Authors Accepted Manuscript

Coping with Climate Change and its Impact on

Productivity, Income, and Poverty: Evidence from
the Himalayan Region of Pakistan

Dil Bahadur Rahut, Akhter Ali

www.elsevier.com/locate/ijdr

PII: S2212-4209(16)30558-1
DOI: http://dx.doi.org/10.1016/j.ijdrr.2017.05.006
Reference: IJDRR554
To appear in: International Journal of Disaster Risk Reduction
Received date: 3 October 2016
Revised date: 5 April 2017
Accepted date: 14 May 2017
Cite this article as: Dil Bahadur Rahut and Akhter Ali, Coping with Climate
Change and its Impact on Productivity, Income, and Poverty: Evidence from the
Himalayan Region of Pakistan, International Journal of Disaster Risk
Reduction, http://dx.doi.org/10.1016/j.ijdrr.2017.05.006
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Coping with Climate Change and its Impact on Productivity, Income, and Poverty:
Evidence from the Himalayan Region of Pakistan

Dil Bahadur Rahut1, Akhter Ali2*

1
International Maize and Wheat Improvement Center (CIMMYT), Carretera Mxico-Veracruz
Km. 45, El Batn, Texcoco, Mxico, C.P. 56237.

2
International Maize and Wheat Improvement Center (CIMMYT), CSI Complex, NARC, Park
Road, Islamabad, Pakistan

*
Corresponding Author. Akhter Ali, Agricultural Economist, International Maize and Wheat
Improvement Center (CIMMYT), CSI Complex, NARC, Park Road, Islamabad, Pakistan. Phone
Number ++92-51-9255052. E-Mail akhter.ali@cgiar.org

Abstract

The current study explores the climate-risk mitigating strategies adopted by farmers and their
impact on household income, poverty levels and wheat yield in the Himalayan region of
Pakistan. This study is based on a primary dataset collected from 500 farmers through a field
survey from the seven districts of the Himalayan region of northern Pakistan: Ghizer, Gilgit,
Diamer, Astore, Skardu, Ghance, and Hunza-Nagar. A multivariate probit was employed to
estimate the determinants of choice of adaptation strategies by farmers and a propensity score
matching approach was used to estimate the impact of coping strategies on wellbeing. The most
common climate-risk management strategies adopted at farm level include an adjustment in the
sowing time, adoption of resistant varieties, tree plantation, non-farm participation as well as
crop-livestock interaction. The impact of all these risk-management strategies was estimated
individually as well as jointly. The empirical results indicated that farmers with higher levels of
education and secured land rights mostly adopt more climate-risk mitigating strategies. The
overall PSM results show that household income levels are high (in the range of PKR 1658-2610
per month), indicating that those households adopting climate-risk management strategies have

Page 1 of 30
higher income levels. The ATT results for the wheat yield are in the range of 42-65 kilogram per
hectare, indicating that farmers adopting climate-risk management strategies have higher wheat
yields. The ATT results for poverty are negative and significant, indicating that households
adopting climate-risk management strategies have lower poverty levels (in the range of 2%-4%).
Increasing public awareness and supporting households in risk-mitigating strategies in the
Himalayan region could be a crucial policy measure in reducing vulnerability to climate change,
thereby increasing wellbeing.

Keywords: Climate Change, Himalayas, Propensity Score Matching, Poverty, Pakistan

JEL Classification: Q54, P36

1. Introduction

Climate change is pervasive in nature, posing huge threats to sustainable development and
poverty reduction (United Nations Framework Convention on Climate Change, 2007, Rod and
Comin, 2003, Stern, 2006)). Detrimental effects of climate changes are manifesting throughout
the world as floods, drought, dry seasons, tornados, erratic rainfall, rising temperature, receding
glaciers, etc. The world is experiencing 400 to 500 abnormal disasters each year now, compared
to only 125 yearly in the 1980s (Maon et al., 2009). Socio-economic conditions and geographical
location of most of the developing countries makes them more vulnerable to the effects of
climate change; their capabilities to adapt to the changes and limited resources will worsen the
situation (Maskrey et al., 2007). Climate change has adversely affected the developmental efforts
of the government and international communities (Rahman and Rahman, 2015; UNFCCC, 2007;
World Bank, 2014, 2013, 2009, 2006). Agricultural productivity has been declining over the
years due to climatic variabilities (Cline, 2007; Hussain, 2011; INCCA, 2010; Lobell et al., 2012,
2011). Though the effect of climate change differs across season and by crop, a negative relation
between climate change and crop yields is found in South Asia (Asseng et al., 2015; Knox et al.,
2012; Lal, 2011). Hence countries heavily dependent on agriculture are most vulnerable to the

Page 2 of 30
impacts of climate change (Foresight, 2011; IPCC, 2014, 2012). Poor rural farm households in
developing countries are the most vulnerable to climate impacts.

In the Himalayas, there is a widespread feeling that the weather is becoming warmer and,
despite the melting snows and glaciers, the water sources are drying up (Chaudhary and Bawa,
2011). Water availability has become inconsistent and extreme. The recurring incident of
drought, severe floods, landslides and mixed effects in crops have been experienced in Nepal
because of climate change (Malla, 2009).

The dominant component of the climate variation was solely the changes in rainfall patterns
(Rod and Comin, 2003).The variation in precipitation patterns straightforwardly influences
water, farming, and the disaster management sectors. In the last two decades, there has been an
increase in the incidence, frequency, and intensity of extreme climatic events: about 40% of
Pakistanis are highly susceptible to multiple recurrent calamities with changes in rainfall
patterns, storms, floods, and droughts (Hussain et al., 2010). In many regions, rainfall patterns
are highly unreliable and unpredictable making it troublesome for individuals to take meaningful
measures for their safety, crops, and animals (Salma et al., 2012).

Establishment of adaptive capacity, knowledge, and governance and the adaptation itself are
crucial for effective human adaptation to climate change (Mirza, 2011). However, climate-
change induced risks at the rate and scale projected in the greater Himalayas cannot be
obliterated by a natural process of gradual adaptation. People must struggle to reduce future
negative consequences. By using traditional ecological knowledge and customs, inhabitants in
the hilly areas can cope with natural hazards (Byg and Salick, 2009). Moreover, adaptations of
mitigation strategies such as conservation agriculture, cover cropping, migration to less weather-
prone areas and usage of weather-resistant varieties were almost negligible. Conservation
agriculture could be the most viable strategy, so farmers have to utilize their existing resources in
judicious ways (Ashraf and Iftikhar, 2013). In recent years, farm households in developing
countries have started adopting available low-cost strategies to cope with the deleterious effect of
climate change. However, the poor farm household in the rural areas are highly vulnerable
because of the inadequate ex-ante risk management and ex-post shock coping capability
(Kochar, 1995)

Page 3 of 30
 Depending upon the capabilities, seasonal and long-term migration is usually sought by poor
farm household to compensate for the income loss due to changes in the climatic condition.
Migration is adopted as a core response to environmental pressure, both as an ex-ante measure of
risk mitigation strategy to cope with the uncertain agricultural production and as an ex-post
coping mechanism in the wake of environmental shocks (Gioli et al., 2014). The ex-ante coping
strategies are those which are implemented ahead of the shock such that the effect of the shocks
are minimized- it includes new agricultural technology, irrigation methods, livelihood
diversification. The ex-post coping strategies include those which is implemented after the
shocks to cope with the damage caused by the shocks- it includes migration, borrowing,
assistance from the government, etc. The adverse effect of the of shocks on depends on the
households capacity to adopt efficient ex-ante mitigation approaches such as livelihood
diversification, adoption of new agricultural technologies, investment in physical, human and
financial capital, as well as building up buffer stocks and savings (Dercon 2002, Morduch 1995,
Townsend 1995). Climate change induces migration and violent conflict, and the decision to
migrate is dependent on mitigation capabilities and the intensity of problems (Reuveny, 2007).

