield and area

74

variances, respectively. The sixth component measures the contribution of change in areayield covariance. The fifth and remaining three components measure the effect of various interaction terms. The final term is the residual that captures the effect of higher order interactions. The changes in mean production can similarly be decomposed into the change in mean yield, change in mean area, change in area-yield covariance, and an interaction term. To apply Hazell's variance decomposition method, the total time span covered by the data was divided into the first (1969-81) and second (1982-94) periods, with each period having 13 years. Data on rice yield and area for each district were detrended separately for each period using linear trends. For each variable, the residuals of the trend equation were added to the mean of the data series to generate detrended data centered on the mean. The analysis thus implicitly assumes that expectations regarding the output remain constant over time at the level denoted by the mean (Bindlish 1988). The detrended production was obtained by multiplying detrended yield and detrended area. The final analyses were conducted using these detrended data.

Eastern India is a large rice-growing area with a high degree of variation in agroclimatic conditions within and across states. The effects of broader variation in agroclimatic conditions were partially accounted for by using agroclimatic zones rather than the state as the basic unit for analysis. Using the agroclimatic zonal classification developed by the Planning Commission of India, the five states of eastem India can be grouped into 10 agroclimatic zones (Planning Commission 1989). Table 1 lists the relative importance of these 10 zones in terms of rice area and production in eastern India as a whole and the major agroclimatic features of each zone. The geographical coverage of these zones is shown in Figure 1. Because the last three regions have small shares in area and output, they were excluded from the analysis. In the third part of this paper, the factors that explain the changes in rice production variability between the two time periods are analyzed. In addition, attempts are also made to explain the differences in variability across districts. For the state-level analysis, all districts in a state were included, except for eastern Uttar Pradesh and eastern Madhya Pradesh. Districts

Table 1. Major characteristics and percentage share of eastern Indian rice area and production by agroclimatic zone (triennium 1992-94). Zone Middle Gangetic Plains Northwest Alluvium Northeast Alluvium Lower Gangetic Plains Eastern Plateau and hills Eastern Plains Eastern Highlands Plateau Eastern Coastal Plains Central Plateau and hills Eastern Himalayas Upper Gangetic Plains Eastern India
Source: Planning Commission (1989).

% share Area 11 14 22 16 8 14 8 2 3 2 100 Production 13 11 32 15 6 10 8 1 2 2 100 Rainfall (mm) 1,211 1,470 1,587 1,271 1,436 1,369 1,287 1,130

Main feature Climate Moist to dry subhumid Dry to moist subhumid Moist and dry subhumid Dry subhumid Moist to dry subhumid Moist to dry subhumid Moist subhumid Dry subhumid Soil type Alluvial and calcareous Alluvial plains Red and yellow alluvial Medium to deep black, red and yellow Red sandy, red and yellow Red and yellow, and red loamy Alluvial laterite and red loamy Medium black

75

belonging to other geographical regions of these two states were excluded. Appendix 2 lists the districts included from each state. For the analysis by agroclimatic zones, various districts belonging to an agroclimatic zone were grouped together. As there are different statistical descriptors of variability, a short digression on the appropriate measure in the context of the present study may be useful. Variance (or standard deviation) is a commonly used indicator of dispersion around the mean. However, it is not a dimensionless measure and estimates of variance between the two periods may change merely because the level of production has also changed. For comparative purposes, we need a measure that represents variability relative to the magnitude of variation. The coefficient of variation (CV) is such a measure. The decomposition of the coefficient of variation into its constituting sources, however, is mathematically cumbersome. In addition, such a 76

decomposition of the CV does not provide any information additional to what the decomposition of variance provides. Accordingly, the comparison of instability over time in this study is based mainly on CV, whereas the analysis of sources of variation is based on the decomposition of variance. Both the variance and CV ignore information about the probability distribution contained in moments higher than the second moment. For this reason, these summary descriptors of variability are inadequate when judgments about “riskiness” of alternative technologies are needed. In the absence of a clearly defined utility function, stochastic dominance analysis (Anderson et al 1977) offers a convenient way of ranking technologies based on the concept of risk efficiency. The stochastic dominance analysis is used in this paper to provide information on the relative riskiness of traditional and modern varieties.

Patterns of growth in rice production in eastern India
Eastern India is a major rice-growing area of India that accounts for more than two-thirds of the rice area and more than half of rice production. Figure 2 shows the trends in area, production, and yield of rice in the period 1961 97. The trends in yield and production show a clear break during the early 1980s. Production growth during 1961-81 was driven primarily by the expansion in area as yield growth was almost negligible. Rice yield increased dramatically after 1981 as the spread of modem varieties in eastern India increased. Table 2 shows the growth rates in area, production, and yield during 1961-81 and 1982-97 for eastern India. The growth rate in yield, which was below 1% during 1961-81, increased to 3.1% during 198297. This increase in productivity has been the main driving force behind the growth in rice output in eastern India.

Of the five states of eastern India analyzed, Uttar Pradesh and West Bengal are the two that have shown a strong growth in rice production since the early 1980s. Output growth in these states during 1982-97 exceeded 4% per year. Growth in yield contributed more than 75% of the growth in output in these states. Other states of eastern India also recorded a higher rate of growth during 1982-97 relative to 1961-81. Even Bihar and Orissa, which had a relatively low productivity, recorded a yield growth of more than 2% per annum during 1982-97. Table 3 summarizes the major characteristics of rice production systems for these five states. Overall, Uttar Pradesh and West Bengal are the leading states in rice productivity, fertilizer use, and adoption of modem varieties. The expansion of tubewell irrigation, rapid adoption of modem HYVs, and expansion of summer (boro) rice cultivation are the major reasons for the increase in productivity in these states. Other states are also experiencing a similar change but at a slower pace.

) in eastern India.

77

Table 2. Growth rates (% y-1) of area, yield, and production of rice in various parts of eastern India. State Assam Area Yield Production Bihar Area Yield Production Madhya Pradesh Area Yield Production Orissa Area Yield Production Uttar Pradesh Area Yield Production West Bengal Area Yield Production Eastern India Area Yield Production
a

1961-81a 1.08** 0.59** 1.67** 0.26 0.79 1.06 0.79** 0.40 1.19 0.03 -0.06 -0.04 1.02** 2.02* 3.04** 0.81** 0.91** 1.73** 0.63** 0.79 1.42**

1982-97a 0.78' 1.81" 2.59" -0.18 2.83* 2.65* 0.61** 1.31 1.91* 0.52** 2.52* 3.04* 0.43 3.93** 4.37** 1.11** 3.81** 4.92** 0.53** 3.07** 3.60**

**, * = significant at 5% and 1% level, respectively.

Analysis of instability
Estimates of changes in variances of rice yield, area, and production for eastern, northern, and southern India are presented in Table 4. The production variance in eastern India increased by 60% between the two periods. A feature of the growth in rice production in eastern India has been the increase in area variance by 157%. This increase in area variance is statistically significant at the 5% level. In both northern and

southern India, the variance of area sown remained almost constant between the two periods. The change in the variance of yield in eastern India is also positive but not statistically significant. A smaller percentage increase in the variance of production than in the change in variance of area is indicative of some compensatory movements in area and yield in eastern India. Although the changes in area and yield variances in northern India are negative and statistically insignificant, the production variance increased by 231%. This indicates the existence of a covariate movement in area and yield in northern India. For southem India, the production variance changed very little. Within eastern India, the variance of area, yield, and production increased significantly in Bihar. In eastern Uttar Pradesh, the variance of area increased but that of yield and production remained unchanged. In all other states, the variances of yield, area, and production remained unchanged between the two periods. Although variance of production increased in eastern India, the relative variability as measured by the coefficient of variation decreased slightly (Table 5). The decline in the CV of production is much larger in northern and southern India. However, these reductions in CVs are not statistically significant. Overall, the absolute variability in production increased, but the relative variability has remained more or less constant between the two periods in all three regions of India. A statistically significant decline in the CV of yield is observed in eastern Uttar Pradesh only (Table 6). The relative variability of area, yield, and production remained unchanged in Orissa, Madhya Pradesh,

Table 3. Features of rice production in eastern India. State Uttar Pradeshf Madhya Pradeshf Bihar Orissa West Bengal Average rice yield (t ha-1)a 3.18 1.52 2.12 1.82 3.33 % rice area irrigatedb 62.3 23.7 40.2 35.5 27.2 % highyielding variety areac 88 63 60 66 74 Fertilizer use (kg ha-1)d 107.6 39.4 79.6 25.8 102.8 Cropping intensitye 149 125 137 134 159

aRice yield is the average of 1996-97 and 1997-98 yield. Yields are expressed year 1995-96 except for Orissa (1993-94) and West Bengal (1988-89). Source: hectare of gross cropped area for the year 1996-97. Source: FAI (1999). eFor defined as the ratio of gross cropped area to net sown area. fData refer to the

in this paper in terms of rough rice. Source: DES (1999). bFor the FAI (1999). cFor 1995-96. Source: FA1 (1999). dTotal of NPK per 1994-95. Source Fertilizer Statistics 1997-98. Cropping intensity is whole state, not just its eastern region.

78

Table 4. Estimates of changes in variance between 196981 and 1982-94. % change in variance relative to the first period Eastern India 156.7 11.4 60.4
a

Item
Area Meld Production
a

Northern India Southern India -29.0 -50.7 a 231.3 19.7 -38.1 -9.0

Statistically significant at the 5% level.

and West Bengal. In Bihar, the CV of production increased mainly due to an increase in the CV of area. When state-level data are classified by agroclimatic zone, the patterns for the Northwest Alluvium (the Uttar Pradesh Plains) and the Northeast Alluvium (the Bihar Plains) were the same as those of the corresponding state-level analysis. Of the seven zones, yield variance increased significantly in the Northeast Alluvium and the Eastern Coastal Plains. For the Eastern Highlands, the yield variance declined significantly. For all other zones, the change in yield variance was not statistically significant. The area variance increased significantly in the Northwest Alluvium and the Northeast Alluvium but declined significantly in the Eastem Highlands. There was no significant change in the area variance in other zones. The production variance increased significantly only in the Northeast Alluvium and the Eastern Coastal Plains. In terms of CV, major changes in production instability occurred in the Northwest Alluvium (-33%), the Northeast Alluvium (+93%), and the Coastal Plains (+68%). The CV of production changed very little in other major

agroclimatic zones (Table 7). A comparison of CV of production for the second period across the agroclimatic zones indicates that the variability is highest in the Northeast Alluvium and in the Eastern Coastal Plains. High instability of yield is the major cause for high CV of production in these two zones. For the Northeast Alluvium, the area variability is also higher than in the other zones. The Lower Gangetic Plains had the lowest CV of production. Surprisingly, the CV of production in the Northwest Alluvium was similar to that in the Plateau, despite the larger irrigation facility in the former. The Northwest Alluvium is also characterized by high instability in area. As the comparison of variability between two time periods can be sensitive to the cutoff point between the periods, a 10-year moving CV of yield, area, and output was estimated to examine the trend in CV. When calculating the moving CV, the first 10-year period was for 1969-78. The data for this period were linearly detrended and the CV was calculated using the detrended data. For the next period, the data for 1969 were dropped and those for 1979 were included. The CV was calculated again using the detrended data. This process was repeated until all data were used. The results indicate that the CV of yield started to decline for most states after the period 1975-84 (Fig. 3A). Eastern Uttar Pradesh had the sharpest decline in yield instability. Yield variability declined in Madhya Pradesh and West Bengal also, although the CV of yield remained almost twice that in West Bengal for the recent periods. Bihar and Orissa are the two states where instability either increased or remained

Table 5. Estimates of coefficients of variation of area, yield, and production by region. a Item Area Yield Production Eastern India 1969-81 1.8 10.0 11.2 1982-94 2.8 (53.6) 7.4 (-26.1) 9.5 (-15.0) Northern India 1969-81 10.7 8.8 12.0 1982-94 4.1 (-62.3) b 4.6 (-48.0) 7.0 (-39.8)
b

Southern India 1969-81 3.7 7.4 7.9 1982-94 4.1 (11.6) 4.2 (-43.4) 5.5 (-30.0)

a Numbers in parentheses measure the percentage change in CV in the second period relative to the first period. 5% level. The statistical test is based on t-statistic as suggested by Anderson and Hazell (1989).

Change in CV significant at the

79

Table 6. Coefficients of variation in area, yield, and production by state. State 1969-81 Eastern Uttar Pradesh Eastern Madhya Pradesh Bihar Orissa West Bengal
a

Area 1982-94 6.6 0.9 7.4 2.8 2.5 (61)b (42) (77)b (-21) (-26) 1969-81 19.2 20.1 10.9 12.5 8.0
b

Yield 1982-94 6.6 (-66)b 11.1 (-44) 15.9 (46) 12.6 (1) 6.8 (-14) 1969-81 20.4 20.3 12.1 13.8 10.0

Production 1982-94 11.7 (-42) 11.6 (-43) 20.4 (68)b 14.7 (6) 8.2 (-18)

4.1 0.7 4.2 3.5 3.4

Numbers in parentheses indicate percentage change relative to the first period.

Statistically significant at the 5% level.

Table 7. Coefficients of variation in area, yield, and production by agroclimatic zone.a Agroclimatic zone Area 1969-81 1982-94 1969-81 Yield 1982-94 Production 1969-81 1982-94

Middle Gangetic Plains Northwest Alluvium (Uttar Pradesh) Northeast Alluvium (Bihar) Lower Gangetic Plains (West Bengal) Eastern Plateau and hills Eastern Plains (Orissa + Madhya Pradesh) Eastern Highlands (Orissa + Madhya Pradesh) Plateau Eastern Coastal Plains Eastern India
a

3.9 4.6 3.2 1.7 2.9 2.9 3.3 1.8

7.4 (88)b 8.6 (90)b 2.5 (-21) 1.5 1.0 2.8 3.7 2.8 (-1 3) (-68)b (-6) (12) (53.6)
b

18.6 11.6 8.2 18.9 18.5 13.6 10.6 10.0

7.2 (-61)b* 16.0 (38) 6.7 (-16) 10.1 (-46) 9.3 (-50) 12.8 (-6) 16.1 (52) 7.4 (26.1)

19.6 11.1 10.1 19.2 19.4 16.0 11.1 11.2

13.1 21.5 8.1 11.4 9.6 14.7 18.6 9.5

(-33) (93)b (-20) (-41) (-50) (-8) (68)b (-15)

Numbers in parentheses indicate percentage change relative to the first period.

Statistically significant at the 5% level.

constant except in the more recent periods. The result for the moving CV of area (Fig. 3B) indicates a clear pattern of increase in area instability in Uttar Pradesh. Area instability in Bihar has remained high throughout the period, although it seems to have increased slightly in recent years. The temporal pattern in production instability follows that of yield instability. Production instability has shown a declining trend in all states except Bihar (Fig. 3C). The moving CVs for different agroclimatic zones also confirm the overall pattern of declining instability in yield over time for all other zones. The moving CV of production mirrors that of yield. The CV of area displays patterns similar to that in the state-level analysis. The Northeast Alluvium zone and the Coastal Plains of Orissa are the only two zones where production instability has increased. Based on these results, we can conclude that the proportionate growth in mean production in most parts of eastern India has more than compensated for the increase in the standard deviation of production, resulting in a decline in its CV. Overall, the output growth in eastern Indian rice production appears to have been 80

achieved at a reduced relative variability. even though the absolute variability has increased in most cases.

Variance decomposition analysis
In eastern India, the change in the variance of yield and the change in the covariance between area and yield account for a major share (i.e., 60%) of the increase in variance between the two periods (Table 8). In contrast, the major share of change in production variance is contributed by the change in mean area in northern India. Change in yield variance had a slightly stabilizing effect in northern India due mainly to the availability of assured irrigation. Increased covariance between area and yield also contributed to a quarter of the increased production variance. For southern India, the change in variance of area had the most stabilizing effect on production variance. The change in the variance of yield had a destabilizing effect. There was also evidence of the stabilizing effect of the change in covariance between yield and area.

81

Table 8. Decomposition of change in production variance by region. Description Eastern India 16.3 17.1 31.4 5.5 28.4 1.3 60.4 Northern India 23.0 74.0 -11.9 -10.1 28.7 -3.8 231 Southern India 37.8 -206.1 -46.6 372.8 52.5 -110.3 -9

Change in mean yield (%) Change in mean area (%) Change in yield variance (%) Change in area variance (%) Change in area-yield covariance (%) Interactions and residual (%) Total change in variance (%)

A change in yield variance affects the change in production variance directly (the third component in Table 8) as well as through a change in the covariance between yield and area (the fifth component in Table 8). If the variance of yield had not changed between the two time periods, the increase in variance of output would have been lower as the third term would be reduced to zero. Furthermore, as covariance is a function of yield variance, area variance, and area-yield correlation, the covariance term would also be affected even if the yield variance did not change. Assuming that the change in yield correlation and the change in area variance were such that the change in area-yield covariance became equal to zero,2 the increase in production variance of eastern India would have been 24% only. It is calculated as 0.604 x (1 .314-.284). This would have led to a reduction

in the CV of production3 by 25%. However, if the change in area-yield covariance remained at its current level, the reduction in CV of production would have been 20%. Thus, if the variance of yield had not changed, the CV of production would have been lower relative to that in the first-period values by 20-25%. The actual reduction in CV was 15% (Table 5). Thus, the effect of an increase in the yield variance on the change in CV of production has been quite small. The decomposition of change in production variance by state for eastern India shows that the change in variance of production within the states contributed to about one-half of the increase in the total production variance of eastern India (Table 9). The other half is contributed by the changes in the interstate production covariances. Thus, even if the production variances within states had remained unchanged, the production variance in eastern India would have increased by as much as 30% (i.e., 0.5 x 60%) as a result of the increased production covariance across states. What are the sources of changes in covariances between yield and area across states? The covariance may change as a result of a change in correlation in area, correlation in yield, or changes in area-yield correlation. It may also change if the area or yield variances

Table 9. Decomposition of change in production variance by state. Total Description Uttar Pradesh 35.9 58.9 -96.6 18.4 25.7 57.7 Madhya Pradesh -3.8 -77.7 154.9 -0.5 -1.9 26.4 variance Bihar Orissa 9.4 -1.5 68.9 -2.5 25.6 0.1 West Bengal 33.0 15.5 54.0 -5.2 4.2 -1.3 Interstate covariance 10.1 11.9 9.0 3.3 12.9 3.2 50.4 All eastern India 16.3 17.1 31.4 5.5 28.4 1.4

Change in mean yield (%) Change in mean area (%) Change in yield variance (%) Change in area variance (%) Change in area-yield covariance (%) Interactions and residual (%) Total contribution to the change in variance of production in eastern India (%)

2.6 -3.1 62.3 7.7 38.2 -7.7

3.1

-3.8

30.4

10.2

9.7

100

2

Calculated using the relationship is the correlation coefficient between yield and area, is the standard deviation of yield, is the standard

where

deviation of area, and A is the change from period 1 and subscript 1 denotes the value for period 1. The conditions above imply that and This relationship is obtained from Hazell (1982).
3

It can be shown that the proportionate change in CV is equal to

where

and

are the proportionate

changes in variance and mean, respectively.

82

have changed. The examination of the Correlation matrix showed that the percentage of negative changes in area-yield correlation (within and among states) was 44%. For yield correlation across states, the percentage of negative changes was 60%. The negative changes in area correlation occurred in only 30% of the cases. Overall, it appears that negative changes in yield correlation may have helped dampen the increase in production variance in eastern India. The change in production variance in Bihar alone contributed to almost one-third of the changes in production variance in eastern India. This is the single largest source of change in production variance in eastern India. Changes in production variances in Orissa and West Bengal have almost equal shares (10%) in the change in production variance of eastern India. The examination of variance decomposition within Bihar, Orissa, and West Bengal indicates that the change in the variance of yield accounts for the major share of change in the within-state production variance. For example, in Bihar and Orissa, it accounts for 62% and 69% of the change in production variances, respectively. If the contribution of the change in within-state covariance between yield and area is also taken into account, this share increases to more than 90%. Uttar Pradesh and Madhya Pradesh are the only states where the changes in yield variance have helped reduce the change production variance. The change in covariance has a destabilizing effect in all states. For each state, the increase in production variance may be the result of increased variance within districts and/or increased covariance across districts. Variance decomposition analysis using district-level data for each state indicates that the change in yield variance is the major source of changes in production variance in most districts (the results of decomposition analysis by district are available from the authors upon request). Table 10 presents the percentages of districts in a state recording negative changes in

Table 10. Percentage of districts in each state recording negative changes in correlations between the two time periods. State Uttar Pradesh Madhya Pradesh Bihar Orissa West Bengal Yield correlation 81 15 13 42 56 Area correlation 69 53 23 41 61 Area and yield correlation 49 49 24 28 48

yield correlation, area correlation, and area-yield correlation between the two periods. The effect is more pronounced in eastem Uttar Pradesh where 81% of the changes in yield correlation were negative. In contrast, Bihar had negative changes in yield correlation among 13% of districts only. Thus, relative to Bihar, districtlevel rice yields in eastern Uttar Pradesh have shown a greater degree of compensatory changes resulting in increased stability in rice production of the state. The patterns are similar for changes in area correlation and area-yield correlation also. Overall, West Bengal and eastern Uttar Pradesh are the two states where agricultural growth is also characterized by changes in correlation that have a dampening effect on production instability. On the other hand, in Bihar and Orissa, such changes in correlation may have reinforced the underlying trend toward increased instability of production. In Bihar, almost all of the increase in production variance is due to the increase in yield variance and the increase in area-yield covariance. Thus, if these two components had not changed, the production variance of Bihar would not have increased much. Even if we assume that the yield variance of Bihar had remained constant with the area variance and the area-yield correlation taking their respective values in the second period, the production variance of Bihar would have increased" by only 58% as opposed to the actual increase of 278%. Thus, the increase in yield variance seems to be the major source of instability in rice production in Bihar.

4

Using the relationship in footnote 2, D cov (A,y) = (A.y) was calculated using this relationship.

when

= 0. The new value of D CoV

83

The variance decomposition analysis by agroclimatic zone indicated that the change in yield variance is the major source of change in production variance within each zone. Changes in the yield variance had stabilizing effects only in the Northwest Alluvium, Eastern Plains, and Eastern Highlands. Of the seven main zones, the Northwest Alluvium, Northeast Alluvium, Lower Gangetic Plains, and Eastem Coastal Plains accounted for almost all the changes in the production variance of eastern India. The Northeast Alluvium zone had the highest share (46%) of the increase in production variance in eastern India. About half of this share was due to the change in production variance within the zone.

Reasons for changes in production stability
Several factors determine variability in rice production and patterns of change over time. These relate to the characteristics of production systems and the nature of changes occurring in these systems. Such changes include expansion of irrigation, increased adoption of modern varieties and crop management practices, changes in the variability of rainfall, changes in the variability of price and production responses to price changes, and shifts in rice-growing environments. The effects of some of these changes on the variability of production and yield of rice in eastern India are examined below.

Under irrigated conditions where irrigation is not assured, production variability can actually be higher than in rainfed conditions. This is due mainly to the possibility of an increase in area variability. When irrigation is based on run-of-the-river schemes, small tanks, and other local storage systems in which the water supply is highly covariate with the local rainfall, area instability tends to be higher than when the water supply is based on schemes backed up by large dams or an assured supply of groundwater. With a less assured supply of water at planting time, farmers adjust the area up or down, depending on the water supply. This results in a considerable fluctuation in area and is the main reason for the high area variability in tank-irrigated areas of southern India relative to that in Punjab where irrigation is based on an assured supply of groundwater. For eastern India, rice is irrigated from both assured and less assured sources of irrigation. The recent expansion of tubewell irrigation (Ballabh and Pandey 1999) in the Gangetic Plains has been a major stabilizing force in rice production, especially in eastern Uttar Pradesh and West Bengal. In West Bengal, the gross cropped area irrigated increased by 122% between 1984 and 1996. The coefficient of variation of production declined drastically in these two states as a result. The stabilizing effect has been so strong that even the variance of yield declined in some districts of these states.

Variability of rainfall
Changes in the variability of rainfall between the two time periods may also result in changes in production variability, even if sensitivity of output with respect to rainfall is constant. This could be part of the explanation for the observed differences in changes in production variability across states. Rainfall variability as measured by the CV of rainfall for the two time periods analyzed increased in the Northeast Alluvium (Bihar Plains) but decreased in eastern Uttar Pradesh and West Bengal. The changes in production variability in these states mirror the changes in rainfall variability. Thus, even without any change in the long-term variability

Nature of irrigation
The nature and extent of irrigation are important determinants of production variability (Pandey 1989). Good control of field hydrology (through both irrigation and drainage) has been an important factor in stabilizing production in irrigated areas such as Punjab. This is also why yield instability tends to be lower for summer rice than for rainy-season rice-the former is grown in fields with an adequate and assured water supply. Area variability also tends to be lower when the water supply is assured. Both these factors reduce overall production instability. 84

of rainfall, the analysis based on different time segments may indicate differences in variability due to different stochastic patterns of rainfall. Could the changes in production variability between the two periods be driven mainly by the cyclical effects of rainfall? To explore this possibility, the CV of yield was regressed on the CV of rainfall. Since there are only two estimates of CV of yield for each state (i.e., one estimate for the first period and the other for the second period), the data for all states were pooled. No state-level fixed effects were specified due to the limited degrees of freedom available for hypothesis testing. The estimated equation is as follows: CVy = -0.57 + 0.80 CVR - 0.9 D (3.2) (-0.4) R2 = 0.5, n = 10.

Table 11. Parameter estimates for regression of the 10year moving coefficient of variation of yield on the IO-year moving coefficient of variation of rainfall. State/zone Eastern Uttar Pradesh Eastern Madhya Pradesh Bihar Plateau Bihar Plains Orissa West Bengal Constant -1.28 1.06 -10.70 1.26 15.21 0.36 Moving CV of rainfalla 0.68 ** (9.19) 0.81 ** (4.38) 1.40 * (2.95) 0.60 (1.61) -0.08 (-0.30) 0.53 (2.96)** R2 0.84 0.53 0.32 0.09 0.01 0.33

*** , * = statistically significant at 1% and 5% level, respectively. Values
in parentheses are 1-ratios.

where CVy, CVR, and D denote the CV of yield, the CV of rainfall, and a dummy variable for the second period (D = 1 for the second period), respectively. The numbers in parentheses are the t-ratios. The results indicate that the CV of rainfall has a positive and significant effect on the CV of yield. A 1 percentage point increase in the CV of rainfall leads to a 0.8 percentage point increase in the CV of yield in the eastern Indian states. Another equation (in which the CV of area was regressed on the CV of rainfall) produced a statistically insignificant coefficient, indicating that the observed area variability is independent of rainfall variability. The response of yield variability to-rainfall variability estimated above is the average value for both time periods and for all states of eastern India. It is interesting to note that the variability of rainfall explains half of the variation in the CV of yield over time and across states. This is despite the tremendous progress in the expansion of irrigation in eastern India. The variability of yield in eastern India thus appears to be still determined mostly by the variability of rainfall.

The sensitivity of yield to rainfall is expected to be high in states with lower proportions of area irrigated. So the CV of yield is likely to be affected more by the CV of rainfall in Orissa, Bihar, and eastern Madhya Pradesh than in eastern Uttar Pradesh and West Bengal. To examine this possibility, a 10-year moving CV of yield of each state was regressed on the corresponding IO-year moving CV of rainfall. The results (Table 11) indicate that the CV of yield is positively influenced by the CV of rainfall for West Bengal, eastern Uttar Pradesh, eastern Madhya Pradesh, and the Bihar Plateau.5 For Orissa and the Bihar Plains, the relationship was statistically insignificant. The size of the coefficient associated with the CV of rainfall is highest for Bihar plateau and lowest for West Bengal, supporting the hypothesis that the effect of rainfall Variability is likely to be greater in regions with higher proportions of rainfed rice. The changes in rice production systems of West Bengal and eastem Uttar Pradesh have not only led to a rapid growth in yield but also have reduced the effect of fluctuations in rainfall on yield instability. This finding also supports the earlier results of Sen (1987).

5

Rainfall data for the Bihar Plateau and Bihar Plains were available. Hence, the relationships for these two zones were estimated separately.

85

Increased sensitivity of output to rainfall The sensitivity of output to rainfall is a function of the pattern of rainfall, crop characteristics, field characteristics, and type of crop production technology used. Fluctuations in rainfall affect production variability through changes in both area and yield. In rainfed areas, a delay in earlyseason rainfall generally leads to a reduction in area under rice. Farm monitoring studies from eastern Uttar Pradesh have found that rice area increases when rainfall during the planting season is adequate, but the area is reduced dramatically when the rains are insufficient or late (Pandey et al 1998). Rainfall fluctuations similarly determine yield instability. Has rice production become more sensitive to rainfall variations in recent years? Rao et al (1988) have found some empirical evidence supporting an increase in sensitivity of output to rainfall in India. The exact cause for the increase in sensitivity however, is somewhat difficult to pinpoint. To the extent that the yield of modem varieties is more responsive to rainfall than that of traditional varieties, the spread of modem varieties under rainfed conditions may lead to a greater sensitivity of output to rainfall. This may be a major factor contributing to increasing production instability in states such as Bihar where modem varieties have also spread in rainfed areas. In eastern Uttar Pradesh and West Bengal, this effect is unlikely to be strong due to the expansion of irrigation. Adoption of modern varieties As mentioned in the introductory section, the adoption of modern varieties and associated crop management practices is often believed to affect production variability. For eastem India, Pal et al (2000) have reported that the CV of state-level yield of rice is negatively correlated with the quantity of fertilizer applied per hectare, indicating that the use of fertilizer has a stabilizing effect. If the quantity of fertilizer used is considered as a proxy for the extent of

adoption of modem varieties, these results indicate that modem varieties have had a stabilizing effect on rice yield. Are modem varieties intrinsically more stable or unstable than traditional varieties? This question remains a topic of debate. Earlier literature suggested that the yield of modem varieties is likely to be more unstable than that of traditional varieties (Mehra 1981). The early modem varieties were susceptible to some pests and diseases and did lead to a yield reduction in some areas. Recent varieties are more tolerant of these biotic stresses. Similarly, early modem varieties performed better under favorable conditions and did not do as well as traditional varieties under unfavorable conditions. However, more recent evidence indicates that, while modem varieties may have a higher variance, they mostly have a lower CV of yield and are not riskier than traditional varieties (Anderson et al 1989, Witcombe 1989). The results of analysis of the moving CV of yield presented earlier also support the view that yield variability (measured by CV) has declined over the years as modem varieties have become more widespread in some parts of eastern India. Further light on this issue can be shed by using farm- and state-level data from eastern India. Stochastic dominance analysis of yield data collected from a panel of farmers during 1994-98 from eastem Uttar Pradesh shows that modem varieties dominate traditional varieties in the sense of first-degree stochastic dominance (Anderson et al 1977). The cumulative probability distribution functions of modem varieties are mostly on the right-hand side of those of traditional varieties, indicating that modem varieties are less risky (Fig. 4). A similar analysis using state-level data for the two periods (1969-81 and 1981-94) also indicates that rice yield has become less risky in the second period relative to the first period in all states6 (Fig. 5). Thus, the adoption of modem varieties in eastern India could be considered to have a stabilizing effect.

6

A limitation of this analysis is that it is based on yield rather than on net return. If the cost of production associated with modem varieties is substantially higher than that of traditional varieties, the conclusions may not be valid.

86

Despite the modem varieties being less “risky,” why are there such large interdistrict differences in changes in yield variance between the two time periods? The change in districtlevel variance of yield in eastem India between the two periods analyzed varied from -77% to 700%. This wide range reflects the differences in agroclimatic conditions and the types of technologies being used for growing rice in these 71 districts of eastem India. Although the change in yield variance was between -50% and 50% in 55% of the districts, almost 25% of the districts experienced a change in production variance of more than 100%. Earlier, it was indicated that at least a part of this difference is caused by the difference in rainfall variability between the two time periods. The districts also vary in the extent of rice area irrigated, adoption of modem varieties, and use of chemical fertilizers. While some districts have experienced a rapid adoption of modem varieties and fertilizers, rice technologies in other districts have changed very little. Variables measuring these technological and climatic differences could be used to explain the interdistrict

variation in changes in production variability. However, a major problem with such an approach is the unavailability of accurate district-level data on major variables such as rainfall, extent of adoption of modem varieties, level of fertilizer use, and extent of irrigation. Rainfall data are available on the basis of meteorological subdivisions, not districts. The data on the adoption of modem varieties, especially in rainfed areas, are subject to high degrees of error as the estimates are not based on sample surveys. The fertilizer-use data available refer to the quantity of fertilizer sold, not the quantity applied to rice. While the data on sales may be a good proxy for fertilizer use at the state level, such an equivalence tends to be less accurate at the district level due to substantial cross-border movements of fertilizer. Data on area irrigated are similarly subject to large errors due to the expansion of private irrigation, which is mostly unreported. Even for publicly sponsored large-scale irrigation projects, data available refer to irrigation capacity created rather than actual area irrigated.

87

These data limitations thus constrain efforts to correlate changes in production variability across districts with climatic and technology variables. A somewhat indirect approach is used here to get around this data constraint. Districts experiencing a rapid adoption of improved seeds and fertilizers are likely to record a higher growth rate in yield than those that have slower rates of adoption of these inputs. Thus, the rate of growth in yield could be used as a proxy for the adoption of improved technologies. 7 A negative relationship is observed between the rate of growth in yield during the second period and the percentage change in yield variance between the two time periods (Fig. 6A). The low explanatory power of the model indicates the presence of many location-specific factors that vary across districts. A model based on a single variable encapsulating a variety of effects is unlikely to be able to account adequately for interdistrict differences, especially when such differences are very small. On the other hand, even such a simplified model may have a better explanatory power when the differences considered are large. Accordingly, a second regression was run on a subset of districts for which the increase in yield variance was more than 100%. The results (Fig.6B) indicate a highly significant negative correlation between the percentage growth rate in yield and the percentage change in yield variance. The explanatory power of the model has also improved substantially. This negative relationship indicates that districts where yield growth rates have been high due to rapid technological change also experienced a relatively smaller increase in yield instability. In other words, technological changes that have resulted in a strong growth in yield have also dampened the tendency toward increased yield variability in eastern India. The increase in yield

variance has been more in those districts that experienced relatively slower growth in yield. This result indicates that, contrary to conventional wisdom, growth and stability can move together.

Increased price elasticity of output
Output may vary from year to year as farmers alter their input use, depending on the prices of inputs and outputs. For a given degree of price instability, output instability will increase with an increase in the price elasticity of output. Thus, a systematic pattern of change in price elasticity over time may induce changes in output instability. One of the factors that determine the price elasticity of the supply of rice is the share of purchased inputs in the total cost. The theory of production economics dictates that the price elasticity of output increases with an increase in the share of purchased inputs. The intuitive reasoning is that a change in output price will induce a compensatory change in purchased variable inputs as farmers attempt to maximize profits. When the share of purchased inputs is greater, a given percentage change in the quantity of such inputs would mean a greater absolute change in output. Thus, the output would appear to be fluctuating more when the share of purchased variable inputs in the total production cost is higher. Since the share of purchased inputs such as fertilizers is generally higher for modem varieties (Kshirsagar et al 1997). the supply elasticity of output can be expected to increase with the expansion of area under modem varieties. It is difficult to test this hypothesis empirically, as estimates of supply elasticity for different rice varieties grown under similar conditions are not available.

The rate of yield growth as a proxy for the extent of adoption of modern technologies can be problematic if area expansion, not yield growth, is the major source of output growth. Similarly, if average productivities across districts differ widely, the growth rate in yield may be a poor proxy for the extent of adoption of modern technologies. A district with a high average yield and ceiling level of adoption of modern technology may record a slower growth rate relative to districts with a low average productivity where technology adoption has just commenced. For eastern India, the major source of growth in rice output has been the increase in yield. Similarly, during the first period when the adoption of modem varieties and fertilizers was limited in eastern India, interdistrict differences in yield were relatively small.
7

89

An additional effect that will tend to increase the price elasticity of rice is the trend toward increasing crop diversification in eastern India. This is evident in eastern Uttar Pradesh where rice contributes only 33% to the gross cropped area. When more cropping choices are 90

available, the price responsiveness of individual crops increases as farmers allocate their resources among alternative crops relatively smoothly. The response to price changes occurs mainly through adjustment in area planted. This is probably why the instability of area under rice

has increased in eastern Uttar Pradesh in recent years. Price-induced fluctuations in output can be expected to increase (decrease) if the instability in price has increased (decreased). For eastern India, the variability of output price may have decreased over time as markets have become more integrated and the domestic output of rice has increased steadily over time. Similarly, input markets are better integrated than they used to be during the early phases of the Green revolution. Any resulting reduction in the variability of prices is thus likely to have a dampening effect on production instability. Although farmers’ supply response may be more closely correlated with farm-gate prices, we have examined the variability of the procurement price for rice as uninterrupted timeseries data on procurement prices (minimum price guaranteed by the government) were available from published sources. Furthermore, the farm-gate price and procurement price tend to he highly correlated. Although the 10-year moving CV has increased in recent periods,

Variability of price

there is some evidence that the variability of the real procurement price8 had an overall negative trend over the period 1970-94 (Fig. 7). Shift in rice areas The shift in rice production from marginal to favorable areas can be expected to decrease instability. As the agricultural systems in eastern India have become more diversified, there is some evidence that other crops such as pulses are replacing for rice in marginal areas. For example, in Orissa and Bihar, pulses and oilseeds have replaced autumn rice, which is mostly grown in marginal areas (Naik et al 2000). With further diversification of agriculture, rice production can be expected to shift from more marginal to less marginal and favorable areas, resulting in greater yield stability. Increased production of summer (boro) rice Eastern India has experienced a rapid expansion of boro rice in recent years. This has occurred

8

Real price obtained by deflating the nominal procurement price of rice by the price index of all commodities as reported in (1995).

