Journal of the Asia Pacific Economy

ISSN: 1354-7860 (Print) 1469-9648 (Online) Journal homepage: https://www.tandfonline.com/loi/rjap20

Crop choice, farm income, and political control in

Myanmar

Takashi Kurosaki

To cite this article: Takashi Kurosaki (2008) Crop choice, farm income, and political
control in Myanmar, Journal of the Asia Pacific Economy, 13:2, 180-203, DOI:
10.1080/13547860801923582

To link to this article: https://doi.org/10.1080/13547860801923582

Published online: 27 Mar 2008.

Submit your article to this journal

Article views: 294

View related articles

Citing articles: 5 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rjap20
Journal of the Asia Pacific Economy
Vol. 13, No. 2, May 2008, 180–203

Crop choice, farm income, and political control in Myanmar

Takashi Kurosaki

The Institute of Economic Research, Hitotsubashi University, Kunitachi, Tokyo, Japan

Myanmar’s agricultural economy has been under transition from a planned to a

market system since the late 1980s and has experienced a substantial increase
in production. However, little research is available on the impact of economic
policies in this country on agricultural production decisions and rural incomes.
Therefore, this paper investigates the impact using a micro dataset collected
in 2001 and covering more than 500 households in eight villages with diverse
agro-ecological environments. Regression analyses focusing on within-village
variations in cropping patterns show that the acreage share under non-lucrative
paddy crops was higher for farmers who were under tighter control of the local
administration due to their political vulnerability. Simulation results based on
the regression estimates show that the loss in rural incomes due to farmers
being forced to grow too much paddy was not negligible.
Keywords: reform; rice policy; transitional economies; Myanmar

JEL Classifications: O53, Q18.

Introduction
The agricultural economy of Myanmar (formerly Burma) has been under transition
from a planned to a market system since the late 1980s. A literature survey by
Rozelle and Swinnen (2004) shows that such a transition is often associated with
a productivity gain in agriculture and such productivity gains were largest in
East Asia, where the gain also resulted in higher agricultural output. Although
Rozelle and Swinnen do not discuss Myanmar in their text, they include it in their
tables in the group of successful East Asian countries and their numbers show
a substantial increase in agricultural output in Myanmar (Rozelle and Swinnen
2004, Tables 1 and 2). However, due to a lack of data, they do not provide evidence
on the productivity/income gain in Myanmar. Even though the country was the
world’s largest exporter of rice in the pre-Second World War period and remains
one of the world’s largest producers of rice today, little research on Myanmar’s
rural economy is available and the effects of agricultural policies on production
and rural incomes are not well documented. A few studies, by the International

Email: kurosaki@ier.hit-u.ac.jp
ISSN: 1354-7860 print / 1469-9648 online

C 2008 Taylor & Francis
DOI: 10.1080/13547860801923582
http://www.tandf.co.uk/journals
 Journal of the Asia Pacific Economy 181

Table 1. Myanmar’s economy and agriculture.

1985/86 1990/91 1996/97 2000/01 2001/02 2002/03 2003/04

Growth rate of real GDP 2.9 2.8 6.4 13.7 11.3 12.0 13.8
Growth rate of 2.2 2.0 3.8 11.1 8.1 4.7 9.3
agricultural sector
Agricultural sector’s 39.7 38.7 36.2 33.6 47.4 44.3 42.6
share in GDP
Agricultural share in 42.4 31.8 36.1 18.2 17.6 14.1 16.6
exports
Agricultural share in 63.4
workforce
Total irrigated area 3.0 2.9 4.6 6.0 6.2 6.2 6.4
(million acres)
Share of paddy in the 70.1 74.8 82.3 76.5 76.5 73.5 73.6
total irrigated area (%)

Note: ‘Agricultural sector’ in this table does not include livestock, fishery, and forestry.
Source: CSO (various issues).

Rice Research Institute (Garcia et al. 2000), by Japanese economists (Takahashi

2000; Okamoto 2004, 2005; Fujita and Okamoto 2000; Fujita 2003), and by social
scientists from Myanmar (Thawnghmung 2004; Dolly Kyaw 2006), do exist, but
compared with the amount of research conducted on other East Asian countries,

Table 2. Survey villages.

Division/ Village
Name State Township tract Topology Irrigation Major crops

DELTA1 Ayeyarwady Myaungmya Kyonethout Deltaic Pump Paddy

Division agric.
DELTA2 Bago Waw Acarick Deltaic Rainfed, Paddy, pulses
Division agric. Canal
DRY1 Mandalay Kyaukse Pyiban Dry zone Canal Paddy,
Division vegetables
DRY2 Magway Magway Kanpyar Dry zone Rainfed Upland crops
Division
DRY3 Magway Taungdwingyi Wetkathay Dry zone Rainfed, Upland crops,
Division Tank paddy
HILL1 Shan State Nyaungshwe Linkin Hilly Rainfed Vegetables,
region paddy,
sugarcane
HILL2 Shan State Kalaw Myinmahti Hilly Rainfed Vegetables,
region paddy
COAST Tanintharyi Myeik Engamaw Coastal Rainfed Paddy, rubber
Division agric.

Source: Author’s survey (ibid. for the tables below).

182 T. Kurosaki

the case of Myanmar remains relatively unexamined. This study tries to fill the gap
by using a primary dataset that is more recent, provides more detailed information,
and covers more diverse regions than previous studies.
The broad objective of this paper is, therefore, to attempt to widen our under-
standing of the impact of economic policies in transitional economies through an
examination of the case of Myanmar in the early 2000s. A descriptive analysis
of the dataset reveals that income levels were lower in villages under the tight
control of the central authorities than in other villages, and farmers and villages
that emphasized a paddy-based, irrigated cropping system had lower farming in-
comes than farmers and villages that did not. This suggests a possibility that when
farmers were under tight political control, they allocated a higher share of their
resources to non-lucrative rice than would otherwise be the case.
The specific aim of this paper is to test this thesis econometrically. For this aim,
this paper focuses on the fact that the level of enforcement of government crop
planning varied not only across villages but also within villages across farmers.
Regression results focusing on the within-village variation of crop choices are
consistent with the theoretical prediction that the share of non-lucrative paddy
should be higher for farmers who are under tighter control of the local authorities.
The paper is organized as follows. The next section describes the background
and the dataset, showing a negative relation between paddy acreage shares and
per-acre farm incomes. To explain this situation, the third section develops a
theoretical model of crop choice where the extent to which the local administration
controls each farmer plays a key role. A prediction of the theoretical model to be
tested empirically is also derived in this section. The fourth section provides the
estimation results. Simulation results to quantify the impact of policy distortions
on household income are also presented. The fifth section concludes the paper.