The temperature increase has a significant negative impact on agriculture production, while
an increase in revenue was envisioned with the increase in rainfall; however, the overall extent of
the adverse effect of changes in temperature is greater than the positive effect of rainfall in the
region (Shakoor et al., 2011). Diffusion of new farming techniques including new irrigation
methods, new methods of farm management and adapted cropping patterns would be appropriate
derivatives of the paradigm shift required in the agriculture sector of arid regions (Shakoor et al.,
2011).

In the past, not many studies have focused on the climate change adaptation strategies in
Pakistan especially in the Himalayan region. This study provide many novel contributions to
existing knowledge on disaster risk reduction with the following objectives; (a) examine the
farmer's adaptation strategies to climate change:- (b) to investigate the determinants of coping
strategies to climate change; (c) evaluate the impact of risk-coping strategies on poverty and
well-being of farmers. Therefore, the main contribution of the current paper is several folds.
Firstly it investigates the climate risk coping strategies adopted by the rural farm household in
the Himalayan region of Pakistan. To the best of our understanding, this is the first attempt that

Page 4 of 30
has been made to explore the climate risk management strategies in the Himalayan region of
Pakistan, hence contributing to the bodies of the literature on the climate change. Secondly, it
estimates the determinants of the adoption of different kind fo climate risk coping strategies.
Thirdly using propensity score matching, it assesses the impact of climate-change risk-mitigating
strategies on the productivity and well-being of the household at the farm level, which is the
unique contribution to the literature on the climate risk coping strategies.

In Section 2, the empirical model is presented; in Section 3, data and a description of

variables are presented; in Section 4, the empirical results are explained, and the paper concludes
with policy recommendations in Section 5.

2. Methodology and Data sampling

Since the dependent variable of the of the adoption of a coping model discrete is discrete,
there are several models available such as logit, probit, multinomial logit and multivariate probit.
As there are five mutually inclusive strategies, the best option would be multivariate probit is the
right model and it is confirmed by the cross-equation correlations reported in Table 3. One can
perform both ex-ante and ex-post analysis on the impact of the climate change on the yield and
income.

Ex-ante impact analysis of the climate change is done with economic model such as (a)
IMPACT model (https://www.ifpri.org/program/impact-model) developed by International Food
Policy Research Institute (IFPRI) and widely used by CGIAR Centres; (b) GEMPACK General
Equilibrium Software (https://www.copsmodels.com/gempack.htm) (Codsi and Pearson, 1988,
Harrison and Pearson, 1996) developed by Victoria University Austalia. Crop modelling
software are also used to analyze the impact of climate change on yield such as (a) Decision
Support System for Agrotechnology Transfer (DSSAT) (http://dssat.net/) (Hoogenboom et al.,
2004, Jones et al., 1998); (b) CropSyst developed by Department of Biological Systems
Engineering, Washington DC (Stckle et al., 2003, Stockle et al., 1994); (c) Agricultural

Page 5 of 30
Production Systems sIMulator (APSIM) created by Commonwealth Scientific and Industrial
Research Organisation (CSIRO) (McCown et al., 1996, McCown et al., 1995).

Ex-post impact analysis can be done with regression analysis, but it does not take into
account the selection biases and endogeneity problems. Hence the most after sought approach for
ex-post impact analysis is propensity score matching (PSM), which was first developed by
(Rosenbaum and Rubin, 1983) and implemented by several study to analyze the impact (Ali et
al., 2014, Ali and Rahut, 2013, Ali et al., 2015, Rahut et al., 2016a, Rahut et al., 2016b)

2.1 Multivariate probit model

Households can use multiple coping strategies at the same time, which are correlated. As the
farm household adopts multiple climates risks coping strategy, a multivariate probit econometric
methods was employed to analyze the determinants of a household's choice of coping strategy. In
the multivariate probit model, different types of coping strategies such as sowing time
adjustment, resistant varieties, tree plantation, non-farm participation and crop-livestock
interaction are considered dependent variables and the independent variables include household
demographic characteristics, labor supply, human and physical capital, and location dummies.
One advantage of the multivariate probit model is that, unlike single-equation probit and logit
models, the multivariate probit model simultaneously analyzes the choice of climate change
adaptation strategy. In formulating the multivariate probit econometrics model, we follow the
standard methods from other studies (Lin et al., 2005, Velandia et al., 2009)

The multivariate probit model is employed as the households use five types of climate risk
coping strategy: drought-resistant varieties, participation in non-farm activities, adjustment in the
sowing time, planting trees and crop-livestock interaction hence the multivariate probit model
will include five equations as follows:

In the five variate case, there are 10 joint probabilities corresponding to the 10 possible
combinations of adoption (yim=1) and non-adoption (yim=0). Where is a dependent variable
that assumes the value 1, when a household i uses sowing time adjustment strategy to adapt to
climate risk; or 0, otherwise, assumes the value 1 when a household uses drought-resistant

Page 6 of 30
varieties to adapt to climate risk; or 0, otherwise; assumes the value 1 when a household uses
planting trees to adapt to climate risk; or 0, otherwise, assumes the value 1 when a household
uses participation in non-farm activities to adapt to climate risk, or 0, otherwise, and assumes the
value 1 when a household uses crop-livestock interaction to adapt to climate risk, or 0, otherwise,
m is the number of equations in this case m=5 as households use five categories of fuel for
domestic works. is the dependent variables, which is explained in the annex. , are the
parameters to be estimated and is the random error term.

For the number of strategies adopted by the farmer at farm level Poisson regression model
has been estimated. For count data mostly Poisson regression and negative binomial logit models
are used and in the current analysis, poison regression model is employed because the normal
distribution is assumed.

2.2 Propensity score matching

In the Propensity score matching (PSM) approach the expected treatment effect for the
treated population is of primary significance. This effect may be given as:

| I 1 E ( | I 1) E ( R1 | I 1) E ( R0 | I 1) (3)

is the average treatment effect for the experiment/treated (ATT), R1 denotes the value of the
outcome for the household adopting a climate-change adaptation strategy and R0 is the value of

the same variable for not adopting a climate-change adaptation strategy. As shown above, the
major problem is that we do not observe E ( R0 | I 1) . Although the difference

[ e E ( R1 | I 1) E ( R0 | I 0] can be projected, it is possibly a biased estimator.

PSM approach can be employed to account for this sample selection bias if experimental
data is not available (Dehejia and Wahba, 2002). The PSM is a conditional probability that a
household adopts a climate change adaptation strategy, given pre-adoption attributes
(Rosenbaum and Rubin, 1983). To build the situation of a randomized experiment, the PSM uses

Page 7 of 30
the unconfoundedness presumption, which is also recognized as conditional independence
assumption (CIA), which implies that once Z is controlled for, adoption of coping strategy is
random and uncorrelated with the outcome variables. The PSM can be expressed as:

p(Z ) Pr{I 1 | Z} E{I | Z} (4)

where I = is the indicator for adoption and Z is the vector of pre-adoption of climate change
coping trait. The conditional distribution of Z, given p(Z) is alike in both categories of non-
adopters and adopters.