CMIE

91

mostly in West Bengal but some expansion of boro rice has also occurred in Bihar and Assam. Since boro rice is grown under climatic conditions more stable than those under which kharif rice is cultivated, an increase in area under boro rice can be expected to have a stabilizing effect on rice production. The data from West Bengal indicate that the stabilizing effect of boro rice comes mainly from the reduced instability of rice area. As boro rice is planted using a relatively assured supply of water, the instability in boro rice area tends to be considerably lower than that of rainy-season (kharif) rice. The area instability in four districts of West Bengal (Howrah, Midnapore, Nadia, and 24-Parganas), which experienced a rapid expansion of boro cultivation, declined dramatically between the two time periods. The area under boro rice in these districts increased by 50% between 1986 and 1994. The decline in the CV of area of these districts was 15-55%. Although yield stabilization was also observed in some cases, the effect was not as strong as for area stabilization.

Increased share of output from states with greater production stability
Eastern Uttar Pradesh and West Bengal are the two major states in eastern India that have shown an increase in stability in rice production. An increase in the contribution of these states to rice output in eastern India would have a stabilizing effect. The average contribution of production from eastern Uttar Pradesh to eastern Indian output increased from 10% to 15% between the two time periods. Similarly, the average share of production from West Bengal increased from 30% to 33%. On the other hand, the average share of Bihar, where production instability is higher, decreased from 24% to 18%. Such shifts in relative contribution can be expected to have a stabilizing effect on rice production. Rising importance of yield growth as a source of production growth The importance of yield in determining the production growth rate has increased in the 92

second period relative to the first period. During the first period, yield growth contributed to 68% of production growth in eastern India. During the second period, the contribution of yield growth increased to 86%. As yield tends to be a relatively more unstable component than area, it is believed that the increased importance of yield growth may lead to greater instability of output (Rao et al 1988). How has the relative size of area and yield variability changed between the two periods? Yield instability can be expected to decline with the expansion of irrigation. This will have a stabilizing effect on output if area instability does not increase. However, the instability of area can increase with commercialization and diversification of agriculture. When farmers have more options in market-oriented diversified systems, they are likely to switch between rice and a nonrice crop more often, depending on climatic conditions and expected relative prices of alternative crops. Thus, area variability is likely to become an important source of variability as rice production systems in eastern India become more diversified over time. In fact, for eastern India as a whole, the ratio of the CV of yield to the CV of area was 5.4 in the first period. This ratio became 2.7 in the second period, indicating that the instability of area has become relatively more important between the two time periods. However, even during the second period, the CV of yield was at least twice as high as the CV of area. As future growth in rice production will arise mostly from an increase in yield, production instability is likely to increase unless yield instability can be reduced through improved technologies. Production variability is likely to rise further if this is accompanied by increased area variability.

Concluding remarks
The analysis presented in this paper indicates that the CV of production and yield in eastem India has declined over time. The decline has been sharper in West Bengal and eastern Uttar Pradesh where groundwater irrigation has expanded rapidly. Even in the Bihar Plains and the Coastal Plains of Orissa, where rice production is characterized by high risk of

flooding, the CV of yield has increased only marginally. Although the CV of yield has shown a declining trend, the absolute variability in production has increased in most states and agroclimatic zones. The major sources of increases in variance in eastern India are the increase in within-state yield variance and areayield covariance as well as an increase in interstate correlations. Judging by the effect on production variance, eastern Indian states can be divided into two categories. The expansion of irrigation, the adoption of modern varieties, and the growth in yield of rice are most prominent in eastern Uttar Pradesh and West Bengal. These factors have reduced not only the yield variance in many districts of these states but also the interdistrict correlation in yields. A reduction in yield variance and an increase in mean yield have thus led to a substantial reduction in the CV of production in these states. On the other hand, production variability has increased only where modem varieties have spread in areas with poorer hydrological control such as in the Northeast Alluvium (Bihar Plains) and Eastern Coastal Plains of Orissa. These two differing patterns indicate that the observed changes in production variability are the result of interplay among several factors, not just of adoption of modem varieties. Although modern varieties are often considered to be riskier than traditional varieties, the analysis based on both farm-level and district-level data indicates that this is not the case. At least a part of the change in the variability of production was found to be due to changes in the instability of rainfall. The analysis showed that CV of yield closely tracks the CV of rainfall. indicating that in spite of some progress in the development of irrigation facilities, rice production in eastern India still depends largely on the vagaries of the monsoon. Other factors that have affected variability are the expansion of irrigation, shifts to summer rice, increased production from more favorable areas, reduced price variability, and shifts in rice production to less marginal areas. The results generally support the hypothesis that yield growth and stability are not mutually inconsistent goals. In fact, production stability has increased only in those districts of eastern

India where yield growth has been very rapid. The expansion of irrigation and the development of varieties suitable to the environment of eastern India are likely to be the causal factors that have reduced instability and increased growth simultaneously. Districts that failed to increase the growth rate in yield for a variety of reasons are the ones where production instability has tended to increase. Thus, emphasis on reducing yield instability in plant breeding programs, even at the cost of reducing yield growth, may be counterproductive. Plant breeders and others involved in technology development may be more successful in increasing and stabilizing rice output by focusing on strategies to increase the average yield. Of course, adequate policy support for irrigation, credit, and marketing of inputs and outputs plays a critical role in realizing the potential benefits of improved varieties.

References
Anderson JR, Hazel1 PB. 1989. Variability in grain yields: implications for agricultural research and policy in developing countries. Baltimore, Md (USA): The Johns Hopkins University Press. Anderson JR, Findlay CJ. Wan GH. 1989. Are modern cultivars more risky? A question of stochastic efficiency. In: Anderson JR, Hazell PB, editors. Variability in grain yields: implications for agricultural research and policy in developing countries. Baltimore, Md (USA): The Johns Hopkins University Press. Anderson JR, Dillon JL, Hardaker JB. 1977. Agricultural decision analysis. Ames, Iowa (USA): Iowa State University Press. Ballabh V, Pandey S. 1999. Transitions in rice production systems in eastern India: evidences from two village in Uttar Pradesh. Econ. Pol. Wkly. 34 (13): All-A16. Barker R, Gabler EC, Winkelmann D. 1981. Long-term consequences of technological change on crop yield stability: the case for cereal grain. In: Valdes A, editor. Food security for developing countries. Boulder, Colorado (USA): Westview Press.

93

Bindlish VL. 1988. An analysis of rice yield variability in India based on district-level data: estimate of the separate variances for the high-yielding and traditional varieties and the role of expectations in measuring variability. PhD thesis, Cornell University. Byerlee D. 1995. Technology transfer systems for improved crop management: lessons for the future. In: Anderson JR, editor. Agricultural technology: policy issues for the international community. Cambridge (UK): Cambridge University Press. CMIE. 1995. India’s agricultural sector: a compendium of statistics. Centre for Monitoring Indian Economy, Economic Intelligence Service, Bombay, India. DES (Directorate of Economics and Statistics). 1999. Agricultural statistics at a glance. Directorate of Economics and Statistics, Department of Agriculture and Cooperation, Ministry of Agriculture, New Delhi, India. FAI (Fertilizer Association of India). 1998. Fertilizer statistics 1997-98. Fertilizer Association of India, New Delhi, India. Griffin K. 1988. Alternative strategies for economic development. Organization for Economic Cooperation and Development, Pans. Hazell PB. 1982. Instability in Indian foodgrain production. International Food Policy Research Institute, Washington, D.C. Hazell PB. 1984. Sources of increased instability in Indian and U.S. cereal production. Am. J. Agric. Econ. 66(3):302-311. Kshirsagar KG, Pandey S, Bellon M. 1997. Farmer perceptions, varietal characteristics and technology adoption: the case of rainfed rice village in eastern India. Social Sciences Division, Internationsl Rice Research Institute, Los Baños, Philippines. Mehra S. 1981. Instability in Indian agriculture in the context of the new technology. Washington, D.C. (USA): International Food Policy Research Institute. Naylor R, Falcon W, Zavaleta E. 1997. Variability and growth in grain yields, 195094. Popul. Dev. Rev. 23(1):41-58. Naik D, Pandey S, Behura D, Villano RA. 2000. Risk and rice production in Orissa, eastern India. In: Pandey S, Barah BC, Villano RA, 94

Pal S, editors. Risk analysis and management in rainfed rice systems. Los Baños (Philippines): International Rice Research Institute. Pal S, Pandey S, Abedullah. 2000. Growth and variability in agriculture revisited: districtlevel evidence of rice production in eastem India. (this volume) Pandey S. 1989. Irrigation and crop yield variability: a review. In: Anderson JR, Hazell PBR, editors. Variability in grain yield: implications for agricultural research and policy in developing countries. Baltimore, Md (USA): The Johns Hopkins University Press. Pandey S, Singh HN, Villano RA. 2000. Rainfed rice and risk coping strategies: some microeconomic evidence from eastern Uttar Pradesh. (this volume) Planning Commission. 1989. Agro-climatic regional planning: an overview. New Delhi, India. Rao CHH. 1968. Fluctuations in agricultural growth: an analysis of unstable increase in productivity. Econ. Pol. Wkly. 3237-94. Rao CHH, Ray SK, Subbarao K. 1988. Unstable agriculture and droughts: implications for policy. Institute of Economic Growth, New Delhi, India. Sen SR. 1967. Growth and instability in Indian agriculture. Agric. Situation India 2(10):827839. Sen SR. 1987. Growth and instability in agriculture. Keynote address at the National Symposium on Growth and Instability in Agriculture organized by the Indian Agricultural Statistics Research Institute, New Delhi, India. Singh AJ, Byerlee D. 1990. Relative variability in wheat yields across countries and over time. J. Agric. Econ. 41 :21-32. Walker TS. 1989. High-yielding varieties and variability in sorghum and pearl millet production in India. In: Anderson JR, Hazell PBR. editors. Variability in grain yields: implications for agricultural research and policy in developing countries. Baltimore, Md (USA): The John Hopkins University Press.

Witcombe JR. 1989. Variability in the yield of pearl millet varieties and hybrids in India and Pakistan. In: Anderson JR, Hazell PBR, editors. Variability in grain yields: implications for agricultural research and policy in developing countries. Baltimore, Md. (USA): Johns Hopkins University Press. p 207-220.
Appendix 1. Components of change in variance of production. Source of change Description Change in mean yield Change in mean area Change in yield variance Change in area variance Interaction between changes in mean yield and mean area Change in area-yield covariance Interaction between changes in mean area and yield covariance Interaction between changes in mean yield and area variance Interaction between changes in mean area and mean yield and changes in area-yield covariance Change in residual
a

Components of change Symbola

A = area sown, y = yield, V = variance. Source: Hazell (1982).

95

Appendix 2. Districts included in the study by state. Eastern Uttar Pradesh Allahabad Azamgarh Bahraich Ballia Basti Deoria Faizabad Ghazipur Gonda Gorakhpur Jaunpur Mirzapur Pratapgarh Sultanpur Varanasi Eastern Madhya Pradesh Balaghat Bastar Bilaspur Durg Jabalpur Mandla Raigarh Raipur Seoni Shahdol Surguja Bihar Bhagalpur Champaran Darbhanga Dhanbad Gaya Hazaribagh Monghyr Muzaffarpur Palamau Patna Purnea Ranchi Saharsa Santhal Pargana Saran Sahabad Singhbhum Orissa Balasore Bolangir Cuttack Dhenkanal Ganjam Kalahandi Keonjhar Koraput Mayurbhanj Phulbani Puri Sambalpur Sundergarh West Bengal 24-Parganas Bankura Birbhum Burdwan Cooch-Behar Dajeeling Hooghly Howarh Jalpaiguri Malda Midnapore Murshidabad Nadia Purulia West Dinajpur

Notes
Authors’ addresses: S. Pandey, International Rice Research Institute, MCPO Box 3127, Makati City, Philippines; S. Pal, National Center for Agricultural Economics and Policy Research, New Delhi, India. Acknowledgment: The authors acknowledge assistance from Abedullah and R.A. Villano. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

96

Growth and variability in agriculture revisited: district-level evidence of rice production in eastern India
S. Pal, S. Pandey, and Abedullah

This study attempts to explain the nature of growth and variability of yield and production of rice in eastern India. A district-level analysis covering diverse agroclimatic conditions for the period 1969-94 is used to investigate the changing pattern of production variability and its primary causes. The results show that the growth in yield of rice in eastern India accelerated in the early 1980s. Yield growth mainly contributes to growth in rice production in the eastern region. Almost 60% of rice area has shown a higher growth with a greater degree of stability in rice yield. Yield variability declined in areas where the adoption of high-yielding varieties and the use of NPK fertilizer has occurred. Based on an empirical analysis, alternative policies to help reduce risk in rice production in eastern India are suggested.

One of the significant changes that has occurred in Indian agriculture during the last decade or so is the diversification of the regional production base. Several new growth areas have emerged, contributing to the growth of national food-grain production. It is believed that the performance of agriculture in these new growth areas would be crucial to national food security in the years to come. However, several questions relating to the nature of agricultural growth have arisen. How is the growth pattern different from that observed in the “Green Revolution areas?” Is growth widespread in the region or concentrated in a few pockets? Is growth accompanied by increased production variability? The issue of production Variability is important because increased variability was considered to be an inevitable consequence of production growth (Hazell 1982). In particular, rainfed agriculture was found to be more unstable than irrigated agriculture (Dhawan 1987, Rao et a1 1988, Pal and Sirohi 1989, Singh and Byerlee 1990). Since the new areas of agricultural growth are predominantly rainfed, it would be interesting to test the hypothesis of a positive association between growth and variability. A district-level analysis covering diverse agroclimatic conditions can provide more insights into the

changing pattern of production variability, its primary causes, and management strategies. Keeping this in view, we studied rice production in eastern India because this area has shown exceptional dynamism and growth in food-grain production. This paper first outlines the analytical approach used to measure variability, followed by a discussion of the regional and temporal patterns of growth and variability in rice production in eastern India. The subsequent section deals with determinants of rice yield variability. Finally, the paper discusses policy options to reduce production variability and mitigate its impact.

Measuring variability
Several measures of variability or instability are used in the literature. These measures have their own merits and demerits (Cuddy and Della Valle 1978). Our choice of measure is decided by three factors: (1) consideration of trend in timeseries data, (2) amenability to apply the test of significance for change over time, and (3) ease with which the measure can be applied and explained. All these criteria are met by using the square root of the mean squared error of the semilog trend function, a relative measure of

97

variability that estimates the average percentage deviation from the trend and eliminates much of the scale effect often encountered in time-series data (Naylor et al 1997). Another advantage of this measure is that it can be directly computed and related to growth rate. The analysis is done for the period 19691 94, which was divided into two equal subperiods, 1969-81 (period I) and 1982-94 (period II), to study temporal changes. Period II corresponds to the period of comparatively higher growth in yield and production of rice in eastern India2 (Fig. 1). Growth and variability in area, production, and yield of rice are analyzed

for both periods in all districts of the eastern states—eastern part of Uttar Pradesh and Madhya Pradesh, Bihar, Orissa, and West Bengal. Appendix I lists the districts in each state covered in the study. The northeast states could not be included in the analysis because of a paucity of data. The study examines regional patterns of variability across administrative regions (states and districts) as well as across agroclimatic regions. The former are important for policy implementation considerations, whereas the latter are essential for explaining differences in variability.

In this paper, all years refer to agricultural years, e.g., 1969 refers to 1969-70, 1970 refers to 1970-71, and so on. All rice data are converted to rough rice. 2 The second period begins with a year (1982-83) of shortfall in rice yield and production: therefore, one may expect the results to be sensitive to selection of time period. This sensitivity was examined but the change in cutoff point for the second time period did not alter the trend in both periods significantly.
1

98

Growth and variability in rice production: regional and temporal patterns
Aggregate analysis

Growth pattern. Rice production in India increased from 63 million t during the triennium ending in 1971 to 118 million t during the triennium ending in 1994. Most of the increased production during the 1970s came from the northern region comprising the states of Punjab and Haryana (Table 1). During 1969-81, the northern region registered an impressive growth of 14% (annual compound growth rate) in production, contributed by growth in yield as well as in area. The growth rate slowed considerably, however, during 1982-94 (Table 2). In the early phase of the Green Revolution, the large increase in rice yield resulted in more

Table 1. Area (000 ha), yield (t ha-1), and production (000 t) of rice in eastern India, 1969-94.a Statelregion Eastern Uttar Pradesh Area Yield Production Eastern Madhya Pradesh Area Yield Production Bihar Area Yield Production Orissa Area Yield Production West Bengal Area Yield Production Eastern region Area Yield Production Northern region Area 'field Production All India Area Yield Production
a

1969-71 2.465 1.08 2,686 3,760 1.27 4,780 5.287 1.22 6,424 4.396 1.28 5.633 4,919 1.60 7.863 20,827 1.31 27,341 585 2.69 1,571 37,439 1.69 63,159

1979-81 2.790 1.15 3,233 4.051 1.09 4,425 5.352 1.26 6,720 4,156 1.35 5,615 5,087 1.89 9,630 21,436 1.38 29,656 1,499 4.03 6,039 40,091 4.03 75,358

1992-94 2.905 2.63 7,627 4,539 1.79 8,142 4.498 1.67 7,574 4,485 2.07 9.271 5,781 3.13 18,077 22,208 2.28 50,733 2,489 4.95 12,320 42,376 4.95 11,8676

Numbers are 3-year averages. State and regional figures are total of the selected districts.

area under rice, giving a very high growth rate. But, at a later stage, with no possibility of area expansion and yield improvement using available varieties, a sharp decrease in growth rate was noted. In fact, rice yield has now become almost stagnant. Several factors stimulated yield growth in the 1970s in favorable areas: homogeneity of the production environment; increasing investment (both public and private) in agriculture, particularly in irrigation, research, and extension; government programs for irrigated areas to increase food production; and price policy. These regions are now encountering second-generation problems relating to sustainability of the production system. Technological intervention is essential to address these problems and to sustain yield growth (Chand 1999). Eastern India, in contrast to the northern region, did not show any significant growth in rice production and yield during the first period. But, in the second period, the growth rate was about 4%, nearly doubling the yield and production and contributing nearly half of the increased production in the country. West Bengal was the only state in the region that had a statistically significant growth rate in yield and production in both periods (Table 2). In the second period, eastern Uttar Pradesh, eastern Madhya Pradesh, and Orissa also showed significant growth. In Bihar, growth was statistically insignificant in each period, but, taking both periods together (1969-94), a significant growth of 1.4% in yield and 1 % in production was noticed. In spite of appreciable yield growth in the eastern region, the current level of rice yield (2.3 t ha-1) is about half of that obtained in the irrigated areas of the northern region (Table 1). Part of the regional differences in growth rates of yield could be attributed to the extreme diversity of rice production environments in eastern India. Erratic weather, coupled with varying soil types, soil depth, rainfall, and topography, results in various kinds of abiotic and biotic stresses. However, an appreciable growth in rice production in West Bengal and Uttar Pradesh can be attributed to the successful implementation of land reform and encouragement of private investment in 99

Table 2. Growth and variability in rice production in eastern India. Statelregion 1969-94 Eastern Uttar Pradesh Area Yield Production Eastewrn Madhya Pradesh Area Yield Production Bihar Area Yield Production Orissa Area Yield Production West Bengal Area Yield Production Eastern region Area Yield Production Northern region Area Yield Production All India Area Yield Production 1.08*** 4.40*** 5.48*** 0.80*** 1.99*** 2.79*** -0.33* 1.37*** 1.05** -0.05 2.24*** 2.20*** 0.64*** 3.10*** 3.74*** 0.36*** 2.61*** 2.97*** 6.62*** 2.54*** 9.16*** 0.56*** 2.46*** 2.96*** Growth rate (%)a 1969-81 1.41*** 1.28 2.70 0.75*** -0.86 -0.11 0.48 0.63 1.11 -0.57** 0.67 0.10 0.35 1.69*** 2.04** 0.40** 0.82 1.22 9.60*** 4.87*** 14.47*** 0.82*** 1.53** 2.33*** 1982-94 0.51 4.92*** 5.43*** 0.99*** 3.08*** 4.07*** -0.57 1.95 1.38 0.85** 3.73*** 4.59*** 1.29*** 4.62*** 5.91*** 0.63** 3.86*** 4.49*** 3.50*** 0.83** 4.33*** 0.63*** 3.08*** 3.73*** F ratioc (-) 1.40 3.22* 2.07 5.65** 2.51 2.58* (-)1.69 0.34 0.03 6.76*** 2.25 3.53** 4.97** 5.04** 6.09*** 1.18 4.46** 3.44** (-)42.34*** (-)13.07*** (-)48.39*** 1.79 0.96 0.21 Annual percentage deviation from trendb 1969-94 5.52 19.25 20.87 0.95 20.03 20.49 6.59 15.01 18.77 4.71 14.43 17.78 3.60 10.16 12.73 2.63 11.25 13.12 13.33 10.84 21.91 2.24 6.84 8.17 1969-81 1982-94 F ratiod

4.12 23.76 25.19 0.66 23.87 24.22 4.45 11.64 12.88 3.60 13.56 14.95 3.48 7.80 9.84 1.89 10.65 12.04 7.57 9.59 11.51 1.65 7.68 8.95

6.47 7.82 12.87 0.94 12.01 12.48 7.90 18.46 24.51 4.12 13.91 17.28 2.71 9.68 11.43 3.17 9.11 11.87 4.75 5.03 7.83 2.59 4.75 7.01

2.46* (-)9.22*** (-)3.83** 2.25* (-)3.95** (-)3.77** 3.14** 2.51* 3.62** 1.31 1.05 1.34 (-)1.66 1.54 1.35 2.78* (-)1.36 (-)1.03 (-)2.54* (-)3.64** (-)2.16 2.47* (-)2.62* (-)1.63

a,bCoefficient and square root of the mean squared error, respectively, of the semilog regression. cChow test F ratio. dRatio of the mean squared error.***,**,*= significant at 1%. 5%, and 10% level. respectively.

agriculture, particularly in tubewell irrigation and land improvement (Rao 1994). Also, infrastructure facilities were comparatively well developed in West Bengal and Uttar Pradesh (Bhatia 1999), placing these states ahead in yield growth. Changes in variability. Table 2 also shows the magnitude of Variability or instability measured as annual percentage deviation from the trend. Variability in rice yield and production was higher in the eastern region than in the northem region or in the country as a whole. The eastern region did not show any significant change in yield variability between the two time periods, whereas it decreased significantly in the northern region. Area variability decreased in the northern region and increased in the eastern region. The lower and declining pattern of yield variability in the northern region is understandable because of assured irrigation and 100

wider adoption of high-yielding varieties (HYVs) developed primarily for the irrigated areas. Among the eastern Indian states, variability was higher in Bihar and lower in West Bengal. In fact, yield variability in Bihar increased from 11.6% in the first period to 18.5% in the second period, and the increase was statistically significant at the 10% level. On the other hand, a significant decline in the variability of yield and production was observed in the eastem parts of Uttar Pradesh and Madhya Pradesh. more so in Uttar Pradesh where yield variability was reduced to one-third of that in the first period. District-level analysis

We have analyzed growth and variability in rice production in 71 districts of eastern India, covering almost the entire rice area in the region,

or half the country's rice area. Tables 3 and 4 summarizes important patterns emerging from the district analysis. Based on the level of rice yield during the triennium ending in 1994-95, the districts were classified into two groups: (1) districts having a yield of 2.28 t ha-1 (average rice yield in eastern India) or more and (2) districts with rice yield less than 2.28 t ha-1. These groups cover 45% and 55%, respectively, of the total rice area in the eastern region during the triennium ending in 1994-95. All the districts are further grouped according to significance of change in growth and variability in yield and production. The Chow test and F test based on the ratio of mean squared errors were applied to test the significance of the change in growth rate and annual percentage deviation, respectively, between the two time periods. Temporal changes. A large number of districts with high yield have shown a significant change in growth and/or variability in rice yield and production between the two time periods (Table 3). Most of these districts have either shown a decrease in variability with no change in growth rate of yield and production or an increase in growth rate with no change in variability. The districts in the former category are mainly from Uttar Pradesh and occupy about 7% of the rice area in the eastern region during the triennium ending in 1994-95, whereas the latter group of districts belongs to West Bengal and contributes about 16-18% of the total rice area. Only three districts of Uttar PradeshGonda, Gorakhpur, and Sultanpur with only 3% of the total rice area-registered significant a increase in yield growth with a significant decrease in its variability. In fact, most of the districts of Uttar Pradesh have registered a sharp and significant decline in yield variability (Fig. 2). In contrast, growth in yield as well as in production was accompanied by an increase in variability in Puri District of Orissa. The districts showing no significant change in growth as well as in variability of yield and production occupy about 18% and 14% of the total rice area. respectively. Most of the districts in this category are from West Bengal, having a fairly low degree of variability in both periods. In the districts with low rice productivity, three important changes were observed. The first

was the increase in growth rate with no change in variability. The districts showing this trend in rice yield contributed 6% to the total rice area, whereas the area with similar changes in production contributed 13%. Second, about 8% of the total rice area, covering five districts from Madhya Pradesh and one from Orissa, showed a significant decline in variability of yield and production with no significant change in growth rate. Third, the districts with an increase in variability but with no change in growth rate of yield and production accounted for 6% and 10%, respectively, of the total rice area. These districts are mainly from Bihar and have not shown significant growth in yield in either period. It is important to note that about half the total rice area did not show any significant change in growth and variability of rice yield between the two time periods, and that rice yield in two-thirds of this area was less than 2.28 t ha-1. If we further add the area with increased Variability in yield, the share of these low-yield districts in the total area increases to 40%. These low-productivity areas are found in the alluvial plains of Bihar and in the plateau and highland areas of Bihar, Orissa, and Madhya Pradesh. Regional pattern. Because of the changes in the growth and variability of rice yield over time, it would be worthwhile to examine how rice yield behaved in terms of growth vis-á-vis the actual level of variability. About 75% of the total rice area had a significant growth in yield during 1982-94 (Table 4). More than half of this area or 41% of the total rice area, had low yield variability (annual percentage deviation from the trend in yield less than 15%). These higher performing areas had rice yields of 2.28 t ha-1 or more and are spread mainly in the alluvial plains of West Bengal and Uttar Pradesh. About onethird of the total rice area had significant growth in yield with a moderate level of variability (1530%). Rice yield was stagnant and moderate to highly unstable in 22% of the total rice area, exclusively in low-productivity areas. In fact, nearly 80% of the area with low rice productivity had a moderate to high degree of variability, versus one-fourth of the highproductivity area showing moderate variability. An examination of the position of various districts under the columns of production and 101

Table 3. Changes in growth and variability in production and yield of rice in eastern India by districts. Change in growth rate Decrease Change in variability (annual percentage deviation from trend) Production No change increase Decrease Yield No change Increase

Districts with yields of 2.28 t ha-1 or higher Increase Sultanpur (0.71) Raipur, Bankura, Burdwan. Darjeeling, Midnapore, Murshidabad, Purulia, West Dinajpur (18.41) Ballia. Gonda, Gorakhpur, Sahabad, Sambalpur, 24-Paragana, Birbhum, Hooghly, Malda (13.78) Puri (1.62) Gonda. Gorakhpur, Sultanpur (3.0) Raipur, Bankura, Darjeeling, Midnapore, Murshidabad. Purulia, West Dinajpur (15.77) Ballia, Sahabad, Ganjam, Sambalpur, 24-Paragana, Birbhum, Burdwan, Hooghly, Howrah, Malda, Nadia (17.48) Puri (1.82)

No change

Allahabad, Azamgarh, Faizabad, Ghazipur, Jaunpur, Mitzapur. Pratapgarh, Nadia, Varanasi, Howrah (7.70) Deoria (1.06)

Ganjam (1.47)

Allahabad, Azamgarh, Deoria, Faizabad, Ghazipur, Jaunpur, Mirzapur, Pratapgarh, Varanasi (6.88)

Decrease

Districts with yields less than 2.28 t ha-1 Increase Basti, Bilaspur, Balasore. Bolangir, Dhenkanal, Cooch-Behar, Jalpaiguri (13.04) Balaghat, Jabalpur, Raigarh, Shahdol, Surguja, Kalahandi (7.71) Durg, Mandla, Seoni, Bhagalpur, Dhanbad, Gaya, Hazaribagh, Palamau, Purnea, Ranchi, Saharsa, Santhal Paragana, Saran, Singhbhum, Cuttack, Keonjhar. Koraput, Mayurbhanj, Phulbani, Sundergarh (25.85) Bahraich. Bastar, Champaran, Darbhanga, Monghyr, Muzaffarpur, Patna (9.56) Basti (1.76) Bahraich, Bolangir, Keonjhar. Cooch-Behar, Jalpaiguri (5.95)

No change

Balaghat, Jabalpur, Raigarh, Shahdol, Surguja, Dhanbad, Kalahandi (7.92)

Bilaspur, Durg, Mandla, Seoni, Bhagalpur, Champaran, Gaya, Hazaribagh, Palamau. Patna, Purnea, Ranchi, Saharsa, Santhal Paragana, Saran, Sighbhum, Balasore. Cuttack, Dhenkanal, Koraput, Mayurbhanj, Phulbani, Sundergarh (33.34)

Bastar, Darbhanga. Monghyr, Muzaffarpur (6.05)

aNumbers in parentheses are rice area of the group as percentage of total rice area in eastern India during 1992-94. Changes in growth and variability are statistically significant at the 10% level. None of the districts in the low-productivity group showed a significant decrease in the growth rate.

Table 4. Variability in production and yield of rice by districts, 1982-94. Growth Variabilitya Significant growth Production No growth Significant growth Yield No growth

Districts with yields of 2.28 t ha-1 or higher 45% Deoria, Faizabad. Varanasi, 24-Paragana, Burdwan, Dajeeling. Hooghly, Malda, Midnapore, Nadia. West Dinajpur (19.17)b Allahabad, Azamgarh. Deoria, Faizabad, Ghazipur, Gorakhpur, Jaunpur, Mirzapur, Pratapgarh, Sultanpur. Varanasi. Sambalpur. 24-Paragana, Bankura, Birbhum, Burdwan, Darjeeling, Hooghly, Malda, Midnapore, Murshidabad, Nadia, West Dinajpur (33.10) Gonda, Sahabad (2.22) Ballia, Gonda, Raipur, Sahabad, Ganjam, Puri. Howrah, Purulia (11.86)

15-30%

Allahabad, Azamgarh. Ballia, Ghazipur, Gorakhpur, Jaunpur, Mirzapur, Pratapgarh, Sultanpur, Raipur, Puri. Sambalpur, Bankura, Birbhum, Howrah, Murshidabad, Purulia (22.09) Ganjam (1.47)

>30%

Districts with yields less than 2.28 t ha-1 45% Raigarh, Mayurbhanj, Cooch-Behar, Jalpaiguri (5.86) Basti, Bastar, Bilaspur, Durg, Mandla, Surguja, Purnea, Saharsa, Balasore, Bolangir. Cuttack, Dhenkanal. Kalahandi, Keonjhar, Koraput. Sundergarh (28.66) Bahraich, Jabalpur (1.58) Balaghat, Shahdol, Bhagalpur, Champaran, Ranchi, Santhal Paragana, Saran, Singhbhum (10.58) Raigarh. Purnea, Mayurbhanj, Cooch-Behar, Jalpaiguri (7.92) Basti, Bilaspur, Durg, Surguja, Balasore, Bolangir, Cuttack, Dhenkanal, Kalahandi, Keonjhar. Sundergarh (20.26) Champaran. (3.63) Koraput

15-30%

Balaghat, Bastar, Mandla, Shahdol, Bhagalpur, Dhanbad. Gaya, Patna, Ranchi. Saharsa, Santhal Paragana, Saran, Singhbhum, Phulbani (16.81) Jabalpur, Seoni, Darbhanga, Hazaribagh, Monghyr, Muzaffarpur, Palamau (5.59)

>30%

Seoni, Darbhanga, Dhanbad, Gaya, Hazaribagh, Monghyr, Muzaffarpur, Palamau, Patna, Phulbani (8.35)

Bahraich

(0.92)

Annual percentage deviation from trend. during 1992-94.
a

b

Figures in parentheses are rice area of the group as percentage of total rice area in eastern India

yield in Table 4 gives some idea about the relative contribution of area and yield to production variability. By and large, variability in rice production was influenced by yield variability in both the low- and highproductivity areas. However, production variability in high-productivity districts was higher than yield variability, indicating the significant contribution of area variability and covariate movements between yield and area.

Year-to-year changes in rice area in highproductivity regions could be due to farmers' responses to changes in output prices, wages, and rainfall. In favorable environments where rice yields tend to be higher, farmers may shift some area to more profitable and less laborintensive activities such as sugarcane growing. The increase in correlation between area and yield may also contribute significantly to the increase in production variability. As a matter of 103

Annual percentage deviation during 1982-94

Annual percentage deviation during 1969-81

Fig. 2. Significant changes in annual percentage deviation from the trend in yield.

fact, much of the increase in absolute variability of crop production until the mid-1980s was attributed to an increase in area-yield covariability (Hazell 1982, Pal and Sirohi 1989). A similar tendency is also observed for rice production in eastern India. In Uttar Pradesh and Bihar, rice areas having positive and significant correlation between area and yield increased sharply from more than 40% in the first period to more than 80% in the second period (Table 5). The correlation of rice yield between districts also increased steeply in Bihar but decreased in Uttar Pradesh and Madhya Pradesh. These changes in covariability are understandable in unirrigated, low-rainfall areas such as in Bihar, where the impact of deficit rainfall on area and yield is stronger on a large area. In Uttar Pradesh, however, a decrease in the number of districts with a positive yield correlation could be due to the rapid expansion of irrigated area and the decreasing sensitivity of rice yield to rainfall. Expansion of irrigation is also expected to reduce area-yield covariability, but it failed to do so in Uttar Pradesh. This implies that other factors, such as fanners’ response to output price

changes, induce area-yield covariability. In the expectation of high prices, farmers allocate more area to rice and use more inputs to increase yield; the reverse holds true when the price is expected to be low.

Determinants of rice yield variability
Crop yield and its variability are governed by interactions between agroclimatic conditions and technology. In a production environment where farmers have more control on crop production methods, production tends to be less sensitive to weather. The availability of infrastructure such as irrigation, inputs, and a technology delivery system helps in the better management of production processes and enables farmers to employ adaptive and compensatory strategies to reduce the likely adverse effect of weather on crop output. Although quantification of all these effects is rather difficult-partly because of the complexity of the relationship and partly because of nonavailability of data-we have analyzed the possible causes of yield variability in this framework.

104

Table 5. Changes in area-yield and yield covariability. State Total number of districts Districts with positive and significant area-yield correlation % of rice area common in both periods % of total rice area Number of positive and significant yield correlations between districts 1969-81 1982-94 Total number of yield correlations

1969-81 Number of districts

1982-94

% of total Number of rice area districts

Eastern a Uttar Pradesh Eastern Madhya Pradesh Bihar Orissa b West Bengal Total
a

15

6

47.1

12

82.5

47.1

72

16

105

11 17 13 15 71

3 9 6 13 38

10.4 42.8 53.1 69.3 45.7

2 15 8 7 44

19.12 90.0 65.7 49.6 59.1

0 36.0 36.0 49.6 33.6

50 22 41 42 227

13 95 49 30 203

55 136 78 105 479

There were two negative and significant yield correlations in the second period. b There was one negative and significant area-yield correlation (in Balasore District) in the first period.