Background and data1

Myanmar’s economy and agricultural policies
Myanmar, whose population is close to 50 million, is in transition from a planned
to a market economy. The military government that has been in power since 1988
has deregulated various economic activities. Industrial development is underway,
but the agricultural sector still remains dominant in the national economy
(Table 1).2 The estimated income level is among the lowest in the world. Rice
is the staple food in Myanmar, accounting for more than 20% of consumption
expenditure of the nation during the late 1990s (CSO various issues).
The government has given high priority to the expansion of paddy production,
because it believes that a stable supply of rice is a prerequisite for political stability.
To achieve this expansion, the government has been promoting the expansion of
paddy areas through irrigation investment. Throughout the 1990s, numerous dams
were constructed in some areas, while private investment in small scale diesel
pumps was promoted in others, in order to increase paddy cultivation in the
 Journal of the Asia Pacific Economy 183

dry season. As a result, both the area under cultivation and paddy production in
Myanmar rose remarkably in the early 1990s (Garcia et al. 2000; Fujita 2003).
Various reforms in agricultural marketing since the late 1980s have also con-
tributed to the increase in paddy production. Among the reforms, the most impor-
tant one was the change in the paddy procurement policy in the late 1980s – the
government reduced the amount of paddy procurement and allowed farmers to sell
the surplus freely in private markets. Since paddy prices in the market during the
late 1980s and early 1990s were usually much higher than the government-fixed
procurement price, the reform initially gave a substantial incentive to produce a
surplus. Because of this impact, Okamoto (2005) calls these reforms ‘the first
phase of marketing liberalization.’ The second phase of marketing liberalization
began in fiscal year 2003/04, in which the abandonment of paddy procurement and
the opening up of rice export to private traders were announced (Okamoto 2005).
However, at the time of writing, the private rice export remains very restricted and
the government’s (partial) enforcement of crop planning continues under the new
regime without the paddy procurement. Because of the uncertainty in the current
policies, this paper analyzes the situation before the second phase only. Note that
the primary dataset for this paper was collected in 2001.
In 2001, there were three elements comprising the paddy policy in rural Myan-
mar. First, the farmland belonged to the state and farmers were given the tillage
right only.3 Farmers did not have the official right to exchange, transfer, lease,
inherit, or mortgage their land, although children were usually given the right
to cultivate their parents’ land and unofficial transfers of tillage rights were not
uncommon (Takahashi 2000). The state had a discretionary power to revoke the
tillage right of a farmer.4
Second, the procurement quota for paddy was set as a fixed quantity per acre
of land designated as a paddy field. Each farmer was obliged to deliver the des-
ignated quantity of paddy to procurement centers run by a state-owned marketing
enterprise. In the main paddy-growing areas, the quota was approximately 20%
of gross produce, while it was lower in other areas. Since the quota was set irre-
spective of the actual acreage devoted to paddy or the actual output of paddy, this
may seem to be a non-distortionary implicit tax. In reality, however, the system
adversely affected paddy production in Myanmar because of the incentive effects
it created. The first of these is a disincentive effect on the quality of rice that was
supplied to the state, which was so low that it was not accepted in foreign markets.
The second effect regards the incentives that influenced farmers’ crop choice (see
the next section).
Third, the government made plans on cropping patterns that were only in-
dicative according to the official policy.5 In reality, however, farmers faced the
threat of seeing their tillage rights revoked if they deviated too much from the
government crop plans, especially with respect to paddy. To retain their tillage
right for paddy fields, farmers thus felt pressure from the local administration to
grow paddy crops on their paddy fields according to the government crop plans.6
According to the official statistics, the national rice output increased by a half
184 T. Kurosaki

during the period from 1990/91 to 2000/01, of which more than two-thirds can
be attributed to the area expansion.7 The macro review of Myanmar’s economy
and agricultural policies thus suggests a possibility that cropping patterns forced
by the local authorities were responsible for the agricultural output growth, at the
cost of farmers’ income. To investigate this possibility, we turn to micro data.

Characteristics of sample villages and households

As a result of the country’s isolationist foreign policy, the availability of micro
data on Myanmar’s rural economy is limited. We therefore conducted a survey
of sample households belonging to eight selected villages8 in June–October 2001
(Figure 1).9
The characteristics of the villages are shown in Table 2. The first two villages
(DELTA1 and DELTA 2) were located in the delta regions of lower Myanmar and
DRY1 was located in the Mandalay Basin, which is one of Myanmar’s centers of
commercial crop production due to its long history of canal irrigation, dating back
to Burma’s dynastic period. In contrast, DRY2 and DRY3 were villages relying
on rainfed agriculture. DRY2 was more typical of a dry zone village since only
rainfed crops and no paddy crops were grown here. HILL1 and HILL2 were
villages relying on vegetable-based development in hilly regions. Both villages
sold their vegetables to major consumption centers such as Yangon and Mandalay,
while their paddy cultivation was oriented toward subsistence. The last village of
the study, COAST, was located in the coastal region of southern Myanmar, where
tropical agro-forestry (rubber, fruits, cashew nuts, etc) prevailed. Peasant farmers
ran both small-scale rubber estates and paddy farms. Among the eight villages
studied, COAST had the most active non-farm sector, which included general
shops, cycle taxis, and fish processing. The eight villages chosen were thus quite
representative of the diverse agro-ecosystems found in Myanmar.
The specific villages were carefully chosen to ensure that they were repre-
sentative of each region in terms of cropping patterns and land distribution. The
sample households were drawn from a complete list of households in each of
the villages studied, making a representative sample in terms of the distribution
of farmland and primary jobs in each village. A total of 521 households were
surveyed in the eight villages and their distribution is shown in Table 3: the 341
households denoted as ‘Farm’ were households with land tillage rights, while the
180 households denoted as ‘Non-farm’ had no tillage rights. Since the sampling
probability differed from village to village and between ‘Farm’ and ‘Non-farm’
households, we correct for the difference by weighted regressions below.
A structured questionnaire was used for all households to establish household
characteristics, household assets, income, consumption, and debt and credit. The
sample households included 2,850 persons, implying that the average household
size was 5.5 persons. This part of the dataset provides the individual characteris-
tics that are used in the analysis of factors that affect crop choice. If households
operated farmland, they were asked to fill in an additional questionnaire on farm
 Journal of the Asia Pacific Economy 185

Figure 1. Map of survey villages.

186 T. Kurosaki

Table 3. Sample households.

Total number of households Number of sample households

Farm Non-Farm Total Farm Non-Farm Total

DELTA1 232 283 515 67 33 100

DELTA2 213 243 456 60 40 100
DRY1 118 101 219 65 37 102
DRY2 326 336 662 24 16 40
DRY3 334 176 510 24 16 40
HILL1 544 298 842 26 12 38
HILL2 422 75 497 34 6 40
COAST 647 520 1167 41 20 61
Total 2836 2032 4868 341 180 521

management. This part of the dataset provides household-level information on

agricultural input, output, and disposal/marketing. Household heads or other rel-
evant persons were interviewed by local research assistants and the information
was cross-checked on the spot by the author to ensure internal consistency and
data quality.

Household income, cropping patterns, and farm productivity

Table 4 reports the asset and income of the sample households. The average
land holding size among the sample households was 5.6 acres, which is large by
South-East Asian standards. Ownership of modern assets, in contrast, compares
poorly: no households owned four-wheel tractors; and motorcycles and four-wheel
vehicles for transportation were very rare. Livestock were the most common and
important assets.
The average incomes were 184,000 Kyats per household and 36,000 Kyats
per person per year. If these figures are converted at the market exchange rate of
650 Kyats/US$ prevailing during the study period, they are equivalent to $283
per household and $55 per person per year. Incomes in the sample villages thus
were indeed low, but not that different from the average village in rural Myanmar.
If these incomes are converted using the price of rice in the Yangon market (56
Kyats/kg) prevailing during the study period, they are equivalent to 3300 kg of
rice per household and 640 kg per person per year.
The composition of income differed across villages (Table 4). The share of
agricultural self-employment income in total household income was highest in
villages HILL2 and DRY1 and lowest in COAST. The table shows that villages
with higher agricultural self-employment incomes and higher non-agricultural
incomes had higher per capita incomes overall. As one of the key factors in
explaining the disparity in household income, this paper further investigates the
disparity in agricultural self-employment income.10
Table 4. Average asset and income of sample households.