Once the propensity scores are estimated, the average treatment effect for the treated
(ATT) can then be estimated as:

E{R1 R0 | I 1} E{E{R1 R0 | I 1, p(Z )}} E{E{R1 | I 1, p(Z )} E{R0 | I 0, p(Z )} | I 0}

(5)

Several techniques have been developed to match households adopting climate change
adaptation strategy and those not adopting climate change adaptation strategy with same
propensity scores. Propensity score matching rests on two strong assumptions, i.e. the
conditional independence assumption and the common support condition. Nearest-neighbor
matching, radius matching, kernel-based matching and mahalnobis metric matching are some of
the commonly used matching algorithms which can be employed to estimate the propensity score
matching. The nearest-neighbor matching matches with the nearest neighbor only, while the
kernel-based matching takes into account the weighted average of the non-adopters and matches.
The radius matching is an extension of nearest-neighbor matching. In the current analysis two
methods i.e. nearest neighbor matching and kernel based matching are used.

After matching, the matching quality is assessed, and the standard errors and treatment
effects are estimated. The matching quality is evaluated by using a number of balancing tests
2
such as reduction in the median absolute bias pre and post matching, the value of R pre and
post matching and the p-value of joint significance of covariates pre and post matching (Becker
and Ichino, 2002, Caliendo and Kopeinig, 2008). The guidelines regarding implementation of the
propensity score matching can also be found in Caliendo and Kopeinig (2008).

2.3 Data and Sampling

Page 8 of 30
 Using a comprehensive questionnaire, information on household, farm level, and climatic
variables was collected from 500 rural households in the Himalayan region of Pakistan. The
enumerators were trained, the questionnaire was pilot tested, and the survey was implemented
between September-December, 2015. The data was collected from all seven districts: Gilgit-
Baltistan - Ghizer, Gilgit, Diamer, Astore, Skardu, Ghance and Hunza-Nagar (Figure 1).
Stratified random sampling approach was adopted for identifying sample and data collection. In
the first stage we selected seven districts, and in the second stage, the sample was randomly
selected. The distribution of the sample is provided in Table 1 and figure 1.

Gilgit-Baltistan (GB) is the meeting point of the worlds famous three high altitude mountain
ranges namely Himalayas, Karakoram, and the Hindu Kush. GB covers an area of 72,971 square
kilometers and has an estimated population of 1.25 million. Gilgit-Baltistan consists of seven
districts, and for the current study, the data was collected from all the seven districts.

3. Results and discussion

3.1 Descriptive Statistics

The data was collected from the person who was directly involved in farming i.e. farmer
himself or the household head. From one household only one person was interviewed. The
description of variables used in the paper is presented in Table 2. The mean age of the farmer
was 52, and the education level of the farmer was about six years of schooling. Average numbers
of male and female in the households are four each.

An overwhelming majority of the farmers, i.e. 82%, owned land and the rest were the
tenants. About 46% of the farmers were members of an organization. As Gilgit-Baltistan is a
mountainous area and the farmers have small land holdings, the most popular unit used for land
measurement is the kanali. The average landholding owned by the farmers are 30 kanals (1.5
hectares) and of this about half of the land, i.e. 14.53 Kanals (0.75 hectares), is wasteland. Only
13% of the households have access to a metal roadii.

Page 9 of 30
 Most of the respondents, i.e. 83%, were experiencing climate changes in the Himalayan
region. About 93% of the interviewees were experiencing a change in the temperature, 68% were
experiencing a change in the winds, and 72% were experiencing changes in rainfall. To combat
the climate challenge, the farmers adopted a number of coping strategies like drought-resistant
varieties, participation in non-farm activities, adjustment in the sowing time, planting trees and
crop-livestock interaction. These strategies were included in the questionnaires based on the pilot
studies.

Information about household assets indicates that about 41% of the households own a
television, 6% own a tractor, 83% have a kasola and 36% own a refrigerator. About 18% of the
households have access to credit, and 22% have access to agricultural extension services. The
details about outcome variables indicate that total wheat produced by the household was 56
maunds. Average monthly household income for the sampled household was rupees 33614. For
poverty, the head count index was estimated and about 24 percent of the population was living
below the poverty line. We employed the descriptive statistical analysis instead of scatter plot of
the scatter plot of adopters and non-adopters for different variables to provide the overview of
the variables used in the study because the descriptive statistics provides less space and more
information and allows quick and easy comparison.

3.2 Empirical result

3.2.1 Determinants of the coping strategies- multivariate probit model

Climate-risk mitigating strategies adopted by households at farm level were estimated jointly
by employing a multivariate probit model. The multivariate probit model was used as the
household may choose to adopt more than one climate-risk coping strategy; hence, the choice
should be estimated jointly. The cross-equation correlations are positive and highly significant at
the 1% level of significance, indicating the robustness of the estimated modeliii. We grouped the
climate-change risk-coping strategies adopted by the farmers into the following main strategies:
sowing time adjustment, resistant varieties, tree plantation, non-farm participation as well as crop
and livestock interaction. Based on the literature review and economic theory, a set of socio-
economic, household and farm-level variables were used in the analysis as explanatory variables.
The results of the multivariate probit are presented in Table 3.

Page 10 of 30
Demographic

The farmer s age was included, and the coefficient is positive and significant for sowing time
adjustment and crop-livestock interaction while negative and significant in the case of the
adoption of resistant varieties and non-farm participation. It shows that the household with a
younger head tends to adopt resistant varieties and non-farm livelihood activities as a coping
mechanism against climate risk because younger individuals have awareness and the necessary
skills. Due to lack of awareness about new seed varieties and skills needed for non-farm
employment, the older household adopts sowing time adjustment and crop-livestock interaction
as climate-risk coping strategies. The coefficient of the number of adult males in the households
is positive and significant for all climatic-risk coping strategies, indicating that households with
more adult males are more likely to adopt climatic-risk coping strategies. Other studies also
confirmed the positive and significant relation between household size and adaptation strategies
(Abid et al., 2015, Deressa et al., 2009, Croppenstedt et al., 2003).

Human capital

The coefficient of the farmers education in years is positive and significant for all four
strategies, indicating that educated farmers are more likely to adopt climate-risk mitigating
strategies which may be due to the awareness of the adverse effect of climate change and the
availability of climate-risk coping strategies. Other studies in Pakistan, and across the globe has
established the significant and positive relation between education and adoption of climate risk
management strategy (Abid et al., 2015, Bryan et al., 2013, Deressa et al., 2009, Hassan and
Nhemachena, 2008, Thomas et al., 2007).

Land assets

The coefficient of the farmer status dummy variable (1 for the owners and 0 for the tenant) is
positive and significant for sowing time adjustment, resistant varieties, non-farm participation
and crop-livestock interaction, while it is negative and significant for tree plantationiv. The
research by (Abid et al., 2015) found that the relationship between adaptation and tenancy were
negative and significant for change of planting dates, change of fertilizer and change of crop

Page 11 of 30
type. In the model, land holdings were included in number of kanals as a measure of wealth
status and the coefficient is negative and significant for sowing time adjustment and non-farm
participation, while it is positive and significant for resistant varieties and tree plantation. In
Punjab province of Pakistan, a study showed a positive and significant association between
landholding and changing crop varieties and crop types(Abid et al., 2015). Households with
more land are less likely to diversify into non-farm livelihoods and are more likely to invest in
tree plantation as it will have long-term benefits.