Rice yield variability and agroclimatic conditions Rice is grown in extremely diverse agroclimatic conditions in eastern India. The type and depth of soil, climate, rainfall, topography, and temperature vary widely. In some districts, all rice production systems–irrigated, upland, lowland, and deepwater–are present. Nevertheless, an analysis of rice yield variability by agroclimatic zones would be useful. The districts covered in this study pertain to four agroclimatic zone: Middle Gangetic Plains, Lower Gangetic Plains, Eastern Plateau and Hills, and East Coast Plains and Hills, as identified by the Planning Commission (1989). Table 6 gives the important agroclimatic conditions of these zones and some of their subzones, along with variability in rice yield. One can easily see from this table that HYVs are widely grown on unirrigated areas in all the zones, and some of these areas, particularly uplands with harsher environmental conditions, have low and unstable rice yield.3 Such areas

appear to be less in the Lower Gangetic Plains comprising the alluvial districts of West Bengal. Therefore, rice yield is high and stable in this zone. Also, in this zone boro or summer rice is widely grown under irrigated conditions using high doses of fertilizer. It is estimated that boro rice has a yield advantage of about 1.5 t ha-1 over irrigated rice in the wet season and about 2.28 t ha-1 over the rainfed lowland rice. A similar yield behavior was observed in eastern Uttar Pradesh, which has a higher proportion of irrigated area, mainly because of more private investments in tubewell irrigation encouraged by land reform and flood control measures. This has facilitated the adoption of HYVs and use of modem inputs (fertilizers), leading to an overall dynamism in agriculture (Ballabh and Pandey 1999). In the northeast alluvial subzone or the Bihar plains of the Middle Gangetic Plains and in coastal areas of Orissa, modem varieties are grown on unirrigated areas with less inputs, giving high yield variability. Rainfall (June to September) in these two zones is comparatively lower but more erratic (Table 7). The northeast

3

The percentages of irrigated and HW area are not strictly comparable, as the former does not refer to rice. In the absence of reliable data on irrigated rice area, the percentage area irrigated is used as an approximation.

105

Table 6. Variability in rice yield by agroclimatic zones. Agroclimatic zone/ subzonea Annual Rice NPK Gross Rice Yieldc rainfall areac usec irrigated area (t ha-1) (mm)b (000 ha) (kg ha-1) areac,f under (%) HYVsc (%) 1.175 4.892 88.3 37.6 42.8d 3.6 Mean yield (t ha-1) 196981 1.8 198294 2.8 Annual %deviation in yield 196981 8.11 198294 9.7 F ratio 1.43

Soil typea

Climatea

Lower Gangetic Plains (or Gangetic West Bengal)

Red and yellow alluvial (new and deltaic)

Moist subhumid to dry subhumid

Middle Gangetic Plains Subzone I Northwest alluvial (or East UP) Subzone II Northeast alluvial (or Plains of Bihar) Alluvial (new), calcareous Alluvial, tarai, calcareous Moist subhumid to dry subhumid Dry subhumid to moist subhumid 901 2467 89.7 52.8 74.5e 2.6 1.2 2.2 22.9 8.5 (-)4.82***

992

3118

74.9

51.3

62.4d

1.8

1.4

1.7

12.9

18.4

2.03

Eastern Plateau and Hills Subzone I Eastern Plains (Plains of MP and Orissa) Subzone II Eastern Highland (MP and Orissa) Subzone III Plateau (north-central, eastem plateau, and tribal areas of Bihar, MP, and Orissa) East Coast Plains and Hills (North Orissa coast) Medium to deep black red and yellow Dry subhumid 1121 3641 36.6 30.3 68.7 2.1 1.3 1.8 22.2 11.7 (-)3.58**

Red sandy, red and yellow

Moist subhumid to dry subhumid Moist subhumid to dry subhumid

1113

1802

16.0

15.0

55.2

1.5

1.1

1.4

22.5

10.2

(44.87***

Red sandy, red and yellow, red loamy laterite

1055

3141

27.1

12.0

30.1**

1.7

1.3

1.5

17.6

14.2

(-)1.07

Deltaic alluvial, coastal alluvial, laterite, red loamy

Moist subhumid

1084

1732

33.5

46.5

69.3

2.3

1.4

2.0

11.6

18.4

2.52'

aMP = Madhya Pradesh, UP = Uttar Pradesh. Source: Planning Commission (1989). bMean rainfall (June to Sep) during 1969-94. cTriennium av ending 1994. Triennium average ending in 1987. e Av of 1988-89. fGross irrigated area as percentage of gross cropped area. ***,**,* = significant at 1%, 5%, and 10% level, respectively.

subzone is also dryer and calcareous, and more prone to waterlogging, floods, and droughts (Ghosh 1991). The eastern plains covering the Chhatisgarh plains of Madhya Pradesh and parts of Orissa have comparatively higher average yields and lower variability. This could be due to deep soils capable of retaining soil moisture for a long time. Rice yields are low and unstable in 106

the plateau region due to poor soil fertility and poor moisture-holding capacity.

Sensitivity of rice yield to rainfall
As rainfall pattern is likely to be the dominant cause of rice yield variability, it would be worthwhile to analyze variation in rainfall and

Table 7. Frequency of deficit rainfall in eastern India. Meteorological subdivision Mean rainfall (June to Sep, in mm) 1969-8 1 Eastern Uttar Pradesh Eastern Madhya Pradesh Bihar Plains Bihar Plateau Orissa Gangetic West Bengal
a

CV (%) of rainfall 1969-81 27.96 20.12 16.76 18.32 11.70 16.07 1982-94 11.39 19.97 22.22 19.39 16.36 12.41

1982-94 863 1,148 988 1,018 1,107 1.158

Direction of change in yield variability Decrease Decrease Increase Constant Constant Decrease

Number of years with actual rainfall less than 80% of mean a 5 (1972, 77, 79, 87, 92) 3 (1974, 79.87) 4 (1972, 82.86, 92) 4 (1979, 82, 86, 92) 2 (1974, 87) 2 (1976, 82)

Number of years with actual rainfall 80-90% of mean a

939 1,3 996 1,070 1,060 1,193

3 (1974, 86, 93) 6 (1973, 76, 81, 86, 89, 91) 5 (1975, 77, 79, 91, 94) 3 (1972, 73, 93) 1 (1979) 4 (1973, 79, 89, 92)

Deficit rainfall years are given in parentheses.

quantify its effect on yield. We have examined the effect of rainfall variation on rice yield in the meteorological subdivisions 4 of eastern India. Since the southwest monsoon is critical to rice production, rainfall during June to September was considered. It is unrealistic to expect a significant change in rainfall pattern over a fairly long period, but year-to-year changes in rainfall and its distribution over the crop growth period are common. These year-to-year fluctuations in rainfall are measured around the average rainfall for a meteorological subdivision. The number of years with severe deficit in rainfall-actual rainfall during June to September is less than more 80% of the average during 1969-94-are in Bihar and Uttar Pradesh (Table 7). The distribution of deficit rainfall years over the period was such that the coefficient of variation (CV) of rainfall was lower during 1969-81 in the plains of Bihar, whereas the CV was higher during 1969-81 in Uttar Pradesh. These differences in the CV of rainfall are similar to those in yield variability in these two states. The CV of rainfall is comparatively low in Gangetic West Bengal and Orissa. In fact, Orissa has the lowest number of deficit rainfall years (severe or

moderate) in eastern India; therefore, rice yield should have been stable in this state. The unexpected high yield variability in Orissa could be attributed to differences in rainfall and topography across the districts. As seen earlier, Orissa forms part of the plateau, highland, and coastal agroclimatic zones, and uplands with shallow soils in these zones may experience early or late-season moisture stress even in a normal rainfall year. To quantify the effect of rainfall on rice yield, yield was regressed on time trend and actual rainfall during the crop growth period (June to September) by the ordinary least squares (OLS) method. Here, rainfall is treated just like any other input and excess rainfall is expected to have a decreasing marginal product. To capture this effect, rainfall variable of the second degree (ie., rainfall2) was also included in the model. The time trend variable was included in the model to capture all other temporal effects such as changes in inputs and technology. The results in Table 8 are consistent with our expectations. During the period 196994, rainfall had a significant effect on rice yield in all the meteorological subdivisions, except for

The meteorological subdivisions are defined based on rainfall and temperature such that there is uniform evapotranspiration area. These subdivisions do not correspond exactly to the states or agroclimatic zones defined by of Planning Commission. Therefore, yield and rainfall analyses from relevant districts were taken.
4

107

Table 8. Response of rice yield to rainfall. Time period/ Variablesa 1969-92 Mean yield (t ha-1) Intercept Trend Eastern Uttar Pradesh Eastern Madhya Pradesh Bihar Plains Bihar Plateau Gangetic West Bengal

Orissa

1.65 -1.55** (3.2) 0.072** (14.5) 0.005** (4.4) -2.12E-6** (3.9)

1.36 -1.88** (2.6) 0.029** (7.0) 0.005** (3.7) -1.52E-6** (3.2) 0.75** 1.17 -2.02* (2.15) 0.015 (1.4) 0.005** (2.93) -1.82E-6** (2.43) 0.70***

1.53 -0.24 (0.2) 0.027** (4.6) 0.003 (1.3) -13.8E-7 (1.3) 0.43** 1.35 3.09** (2.44) 0.021 (1.73) -0.004 (1.35) 1.82E-6 (1.25) 0.13 1.69 -2.95** (2.34) 0.033* (2.05) 0.008*** (3.07) -3.79E-6** (2.89) 0.52** 3.60***'

1.30 -1.12 (1.4) 0.009* (2.2) 0.003* (2.4) -15.0E-7* (2.0) 0.41* 1.24 -1.47 (1.48) -0.007 (0.73) 0.005** (2.54) -1.82E-6** (2.25) 0.47**

1.58 -1.68 (1.7) 0.022** (7.2) 0.005* (2.5) -14.24E-7* (2.0) 0.78** 1.33 -0.73 (0.38) 0.013 (1.34) 0.003 (0.78) -9.24E-7 (0.52) 0.40* 1.80 -2.73* (1.89) 0.056*** (3.36) 0.006" (2.32) -2.12E-6* (2.06) 0.69*** 1.45

2.30 -4.05* (2.0) 0.074** (12.8) 0.009** (2.7) -3.45E-6** (2.6) 0.89** 1.82 -0.29 (0.68) 0.04*** (3.56) 0.005 (1.51) -1.97E-6 (1.52) 0.49** 2.80 -4.42 (1.51) 0.115*** (7.71) 0.008 (1.55) -3.18E-6 (1.44) 0.88*** 4.75***

Rainfall (June to Sep, in mm) Rainfall2

Adjusted R2 1969-81 Mean yield (t ha-1) Intercept Trend

0.91**

1.15 -1.04*** (3.09) 0.041*** (5.05) 0.004*** (5.65) -1.82E-6*** (5.01) 0.81*** 2.15 0.1 8 (0.06) 0.12*** (9.85) -0.002 (0.24) 1.82E-6 (0.37)

Rainfall (June to Sep, in mm) Rainfall2

Adjusted

R2

1982-92 Mean yield (t ha-1) intercept Trend

1.56 -0.67 (0.34) 0.043*** (3.20) 0.002 (0.63) -6.52E-7 (0.53) 0.49** 0.81

1.34 -0.74 (0.42) 0.021 (1.15) 0.003 (0.69) -9.85E-7 (0.50) 0.28 0.64

Rainfall (June to Sep, in mm) Rainfall2 Adjusted R2

0.90*** 8.50***

Chow F (22. 18) valueb
a

Dependent variable: yield (t ha-1). Figures in parentheses are t values. ***, **, * = significant at 1%. 5%. and 10% level, respectively. bFor the test of significance of the response function in the two periods (1969-81 and 1982-92).

the Bihar Plains. The first rainfall variable had positive and significant effect on yield, whereas the second variable had a negative and significant coefficient. In the Bihar Plains, the coefficients of rainfall variables had the expected sign, but they were nonsignificant. An estimation of the equation by time period reveals 108

that rainfall had a significant effect on rice yield in Bihar and Orissa in the second period, but the effect became nonsignificant in Uttar Pradesh, Madhya Pradesh, and the Bihar Plateau. However, the change in responsiveness of yield to rainfall was statistically significant in Uttar Pradesh and the Bihar Plains, which is expected

in view of the change in the CV of rainfall in both regions and the expansion of irrigation during the second period in eastern Uttar Pradesh. To estimate the area's responsiveness to rainfall, rice area was also regressed on time trend and rainfall variables. However, the effect was significant only in the Bihar Plains and West Bengal (equations 1 and 2). The differences in area response to rainfall require further analysis. Bihar Plains
RAt = 1,004.14 - 2.067 T + 4.695** Rt0.00203* R2t; (1.06) (0.38) (2.39) (2.02) (1) Adjusted R2=0.35***

West Bengal RAt= 124.75 +28.727***T+6.588***Rt-0.0027***R2t; (0.09) (7.41) (2.95) (2.96) 2= 0.78*** (2) Adjusted R

where RAt is rice area in year t (in 000 ha) Rt is rainfall in year t (in mm), and T is the time trend variable. Numbers in parentheses are t values and ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

attributes of technology adoption and farmers' adjustment behavior. Irrigated area does not indicate the reliability of irrigation water and the statistics in area under HYVs tends to be less reliable. The use of NPK not only captures these effects but also shows how much control farmers have on input use and crop management practices. Second, NPK use also represents indirectly the effect of infrastructure facilities necessary for the delivery of farm inputs. Third, NPK use data are more consistent than HYVs or irrigated area. The only limitation of NPK use is that it does not pertain to a rice crop, but to all crops grown in the district. In the absence of crop fertilizer-use data, this is an approximation. One significant omission in this analysis is of rainfall variability because the CV computed for the meteorological subdivision level rainfall does not capture interdistrict differences in rainfall variability, making this variable redundant. The dummy variables, however, capture part of the rainfall effect. The OLS estimates of the model are given in Table 9. The coefficients of the zonal dummy
Table 9. Results of yield variability model. a Explanatory variable Constant Mean NPK use (kg ha-1) Dummy variables West Bengal Gangetic Plains Bihar Plains Eastern Plains Eastern Plateau and Highlands (combined) Eastern Coastal Plains Adjusted R2 Dependent variable Annual percentage Mean yield deviation from trend in yield (t ha-1) 2.12*** 22.82*** (15.32) (11.68) 0.008*** -0.08*** (3.35) (3.69) 0.52** (3.18) -0.80*** (4.81) -0.44** (2.23) -0.67*** (3.65) 0.06 (0.22) 0.69*** 0.13*** 0.30*** 21.93*** (16.18) -0.07*** (2.99) -4.37* (1.78) 7.82*** (3.33) 18.20** (5.98) -0.04 (1.39) -2.96 (1.10) 9.88*** (3.53) 4.25 (1.31) 3.54 (1.16) 4.04 (0.95) 0.29***

Yield variability model
Differences in rice yield variability across districts are examined to identify factors contributing to these differences. The analysis was done for the second time period as the current level of variability is of greater practical significance. Annual percentage deviation in yield from the trend was regressed on the level of HYV adoption, irrigation, other infrastructure facilities, share of boro rice in total rice area, and dummy variables for agroclimatic zones to capture zone-specific effects. All coefficients in the full model were statistically insignificant, possibly due to multicollinearity. Variables such as percentage of rice area under HYVs, gross irrigated area as a percentage of gross cropped area, per hectare use of NPK, etc., are highly correlated, indicating high complementarity among the uses of these inputs. When these variables were used one at a time, coefficient estimates were negative and significant. We opted for NPK use (kg ha") for three reasons. First, the use of NPK also represents the quality

aEastern Uttar Pradesh is the benchmark region for dummy variables. The Chow test F ratio for the equations in columns 3 and 4 was 9.2 at (2,66) degrees of freedom, which IS significant at the 1% level. ***, * = significant at 1% and 10% significance level, respectively; n = 70.

109

variables have to be seen with Uttar Pradesh as a basis for which no dummy variable was used. The model has reasonably explained the interdistrict differences in annual percentage deviation from the trend in rice yield during 1982-94. The dummy variable for Gangetic West Bengal had a negative and statistically significant coefficient, whereas the coefficient was positive and statistically significant for the plains of Bihar. This showed that in comparison to Uttar Pradesh, yield variability is lower in West Bengal and higher in the Bihar Plains. The quantity of NPK had a negative and highly significant coefficient, indicating that the adoption of modem varieties and associated crop management practices helped achieve higher and more stable yield. When NPK use alone was retained in the model, it explained 13% of the variation in yield. However, the Chow test indicated that both models were statistically different, and therefore both dummy variables captured a significant effect of yield variability. In the final iteration, when all the zonal dummy variables were included in the model, all effects were captured by these dummy variables and NPK use became nonsignificant. Sensitivity of yield variability to selection of time period Sometimes, yield variability can be sensitive to the selection of time period, particularly when shortfall years are close together. The temporal pattern of yield variability and its underlying causes thus obtained could be misleading. We have therefore tested this sensitivity by computing a 5-year moving CV of rice yield and examined the significance of the trend. Then, we tested the sensitivity of annual percentage deviation from the trend in yield by changing the subperiods. In particular, the test shows how sensitive the results are to changes in cutoff year for the second period. The moving CV, like the moving average, was computed around the mean for successive 5year periods by including one year and excluding the first year. For example, the first moving CV was computed for 1969-73 and is referred to as 1973. In the next, the CV was was included and computed for 1970-74-1974 110

1969 was excluded. and so on. The moving CV of rice yield plotted against time for different agroclimatic zones (Fig. 3A,B) confirmed the results discussed above. The variability in rice yield maintained an increasing trend until the mid-1980s and declined sharply thereafter in the zones of the lower and northwest Middle Gangetic Plains, Eastern Plains, and Plateau. Variability was markedly high in the northwest Middle Gangetic Plains and the Eastern Plains during the late 1970s to the early 1980s. A similar increasing pattern was also observed for the eastern region as a whole, primarily because these zones contribute nearly 75% of rice production in eastern India. The yield in the coastal zone remained quite unstable in the second period also. But, in the northeast Middle Gangetic Plains, rice yield, after showing some degree of stability in the late 1980s, became more unstable in the early 1990s because of deficit rainfall in three consecutive years since 1991. To test the significance of the changing pattern of rice yield variability in eastern India, we fitted a quadratic trend line for the moving CV of rice yield. and trend coefficients were found to be highly significant (equation 3). This clearly indicates that rice yield has become more stable in recent years in a large part of eastern India. MCVt = 0.50 + 2.43 T*** - 0. 11 T2***; (6.07) (6.31) Adjusted R2 = 0.64***

(3)

where MCV, is the moving CV of rice yield for the 5-year period ending in year t, T is the time variable, and *** indicates significance at the 1 % level. Numbers in parentheses are t values. Growth rate and annual percentage deviation from the trend in yield were recomputed by changing the cutoff year for the second time period: (1) 1982-83 was included in the first period and the second period started from 198384, and (2) the second period started from 198182. The change in cutoff point did not alter the results (Table 10) and the conclusions drawn were still valid. The only difference in the results was that inclusion of 1982-83 in the first period gave a significant decline in the variability of rice yield in the Lower Gangetic

Plains and the increase in variability became nonsignificant in the coastal zone. Rice yield variability continued to show a significant decrease during the second period in the northwest alluvial part of the Middle Gangetic Plains and the Eastern Plains and Highlands.

Conclusions and policy implications
This study has clearly shown that growth in rice yield and production in eastern India has accelerated since the early 1980s. Unlike in the northern irrigated areas, yield growth mainly contributed to growth in rice production in the eastern region. Owing to differences in

agroclimatic factors and technology adoption, the patterns of yield growth and variability are rather diverse. Nearly 60% of the total rice area in the eastern region having high and assured rainfall or rapid expansion in irrigated area has shown higher growth with a greater degree of stability in rice yield. About 15% of the area has not shown any change in either growth or variability of yield, and one-fourth of the area has a tendency to increase yield variability with a moderate yield gain. The adoption of HYVs primarily bred for irrigated areas in the erratic and low-rainfall areas with low input use might have increased yield variability. A negative and highly significant effect of NPK fertilizer use, a

111

Table 10. Sensitivity of yield variability to time period: Agroclimatic zone Lower Gangetic Plains Time period I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 Growth rate (%) of yield Period I 2.16*** 1.42** 2.79*** Period II 4.77*** 3.65*** 5.06*** Annual percentage deviation from trend in yield Period I 8.11 10.10 6.71 Period II 9.73 4.58 9.65 F ratio 1.43 (-)4.82*** 2.07

Middle Gangetic Plains Northwest Alluvial I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 1.37 1.73 0.64 1.37 0.65 1.88 5.11*** 4.70*** 5.17*** 1.97 0.79 2.30* 22.92 22.18 23.28 12.92 13.88 12.89 8.53 8.27 8.1 9 18.41 16.71 17.84 (-)7.21*** (-)7.18*** (-)8.0V** 2.03 1.45 1.92

Northeast Alluvial

Eastern Plateau and Hills Eastern Plains I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 I: 1969-81; II: 1982-94 I: 1969-82; II: 1983-94 I: 1969-80; II: 1981-94 -0.43 -0.51 -1.21 -0.39 -0.62 -0.96 -0.66 -1.30 -0.36 1.32 0.31 1.74' 0.82 0.93 0.44 3.66*** 2.79*** 3.53*** 3.24*** 2.68*** 2.89*** 2.27** 1.02 2.66*** 4.59*** 3.24** 4.60*** 3.86** 3.99*** 3.99*** 22.19 21.26 22.38 22.50 21.63 23.13 17.57 14.93 15.16 11.57 14.19 11.65 10.65 11.06 10.66 11.71 9.98 11.27 10.19 9.69 10.21 14.24 10.91 14.07 18.39 15.80 17.61 9.11 8.73 8.73 (-)3.58** (-)4.54*** (-)3.94** (-)4.87*** (-)4.98*** (-)5.13*** (-)1.07 (-)1.87 (-)1.16 2.52* 1.34 2.28* (-)1.36 (-)1.60 (-)1.49

Eastern Highlands

Plateau

East Coast Plains and Hills Eastern India

a

***, **, * = significant at 1%, 5%. and 10% level, respectively..

proxy for technology-related factors, indicates that yield variability declined in areas where the adoption of HYVs and the use of NPK has occurred. Although the level of relative variability in yield and production does not appear to be very high at the regional level, it is high at the district level and, perhaps, could be even higher at the farm level in absolute terms, adversely affecting farmers and the rural poor, The effect of a shortfall in yield would be far more serious in a year of drought or flood. Since rice is a major crop of the region, both ex ante and ex post measures to deal with a drastic fall in production are essential. Ex post management measures such as public distribution of food grains and employment generation programs are already operating, albeit with varying degrees of 112

success. These should be strengthened and made more pro-poor. In particular, regional allocation of food grains for public distribution should be congruent with the number of poor people in the region, and the distribution should target the rural poor (for details, see Pal et al 1993). Crop insurance is another option to cope with risk, but, in the absence of a detailed analysis, it is rather unrealistic to talk about its feasibility or effectiveness for rainfed production systems. A long-term strategy to check a sharp shortfall in production should incorporate appropriate measures in the development strategy. Product diversification toward highvalue horticultural crops in low-potential rainfed areas is often talked about to minimize risk and raise farm income. This option should be viewed from two considerations, besides assessing the

availability of postharvest product-handling facilities. First, scope for diversification in low and deepwater lands is limited, as rice is the most suitable crop for these environments. Second, one of the preconditions of diversification is that rice yield should be increased in irrigated areas to meet demand and to check the increase in rice prices. This calls for accelerating research efforts to push the rice yield potential up in irrigated areas. A more effective strategy would be to strengthen research and development efforts for the rainfed rice production systems. Undoubtedly, the benefits of crop improvement research for developing tolerance for biotic and abiotic stresses in varieties would be comparable with those achieved in irrigated areas. In addition, the impact of research on poverty alleviation would be much more visible in rainfed areas. Given the availability of rainwater and the unabated land degradation in rainfed areas, research on resource management deserves equal priority. Considering the diversity of the production system, research investment should be enhanced and the focus should be on on-farm or farmer participatory research. This calls for the strengthening of regional agricultural research stations in the region. Similarly, technology, input, and information delivery systems have to be more organized and responsive to meet the diverse needs of the farmers. Finally, development priorities such as public investment in land improvement including flood control and development of irrigation, rural institutions, and other infrastructure facilities should be commensurate with the problems and potential of the rainfed areas.

References
Ballabh V, Pandey S. 1999. Transitions in rice production systems in eastern India. Econ. Pol. Wkly. 34(13):A11-A16. Bhatia MS. 1999. Rural infrastructure and growth in agriculture. Econ. Pol. Wkly. 34(13):A43-A48. Chand R. 1999. Emerging crisis in Punjab agriculture: severity and options for future. Econ. Pol. Wkly. 34(13):A2-A16.

Cuddy JDA, Della Valle PA. 1978. Measuring the instability of time series data. Oxford Bull. Econ. Stat. 40(1):79-85. Dhawan BD. 1987. How stable is Indian irrigated agriculture. Econ. Pol. Wkly. A9396. Ghosh SP. 1991. Agro-climatic zone-specific research: Indian perspective under NARP. Indian Council of Agricultural Research, New Delhi, India. Hazell PBR. 1982. Instability in Indian foodgrain production. Research report 30, IFPRI, Washington, D.C. (USA): International Food Policy Research Institute. Naylor R, Falcon W, Zavaleta E. 1997. Variability and growth in grain yields, 195094: Does the record point to greater instability? Popul. Dev. Rev. 23(1):41-58. Pal S, Sirohi AS. 1989. Instability in Indian crop production: its magnitude and sources. Artha Vijnana 31(3):241-256. Pal S, Bahl DK, Mruthyunjaya. 1993. Government interventions in foodgrain markets: the case of India. Food Policy. Planning Commission. 1989. Agro-climatic regional planning: an overview. New Delhi, India. Rao CHH. 1994. Agricultural growth, rural poverty and environmental degradation in India. Delhi (India): Oxford University Press. Rao CHH, Ray SK, Subbarao K. 1988. Unstable agriculture and droughts. New Delhi (India): Vikas Publishing House. Ray SK. 1983. An empirical investigation of the nature and causes for growth and instability in Indian agriculture. Indian J. Agric. Econ. 38(4):459-474. Singh AJ, Byerlee D. 1990. Relative variability in wheat yields across countries and over time. J. Agric. Econ. 41:21-32.

113

Discussion
Question: If you include other variables like rainfall, radiation, area under HYVs, besides fertilizer use, it would further explain differences in yield variability. Answer: Yes, I agree. The explanatory power must increase as more variables that determine yield fluctuations are included. However, there are serious constraints on the availability and accuracy of these data at the district level. Question: The analysis of changes in the variance of production between the two time periods indicates increase in the variability. On the contrary, the CV measure used in the later part of the paper shows a different picture. Answer: Variance and CV are two different measures of variability (absolute and relative measures, respectively) which have different meanings for policy recommendations. In terms of variance, the variability has increased over time. But for cross-sectional comparison, relative measure, i.e. CV, is more appropriate, which shows a diversity of patterns. For the national authorities in charge of managing food supplies, the CV may be more relevant as it is a dimensionless measure.

Appendix 1. list of districts included in the study by region. Northern Punjab Amritsar Ferozpur Gurdaspur Kapurthala Jallandhar Ludhiana Patiala Sangrur Haryana Karnal Ambala Eastern region region Bihar Bhagalpur Champaran Darbhanga Dhanbad Gaya Hazaribagh Monghyr Muzaffarpur Palamau Patna Purnea Ranchi Saharsa Santhal Pargana Saran Sahabad Singhbhum Orissa Balasore Bolangir Cuttack Dhenkanal Ganjam Kalahandi Keonjhar Koraput Mayurbhanj Phulbani Puri Sambalpur Sundergarh West Bengal 24-Parganas Bankura Birbhum Burdwan Cooch-Behar Darjeeling Hooghly Howarh Jalpaiguri Malda Midnapore Murshidabad Nadia Purulia West Dinajpur

Eastern Uttar Pradesh Allahabad Azamgarh Bahraich Ballia Faizabad Ghazipur Jaunpur Mirzapur Pratapgarh Sultanpur Varanasi Eastern Madhya Balaghat Bastar Bilaspur Durg Jabalpur Mandla Raigarh Raipur Seoni Shahdol Surguja Pradesh

Notes
Authors’ addresses: S. Pal, National Centre for Agricultural Economics and Policy Research, New Delhi, India; S. Pandey, and Abedullah, International Rice Research Institute, MCPO Box 3127, Makati City 1271, Philippines.

Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

114

Rainfed rice and risk-coping strategies: some microeconomic evidence from eastern Uttar Pradesh
S. Pandey, H.N. Singh, and R.A. Villano

Risk management is one important aspect of rainfed rice production activities. In rainfed environments, variability of rainfall can lead to a wide variability of output and production. Farmers in this environment have evolved several coping strategies to deal with risk. This paper examines variability in rainfed rice production and farmers coping mechanisms using farm-level panel data from eastern Uttar Pradesh. The results of this study showed that farmers use a range of strategies to cope with risk. These include changing the proportion of rice area, adjusting the area under traditional varieties, growing several rice varieties, and changing rice establishment methods. Crop and income diversification was found to be another important farmer strategy for dealing with risk. Through these measures, rainfed rice farmers in eastern Uttar Pradesh have been able to maintain their total income even in the face of a sharp decline in rice income. The implications of these results for technology design and policy reforms are discussed in this paper.

An intrinsic feature of agricultural production is its riskiness. The major sources of risk in agriculture are the unpredictable variations in yield and prices of inputs and outputs. Risk under rainfed conditions generally tends to be high as variability of rainfall can lead to wide swings in yield and output. Over time, farmers have developed various strategies to avoid the negative consequences of unpredictable variations in agricultural output. A good understanding of these strategies is required to assess the likely responses of farmers to new technologies or policies. The uptake of technologies that complement and reinforce farmers’ coping strategies is to be quite rapid. On the other hand, interventions that undermine the key components of risk management strategies are likely to be rejected. Coping strategies can be classified into ex ante or ex post, depending on whether they help reduce risk or reduce the impact of risk after a production shortfall has occurred. Because of a lack of efficient market-based mechanisms for diffusing risk, farmers modify their production practices to provide “self-insurance” such that

the chances of negative consequences are reduced to an acceptable level. Ex ante strategies are designed to provide such self-insurance, which, however, can be costly to farmers in terms of opportunities forgone. Ex ante strategies can be broadly grouped under two categories: those that reduce risk by diversification and those that do so by maintaining flexibility. Diversification is simply captured in the principle of “not putting all eggs in one basket.” Being involved in several economic activities that have negatively or weakly correlated returns reduces the risk of income shortfall. This principle is used in different types of diversification common in rural societies. Examples are spatial diversification of farms, diversification of agricultural enterprises, and diversification between farm and nonfarm activities. Maintaining flexibility is an adaptive strategy that allows farmers to switch resources between activities as the situation demands. Flexibility not only allows farmers to reduce the chances of low income but also helps to capture income-increasing opportunities when they do

115

arise. Flexibility is high when losses associated with reallocation of resources are low. Instead of committing all resources prior to the occurrence of uncertain events, flexibility is maintained by being able to adjust decisions to current values of uncertain events. Examples are using a split dose of fertilizer, temporal adjustment of input use to crop conditions, and expanding the area allocated to a crop when climatic conditions are more favorable to that crop. Ex post strategies are designed to prevent a shortfall in consumption when family income drops below what is necessary for maintaining consumption at its normal level. These include seasonal migration, consumption loans, asset liquidation, and charity. A consumption shortfall can occur despite these ex post strategies if the drop in income is substantial. The objective of this paper is to characterize risk in rice production and document farmers’ risk management strategies in eastern Uttar Pradesh. The analysis is based on panel data collected from a group of farmers from two adjacent villages that differ mainly in risk in rice production because of their differential access to irrigation. Three types of analyses are conducted. First, intervillage comparisons of production systems are made to assess the differential impact of risk on the broader characteristics of the village economy and the rice production systems. The land use patterns and the structure of the village economy can be expected to capture the long-term adjustments that may have occurred over rime in these villages. Some of these adjustments could at least be partly in response to risk. Within a village, changes in rice production practices and income strategies over time are also examined to document how farmers deal with uncertainty in rainfall. As the nature of risk responses may be conditional on economic status, risk responses are also examined by farm size category.

Study approach
Even though the potential impact of risk and risk aversion on farmer behavior is well recognized, insufficient attention has been given to characterizing the nature of risk faced by rainfed

rice farmers and their responses. The study of risk and risk management strategies requires temporal data as it is the fluctuation in production and income over time that is of concern to individual farmers. Such farm-level data, however, are generally not available. As a result, analysts are often forced to use crosssectional farm-level data or aggregated timeseries data. Although useful in some contexts, such analyses cannot adequately describe the nature of risk facing farmers and how they manage the risk. The study approach consists of a microlevel analysis of risk and its management by rainfed rice farmers using temporal data. To this end, the collection of panel data in eastern Uttar Pradesh started in 1994 in two villages of Faizabad District. The selected villages were Mungeshpur and Itgaon. These two villages differ in several aspects, the major difference being with respect to access to irrigation. Mungeshpur farmers have a greater access to irrigation than farmers in Itgaon. Since irrigation reduces risk associated with rice production, the production system of Mungeshpur serves as a reference point against which the effect of risk in Itgaon can be assessed. The two villages are only a few kilometers apart; hence, the effects of other variables such as differences in market access and climate on the nature of production systems are likely to be minimal. Farmers were first stratified by farm size category into small (up to 1 ha), medium (between 1 and 2 ha), and large (more than 2 ha). The number of farmers selected randomly from each stratum corresponded to their proportion in the overall population. Altogether, 30 farmers from each village were selected and detailed data on rice production practices and other farm and nonfarm activities were collected using the interview method. The sample size was increased from 30 to 45 in 1996. Most of the results presented below pertain to the 5 years spanning the period 1994-98. Table 1 summarizes the major characteristics of the study villages. The average operational holding is 50% smaller in Mungeshpur than in Itgaon. Farmers in Itgaon also hold proportionately more upland fields, which are

116

Table 1. Village characteristics. Mungeshpur Characteristic Small farmers 36 93 0.5 38 6 57 5 7 Medium farmers 7 74 1.5 44 16 40 8 9 Large farmers 2 75 2.2 24 43 33 12 7 All farms Small farmers 22 34 0.5 59 15 26 3 7 ltgaon Medium farmers 12 35 1.4 60 21 20 8 8 Large farmers 11 39 2.9 48 34 18 9 9 All farms

Number of households Irrigated area (%) Av operational holding (ha) Proportion of land type (%) Upland Medium land Lowland Av years of schooling of household head Av household size (no. of members) "

45 85 0.7 34 14 52 5 7

45 37 1.4 55 24 21 6 8

Table 2. Kharif-season rainfall for Kumarganj, Fairabad, 199498. June Year No. of rainy days 8 8 11 4 4 Rainfall (mm) 145 109 99 121 29 No. of rainy days 15 7 5 19 19 July Rainfall (mm) 224 60 133 485 337 August No. of rainy days 16 23 18 15 16 Rainfall (mm) 341 355 337 264 276 September No. of rainy days 6 10 14 12 6 Rainfall (mm) 153 395 168 294 110 October No. of rainy days 1 0 3 2 2 Rainfall (mm) Total No. of rainy days 46 48 51 52 47 Rainfall (mm) 865 919 902 1,207 799

1994 1995 1996 1997 1998

2 0 165 43 47

Central Research Station, Masodha. Faizabad (1968-98) Long-term mean rainfall (mm) 145 314 CV of long-term rainfall (%) 87 45

300

141

74

1,088

40

56

115

22

fields in the upper part of the toposequence. Since the hydraulic gradient decreases and the clay content of the soil generally increases with a move down the toposequence, upland fields tend to be relatively drought-prone, whereas lowland fields tend to be submergence-prone. The classification of field types was based purely on farmers' classifications. Farmer classification of field characteristics has been found to correspond well with scientific classification (Talawar 1996), indicating that this is a rapid and cost-effective method of characterizing fields. In Mungeshpur, nearly 85% of the fields have access to irrigation, whereas less than 40% of the area in Itgaon is irrigated. In addition, the reliability of irrigation also differs between these two villages as Mungeshpur has more private tubewells, which supply water with a greater reliability than public tubewells that are the major source of irrigation in Itgaon. Except for farm size, extent

and nature of irrigation, and composition of land types, other characteristics are similar in these two villages.

Rainfall analysis
The average total rainfall based on weather records from Masodha (about 25 km from the study villages) for 1968-98 for June to December is 1,088 mm with a CV of 22% (Table 2). The rice production environment of the region can be considered to be drought-prone, especially in October, when rice may be in the critical growth stage of heading (or grain fill). Rainfall at the beginning of the season (i.e., June) is also characterized by a high degree of uncertainty as attested by its high CV. The analysis of rainfall data indicates that farmers in these villages have to deal with the uncertainty of moisture conditions at both the initial stage of land preparation and rice 117

establishment (June-July) and during the later stage of crop development (October). While the uncertainty of rains during the start of monsoon may encourage farmers to wait for land preparation and planting until sufficient moisture has been accumulated, such a strategy has to be balanced against the possibility of the crop suffering from a terminal drought if planting is too late. In other words, the rainfall pattern would dictate a strategy of rapid early planting as soon as conditions become favorable so that terminal drought could be avoided. Under predominantly rainfed conditions of Itgaon, the optimal planting window will therefore be narrow. In Mungeshpur, this constraint may be less severe as farmers can establish rice earlier using irrigation even if rains are late. Figure I presents the rainfall data from a monitoring station closer to the surveyed villages for 1994-98. Compared with 1994 and 1997, rainfall in June and July was much lower in 1995 and 1996. Of the two drought years, 1995 and 1996, the early-season (June and July) drought in 1995 appears to be more serious than in 1996. In 1998, rain began relatively late and

terminated quite early. Judging by the distribution and amount of rainfall, 1994 and 1998 may be classified as “normal” years, 1995 and 1996 as “drought” years, with drought being more severe in 1995, and 1997 as a “good” year. Farmers’ responses to these temporal differences in rainfall patterns are explored later in the paper.