Household incomea (Kyats) Composition of income sourcess (%)

Total current value of
Farmland production assetsb Average of total Average of per-capita Self-employment Agricultural Non-agricultural
(acres) (1000 Kyats) household income household income income from agric. wage income income

DELTA1 5.97 218.2 134,535 30,065 61.5 14.8 23.6

DELTA2 7.17 207.8 155,423 29,745 57.3 24.7 18.0
DRY1 3.32 232.7 209,661 49,378 61.3 11.9 26.8
DRY2 6.13 282.0 216,482 43,975 69.0 10.4 20.6
DRY3 6.06 188.5 87,591 17,084 60.5 29.3 10.2
HILL1 7.06 225.7 194,807 36,447 53.9 22.7 23.4
HILL2 3.92 172.9 169,477 32,147 70.2 11.7 18.1
COAST 5.81 579.0 314,478 44,547 33.8 9.7 56.4
Total 5.62 261.0 184,086 36,177 55.9 15.3 28.8

Notes: a Household income is defined as the sum of wage/salary receipts including the imputed value of in-kind payment such as meals and rice, non-agricultural
Journal of the Asia Pacific Economy

self-employment earnings (gross revenue minus actually paid costs), agricultural self-employment earnings (sum of the value of output minus actually paid costs),
and net receipts of non-earned income. Median market prices within each village were used to impute the value of non-cash transactions such as the paddy produced
by farmers and consumed by themselves and in-kind payment to workers.
b The sum of the values of livestock, agricultural equipment and machinery, and transportation equipment. The year-end stock value in the reference year is reported
in this table.
187
 188

Table 5. Cropping patterns of sample households.

Acreage share of major crop groups in the

gross cultivated area (%)
Average gross
Number Average farm cultivated Paddy, Summer Oilseed Industrial Other
of house-holdsa size (acres) area (acres) total paddy Pulses crops Vegetables cropsb crops

DELTA1 67 8.97 15.08 99.5 42.3 0.1 0.0 0.1 0.0 0.2
DELTA2 60 12.10 17.14 74.0 8.6 25.5 0.4 0.0 0.0 0.1
DRY1 71 5.38 8.75 62.5 22.5 1.8 16.2 17.4 0.8 1.3
T. Kurosaki

DRY2 24 10.45 21.42 0.0 0.0 35.6 46.7 0.2 0.0 17.4
DRY3 26 9.51 12.27 45.6 1.1 15.9 30.9 2.6 0.2 4.7
HILL1 26 10.44 9.18 15.4 11.4 9.7 12.2 6.4 22.3 34.1
HILL2 32 4.53 5.24 32.1 0.0 6.9 9.4 50.6 0.0 1.0
COAST 44 8.21 7.77 51.7 1.0 0.3 0.0 2.4 33.6 12.0

Notes: a Only those households with positive areas under cultivation in the survey year are included in this table.
b Industrial crops include sugarcane, cotton, and natural rubber.
 Journal of the Asia Pacific Economy 189

Since crop income accounted for about 99% of agricultural self-employment

income and farm size is pre-determined in the short run, the focus below is on the
determinants of crop income per acre. The cropping patterns are shown in Table 5.
Of the major crop groups, paddy occupied more than 60% in DELTA1, DELTA2,
and DRY1. Among these villages, DELTA1 had the least diversified cropping
pattern: monsoon paddy followed by summer paddy. In contrast, in DELTA2 and
DRY1, not all of the paddy fields were cropped with summer paddy but some
fields were cropped with pulses (DELTA2) and vegetables (DRY1). The other five
villages had a more diversified agriculture. Among these five villages, DRY3 and
COAST had higher paddy shares than the other three. There were no paddy fields
in DRY2.
Table 6 relates the cropping patterns with per-acre crop income. Crop income
per acre was highest in DRY1 and HILL2, followed by DRY2, and lowest in DRY3,
followed by DELTA2, and DELTA1. A comparison of the first two columns in
Table 6 suggests that farm income per acre was lower in villages where paddy
cropping was more dominant than in other villages because per-acre income was
lowest for paddy and highest for vegetables.
This negative relation between the paddy share and per-acre crop income was
observed within villages as well. Table 6 also shows the intra-village correlation
coefficients between the crop income per acre of a farm and cropping patterns (the
acreage share of each crop group in the gross cropped area). In all villages, the
correlation coefficient between the per-acre crop income and the paddy share was
negative. It was statistically significant in DELTA2, DRY1, DRY3, HILL2, and
COAST. In DELTA2, the correlation coefficient between the per-acre crop income

Table 6. Correlation between crop acreage shares and per-acre crop income.

Intra-village correlation coefficients

between per-acre crop income and
the acreage share of a crop group
Village-average Village-average
of crop of paddy Acreage
income per shares in Paddy shares of (Name
acre of farmland the gross acreage crop of crop
(Kyats/acre) cultivated areas (%) shares group i group i)

DELTA1 11,222 99.5 −0.019 0.162 (Vegetables)

DELTA2 12,958 74.0 −0.443* 0.448* (Pulses)
DRY1 33,305 62.5 −0.385* 0.555* (Vegetables)
DRY2 25,718 0.0 n.a. 0.599* (Other crops)
DRY3 9,582 45.6 −0.529* 0.349* (Oilseed crops)
HILL1 13,200 15.4 −0.094 0.319* (Industrial crops)
HILL2 33,313 32.1 −0.364* 0.308* (Vegetables)
COAST 14,256 51.7 −0.473* 0.810* (Vegetables)

Note: ∗ indicates that the coefficient is statistically significant at the 5% level. See Table 5 for the
number of observations.
190 T. Kurosaki

and the pulses share was 0.448. In DRY1, the correlation coefficient between the
per-acre crop income and the vegetables share was 0.555. Therefore, in DELTA2
and DRY1, villages located in the major paddy growing regions, farmers who
did not grow much paddy on paddy fields during the summer season but grew
more commercial crops instead were better off. This indicates that the policy of
maximizing paddy output placed a heavy burden on farmers in the major paddy-
growing regions. In the other five villages, where agriculture was more diversified,
each village had non-paddy crops whose acreage share was positively correlated
with the per-acre crop income. In these villages, it was not always the case that
these non-paddy crops directly competed with paddy for land, because these crops
were usually grown on farmland not designated as paddy fields. Even then, the
allocation of labor and efforts expended on non-paddy crops should have been
affected adversely when the paddy acreage was increased.
It thus seems that the farmers grew too much paddy in the sense that its
cultivation replaced more lucrative crops, resulting in lower crop income than the
highest level achievable from the same farmland.11 Differences in the enforcement
of the government crop plans could have been responsible for this situation. It was
observed in the field that enforcement along procedures inherited from the socialist
period was attempted in the three villages located in the core regions of paddy-
based agriculture (DELTA1, DELTA2, DRY1), with the strictest enforcement in
DELTA1; similar procedures were being extended to the other two villages in the
dry zone (DRY2, DRY3); and the other three villages (HILL1, HILL2, COAST)
were subject to the weakest enforcement of the government crop plans. Similar
differences could have existed within villages. The next section formalizes this
idea and derives testable predictions.