Household and agricultural assets (wealth)

A number of household assets like television, refrigerator, tractor, and kasola were included
in the model. The television ownership and tractor ownership dummy is positive and significant
for sowing time adjustment, resistant varieties, tree plantation, non-farm participation and crop-
livestock interaction while it was insignificant for refrigerator and kasola. The positive and
significant relationship between wealth and adoption of the climate risk coping strategies
indicates that richer farm households are able and more likely to invest in climate risk coping
strategies.

Social capital

Membership in an organization indicates the social capital/network of the household. The

coefficient of the organization membership dummy is positive and significant only for resistant
varieties, non-farm participation, and crop-livestock interaction, indicating those households with
membership in organizations are more likely to have a network and are able to access resistant
varieties and non-farm employment opportunities.

Access to facilities

Road access was included as a dummy variable, and the coefficient is positive and significant
for resistant varieties and non-farm participation, while it is negative and significant for sowing
time adjustment and tree plantation. The coefficient of the credit access dummy variable is
positive and significant, indicating that households that have access to credit mostly adopt these
climatic-risk coping strategies. Other studies in Pakistan indicated a positive relationship
between access to credit and climate risk adaptation strategies although non-significant (Abid et

Page 12 of 30
al., 2015). The agricultural extension dummy was included, and the results show that the sowing
time adjustment and adoption of resistant varieties were positive and significant. The agricultural
extension services provide information on new farming methods, inputs, irrigation methods
including methods to adapt to the changing climatic condition. To control for the locational fixed
effect, six district dummies were included in the model.

3.2.2 Determinants of the number of coping strategies- Poisson regression model

Farmers may adopt more than one strategies to minimize the impact of the climate change
related risk. As farmers adopt a number of climate-risk coping strategies at the farm level, the
Poisson regression model was estimated to ascertain the determinants of the number of climate-
risk management strategies adopted and the results are presented in Table 4.

Demographic

The farmers age was negative and significant at the 5% level of significance, indicating that
young farmers were adopting more climate-mitigating strategies, which can be due to awareness
about the impact of climate change and adaptation strategies among the younger generation. The
number of adult males is positive and significant, indicating that more the number of adult males
in a household, the more climate-risk coping strategies are adopted at the farm level.

Human capital

The coefficient of farmers education in years of schooling is positive and significant at the
1% level of significance, indicating that educated farmers were adopting more climate-risk
coping strategies. Farmers with more years of schooling are aware of the adverse impact of the
climate change, availability of the coping strategies and the benefit of adopting the climate risk
coping strategies, hence household with better education are more likely to adopt more number
of strategies.

Land assets

The farmers status dummy variable is positive and significant, pointing out that owners were
adopting more climate-risk mitigating strategies compared to tenants. The land holding is

Page 13 of 30
positive and significant at the 1% level of significance, indicating that households with more
land holdings adopt more climate risk coping strategies.

Household and agricultural assets (wealth)

A number of household assets were included in the model, i.e. television, refrigerator, tractor and
kasola were included in the model. The coefficients are positive and significant, indicating that
households having more assets adopt more climate-risk mitigating strategies. The tractor
coefficient is positive but non-significant.

Social capital

The farmers membership dummy variable is positive and significant, indicating that farmers
having organizational membership likely adopt more climate-risk coping strategies. Through the
social network, farmers are able to obtain information about different aspect of farming including
the knowledge about the climate change, the risk of climate change and the new technology to
combat the adverse consequences of the climate change. Hence the household who are the
member of the village organization are aware of the climate risk and tool to manage the risk and
are more likely to adopt them.

Access to facilities

The road access dummy is positive but non-significant. The coefficient of credit access
dummy is positive and significant, indicating that households with access to credit facilities
normally adopt more coping strategies. The agricultural extension dummy is positive and
significant, indicating that farmers having access to agricultural extension services normally
adopt more coping strategies. The district dummies were included in the model to control for the
districts fixed effect.

The R-square value is quite high, indicating the high explanatory power of the explanatory
variables.

3.2.3 Impact of risk-coping strategies on poverty and wellbeing

The impact of risk-coping strategies was estimated on household income, yield and poverty
levels. The impact of risk-coping strategies, i.e. sowing time adjustment, resistant varieties, tree

Page 14 of 30
plantation, non-farm participation and crop-livestock interaction, was estimated by employing
the propensity score matching (PSM) approach. The PSM approach was employed to correct for
the potential sample selection biasedness that might arise because of systematic differences
between the farmers having adopted these climatic-risk coping strategies and not adopting them.
The PSM analysis was estimated by employing the nearest-neighbor matching algorithmv and the
result is presented in Table 5. For the PSM analysis, the average treatment effect for the treated
(ATT) is the outcome between the household having adopted climate-risk management strategies
and not having adopted them.

The impact of all these strategies was estimated individually as well as jointly. The impact
of sowing time adjustment on household income is positive and significant. The ATT value of
the household income is PKR 1658, indicating that households having adjusted the sowing time
to cope with climate risks had higher income levels. Similarly, the ATT value for wheat yield
was 20.51 kilogram per hectare, indicating that households having adjusted the sowing time had
higher wheat yield levels. The impact on household poverty was negative and significant,
pointing out that poverty levels were less, down as much as 2% among the households that have
adopted adjusted the sowing time as the climate risk adaptation strategy.

Adoption of resistant varieties was another important climate-risk management strategy.

The impact of adopting resistant varieties on household income was positive and significant with
the ATT value of PKR 2610, indicating that households adopting the resistant varieties have
higher income levels. The wheat yield was higher in the range of 34.28 kilograms per hectare for
households adopting resistant varieties. The ATT results for poverty are negative and significant,
indicating that due to the adoption of resistant varieties the poverty levels are less, down by as
much as 4%.

The results were not significant for tree plantation, indicating the non-significant impact of
adopting this climate-risk coping strategy on household income levels, as well as wheat crop
yields and poverty levels. In the long run, this strategy might help to deal with the climate risk,
but there is needs to explore further in future research.

Page 15 of 30
 Non-farm participation was used as another climate-risk coping strategy, and the results are
positive and significant for household income levels. The ATT value of PKR 2409 indicates
higher income levels for households adopting participation in non-farm activities as their
climate-risk coping strategy compared to households not adopting it.

The impact of the crop-livestock interaction strategy on household income levels is positive
and significant at the 10% level of significance. The ATT value for income is PKR 1348,
indicating higher income levels for households adopting the crop-livestock interaction strategy.
However the ATT value for wheat yield is negative and non-significant, indicating lower wheat
yields of up to 17.29 kilogram per hectare, but it is not significant. Similarly, the results for
poverty are negative and non-significant. The critical level of hidden bias, also reported in Table
5, indicates the levels up to which the farmers having practiced climate-risk management
strategies and those not having practiced them differ from each other. The number of treated and
number of controls is also reported in Table 5.