Intervillage comparison
Spatial diversification of fields
Scattering of fields over different landscape positions can help reduce the overall production risk if yields across field types are negatively or poorly correlated. Scattering of fields has been an important mechanism used by farmers, especially in the semiarid areas (Walker and Jodha 1986). Most farmers in both villages have their fields distributed over all land types (upland, medium, and lowland). In terms of fragmentation, nearly 80% of the farmers in Itgaon have their fields scattered in three or

118

more parcels, whereas the corresponding figure for Mungeshpur is only 32% (Fig. 2). Land is thus more scattered in Itgaon than in Mungeshpur. It is, however, not clear whether greater fragmentation in Itgaon is the result of its higher level of production risk. The extent of land fragmentation depends on many other factors such as farm size, inheritance law, and land reform regulations. Given the proximity of these villages and their similarity in social structure, the effect of inheritance law and land reform regulations is likely to be similar. A greater fragmentation in Itgaon is probably a function of its larger average farm size.

ranges between zero and unity and is calculated as (1) DIi = 1 - S (a ij /Saij)2
j j

where aij is the area under the jth crop in the ith village (or season). The diversification indices (Table 3) indicate that cropping pattern in Itgaon is more diversified than in Mungeshpur for both the kharif and rabi seasons. In Itgaon, farmers grow rice in a proportionately smaller area than in Mungeshpur and rely more on pulses and maize, which grow well under rainfed conditions. Mixed cropping and intercropping

Diversification of cropping pattern
Diversification of cropping pattern is a way of reducing risk. If crop yields are poorly correlated, increasing the number of crops can reduce the variability of total output. We have used the Simpson diversification index (DI) to measure the extent of diversification. The index

Table 3. Crop diversification index. Year 1994 1995 1996 1997 1998 Mungeshpur Kharif 0.47 0.53 0.58 0.50 0.45 Rabi 0.56 0.55 0.57 0.53 0.53 Kharif 0.72 0.90 0.82 0.75 0.78 ltgaon Rabi 0.71 0.77 0.67 0.68 0.62

119

Table 4. Percentage of area by variety. Farm size Marginal Medium Large All farms Variety Mungeshpur 1994 84 16 81 19 95 5 89 11 1995 81 19 88 12 97 3 92 8 1996 93 7 92 8 93 7 93 7 1997 94 6 90 10 90 10 91 9 1998 100 0 97 3 96 4 97 3 Av 91 9 90 10 94 6 93 7 1994 31 69 61 39 77 23 67 33 1995 60 40 41 59 73 27 64 36 ltgaon 1996 38 62 42 58 62 38 53 47 1997 33 67 40 60 73 27 61 39 1998 32 68 67 33 78 22 67 33 Av 37 63 51 49 73 27 62 38

Modern Traditional Modern Traditional Modern Traditional Modem Traditional

Table 5. Varietal composition (%) by farm size and duration. Farm size Small Durationa Short Medium Long Short Medium Long Short Medium Long Short Medium Long Mungeshpur 1994 27 22 51 36 26 37 19 31 50 26 28 46 1995 30 34 36 44 40 17 15 14 70 25 24 51 1996 34 48 19 26 57 17 30 36 35 30 44 27 1997 19 67 14 27 70 3 33 47 20 30 56 15 1998 20 64 16 12 88 1 23 55 22 21 63 17 Av 26 47 27 28 58 14 25 40 35 26 46 28 1994 79 5 16 62 20 18 52 31 18 58 25 18 1995 88 12 0 53 6 41 58 24 18 66 17 16 ltgaon 1996 79 21 0 71 28 2 40 45 14 55 36 9 1997 92 6 2 85 15 0 58 41 2 67 31 2 1998 73 21 5 87 11 2 46 54 0 61 37 1 Av 81 14 5 77 17 6 51 41 9 61 31 8

Medium

Large

All farms

a

Short = <120 d. medium = 120-135 d. long = >135 d.

are also more common in this village. The average number of intercrops over the 5 years in the kharif season in Itgaon was 19, while the corresponding figure for Mungeshpur was only 11. In terms of land type, upland fields generally have a more diversified cropping pattern than medium and lowland fields as upland fields are less prone to submergence. Crop diversification may occur for several reasons, in addition to risk reduction (Smale et a1 1994). Differences in land quality can lead to crop diversification even if yields are deterministic. For crop diversification to be an effective strategy for risk reduction, the temporal correlation of yield between different crops must be low. We have not examined yield correlation as the precision of correlation estimated on the basis of 5 years of data is likely to be low. Furthermore, yield data for several intercrops occupying small areas are likely to be imprecise. Instead of yield correlation, we later assess how the variability of crop income is affected by crop diversification. 120

Rice varieties In Mungeshpur, about 93% of the total rice area is planted to modern varieties. The adoption of modem varieties in Itgaon is lower and varies between 37% and 73% (Table 4). In Itgaon, the adoption of modem varieties is correlated with farm size, whereas adoption in Mungeshpur is high, irrespective of farm size. Short-duration varieties occupy proportionately more area in Itgaon than in Mungeshpur (Table 5). The lower rate of adoption of modem varieties and the greater importance of short-duration varieties in Itgaon are probably adaptations to the more moisture-deficient environment in this village. To the extent that yields of different rice varieties are poorly correlated, growing of several varieties can reduce the overall risk. To examine whether there are any systematic differences in varietal diversity of rice, we calculated the diversity index using equation (1) where a,, in this case refers to area planted for the jth variety in the ith village. The varietal

diversity index for both villages averaged 0.85, indicating that diversification into several varieties as a way to reduce risk is not necessarily more important in Itgaon than in Mungeshpur. Rice establishment methods Farmers in the study villages practice all three methods of crop establishment—dry seeding, wet seeding, and transplanting. Dry seeding consists of broadcasting rice seeds in dry soils. Wet seeding consists of broadcasting pregerminated rice seedlings in puddled soils. Transplanting consists of pulling the seedlings established in the seedbed and replanting them in the main field. As the field's hydrological conditions that are optimally suited for these three methods are different, farmers can be expected to switch from one method to another as well as expand or decrease the area under a specific method in response to climatic conditions. Under purely rainfed conditions, preparation of the seedbed and sowing of seeds cannot be done without sufficient rain. Similarly, preparation of the main field for transplanting requires a sufficient amount of rain. Farmers who do not have access to irrigation have to wait until sufficient rains have been received to prepare the seedbed. If the rainfall is higher than expected, the transplanted area may be expanded if sufficient seedlings are available (or by

purchasing seedlings from those who may have excess seedlings). However, if the rainfall is late and/or lower than expected so that transplanting is not possible, the field may be dry- or wetseeded. Transplanting is by far the most dominant method of crop establishment in Mungeshpur (Table 6). Availability of supplemental irrigation has made transplanting in more than 75% of the area feasible in Mungeshpur, even in drought years such as 1995 and 1996. In Itgaon, however, even in the most favorable rainfall condition of 1997, the area transplanted was only 53%. The favorable rainfall in 1998 allowed farmers to increase the area transplanted. Farmers often use more than one method of crop establishment within a season for a variety of reasons such as risk reduction, staggering of labor demand, and better matching with field conditions. In Mungeshpur, about 60% of the farmers in each of the years practiced only one method in all of their fields (Table 7). The remaining used a combination of methods, with transplanting being the dominant one. A relatively favorable environmental condition in Mungeshpur probably led to the dominance of the use of the single most profitable method. One would expect more farmers in Itgaon to establish their rice crop by a combination of methods. This expectation is supported by the data.

Table 6. Percentage of area, by crop establishment method and farm size. Farm size Small Crop establishment methoda DSR TPR WSR DSR TPR WSR DSR TPR WSR DSR TPR WSR Mungeshpur 1994 12 94 6 5 65 30 7 78 15 5 78 17 1995 12 70 18 1 70 28 79 21 2 76 23 1996 1997 1998 Av 4 90 5 3 81 16 1 87 12 2 86 12 1994 46 20 34 16 37 48 11 27 62 17 27 55 1995 44 41 14 94 0 6 47 49 4 54 39 7 ltgaon 1996 52 36 12 43 46 10 35 55 10 40 49 11 1997 9 15 76 0 42 58 3 64 33 3 53 44 1998 10 58 32 4 64 32 8 76 16 8 70 23 Av 33 35 33 17 46 37 15 55 31 18 49 32

88 11 75 14

92 8 90 10 85 15

98 2 99 1

Medium

Large

91 8 5 86 8

93 7

All farms

88 12

95 5

a

DSR = dry-seeded rice, WSR = wet-seeded rice, TPR = transplanted rice.

121

Table 7. Percentage of rice farmers using different methods of crop establishment. Farm size Small No. of methods used One Two Three One Two Three One Two Three One Two Three Mungeshpur 1994 73 27 57 29 14 38 50 12 60 33 7 1995 80 20 55 44 73 27 70 30 1996 100 1997 93 7 85 15 28 72 64 36 1998 93 7 90 10 68 32 82 18 Av 89 11 73 25 2 54 42 4 72 27 2 1994 51 50 17 83 25 75 33 66 1995 90 10 50 50 80 20 a3 18 ltgaon 1996 84 15 44 33 22 50 42 8 65 27 8 1997 80 20 21 79 37 58 5 42 57 2 1998 84 15 46 54 54 37 9 62 37 3 Av 78 22 34 61 5 40 46 5 54 44 3

Medium

70 31 64 29 7 79 17 2

Large

All farms

Variability of yield and net returns
The average yield and variability of rice were estimated by pooling the cross-sectional and time-series data. Such data capture two sources of variability. Yield variability from plot to plot within a year is due mainly to differences in soil type and management practices. For a given plot, yields vary over time because of fluctuations in climatic events and changes in management practices. Variability over time is the main concern of farmers, as it influences their decisions regarding rice production. Thus, it is essential to separate out temporal variability from spatial variability. Although a range of methods are available to achieve this (Judge et al 1985), we used the simple dummy variable model by assuming that the farmer and land types have fixed effects. Variability was
Table 8. Measures of probability distribution of plot-level rice yield, net returns, and nitrogen use in rice, 1994-98. Item Yield Mean (t ha-1) CV (%) Skewness Gross returns Mean (Rs ha-1) CV (%) Skewness Net returns Mean (Rs ha-1) CV (%) Skewness Nitrogen Mean (kg ha-1) CV (%) Skewness Mungeshpur 2.2 52 2.86 6.592 54 3.1 8 5.343 67 2.55 35 60 0.53 ltgaon 1.7 60 0.8 4.426 62 0.92 3.250 86 1.00 30 74 0.9

estimated after purging the data of these fixed effects. The average rice yield in Mungeshpur is about 0.5 t ha-1 more than in Itgaon (Table 8). The variability in rice yield, as measured by the CV, however, is lower. A higher mean and a lower CV indicate a more favorable environmental condition for rice production in Mungeshpur. Farmers grow modem varieties in a bigger area and use more fertilizer in Mungeshpur than in Itgaon. Lower risk because of greater availability of irrigation may have encouraged the adoption of modern varieties and fertilizers in Mungeshpur. The average net return from rice in Itgaon is almost half its value in Mungeshpur. The CVs of net return in Mungeshpur (67%) and Itgaon (86%) suggest high profit variability associated with rice production in these villages. The coefficient of skewness of yield, gross returns, and net returns indicates an approximately symmetrical probability distribution. The approximate normality of the distributions of yield and net returns indicates that risk analysis based on mean and variance may be sufficient for evaluating risk effects. The CVs of nitrogen application from chemical fertilizers are similar to those of net returns. The coefficient of skewness is higher in Itgaon, indicating a more skewed use of fertilizer in this village. A positive skewness indicates that only a few farmers are using larger quantities of fertilizer, while the majority use very little. Small and medium farmers may face

122

more parcels, whereas the corresponding figure for Mungeshpur is only 32% (Fig. 2). Land is thus more scattered in Itgaon than in Mungeshpur. It is, however, not clear whether greater fragmentation in Itgaon is the result of its higher level of production risk. The extent of land fragmentation depends on many other factors such as farm size, inheritance law, and land reform regulations. Given the proximity of these villages and their similarity in social structure, the effect of inheritance law and land reform regulations is likely to be similar. A greater fragmentation in Itgaon is probably a function of its larger average farm size.

ranges between zero and unity and is calculated as (1) DIi = 1 - S (a ij /Saij)2
j j

where aij is the area under the jth crop in the ith village (or season). The diversification indices (Table 3) indicate that cropping pattern in Itgaon is more diversified than in Mungeshpur for both the kharif and rabi seasons. In Itgaon, farmers grow rice in a proportionately smaller area than in Mungeshpur and rely more on pulses and maize, which grow well under rainfed conditions. Mixed cropping and intercropping

Diversification of cropping pattern
Diversification of cropping pattern is a way of reducing risk. If crop yields are poorly correlated, increasing the number of crops can reduce the variability of total output. We have used the Simpson diversification index (DI) to measure the extent of diversification. The index

Table 3. Crop diversification index. Year 1994 1995 1996 1997 1998 Mungeshpur Kharif 0.47 0.53 0.58 0.50 0.45 Rabi 0.56 0.55 0.57 0.53 0.53 Kharif 0.72 0.90 0.82 0.75 0.78 ltgaon Rabi 0.71 0.77 0.67 0.68 0.62

119

Fig. 3. Relationship between the coefficient of variation (CV) of rice income and the CV of kharif income.

124

125

Table 9. Different sources of household income. Source Mungeshpur 1994 1995 1996 19.3 11 26 18 3 1 41 0 100 1997 15.9 20 17 13 3 1 45 0 100 1998 16.4 22 20 17 3 1 35 1 100 AV 17.4 15 21 19 4 2 38 1 100 1994 25.1 3 14 28 5 1 48 1 100 1995 21.6 1 18 24 1 1 55 0 100 1996 26.8 4 23 17 3 1 52 1 100 ltgaon 1997 21.0 13 21 14 7 1 43 1 100 1998 20.2 6 20 24 0 0 50 0 100 Av 22.9 5 19 21 3 1 50 0 100

Total household income 18.8 16.6 (000 Rs household-1) % Share Rice 15 9 17 Wheata 23 28 Other cropsb 19 4 Livestock 9 2 Farm labor 5 Nonfarm activitiesc 35 35 2 0 Othersd Total 100 100
a

Includes the value of intercrops such as mustard. blncludes other rabi and zaid (third-season) crops. “Includes services, laborer, business, barber, blacksmith, and pensions. dlncludes the sale of fruits and timber.

important in Itgaon than in Mungeshpur as indicated by the relatively larger number of scatter points below the 1:1 line.

Tactical adjustments
This section focuses on how farmers responded to different patterns of rainfall experienced in the survey years. We examine the responses mostly in the context of rice production. Comparisons are made for a village over time as well as between the two villages. We analyze the response in terms of area of rice planted, method of rice establishment, and changes in rice varieties.

Area of rice planted
In both Mungeshpur and Itgaon, rice area decreased in 1995 and I996 compared with 1994 and 1997 (Table 10). The decrease in rice area in Itgaon is much sharper than in Mungeshpur. Delayed rainfall in 1995 forced many farmers in Itgaon to forgo rice completely, whereas farmers in Mungeshpur were able to maintain the rice area by using irrigation. Although some of the rice area was diverted to less moisture-sensitive crops such as pulses, maize, and various types of intercrop, most of the rice land was left fallow in Itgaon. The area under fallow declined in 1996 when the early-season drought was less severe. In 1997, when the rainy season was favorable, the rice area increased beyond its share in 1994.

The relative importance of area under various crops and fallow during 1994-98 indicates that the initial response to the earlyseason drought is to reduce the area under rice and allocate the area to less moisture-sensitive crops. Most of this type of adjustment occurs in the upland and medium land fields. When the early-season drought is more severe, farmers fallow the land completely, especially the lowlands that have poor access to irrigation. Unlike with the upland fields, alternative options in these lowland fields are very limited as poor drainage of these fields results in waterlogging once the rains commence. Large fluctuations in area planted to rice in rainfed environments can be a major source of variability in rice production unless there are corresponding offsetting movements in yields. The relative contribution of variations in area and yield to the total production variance was examined using village-level data for 1994-98 (Table 11). In Itgaon, the variability in area accounts for about 42% of the total production variation. In addition, about a quarter of the production variance can be attributed to the positive correlation between yield and area. Thus, in drought years, rice output declines not only because of a reduction in area but also because of the lower yield in the area planted. These effects are much smaller in Mungeshpur where planting decisions are less dependent on rainfall. In addition to technologies for greater yield stability, the high risk of failure to plant in

126

127

Table 10. Percentage of area under different crops. Cropping pattern 1994 Kharif Fallow Rice Fodder Sugarcane Maize Othersa Rabi Fallow Wheat + oilseeds Pulses + oilseeds Sugarcane Vegetables Pulses Othersb Zaid Fallow Pulses Sugarcane Othersc
a c

Mungeshpur 1995 20 53 4 13 2 8 1996 15 54 6 10 4 11 1997 16 59 6 8 3 8 1998 12 65 7 7 1 8 Av 16 58 6 9 2 8 1994 33 32 7 2 6 19 18 40 22 2 5 4 8 1995 1996

ltgaon 1997 1998 Av

17 60 7 6 2 7

48 10 5 7 7 22

33 24 6 5 8 24

27 35 5 3 10 21

29 26 6 5 8 26

34 25 6 4 8 22

12 58 11 8 3 0 7

8 52 11 14 2 2 10

9 56 9 9 7 6 4

15 60 8 1 5 5 6 99 0 0 1

5 63 12 7 6 5 2

10 58 10 8 5 4 6 85 6 7 2
b

22 41 17 6 5 5 4

24 46 17 0 4 7 2 100 0 0 0

18 46 15 2 5 8 5 100 0 0 0

7 52 20 2 7 5 7

18 45 18 2 5 6 5

80 8 8 5

77

7 13 2

82 7 9 2

85 6 7 2

97 1 2 0

92 1 7 1

99 1 0 0

98 1 2 0

Includes sweet potato, cereal and pulses. pigeonpea and mixed cereals. Includes vegetables, fodder, pulse + oilseed. cereal + oilseeds.

lncludes pulses, fodder, vegetables, oilseeds, vegetables + oilseeds.

Table 11. Decomposition of rice production variance, 199498. Source Area variance (%) Yield variance (%) Area and yield covariance (%) Residual (%) Mungeshpur 16 52 23 9 ltgaon 42 26 25 7

Method of rice establishment As seasonal conditions vary from year to year, farmers respond by adjusting their crop establishment methods. A farmer who transplanted all rice fields in one year may choose to wet-seed the following year if the field hydrology is suboptimal for transplanting. Similarly, a farmer who dry-seeded all rice fields in one year may subsequentIy wet-seed one part and transplant the other. In addition to seasonal conditions, variations in other factors such as labor supply, working capital, and pest problems also play a role in the decision-making process. Temporal variations in each of the crop establishment methods are higher in Itgaon than in Mungeshpur. Neary 20% of the farmers in Mungeshpur used the transplanting method in all their fields during 5 years. In contrast, none of the Itgaon farmers maintained the same method of crop establishment in a field over time. In Mungeshpur, farmers appear to be switching out of wet seeding to transplanting in more favorable years and vice versa in less favorable years. In Itgaon, the switch seems to occur

rainfed environments highlights the need to deal with that type of risk also. Research to reduce area variability and the correlation between area and yield has not been given as much attention as research to reduce yield variability. Technologies that permit late planting (such as short-duration varieties, tolerance for transplanting of older seedlings) can be useful in this regard. The diversification index also indicated that crop diversification in the kharif season in Itgaon was higher in the unfavorable years of 1995 and 1996 compared with 1994 and 1997 (Table 3). Farmers may have used a greater degree of crop diversification to reduce risk in these unfavorable years.

128

Table 12. Area (%) under different duration of varieties and crop establishment methods.a Village Mungeshpur ltgaon Duration 1994 DS+WS 53 10 37 63 14 24 TP 18 33 49 44 55 1 1995 DS+WS 40 12 49 84 5 11 TP 21 28 51 38 37 25 1996 DS+WS 65 25 10 79 6 16 TP 24 47 30 30 68 1 1997 DS+WS 78 13 9 98 1 1 TP 23 62 15 41 57 2 1998 DS+WS 59 39 2 93 7 0 TP 19 64 17 48 50 2 Av DS+WS 58 17 26 81 7 12 TP 21 51 29 41 56 3

Early Medium Long Early Medium Long

a

DS = dry seeding, WS = wet seeding, TP = transplanting.

between dry-seeded rice and wet-seeded rice (depending on whether the soil is wet or dry) and also between transplanted rice and wetseeded rice (depending on availability of seedlings and labor constraint). These adjustments in crop establishment methods are manifestations of flexibility that farmers exploit to deal with the uncertainty of the rainy season. Implications are that, if rainfed farmers were to rely on a single method, they would lose such flexibility in adapting to environmental conditions. Crop breeding programs can help maintain flexibility by developing varieties that are well adapted to different methods of establishment.

Varieties
The adoption of modem varieties (MVs) in Itgaon is not only lower than in Mungeshpur; the temporal variations in the proportion of area under MVs are also higher. The data for Itgaon indicate that, in years with low rainfall during the planting period ( 1995 and 1996) when the overall rice area was reduced, area under MVs decreased more than proportionately (Table 4). The share of traditional varieties (TVs) in rice area planted thus increased. When rice area expanded in 1997, the area under MVs increased more than proportionately. Thus, an important coping mechanism appears to be reliance on TVs in drought years. By switching to TVs, farmers may be able to cushion the effect of early-season drought to a certain extent. Another coping mechanism is increasing the area under short-duration varieties in years with poor rainfall to partially compensate for late crop establishment. If this is the case, the proportion of area under short-duration varieties

should be higher in 1995 and 1996 than in 1994 and 1997. The data show that this is only true in Itgaon (Table 5). This potential risk-coping mechanism thus appears to be unimportant in Mungeshpur. Overall, farmers probably rely more on adjusting crop establishment method to deal with early-season drought than on manipulating varietal duration. There is also an interactive effect between variety and crop establishment method. In response to early-season drought, farmers not only increase the area under TVs but they also establish a greater proportion by direct seeding. For direct-seeding culture, farmers prefer shorter duration varieties (Table 12). In Itgaon, the proportion of short-duration varieties increased in 1995 and 1996 compared with 1994, indicating that an adjustment in phenology could be an important mechanism for direct-seeded rice. For transplanted rice, the pattern of adjustment is not very clear.

Total household income
Changes in the proportion of income in Itgaon over the 5 years indicate that a decline in rice income in 1995 and 1996 relative to that in 1994 has been more than offset by an increase in income from wheat and other crops (Table 9). In fact, income from wheat increased over time because of expansion in area in both villages. Income from other crops appears to be negatively correlated with income from rice. Farmers may be dealing with the shortfall in rice production by expanding the area under other crops. As mentioned earlier, the area under maize, pulses, and other nonrice crops increased significantly in Itgaon in response to the 1995 drought. 129

Despite the shortfall in rice production in 1995, total household income has remained similar over the 5 years in each village. This indicates the ability of the households to efficiently cope with the risk associated with rice production. As rice contributes to only a small proportion of total income, variability in rice income does not affect variability of total income appreciably.

Farm size and production system
We examine the effect of farm size (or wealth status) in terms of three variables-land use pattern, extent of adoption of MVs, and income. These indicators are expected to partially capture the effects of risk that may differ by wealth status. In terms of cropping pattern, the major difference was that the proportionate area under rice decreased and that under fallow increased as farm size increased. Thus, judging by the area planted, the importance of rice to smaller farmers is greater. Areas under other crops were more or less similar. The responses to drought in 1995 and 1996 were comparable in that farmers in all categories reduced their rice area and increased their fallow area by approximately the same proportion. A major difference was found in terms of the extent of adoption of MVs. The area under MVs increased with farm size (Table 4). Large farmers in Mungeshpur planted MVs only. The differences in adoption rate for MVs, by farm size category, are clearer in Itgaon where large farmers had almost double the proportionate area under MVs compared with small farmers. In terms of temporal variations, large farmers maintained the proportionate area under MVs and TVs throughout the 5 years. Small farmers, on the other hand, shifted from MVs to TVs in response to drought conditions in 1995 and 1996. In 1997, they expanded the area under MVs in response to favorable rainfall. The reliance of small farmers on TVs for risk reduction is thus greater. Such a behavior probably reflects strategies that medium farmers use to protect themselves against risk, which will thus have positive welfare effects.

Characterization of these traditional varieties and identification of traits that impart adaptation to early-season drought would be important initial steps. The average level of income is positively correlated with farm size but, for a given size class, income levels are similar across the two villages (Table 13). The relative share of rice in total income increases with farm size and is substantially higher for all size classes in Mungeshpur than in Itgaon. Even though large farmers have proportionately less area under rice than small farmers, their share of rice income is higher due to the larger absolute area under rice as well as to the greater adoption of MVs. The share of nonfarm income in total income also decreases with an increase in farm size, indicating that small farmers depend more on nonfarm income.

Table 13. Percentage share of different sources of income in total household income, 1994-98. Item Small farms Income household-1 (000 Rs) % Share Rice Rabi cropsa Nonfarm Others Medium farms Income household-1 (000 Rs) % Share Rice Rabi cropsa Nonfarm Othersb Large farms Income household-1 (000 Rs) % Share Rice Rabi cropsa Nonfarm Othersb All farms Income household-1 (000 Rs) % Share Rice Rabi cropsa Nonfarm Othersb
a Rabi

Mungeshpur

ltgaon

Both villages

13.0 13 23 50 14 26.0 17 37 28 19 36.1 22 37 13 28 17.4 15 30 37 17

11.1 6 19 65 9 22.1 3 27 63 7 34.5 6 35 43 16 22.9 5 30 52 13

12.4 11 22 54 13 23.8 10 32 46 12 34.7 8 35 39 18 20.2 9 30 46 15

crops are mostly wheat intercropped with mustard. b Includes income from other kharif and summer crops, income from livestock, and income as a farm laborer.

130

Risk benefit and rice research What opportunities exist for rice research to reduce fluctuations in income and consumption of farmers? What is the size of the economic benefit if rice yield and production could be stabilized? Answers to these questions are critical for designing suitable technological and policy interventions to reduce the ‘cost’ of risk. Before proceeding further, it is essential to define what we mean by cost of risk and to develop a device to measure it quantitatively. For this, we use the expected utility model of decision-making. The model postulates that, under risky situations, decision makers evaluate decisions in terms of expected utility of income and choose the decision that maximizes the expected utility. For risk-neutral decision makers, the decision that maximizes the expected utility is also the decision that maximizes the expected income gain. A riskaverse decision maker, on the other hand, would be willing to sacrifice some income to avoid taking risk. The cost of risk is the amount of income sacrificed to protect or insure against risk. Using the expected utility theory, the cost of risk can be approximated as (Pandey et al 1999) D = 0.5 R [a2Cr2 + 2 a (1 -a) g CrCy] (2)

Table 14. Cost of instability in rice income. Item CV of rice income (%) CV of nonrice income (%) Ratio of rice income to total income (%) Av cost of risk (% of mean income) if correlation between rice and nonrice income = 0 if correlation between rice and nonrice income = 0.2 Mungeshpur 30 15 15 0.3 0.4 ltgaon 76 15 5 0.1 0.3

close to zero. If the applicable value of the coefficient of relative risk aversion is 2, the cost of risk will be D = a 2 [Cr]2 (3)

where D is the cost of risk (or risk deduction) expressed as a proportion of mean income, R is the coefficient of relative risk aversion, -Cr is the coefficient of variation (CV) of rice income, a is the share of rice income in total income, Cy is the CV of nonrice income, and g is the correlation coefficient between rice and nonrice income. The proportional risk premium measured in equation 2 provides an estimate of the cost of risk currently borne by farmers relative to the situation in which the variability of rice income is completely eliminated. As there will always be some variability of rice income that cannot be eliminated, the estimate obtained from equation 2 can be considered an upper-bound value. For simplicity, let us assume that the correlation between rice and nonrice income is

Thus, the cost of risk is directly proportional to the CV of rice income and the share of rice income in total income. The average costs of risk in rice income are low for both villages (Table 14). Although the CV of rice income for Itgaon is as high as 76%, a low ratio of rice income to nonrice income has made the cost of risk low. This basically means that in farming communities where income diversification has reduced the importance of rice as a source of income, the economic cost of variability in rice income will be low. It is the variability of total income, not that of rice income alone, that is of concern to farmers. The benefit of stabilizing rice income thus decreases as the contribution of rice to total income shrinks. A major factor determining the cost of risk is the ratio of rice income to total income. To illustrate the effect of this ratio, we recalculated the cost of risk in both villages by varying this ratio from close to zero to 0.5 but assuming the CV of rice income to remain constant as in Table 14. The results indicate that the cost of risk in Itgaon increases rapidly with the increase in the share of rice income (Fig. 6). By keeping the share of rice income in total income low, farmers in Itgaon thus seem to have been able to avoid the cost of risk. The above estimates of the cost of risk were obtained assuming that rice and nonrice incomes have a zero correlation. If fluctuations in rice income are independent of fluctuations in

131

Fig. 6. Relationship between the risk premium and share of rice in total income.

nonrice income, the latter will not contribute to the income risk of rice. Although the share of rice is small in both villages, the correlation coefficient is likely to be positive. If the correlation coefficient is 0.2, the risk increases by about 0.2 percentage points. This additional cost of risk cannot be reduced by lowering the instability in rice income but will require ways of making the level of nonrice income less dependent on rice income. Diversification of the income base away from rice can result in such an outcome. Policies such as increased investment in rural education and improvements in rural infrastructure promote such diversification and thereby help reduce the cost of variability in rice income. Such policies can also simultaneously result in efficiency gains. The above picture indicates that risk benefits derived from stabilizing rice income are fairly small. This does not, however, mean that investment in technology development for risk reduction has a low return. The total benefit of an intervention can be viewed as the sum of the gain in mean income and the value of reduction in risk. Even if the latter benefit is small, the efficiency gains resulting from an improvement in rice productivity are likely to be large enough to justify such investments. In addition, improvements in rice productivity could be an effective strategy for inducing income diversification and overall rural economic growth. 132

Concluding remarks
To manage risk in rice production, farmers use a range of methods such as altering the proportion of rice area, adjusting the area under traditional varieties, growing several rice varieties, and changing the rice establishment method. Some of these responses are seasonal, whereas others are more permanent and have been incorporated into the nature of rice production systems. In addition to these adjustments, crop and income diversifications are features that have helped reduce the overall risk in income. These riskcoping mechanisms are based on the strategy of diversification and maintenance of flexibility. Researchers and policymakers need to be cognizant of these risk-coping mechanisms so that interventions that complement these mechanisms can be developed. A lot of research effort has been made to improve yield stability of rice through breeding and agronomic manipulations. Although yield stabilization is an important research objective. our data indicate that variability in rice area is often higher than variability in yield. When initial rains are delayed, farmers fail to plant rice and divert the land to other crops or simply fallow it. Technologies that permit late establishment of rice are needed to reduce variability in rice area. To the extent that such technologies are also capable of stabilizing yield, the overall variability in rice production

could be substantially reduced. Although we analyzed the situation where rice planting was not possible due to early-season drought, submergence can similarly create wide fluctuations in rice area. Stabilization of rice area should be an important objective of technical research in rainfed systems. Another finding of the study is that farmers rely more on TVs in unfavorable planting seasons to reduce yield losses. These varieties are presumably better adapted to abiotic stresses, especially during planting time. Identifying and incorporating such adaptive traits in varietal improvement programs could help reduce losses under adverse conditions. Overall, rainfed rice farmers of eastern Uttar Pradesh appear to be reasonably risk-efficient in the sense of being able to maintain total income, even in the face of a sharp decline in rice income. Even though variations in rice yield, net returns from rice per unit area, and rice income per household are high, the overall economic cost of variability in rice income is low due to a small share of rice in total income. Farmers are apparently able to meet deficits in rice production by allocating their resources to alternative activities. The existence of these relatively low-cost coping mechanisms does not mean, however, that research to reduce the probability of low yields should not be undertaken. Even though risk benefits per se may be small, the resulting gain in productivity can benefit farmers by raising their average incomes. Although methods and tools for microeconomic analysis of risk are generally available, the lack of temporal farm-level data covering a sufficient number of periods remains a problem for the analyst. While an ingenious use of cross-sectional data and weather-driven crop growth models can help shed some light on the nature of risk, temporal data are essential for conducting a more complete analysis at the farm-household level. Collection of such data should be a part of the research program of risk analysis.

References
Judge GG, Griffiths WE, Hill CR, Lutkepohl H, Lee T-C. 1985. The theory and practice of econometrics. New York (USA): John Wiley and Sons. Pandey S, Singh HN, Villano RA. 1999. Rainfed rice and risk-coping strategies: some microeconomic evidences from eastern India. Selected paper at the Annual Meeting of the American Agricultural Economics Association, 8-11 August 1999, Nashville, Tennessee, USA. Smale M, Just RE, Leathers CV. 1994. Land allocation in HYV adoption models: an investigation of alternative explanations. Am. J. Agric. Econ. 76:533-546. Talawar S. 1996. Local soil classification and management practices: a bibliographic review. Research Paper 2. Laboratory of Agricultural and Natural Resource Anthropology, Department of Anthropology, University of Georgia, Athens, Georgia, USA. Walker TS, Jodha NS. 1986. How small farm households adapt to risk. In: Hazell PBR, Pomareda C, Valdes A, editors. Crop insurance for technology development: issues and experiences. Baltimore, Md. (USA): Johns Hopkins University Press. p 17-34.

Discussion
Question: How the implications of this study will be useful when rice accounts for only a small proportion of the total income of sample farmers. The share of rice in the total income is small, but rice is still the dominant crop in the kharif season and many people are involved in rice production activities. An increase in the productivity of rice can improve food security, especially of the poorer groups. While the effect of risk to individual farmers may be small, a substantial drop in rice yield over a large area is

Answer:

133

likely to reduce labor employment in farm and other sectors that are dependent on farm production. Thus the finding of the study that risk-efficient technologies are needed is still valid. The results, however, imply that the demand for crop insurance for rice may be low in these areas as farmers are willing to pay only a small proportion of income to purchase insurance that stabilizes rice income at the current mean level. Question: Some differences in the two samples like size of holding and land types might have influenced the results. It would be better to make the samples homogeneous and take out the effect of these factors. As regards the share of rice in total income, I feel it should be higher if you consider net income. This is because the cost is much higher for cash crops like sugarcane. Answer: Yes, there are differences in size of holdings and number of parcels per holding between the two villages. We have tried to control for these differences by conducting separate analysis by farm-size categories. For the temporal analysis, the comparison is not across villages but over time for each village. Question: If the major share is coming from non-farm income, why should farmers bother about risk in rice and therefore, crop management practices may not be really seen in the context of risk management. Answer: Yes, opportunities for income gains through agronomic management that reduce fluctuations in rice yield may not be so important in this situation. The current structure of household income has evolved over a period of time in such a way that non-farm income is contributing maximum to the total household income. This evolution itself may

be a response to risk as diversification out of rice may have helped reduce the overall risk. The majority of income now is coming from rabi crops and non-farm employment. The availability of irrigation and proximity to urban centers may have facilitated such diversification in these villages. However, environmental conditions and accessibility in other areas may constrain such diversification out of rice. Risk management strategies may be more important in those areas. We have started similar studies in other locations so that some generalizations can be made.