Analytical framework
A theoretical model of farmers’ crop choice
To explain the situation described above, a simple model of a farmer’s crop choice
is presented first. Since paddy is the backbone of Myanmar’s economy, it is treated
as the numeraire. The farmer has a fixed acreage x0 of farmland that can be
allocated to either a paddy or a non-paddy crop. Let x denote the paddy acreage.
The allocation x results in farm income net of production costs, which is denoted
by a function f (x). This function reflects the farming technology available to
the farmer and the resource constraints he faces. The function is assumed to be
continuous, differentiable, and single-peaked. Thus, in the absence of procurement
and land policies, a profit-maximizing farmer chooses x ∗ that maximizes f (x).
Under the policies prevailing in Myanmar during the survey period, however,
the farmer was required to deliver to the government a fixed quantity of paddy
proportional to z (the planned acreage under paddy). To reflect the paddy output
maximization policy adopted by the state, it is assumed that x ∗ < z.12 The total
procurement quota is bqz, where b is the procurement ratio and q is the average
yield of paddy per acre. As explained in the previous section, the government
 Journal of the Asia Pacific Economy 191

procurement price of paddy was below its market price, resulting in implicit
taxation on paddy farmers. The implicit tax amount is denoted by tqz, where t is
equal to b times the price disparity. The government fixes z, b, and t.
Although the farmer is expected to allocate z acres of his farmland to paddy,
this may not be enforceable, i.e. x may deviate from z. However, when the deviation
is large and detected by the local administration, the administration may threaten
the farmer with revoking the farmland. The expected welfare cost of this threat
is assumed to be a quadratic function, A(z − x)2 , where A is a non-negative
parameter characterizing the political control exercised by the local administration
with respect to the farmer. Given these assumptions, the payoff of the farmer is
expressed as

y = f (x) − tqz − A(z − x)2 (1)

The first-order necessary condition for the optimal solution is given by

f  (x) = −2A(z − x) (2)

Let x ∗∗ be the optimal solution for the farmer that satisfies equation (2). When it
is an interior solution, the following relations hold

x ∗ ≤x ∗∗ < z (3)
∗∗
∂x /∂A > 0 (4)
∂x ∗∗ /∂z > 0 (5)

Equation (4) shows the basic relation for the empirical analysis below. The gov-
ernment’s objective function is increasing not only in tqz (the implicit tax revenue)
but also in x, since the government is concerned with the total supply of paddy
to the nation. Thus, the government attempts to affect x through its choice of z
according to equation (5) when it fixes z, b, and t.13
When A = 0 (the case where the local administration has no ability to force
farmers to follow its directions regarding the planned acreage), x ∗ = x ∗∗ holds. In
this case, the procurement system does not affect farmers’ production decisions at
all. When A >0, however, x ∗ < x ∗∗ holds, so that the procurement quota becomes
a distortionary implicit tax. As A approaches infinity, x ∗∗ also approaches z. The
last case may characterize the situation in Myanmar during the Burmese Socialism
period that ended in the late 1980s.
Thus, the divergence between the actual acreage and the income-maximizing
acreage is an increasing function of parameter A. It is expected that A is higher
for a farmer who has some political reason to please the local administration.
Although it is possible that A may be higher for a farmer who derives greater
moral satisfaction from following the rule or who is more averse to the risk of
penalty, this possibility is not analyzed since such preferences are not directly
192 T. Kurosaki

observable. Instead, it is assumed that preferences do not systematically differ

across individuals within villages. Although parameter A is higher for a farmer
who lives in a village under tighter political control, this aspect is not analyzed
directly but controlled indirectly by village fixed effects in the empirical analysis
(see below). Although parameter A should be treated as endogenous in the long
run, since it is determined as a result of strategic interactions of individual farmers
with the local administration, it is treated as exogenous in the empirical analysis,
because the focus of the analysis is on the short-run determinants of crop choice.

Empirical specification
The share of paddy acreage in the gross cultivated area (see Table 5) is employed
as the main dependent variable for regression analysis, as a proxy for x/x0 . As ex-
planatory variables, household/individual characteristics and village dummies are
included. Although villages differ with respect to their market conditions includ-
ing local demand, transportation, and infrastructure, the strictness of enforcement
of the government crop plans, production conditions such as technologies, soil
quality, and environmental suitability of growing conditions for various kinds of
crops, and other factors, it is impossible to identify the effects of these factors
separately using the dataset adopted in this study. The number of sample villages
is too small for such an analysis. Therefore, village fixed effects are included
to control for these factors collectively and to enable us to concentrate on the
intra-village variations in cropping patterns.
The main prediction of the theoretical model to be tested empirically is the
effect of A, the political control exercised by the local administration on each
farmer. The focus of the model on this parameter as a determinant of the amount
of paddy a farmer grows makes it possible to examine an important aspect of
crop choice that other models fail to address. For example, models emphasizing
credit constraints (Feder et al. 1990), risk management (Kurosaki and Fafchamps
2002), and resource constraints (Gotsch et al. 1975) do not address the response
of farmers’ crop choice to the political control. These models instead predict
that human capital and household resource availability affect farmers’ crop
choice.
However, it is possible that, in a rural setting like Myanmar, farmers with
a higher A are those with superior access to technology, resources, or markets.
Therefore, even if it is found that a proxy variable for A is associated with a higher
paddy share, this association could be the result of these indirect effects and not of
the political mechanism modeled above. Furthermore, there are other endogenous
variables that are chosen simultaneously with x/x0 , which were assumed away in
the theoretical model for simplicity. The empirical specification should take into
consideration that households simultaneously choose several variables (including
x/x0 ) to determine the level of each income source, including agricultural wage
income and non-agricultural income. Therefore, the following four strategies are
adopted to identify the effect of parameter A.
 Journal of the Asia Pacific Economy 193

First, a completely reduced-form approach is adopted. Although it is of interest

to quantify the effects on crop choice of the level of non-agricultural income or
agricultural wage income, valid instrumental variables were not found in the
dataset to identify these effects.
Second, to control for other factors that affect crop choice, human capital
and household resource availability are included as explanatory variables. Hu-
man capital and household resource availability, such as education, household
demography, land size, and other assets, may be important determinants of crop
choice because they influence farmers’ access to markets, technology, and credit.
Although these variables may include information on the political vulnerability of
a farmer, this aspect is ignored to set a higher hurdle for the theoretical model to
pass the empirical test.
Third, an attempt is made to search for variables to represent A that are not
directly related to households’ access to markets and technologies. Based on the
discussion with sample farmers during the field work, a variable representing
expectations regarding future land inheritance from parents is chosen as the main
proxy for A. As explained in the second section, the land tillage right was not
officially inheritable, but children were usually given the right to cultivate their
parents’ land if the local administrator found no reason to deny such a transaction.
This implies that a farmer who had parents with farmland and expected them to
transfer the tillage right to him/her in the near future may have had an additional
incentive to follow the directions of the local administration at the time of our
survey so that the administration would not object to the transfer of the cultivation
right. As the second proxy for A, a variable representing the early adoption of new
agricultural technology during the pre-transition period is also chosen. During
the pre-transition period, the local administration designated several villagers as
‘contact farmers’ and implemented policies through them. The early adoption of
new technology during the planned economy period could be an indicator for
the status of a contact farmer because such adoption of new technology was not
chosen by the farmer but ordered by the administration.
However, it is still possible to argue that these variables capture not only the
political control but also reflect unobservable characteristics of the farmer that
enhance the efficiency of paddy production. A farmers’ expectation to inherit the
parents’ paddy field14 or having been directed in the past to try out a new technology
may be linked to her/his unobservable ability to produce paddy more efficiently.
If this is the case, these variables should not only be associated with a larger
paddy acreage share but also higher paddy productivity, even after controlling
for other observable characteristics that affect productivity. The fourth strategy
of this paper is, therefore, to regress variables representing paddy productivity on
the same set of variables adopted for the paddy acreage share. If it is found that
the variables proxying for A are associated with a higher paddy acreage share but
with no increase in paddy productivity, the argument in favor of the unobservable
productivity superiority is not supported. As the second dependent variable, paddy
value-added per acre is adopted.
194 T. Kurosaki