The overall impact of the climate-change adaptation strategies on household income, wheat
yield, and poverty is presented in Table 6. The overall impact is estimated by using two different
matching algorithms, i.e. nearest-neighbor matching and kernel-based matching. The overall
impact of all the climate-risk coping strategies on household income is positive and highly
significant at the 1% level of significance both for NNM and KBM. The ATT value is in the
range of PKR 1658-2610, indicating that, due to the adoption of the coping strategies, household
income levels are high. The ATT results for the wheat yield are in the range of 42-65 kilogram
per hectare, indicating that farmers having adopted these coping strategies have higher wheat
yields. The ATT results for poverty are negative and significant, indicating that poverty levels
are lower by 2% to 4%, due to the adoption of climate risk coping strategies. This result in Table
5 clearly shows that adoption of the climate-risk coping strategies helps in mitigating the adverse
consequences of the climate change. Those households which implement climate risk have
higher income, higher yields, and lower poverty level compared to those who are not adopting
the climate risk mitigating strategy. A study in India found that with a rise in temperature by
2C, the wheat yield would decrease in all regions of India (Aggarwal & Sinha, 1993). A 4 C
rise in temperature in Terai region of Nepal will lead to a decline in wheat yield by 1.8% in Terai
region (Malla, 2008). An increase in temperature by 2-4C could increase a loss of, wheat

Page 16 of 30
production loss by more than 60% in Bangladesh (Karim, 1993). The current study shows that
the positive effect of the adoption of climate risk mitigating methods of the climate on yield,
income and poverty reduction, hence adverse consequences of the climate change could be
reduced. Similarly, other studies have revealed that the adoption of agricultural climate risk
coping mechanism reduces production losses due to decline in rainfall and increase in
temperature (Hassan and Nhemachena, 2008)

After carrying out the matching, the matching quality was checked by employing a number
of tests:- (a) median absolute bias before and after matching; (b) percentage decrease in bias
reduction; (c) The value of R-square before and after matching; and (d) the joint significance of
covariates before and after matching. The results from the matching quality are presented in
Table 6.

The median absolute bias is quite high before matching and is quite low after matching; the
bias before is in the range of 18-25 and, after matching, the bias is in the range of 4-6, indicating
a percentage bias reduction in the range of 68%-83%. Such a high percentage bias reduction
shows that, after matching, the group adopting climate risk management strategies and the group
not adopting them is very similar to each other.

The value of R-square, used as another measure to check the matching quality, is high
before matching and is much lower after matching. The value of R-square indicates that, after
matching, both groups are very similar to each other. Figure 2 clearly indicates that the matching
quality is excellent. The result from Table 7 and figure 2 on the matching quality shows that after
matching the group adopting climate risk management strategies and the group not adopting
them are very similar to each other. All the matching estimators compare the outcome of the
treated and non-treated (comparison) groups. Hence, the similarity between the treated and non-
treated groups indicates good matching quality and the impact result obtained in Table 4 are
reliable. The matching results are in line with previous studies(Ali and Abdulai, 2010, Ali and
Sharif, 2011).

Page 17 of 30
 4. Conclusion

Using comprehensive primary data collected from 500 farm households in the seven districts
of the Gilgit-Baltistan province in the Himalayan region of Pakistan, this paper adds many novel
contributions to the research on climate-change risk-coping strategies. First, the study focuses on
the climate-risk mitigating strategies adopted by the households in the Himalayan region of
Pakistan. Second, using primary data, it assesses the determinants of these coping strategies by
rural farm households in this area. Third, it assesses the impact of climate-risk coping strategies
on household income, wheat yield, and poverty levels.

The present study provides understanding about climate-related risks faced by farmers,
climate risk coping methods practiced by farmers, determinants of climate risk coping strategies
and its impact on the yield, income and poverty at farm household level in the Himalayan region
of Pakitan. Large majority of farmers were aware of the climate changes and changes in
temperature, variations in the rainfall pattern and wind as some of the major changes observed.
Predominant climate risk adaptation strategy adopted by the farm households in the Himalayan
regions adopt a number of climate risk management strategies like adjustment in the sowing
time, adoption of resistant varieties, non-farm participation and crop-livestock interaction.'
The empirical results indicated that educated farmers having land rights, large land holdings, and
more household assets are most likely to adopt a higher number of strategies at the farm level.
The impact of all these climate-risk management strategies was estimated by employing the
propensity score matching approach. The individual, as well as joint, impact of all these climate-
risk management strategies, is encouraging, indicating that households having adopted these
strategies have higher income levels, have higher wheat yields and less poverty. After matching
to check the matching quality, a number of balancing tests were employed which indicated the
robustness of the model estimated

Based on the finding of the current study we can draw the following policy recommendation
to reduce the farm household vulnerability to climate shocks. Firstly the result suggests the
importance of increasing awareness among the farming community regarding the climate risk,
adverse consequences of the climate change and benefit of adopting climate risk coping

Page 18 of 30
strategies through education, social network and extension services. Unless farmers are aware of
the climate risk, its consequences and mechanism to cope with it, it is unlikely that farmers will
adopt coping strategies. Secondly, the study suggests the significant role of the outreach and
extension services in the adoption of climate risk coping strategies. The outreach and extension
services will be able to train farmers on the issues related to climate and the coping mechanism
suitable for the particular location. Thirdly, the land holding and farmers status (tenancy versus
owner) play a major role in the adoption of the climate risk coping mechanism. Hence the land
reform policy focusing on land entitlement and distribution would be of paramount importance
in increasing the adoption of the climate risk coping strategies. Fourthly, the wealth of the farm
household (agricultural and non-agricultural) is also crucial determinants of adoption of climate
risk coping mechanism, hence to enable the poor farm household to adopt the climate risk coping
strategies, the government should aim at augmenting the capacity of resource-poor farm
household in implementing the climate risk coping mechanism. Finally, this study supports the
important role of climate risk coping strategies in reducing the vulnerability of poor farm
household by providing evidence that it reduces poverty, increase yield and income.

References
ABID, M., SCHEFFRAN, J., SCHNEIDER, U.A., ASHFAQ, M., 2015. Farmers' perceptions of
and adaptation strategies to climate change and their determinants: the case of Punjab
province, Pakistan. Earth System Dynamics, 6(1):225.
ALI, A., ABDULAI, A., 2010. The adoption of genetically modified cotton and poverty
reduction in Pakistan. Journal of Agricultural Economics, 61(1):175-192.
ALI, A., ERENSTEIN, O., RAHUT, D.B., 2014. Impact of direct rice-sowing technology on rice
producers earnings: empirical evidence from Pakistan. Development Studies Research,
1(1):244-254.
ALI, A., ERENSTEIN, O., RAHUT, D.B., 2015. Impact of gender participation in non-farming
activities on household income and poverty levels in Pakistan. Work, 52:345351.
ALI, A., RAHUT, D.B., 2013. Impact of Agricultural Extension Services on Technology
Adoption and Crops Yield: Empirical Evidence from Pakistan. Asian Journal of
Agriculture and rural Development, 3(11):801-812.
ALI, A., SHARIF, M., 2011. Impact of Integrated Weed Management on Cotton Producers'
Earnings in Pakistan*. Asian Economic Journal, 25(4):413-428.