Notes
Authors’ addresses: S. Pandey and R. Villano, International Rice Research Institute, Los Baños, Philippines; H.N. Singh, Narendra Deva University of Agricultural Technology, Faizabad, India. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAPDRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

134

Risk and the value of rainfall forecast for rainfed rice in the Philippines
Abedullah and S. Pandey

The value of a rainfall forecast for rainfed rice production in the Philippines is estimated under the assumption that farmers adjust the quantities of fertilizer and labor if rainfall forecasts are available. Using a panel data of 46 rice farmers in Tarlac, Philippines, a heteroscedastic production function with growing-season rainfall (July to October) as one of the independent variables is estimated. The value of a perfect predictor of rainfall is subsequently estimated for risk-neutral and risk-averse farmers. The effect of risk aversion on the value of rainfall forecast was found to be minimal. The expected value of the rainfall forecast was higher when farmers were assumed to be adjusting labor only than when farmers were assumed to adjust fertilizer only. The expected value of the rainfall forecast under the assumption of simultaneous adjustment of fertilizer and labor was estimated to be slightly more than 1% of the net return from rice production. Taking the rainfed rice area in the Philippines of 1.2 million ha and a net return of $446 ha-1, the total value of the forecast is estimated to be $6.6 million per year. The expected value was also estimated under the assumption that, instead of a forecast of rainfall amount for each year, forecasts made were for rainfall “above average,” “average,” or “below average.” The value of rainfall forecast was the highest and ranged between 1.4% and 4.5% of the net return when the forecast was “above average.”This implies that investment by the Philippines weather bureau to predict an above average rainfall situation will be more valuable to Filipino rice farmers than predicting other rainfall categories.

Agriculture is risky because various kinds of risks are involved in the production and marketing of agricultural products (Anderson and Dillon 1992). Since risk arises out of the uncertainty about variables that affect production and profits, a reliable prediction of these uncertain variables will reduce risk. When information about uncertain variables is available, input use and productivity are likely to be different compared with a situation when no such information is available. Although risk emanates from several sources, climatic uncertainty is the dominant source of risk, especially under rainfed conditions. Rainfall forecasts can partially help resolve uncertainties in rice production. In the Philippines, the Philippine Atmospheric, Geophysical and Astronomical Services

Administration (PAGASA) provides seasonal rainfall forecasts that are classified as “above average,” “below average,” or “average.” Although these forecasts are provided to help improve farmers’ decisions regarding rice production, it is uncertain how valuable these forecasts are to rice farmers in economic terms. Allocation of more resources for forecasting of rainfall is desirable if the additional value of the rainfall forecast is higher than the additional cost of providing such a forecast. Estimates of the value of the rainfall forecast can be a useful guide in determining the optimal resource allocation for generating forecasts. To aid in this task, this study estimated the potential value of rainfall forecast to rice farmers of Tarlac, Central Luzon, Philippines.

135

A conceptual model for estimating the value of information
The value of information can be derived using the standard model of agricultural risk analysis (Anderson et al 1977). Let f be the stochastic variable (i.e., state of nature) beyond the control of the decision-maker. If X is a vector of variable inputs that are manipulated by the decision-maker, the return g( f ,X) earned depends on the state of nature and the vector of inputs. The function g( f ,X) embodies input, output, and price relationships. In the absence of forecast information, decisions are based on the prior belief about the probability distribution of the stochastic variable. Let this prior probability density function be denoted by f (X). A riskneutral decision-maker selects % to maximize the expected return òg(f ,X)f (X)dX from the production process. The optimal decision and profits based on prior information only are the prior optimal level of inputs and the prior optimal profits, respectively. On the other hand, if the economic agent has a forecast of the value of f (i.e., about the state of nature) before the selection of the input vector, the decision-maker will select X to maximize g( f ,X) for each X. Let the maximized value of the profit be represented by g(f ,X). The expected value of information “V” for risk-neutral farmers is the difference between expected profit derived with and without the information and is obtained as (1) where V stands for the value of information to risk-neutral farmers and h(f ) is the probability density of an uncertain event revised using the forecast information. Following the expected utility model of decision-making under risk, risk-averse farmers are assumed to maximize the expected utility of profit (Anderson et al 1977). Analogous to the model above for risk-neutral farmers, expected utility is estimated “with” and “without” information and the difference in the expected utility can be regarded as an indicator of the expected value (in utility terms) of a rainfall predictor to risk-averse farmers. As the differences in utility, which are ordinal in scale, 136

are meaningless, we have used the method followed by Byerlee and Anderson (1982) to obtain the value of information in money terms: (2)
* where E[U(p) is the expected utility of the prior optimal act and E[U(p - ] is the expected utility of the optimal act derived using the prediction that costs $v to acquire. To implement model 1, it is essential to quantify the effect of inputs (X) and stochastic variables (f ) on agricultural output. Such a relationship can be quantified using a production function. To implement model 2, the utility function that relates the level of profit to utility is also needed. A convenient form of the utility function is the constant partial risk-averse function (CPRA). It is specified as

(3) where S is the risk aversion coefficient. This form of utility function has been widely used in applied research (Sillers 1980, Smith and Umali 1985, Rosegrant and Roumasset 1985).

Description of study area
Socioeconomic monitoring of rice production practices of 46 farmers from the municipality of Victoria, Tarlac, Philippines, began in 1990. Rice is grown in the rainy season with most of the land being left fallow in the dry season. The area has good market access and is well linked with the town economy of Tarlac. Farmers are engaged in various off-farm and nonfarm activities during the dry season to supplement their incomes. Table 1 shows the means and coefficients of variation (CVs) of output and input variables. Based on the long-term weather record (1976-93, the mean annual rainfall in Victoria is 1,649 mm. Figure 1 shows the variation in rainfall during the rice-growing season (July-October). Plot-level data from the sample of 46 farmers were collected for the period 1990-95. All inputs and output were recorded in a survey questionnaire, which was administered every year to the same group of farmers. Unbroken panel data for 420 plots for

each of the six years were used to estimate the production function. The only source of uncertainty considered was rainfall, which was specified in the model as the total rainfall during the rice-growing season. For biological reasons, it would have been more appropriate to specify rainfall as a weekly or monthly total compared with a seasonal total. However, we used the seasonal total because a reliable estimation of the production responses for weekly or monthly rainfall using production data for only six years
Table 1. Means of per hectare input and output use in Victoria, Tarlac, Philippines. Input/output Total labor b (d ha-1) Seed (kg ha-1) Fertilizer (NPK) (kg ha-1) Herbicide (kg ai ha-1) Pesticide (kg ai ha-1) Yield (t ha-1)
a

would have been constrained by the limited degrees of freedom. Production function estimation When production functions are estimated using a combination of cross-sectional and time-series data, heteroscedasticity may lead to asymptotically inefficient parameter estimates (Just and Pope 1978). The Breusch-Pagan test rejected the null hypothesis of homoscedasticity at the 5% level. To correct for heteroscedasticity, a multistage production function estimation technique suggested by Antle (1983) was used. A quadratic production function as specified in 1 2 equation 4 was used.

Mean 55 110 94 0.14 0.09 3.35
b

Standard deviation 15 47 42 0.21 0.14 1.07

CV

a

(%)

27 43 45 150 155 32

(4) where Y, X1, X2, and Z represent yield, total labor, fertilizer (total of N, P, and K), and total

CV = coefficient of variation.

lnciudes family labor and hired labor.

137

Table 2. Quadratic production function estimates Antle s technique in Victoria, Tarlac, Philippines.a Explanatory variables Intercept Labor Fertilizer Rain Labor2 Fertilizer2 Rain2 Labor x rain Fertilizer x rain R2 n
a

with

Coefficients

Standard errors 134.1E-02 18.2E-03 7.6E-03 1.6E-03 9.1E-05 1.7E-05 5.6E-07 7.1E-05 1 E-05 4.4E-07

-72.9E-03ns 30.2E-03* 18.2E-03*** 1.4E-03ns -0.2E-03*** -3.8E-05*** -1.1E-06** -0.2E-03*** 1.7E-05* 5.1E-06** 0.25 420

... = significant at 1%, ** = significant at 5%. * = significant at 10%, ns = not significant.

rainfall during the rice production period (from July to October), respectively. The stochastic error term is represented by µ. Table 1 shows the per hectare input use (labor, seed, herbicide, pesticide, and active ingredients of fertilizer i.e., sum of N, P, and K) and output with descriptive statistics. Table 2 presents the parameter estimates using Antle’s method. The coefficients of labor, labor’, fertilizer, fertilizer’, rain, and rain2 all have the expected signs. The joint effects of rain and labor and those of rain and fertilizer are positive and significant. The interaction term between fertilizer and labor is negative, indicating that these two inputs have a substitute relationship in rice production. Procedure for estimating the value of rainfall forecast In valuing the forecasts, the prior probability distribution of rainfall is required. We assumed the farmers’ estimates of the probability of rainfall to be equivalent to the historical distribution of total rainfall during the rice production period. The probability distribution was estimated by applying the sparse data rule (Anderson et al 1977) to the historical rainfall for the period 1977-95. For a given value of the decision variable, profits were generated for each year by substituting the rainfall for that particular year into the production function. The expected profit was then calculated by using the corresponding rainfall probability weights. This 138

process was repeated for all possible values of the decision variable and the value of the decision variable that generated the maximum expected profit was taken as the prior optimal decision. For risk-averse farmers, the prior maximal expected utility was similarly calculated by substituting the profit for each decision in equation 3 and using the corresponding probability weights. Sillers (1980) concluded that 78% of rice farmers in Nueva Ecija, Philippines, belong to the two intermediate categories of risk aversion. The common endpoint of these two categories is S = 0.8. The value of the risk aversion coefficient used in this study was 0.8. The rainfall forecast was generated by random sampling from the discrete probability distribution of the historical rainfall data. Assuming that the prediction is perfect, optimal profit for this forecast was then obtained using the estimated production function. As prediction is assumed to be perfect, the probability distribution of the forecast is also the historical probability distribution of rainfall. Using the historical distribution, the expected profit when a perfect predictor of rainfall was available was then calculated. The difference between this expected profit and the expected profit of the prior optimal act is the expected value of the perfect predictor of rainfall. A similar procedure was used for the risk-averse case and equation 2 was used to obtain the value of a perfect predictor to a risk-averse farmer. Estimation of the value of the perfect predictor is convenient as it avoids the need to explicitly obtain the likelihood function, which is an indicator of the accuracy of the predictor. Since no rainfall predictor can be perfect, the estimated value of the predictor can be considered as the upper limit of the value of rainfall forecast. As rainfall forecasts in the Philippines are provided as “average,” “below average,“ or “ above average,” we also estimated the value of these forecasts. The 19-year historical JulyOctober total rainfall fluctuated between 700 and 1,650 mm (IRRI various years). We divided this total range of rainfall into three categories”below average” (between 700 and 1.000 mm of rainfall), “average” (between 1,050 and 1,300 mm), and “above average” (between 1,350 and

1,650 mm). To calculate the value of the perfect predictor, predicted rainfall values were limited within the range defined by these prediction categories. The prior optimal expected profit was obtained as before. The posterior expected profit was estimated by using the optimal profits for each perfect prediction (randomly selected, within the particular rainfall category) and the corresponding conditional probabilities. The difference between these two expected profits is the value of a particular category of forecast for a risk-neutral farmer. A similar procedure was used for the risk-averse farmer. Labor and fertilizer are the two decision variables considered. Farmers may adjust either or both of them from their prior optimal values if rainfall forecasts are available. We estimated the value of the rainfall forecast under the assumptions that (1) only labor is adjusted to its posterior optimal value while fertilizer is fixed at its sample average, (2) only fertilizer is adjusted to the posterior optimal value while labor is kept fixed at its sample average, and (3) both labor and fertilizer are adjusted simultaneously to their posterior optimal values. Value of rainfall forecast Table 3 presents the estimated values of a perfect rainfail predictor for risk-neutral and risk-averse farmers. The expected value of the rainfall forecast, if farmers are assumed to adjust fertilizer application when rainfall predictions are available, is $1.92 and $2 ha-1 for riskneutral and risk-averse farmers, respectively. These values account for 0.43% and 0.46% of
Table 3. Total expected value of forecast ($ ha-1) under different input adjustments in Victoria, Tarlac, Philippines. Item Fertilizer adjustment only Labor adjustment only Simultaneous adjustment of fertilizer and labor Value of rainfall forecast” Risk-neutral 1.92 (0.43) 4.08 (0.92) 5.50 (1.23) Risk-averse 2.04 (0.45) 4.29 (0.96) 5.79 (1.29)

Value in parentheses represents percentage of net return.

the net return from rice production in the study are a. The expected value of the rainfall forecast under the assumption that only labor is adjusted is $4.08 and $4.29 ha-1 for risk-neutral and riskaverse farmers, respectively. These account for 0.92% and 0.96%, respectively, of the net return from rice production. The expected values of a perfect predictor of rainfall (under the assumption that farmers adjust fertilizer and labor simultaneously) for risk-neutral and riskaverse farmers are $5.50 and $5.79 ha-1, respectively. In terms of percentage of net return, these values are 1.23% and 1.29%, respectively. The effect of risk aversion on the value of the rainfall forecast was minimal. The cost of acquiring information was not included in the above calculation, implying that the results indicate the expected gross benefits of the forecast. Net benefits depend on the cost of obtaining and using information. Gross benefits, as estimated here, are useful to indicate the maximum amount a farmer would be willing to pay to obtain the forecast. The expected values of the forecast under the assumption that forecasts are “below average,” “average,” or “above average” (according to categories) were also estimated (Table 4). In the case of simultaneous adjustment of fertilizer and labor, values vary from $1.58 to $20 ha-1 depending on the type of the forecast and the assumption about risk attitude of the farmers. In terms of percentage of net return, these values range from 0.35% to 4.52%. The “above average” forecasts are seemingly more valuable to farmers than “below average” and “average” forecasts. PAGASA could help farmers more by investing more of its resources in accurate predictions of above average rainfall events. The estimates of the value of the rainfall forecast obtained here are only a small fraction of the net return. In rainfed agriculture where rainfall is the major source of uncertainty, such a low value may appear to be somewhat surprising. This may be partly the result of the model specification in which rainfall is included as the seasonal total and the only two decision variables considered are labor and fertilizer

139

Table 4. Expected value of different rainfall forecasts for rice production period ($ ha-1) under different input adjustments in Victoria, Tarlac, Philippines. Forecast and adjustment of inputs Below average (700-1,000 mm) Fertilizer adjustment only Labor adjustment only Simultaneous adjustment of fertilizer and labor Average(1,050-1300 mm) Fertilizer adjustment only Labor adjustment only Simultaneous adjustment of fertilizer and labor Above average (1,350-1.650 mm) Fertilizer adjustment only Labor adjustment only Simultaneous adjustment of fertilizer and labor
a

Concluding remarks
Overall, the expected value of a hypothetical perfect predictor of rainfall is between $2 and $6 ha-1 for moderate risk-averse farmers, which is a little higher than the 1% of the net return earned by farmers from rice production in the study area. This is an upper-bound estimate of the value of the rainfall predictor as the prediction accuracy of a realistic predictor is likely to be less than 100%. As rainfall prediction in the Philippines is provided costfree, this is also an estimate of its net value to farmers. The effect of risk aversion on the value of the forecast was minimal. Our results also indicate that the value of information is asymmetrical, with the “above average” forecast being four times more valuable than the “average” forecast and about two times more valuable than the “below average” forecast. An important implication of this finding is that additional effort by PAGASA to predict “above average” rainfall events may be justifiable. Overall, the average value of a perfect predictor of seasonal rainfall to rainfed rice farmers of the Philippines was estimated to be $6.6 million per year. The value of the forecast depends critically on its quality and timeliness (Mjelde et al 1988). Forecasts are valuable only if they are received before inputs have been applied. We did not investigate timeliness because of limitations of data for estimating production functions that adequately capture temporal interactions between managed inputs and rainfall. Similarly, we have considered rainfall as the only source of uncertainty on the assumption that rainfall variability is the major source of risk in rainfed rice production. Further expansion of the approach used to include these refinements is suggested.

Value of rainfall forecast Risk-neutral 1.42 (0.32) 2.71 (0.61) 3.42 (0.77) 0.50 (0.11) 1.25 (0.28) 1.58 (0.35) 6.1 7 (1.38) 13.75 (3.08) 19.33 (4.33)

a

Risk-averse 1.50 (0.34) 2.92 (0.65) 3.67 (0.82) 0.58 (0.13) 1.33 (0.30) 1.71 (0.38) 6.50 (1.46) 14.17 (3.18) 20.17 (4.52)

Value in parentheses represents percentage of net return.

application. Nevertheless, estimates derived here are comparable with those obtained for other countries. Mjelde et al (1988) reported the value of rainfall forecast varying from 5% to 13% of the net return in maize production in Illinois (USA). Pannell (1994) estimated the value of information from herbicide decision making in wheat production in Australia to be between 0 and 15% of the gross margin from the crop. Marshall et al (1996) estimated the value of the seasonal forecast for dryland wheat production in Australia to be between 0 and 6% of the net return. Even though the value of the forecast expressed as a percentage of net return may be small, the total absolute value can be quite large, depending on the size of the area covered by the forecast. Taking the rainfed rice area in the Philippines of 1.2 million ha and a net return of $446 ha-1 (Abedullah 1998), the total benefit to rainfed rice farmers would be $6.6 million y-l even if the value of the forecast is 1.23% of the net return.

References
Abedullah. 1998. Risk, information, and fertilizer use in rainfed lowland rice areas of Tarlac, Philippines. Unpublished PhD dissertation, Graduate School, University of the Philippines Los Baños.

140

Anderson JR, Dillon JD, Hardaker JB. 1977. Agricultural decision analysis. Iowa State University Press, Ames, Iowa. Anderson JR, Dillon JL. 1992. Risk analysis in dryland farming systems. Food and Agriculture Organization of the United Nations, Farm Systems Management Series 2. Antle JM. 1983. Testing the stochastic structure of production: a flexible moment-based approach. J. Business Econ. Stat. 1(3):192201. Byerlee DR, Anderson JR. 1982. Risk, utility and the value of information in farmer decision making. Rev. Marketing Agric. Econ. 50(3):231-246. IRRI. Various years. Climate Data from Agronomy, Plant Physiology, and Agroecology Division. International Rice Research Institute (IRRI). Just RE, Pope RD. 1978. Stochastic specification of production functions and economic implications. J. Econ. 7(1):67-86. Marshall GR, Parton KA, Hammer GL. 1996. Risk attitude, planting conditions and the value of seasonal forecasts to a dryland wheat grower. Aust. J. Agric. Econ. 40(3):211-233. Mjelde JW, Sonka ST, Dixon BL, Lamb PJ. 1988. Valuing forecast characteristics in a dynamic agricultural production system. Am. J. Agric. Econ. 70(3):674-684. Pannell JD. 1994. The value of information in herbicide decision making for weed control in Australian wheat crops. J. Agric. Resour. Econ. 19(2):366-381. Rosegrant MW, Roumasset JA. 1985. The effect of fertilizer on risk: a heteroscedastic production function with measurable stochastic inputs. Aust. J. Agric. Econ. 29(2):107-121. Sillers DA. 1980. Measuring risk preferences of rice farmers in Nueva Ecija (Philippines), an experimental approach. PhD dissertation, Yale University. Smith J, Umali G. 1985. Production risk and optimal fertilizer rates: an application of the random coefficient model. Am. J. Agric. Econ. 67(3):654-659.

Discussion
Question: Answer: What is the nature of forecast monthly or daily rainfall? It is a seasonal rainfall forecast in terms of average, above average or below average. How will the value of rainfall forecast change if forecast is inaccurate? Also, have you discounted the value or benefits on account of farmers’ risk adjustment behavior and changes in input use in the absence of rainfall forecast? We have estimated the value of forecast assuming that all forecasts were correct and there are two inputs, viz. labor and fertilizer, which can be adjusted depending on the forecast. As real world forecasts are never perfect, the estimates provides an upper bound value of forecast. Yes, the farmers’ responses you mention are built into the model through the use of a production function and optimizing conditions. Is this value of 13 per cent a large number in comparison to the value of other information like variety selection and fertilizer application. We have applied this model to rainfall forecast only. The model can be potentially applied to other kinds of forecasts also such as pest attack. Fertilizer application depends upon moisture conditions and this effect has been captured in the production function. If farmers have better scientific advice on fertilizer management, the value of forecast may go up.

Question:

Answer:

Question:

Answer:

141

Notes
Authors’ addresses: Abedullah and S. Pandey, International Rice Research Institute, MCPO Box 3 127, Makati City 1271, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

142

Characterizing risk and strategies for managing risk in flood-prone rice cultivation in Assam
B.C. Bhowmick, S. Pandey, R.A. Villano, and J.K. Gogoi

High risk of flood and low yields characterize the rice production systems in Assam. This paper analyses the changes in rice productivity and variability over the period 1971-97 using district level data. In addition, some farm-level data are used to highlight the differential effects of risk on the adoption of modern varieties and features of rice production systems. The analysis of instability in rice production using district-level data indicated that modest production growth has not led to an increase in production instability in Assam. Despite the increase in production variance, coefficients of variation of area, yield and production did not record any significant changes. The farm-level data showed that the share of rice income in the total household income in Assam is relatively higher than in other states of eastern India. The economic benefits from yield stabilization are therefore likely to be substantial in Assam.

Eastern India, which comprises Assam, West Bengal, Orissa, Bihar, eastern Uttar Pradesh, and eastern Madhya Pradesh is predominantly a ricegrowing region. Rice in eastern India is grown mainly under rainfed conditions. In spite the importance of rice in the economy of Assam, rice productivity here is the lowest of all eastern Indian states. Rice production heavily depends on the monsoon rains that cause severe flooding and submergence. The high level of climatic risk is a major factor constraining the adoption of improved technologies in this state. Most past studies on eastern India excluded Assam from their analyses because of data accessibility problems in this somewhat remote and politically sensitive state. The major objective of this paper is to analyze the changes in rice productivity and variability over the period 197 1-97 using district-level data. Attempts are also made to explain the differences in the variability of rice area, yield, and production across districts. In addition, some farm-level evidence on the possible effects of differential levels of risk on the characteristics of rice production systems is investigated. Under risky situations, the main objective of farmers may not be to maximize crop productivity but to

assure survival. Over time, farmers develop strategies to ensure survival when faced with such risky situations. Such strategies can be classified as ex ante or ex post depending on whether they reduce risk or help cope with the situation when losses do occur. A good understanding of such strategies can be helpful in developing interventions that complement farmers’ strategies. A brief discussion of strategies used by Assam rice farmers to cope with risk is also provided.

Data and methods
The analysis of production variability in Assam is based on district-level time series data (197197) of the state. In as much as these data were not easily available from the national statistical agencies, some efforts were made to collect and collate data from the state and district authorities. Data have been grouped according to the original districts. The coefficient of variation (CV) of rice area, production, and yield was computed after linearly detrending the timeseries data. Farm-level analysis was carried out from primary data collected by the Department of Agricultural Economics, Faculty of

143

Agriculture, Assam Agricultural University, Jorhat (Bhowmick 1992, Das 1995, Gogoi 1998). To analyze the change in instability over time, the variance decomposition analysis developed by Hazell (1982) was used. The timeseries data were divided into two periods of equal length-1971-1982and 1983-19941. The data on area and yield for each period were linearly detrended and the detrended data were centered on the respective means. Detrended yield and detrended area were multiplied to obtain detrended data on production. These detrended data were used for the variance decomposition analysis. The three zones with highest number of flood-affected areas were considered for the microlevel analysis. The zones selected were North Bank Plains Zone (NBPZ), Lower Brahmaputra Valley Zone (LBVZ), and Upper Brahmaputra Valley Zone (UBVZ). The districts selected from these zones were Lakhimpur, Kamrup, and Sibsagar, respectively. More than 50% of the rice area in Lakhimpur is chronically flood-prone, whereas a similar area in Kamrup and Sibsagar is about 10%. The production environment in Sibsagar is more favorable, as indicated by its higher average yield and more rapid expansion of summer rice. A total of 420 households were surveyed and a brief analysis of farm-level data in terms of input use, yield, nature of tenancy, and the income structure is presented.

The average annual rainfall in the state is 2,586 mm. About 60-70% of the total. rainfall is received within a span of 3-4 months (May to August). Because of high rainfall, flooding is a typical feature of agriculture in Assam. Flood damage is more intense on either side of the rivers Brahmaputra and Barak and their tributaries. Between 1957 and 1996, flood affected an average of 236,000 ha of cropped area (Table 2). In heavily flooded years, such as 1988, more than 1 million ha of cropped area were affected. The occurrence of multiple waves of flood during the monsoon period extensively damages crops. The number of people affected by flood and the amount of damage have increased steadily over time. Based on the incidence and extent of flood, the state is categorized into two broad areas: chronically flood-prone area and occasionally flood-prone area (Table 3). Nearly 73% of the total flood-prone area is chronically flood-prone. The proportion of the flood-prone area is highest in Lakhimpur at 63%. Among the agroclimatic zones, the NBPZ occupies 39% of the floodprone area of the state. High variability of crop
Table 1. Some basic features of agriculture of Assam. Features Population density (persons km-1) Average operational holding (ha) Average annual rainfall (mm) Cropping intensity (%) Area under important crops (000 ha) Rice Wheat Pulses Oilseeds Fiber crops Yield of some important crops (t ha-1) Rice" Wheat Pulses Oilseeds Fiber crops
a Rice

Year 1991 1991 1994 1991 1997

Particulars 286 1.27 2,586 144 2,526 80 120 320 100 2.03 1.30 0.55 0.55 0.89

Major features of rice production systems of Assam
Table 1 lists some of the important features of Assam. The mainstay of the economy is agriculture, which accounts for about 40% of the state's domestic product (Phukan 1996). Assam is divided into 6 agroclimatic zones, whose particulars appear in Appendix 1.

1997

data are reported in terms of rough rice (conversion factor 1 kg rough rice = 0.66 kg milled rice). Source: Directorate of Economics and Statistics (various years).

1For variance decomposition analysis, data for the recent years were not used. To make use of the data up until 1996, the data span could have been divided into 1971-83 and 1984-96. But for the sake of consistency of analysis for other states (included in this volume), for which some of the more recent data were not available for all districts, we decided to use the time segments 1971-82 and 1983-94 for all states.

144

Table 2. Cropped area and population affected by flood in Assam (1957-96). Cropped area affected by flood (000 ha) 104 206 122 21 0 226 226 1,060 226 236 Cropped area affected as % of total area affected 17 17 26 38 29 31 29 46 28 Population Total affected damages a (000) (million Rs)

Year

1957-61 1962-66 1967-71 1972-76 1977-81 1982-86 1987-91 1992-96 Average
a

82 210 109 197 251 282 1,107 271 248

48 96 71 144 221 834 3,797 31 3 438

Value of crops, livestock and house and property. Source: Centre for Science and Environment (1996).

ha-1.2

yields and low productivity are the characteristic features, especially in the chronically floodprone areas (ICAR 1981). Assam is one of the important northeastern states of the country for rice cultivation, which covers more than 63% of the gross cropped area in the state. Assam’s rice area for 1997 was 2.5 million ha with an average productivity of 2.03 t Basically three distinct types of rice are grown in the state. Of these, sali (winter rice) is the major rice grown during the kharif season

(June–October). Ahu (autumn rice) is early rice grown between February and June. Boro rice (summer rice) is grown between November and May. Autumn rice is grown as a normal ahu crop under rainfed conditions and as an early ahu under irrigated conditions. With the expansion of irrigation, the early ahu has become popular in Cachar, Nagaon, Dibrugarh, Kamrup, and Goalpara districts. The varieties used for early ahu are mostly improved. This rice type is established in early February and harvested in late May. Because of its higher productivity, there has been some shift in area from normal ahu to early ahu in recent years. Flood generally affects ahu rice during the maturity stage and sali rice during the early and late period of its growth. Medium land and medium lowlands are the important land types where ahu and sali rice are usually grown. As with ahu, boro rice in Assam is also of two types. Boro grown in lowlying rainfed areas that contain enough soil moisture is known as “typical” boro. Boro rice grown under irrigated conditions is known as irrigated boro. While typical boro is limited to some pockets, the area under irrigated boro has increased since 1980. Deepwater rice (buo) is another type of rice grown in areas subject to deep inundation by floodwater. However, timeseries data on area, production, and yield of the

Table 3. Estimates of flood-prone areas of Assam. Net sown areaa (000 ha) Chronically flood-prone area 1000 ha) 65 28 16 29 16 40 52 4 248 Occasionally flood-prone area 1000 ha) 36 4 9 7 8 12 5 10 90 Total flood-prone area (000 ha) 101 32 25 36 24 52 57 14 338 Percent of flood-prone area to total net sown area 63 10 15 11 8 10 14 5 14

Agroclimatic zone/district North Bank Plains Zone Lakhimpur Darrang Upper Brahmaputra Valley Zone Dibrugarh Sibsagar Central Brahmaputra Valley Zone Nagaon Lower Brahmaputra Valley Zone Kamrup Goalpara Barak Valley Zone Cachar All of Assam
a

160 310 160 320 295 510 410 247 2,412

Av of data from 1968 to 1978. Source: Hydrogeology and Ground Water Resources Development (1980).

2

All production and yield data for rice are in terms of rough rice, not milled rice.

145

Table 4. Relative importance of various rice cultures in Assam. Item Share of rice area (%) Autumn Winter Summer Share of production (%) Autumn Winter Summer Triennium average 1971-73 27 71 2 19 78 3 1981-83 27 72 2 18 80 2 1991 -93 25 70 5 17 76 7 1995-97 24 69 7 16 75 8 Overall average

26 71 3 18 78 4

Sources of data: Directorate of Agriculture (various years)

bao crop are not available as it is considered to be a minor crop mostly confined to small pockets of very low lying areas.

Results and discussion
State-level analysis
Winter rice is the major rice culture of Assam, accounting for 71% of the rice area planted and 78% of the output (Table 4). Autumn and summer rice account for 26% and 3% of the total rice area, respectively. Over time, the share of autumn rice has shown a slight decline in both area and output. The share of summer rice has 146

increased, but its contribution to output even in more recent years remains below 10%. The average yields of autumn, winter, and summer rice for the period 1971-97 are 1.17 t ha-1, 1.86 t ha-1 and 2.25 t ha-1, respectively. Figure 1 depicts trends in area, yield, and production of rice in Assam. Area has expanded gradually over the past 25 years. The average yield of rice remained more or less stationary until 1989, after which it increased. The overall annual growth in total rice area, yield, and production in the state is estimated to be 0.9%, 1.4%, and 2.3%, respectively (Table 5). Thus, nearly 40% of the production growth over the period 1971-97 resulted from growth in area.

Table 5. Growth rates of area, production and yield of rice in Assam 1971-97. Item Autumn rice Area Yield Production Winter rice Area Yield Production Summer rice Area Yield Production Total Area Yield Production
a

% 0.43 1.11 ***a 1.54 *** 0.79 *** 1.34 *** 2.13 *** 6.95 *** 1.52 *** 8.47 *** 0.90 *** 1.39 *** 2.29 ***

*** = significant at 1% level. Source of data: Directorate of Agriculture

(various years).

Table 6. Mean and coefficient of variation of different rice types in Assam 1971-97. Item Mean Area (000 ha) Yield (t ha-1) Production (000 t) CVa (%) Area Yield Production Autumn rice 603 1.17 701 4 10 12 Winter rice 1,656 1.86 3,112 3 6 6 Summer rice 73 2.25 164 33 19 46 Total rice 2,330 1.73 4,024 2 6 7

1Calculated from the Iinearly detrended data. Source: Directorate of Agriculture (various years).

the yield growth of summer rice is quite modest, contributing to only 18% of production growth. Irrigation facilities in the state, mostly created by private, investment, have facilitated the expansion of area under summer rice. The expansion is also partly the result of a shift of autumn rice to summer rice. Table 6 shows the coefficient of variation (CV) of area and yield of different types of rice in Assam. The CVs of area of autumn and winter rice are lower than those of summer rice. The higher variability in summer rice production is due mainly to high variability of area. As area under summer rice is very small, variability of area can be high because of the lack of compensating effects over a wider region that may have helped stabilize the total area of winter rice. It is somewhat surprising to note that the average yield of summer rice is only slightly higher than that of winter rice. Despite the high level of risk associated with producing winter rice, this crop seems as productive as, and perhaps more stable than, summer rice. Yield trends show that yield fluctuations of summer rice declined after 1987 (Fig. 2) probably due mainly to the expansion of irrigation. Similarly. fluctuations in the yield of winter rice have also been lower after the mid- 1980s. Meanwhile, the yield of autumn rice appears unchanged over time.

Yield growth has contributed less in Assam than in eastern India as a whole, where its contribution was 86% over the period 197 1-97. In more recent years, the relative importance of yield growth in Assam has increased, however, with the leveling off of the growth in area. Of the three types of rice culture, winter rice recorded the yield growth of 1.34% per year. Because of the dominance of this rice culture, the overall yield growth in rice for Assam of 1.39% approximates that of winter rice. The area of winter rice has remained virtually constant, growing at less than 1% per year. The growth of area of autumn rice is also very small (less than 0.5% per year) and statistically insignificant. Only the area under summer rice has increased rapidly, more so in recent years. The expansion of rice production in the summer season is driven mainly by growth in area. In comparison,

Assam has 10 districts, of which the five (Darrang, Goalpara, Kamrup, Nagaon, and Sibsagar) account for more than 70% of the rice area (Table 7). Kamrup is the largest district, accounting for 20% of the state's total rice area. Of these major districts, Darrang, Goalpara, and Kamrup have a relatively higher proportion of area under low-lying autumn rice. As a result, the average yields of rice in these districts are lower than those in Nagaon and Sibsagar, which have a much lower proportion of autumn rice area. In Nagaon, summer rice production expanded rapidly after mid- 1980s and accounted for nearly 27% of the total output of the district for the triennium 1995-97. Judging by the average yield of winter rice, Sibsagar has the most productive rice environment with Kamrup and Goalpara the least productive. 147

District-level analysis

148

Table 7. Rice area and yield by districts. Triennium average (1 995-97) District Rice area (000 ha) Cachar Darrang Dibrugarh Goalpara K. Anglong Kamrup Lakhimpur NC Hills Nagaon Sibsagar All of Assam 222 319 152 390 122 516 178 14 336 281 2,531 (%) 9 13 6 15 5 20 7 1 13 11 100 2.61 1.80 2.35 1.53 2.23 1.67 1.86 2.75 2.21 2.65 2.01 Rice yield (t ha-1)

Sources: Directorate of Agriculture (various years).

In terms of shift in rice area, the overall trend is either stagnation or a decline in the share of autumn rice (Table 8). The share of autumn rice area declined in Cachar, K Anglong, Kamrup, NC Hills, and Sibsagar. In other
Table 8. Percentage share of rice area by districts. District Cachar Darrang

districts, the share has remained constant or increased slightly. In most districts, the share of summer rice area has increased over time. The growth rates of area and yield of winter and summer rice in almost all districts were statistically significant (Table 9). Except for Lakhimpur District, the growth of yield of total rice was highly significant in all districts, the highest being 2.8% in NC Hills. The growth in yield of winter rice in NC Hills (2.5%) and Cachar (2.3%) was higher as these districts are less affected by flood. The negative growth in area under autumn rice in Cachar (-3.0%) and Karbi Anglong (-0.5%) could be attributed mainly to a shift in area from this crop to summer rice. There is a tendency to shift from autumn rice to summer rice because of the latter's comparative advantage in yield and profitability and its relatively lower risk from flood. In Kamrup, the yield of autumn rice increased but the area planted to it declined,

Triennium Season Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer Autumn Winter Summer 1971 -73 22 73 6 28 71 1 10 90 0 39 59 . 2 16 83 1 38 60 2 15 85 0 64 36 0 21 77 2 10 89 0 27 71 2

average 1991-93 13 80 7 32 65 3 13 87 0 38 57 5 9 91 0 32 63 5 20 80 0 45 55 0 24 57 19 11 89 0 25 70 5 1995-97 11 84 6 31 65 4 13 87 0 38 54 8 9 90 1 33 61 6 19 77 4 42 58 0 25 55 20 6 92 1 24 69 7

Overall 17 77 6 31 68 1 12 88 0 39 58 4 11 89 0 36 62 3 19 80 1 50 50 0 24 66 10 10 89 0 26 71 3

1981-83 21 75 4 30 70 0 14 86 0 40 58 2 9 91 0 38 61 1 19 81 0 48 52 0 25 70 4 9 90 0 27 72 2

Dibrugarh

Goalpara K.Anglong Kamrup Lakhimpur NC Hills

Nagaon

Sibsagar

Assam

149

Table 9. Growth in rice area and yield by season and district, 1971-97. Districts Autumn rice (%) Area Cachar Darrang Dibrugarh Goalpara K.Anglong Kamrup Lakhimpur N.C.Hills Nagaon Sibsagar All Assam -3.0*** 1.3*** 2.6*** 0.3* -0.5 -0.5* 3.4*** -0.3 2.6*** -0.2 0.4 Yield 0.8 1.0* 2.3* 0.9* 0.1 1.2* 0.3 2.7*** 2.7*** 1.2** 1.1*** Winter rice (%) Area 0.9*** 0.5*** 1.2*** 0.1 2.3*** 0.8*** 1.8*** 1.6*** 0.2 0.8*** 0.8*** Yield 2.3*** 1.0*** 1.2*** 1.2*** 0.6* 1.3*** 0.4* 2.5*** 1.3*** 1.7*** 1.3*** Summer rice (%) Area 0.8* 10.5*** 8.9* 6.7*** 0.4 6.7*** 16.3*** 13.3*** 10.3*** 6.3* 7.0*** Yield 0.7 0.7 1.7** 0.0 1.9** 2.1** 3.2 3.1*** 2.0** 1.7** 1.5*** Total Rice (%) Area 0.2 0.9*** 1.3*** 0.4*** 2.0*** 0.5*** 2.2*** 0.7*** 1.7*** 0.7*** 0.9*** Yield 1.9*** 0.9*** 1.3*** 1.7*** 0.6* 1.6*** 0.3 2.8*** 1.8*** 1.7*** 1.4***

*** ** * , , = significant at 1%, 5%, and 10% level, respectively.

Table 10. District showing patterns of yield variability in Assam (1971-97). CV range <10% Autumn rice Winter rice Darrang Dibrugarh, Goalpara, Kamrup, Lakhimpur Sibsagar Goalpara Cachar Karbi Anglong Nagaon N.C. Hills Dibrugarh, Goalpara, Karbi Anglong, Kamrup, Nagaon, Sibsagar Summer rice Total rice Darrang Dibrugarh Goalpara Kamrup Lakhimpur Sibsagar Cachar Karbi Anglong Nagaon NC Hills

10-1 5%

16-20%

Cachar Darrang Dibrugarh, Kamrup, Lakhimpur, NC Hills, Nagaon, Sibsagar Karbi Anglong

>20%

Cachar, Darrang, Lakhimpur, NC Hills

indicating a shift from autumn rice to other crops in less productive areas. Some changes in normal ahu area in this district occurred-from to early ahu, from ahu to summer rice and from ahu to jute. After the 1980s, a strong growth in demand for jute led an increase in the price of jute relative to that of autumn rice. As a result, the area of jute expanded, while that of ahu rice decreased, especially in Kamrup, Goalpara, and Nagaon. Similarly, a shift from normal to early ahu would have increased the average yield as the latter is grown mostly under irrigated conditions using HYVs and fertilizer.