Estimation results
Determinants of farmers’ crop choice
Table 7 reports the definition and summary statistics of the empirical variables.15
To avoid the reverse causality from crop choices to the asset formation, the asset
variables corresponding to the values preceding the reference period of the survey
are employed.
The estimation results are shown in Table 8. They show that the dummy variable
for the expectation of inheriting land (Future inh) and the dummy variable for the
early adoption of new technology in the past (Past techn) are both positive and
statistically significant predictors of the paddy acreage share (Paddy gca).
Table 8 also shows that larger lowland fields are associated with a higher paddy
acreage share while larger upland fields are associated with a lower share. Most of
the demographic and asset variables are not statistically significant, except for the
thresher ownership. Farmers owning a paddy thresher tended to grow more paddy,
which suggests that the rental market for thresher service in the study region did
not work well.
To investigate whether the positive correlation between the paddy acreage
share and the two variables proxying for A is due to unobservable superiority
in paddy production, paddy value-added per acre (VA paddy) is regressed on the
same variables. The results in Table 8 show that both Future inh and Past techn
have coefficients whose absolute values are small and whose standard errors are
large so that the null hypothesis that the coefficient is zero is accepted. Thus, these
variables do not enhance paddy productivity.
Table 8 also shows that the education level of the household head does not affect
paddy productivity. This seems to be against the view that education enhances farm
productivity, a fact that has been demonstrated empirically for other developing
countries (see, for example, Jamison and Lau 1982; Kurosaki and Khan 2006). It
will be shown later that the positive effect of education on agricultural productivity
does exist in Myanmar as well, at the farm level rather than at the crop level.
To examine the robustness of the results reported in Table 8, different specifica-
tions are also tried. First, as the dependent variable for the crop choice regression,
paddy acreage divided by the acreage of the net cultivated area (Paddy nca) or
by the planned acreage under paddy (Paddy z) is adopted, and as the dependent
variable for the paddy productivity, paddy output per acre (VQ paddy) is adopted.
Second, a different set of explanatory variables is employed, replacing the vari-
ous wealth variables into a single variable of the total wealth value and replacing
the household head education variable by two variables that represent the highest
and the average education among the adult household members. Third, for the
determinants of the paddy acreage share, tobit models are also attempted since the
dependent variable is truncated: Paddy gca ranges from zero to one, Paddy nca
ranges from zero to two, and Paddy z takes non-negative values. The main result
remained qualitatively unchanged that Future inh and Past techn are positive pre-
dictors of the paddy acreage share but they are insignificant in predicting paddy
productivity (Table 9). The result is thus highly robust.
 Journal of the Asia Pacific Economy 195

Table 7. Variables used in the regression analysesa .

Variable Definition Meanb Std. Dev. Min. Max.

1. Dependent variables
Crop choice model
Paddy gca The share of paddy acreage in the gross cultivated 0.648 0.341 0 1
area (i.e. the total of acreage under various crops
cultivated in the farm).
Paddy productivity model
VA paddy Value-added of paddy divided by the acreage under 0.135 0.140 −0.434 0.705
paddy crops (100,000 Kyats/acre).
Household income model
Y crop Household income from self-employment in 1.375 1.328 −0.640 10.757
agriculture (crop income only) (100,000 Kyats).
Y hh Total household income (100,000 Kyats). 1.941 2.745 −1.275 42.481
2. Explanatory variables
Lowland Acreage of lowland fields managed by the household. 6.262 6.029 0 44
Upland Acreage of other types of fields (mainly upland fields) 2.124 4.230 0 34.11
managed by the household.
Future inh Dummy variable for the expectation of the household 0.163
to inherit cultivation rights from parents’
households.
Past techn Dummy variable for the past adoption of new 0.276
technologyc before the median year of adoption in
each village (the reference is ‘adopted later than or
equal to the median’ or ‘the household began
farming after the median’).
Educ head Education status of the household head in terms of 3.095 3.074 0 16
completed years of schooling (years).
Age head Age of the household head (years). 47.414 13.067 21 93
Hh size The number of household members living together. 5.521 1.979 2 12
Hh deprate The dependency ratio: the number of household 0.487 0.231 0 0.875
members who are out of labor force divided by
‘Hh size’.
Extended Dummy variable for a household that has extended 0.291
family members.
Asset house The value of household durable consumption goods 0.223 0.488 0 4.741
such as radio-cassette, TV, sewing machine, etc.
d
(100,000 Kyats).
Asset transp The values of transportation equipment such as 1.511 13.861 0 248.000
bicycle, bullock carts, motorcycle, etc. (100,000
Kyats).d
Asset lvstk The value of livestock such as cattle, buffaloes, goats, 1.453 1.479 0 8.800
chickens, etc. (100,000 Kyats).d
Powertiller The number of owned power tillers for farming.d 0.110 0.351 0 3
Pump The number of owned diesel pumps for irrigation.d 0.212 0.438 0 2
Thresher The number of owned threshing machines for paddy.d 0.021 0.145 0 1
Village fixed Dummy variable for each village.e
effects

Notes: a The unit of observation is a farm household with positive areas under cultivation in the survey
year and living in villages other than DRY2. The number of observations (NOB) is 326, except for
VA paddy that is not defined for 35 households that did not grow paddy crop at all (NOB is 291).
b When the variable is a dummy, the percentage of those observations taking one is reported.
c The key technology is: DELTA1 = summer paddy production using diesel pumps, DELTA2 = summer
paddy production using perennial canals, DRY1 = summer paddy production using Manawthakha
variety, DRY3 = IR variety for monsoon paddy production, HILL1 = Shwewar-tun (Ywe) variety for
monsoon paddy production, HILL2 = Kauk Phwar Phy variety for monsoon paddy production, COAST
= summer paddy production using artificial irrigation.
d The year-end stock in the year preceding the reference period of the survey.
e The distribution is: DELTA1=0.206 (reference), DELTA2 = 0.184, DRY1 = 0.218, DRY3 = 0.080,
HILL1 = 0.080, HILL2 = 0.098, COAST = 0.135.
196 T. Kurosaki

Table 8. Determinants of crop choices and paddy productivity.

Paddy gca VA paddy

Dep. variable= Coeff. Std. Err Coeff. Std. Err

Individual and household attributes

Lowland 0.0134∗∗∗ (0.0046) −0.0041∗∗∗ (0.0015)
Upland −0.0129∗∗ (0.0065) −0.0023 (0.0046)
Future inh 0.0996∗∗ (0.0446) 0.0180 (0.0173)
Past techn 0.1386∗∗∗ (0.0521) −0.0026 (0.0271)
Educ head 0.0013 (0.0079) −0.0013 (0.0033)
Age head 0.0032∗ (0.0017) 0.0008 (0.0008)
Hh size −0.0140 (0.0109) 0.0023 (0.0051)
Hh deprate 0.0386 (0.0861) 0.0352 (0.0372)
Extended 0.0247 (0.0362) −0.0454∗∗ (0.0206)
Asset house 0.0219 (0.0391) 0.0629∗ (0.0369)
Asset transp −0.0011 (0.0007) −0.0023 (0.0064)
Asset lvstk 0.0036 (0.0153) 0.0035 (0.0066)
Powertiller −0.0395 (0.0706) 0.0116 (0.0461)
Pump −0.0375 (0.0488) 0.0206 (0.0435)
Thresher 0.2917∗∗∗ (0.0903) −0.0453 (0.0327)
Village fixed effects
DELTA2 −0.2577∗∗∗ (0.0599) 0.0518∗ (0.0309)
∗∗∗
DRY1 −0.3101 (0.0536) 0.0932∗∗∗ (0.0315)
DRY3 −0.3852∗∗∗ (0.0771) −0.0615 (0.0413)
HILL1 −0.5885∗∗∗ (0.0760) 0.1390∗∗ (0.0674)
HILL2 −0.5077∗∗∗ (0.0817) −0.2519∗∗∗ (0.0433)
COAST −0.2942∗∗∗ (0.0858) 0.0891∗∗ (0.0448)
Intercept 0.7034∗∗∗ (0.1168) 0.0712∗ (0.0423)
F-stat for zero slopes 69.96∗∗∗ 13.60∗∗∗
R2 0.563 0.632
Number of observations 326 291