Page 19 of 30
ASHRAF, S., IFTIKHAR, M., 2013. Mitigation and adaptation strategies for climate variability:
a case of cotton growers in the Punjab, Pakistan. International Journal of Agricultural
Extension, 1(1):30-35.
BECKER, S.O., ICHINO, A., 2002. Estimation of average treatment effects based on propensity
scores. The stata journal, 2(4):358-377.
BRYAN, E., RINGLER, C., OKOBA, B., RONCOLI, C., SILVESTRI, S., HERRERO, M.,
2013. Adapting agriculture to climate change in Kenya: Household strategies and
determinants. Journal of environmental management, 114:26-35.
BYG, A., SALICK, J., 2009. Local perspectives on a global phenomenonclimate change in
Eastern Tibetan villages. Global Environmental Change, 19(2):156-166.
CALIENDO, M., KOPEINIG, S., 2008. Some practical guidance for the implementation of
propensity score matching. Journal of economic surveys, 22(1):31-72.
CHAUDHARY, P., BAWA, K.S., 2011. Local perceptions of climate change validated by
scientific evidence in the Himalayas. Biology Letters:rsbl20110269.
CODSI, G., PEARSON, K.R., 1988. GEMPACK: General-purpose software for applied general
equilibrium and other economic modellers. Computational Economics, 1(3):189-207.
CROPPENSTEDT, A., DEMEKE, M., MESCHI, M.M., 2003. Technology adoption in the
presence of constraints: the case of fertilizer demand in Ethiopia. Review of Development
Economics, 7(1):58-70.
DEHEJIA, R.H., WAHBA, S., 2002. Propensity score-matching methods for nonexperimental
causal studies. Review of Economics and statistics, 84(1):151-161.
DERESSA, T.T., HASSAN, R.M., RINGLER, C., ALEMU, T., YESUF, M., 2009.
Determinants of farmers choice of adaptation methods to climate change in the Nile
Basin of Ethiopia. Global environmental change, 19(2):248-255.
GIOLI, G., KHAN, T., BISHT, S., SCHEFFRAN, J., 2014. Migration as an adaptation strategy
and its gendered implications: A case study from the Upper Indus Basin. Mountain
Research and Development, 34(3):255-265.
HARRISON, W.J., PEARSON, K.R., 1996. Computing solutions for large general equilibrium
models using GEMPACK. Computational Economics, 9(2):83-127.
HASSAN, R., NHEMACHENA, C., 2008. Determinants of African farmers strategies for
adapting to climate change: Multinomial choice analysis. African Journal of Agricultural
and Resource Economics, 2(1):83-104.
HOOGENBOOM, G., JONES, J., WILKENS, P., PORTER, C., BATCHELOR, W., HUNT, L.,
BOOTE, K., SINGH, U., URYASEV, O., BOWEN, W., 2004. Decision support system
for agrotechnology transfer version 4.0. University of Hawaii, Honolulu, HI (CD-ROM).
HUSSAIN, A., ZULQARNAIN, M., HUSSAIN, J., 2010. Catastrophes in the South Punjab Due
to Climate Change and the Role of PIDEANS. Center for Environmental Economics and
Climate Change (CEECC), PIDE; http://www.pide.org.pk/, Islamabad, Pakistan.
JONES, J., TSUJI, G., HOOGENBOOM, G., HUNT, L., THORNTON, P., WILKENS, P.,
IMAMURA, D., BOWEN, W., SINGH, U., 1998. Decision support system for
agrotechnology transfer: DSSAT v3. Understanding options for agricultural production.
Springer.
KOCHAR, A., 1995. Explaining household vulnerability to idiosyncratic income shocks. The
American Economic Review, 85(2):159-164.
LIN, C.-T.J., JENSEN, K.L., YEN, S.T., 2005. Awareness of foodborne pathogens among US
consumers. Food Quality and Preference, 16(5):401-412.

Page 20 of 30
MALLA, G., 2009. Climate change and its impact on Nepalese agriculture. Journal of
agriculture and environment, 9:62-71.
MAON, F., LINDGREEN, A., VANHAMME, J., 2009. Cross-Sector collaboration for disaster
relief supply chain enhancement: Mingling corporate expertise with humanitarians'
Willpower. Willpower, IAG-Louvain School of Management Working Paper.
http://www. uclouvain. be/cps/ucl/doc/iag/documents/WP_21_Maon_Lindgreen_
Vanhamme. pdf.
MASKREY, A., BUESCHER, G., PEDUZZI, P., SCHAERPF, C., 2007. Disaster Risk
Reduction: 2007 Global Review. Consultation Edition. Prepared for the Global Platform
for Disaster Risk Reduction First Session, Geneva, Switzerland:5-7.
MCCOWN, R., HAMMER, G., HARGREAVES, J., HOLZWORTH, D., FREEBAIRN, D.,
1996. APSIM: a novel software system for model development, model testing and
simulation in agricultural systems research. Agricultural systems, 50(3):255-271.
MCCOWN, R., HAMMER, G., HARGREAVES, J., HOLZWORTH, D., HUTH, N., 1995.
APSIM: an agricultural production system simulation model for operational research.
Mathematics and computers in simulation, 39(3-4):225-231.
MIRZA, M.M.Q., 2011. Climate change, flooding in South Asia and implications. Regional
Environmental Change, 11(1):95-107.
RAHUT, D.B., ALI, A., IMTIAZ, M., MOTTALEB, K.A., ERENSTEIN, O., 2016a. Impact of
irrigation water scarcity on rural household food security and income in Pakistan. Water
Science and Technology: Water Supply, 16(3):675-683.
RAHUT, D.B., BEHERA, B., ALI, A., 2016b. Do forest resources help increase rural household
income and alleviate rural poverty? Empirical evidence from Bhutan. Forests, Trees and
Livelihoods, 25(3):187-198.
REUVENY, R., 2007. Climate change-induced migration and violent conflict. Political
Geography, 26(6):656-673.
ROD, X., COMIN, F., 2003. Global climate: current research and uncertainties in the climate
system.
ROSENBAUM, P.R., RUBIN, D.B., 1983. The central role of the propensity score in
observational studies for causal effects. Biometrika, 70(1):41-55.
SALMA, S., REHMAN, S., SHAH, M., 2012. Rainfall trends in different climate zones of
Pakistan. Pakistan Journal of Meteorology (Pakistan), 9(17):37-47.
SHAKOOR, U., SABOOR, A., ALI, I., MOHSIN, A., 2011. Impact of climate change on
agriculture: empirical evidence from arid region. Pak. J. Agri. Sci, 48(4):327-333.
STERN, N., 2006. Stern review report on the economics of climate change.
STCKLE, C.O., DONATELLI, M., NELSON, R., 2003. CropSyst, a cropping systems
simulation model. European journal of agronomy, 18(3):289-307.
STOCKLE, C.O., MARTIN, S.A., CAMPBELL, G.S., 1994. CropSyst, a cropping systems
simulation model: water/nitrogen budgets and crop yield. Agricultural Systems,
46(3):335-359.
THOMAS, D.S., TWYMAN, C., OSBAHR, H., HEWITSON, B., 2007. Adaptation to climate
change and variability: farmer responses to intra-seasonal precipitation trends in South
Africa. Climatic change, 83(3):301-322.
UNITED NATIONS FRAMEWORK CONVENTION ON CLIMATE CHANGE, 2007. Climate
Change: Impacts, Vulnerabilities, and Adaptation in Developing Countries. Bonn,
Germany. United Nations Framework Convention on Climate Change Bonn.

Page 21 of 30
VELANDIA, M., REJESUS, R.M., KNIGHT, T.O., SHERRICK, B.J., 2009. Factors affecting
farmers' utilization of agricultural risk management tools: the case of crop insurance,
forward contracting, and spreading sales. Journal of agricultural and applied economics,
41(01):107-123.