The variability in area, production, and productivity of the different rice types across districts did not show any uniform pattern (Table 10). For autumn rice, Karbi Anglong and Goalpara showed the highest and the lowest CV values, respectively. Cachar, Darrang, Dibrugarh, Kamrup, Lakhimpur, NC Hills, Nagaon, and Sibsagar had the CV range of 1620%. For summer rice, Dibrugarh, Goalpara, Karbi Anglong, Kamrup, Nagaon, and Sibsagar showed similar levels of variability (CV range of 16-20%). whereas Cachar, Darrang, Lakhimpur, and NC Hills showed a higher variability (above

150

Table 11. Extent of irrigation and adoption of high-yielding varieties (HYV). District % irrigated % area under HYV area (triennium (triennium 1995-97) 1990-92) Autumn Winter Summer 2 16 2 5 12 10 1 30 17 5 9 92 57 65 27 72 31 12 31 45 33 39 44 47 38 37 61 59 30 54 57 47 48 44 82 62 67 100 75 22 61 93 32 75

All 49 51 42 36 62 51 26 44 61 46 47

Cachar Darrang Dibrugarh Goalpara K.Anglong Kamrup Lakhimpur NC Hills Nagaon Sibsagar Assam

To know whether there is any effect of modem technology (HYV, fertilizer, and irrigation) on yield variability of rice, the CV of rice yield was regressed on the percentage of area under HYVs (HYV), percentage of area under irrigation (IRRIG), and per hectare fertilizer use (FERT). None of the coefficients were significant indicating that adoption of modem technology has not affected yield variability in Assam. CV = 13.91 -0.15HYV + 0.06 FERT + 1.51 IRRIG (0.71) (0.10) (1.94) Adj. R2 = 0.14 As the data on area irrigated, fertilizer use, and area under HYVs are likely to be somewhat unreliable, a second regression equation was estimated with the percentage growth rate of yield as the explanatory variable. Growth rate in yield is likely to be higher in districts where the adoption of modem varieties, irrigation, and fertilizer is rapid. Thus, growth rate in yield can be considered to be a proxy for adoption of improved technologies. A positive but statistically insignificant slope coefficient of the regression of district-level CV of yield on the growth rate of yield indicates that adoption of improved technology has tended to increase the instability of rice yield, but that this effect is not strong (Fig. 3). The differences in the CV of yield across districts are probably mainly due to environmental factors rather than to a differential degree of adoption of improved technologies. Variance decomposition analysis The analysis revealed that the mean area, yield, and production of rice for the state changed between the two periods by 14%, 19%, and 35%, respectively. Average yield increased by more than 20% in Cachar, Kamrup, Nagaon, and Sibsagar. A major expansion in mean area occurred in Dibrugarh (26%), K Anglong (39%), Lakhimpur (40%), and Nagaon (25%). In Nagaon, an expansion in area under summer rice is the major reason for the increase in average rice area. In other districts, increases in autumn

Source of data: Directorate of Agriculture (various years), IRRl (1995).

20%). For winter rice, however, 6 districtsDarrang, Dibrugarh, Goalpara, Kamrup, the least Lakhimpur, and Sibsagar-revealed variability (<10 %) while Cachar, Karbi Anglong, and Nagaon had CV ranging between 10% to 15%. Extent of technology adoption Rice productivity depends on a variety of type of land on which it is grown, factors-the agroclimatic conditions, and level of adoption of improved agricultural practices including the use of HYV seed, fertilizer, and irrigation water. Although private investment expanded the irrigation potential in Assam, the overall irrigated area in this state remains low at 7%, with only Darrang, K Anglong, NC Hills, and Nagaon having more than 10% of their rice area irrigated. The percentage of area under HYVs in different districts is fairly high (the state average was 47% for the triennium 1995-97 (Table 11). The variation in percentage area under HYVs across districts is quite small with the adoption rate for all types of rice being in the range 4050% for most districts. The adoption of highyielding summer rice is high at 70%, with almost 90% and 75% of the summer rice area in Nagaon and Kamrup being planted to HYVs, respectively. These two districts account for over 60% of the summer rice area of the state. The adoption rates of HYVs for autumn and winter rice are below 50%.

151

and winter rice area accounted for most of the area expansion. The variance of area for the state increased statistically significantly, by 190%. However, the variance of area in most districts declined, indicating that the increase in variance of area for the state must have resulted from an increase in area correlation across districts. The increase in yield variance was statistically insignificant for the state as well as for individual districts. The production variance for the state increased by 172%, and this increase was statistically significant. The only district that recorded a significant increase in production variance was Kamrup. Overall, about 16% of the increase in the variance of production for the state was due to the increase in the variance of production within districts and the remaining 84% was due to the increase in covariance among districts. For the state as a whole, changes in area-yield covariances and changes in yield variances accounted for more than 50% of the total change in production variance. Despite the changes in production variance, there was no significant change in relative variability of area, yield, and production (as measured by the respective CVs). This indicates that any change in variance of production was compensated for by a proportional change in the mean. Overall, the growth in rice production in

Assam has not been at the cost of increased relative variability, even though the absolute variability has increased.

Farm-level analysis
In spite of several pieces of land reform legislation, share tenancy is quite widespread in Assam. Table 12 shows the types of tenancy and different tenurial arrangements of the tenant farm households in the three districts under
Table 12. Some features of farm households in flood-prone areas of the sampled districts. Feature Lakhimpur Kamrup 40 2 56 13 29 2,155 4,518 74 22 4 30,533 Sibsagar 29 70 7 23 1,973 4,729 52 18 30 33,615

38 Percent of tenant farmers Tenurial arrangement (% of farmers) Fixed share with cost Fixed share without cost 38 Variable share with cost 9 Variable share without cost 53 Cost of cultivation (Rs ha-1) Local variety 2,628 High-yielding variety 4,737 Share of different sources of income (%) 48 Rice Other crops a 21 Homesteadb 31 Net farm income (Rs 34,233 household-1)

aKitchen garden, coconut, bamboo and others. bMustard, summer and winter vegetables, sesamum, jute, blackgram, lentil, and potato.

152

study. The sampled farmers were categorized under three distinct types of tenancy based on the proportion of land leased. Four kinds of tenurial arrangements were observed. In the case of “fixed share with cost,” tenants absorb the cost of inputs and pay the landowners a fixed predetermined amount of their produce irrespective of crop success or failure. But in a fixed share without cost, the landowners provide all inputs, except labor, but require a comparatively larger amount of (predetermined) produce from the tenant, irrespective of crop success or failure. Similarly, under the categories of variable share with cost and variable share without cost, the proportion of the share of produce given to the landowners varies from one-tenth to one-fourth and one-third to one-half of the actual produce, respectively. The table shows that the without-cost tenurial arrangement were more prevalent than the with-cost arrangements. This might be due to apprehension of crop losses from floods and the tenant’s intention of not shouldering the input cost in the event of crop failure. In the relatively flood-prone Lakhimpur District, the proportion of farmers with a fixed share contract is lower than that in Sibsagar, where the production environment is more stable. The prevalence of variable share arrangements in the more floodprone district may be due to its risk-reducing feature as the share is renegotiated in the event of crop loss. The percentage share of income from rice to total income’ was very high in all the districts, ranging from 52% in Lakhimpur to 61% in Kamrup. Homestead income likewise plays an important role in compensating losses in farm income in flood-prone areas. In general, the homestead is located in uplands, and such lands are not generally affected by flood. The activities included in the homestead provide safety net by compensating for losses in farm income and by supplementing farm income. A relatively higher share of rice income in the household income in Assam in comparison

with that in other states is indicative of the importance of rice in this state. As the economic benefit from stabilization of rice yield is directly proportional to the share of rice income in the total household income (Pandey et al, this volume), technologies that stabilize rice yield can be expected to be more valuable in Assam than in other states of eastern India.

Risk management strategies
Ex ante strategies These strategies were designed primarily to avoid or minimize losses from flood damage. Farmers developed them on the basis of their experiences, depending on the time and extent of flood in a particular region. The following are the examples of major ex ante strategies used by farmers. Flood escape. Establishment of rice before flood arrives is one of the strategies followed in flood-prone areas. The Regional Agricultural Research Station, Titabar, Assam Agricultural University developed two “flood-escaping” varieties-Luitand Kapili-thatcould be harvested within 90 d. Planting these varieties to enable harvest before flood arrives is a good escape strategy. Flood-tolerant varieties for lowlands. Usually, tall local varieties are grown on medium lowlands and lowlands. These varieties can withstand moderate degrees of submergence because of their tallness. Staggered planting. Planting rice over a period of time with adjustments made such that it does not directly confront flood is also a common strategy. Seedlings are subsequently transplanted in staggered fashion as the flood recedes. The State Agricultural University has developed several varieties with higher production potential for the staggered planting schedule. Fragmentation of landholding. Tiny farms in the region are mostly composed of several parcels. Although these parcels maybe the result

3As the opportunities for earning nonfarm income in the survey area are rather limited, we have ignored non-farm income in these calculations.

153

of inheritance over the years, they may have played a vital role in pooling risk and thereby reducing overall risk. Crop diversification. Crop diversification is one of the mechanisms adopted by farmers to guard against heavy losses under stress conditions. It primarily depends on resource availability and climatic conditions. Farmers usually grow vegetables within the homestead, but in years when sali rice is lost from floods, they do large-scale vegetable cultivation even in the main crop field. Mixed cropping. In lowlands, farmers practice mixed cultivation of ahu and bao rice. Ahu is grown as chance crop. If ahu rice escapes flood, farmers can grow both ahu and bao crop. Even if there is early flood damage to the ahu crop, farmers can at least have some harvest from the bao crop. Ex post strategies Asset disposal. Disposal of assets during difficult times is one important survival strategy used when losses do occur. Asset disposal, although essential for survival in stress conditions, can increase the incidence of poverty in the long run. Farmers may not be able to reacquire productive assets. For example, bullocks cannot be easily acquired after the situation returns to normal. This is likely to affect farmers’ long-term productivity adversely, thus further perpetuating poverty. Off -farm income. Generating income from off-farm jobs is an important strategy for survival under risky situations. Affected farmers, especially marginal and small farmers, temporarily migrate to nearby towns and cities for work, mostly in the unorganized (private) sector. Government help. Farmers also receive help from the governmental and non-governmental agencies during and immediately after floods. But this assistance is primarily aimed at helping them survive during the emergency period, not at increasing their long-term productivity.

Concluding remarks
The rice production system of Assam is characterized by high levels of flood risk and low yields. In spite of the importance of rice in the state’s economy, the yield growth remains lowest among the eastern Indian states. The major features of rice production systems have changed very little over the past 25 years, although the expansion of summer rice has increased in some districts. Expansion of irrigation since mid-1 980s has been instrumental to the rapid expansion of summer rice. Autumn and winter rice, however, dominate rice production as these two contribute to more than 90% of Assam’s total output. Although the area under autumn rice has shown a tendency to decline as a result of a shift to summer rice and to other crops, this rice culture remains very important in Assam and accounts for over 25% of the total output. Productivity of autumn rice is low and technologies are needed to improve its average yield. As the crop generally suffers from flood during its later growth stage, use of shorter duration varieties that mature before floods arrive can reduce risk. Similarly, the early establishment of autumn rice can be an important risk-avoiding strategy. But the success of such a strategy is conditional on the availability of supplemental irrigation. A shift to early autumn rice (early ahu) has occurred only in areas with access to irrigation. Government policies to further expand public and private irrigation systems in Assam can thus help reduce the risk associated with the production of autumn rice. For winter rice, which accounts for more than 70% of the total rice output, yield growth over the past 25 years has been very modest. Several factors. including risk, limit the adoption of modem varieties and fertilizer. One important constraint in Assam is its poorly developed infrastructure. As a result, costs of inputs increase while farm-gate prices remain low. Investments in infrastructure development in

154

Assam are needed as they not only improve the economics of rice production but also reduce overall risk by providing opportunities to diversify agriculture and income. An analysis of changes in instability in rice production in Assam indicates that production growth has not resulted in an increase in production instability at the district and state level. However, the tendency toward an increase in interdistrict production covariance is a cause for concern, as this can lead to increased amplitude of fluctuations in total output. Correlated climatic patterns and the use of similar varieties across districts may have resulted in a covariate movement in production across districts. If the increased covariate movement is caused by similarity of varieties grown, a more decentralized breeding and testing program may be needed to take advantage of local adaptation and thereby reduce output fluctuations through compensating movements across locations. This paper did not address this issue because of a lack of data. The farm-level study indicates that the share of rice income in total income of Assam is around 50%. This is quite high relative to other states of eastern India (Pandey et al, this volume). With such a high share of rice income, economic benefits from stabilization of rice yield can be substantial. Varieties and crop management practices that reduce risk can be instrumental in improving rice yield in Assam. To help develop and target such technologies, a detailed microeconomic analysis of the patterns of adoption of rice varieties in representative locations of Assam is needed. As shown in the farm-level analysis, the yield of boro rice is substantially higher than that of ahu and sali rice. Therefore, if irrigation potential is created in the flood- prone area, especially during the rabi and summer seasons, farmers can use large-scale boro and early ahu rice cultivation and thus avoid planting sali rice in flood-prone situations. These strategies and shift in area will enable the state to increase rice production substantially.

To have systematic information on farmers’ perception of risk and risk management strategies adopted by the farmers in the floodprone areas, a long-term project needs to be initiated. This will enable the researchers to obtain relevant and reliable data that will be of great use for making appropriate policy decisions.

References
Bhowmick BC. 1992. Identification and optimization of resources in major farming systems. Agric. Econ. Res. Rev. 5(1): Centre for Science and Environment. 1996. Floods, flood plains and environmental myths. State of India’s Environment, A Citizen Report, Centre for Science and Environment, New Delhi. Das R. 1995. Farming system approach in developing ideal agricultural production plan under different agro-ecological situations in the North Bank Plains Zone of Assam. PhD thesis, Department of Agricultural Economics, Assam Agricultural University, Jorhat, India. Directorate of Agriculture. Various years. Basic agricultural statistics. Statistical Wing, Directorate of Agriculture, Assam, Khanapara, Guwahati, India. Directorate of Economics and Statistics. Various years. Statistical hand book of Assam. Directorate of Economics and Statistics, Government of Assam, Guwahati, India. Gogoi JK. 1998. An economic estimation of risk in rice cultivation in Upper Brahmaputra Valley Zone of Assam. PhD thesis, Department of Agricultural Economics, Assam Agricultural University, Jorhat, India. Hazell PBR. 1982. Instability in foodgrain production. Research Report 30. Washington, D.C. (USA): International Food Policy Research Institute. Hydrogeology and Ground Water Resources Development. 1980. Perspective of Assam North Eastern Region, March 1980, Central Ground Water Board, Regional Office, Guwahati, Government of India.

155

ICAR. 1981. National agricultural research project. Report of the ICAR Research Review Committee for AAU, Jorhat, Assam, Krishi Bhavan, New Delhi, India. IRRI (International Rice Research Institute). 1995. World rice statistics 1993-94. Los Baños (Philippines): IRRI. Phukan N. 1996. Statistical handbook. Government of Assam, Guwahati, India: Directorate of Economics and Statistics.

Notes
Authors' addresses: B.C. Bhowmick and J.K. Gogoi, Assam Agricultural University, India; S. Pandey and R.A. Villano, International Rice Research Institute, MCPO box 3 127, Makati City 1271, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

Appendix 1. Particulars of agroclimatic zones of Assam. Rice area Area km2 1422 1 5646 8775 16013 7024 8989 5561 5561 20222 10359 9863 6962 6962 15222 10332 4890 78201 % of state share 18 7 11 20 9 11 7 7 26 13 13 9 9 19 13 6 (triennium ending 1997) 000 ha 497 178 319 433 152 281 336 336 906 390 516 222 222 136 122 14 2531 % 20 7 13 17 6 11 13 13 35 15 20 9 9 6 5 1

Agroclimatic

zone/district

North Bank Plains Zone Lakhimpur Darrang Upper Brahmaputra Valley Zone Dibrugarh Sibsagar Central Brahmaputra Valley Zone Nagaon Lower Brahmaputra Valley Zone Goalpara Kamrup Barak Valley Zone Cachar Hills Zone Karbi Anglong N.C.Hills Assam
a

Geographical area as per 1971 census. Source: CAR (1981).

156

Risk and its management in the rainfed rice ecosystem of Bihar
J. Thakur

Rice production in Bihar is characterized by high yield and area variability. The high risk associated with rice production in Bihar has led to sluggish growth in productivity and even an absolute decline in the area under rice. Based on the analysis of time-series data covering the period 1973-74 to 1995-96, the district-level coefficients of variation of yield and area were as high as 51% and 36%, respectively. Production variability was quite high in all agroclimatic zones but more so in the South Bihar Plains, which is characterized by submergence. The adoption of modern varieties was limited except in irrigated areas. The analysis indicates that rice technologies that can impart yield stability in the highly unstable hydrological conditions of Bihar are needed to increase rice productivity in this relatively poorer state of India.

Agricultural production is subject to a large degree of risk, both natural and institutional. This is more so in some of the underdeveloped states, such as Bihar, where vagaries of nature are more pronounced and where less has been achieved in providing a secure base for agriculture through irrigation facilities, crop protection measures, and the like. Further, a weak and inadequate marketing system exacerbates the situation. Besides, no organized insurance exists to distribute risks and protect individual farmers from huge losses. The nature of these risks can be classified into three categories (1) flood and drought caused by inadequate and uneven distribution of rainfall, (2) occurrence of pests and diseases, and (3) economic and marketing factors such as instability of prices, marginal profits, parity prices among competing crops, market demand, availability of storage and transport facilities, and others. In addition, fluctuating crop yields have become a regular feature and farmers have no alternative except to succumb to the vagaries of nature. Bihar is the fifth largest state of India in size (5.3% of the total geographical area) and the second most populous state (more than 10% of the total population). It has remained by and

large rural and about 85% of the population lives in villages. Agriculture is the main occupation in the rural areas. About 79% of the working population directly depends on agriculture. Agriculture alone contributes nearly 45% to the state's net domestic product. The main objectives of this study are (1) to measure variability in area, production, and yield of rice in Bihar, (2) to estimate the growth rate in area, production, and yield of rice, (3) to quantify the extent of yield losses caused by various factors, and (4) to identify farmers' risk management options.

Rice production scenario
Rice is the staple food of the people of Bihar and an important component in shaping the economy of the state. Bihar grows nearly 5.0 million ha of rice, which accounts for about 12% of the total rice area of the country. Rice covers about 56% of the total gross cropped area of the state. However, crop productivity averages 1.9 t ha" (rough rice) in comparison with the national average of 2.5 t ha". About 66% of the rice area still depends on rainfall and is vulnerable to droughts and floods. In addition, it is cultivated under highly variable and heterogeneous

157

conditions. If we look at the growth trend of area, production, and yield of rice over the period and compare it with that of population growth in the state, the picture is gloomy. Growth in area has declined by 0.42% during 1973-74 to 1995-96. This is attributed to the very unstable nature of rice cultivation in the state because of uncertain monsoon. Farmers sometimes experience severe droughts; other times, they contend with devastating floods. Yield growth consequently remained at a low 1.3% y-1 during 1973-74 to 1995-96. Population, on the other hand, increased at a much faster rate (2.4% y-1), with the gap between rice production and population becoming wider over the years. During the entire period, rice production failed to keep pace with population growth.

the natural levee of the Ganga is a vast stretch of backwaters known as tal lands. The floodplains of Ganga, which get reworked, eroded, and deposited at regular intervals, are higher than tal lands and known locally as diara lands. Average annual rainfall is reported to be only 1,110 mm in this zone. The Chotanagpur Plateau comprises the districts of Hazaribagh, Giridih, Santhal Parganas, Dhanbad, Palamu, Singhbhum, and Ranchi. This zone is characterized by a humid to subhumid tropical monsoon-type climate. Average annual rainfall is 1,321 mm.

The rice ecosystems in Bihar
Rainfed uplands
Tanr lands: Plain to slopy (0-30%) fields without surface water accumulation; loose fragile soils with poor waterholding capacity. Tanr I: Land adjacent to dwellings and used mainly for growing maize, vegetables, and rice nursery; an extended kitchen garden with supplemental irrigation from wells. Tanr II: Leveled land away from the village but with good soil depth; unbunded uplands. Tanr III: Sloping and stony land away from villages with very little soil depth and little or no capacity to retain moisture; the most fragile upland rice/millet system.

Characteristic features of agroclimatic zones
Bihar is divided into four agroclimatic zones: the Northwest Plains, the Northeast Plains, the South Bihar Plains, and the Chotanagpur Plateau. The Northwest Plains comprise the districts of West and East Champaran, Gopalganj, Siwan, and Saran. The climate is tropical humid to subhumid. The northern districts of East and West Champaran receive more than 1,300 mm of rainfall. The land is mostly alluvial plains sloping toward the southeast with a very low gradient as evidenced by the direction in which the rivers flow. The soils under the influence of Gandak, Burhi Gandak, and Ghaghra rivers are mostly calcareous with varying amounts of lime in them. The Northeast Plains comprise the districts of Sitamarhi, Muzaffarpur, Vaishali, Darbhanga, Madhubani, Samastipur, Pumea, Katihar, Saharsa, and Begusarai. This zone has a tropical monsoon climate and can be classified as humid to subhumid. Average annual rainfall is 1,405 mm. The South Bihar Plains comprise the districts of Gaya, Aurangabad, Rohtas, Bhojpur, Patna, Nalanda, Nawada, Monghyr, and Bhagalpur. This zone covers the alluvial plains of the Ganga River on its southern side and sediments are received from the river. South of 158

Rainfed lowlands
Shallow rainfed: Flat lands as well as terraced rice fields with a water regime of 025 cm. Has two subcategories: (1) drought-prone and (2) drought- and flood- prone. Intermediate rainfed: Water depth ranges from 25 to 50 cm. Areas receiving normal rainfall, where intermittent flooding is a problem, and areas receiving erratic rainfall, where droughts and floods are major constraints.

Semideep rainfed: Low-lying, flat, unbunded illdrained areas, where water stagnates for 4-5 mo during the wet season; water regimes vary from 50 to 100 cm. Deepwater: Water depth varies from 50 to 200 cm. Soils are heavy and rich in organic matter. Direct seeding is generally followed, but, in areas where water does not dry out even in May, transplanting is done. Irrigated lands: In much of this system, rainfall supplements irrigation water. Drainage in the irrigated system is generally good, but, in certain areas, waterlogging becomes a constraint.

Table 1. Mean value of rice area, production, and yield by agroecological zones (1973-74 to 1995-96). Zone Period I Period II 1973-74 to 1983-84 to 1982-83 1995-96 Percentage change over period I 2.9 -2.8 0.8 -12.3 -4.2

Area (000 ha) Northwest Plains Northeast Plains South Bihar Plains Chotanagpur Plateau Bihar Yield (t ha-1) Northwest Plains Northeast Plains South Bihar Plains Chotanagpur Plateau Bihar Production (000 t) Northwest Plains Northeast Plains South Bihar Plains Chotanagpur Plateau Bihar

673 1,597 1,342 1,667 5,279

692 1,552 1,353 1,462 5,059 1.8 1.4 2.0 1.5 1.6

1.3 1.2 1.5 1.3 1.3 852 1,873 2,036 2,155 6,918

39.8 21.4 33.6 13.9 25.0 43. 17.98 18.0 33.5 -1.1 19.8

Data and methods
The analysis is divided into two sections. The first section is based on the analysis of secondary data at state, district, and agroclimatic zone levels from 1973-74 to 1995-96. The timeseries data were divided into two segments with the first covering 1973-74 to 1982-82 and the second covering 1983-84 to 1995-96. Changes in production variability and growth rates between these periods are considered. The choice of this period was based mainly on the availability of a consistent set of data for 31 districts. The second section presents the findings of microlevel analysis based on farmlevel data from different ecosystems of the state. Time-series data for various years were obtained from the Directorate of Economics and Statistics, Government of Bihar.

1,224 2,209 2,719 2.1 30 8,286

Results and discussion
Estimation of growth trend The contribution of the Northwest Plains is the lowest in terms of both rice area and production, although its share has increased in the more recent periods (Table 1). The other three zones account for most of the rice area and production in Bihar. The average rice area declined by about 4.2% during the second period in comparison with the first period. This decline was mainly

attributed to a decline in area in the Northeast Plains (2.8%) and in Chotanagpur Plateau (12.3%). For yield, the Northwest Plains and South Bihar Plains recorded increases of about 40% and 33%, respectively, during the second period. A moderate increase in yield was noticed for the Northeast Plains and Chotanagpur Plateau during the same period. A similar trend was found for rice production, except in the Chotanagpur Plateau, where average production declined by 1.1% compared with the first period (mainly because of a decline in rice area). Table 2 shows that the rice area of the 18 districts (accounting for 58.1% of the total rice area) had negative growth ranging from -0.04% in Aurangabad to -1.97% in Dhanbad. Only five districts recorded a positive area growth in the range of 1-2% annum-1; the remaining eight districts experienced area growth of less than 1 % annum-1. Yield growth was estimated to be above 2% annum-1 for only five districts: West Champaran, Gopalganj, Pumea, Bhojpur, and Rohtash. All these districts represent the irrigated rice ecosystem of Bihar. A significant positive yield growth in Pumea District is caused by the expansion of area under boro (summer) rice, which gives a higher yield per unit area. In fact, boro rice is expanding to a certain extent also in Saharsa, Pumea, Katihar, 159

Table 2. Growth rates of rice area, production, and yield in Bihar (1973-74 to 1995-96): District Northwest Plains West Champaran West Champaran Gopalganj Siwan Saran Northeast Plains Sitamarhi Vaishali Muzaffarpur Darbhanga Madhubani Samastipur Begusarai Saharsa Purnea Katihar South Bihar Plains Munger Bhojpur Rohtash Patna Gaya Nawada Aurangabad Nalanda Bhagalpur Chotanagpur Plateau Hazaribagh Ranchi Santhalpargana Dhanbad Palamu Singhbhum Giridih
a ***,

Table 3. Growth rates of rice area, production, and yield by agroecological zone (1973-74 to 1995-96).a Zone Northwest Plains Northeast Plains South Bihar Plains Chotanagpur Plateau Bihar
a***,**

Area

Production (%) 1.96*** 1.25** 3.97*** 3.37*** 2.11* -0.47 ns 0.34 ns 1.01 ns 0.87 ns -0.1 7 ns 1.92* 3.48** 1.88** 2.83*** 0.78 ns -0.38 ns 2.19*** 2.98 -0.30 ns 1.05 ns -1.32* 1.78*** -0.68ns 1.17* -1.24 ns -1.05 ns -1.32** -1.69* -0.70 ns -0.30 ns -0.38 ns

Yield

Area 0.1 ns -0.4 ns 1.1 ns -1.1*** -0.4**

Production (%) 2.2*** 0.7 ns 2.0*** 0.8 ns 0.9 ns

Yield 2.2.9. 1.1** 1.9*** 0.3 ns 1.3***

-0.95*** -0.12 ns 1.33*** 1.56*** 0.25 ns -1.07*** -0.40** 0.61* .023 ns -0.60*** 1.08*** 2.36*** 0.64* -0.70*** 0.49* -1.32*** 0.02 ns 1.04* -1.28*** -0.16 ns -1.02** -0.0 4 ns 0.22 ns -0.43* -1.05*** -1.35*** -1.28*** -1.97*** 0.42 ns -0.68*** -0.70*

2.85*** 1.43*** 2.68*** 1.70** 1.87** 0.56 ns 0.84 ns 0.41 ns 0.43 ns 0.82** 0.88 ns 0.92 ns 1.33** 2.10*** 0.47 ns 0.90 ns 2.23*** 2.10*** 1.6** 1.08* -0.47 ns 1.89*** -0.89 ns 1.54** -0.27 ns 0.41** -0.02 ns 0.20 ns -0.30 ns 0.33 ns 0.30 ns

= significant at 1% and 5% level, respectively. ns = not signifcant.

population growth rates, even for the Northwest Plains and South Bihar Plains, both of which registered positive yield growth rates.

Area, production, and yield variability
The coefficient of variation, which gives the relative change in variability, was estimated to examine the area, production, and yield variability of rice. The basic data comprise timeseries information on rice area, production, and yield for the districts of Bihar during 1973-74 to 1995-96. Here, yields are estimated by dividing production by area sown. To estimate the coefficient of variation, the data were first linearly detrended. The coefficient of variation of rice area of the 14 districts (46% of total rice area of the state) was less than 10% (Table 4). However, area variability was high and exceeded 20% in Madhubani, Rohtash, and Nalanda. Madhubani is a submergence-prone district; both Rohtash and Nalanda have unreliable irrigation based mainly on canal and lift irrigation, respectively. Variability of rice yield was between 10% and 20% for two districts (8% of total rice area), East and West Champaran (Table 4). The yield Variability for 23 districts covering 80% of the total rice area was between 20% and 30% and about 12% of the state's rice area experienced a yield variability of more than 40%. Overall, the district-level yield variability is quite high in Bihar, partly due to the high variability of rainfall (Table 5). Table 6 presents analyses of variability for agroclimatic zones. The South Bihar Plains exhibited the highest variability in area and quite a high variability in yield also. Production variability across the agroecological zones was highest (23%) for the Northeast Plains and

**, * =significant at 1%. 5%. and 10% level, respectively. ns = not significant.

Darbhanga, and Madhubani districts of the Northeast Plains. Twenty-one other districts reported positive yield growth, ranging from 0.2% in Dhanbad to about 1.9% in Saran and Aurangabad. Again, the latter district having a slightly higher yield growth represents the irrigated rice ecosystem of the state. A similar trend was found for production growth. At the aggregate level, the area growth for the state as a whole was negative (-0.4%) because of the negative area growth in the Chotanagpur Plateau (Table 3). The yield growths for the Northwest Plains and South Bihar Plains were positive and significant. The Chotanagpur plateau experienced insignificant yield growth during the same period and the Northeast Plains registered a yield growth of 1.1 %. The yield growth rates were less than the 160

Table 4. Mean and coefficient of variation (CV) of rice area and yield in Bihar (1973-74 to 1995-96). District Area Mean (000 ha) 201 213 90 93 86 142 59 141 103 176 75 18 203 270 123 161 192 300 112 232 97 132 130 152 123 342 349 56 89 335 95 CV (%) 7 9 8 10 8 9 6 12 17 23 12 15 11 15 10 17 6 31 9 17 16 13 36 8 8 11 8 17 20 5 12 Mean (t ha-1) 1.8 1.4 1.5 1.5 1.5 1.3 1.4 1.2 1.1 1.4 1.5 1.3 1.2 1.3 1.3 1.5 2.3 2.1 1.5 1.7 1.4 1.8 1.5 1.3 1.4 1.5 1.4 1.3 1.3 1.3 1.4 Yield CV (%) 18 18 23 26 26 29 43 25 29 51 47 29 26 37 34 32 32 29 26 32 45 25 30 36 35 31 23 26 50 24 42

Table 5. Coefficient of variation of rainfall in Bihar (197374 to 1992-93). Month June July August September
a

Agroecological zone a NWP 61 42 50 55 NEP 56 38 57 59 SBP 73 44 45 59 CHP 92 76 77 83 Bihar 71 50 58 64

Northwest Plains West Champaran East Champaran Gopalganj Siwan Saran Northeast Plains Sitamarhi Vaishali Muzaffarpur Darbhanga Madhubani Samastipur Begusarai Saharsa Purnea Katihar South Bihar Plains Munger Bhojpur Rohtash Patna Gaya Nawada Aurangabad Nalanda Bhagalpur Chotanagpur Plateau Hazaribagh Ranchi Santhalpargana Dhanbad Palamu Singhbhum Giridih

NWP = Northwest Plains, NEP = Northeast Plains, SBP = South Bihar Plains, CHP = Chotanagpur Plateau.

Table 6. Mean, standard deviation, and coefficient of variation (CV) of rice area, production, and yield by agroecological zone (1973-74 to 1995-96). Zone Area (000 ha) Northwest Plains Northeast Plains South Bihar Plains Chotanagpur Plateau Bihar Yield (kg ha-1) Northwest Plains Northeast Plains South Bihar plains Chotanagpur Plateau Bihar Production ( 000 t) Northwest Plains Northeast Plains South Bihar Plains Chotanagpur plateau Bihar
a

Mean a (000 ha) 684 (13) 1,511 (30) 1 ,348 (27) 1,507 (29) 5,051 1,558 1,309 1,797 1,380 1493 1,070 (14) 1,979 (26) 2,435 (32) 2,088 (27) 7,565

CV (%)

5 9 10 6 6 15 17 17 18 14

lowest ( 18%) for the Northwest Plains. While the Chotanagpur Plateau represents droughtprone upland environments, the South Bihar Plains and Northeast Plains represent submergence-prone environments. Both the area and yield variability of these zones are accordingly higher than those of the Northwest Plains, which represent mostly the irrigated environment. Adoption of modern varieties Farmers of Bihar grow some modem varieties, although the rates of adoption are quite low except in irrigated environments (Table 7). A majority of farmers (48-60%) still grow local varieties of rice under rainfed and upland ecosystems. This suggests that farmers in these ecosystems are unwilling to adopt modem

18 23 21 21 18

Numbers in parentheses indicate percent of the respective total.

varieties probably because of the greater risks involved. Estimating yield loss Yield loss is estimated based on the average loss associated with the presence of technical 161

Table 7. Adoption of modern rice varieties in Bihar (% of farmers). Ecosystem Local variety only (LV) 60 48 10 39 Modern variety only (MV) 5 63 23 Both LV and MV 40 48 28 38

Risk management options Risk under a rainfed rice ecosystem refers to the situation when the crop suffers losses as a result of flood and/or drought. Apart from this, sudden outbreaks of pests or diseases cause severe losses to the standing rice crop. In this situation, farmers seek risk management options to recover from such losses. Farmers surveyed during the cropping year 1996-97 used the following risk management options. Replanting of rice after early recession of floodwater from the field. Replanting is a common practice to cope with early loss of seedlings. Farmers grow more seedlings than necessary in their nurseries so that these extra seedlings can be used for replanting if the need arises. Seedlings are also borrowed or purchased from neighboring and sometimes even distant villages. Uprooting of transplanted rice from the unaffected plot to other plots is another method used. This practice is popularly known as kharuhan. Changing the cropping pattern. When floodwater recedes late, resource-rich farmers plant a wheat crop after proper tillage operations, whereas resource-poor farmers prefer to grow a para crop such as lathyrus, lentil, linseed, rapeseed, and mustard. Temporary migration. As agriculture is the main source of income for the rural poor of Bihar, marginal farmers and landless laborers migrate to cities to supplement their family incomes. However, family income is also generated through dairy enterprises.