Notes: Estimated by weighted least squares with Huber-White heteroscedastic robust standard errors
in parentheses. The weights reflect the difference in sampling probability shown in Table 3. Coefficient
estimates are statistically significant at 1% (***), 5% (**), and 10% (*).

It can be concluded therefore that farmers who expected to inherit land or

possessed past experience of adopting new technology grew more paddy crops,
although their paddy productivity was not higher than that of other farmers. If
crop choice were solely determined by the relative profitability of competing
crops, this would be a puzzling result. But if we borrow the insights from the
theoretical model in the previous section, the paradox can be resolved: farmers
who were more susceptible to pressure from the local administration to follow the
government crop plans grew more paddy crops even though paddy crops were less
profitable than other crops.
An alternative explanation could be farmer’s risk aversion. If Future inh and
Past techn capture a difference in risk preferences of farmers, independently of
the paddy productivity, we expect the results reported here. But if the risk aversion
 Journal of the Asia Pacific Economy 197

Table 9. Robustness of the estimation results.

Coefficient estimates on proxy variables forA

Future inh Past techn

Coeff. Std. Err Coeff. Std. Err

1. Different dependent variables (estimated by weighted least squares [WLS])

Determinants of crop choices
Paddy gca (Table 8) 0.0996∗∗ (0.0446) 0.1386∗∗∗ (0.0521)
Paddy nca 1 0.1212∗∗∗ (0.0448) 0.1994∗∗∗ (0.0517)
Paddy z 2 0.0969∗∗ (0.0458) 0.0971∗∗∗ (0.0324)
Determinants of paddy productivity
VA paddy (Table 8) 0.0180 (0.0173) −0.0026 (0.0271)
VQ paddy 3 0.0162 (0.0141) −0.0028 (0.0288)
2. Different definitions for asset and education variables (WLS)
Determinants of crop choices
Paddy gca 0.0926∗∗ (0.0430) 0.1275∗∗ (0.0549)
Paddy nca 0.1180∗∗∗ (0.0439) 0.1950∗∗∗ (0.0522)
Paddy z 0.0944∗∗ (0.0439) 0.0883∗∗ (0.0352)
Determinants of paddy productivity
VA paddy 0.0184 (0.0165) −0.0116 (0.0325)
VQ paddy 0.0135 (0.0130) −0.0111 (0.0319)
3. Estimated by weighted tobit
Determinants of crop choices
Paddy gca 0.1438∗∗ (0.0584) 0.2241∗∗∗ (0.0674)
Paddy nca 0.1552∗∗∗ (0.0557) 0.2420∗∗∗ (0.0610)
Paddy z 0.1050∗∗∗ (0.0396) 0.0989∗∗ (0.0426)

Notes: Each row corresponds to a regression model for the dependent variable shown in the first column.
Only the coefficient estimates on Future inh and Past techn are reported in this table to save space. All
the regressions include a full set of explanatory variables similar to those reported in Table 8. Complete
estimation results are available on request.
1. Paddy nca = The share of paddy acreage in the net cultivated area (i.e., the total of acreage of farm
plots that were cultivated during the reference year); its mean is 0.954 (its standard deviation is 0.608).
2. Paddy z = Acreage under paddy crops divided by the paddy land acreage subject to procurement;
0.703 (0.280).
3. VQ paddy = Output value of paddy divided by the acreage under paddy crops (100,000 Kyats/acre);
0.233 (0.123).

factor is important in crop choice, wealth variables should also affect crop choice,
which is not the case. Furthermore, as discussed in the theoretical section, the
parameter of political susceptibility A is likely to be higher for a farmer who is
more averse to the risk of penalty. Thus, the alternative explanation based on risk
aversion can be partly accommodated in the theoretical model of this paper.

The welfare impact of paddy policies

The regression results above suggest that the political vulnerability of a farmer
affected the paddy acreage share and the correlation analysis (Table 6) shows
198 T. Kurosaki

that the paddy acreage share was negatively correlated with per-acre crop income.
These findings raise the question: how large was the negative welfare impact of the
paddy output maximization policies, functioning through parameter A? It should
be noted that the income loss functioning through parameter A is likely to be a
net loss to the economy since the farmer’s production deviates from the efficient
point due to the policy distortions.
To simulate the impact, household income is regressed on the paddy acreage
share (Paddy gca) and other variables representing human capital and household
resource availability. The dummy variables representing the expectation of in-
heriting land (Future inh) and the early adoption of new technology in the past
(Past techn) are deleted from the regression since they do not affect paddy pro-
ductivity as shown in Table 8 and they do not affect other sources of household
income, either.16 Since the endogeneity test did not reject the statistical exogene-
ity of Paddy gca in determining both crop and household income even at the
20% level, Table 10 reports the OLS estimates, which are more efficient than the
instrumental variable estimates when the exogeneity condition is satisfied.17

Table 10. Impacts of paddy policies on household income.

Y crop Y hh

Dep. variable= Coeff. Std. Err Coeff. Std. Err

Individual and household attributes

Paddy gca −0.4978∗∗∗ (0.1853) −1.1556∗∗∗ (0.3725)
Lowland 0.0380∗∗ (0.0181) −0.0289 (0.0210)
Upland 0.0755∗∗∗ (0.0192) 0.0536∗ (0.0298)
Educ head 0.0463∗∗ (0.0263) 0.0560 (0.0367)
Age head 0.0089 (0.0055) 0.0109 (0.0081)
Hh size 0.0057 (0.0349) 0.1368∗∗ (0.0542)
Hh deprate −0.0611 (0.2753) −0.5128 (0.3883)
Extended −0.0985 (0.1539) −0.0081 (0.2115)
Asset house 0.1960 (0.2103) 0.5567∗ (0.3136)
Asset transp 0.0071 (0.0049) 0.1383∗∗∗ (0.0063)
Asset lvstk 0.2045∗∗∗ (0.0629) 0.1126 (0.0750)
Powertiller −0.2715 (0.2594) 0.1225 (0.4250)
Pump 0.3764 (0.2804) 0.1460 (0.3590)
Thresher 0.1652 (0.2351) 2.1466∗ (1.2108)
Village fixed effects
DELTA2 0.1071 (0.2711) −0.0060 (0.3288)
DRY1 0.6386∗∗ (0.2928) 0.2503 (0.3278)
DRY3 −0.7468∗∗ (0.3388) −1.5954∗∗∗ (0.4310)
HILL1 −0.2315 (0.4108) −1.0335∗ (0.5261)
HILL2 0.1385 (0.3251) −0.8608∗∗ (0.4184)
COAST −0.2463 (0.2879) −0.1283 (0.3729)
Intercept 0.2842 (0.3669) 1.2227∗∗ (0.4839)
F -stat for zero slopes 126.63∗∗∗ 1578.15∗∗∗
R2 0.421 0.874