Figure 1: Map showing sample distribution

Page 22 of 30
 0 .2 .4 .6 .8 1
Propensity Score
0 .2 .4 .6 .8 1
Propensity Score Untreated: Off support Untreated: On support
Treated: On support Treated: Off support
Untreated Treated

0 .2 .4 .6 .8 1
Propensity Score

Untreated: Off support Untreated: On support

Treated: On support Treated: Off support

Figure 2: Indicators of covariates balancing before and after matching

Table 1: Sample
Population Projected Sample size
Area (Square Census Population (Percentage)
kilometers) (1998-03-01) (2013-07-01)
Ghizer 12,381 120,218 190,000 15
Gilgit 4,208 145,272 222,000 13
Diamer 7,234 131,925 214,000 6
Astore 5,411 71,666 114,000 22

Page 23 of 30
 Skardu 19,697 214,848 305,000 19
Ghanche 8,531 88,366 108,000 18
Hunza-Nagar 15034 98,052 148,000 7
*In total 500 farmers were interviewed from the seven districts of Gilgit-Baltistan.
Source: City andPopulation (https://www.citypopulation.de/GilgitBaltistan.html)
Note:
Pupulation (1998) Government of Pakistan and IUCN (web).
Population (2013) Government of Gilgit-Baltistan, Planning and Development Department,
Statistical Cell.
In 2015 Hunza-Nagar district was split into Hunza district (46,665 inhabitants in 1998) and
Nagar district (51,387) as well as Skardu district was split into Kharmang district, Shigar district
and a remaining Skardu district.

Table 2: Data and description of variables

Variable Description Mean Std.
Dev
Dependent variables
Drought-resistant 1 if the household has planted drought-resistant 0.31 0.24
varieties varieties, 0 otherwise
Non-farm 1 if the household participated in non-farm activities, 0.38 0.29
participation 0 otherwise
Sowing time 1 if the household has adjusted sowing time, 0 0.27 0.21
adjustment otherwise
Tree plantation 1 if the household has planted trees, 0 otherwise 0.45 0.34
Crop-livestock 1 if the household have both crops and livestock, 0 0.69 0.37
interaction otherwise
Independent variable
Demographic
Farmers age Age of the farmers in years 52.09 13.05
Head age Age of the household head in years 56 13.67
Adult males Number of adult males in the household 4.12 0.26
Adult females Number of adult females in the household 4.05 0.34
Human capital
Farmers education Education of the farmer in years 6.07 5.75
Head education Education of the head in years 5.8 5.8
Land assets
Farmer status 1 if the farmer is owner, 0 otherwise 0.82 0.51
Own land holdings Land owned by the farmers in number of Kanals 29.88 24.69
Household and agricultural assets
Television 1 if the household owns a TV, 0 otherwise 0.41 0.33
Refrigerator 1 if the household owns a refrigerator, 0 otherwise 0.36 0.24

Page 24 of 30
Tractor 1 if the household owns a tractor, 0 otherwise 0.06 0.17
Trolley 1 if the household owns a trolley, 0 otherwise 0.10 0.32
Kasola 1 if the household owns a kasola, 0 otherwise 0.83 0.27

Social capital
Membership 1 if the farmer is a member of organization, 0 0.46 0.31
otherwise
Access to facilities
Road access 1 if the household has access to road, 0 otherwise 0.13 0.28
Access to credit 1 if the household has access to credit facility, 0 0.18 0.23
otherwise
Agri. extension 1 if the household has access to agri. extension 0.22 0.29
facility, 0 otherwise
Welfare
Wheat yield Total wheat produced by the household in maunds 56.28 26.76
Income Monthly income of the household in Pak rupees 33614 2247.3
Poverty 1 if the household is living below the poverty line, 0 0.24 0.19
otherwise
Districts
Gilgit 1 if the farmer is from Gilgit district, 0 otherwise 0.13 0.17
Skardu 1 if the farmer is from Skardu district, 0 otherwise 0.19 0.12
Astore 1 if the farmer is from Astore district, 0 otherwise 0.22 0.18
Ghizer 1 if the farmer is from Ghizer district, 0 otherwise 0.15 0.14
Ghanche 1 if the farmer is from Ghanche district, 0 otherwise 0.18 0.13
Hunza Nagar 1 if the farmer is from Hunza-Nagar district, 0 0.07 0.14
otherwise
Diamer 1 if the farmer is from Diamer district, 0 otherwise 0.06 0.24
Other climate-related information (not used in the econometrics models)
Climate change 1 if the household is experiencing climate change, 0 0.83 0.34
otherwise
Temperature 1 if the household is experiencing change in 0.93 0.29
temperature, 0 otherwise
Winds 1 if the household is experiencing changes in winds 0.68 0.37
speed and outset, 0 otherwise
Rainfall 1 if the household is experiencing changes in rainfall, 0.72 0.29
0 otherwise
Source: Authors calculations.

Table 3: Climate-risk mitigating strategies adopted by the households (multivariate probit)

Variable Sowing time Resistant Trees Non-farm Crop-
adjustment varieties plantation participation livestock
interaction
Demographic
Farmers age 0.03***(3.17) -0.01**(2.11) 0.01(0.64) -0.04**(-2.18) 0.02**(2.13)
Adult males 0.02***(2.19) 0.02***(3.02) 0.03***(3.17) 0.01**(2.16) 0.02**(2.11)

Page 25 of 30
Human capital
Farmers education 0.01**(2.19) 0.03***(2.54) 0.01**(2.19) 0.01*(1.75) 0.01(1.26)
Land assets
Farmers statusa,b 0.01***(2.75) 0.01**(2.13) -0.02*(1.85) 0.01**(2.14) 0.03***(2.89)
Own land holdings -0.02***(-3.11) 0.03*(1.68) 0.04**(2.05) -0.01*(-1.63) -0.02(-1.30)
Household and agricultural assets
Televisiona,c 0.03**(2.01) 0.01**(2.26) 0.03***(2.58) -0.02(-1.05) 0.03(1.24)
a,d
Refrigerator 0.03(1.45) 0.02(1.31) -0.01(-1.26) -0.03(-1.46) 0.02(1.38)
Tractora,e 0.02**(3.16) 0.02**(2.49) 0.01(1.29) 0.03***(2.91) 0.02**(2.18)
a,f
Kasola 0.01(1.41) -0.02(-1.63) 0.03(1.50) 0.02(1.39) -0.01(-1.37)
Social capital
Membershipa,g 0.01(1.32) 0.03*(1.74) 0.03(1.24) 0.02**(2.31) 0.04*(1.67)
Access to facilities
Road accessa,h -0.03**(-2.08) 0.01**(2.17) -0.02*(-1.70) 0.03***(3.05) 0.01**(2.04)
Credita,i 0.01***(2.78) 0.03**(2.10) 0.03*(1.67) 0.01(1.28) 0.03*(1.70)
a,j
Agri. extension 0.01***(2.56) 0.02**(3.17) 0.03(0.82) 0.01(1.25) -0.03(-0.26)
Districts
Ghizera,k 0.02**(2.17) 0.01***(2.84) 0.01*(1.85) 0.02(0.74) 0.01(1.49)
Gilgita,k 0.01(1.60) 0.02(1.65) 0.02*(1.97) 0.03(1.46) 0.02(1.32)
a,k
Ghanche 0.02(1.35) 0.01*(1.73) 0.01**(2.10) 0.03(1.90) 0.01(1.76)
Diamera,k 0.03(1.26) 0.02(1.24) 0.04*(1.78) 0.02(1.44) 0.03*(1.75)
a,k
Skardu 0.04*(1.85) 0.01**(2.17) 0.03***(3.05) 0.02*(1.83) 0.01**(2.16)
Hunza-Nagara,k 0.02*(1.86) 0.02**(2.15) 0.01(1.30) 0.02(1.49) 0.02*(1.84)
Constant 0.02(1.47) 0.01(1.33) 0.02(1.67) 0.03(1.03) 0.02**(2.17)
Cross Equation 12 13 14 15 23
Correlations 0.27***(2.64) 0.21***(2.92) 0.35***(3.18) 0.26***(2.48) 0.23**(2.17)
24 25 34 35 45
0.22***(3.08) 0.20*(1.94) 0.24***(2.64) 0.31**(2.37) 0.26***(2.58)
R 2 0.28 0.23 0.27 0.34 0.29
LR 2 135.27*** 146.91*** 121.26*** 173.85*** 163.20***
Prob> 2 0.000 0.000 0.000 0.000 0.000
*** ** *
Notes: = 1% level of significance, = 5% level of significance, = 10% level of significance. t-values
are reported in parenthesis.
a
dummy variables; b excluded category: Tenant farmer; cexcluded category: Households do not own
television; dexcluded category: Households do not own refrigerator; eexcluded category: Households do
not own tractor; fexcluded category: Households do not own kasola; gexcluded category: Household is
not a member of any organization; hexcluded category: No access to road; iexcluded category: Households
do have access to credit; jexcluded category: Households do have access to agricultural extension agent;
k
excluded category:Astore

Page 26 of 30
Source: Authors calculations.