Rainfed upland Rainfed lowland Irrigated Bihar

Numbers in parentheses indicate percent of total farms. Source: Field survey.

constraints, the percentage of area affected by a constraint, and the probability of occurrence of a constraint. Total loss was estimated by multiplying average yield loss by total area under a particular production system. Loss estimates were obtained through a survey of researchers, farmers, and extension workers. The methodology used to generate and analyze the data is detailed in Thakur and Hossain (1998). Table 8 shows estimates of yield loss from technical constraints for the three different rice ecosystems in Bihar. Results showed maximum yield loss (972 kg ha") in the rainfed lowland ecosystem followed by the irrigated ecosystem (943 kg ha"). The minimum loss (785 kg ha") was noted in the rainfed upland ecosystem. The results further indicated that the estimated yield losses from technical constraints across ecosystems are substantial and account for a large portion of yield forgone. The loss estimates also showed big differences in magnitude among the constraints. The yield loss caused by insects, pests, and weeds was maximum (39%), followed by diseases (24%), climate and environment (15%), soil-related problems (11%), and physiological and genetic problems (10%).

Table 8. Estimates of yield loss (kg ha-1) from technical constraints by constraint group and ecosystem.a Ecosystem Insects, pests, and weeds 363 310 41 3 358 (39.0) Diseases 186 234 245 227 (24.5) Climate/ environment 80 194 100 136 (15.0) Soils 78 106 119 104 (11.0) Physiological and genetic 78 128 66 95 (10.5) Total 785 972 943 920 (100.0)

Rainfed Rainfed Irrigated Bihar

upland lowland

a

Rough rice. Source: Thakur and Hossain (1998). Numbers in parentheses indicate percent of total yield loss. Source: Field survey.

162

Conclusions
This investigation on risk and its management in the rainfed ecosystem of Bihar indicates that the state's rice area has declined by about 4.17% during 1983-84 to 1994-96 mainly because of the decline in rice area in the Northeast Plains and Chatanagpur Plateau. Production variability of rice was much higher than area variability, which was also quite large. Yield variability across the agroecological zones was the highest for the Northeast Plains. A yield growth of 2% annum-1 was observed only for the five irrigated districts of West Champaran, Gopalganj, Purnea, Bhojpur, and Rohtash. As to the extent of adoption of modem rice varieties, 62.5% of the farmers in the irrigated ecosystem cultivate only modem varieties in their fields. Many farmers (47% to 60%) still grow local varieties under the rainfed and upland rice ecosystems. Estimation of yield loss from biotic and abiotic constraints showed that the highest loss occurred in the rainfed lowlands. These findings suggest the need for more stable varieties and crop management technologies that can stabilize rice yields in the rainfed environments of Bihar.

Reference
Thakur J, Hossain M. 1998. Yield gaps and constraints to rice production in Bihar, eastern India. Paper presented at the International Workshop on Prioritization of Rice Research in Asia, 20-22 April 1998, International Rice Research Institute, Los Baños, Philippines.

Notes
Author's address: Rajendra Agricultural University, Bihar, India. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 2 1-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute. 163

Risk and its management in rainfed rice ecosystem of West Bengal
N. K. Saha, S. K. Bardhan Roy, and U. S. Aich

West Bengal experienced a rapid change in rice production as the use of modern varieties expanded with the increase in tubewell irrigation and shift to boro rice systems. This paper analyzes the pattern of variability in rice production in West Bengal and documents evidence on farmers risk- coping strategies that are based on manipulating cropping systems. The paper highlights the need to further improve rice productivity and diversify rice production systems to increase farmers' income as well as reduce overall risk.

West Bengal is the most densely populated State in India (population density 767 persons km2). It covers 8.88 million ha (2.7% of nation's geographical area) and has a population of 68.07 million (more than 8% of country's population) (1991 census). Seventy-five percent of population lives in rural areas (in 37,910 villages) and depends on agriculture, directly or indirectly, for livelihood. The agricultural sector
Table 1. Agroclimatic zones in West Bengal, India. Agroclimatic zone/subregion Districts covered

Eastern Hills and Plateau Hills Terai Plateau Lower Gangetic Plains Old Alluvium New Alluvium Lateritic Coastal saline

Darjeeling Jalpaiguri and Purulia

Coochbehar

North Dinajpur, South Dinajpur. and Malda Murshidabad. Nadia. 24-Parganas (North), Hooghly, and Burdwan Birbhurn, Bankura, and Midnapore (West) 24-Parganas (South), Midnapore (East), and Howrah

Table 2. Rice harvested area by season in West Bengal, India, 1996-97. Season Autumn (aus) rice Winter (aman) rice Summer (boro) rice Total rice Area (million ha) 0.462 4.282 1.056 5.800

is characterized by a predominance of marginal and small farmers (more than 91%) and scattered and fragmented holdings (average size 0.90 ha). The state is divided into two broad agroecological zones and seven subregions on the basis of differences in soil characteristics, topography, rainfall, and temperature. Table 1 shows the geographical boundaries covered by these zones/subzones. Rice covers 5.8 million ha of land in West Bengal. It grows across different agroecological environments in three seasons-prekharif (aus), kharif (aman), and summer (boro). The major area (74%) is planted with wet-season rice (aman), followed by dry-season boro (1 8%) and prekharif aus (8%). The dry-season rice grows under irrigated conditions, while major area during prekharif and kharif are rainfed (Table 2). In the wet season, 32% of the rice area is irrigated. Of the remaining area, which is rainfed, 52% is lowland and 16% is flood-prone. In the dry season, irrigated area consists of 30% upland (Table 3). Rice yield varies widely across seasons and ecosystems in different agroclimatic zones in West Bengal (Table 4). Growth and variability in rice production The annual compound growth rate of area, productivity, and production of rice by

165

Table 3. Distribution of rice area (as percentage) by culture and season in West Bengal, India, 1997. Rice culture Irrigated Rainfed Lowland Flood-prone Upland Wet season 31.7 68.3 52.5 15.8 Dry season 69.6 30.4 30.4

agroclimatic zone revealed a differential pattern of growth (Table 5). During the last one decade (1986-87 to 1996-97), the growth in rice harvested area was only 0.79% in the State. However, the harvested area in the Hills, Terai, and Old Alluvium regions has declined as farmers diversified to other crops. Overall productivity in the state during this period
Table 4. Rice productivity (t ha-1) Agroclimatic zone Eastern Hills and Plateau Hills Terai Plateau Lower Gangetic Plains Old Alluvium New Alluvium Lateritic Coastal West Bengal
a

registered growth of 2.65%, with production growth of 3.46%. The maximum production and productivity growth came from the Old Alluvium region. followed by the New Alluvium and coastal saline regions. Table 6 shows the growth rate in harvested area, productivity, and production (by district). Table 7 presents the instability in rice harvested area and productivity at the district level from 1986-87 to 1996-97 as measured by the coefficient of variation (CV). Based on the aggregated CV (area and productivity), West Bengal can be broadly classified into two risk levels. The districts having a CV of harvested area more than 3% and the yield CV of more than 7% have been defined as high-risk areas.

a

by ecosystem and agroclimatic zones in West Bengal, India, 1997. Irrigated Wet 2.9 3.9 3.0 3.4 3.6 3.4 3.6 3.5 Dry 4.0 4.6 4.0 4.6 4.7 4.7 4.7 4.7 Upland 2.4 2.5 2.6 2.5 2.5 2.5 2.6 2.5 Rainfed Lowland 2.6 2.8 2.8 2.9 2.9 2.9 2.9 Flood-prone 2.5 1.9 2.2 2.1 2.4 2.3 2.8 2.9 2.9 3.3 3.5 3.2 3.4 3.3 Total

In terms of rough rice.

Table 5. Annual average compound growth rate of area, productivity, and production of rice by agroclimatic zones in West Bengal during 1986-87 to 1996-97. Agroclimatic zone Growth rate (%) Area Productivity 1.95 1.98 1.22 4.39 2.64 1.99 2.58 2.65 Production -1.35 1.04 2.52 4.33 3.63 3.44 3.94 3.46

Table 6. Annual average compound growth rate of rice area, productivity, and production during 1986-87 to 1996-97. District Jalpaiguri Dinajpur (N&S) 24-Parganas (N) Burdwan Midnapore (W) Midnapore (E) Malda Murshidabad Nadia Hooghly Bankura Purulia Darjeeling Coochbehar 24-Parganas (S) Howrah Birbhum West Bengal Growth rate (%) Area -1.12 0.12 -0.11 1.61 1.57 1.3: -0.40 0.24 1.09 1.54 1.61 1.28 -3.23 -0.74 1.51 0.60 0.97 0.79 Productivity 2.17 4.34 2.38 2.04 1.90 2.87 4.54 4.38 2.57 2.23 3.10 1.22 1.95 1.84 3.42 -0.78 0.99 2.65 Production I .03 4.47 2.27 3.68 3.50 4.29 4.12 4.64 3.68 3.80 4.76 2.52 -1.35 1.09 4.98 -0.18 1.97 3.46

Eastern Hills and Plateau Hills -3.23 -0.92 Terai Plateau 1.28 Lower Gangetic Plains Old Alluvium New Alluvium Lateritic Coastal saline West Bengal -0.05 0.96 1.42 1.33 0.79

Sources of data: Directorate of Agriculture (various issues)

166

Table 7. Coefficient of variation (CV) of area and productivity of rainfed rice from 1986-87 to 1996-97. District Darjeeling Jalpaiguri Coochbehar Dinajpur (N & S) Malda Murshidabad Nadia 24-Parganas (N) 24-Parganas (S) Howrah Hooghly Burdwan Birbhum Bankura Purulia Midnapore (W) Midnapore (E) West Bengal Coefficient of variation (%) Area 10.11 2.44 5.69 1.72 4.26 4.75 3.09 1.13 3.35 7.17 4.89 2.79 5.07 3.48 3.17 1.78 2.41 1.45 Productivity 14.28 11.60 5.35 8.24 6.84 8.51 7.17 8.62 12.05 12.25 5.25 3.71 6.41 3.75 6.64 7.91 10.01 4.62

Table 9. Harvest prices of rice, in West Bengal, India, 198081 to 1995-96. Year 1980-81 1985-86 1990-91 1995-96 Harvest price a (Rs t-1) 2,104 3,169 4,669 7.584 Real price (Rs) 2,758 3,686 5,147

a In terms of rough rice. Sources: Farm Household Surveys, Evaluation Wing, Directorate of Agriculture, Government of West Bengal.

Table 10. Yield loss caused by abiotic stresses in rainfed rice (kharif) a in West Bengal. Abiotic stress Kharif rice Probability Production loss Av yield of in affected loss occurence area (kg) (kg ha-1) 0.35 0.46 0.35 0.40 0.44 0.38 0.15 0.31 541 408 682 345 467 615 726 837 189 188 239 138 205 234 109 260

Drought At seedling stage At vegetative stage At anthesis Submergence At seedling stage At vegetative stage At anthesis Cold At anthesis Cyclone/strongwind

Table 8. Classification of district by risk level in West Bengal, India. Risk levela High 1. Dajeeling 3. Nadia 5. Howrah Moderate to low Districts 2. Murshidabad 4. 24-Parganas (South)

2. Jalpaiguri 1. Coochbehar 3. Dinajpur (N & S) 4. Malda 5. 24-Parganas 6. Purulia (North) 7. Midnapore (West) 8. Midnapore (East) 9. Hooghly 10. Burdwan 11. Birbhum 12. Bankura

a In terms of rough rice. Source: Farm Household Survey. Evaluation Wing, Directorate of Agriculture, Government of West Bengal.

a CV of harvested area >3% and CV of productivity more than 7% = high: if otherwise = moderate to low.

districts not falling in this category are considered moderate to low-risk areas. Of 17 districts, five have been identified as high-risk and the rest moderate to low risk (Table 8). Output prices of rice also vary over the years. Output prices have increased more than 260% during last 15 years. However, a sharp increase in output prices (62%) occurred from 1990-91 to 1995-96 (Table 9). The sharp rise in input prices, particularly labor wages and fertilizer, as well as the increased demand for rice contributed to the increase in output prices. Risk in rainfed rice has mainly been analyzed on the basis of yield losses from

various abiotic stresses. Estimates of yield losses were taken from the household survey conducted in 1997 (Table 10). Among the abiotic stresses, drought at the seedling and anthesis stages caused a severe yield loss. Submergence at vegetative and anthesis stages also contributed to yield losses. Cold temperature at the anthesis stage in late-planted aman rice also resulted in a substantial yield loss in flood-affected areas. Cyclones/strong winds in the coastal areas are considered major causes of yield loss in rainfed rice. In the monocropped rainfed rice area, changes in land use pattern took place over the decade. Exploitation of surface water and groundwater through shallow tubewells using backfeed river tide water through canals, particularly in the coastal part, contributed to a rapid shift to boro rice cultivation. Boro rice is more productive and its share in the total rice production of West Bengal has increased over 167

time, thus helping stabilize production somewhat, as it takes advantage of the early premonsoon rains. This practice also minimizes irrigation cost. Managing risk Rainfed rice farmers generally adopt certain risk-coping mechanisms to obtain the minimum grain yield under varying levels of risks. Some of these are adaptations to particular characteristics of the ecosystem that do not change from year to year. Moisture stress in uplands, water stagnation and submergence in lowlands are examples of such characteristics. The farmer's management strategy to cope with uncertainties of rainfall and abnormal flooding may be considered as an adaptation to risk. One of the risk management strategies is to grow varieties suitable to specific environmental conditions (Table 11). Rainfed uplands The risk-prone unfavorable uplands of West Bengal are undulating red and lateritic soils of Midnapore (West), Bankura, part of Birbhum, and Burdwan districts. Rainfall in these areas varies from 1,125 to 1,410 mm spread over 90100 d (mid-June to September) with a short drought spell in between. Unlike upland farmers in the adjacent states of Bihar and Orissa, farmers in these areas do not practice mixed cropping during kharif, but they prefer to transplant a sole crop of drought-tolerant earlyduration rice varieties. Farmers themselves cope with risk by adjusting the establishment of nursery bed at different dates to match the onset of rain. But nursery sowing cannot be extended beyond 15-20 d from the optimum time (early June) since the distribution period of rain is short and delayed planted rice fails to produce grain. In such cases, farmers have to abandon kharif rice cultivation. This happened in 1998 kharif, when 70% of the upland area was not planted to rice. The rainfall figures for June and July of 1998 were 58% and 43% below the average of 1992-97.

Rainfed lowlands Rainfed lowlands are distributed mainly in Gangetic floodplains of the state. Rainfall varies between 1,300 and 1,600 mm with 80% precipitation occuring from June to October. Intermittent flash flooding, and sudden submergence/water stagnation are the risk factors associated with heavy precipitation, causing total or partial damage to the established crop. These contribute substantially to yield loss. Nursery sowing in different dates (so that farmers can replant after floods) and the use of overaged rice seedlings are some of the riskcoping mechanisms used (Table 12). Double transplanting/splitting of tillers from the crop that survived is a compensatoty strategy.

Table 11. Varietal condition. Varietal option 1.

management for environmental stress Rice varieties that can cope with isk Aditya, Lalat, Ananda, Sneha. Sati, Jhular, Satia, Rasi Heera, Latat, IET4786, IR36

management

Extra early duration rice varieties for upland Early duration rice varieties for preflood harvest Long duration submergence tolerant varieties for flash-flood areas (30-50 cm water depth) Varieties for deep flooding/stagnant water areas Saline tolerant varieties for coastal areas Varieties for flash flood areas

2.

3.

Pankaj, Swarna, Bipasa, Suresh, Biraj, Jogen

4.

Sabita, Dinesh, Purnendu. Jitendra Mohan, Utkalprava, Suresh, SR-26B, Non-bohra A mixture of aus and aman rice varieties to ensure harvest of at least a single crop, depending on early and late flood Boro harvest release land late for floating rice establishment. Seeding floating rice in boro crop improves stand establishment.

5.

6.

7.

Boro-floating rice integration for proper stand establishment of floating rice before arrival of flood in flood-prone areas

168

Table 12. Managing risk in rainfed rice culture in West Bengal, India. Rainfed rice culture Upland Broadcasting Drought at different stages of crop growth I. Use of high seed rate 2. Sowing very early duration rice varieties 1. Better plant population density. Terminal drought may worsen drought effect 2. To avoid soil moisture stress in shorter rainfall span 1. To avoid soil moisture stress in reproductive stage 2. To compensate early drought loss Risk element Risk management Effect

Transplanting

Soil moisture stress at reproductive stage

1. Use of short duration rice variety 2. Establishment of Nursery bed at different time to match on set of rain

Lowland Flash flood Intermittent flooding damage: the established crop in different stages, resulting in yield loss 1. Growing very early rice variety and harvest it before onset of flooding 2. Growing a mixture of Aus and aman types to ensure at least harvest of a single crop, depending on early or late flood 3. Nursery sowing on different dates to enable replanting following flood damage. Double transplanting/ splitting of tillers from surviving crop to compensate for the damaged area 4. Sowing of extra early rice varieties/photoperiodsensitive varieties in postflood regime 1. Grain yield security and avoidance of flood damage 2. Lessening the early or late flood damage and ensuring grain yield of either rice crop 3. Proper crop establishment and higher land coverage in receeding floodwater 4. Use of residual soil moisture to obtain grain yield

5. Early crop establishment to reduce flood and submergence damage 6. Enhance survival percentage and maintenance of a good population density

Submergence/water stagnation

Reduction in plant population and thereby loss of yield

1. Using long-duration submergence-tolerant varieties 2. Advancement of planting with the help of subsurface water 3. Planting of overaged seedlings/ split tillers to cope with complete or partial submergence 1. Use of flood-tolerant variety 2. Proper land preparation for uniform germination 3. Proper stand establishment before flooding through application of bone meal, farmyard manure, and ash 4. Shift to boro rice cultivation in flood-free dry season 5. Boro-floating rice integration 1. Planting of salinitytolerant varieties 2. Transplanting of overaged seedlings in submerged soil 3. Extensive rice-fish cultivation 4. Planting of boro rice with backfeed water 1. Better grain yield 2. Better crop establishment and proper plant population stand 3. Lighter soil and early vigor of rice plant 4. Minimized environmental hazards, assured higher grain yield and economic returns 5. Early establishment of floating rice to minimize flooding damage

Flood-prone

Total damage of rice plantfollowing deep flooding

Coastal

flooding

Salinity reduces grain yield

1. Higher grain yield 2. Better crop establishment 3. Maximization of land use and higher economic returns using natural resources 4. Higher grain yield and economic returns using lowcost irrigation water

169

Flood-prone land Deep flooding causes total damage to rice. Farmers do not have any risk management practices to minimize grain loss. But fish culture with floating rice is an adaptation to cope with the risk of economic loss. Changes in land use pattern to reduce perennial risk in such a fragile environment as deep flooding have been occurring in West Bengal for more than a decade. Instead of growing risky floating rice, farmers grow boro rice in deepwater as the crop during the floodfree period is more assured. This shift has enabled farmers to meet their domestic consumption and reap good returns . Table 12 summarizes the risks involved in rainfed rice cultivation under different stress situations and farmers’ strategies to cope with them. The magnitude of risk in the rainfed rice ecosystem in West Bengal, as analyzed in this paper, emphasizes the need for appropriate management practices to combat stress situations.

References
Directorate of Agriculture, Government of West Bengal. Estimates of area and production of principal Ccops in West Bengal, 1986-87 to 1996-97. Directorate of Agriculture, Government of West Bengal. Harvest prices of crops. Directorate of Agriculture, Government of West Bengal. Land use statistics of West Bengal.

Discussion
Question: Have you examined the cost of risk management strategies in floodaffected areas? Answer: Not really, but these are crop management practices (variety, gap filling by splitting of tillers) that do not involve much cost. Question: How have you actually classified into high, and low risk areas and what are the possible strategies to minimize risk? Answer: The classification of areas is on the basis of coefficient of variation of area and productivity estimated for the last decade. Availability and certainty of irrigation can largely explain these variations. For example, high risk areas are subject to moisture stresses at different stages of crop growth due to inadequacy of irrigation. Comment: Instead of analyzing in terms of broad, subjective groups of risk level, it would be better if you could analyze risk as a continuous variable (CV of yield) in relation to variables such as irrigation, fertilizer. diversification, etc.

Concluding remarks
While the expansion of tubewell irrigation and the shift to boro rice have led to a rapid growth in rice productivity and stabilization of rice production in West Bengal, the following priority areas need to be addressed to further improve rice production. 1. Develop appropriate location-specific rice varieties to combat stresses in rainfed rice culture. 2. Identify appropriate agronomic practices that are economically viable and socially acceptable. 3. Develop a suitable technology for raising a nonrice crop with upland rice culture to reduce risk. 4. Intensify a crop substitution program, in which a crop of wet-season floating rice is converted to a crop of dry-season irrigated rice. 5. Assess farmers’ resource management practices, the adoption and economic viability of improved technology, and their extrapolation domains and ex ante impact on income distribution and poverty. 170

Authors’ addresses: N. Saha, Directorate of Agriculture (Evaluation), West Bengal; S. Bardhan Roy, economic botanist-III, West Bengal; U. Aich, Directorate of Agriculture (Agricultural Statistics), West Bengal. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

Notes

171

Risk and rice production in Orissa, eastern India
D. Naik, S. Pandey, D. Behura, and R. A. Villano

Orissa is an important rice-producing state in eastern India. Flood and drought are the major sources of risk that constrain growth in rice productivity in Orissa. This paper analyzes the nature and magnitude of risk in rice production there. Both district- and farm-level data are used. The analysis of district-level data showed that the average rice yield is positively correlated with the adoption of modern varieties. However, the variability of rice yield is not correlated with the adoption of modern varieties. The differences in variability across districts could be attributed mainly to soil/climatic factors but not to modern technology. Overall, the variability of rice production has changed little, except in coastal districts where variability has increased. Farm-level data showed various types of coping strategies used by farmers to deal with risk. Opportunities for reducing risk by manipulating crop management practices appear circumscribed by hydrological factors in the flood-prone village studied. Diversified sources of income in such a village may have helped stabilize household income. The share of rice income in the total household income was low, indicating that stabilization of rice yield per se will not generate a substantial economic advantage. Rice technologies that produce higher yield on average while reducing the chances of crop failure are needed to improve the performance of the rice economy.

Orissa is an important agricultural state of eastern India. It is predominantly an agricultural state with nearly 86% of its total population of 35 million residing in rural areas. Farmers and agricultural labor account for 73% of the economically active labor force in the state. Rice is an important crop of Orissa, where it is grown on about 4.5 million ha annually. It accounts for more than 89% of the total area under cereals and contributes to about 93% of total cereal production. The average productivity of rice in 1995-97 was 1.89 t ha". Almost 75% of the total area under rice in Orissa is rainfed. Uplands account for 21 % of the rainfed area. For 1969-97, rice production in Orissa increased at the compound growth rate of 2.03% per year. The compound growth rate in yield for the same period was 2.01%. The population growth rate for the state during 1981-91 was 1.84% per annum. Thus. the growth rate in rice production has been just adequate to maintain the current consumption of about 157 kg of milled rice per capita. As the predominant rice ecosystem in

Orissa is rainfed, rice production is subject to the vagaries of the monsoon and reflects a high level of yield and production instability (Fig. 1). The majority of farmers who are very poor may lack incentives and the capacity to invest in productivity-enhancing technology because of high risk. This paper analyzes the nature of risk in rice production in Orissa and draws out implications for technology design and policy improvement to help farmers better manage climatic risk in rice production. Although the risk associated with uncertainty in prices may also be important to some farmers, this paper focuse on climatic risk and how farmers have adapted to it.

Data and methods
The analysis of production variability in Orissa is based on district-level, time-series data (196997) for the state. These data were obtained from the official publication of the Government of Orissa. Although more districts were created in 173

199 1 by subdividing several districts, the data were grouped according to the original districts. Growth rates were calculated by fitting a semilogarithmic trend equation and the coefficients of variation (CVs) were estimated using linearly detrended data. Using these district-level data, attempts were made to explain the cross-district differences in variability of rice area, yield, and production. To analyze the change in instability over time, the district-level, time-series data were divided into two periods of equal length—19698 1 and 1982-94. The data on area and yield for each period were linearly detrended and the detrended data were centered on their respective means. Detrended yield and detrended area were multiplied to obtain detrended data on production. The coefficients of variation of area, yield, and production for these two time segments were compared to assess the nature of changes in variability. A variance decomposition analysis developed by Hazell (1982) was used to identify the components that have contributed the most to the change in variance of production.

The district-level analyses were complemented by farm-level analysis of coping mechanisms. Collection of panel data for analyzing farmers’ risk-coping mechanisms began in 1996. Two villages, Taraboi and Thailo, were selected for the study. Taraboi belongs to Khurda District, whereas Thailo is in Jagatsinghpur District. Taraboi is an inland village that is less prone to submergence than the coastal village Thailo. Thus, production systems in the former serve as a benchmark against which the nature of risk and risk management practices of farmers in the latter can be analyzed. Sixty-five farmers from Taraboi and 70 from Thailo were included in the sample. Data collection began in Taraboi in 1996 and in Thailo in 1997. As comparative data for both villages were available for 1997 only, the analysis reported in this paper was based on 1997 data. Data for 1998 and 1999 were not included in the analysis as these are being processed. Because data for only one year were used, discussion of risk and risk management touches only on the broader aspects.

174

An adequate analysis of risk adjustment mechanisms requires data other than those reported in this paper. Nevertheless, some broader features of adjustment mechanisms can be discerned. Adjustment mechanisms can be grouped into two types: ex ante and ex post. The former refers to strategies used to reduce production and income risk, whereas the latter refers to those that prevent a drastic reduction in consumption when losses do occur. These strategies were elicited during the interview process.

Table 2. Growth rates of area, yield, and production of rice in Orissa. Item Autumn rice Area Yield Production Winter rice Area Yield Production Summer rice Area Yield Production Total rice Area Yield Production 1980-97 -0.62** 2.47 1.85 0.67*** 2.13** 2.80*** 2.80*** 2.19*** 4.99*** 0.51** 2.34*** 2.85*** 0.02 2.01*** 2.03*** 1969-97
a

Results and discussion
District-level analysis

Rice is grown in Orissa in autumn, winter, and summer. Autumn, winter, and summer rice accounted for 19%, 76%, and 6% of the total rice area, respectively, in 1995-97 (Table 1). Their respective contributions to total output were 12%, 78%, and 10%. Winter rice, grown during July-November, predominates in Orissa. Autumn rice is mostly grown in the rainfed upland area and is established by direct seeding. Although the area under autumn rice expanded from about 0.5 million ha in 1950 to about 1 million ha in 1985, it has now shown a declining trend. The compound growth rates of area, production, and productivity of autumn rice from 1980 to 1997 were estimated to be -0.62%, 1.85%, and 2.47%, respectively (Table 2). The reduction in autumn rice area was mainly due to substitution by pulses, oilseeds, and fruit crops, which have become more profitable in recent

a Rice data, by season, only available starting from 1980. ***.** significant at 1% and 5% level, respectively.

Table 1. Relative importance of various rice cultures in Orissa. Item Season 1980-82 1990-92 1995-97 Overall average 21 74 5 13 79 8

Share of rice area (%) Autumn Winter Summer

21 75 4

20 75 5 14 77 9

19 76 6 12 78 10

Share of production (%) 14 Autumn 79 Winter 7 Summer

years. At least a part of the increase in the productivity of autumn rice is attributed to the replacement of autumn rice in more marginal areas by these crops. The area under winter rice was 3.3 million ha in 1950, but increased marginally to 3.4 million ha in 1997. The compound growth rates of area, production, and productivity of winter rice were 0.67%, 2.8%, and 2.1396, respectively. Thus, growth in the production of winter rice was driven mainly by growth in productivity. Summer rice is grown mostly in the irrigated areas of Balasore, Bolangir, Puri, Cuttack, and Sambalpur. Production in Orissa grew at almost 5% per year, with the contribution of area growth (2.8%) being slightly higher than that of yield growth (2.2%). In spite of a strong growth performance, the crop is of minor importance as its share in total area and production of Orissa is still quite small. Balasore, Cuttack, Koraput, Sarnbalpur, and Puri are the five major rice-producing districts of Orissa. These districts together accounted for 53% of the area and 57% of the production during 1995-97. Over the period 1969-97, Cuttack, Dhenkanal, Ganjam, and Mayurbhanj exhibited a declining trend in rice area (Table 3). Growth in rice productivity was higher in Sambalpur (2.58%) and Balasore (2.54%).

175

Table 3. Growth rates of rice area, yield, and production in Orissa, 1969-97. District Balasore Bolangir Cuttack Dhenkanal Ganjam Kalahandi Keonjhar Koraput Mayurbhanj Phulbani Puri Sambalpur Sundergarh Orissa
a

Area 0.23** 0.47*** -0.40** -1.34*** -0.52** 0.75*** 0.13 0.32* -0.02*** 1.56*** 0.14 -0.03 0.12 0.02

Yield 2.54*** 2.27*** 2.09*** 2.03*** 2.42*** 1.19** 1.11** 1.93*** 1.33 2.02*** 1.61*** 2.58*** 0.33 2.01***

Production 2.78*** 2.74*** 1.70*** 0.69 1.89** 1.94*** 1.24** 2.25*** 1.31*** 3.57*** 1.74*** 2.56*** 0.45 2.03***

***,**,*= significant at 1%. 5%. and 10% level, respectively.

Overall, most districts have been able to maintain a productivity growth of around 2% per year, except for Kalahandi (1.19%). Keonjhar (1.11 %), Mayurbhanj (1.33%). and Sundergarh (0.33%). These districts are characterized by a high incidence of poverty, poorer infrastructure, and less fertile soils, and are inhabited mostly by scheduled tribes. Autumn rice accounts for a higher proportion of total rice area in these districts. The average district-level rice yield

was highly correlated with percentage area under modem varieties (Fig. 2),. indicating that further spread of modem varieties is needed to raise yield in districts where adoption has been limited for a variety of reasons. For Orissa, the coefficients of variation of area, yield, and production for the period 196997 were estimated to be 4%, 15%, and 17%, respectively (Table 4). Variability of yield was quite high relative to that of other states such as Assam and West Bengal, which had yield CVs of less than 10%. The CVs of yield in Bolangir, Dhenkanal, Ganjam, Kalahandi, Phulbani, and Puri exceeded 20%. The scatter diagram (Fig. 3) indicates no clear relationship between district yield CV and district mean yield. As winter rice is the dominant crop in all districts, variations in CV probably reflect climatic fluctuations that characterize winter rice production in Orissa. To examine whether instability of yield across districts is related to the adoption of modem varieties, the CV of district-level yield for 1982-94 was regressed on the mean percentage area under modem varieties for the same period. The result indicates a poor correlation between these two variables (Fig.

176

Table 4. Mean and coefficient of variation of rice area, yield, and production in Orissa, 1969-97. Mean Area (000 ha) 416 307 575 266 323 298 211 395 336 102 395 546 219 4,387 Yield Pro(t ha-1) duction (000 t) 1.54 1.58 1.68 1.49 2.02 1.25 1.35 1.57 1.49 1.53 1.78 1.80 1.22 1.60 643 490 964 388 652 376 287 624 498 162 707 986 266 7,043 Coefficient of variation Area (%) 4 4 7 9 8 6 5 7 4 9 5 6 7 4 Yield Pro(%) duction (%) 18 21 17 25 28 21 19 15 15 23 20 15 17 15 19 23 19 26 31 24 22 18 15 29 23 18 18 17

District

Balasore Bolangir Cuttack Dhenkanal Ganjam Kalahandi Keonjhar Koraput Mayurbhanj Phulbani Puri Sambalpur Sundergarh All Orissa

4A). Regression of the CV of yield on the percentage area irrigated and the percentage area under modem varieties also produced statistically insignificant coefficients. As data on area irrigated and area under modem varieties may be somewhat less reliable, a second

regression of yield CV was run with percentage growth of yield as the explanatory variable. The growth rate of yield is expected to be higher in districts where modem technologies have been adopted rapidly. Thus, growth rate of yield can serve as a proxy for the adoption of modem technologies. This regression also produced a statistically insignificant slope coefficient (Fig. 4B), indicating that variations in CV across districts cannot be attributed to modem technologies. For Orissa as a whole, the mean area, yield, and production of rice changed in the second period (1982-94) relative to the first period (1969-81) by -1%, 35%, and 33%, respectively. All districts showed an increase in average yield, with the increases being more pronounced in Bolangir (43%), Ganjam (61%), Phulbani (40%), and Sambalpur (43%). Although Phulabani experienced a major increase in area (27%), Cuttack (9%), Dhenkanal (20%), and Ganjam (8%) had a decline. The changes in area in other districts were quite small.

177

The variance of rice area declined slightly (by 39%) between 1969-81 and 1982-94, but this decline was not statistically significant. Yield and production variance increased by 67% and 79%, respectively, but these increases were also statistically insignificant. Overall, the changes in variance of area, yield, and production at the state level have been small. At the district level, area variance in most districts declined, with Ganjam being the only district 178

where the area variance increased by 280%. The picture is that of a greater stability in area planted in most districts in the second period than in the first. Yield variance increased in most districts but only Ganjam (275%) and Pun (3 13%) registered a statistically significant increase. Thus, Ganjam is the only district in which both area and yield variance increased between the two periods. As a result, production variance in this district increased by 346%.