Notes: See notes to Table 8 for the estimation method. The number of observations is 326.
 Journal of the Asia Pacific Economy 199

The estimation results in the table show that crop income increased with the
size of farmland, the level of education, and the livestock asset value. The positive
effect of education on crop income is either due to higher per-acre yields of non-
paddy crops or due to the superiority of educated farmers in allocating land across
crops, since education does not affect per-acre yields of paddy crops (Table 8).
A model similar to the one reported in Table 8 was estimated with non-paddy
output per acre as the dependent variable, yielding a very small and statistically
insignificant coefficient on education. Thus, the first route is unlikely. Instead,
these results support the view that education enhances farm productivity mainly
through improvement in the allocative efficiency in Myanmar, which is similar
to the finding for Pakistan reported by Kurosaki and Khan (2006).
The parameter of concern is the effect of the paddy acreage share (Paddy gca)
on income. The effect is significantly negative both on crop income and household
income. An interesting finding is that the ill effect of Paddy gca is much stronger
on total household income than on crop income. This indicates that increasing the
paddy acreage share reduces household income not only through the reduction of
crop income due to a greater deviation from the crop income maximization point,
but also through the reduction of income from other sources due to a decreased
allocation of household efforts and resources to these sources.
Based on the coefficients in Table 10, the loss of household income through the
political control parameter A could be calculated as follows. When a farmer feels
less pressure to satisfy the government crop plans due to a change in the household
status with regard to the expectation of inheriting land (Future inh is changed from
one to zero), the paddy acreage share in gross cultivated areas decreases by 10.0
percentage points (see Table 8), resulting in an increase in crop income of 4,200
Kyats and an increase in household income of 9,800 Kyats. These values are 3.1%
of the average crop income per household and 5.0% of the average household
income respectively. When the local administration stops pressuring a farmer with
past technology experiments to follow its crop plans (Past techn is changed from
one to zero), the paddy acreage share decreases by 13.9 percentage points (see
Table 8), resulting in an increase in crop income of 4,900 Kyats (3.5% of the
average crop income) and an increase in household income of 11,300 Kyats (5.8%
of the average household income). Thus, the welfare loss from forcing farmers to
grow too much paddy is not negligible.18
The loss simulated above can be compared with the amount of implicit taxation
realized through the paddy procurement system. How much does a change in the
procurement quota imposed affect household incomes? Its direct effect can be
calculated by multiplying the difference between the market and procurement
prices by the procurement quantity.19 When the procurement quota is set at zero,
paddy income increases by 8,600 Kyats (8.3% of the average crop income or
4.7% of the average household income). The amount of implicit taxation is largest
in DELTA1 where the paddy procurement puts the heaviest burden on farmers:
14,000 Kyats (16.9% of the average crop income or 10.4% of the average household
income in the village). Although the amount is large, this income is not lost but
200 T. Kurosaki

appropriated by the state as implicit taxation. On the other hand, the income loss
simulated by changing A quantifies a net loss and its amount relative to the average
household income is comparable to the amount of implicit taxation.

Conclusion
Based on a sample household survey conducted in 2001 and covering diverse agro-
ecological environments, this paper investigated the impact of agricultural policies
on farm production decisions and rural incomes in Myanmar as a case study of
low-income transitional economies in Asia. Since the enforcement of agricultural
policies varied not only across villages but also within villages, a theoretical
model focusing on the political control exercised by the local administration on
each farmer was developed to derive implications for the empirical analysis of
intra-village variations in cropping patterns. The theoretical model predicted that
the acreage share of non-lucrative paddy should be higher for farmers who are
under tighter control of the local authorities. It was found that the paddy acreage
share was indeed higher for such farmers, approximated by their expectation
of future land inheritance from their parents and by their previous experience
in adopting new technology during the period of socialist planning. Since the
inheritance expectation and the technology adoption variables were not found to
increase paddy productivity, the argument that these variables reflect unobservable
productivity superiority in paddy production was not supported. Therefore, the
estimation results were consistent with the prediction of the theoretical model.
The simulation results based on the regression estimates showed that the loss
in rural incomes due to farmers being forced to grow too much paddy was not
negligible. The government policies that put too much production emphasis on
paddy crops with an income per acre that was lower than that of other crops thus led
to sub-optimal use of agricultural resources in Myanmar. This finding, therefore,
suggests a possibility that agricultural output gains in Myanmar experienced since
the late 1980s (Rozelle and Swinnen 2004, Tables 1 and 2) were not accompanied
by gains in agricultural productivity and rural incomes. Since this paper uses cross-
section data, it cannot directly analyze the change in agricultural productivity or
rural income. Investigating this possibility is thus left for further research.
Finally, this paper found that, at the time the survey was conducted, there
remained vast room for an expansion of agricultural output and rural incomes in
Myanmar, even without innovation in technology or further investment in irriga-
tion. All that was needed to tap this potential was to give farmers more freedom in
land use and liberalize paddy/rice marketing. Regarding the clear and transparent
land use policies, there are several examples from other transitional economies,
from which Myanmar can learn.20 Regarding the more liberalized agricultural
marketing policies, the second phase of liberalization that began in 2003/04 seems
to be a step in the right direction. A follow-up survey under this new regime is a
task that remains for further investigations.
 Journal of the Asia Pacific Economy 201

Acknowledgment
The author is grateful to anonymous referees, Ikuko Okamoto, Takeshi Sakurai, Yujiro
Hayami, Shigeru Ishikawa, Konosuke Odaka, Kei Otsuka, Tetsushi Sonobe and Naohito
Abe for their useful comments on earlier versions of this paper.

Notes
1. See Myat Thein (2004) and Thawnghmung (2004) for more background information
on recent economic developments in Myanmar.
2. Table 1 is based on the official data, which are regarded by several authors as overes-
timates of the actual agricultural production (see, for example, Thawnghmung 2004,
pp. 131–141).
3. This is a system inherited from previous periods. Before General Ne Win’s military
government overthrew the elected government in 1962, the 1948 Constitution stipu-
lated that: (i) the State is the ultimate owner of all lands; (ii) the State will have the
right to regulate, alter or abolish land tenures or resume possession of any land and
distribute the same for collective or cooperative farming or to agricultural tenants;
(iii) there can be no large land holdings on any basis whatsoever. Under General Ne
Win’s ‘Burmese Socialism,’ a tenancy law was enacted, by which the rights of tenancy
was vested solely in the State in the form of Agrarian Committees, and another law
was enacted to protect the cultivators’ rights so that no one could confiscate or seize
any of the farmer’s means of production such as land, livestock, farm implements, and
agricultural produce as payments for debts. The 1974 Constitution then confirmed the
fact that all lands belong to the state. Although the present regime suspended both
constitutions of 1948 and 1974, it continued to follow the land policies adopted in both
constitutions. Therefore, under the land policies during our survey period and at the
time of this writing, paddy land cannot be sold or rented or mortgaged or transferred
or exchanged, although land use rights for plantations, orchards, and non-paddy can
be sold or leased or transferred. The author is grateful to the anonymous referee for
this historical information.
4. In addition to those farmers with the tillage right, a large pool of landless, non-farm
households exists in rural Myanmar, who do not have the tillage right. The share
of landless, non-farm households in villages typically ranges from 20 to 50%. The
majority of landless households depend on income earned as agricultural wages and
their wealth level tends to be lower than that of landed households.
5. The government crop plans were strictly enforced on farmers during the Burmese
Socialism period by force. The system was changed into indicative planning in 1987.
6. At the time of this writing, the government crop plans still prevail in rural Myanmar.
7. Data are three year moving averages of rice statistics reported in CSO (various issues).
We use the decomposition: Outputt+1 = Outputt + (Areat+1 – Areat ) Yieldt + Areat
(Yieldt+1 – Yieldt )+ (Areat+1 – Areat )(Yieldt+1 – Yieldt ), where the second term
on the right-hand side shows area effect, the third shows yield effect, and the fourth
shows interaction effect.
8. The smallest administrative unit in Myanmar is the ‘village tract,’ which usually
consists of several hamlets or natural villages. While Table 2 refers to ‘village tracts,’
in the text and the following tables, they are simply referred to as ‘villages’ for brevity.
9. The dataset used in this paper is available on request for verification.
10. Preliminary analysis of the determinants of non-agricultural income using a probit
regression model reveals that household members were more likely to work in non-
agricultural jobs when they were more educated.
11. The negative relation within villages between the paddy crop share and the farm
income was observed even after controlling for other factors. See Table 10 for the
202 T. Kurosaki