Table 4: Number of climate-mitigating strategies adopted at farm level (Poisson estimates)

Variable Coefficient t-values
Demographic
Farmers age -0.02** -2.04
Adult males 0.01*** 2.57
Human capital
Farmers education 0.04*** 3.14
Land assets
Farmers statusa,b 0.01*** 2.65
Own land holdings 0.03*** 2.91
Household and agricultural assets
Televisiona,c 0.02*** 3.19
a,d
Refrigerator 0.01*** 3.05
Tractora,e 0.04 1.54
Kasolaa,f -0.02** -2.15
Social capital
Membershipa,g 0.02** 2.13
Access to facilities
Road accessa,h 0.02 1.35
a,i
Credit 0.02** 2.16
Agri. extensiona,j 0.01*** 2.73
Districts
Ghizera,k -0.02*** -2.14
a,k
Gilgit -0.01** -2.08
Ghanchea,k 0.02*** 2.84
Diamera,k 0.01* 1.75
Skardua,k -0.03*** -2.94
Hunza-Nagara,k 0.02 1.45
Constant 0.03** 2.16
R2 0.47
LR 2 138.29
Prob> 2 0.000
***
Notes: = 1% level of significance, = 5% level of significance, * = 10% level of significance. t-values are
**

reported in parenthesis.
a
dummy variables; b excluded category: Tenant farmer; cexcluded category: Households do not own television;
d
excluded category: Households do not own refrigerator; eexcluded category: Households do not own tractor;

Page 27 of 30
f
excluded category: Households do not own kasola; gexcluded category: Household is not a member of any
organization; hexcluded category: No access to road; iexcluded category: Households do have access to credit;
j
excluded category: Households do have access to agricultural extension agent; kexcluded category:Astore
Source: Authors calculations.

Table 5: Impact of climate change adaptation strategy on crop yields, household income
and poverty levels (matching algorithm nearest-neighbor matching)
Risk Outcome ATT t- Critical Number Number
management values level of of treated of control
strategy hidden bias
(A) (B)

Income 1658** 2.15 1.30-1.35 265 223

Sowing time Wheat yield 20.51*** 2.72 1.25-1.30 265 223
adjustment
Poverty -0.02*** 3.16 1.15-1.20 265 223
Income 2610*** 3.04 1.05-1.10 218 249
Resistant Wheat yield 34.28*** 3.14 1.20-1.25 218 249
varieties
Poverty -0.04*** -3.12 1.30-1.35 218 249
Income 873 1.26 - 257 180
Tree plantation Wheat yield 12.26 1.20 1.35-1.40 257 180
Poverty -0.01 0.53 - 257 180
Income 2409** 2.16 1.30-1.35 193 216
Non-farm Wheat yield 24.15*** 2.62 1.45-1.50 193 216
participation
Poverty -0.03** -2.17 1.20-1.25 193 216
Income 1348* 1.75 1.35-1.40 214 227
Crop-livestock Wheat yield -17.29** -0.93 - 214 227
interaction
Poverty -0.01 -0.62 - 214 227
Note: ATT stands for the average treatment affect for the treated. The results (***, **, * ) are significant at the 1%,
5% and 10% levels, respectively. The numbers in each section of the column (A) and (B) are different because
different calipers and bandwiths are used. Income is given in Pakistani rupees. 1 US$=106.30 Pakistani rupees.

Page 28 of 30
Table 6: Overall impact of the climate-change adaptation strategies on crop yields,
household income and poverty levels
Matching Outcome ATT t- Critical level Number Number
algorithm values of hidden bias of treated of control
(A) (B)

Nearest- Income 1658** 2.15 1.30-1.35 265 223

neighbor Wheat yield 20.51*** 2.72 1.25-1.30 265 223
matching Poverty -0.02*** 3.16 1.15-1.20 265 223
Kernel-based Income 2610*** 3.04 1.05-1.10 218 249
matching Wheat yield 34.28*** 3.14 1.20-1.25 218 249
Poverty -0.04*** -3.12 1.30-1.35 218 249
Note: ATT stands for the average treatment affect for the treated. The results (***, **, * ) are significant at the 1%,
5% and 10% levels, respectively. For the nearest-neighbor matching, the calipers are reported, while for the kernel-
based matching, the band widths are reported.The numbers in each section of the column (A) and (B) are different
because different calipers and bandwiths are used.

Table 7: Indicators of covariates balancing before and after matching (nearest-neighbor

matching-matching algorithm)
Risk Out- Median Median Percen- Value Value Joint Joint
man- come abso- abso- tage bias of R- of R- signifi- signifi-
agement lute lute reduc- square square cance of caace of
covari
strategy bias bias tion before after covari-
ates after
before after match- match- ates
matching
match- match- ing ing before
ing ing matching

Sowing Income 24.35 4.06 83 0.257 0.002 0.001 0.294

time Wheat 20.91 5.47 73 0.334 0.001 0.003 0.236
adjust- Yield
ment Poverty 23.69 5.11 78 0.336 0.003 0.002 0.275
Resistant Income 19.37 4.28 77 0.275 0.002 0.001 0.386
varieties Wheat 18.05 5.30 71 0.249 0.001 0.002 0.390
Yield
Poverty 19.83 4.27 78 0.340 0.003 0.001 0.450
Tree Income 21.70 5.69 74 0.257 0.002 0.003 0.431
plantation Wheat 23.11 4.32 81 0.428 0.001 0.001 0.362
Yield
Poverty 18.09 5.76 68 0.276 0.003 0.002 0.720
Non-farm Income 20.16 4.28 78 0.329 0.002 0.001 0.341
participa- Wheat 18.10 6.39 64 0.164 0.001 0.004 0.426
tion Yield
Poverty 19.20 4.25 77 0.248 0.002 0.003 0.382
Crop- Income 18.07 5.13 71 0.362 0.003 0.002 0.413
livestock Wheat 20.45 5.38 74 0.253 0.004 0.001 0.528

Page 29 of 30
interac- Yield
tion Poverty 25.19 4.19 83 0.284 0.002 0.004 0.379

Notes

i
Generally in Pakistan the unit used for land measurement is the number of acres. In 1 acre there are 8
Kanals and in 1 hectare there are 20 kanals.
ii
Due to hilly area and scattered establishments only few households have road access in the hilly areas.
iii
In multivariate probit model the cross equation correlations significance is an indication that these
needs to be estimated jointly.
iv
The tenants mostly plant fewer trees as they dont have the secured land rights.
v
There are number of matching algorithms i.e. nearest neighbor matching matches the similar farmer
having adopted the climatic risk management strategies with the similar farmer having not adopted the
climatic risk management strategies.

Page 30 of 30