Table 5. Decomposition of production variance in the state of Orissa. Description Change in Change in Change in Change in Change in Interactions mean yield mean area yield variance area variance area-yield covariance and residual Percent 10.22 -1.08 72.36 -2.27 26.65 -5.88

Phulbani and Puri are the other two districts registering a statistically significant increase in production variance, 248% and 339%, respectively. The decomposition of change in variance between the two periods indicated that almost all of the changes in variance at the state level are due to the change in yield variance and the change in area-yield covariance (Table 5). These two components accounted for 72% and 26%, respectively, of the total change in production variance. As the change in the area-yield covariance depends partly on the change in yield variance, the total contribution of the change in yield variance will be higher than 72%. Thus, the changes that are occurring in the rice production systems of Orissa appear to have increased production variance mainly through an increase in yield variance. It should be bome in mind, however, that the overall increase in production variance for Orissa as a whole is statistically insignificant. The destabilizing effect seems to be concentrated mainly in the coastal districts of Cuttack, Puri, and Ganjam. What are the reasons for such large variability, especially in the coastal zone? The adoption of modem rice varieties in these districts increased rapidly, with the area under modern varieties in 1997 being more than 80%. These districts also have relatively more area under irrigation (38-52% of gross cropped area). However, Sambalpur, which also has a high rate of adoption of modem varieties (>80%) and a higher proportion of irrigation (>50%), has recorded very little change in variance. Thus, the adoption of varieties cannot by itself explain changes in production variability between the two time periods. Other interacting factors must have brought about changes in production variability. The increase in production variance

in the coastal belt is probably due to the spread of modem varieties in areas with poorer quality of irrigation and drainage and higher frequency of flooding. The supply of irrigation in the coastal belt has also become less reliable due to inadequate repair and maintenance of irrigation canals. Sambalpur, on the other hand, has a relatively assured source of irrigation. Thus, interactions between modem varieties and hydrological fluctuations in the coastal belt could be a major reason for increased production instability in this belt. Analysis of farm-level data The long-term rainfall patterns for the two study districts were similar (Fig. 5). Despite these similarities, submergence is more common in Thailo due to its low elevation and proximity to the coast. Jagatsinghpur received very high rainfall in June and July 1997 (Fig. 5), causing massive flooding and destroying almost its entire rice crop for the year. Information on various sources of risk was elicited from farmers. The results indicate that the probability of flood in Thailo is quite high (44%) (Table 6A). As strong winds are often associated with heavy rains, the probability of flood occurrence in Thailo may be even higher. Taraboi has no single dominant source of risk. Several sources contribute to overall risk in Taraboi, although each has a relatively low probability of occurrence. Table 6B shows farmers’ estimates of the probabilities of two levels of losses. Again, in Thailo, the probability that more than 50% of the output will be lost due to submergence is 27%. This provides an indication of the severity of flood risk. Overall, the probability of production loss being greater than 50% from different sources of risk is lower in Taraboi than in Thailo. Although the elicited probabilities presented above approximate the marginal probabilities of each stress event, more than one stress event may occur within the same growing season. Such joint occurrences can result in even greater losses to farmers. Although the average farm size is similar, the degree of fragmentation of landholding is greater in Thailo than in Taraboi (Table 7). In 179

Table 6A. Incidence (%) of stresses. Nature of stress Strong wind Flood/submergence Drought insects Diseases Taraboi 12 7 11 10 7 Thailo 24 44 10 10

Table 6B. Farmers’ estimates of probabilities (%) of yield losses. Taraboi Nature of stress Strong wind Flood/submergence Drought insects 60% 8 5 6 9 >50% 4 2 5 1 40% 10 17 7 9 Thailo >50% 14 27 3 1

Thailo, farmers have as many as 22 plots per household. The average plot size is also much lower in Thailo than in Taraboi. Although land fragmentation can help reduce the overall production risk, the available data do not permit us to judge the extent to which risk reduction is a motive for land fragmentation in the study villages. A distinct agrarian feature of the study area is a relatively high proportion of rented-in land among marginal farmers (30-35%). In term of rice area, lowland fields where submergence is a common problem are proportionately more important in Thailo (53%) than in Taraboi (28%). The distribution of various types of fields across households of

180

Table 7. General characteristics of the surveyed villages. Characteristic No. of households Av area operated (ha) Av owned area (ha) Av area of parcel (ha) Av no. of parcels/households Minimum Maximum % area by land type Upland Medium land Lowland Taraboi Marginal 29 1.18 0.72 0.27 5 1 13 29 29 42 Small 22 1.48 1.34 0.38 4 1 9 6 65 31 Large 14 3.03 2.86 0.49 6 2 15 14 72 14 All farms 65 1.68 1.39 0.36 5 1 15 15 57 28 Marginal 37 0.91 0.67 0.11 9 1 22 2 41 57 Small 19 1.62 1.48 0.1 7 10 1 19 7 40 53 Thailo Large 14 2.82 2.82 0.23 13 6 21 5 45 50 All farms 70 1.48 1.32 0.1 5 10 1 22 5 42 53

different size categories is similar in Thailo. In Taraboi, the proportion of medium land is higher among small and large farmers, while the proportions of lowland and upland fields are higher among marginal farmers. As Thailo represents predominantly a submergence-prone environment, dry seeding is the most common method employed for establishing rice in more than 95% of the area. Farmers establish rice by dry seeding ahead of rains, so that, by the time the area gets flooded, the crop is tall enough to avoid serious damage. Dry seeding in submergence-prone areas is a common practice in South and Southeast Asia. In Taraboi, the major crop establishment method is dry seeding as it accounts for 80% of the rice area. However, transplanting is also practiced, especially in the medium land where submergence is not a serious problem. An interesting feature of the study villages is the dominance of rice in the rainy season. Because rice occupies almost 100% of the land, no other crops are grown on any of the land types. This limited diversification, even in the uplands, is an indication that the overall ecosystem is lowland, although farmers classify fields with slight elevation as upland. Because of excessive accumulation of water in the field, opportunities for risk reduction through crop diversification seem rather limited. Although several crops are grown during the dry season, almost half the area is left fallow in both villages. The fallow land serves as grazing area for cattle. Modem varieties are adopted in only onethird of the rice area in both villages. The

adoption of modem varieties is therefore far below the state average of 68% for 1997. Mahsuri and CR1009 are the two major improved varieties grown in the study villages. CR1009 is a medium tall variety of 155-d duration and is suitable for lowland conditions. Mahsuri, a 145-d variety, has become popular in many parts of eastern India because of its good grain quality. Both villages showed a high degree of varietal diversification with as many as 93 varieties grown by the sampled farmers. Varietal diversity was higher in submergence-prone Thailo. Rice varietal diversification indicates environmental diversity as well as farmers' mechanisms for coping with high risk. Farmers attempt to reduce production risk by growing several varieties with different characteristics so that at least some production is assured. There may be other reasons, however, for growing several varieties such as staggering of labor use and producing output of different quality (grain and straw). Submergence was a major problem that severely affected rice production in Thailo in 1997. More than 90% of the crop was destroyed due to submergence. Rainfall during July was very high and resulted in widespread flood in the village (Fig. 5). Land was mostly under water and no other crops could be grown. As a result, only a small proportion of the planted area could be successfully harvested. Rice yield averaged 1.8 t ha-1 in Taraboi but almost zero in Thailo because of flood damage. Crop production practices are similar in both villages but flood damage resulted in Thailo 181

Table 8. Yield, cost, and return of rice in the surveyed villages, 1997. Item Taraboi Marginal 1.8 3,429 8,283 4,853 Small 1.7 4,619 8,077 3,458 Large 1.9 4,445 8,788 4,342 All farms 1.8 4,058 8,358 4,300 Marginal 0.0 1,213 52 -1,160 Small 0.0 1,619 0 -1,619 Thailo. Large 0.0 1,872 123 -1,750 All farms 0.0 1,455 52 -1,403

Yield (t ha-1) Total cash cost (Rs ha-1) Total gross return (Rs ha-1) Net returns (Rs ha-1)

Table 9. Sources of income in 1997. Source Total income (Rs household-1) Share (%) Rice Other cropsa Fruits and timber Livestock and poultryb Nonfarm labor Off-farm labor Other sector employment Business Pension /remittances Taraboi Marginal 26.527 16 2 0 4 5 7 44 15 6 Small 42,397 9 2 1 6 1 0 64 8 7 Large 82,007 15 2 1 3 0 0 62 10 7 All farms 43,865 13 2 1 4 2 2 58 10 7 Marginal 17,300 5 7 0 17 9 10 42 8 2 Thailo Small 23,929 6 11 1 15 3 3 39 10 13 Large 59.359 3 20 1 8 1 1 43 13 10 All farms 27,197 4 14 1 13 4 5 41 10 8 100

a

Includes sugarcane, mungbean, horsegram and vegetables. chili. and colocasia.

b

lncludes sale of livestock and poultry and products such as milk.

farmers applying virtually no chemical fertilizers. The average rate of application of N, P, and K was 25, 1, and 13 kg ha-1, respectively, in Taraboi. Overall, the net returns (gross returns minus cash costs) were similar across farm size categories and averaged Rs 4,300 ha-1 (Table 8). The average farm household income in 1997 was Rs 44,000 in Taraboi and Rs 27,000 in Thailo (Table 9). Although the average operational holding was only 13% higher, the farm household income in Taraboi was 62% higher. The major part of this difference is not due to income from rice or agricultural income in general but to a substantially higher nonfarm income in Taraboi. Even if farmers in Thailo had obtained the same income from rice as Taraboi farmers did, their total income would have been only 17% higher. Rice accounted for only 13% of the total household income in Taraboi, with government employment accounting for nearly 60% of the total income. Although public- sector employment was also a major source of income, accounting for 40% of the total income, other important sources of income in Thailo were livestock and crops such as mungbean, horsegram, and vegetables. Overall, the income 182

sources in Thailo are a bit more diversified than in Taraboi, where public-sector employment is the single major source of income. With rice accounting for only about 13% of the income, the effect of stabilization of rice income on the stability of total household income is likely to be small (Pandey et al 1999). The contributions of rice income to total income of small and marginal farmers are even smaller. Thus, the usual argument that stabilization of rice yield will benefit small farmers proportionately more is untenable. Such farmers are likely to benefit more from technologies that increase the average rice yield than from those that reduce the variability of rice around a given mean yield. The current income structure also highlights the need to improve the productivity of other crops as well as to generate nonfarm employment opportunities to raise the income of these households. The major ex ante strategies used to reduce risk in rice production in the study area are ( 1) resowing when the crop is damaged early in the growing season, (2) reducing fertilizer application when crop damage is anticipated, (3) using transplanting as a fall-back strategy when the dry-seeded crop is damaged or dry seeding is

Table 10. Ways of augmenting family income (% of farmers). Item Nonfarm labor Agricultural labor Selling of animals Borrowing Bank Other sources a Others b
a

Taraboi Marginal 34 24 1 1 20 3 Small 10 14 14 5 14 Large 7 All farms 2 15 9 9 12 2
b

Thailo Marginal 65 49 32 22 33 11 Small 22 22 22 16 37 5 Large 36 21 14 21 All farms 47 36 26 2 27 8

14

Includes friends and relatives, money lenders, traders, and big farmers.

Includes mortgaging of jewelry and land and remittances.

not possible due to early rains, and (4) changing the crop establishment method in the subsequent year to transplanting when weed infestation is high. Farmers in Taraboi, where submergence is a less serious problem, changed crop establishment methods from dry seeding to transplanting (and vice versa) more often to adapt to varying patterns of rainfall. High incidence of submergence in Thailo constrains opportunities for changing the crop establishment methods in this village. Opportunities for other kinds of adjustments such as changes in rice varieties are also constrained because of the need to dry-seed fairly early in the season. Farmers in this village, however, transplant late if the dry-seeded crop is damaged. They use surplus seedlings from neighboring villages for this purpose. Income diversification is another strategy used to reduce instability in household income. As discussed earlier, income sources are more diversified in Thailo than in Taraboi. Nonrice crops and livestock are more important sources of income in Thailo. However, income diversification may also be the result of nonrisk factors such as differences in resource base. The single-year data, however, do not permit an analysis of the stabilization effect of income diversification. When losses do occur, as they did in 1997, reliance on income from nonfarm and agricultural employment is a major coping strategy, especially for marginal and small farmers. In Thailo, 65% of the farmers relied on nonfarm labor earning and 49% relied on earning as agricultural labor (Table 10). In Taraboi, these are also the major coping strategies, but reliance on these strategies was

probably not so essential as crop losses were not as severe. It is interesting to note that quite a number of even the so-called “large” farmers in Thailo resorted to nonfarm employment and agricultural wages to supplement their income in 1997. While nonfarm employment can be an important coping mechanism in times of stress, employment as agricultural labor is less effective in dealing with an income shortfall if the adverse effect is spread over a large area as demand for agricultural labor in the locality is reduced, making it difficult to rely on this source of income as a coping mechanism. The other two major coping mechanisms are borrowing money and selling assets. Farmers borrowed money from both noninstitutional and institutional sources as well as sold animals and mortgaged land. The sale of productive assets such as bullocks in these stress events may be an important coping mechanism, but it can reduce long-term productivity because of a long delay in reacquiring the lost assets (Pandey et al 2000).

Concluding remarks
Despite the importance of rice in Orissa, the average state-level yield has remained low and its growth has been just enough to maintain the current level of per capita consumption. Clearly, rice productivity needs to be increased to ensure food security and incomes need to be raised to reduce poverty. The adoption of modem varieties in Orissa has increased over time and this has had a positive impact on the average yield. Fortunately, production variability does not seem to have increased appreciably except in the coastal belt. Increases in yield variance and the 183

area-yield correlation seem to be the major contributors of increased production variance but these increases are not correlated with the adoption of modem varieties. Analysis of farm-level data indicates that rice can be lost completely in some years due to flooding. Obviously, the impact on poor farmers of such a major loss can be catastrophic. But nonfarm income and agricultural labor earnings are important safety nets employed during these stress years. Fortunately, diversified sources of income have evolved over time, with nonfarm income contributing a major source of livelihood to most farmers. This has helped cushion the adverse impact of stress events such as flood. Policies that encourage further diversification of income sources could play an important role in reducing the impact of risk in rice production. However, per capita growth of the state domestic product of Orissa over the last decade has remained quite low, thus limiting opportunities for further income diversification. The share of rice income in the total household income was quite low. This implies that stabilization of rice yield per se will generate only a small economic benefit. What is needed is to improve rice yield and to stabilize it simultaneously. Poor soil fertility in several districts of Orissa and rainfed environments make the task of developing suitable technologies challenging indeed. A more thorough understanding of farmers' practices and the features of technology that would best suit these harsh environments can play a critical role in this task.

Pandey S, Behura DD, Villano RA, Naik D. 2000. Economic cost of drought and farmers' coping mechanisms: a study of rainfed rice systems in eastern India. Discussion Paper Series No. 39. Los Baños (Philippines): International Rice Research Institute. 35 p. Pandey S, Singh HN, Villano RA. 1999. Rainfed rice and risk coping strategies: some microeconomic evidences from eastern India. Selected paper for presentation at the Annual Meeting of the American Agricultural Economics Association, 8- 11 August 1999, Nashville, Tennessee, USA.

Discussion
Question: Could you tell us the reasons for increasing instability and growth in rice productivity in Orissa? Answer: Rice is grown in very diverse agroclimatic conditions in Orissa. It so happens that there are floods in some areas and drought in other areas in the same year. We have good technology in medium lands that has pushed up the yield. But there are no suitable, high-yielding variety/ technology for the uplands and lowlands. About 70% of the total rice production in Orissa comes from traditional varieties. Question: Farmers are growing 30-40 varieties in a year. Are these mostly officially recommended or are these local varieties? Answer: In 1997, farmers grew about 45 varieties and almost 23 of them were local. Further, seed replacement rate was very low (2%) in Orissa, and therefore, improved varieties grown by farmers are not really good and are losing seed vigor over time.

Government of Orissa. Various years. Orissa agricultural statistics. Orissa, Bhubaneshwar: Directorate of Agriculture and Food Production. Hazell PB. 1982. Instability in Indian foodgrain production. Washington, D.C.: International Food Policy Research Institute.

References

184

Notes
Authors’ addresses: D. Naik and D. Behura, Orissa University of Agriculture and Technology, Bhubaneshwar, Orissa, India; S. Pandey and R.A. Villano, International Rice Research Institute, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

185

Risk and rice technology design
L.J. Wade

This paper considers approaches to quantitative risk assessment using experimental data and crop simulation as a basis for discussing promising technology interventions and pathways to adoption in rainfed lowland rice. Results from pattern analysis of genotype by environment interaction are used to demonstrate that risk is not random, but there are patterns of risk, and these can be quantified with cumulative distribution functions. Promising technologies include reduction of percolation loss of water in northeast Thailand, rotation of short-duration crops in northwest Bangladesh, use of short-duration cultivars with better root systems and on-farm reservoirs to reduce the impact of drought, better cultivars and improved nutrition to improve submergence tolerance, and use of weed-suppressive cultivars with improved management in direct seeding. On-farm development of such technologies is essential, with partnership among research, extension and farmers, so that all can benefit from technical knowledge, understanding, and feedback. Because patterns are repeatable, we do not need a new technology for every field. Our technologies should be systems-based and suitable for local adaptation.

This paper is written with the philosophy that risk is not fully random, but there are patterns of risk. We should be able to identify those patterns and use that knowledge in risk management. A biologist’s perspective on quantitative risk assessment using experimental data and crop simulation is provided as a basis for considering some technologies in the pipeline and some issues in relation to adoption. Results from research on genotype by environment (G x E) interaction are used to illustrate the approach for research and technology transfer in rainfed lowland rice. Quantification of risk is illustrated using cumulative distribution functions for the effects of different cultural practices on yield expectations. Some promising interventions in agronomy and plant breeding are considered for rainfed lowland rice. The paper concludes that targeted strategies are needed for targeted environments. In this approach, trait combinations and management practices consistent with the critical characteristics of the target population of environments are more likely to be successful.

Our research in rainfed lowland rice
Drought stress is commonly considered the most important limitation to yield in rainfed lowland rice (Widawsky and O’Toole 1990). Consequently, we commenced research on drought tolerance at various levels of integration. The core of the research was crop physiology, to better understand plant response to drought (Samson et a1 1995, Wade et a1 1997b). At a higher level of integration, G x E interaction was used to integrate this understanding of drought response with patterns of adaptation of lines to target populations of environments in the rainfed lowlands (Wade et a1 1995, 1997a). Agronomic management provides prospects for improving grain yield and yield stability quickly, although the largest gains are likely when cultivar and management are manipulated together (Wade et al 1998a). Consequently, we examined seedling vigor (Douthwaite et a1 1995, Bhuiyan et al 1998) and nutrient by water interactions (Wade et al 1998a,b).

187

An example for G x E interaction
To illustrate the principle that there are patterns of response in relation to risk, let us consider the results of Wade et al (1997a), who examined the performance of 31 genotypes over nine locations in the rainfed lowlands of Thailand, India, and the Philippines in 1995. Pattern analysis was used to group the responses of genotypes in relation to different types of environment in order to identify varieties with different patterns of adaptation. The analysis revealed that 48% of the variation in yield response of the 31 genotypes over the nine locations was due to site conditions, 14% to the main effect of genotype,

and 38% to the G × E interaction. This illustrates the challenge for plant breeders in trying to develop better cultivars for variable and heterogeneous rainfed environments; G x E was three times the magnitude of G. Nevertheless, there were clear and repeatable patterns in the data. Figure 1 presents a biplot of the G × E interaction. Axis 1 accounted for 51% of the G × E sum of squares and axis 2 for 27%. The biplot is a spatial representation of the associations in the data, with sites shown as spokes from the origin and genotypes shown as symbols. Groups of genotypes with a common symbol are circled to emphasize the similarity of their response over environments. Where a genotype group

188

maps close to a site, that indicates a preferential adaptation to those conditions. Where a genotype group maps to the opposite side, this indicates a poor adaptation to those conditions. The strength of the association is quantified by the intercept between the environment spoke (or its extension) and a perpendicular line to the genotype. The further the intercept is from the origin, the stronger is the association, and this may be positive or negative. For axis 1, the Philippine sites are to the left and the Thai sites to the right (Fig. 1). For axis 2, the majority of sites were neutral, with Phimae strongly positive and Raipur B strongly negative. From knowledge of the site conditions, Wade et a1 (1997a) interpreted axis 1 as general site favorability, with the low-fertility and drought-stressed Thai sites of Ubon and Chum Phae to the right and the more fertile Philippine sites with favorable hydrology to the left. Raipur A was intermediate between those extremes, whereas the late-planted Raipur C, where drought was severe, mapped closer to Ubon and Chum Phae. Wade et al ( 1997a) noted that the crop at Phimae was fully submerged at flowering, whereas the crop at Raipur B had favorable conditions early, followed by fastonset late-season drought. Consequently, axis B was interpreted as indicating significant fluctuations in the water regime later in the season, with either too much or too little water. Responses of genotypes may be interpreted relative to these site conditions. The group comprising CT9897 and several semidwarf IR lines was preferentially adapted to the more favorable conditions occurring at the Philippine sites, whereas Sabita and KDML105 were better adapted to the drought conditions of Ubon and Chum Phae. The group comprising several breeding lines from IR66516 was better adapted to submergence, whereas NSG19 was better adapted to fast-onset drought. Groups including Mahsuri and IR99879 were more neutral and more stable over environments.

Plant breeding and agronomic philosophy
What adaptive strategy is desirable for the best cultivar? Do you choose a cultivar with high and

stable yields over many environments, a strategy referred to as general adaptation? Or do you choose the highest yielding cultivar in particular conditions, for example, for fully irrigated conditions with adequate nutrients, or for a particular type of drought or specific soil condition such as low pH, examples of more specific adaptations? For a well-defined target with specific adaptation, the preferred cultivar would have high and stable yields and low G X E interaction within the target population of environments, but may do very poorly elsewhere. My personal philosophy is that the scientific breeding program should seek to lock in adaptation to the dominant constraints of the major target subecosystems. This involves addressing difficult traits such as drought or submergence tolerance, in collaboration with physiologists, molecular biologists, and soil scientists. Rainfed lowland rice may have several dominant targets as indicated by Wade et a1 ( 1997a) and discussed briefly above: favorable, late drought, early drought, submergence, fast-onset late drought, and combinations of these. Each of these broad targets represents substantial rainfed lowland areas, with specific trait requirements. So the core breeding program would seek to develop populations adapted to the broad constraints of these major target subecosystems. Within each population, however, it is important to retain variance for other nontarget traits, such as phenology, grain quality, and others. These materials may then be subjected to further selection in more specific conditions, such as in farmer participatory selection. The objective then is to identify plants with specific adaptations to local conditions (e.g., growing season duration) and farmer preferences (plant type, grain quality). For success, 1 believe that the core program must address complex but essential traits such as drought tolerance, using all the tools now available. Farmer participation and local phenotypic selection are most likely to be successful with simply inherited traits of clear expression, which confer local adaptation to specific conditions, especially if repeatability is high, such as low pH and flowering time. Such a program should result in cultivars with broad 189

adaptation to the principal constraints of target subecosystems and with specific adaptations to local variants. A similar argument may be provided for agronomic management. Broader strategies should be developed initially in relation to the same target subecosystems, ideally using the adapted cultivars developed above. Such strategies may then be locally modified according to the specific characteristics of local conditions and farmer preferences. This is an important distinction. In practice, successful cultivars are grown across several local environmental variants, but with adjustment of agronomic practices. The task of strategic research is to understand the principles. There is opportunity for research and extension, and for farmers to participate together in adapting such strategies to local conditions. It is important to note that such tactics may be derived locally, then applied more widely with extension or NGO support. A common-sense evaluation of a local problem may lead to a real solution. The contribution of research may be in providing further options to farmers, in checking and ensuring sustainability of the resulting systems, and in quantifying risk associated with alternative strategies.

wide rows at low density to provide a store of water between the rows as a buffer against dry periods. Standard practice is a compromise between these extremes. The yield probabilities shown in Figure 2 quantify these concepts, so that the number of crop failures, likely yields, and economic returns can all be evaluated (Lansigan et al 1997).

Quantifying risk associated with alternative strategies
Risk can be quantified by analyzing data collected over sites and years or through crop simulation. Figure 2 shows an example of this, in which cumulative distribution functions are used to illustrate the response of sorghum to plant density and row spacing at Katherine, Emerald, and Dalby in Australia (Wade et al 1991, Wade 1995). Strategies are compared for three possible farmer attitudes to risk. “Risk takers” who own their property may be willing to lose a crop in a poor season if yield and profit can be maximized in a favorable season. High density in narrow rows would be the strategy employed in this instance. In contrast, “risk avoiders” may wish to minimize the risk of crop failure so they can always pay their debts, even if it means sacrificing yield and profit in a good season. These farmers would grow the crop in 190

Emerging technologies in rainfed lowland
In crop improvement, the principal activities are the national breeding programs, shuttle breeding for exchange of knowledge and materials, G x E interaction to understand patterns of response and adaptation to target environments, and participatory research with farmers for more specific adaptations. In agronomy, research is conducted in a cropping systems context, with concerns for nutrient cycling and nutrient requirement, weed ecology and management, and the use of on-farm reservoirs, direct seeding, and practices to alter patterns of water extraction to buffer effects of drought. Some examples of emerging technologies follow. In northeast Thailand, crops are exposed to severe drought on coarse-textured soils of low fertility, where percolation loss of water is high, and water and nutrient retention is low because of low content of clay and organic matter and low nutrient-buffering capacity. Three options offer prospects to improve the situation by increasing water- and nutrient-use efficiency: leveling fields, compacting subsoil, and adding recalcitrant organic matter (Sharma et a1 1995, Trebuil et a1 1998). In northwest Bangladesh, a change to earlier maturing cultivars, especially if direct-seeded, avoids late drought for rice and may permit a short-duration crop of high value to be grown on stored water, such as chickpea, linseed, or mustard. Direct seeding may result in increased weed problems, but pre-/postemergent herbicides or the use of stale seedbeds should reduce this threat (Mazid et al 1997, 1998). The use of on-farm reservoirs and early cultivars can reduce the impact of drought (Saleh and Bhuiyan 1995). For submergence, establishing vigorous seedlings with adequate nutrient supply, especially with controlledrelease nutrients, improves crop survival and recovery (Wade et a1 1998a,b). For rice blast, supplying potassium and silicon by applying fly ash has reduced infestation severity (Pinnschmidt et a1 1996). In crop improvement, adjustment of crop phenology to match growing-season duration is

fundamental to improved adaptation. For drought-prone areas, this may involve selection for early maturity to escape drought or the use of photoperiod sensitivity to ensure that anthesis occurs during the most favorable part of the growing season (Wade et a1 1997a). For rainfed lowlands, where soil conditions fluctuate from anaerobic to aerobic, a root system with a capacity to extract water from deeper soil layers should be advantageous for drought-prone environments (Wade et a1 1997b). For short-term submergence, seedling vigor for greater reserves of assimilate; enzymatic machinery for anaerobic respiration and protection from aldehydes, active oxygen species, and other harmful by-products; and a nonelongation strategy should be effective (Setter et a1 1997). A rapid increase in plant height and a rapid expansion of leaf area are considered advantageous for weed suppression (Bastiaans et al 1997).

Issues in the adoption of promising technologies
On-farm development of promising technologies is conducted to link research at key sites of the Rainfed Lowland Rice Research Consortium (RLRRC) with the activities of applied research and extension staff of the national systems and, more recently, with staff of nongovernment organizations that work closely within their villages. In Bangladesh, research has demonstrated benefits of a transition from a long-duration, T. aman sole rice crop to a short-duration rice/ chickpea rotation. At Rajshahi, researchers from the Bangladesh Rice Research Institute have active collaboration with extension block supervisors of Bangladesh Agricultural Extension Service, technicians from the Bangladesh Rural Advancement Committee (BRAC), and local farmers from Nachole, Thanore, and Ghadagari. The on-farm demonstrations provide an opportunity for research and extension, and for farmers to discuss and evaluate alternative agronomic approaches to implementing the new rotation in local conditions. BRAC provided staff to

191

support the local demonstrations, brochures in local languages, and venues for regular farmer meetings and discussions. In Thailand, researchers from Rice research Institute of Thailand (RRIT) collaborate with technicians from the Population and Community Development Association and farmers from Nang Rong and Lamplaimat in defining production constraints and setting up appropriate demonstrations. These include attention to field leveling for improved rainfall management, nutrient management, weed management, and their combination. The approach is now being considered for all RLRRC key sites, as an additional activity to supplement the research activities. This involves analysis of problems with farmers, identifying likely technologies from anywhere (RLRRC, NARES, farmers, literature), and demonstrating and evaluating the technology on-farm. It should be emphasized that this is a two-way learning process. Farmers, extension agents, and NGO staff gain from the technical knowledge of the researchers, while the researchers learn from problem identification and feedback from their clients.

process so we can learn and better target the next generation of research and adoption.

References
Bastiaans L. Kropff MJ, Kempuchetty N, Rajan A, Migo TR. 1997. Can simulation models help design rice cultivars that are more competitive against weeds? Field Crops Res. 51:101-111. Bhuiyan SI, Tuong TP, Wade LJ. 1998. Management of water as a scarce resource: issues and options in rice culture. In: Dowling N, Greenfield S, Fischer KS, editors. Sustainability of rice in the global food system. Pacific Basin Study Center, California, and IRRI, Philippines. p 175192. Douthwaite B, Tado CJM, Calendacion AN, Wade LJ, Cassman KG, Quick GR. 1995. Effect of stubble treatment on performance of ratoon rice. In: Fragile lives in fragile ecosystems. Proceedings of the International Rice Research Conference, 13-17 Feb 1995. Manila (Philippines): International Rice Research Institute. p 421-435. Lansigan FP, Pandey S, Bouman BAM. 1997. Combining crop modeling with economic risk analysis for the evaluation of crop management strategies. Field Crop Res. 51: 133-145. Mazid MA, Mollah MIU, Mannan MA, Elahi NE, Ali A, Ahmed HU, Hasan M, Rashid MA, Siddique SB, Kumar J, Wade LJ. 1997. Rainfed rice-chickpea cropping system to increase productivity for high Barind tract of Bangladesh. International Food Legume Research Conference, Adelaide, Australia. Mazid MA, Wade LJ, Saleque MA, Sarkar ABS, Mollah MIU, Olea AB, Amarante ST, McLaren CG. 1998. Nutrient management in rainfed lowland rice for the high Barind tract of Bangladesh. In: Ladha JK et al, editors. Rainfed lowland rice: advances in nutrient management research. Manila (Philippines): International Rice Research Institute. p 217-227. Pinnschmidt HO, de la Pena F, Suriya-Arunroj D, Don LD, Teng PS. 1996. Approaches for the characterization of cropping situations

Conclusions
Dr. R.B. Singh stated, “Our technology has to be location-specific, with systems devised at the local level. The technology must be considered in its ecosystem context. This is the crux of the problem.” In the end, I agree completely with Dr. Singh. We should identify our technology interventions for on-farm development with the farmers, drawing from our research and experience, and from theirs. We should modify, evaluate, and adapt the technology according to local needs. But I do not believe this job is impossible. We do not need a new technology for every field. There are repeatable patterns over sites and years. Our technology should be systemsbased and suitable for local adaptation. I believe that the farmer is an innovator who can devise novel solutions for a particular situation, given the vision (from wherever) of how a new technology can help. We as scientists should be involved in the modification-adoption-feedback 192

and yield constraints in rainfed lowland rice agroecosystems of some Southeast Asian countries and consequences for crop protection. J. Plant Dis. Prot. 103:620-643. Saleh AFM, Bhuiyan SI. 1995. Crop and rainwater management strategies for increasing productivity of rainfed lowland rice systems. Agric. Syst. 49:259-276. Samson BK, Wade LJ, Sarkarung S, Hasan M, Amin R, Hampichitvitaya D, Pantuwan G, Rodriguez R, Sigari T, Calendacion AN. 1995. Examining genotypic variation in root traits for drought resistance. In: Fragile lives in fragile ecosystems. International Rice Research Conference, Manila. Manila (Philippines): International Rice Research Institute. p 52 1-534. Setter TL, Ellis M, Laureles EV, Ella E, Senadhira SB, Mishra SB, Sarkarung S, Datta S. 1997. Physiology and genetics of submergence tolerance in rice. Ann. Bot. 79:67-77. Shanna PK, Ingram KT, Hampichitvitaya D. 1995. Subsoil compaction to improve water use efficiency and yields of rainfed lowland rice in coarse-textured soils. Soil Till. Res. 36:33-44. Trebuil G, Hampichitvitaya D, Tuong TP, Pantuwan G, Wade LJ, Wonprasaid S. 1998. Improved water conservation and nutrient use efficiency via subsoil compaction and mineral fertilization. In: Ladha JK et al, editors. Rainfed lowland rice: advances in nutrient management research. Manila (Philippines): International Rice Research Institute. p 245-257. Wade L.J. 1995. Genotype by environment interaction and selection-experiences in sorghum and expectations for rainfed lowland rice. In: Aggarwal PK, Matthews RB, Kropff MJ, van Laar HH, editors. Applications of systems approaches in plant breeding. TPEV-WAU Wageningen and IRRI, Philippines. p 31-39. Wade LJ, George T, Ladha JK, Singh U, Bhuiyan SI, Pandey S. 1998a. Opportunities to manipulate nutrient by water interactions in rainfed lowland rice systems. Field Crops Res. 56:93-112.

Wade LJ, McLaren CG, Criseno L, Amarante ST, Sarawgi AK, Kumar R, Bhambri MC, Singh ON, Ahmed HU, Rajatasereekul S, Porn-uraisanit P, Boonwite C, Harnpichitvitaya D, Sarkarung S. 1997a. Genotype by environment interactions-RLRRC experience. In: Breeding strategies for rainfed lowland rice in drought prone environments. Proc. 77, Canberra (Australia): Australian Centre for International Agricultural Research. p 115125. Wade LJ, Moya TB, Pantuwan G, Regmi KR, Samson BK. 1997b. Research at IRRI on rice root systems for drought resistance. In: Morita S, Abe J, editors. Perspective on ideotype of rice root system. Singapore: World Scientific Publishing. Wade LJ, Myers RJK, Foale MA. 1991. Optimizing plant stand in response to climatic risk. In: Muchow RC, Bellamy JA, editors. Climatic risk in crop production: models and management for the semiarid tropics and subtropics. Wallingford (UK): CAB International. p 263-282. Wade LJ, Quintana L, Amarante ST, Naklang K, Hampichitvitaya D, Singh AP, Sengar SS, Parihar SS, Singh G, Wihardjaka A, Mazid MA. 1998b. Nutrient-water interactions in diverse soils of the rainfed lowland rice ecosystem in Asia. World Congress of Soil Science, Montpellier. CD-Rom Proceedings, Symposium 14, Paper 7. Wade LJ, Sarkarung S, McLaren CG, Guhey A, Quader B, Boonvite C, Amarante ST, Sarawgi AK, Haque A, Hampichitvitaya D, Pamplona A, Bhamri MC. 1995. Genotype by environment interaction and selection methods for identifying improved rainfed lowland rice genotypes. In: Fragile lives in fragile ecosystems. Proceedings of the International Rice Research Conference, 1317 Feb 1995. Manila (Philippines): International Rice Research Institute. p 885900. Widawsky DA, O’Toole JC. 1990. Prioritizing the rice biotechnology research agenda for Eastern India. New York: The Rockefeller Foundation.

193

Discussion
Question: You have mentioned the need for participatory research and involvement of NGOs in technology transfer. Could you please tell us about the experience of IRRI in this regard. IRRI has just started work on rainfed lowland rice in Thailand and Bangladesh, where NARS collaborators, including NGOs, are involved. Participation of NGOs is important as they work very closely with farmers and understand farm situations and problems. Breeding materials developed on research stations may not satisfy all requirements of farmers. Their involvement through on-farm research even in segregation stage enables realistic evaluation of breeding material. There may be several problems relating to farmers’ participation in on-farm research. For example, costs involved in terms of low returns from experiments may not attract farmers, particularly small and marginal farmers, and eventually, there would be some kind of contractual participation by farmers, i.e. giving some land for experiments. How are these problems taken care of in the IRRI’s programmes. On-farm research involves a range of arrangements including putting few lines on farmers’ fields to having breeding nurseries right there. One can minimise the cost to farmers by bringing farmers to the nursery and obtain their feedback. Social scientists may further supplement on the issue of poor farmers’ participation, but breeders may be interested more in soil types

Answer:

Question:

and environment that are represented in farmers’ fields rather than the issue of small versus large farmers. Comment: In India, we have selected farmers based on several criteria like farm size, ethnicity, access to markets, etc. to assess impact of these factors on breeding criteria. The selection of farmers by farm size was done randomly, and selected farmers volunteered for participation. Farmers are brought to research stations to select lines which are subsequently put on farmers’ fields. Comment: By and large, participation of farmers and NGOs in research is good, and there are some very big NGOs like those in Brazil, which are participating effectively. To address some of these participatory issues, the CGIAR has constituted the CG-NGO Committee.

Notes
Author’s address: International Rice Research Institute, MCPO Box 3127, Makati City 1271, Philippines. Citation: Pandey S, Barah BC, Villano RA, Pal S. 2000. Risk analysis and management in rainfed rice systems. Limited Proceedings of the NCAP/IRRI Workshop on Risk Analysis and Management in Rainfed Rice Systems, 21-23 September 1998, National Centre for Agricultural Economics and Policy Research, New Delhi, India. Los Baños (Philippines): International Rice Research Institute.

Answer:

194

Participants

Adlakha, A Ballabh, Vishwa Dr. Barah, BC Dr. Bardhan Roy, SK Dr. Behura, DD Mr. Bhowmick, BC Dr. Chadha, GK Dr. Chandra, R Dr. Gogoi, J Dr. Guru, A Dr. Haque, T Dr Hossain, Mahabub Dr. Jain, R Dr. Janaiah, Aldas Dr. Jha, D. Dr. Joshi, PK Dr. Kaul, GL Dr. Krishnaiah, J Dr. Kumar, Anjani Dr. Kumar, P Dr. Madan, ML Dr. Marothia, D Dr. Mishra, PK Dr. Morin, Steve Dr. Mruthyunjaya, Dr. Naik, D Dr. O’Toole, J Dr. Pal, Suresh Dr. Palanisami, K Dr. Pandey, RK Dr Pandey, Sushil Dr. Prasana, L Dr Ram, Salik Dr. Ramasamy, C Dr.

Indian Rice Exporters Association Indian Institute of Rural Management National Centre for Agricultural Economics and Policy Research Directorate of Agriculture, West Bengal Orissa University of Agriculture and Technology Assam Agricultural University Jawaharlal Nehru University National Centre for Agricultural Economics and Policy Research Assam Agricultural University National Centre for Agricultural Economics and Policy Research National Centre for Agricultural Economics International Rice Research Institute National Centre for Agricultural Economics Directorate of Rice Research National Centre for Agricultural Economics National Centre for Agricultural Economics Indian Council for Agricultural Research and Policy Research and Policy Research and Policy Research and Policy Research

Andhra Pradesh Agricultural University National Centre for Agricultural Economics and Policy Research Indian Agricultural Research Institute Indian Council for Agricultural Research Commission on Agricultural Costs and Prices Government of Gujarat International Rice Research Institute Indian Council for Agricultural Research Orissa University of Agriculture and Technology Rockefeller Foundation National Centre for Agricultural Economics and Policy Research Tamil Nadu Agricultural University Indian Agricultural Statistics and Research Institute International Rice Research Institute National Centre for Agricultural Economics and Policy Research Central Rice Research Institute Tamil Nadu Agricultural University 195

Rana, Subrata Mr. Roy, BC Dr. Saha, NK Dr. Samal, P Dr. Sehara, DBS Dr. Selvarajan, S Dr. Sen, A Dr. Sharma, SD Dr. Sharma, SK Dr. Singh, G Dr. Singh, HN Dr. Singh, Parmanta Dr. Singh, Pratap Dr. Singh, RB Dr. Singh, RP Dr Thakur, Jawahar Dr. Villano, Renato Mr. Wade, Len Dr.

International Institute for Rural Reconstruction National Centre for Agricultural Economics and Policy Research Directorate of Agriculture, West Bengal Central Rice Research Institute Indian Council for Agricultural Research National Centre for Agricultural Economics and Policy Research Commission on Agricultural Costs and Prices Indian Agricultural Statistics and Research Institute Indira Ghandi Agricultural University National Centre for Agricultural Economics and Policy Research Narendra Deva University of Agriculture and Technology Indian Agricultural Research Institute National Centre for Agricultural Economics and Policy Research Indian Agricultural Research Institute Indian Agricultural Research institute Rajendra Agricultural University International Rice Research Institute International Rice Research Institute

196

Sign up to vote on this title
UsefulNot useful