multivariate analysis results. Whether the negative relation across villages existed even
after controlling for other village-level factors cannot be assessed using the current
dataset since the small number of sample villages precludes multivariate analysis at
the village level. This is the reason why village fixed effects are included in regression
models below.
12. This relation holds when the model is extended into an agency model with the local
administration as the principal, who determines the level of z (see Note 13). Fur-
thermore, among the 301 sample farm households for which the information on z is
available, only four cultivated paddy crops in excess of z. Villagers we interviewed
during the field work expressed their opinion that the government’s planned acreage
for paddy is much larger than the acreage they want to grow in the absence of the
government’s pressure.
13. This can be shown formally by an agency model with the local administration as the
principal. An example of such a model is available on request.
14. Since expectations to inherit land from parents may also depend on a number of
other factors (e.g. younger household heads are more likely to expect to inherit land
from their parents), a set of demographic variables are included and retained in the
regression even when they are not significant. The prerequisite to expect to inherit land
from parents is that there should be parents with land, implying a potential selection
problem if the proxy variable applies only to a subset of the population. However, all
the sample farmers had parents with the cultivation right, suggesting that the selection
bias is not likely to be serious.
15. All empirical variables used in the analysis are available on request for verification.
16. As a reduced-form approach, non-paddy agricultural self-employment income, agri-
cultural wage income, and non-agricultural income were each regressed on the ex-
planatory variables listed in Table 8. The two variables Future inh and Past techn
were not statistically significant even at the 20% level in all of the three regressions.
17. In the specification tests, the variables Future inh and Past techn are used as the iden-
tifying instrumental variables for Paddy gca. Since the two variables are significant
predictors of Paddy gca as shown in Table 8, the acceptance of exogeneity suggests
that the error terms in the paddy acreage equation and in the household income
equations are not significantly correlated.
18. It should be noted, however, that these figures ignore market equilibrium effects. A
decrease in paddy acreage should increase the market price of paddy while an increase
in the production of non-paddy crops may decrease their market prices. Incorporating
equilibrium effects is left for further analysis.
19. There are also indirect effects of the paddy procurement quota on rural incomes. A
decrease in the quota should lead to an increase in crop income through the adjustment
of the optimal paddy acreage described in the theoretical section. Unfortunately, this
effect cannot be assessed using the current dataset because the procurement ratio
within a village had very little variation so that the variable cannot be included as an
explanatory variable in the paddy acreage share regressions. When the paddy acreage
responds to the decrease in the quota, the market equilibrium effects discussed in the
previous endnote may occur as well.
20. For instance, Asian Development Bank (1996) documents that ‘Clear land-use and
ownership rights, protected by a legal system, are essential for production and invest-
ment incentives and for the use of land as loan collateral. In the Kyrgyz Republic, land
leases with a term of 49 years are available with the right to sell, exchange, transfer
by inheritance and use as a security deposit. Vietnam, through the Land Law of 1993,
extended land tenure rights to 20 to 50 years, depending on land use. Kazakhstan has
gone the farthest with 99-year leases and introduction of private ownership for small
garden plots’ (p.31). The author thanks the referee for this reference.
 Journal of the Asia Pacific Economy 203

References
Asian Development Bank, 1996. Asian Development Bank Annual Report 1995. Manila:
Asian Development Bank.
Central Statistical Office [CSO], various issues. Statistical Yearbook. Yangon: CSO, Gov-
ernment of Myanmar.
Dolly Kyaw, 2006. Rural poverty analysis in Myanmar: a micro level study in the dry zone.
PhD dissertation, School of Environment, Resources and Development, Asian Institute
of Technology, Bangkok, Thailand.
Feder, G., Lau, L., Lin, J., and Luo, X., 1990. The relationship between credit and pro-
ductivity in Chinese agriculture: a microeconomic model of disequilibrium. American
Journal of Agricultural Economics, 72, 1151–1157.
Fujita, K., 2003. Policy-initiated expansion of summer rice under constraints of rural credit
in Myanmar in the 1990s: perspectives from a village study in Ayeyarwaddy division.
Keizai Kenkyu, 54(4), 300–314 (in Japanese).
Fujita, K., and Okamoto, I., 2000. An economic study on irrigated summer rice production
in Myanmar: the case of a village near Yangon. Southeast Asian Studies, 38(1), 22–49
(in Japanese).
Garcia, Y.T., Garcia, A.G., Oo, M., and Hossain, M., 2000. Income distribution and poverty
in irrigated and rainfed ecosystems: the Myanmar case. Economic and Political Weekly,
35(52/53), 4670–4676.
Gotsch, C.H., Ahmed, B., Falcon, W.P., Naseem, M., and Yusuf, S., 1975. Linear program-
ming and agricultural policy: micro studies of the Pakistan Punjab. Food Research
Institute Studies, 14(1), 3–105.
Jamison, D.T., and Lau, L.J., 1982. Farmer Education and Farm Efficiency. Baltimore:
Johns Hopkins University Press.
Kurosaki, T., and Fafchamps, M., 2002. Insurance market efficiency and crop choices in
Pakistan. Journal of Development Economics, 67(2), 419–453.
Kurosaki, T., and Khan, H., 2006. Human capital, productivity, and stratification in rural
Pakistan. Review of Development Economics, 10(1), 116–134.
Myat Thein, 2004. Economic Development of Myanmar. Singapore: Institute of Southeast
Asian Studies.
Okamoto, I., 2004. Agricultural marketing reform and the rural economy in Myanmar: a
case study of a pulse producing area in lower Myanmar. Paper presented at the 5th
Global Development Network Conference, New Delhi, January 28.
Okamoto, I., 2005. Rice marketing in the transition to a market economy in Myanmar.
In: K. Fujita (ed.) Myanmar’s Economy in Transition: Market versus Control. Chiba,
Japan: Institute of Developing Economics, pp. 231–271 (in Japanese).
Rozelle, S., and Swinnen, J.F.M., 2004. Success and failure of reform: insights from the
transition of agriculture. Journal of Economic Literature, 42(2), 404–456.
Takahashi, A., 2000. Myanmar’s Village Economy in Transition: Peasants’ Lives under the
Market-Oriented Economy. Tokyo: University of Tokyo Press (in Japanese).
Thawnghmung, A.M., 2004. Behind the Teak Curtain: Authoritarianism, Agricultural Poli-
cies and Political Legitimacy in Rural Burma. London: Kegan Paul.