University of Chicago Press Is Collaborating With JSTOR To Digitize, Preserve and Extend Access To Economic Development and Cultural Change

Green Revolution, Land Reform, and Household Income Distribution in the Philippines
Author(s): Keijiro Otsuka, Violeta Cordova and Cristina C. David

Source: Economic Development and Cultural Change, Vol. 40, No. 4 (Jul., 1992), pp. 719-741
Published by: University of Chicago Press
Stable URL: http://www.jstor.org/stable/1154631
Accessed: 12-12-2015 05:38 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1154631?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Economic Development and
Cultural Change.
http://www.jstor.org
This content downloaded from 128.143.23.241 on Sat, 12 Dec 2015 05:38:14 UTC
All use subject to JSTOR Terms and Conditions
Green Revolution, Land Reform, and Household
Income Distribution in the Philippines*
Keijiro Otsuka
Tokyo Metropolitan University
Violeta Cordova and Cristina C. David

International Rice Research Institute
I. Introduction
Income of agricultural households consists of returns to labor, capital,
and land from farming supplemented mainly by labor earnings from
nonfarm sources. Agricultural technology affects farmers' income
through its effect on factor demands and factor prices, which, in turn,
induce changes in the allocation of farmers' own resources to different
uses. Income from farming is also affected by village institutions such
as land tenure, which governs the distribution of income between land-
lords and tenants as well as resource allocations.
The introduction of fertilizer-responsive and high yielding modern
rice varieties (MVs) in the late 1960s, commonly referred to as the
Green Revolution, has been the most important technological change
in Philippine agriculture during the postwar period. Adoption of MVs,
however, has been concentrated in irrigated and favorable rain-fed
areas with adequate water control.' As a result, the productivity gap
between favorable and unfavorable rice production environments wid-
ened, and fears have been expressed that regional income distribution
worsened.2
The impact of differential technology adoption on regional factor
prices and factor incomes depends critically on the mobility of factors
of production. A recent study on rural labor markets in the Philippines
found that differential adoption of MVs was followed by interregional
labor migration from unfavorable to favorable areas, which signifi-
cantly contributed to the equalization of agricultural wages across pro-
duction environments.3 On the other hand, since land is an immobile
factor of production, productivity growth in the favorable rice growing
? 1992 by The University of Chicago. All rights reserved.

0013-0079/92/4004-0045$01.00
720 Economic Development and Cultural Change
areas will be expected to widen returns to land from rice production

across production environments. This can be mitigated by a shift to
alternative crops or institutional changes that alter the distribution of
returns to land.
In the Philippines, a land reform program was instituted in the
early 1970s soon after the widespread adoption of MVs. By converting
share tenants to leaseholders and amortizing owners with rents and
amortization fees fixed at about 25% of yield in the early 1970s, land
reform enabled tenant farmers to receive the major part of the benefits
of technological change in the 1970s. Land reform was particularly
well implemented in favorable rice growing areas, where returns to
land significantly increased because of MV adoption.4
The purpose of this article is to explore the differential impact of
MVs on factor incomes of rice farmers and to identify the extent to
which land reform redistributed the higher returns to land in favor of
tenants, which otherwise would have accrued to landlords. The analy-
sis is based on an intensive survey of 296 farm households and 82
landless agricultural households in five selected Philippine villages rep-
resentative of different production environments. After describing
technology adoption and land tenure in the survey villages in Section
II and factor incomes in rice production in Section III, we estimate
income determination functions by factor components to quantify the
effects of MV adoption, production environments, and land reform on
factor incomes. Based on the regression analysis of determinants of
household income, a counterfactual Gini decomposition analysis is
conducted to assess the distributional impact of MV adoption by com-
paring actual with counterfactual income distributions by factor com-
ponents.
II. Technology Adoption and Land Tenure in Survey Villages

The study villages are located in the two top rice growing areas in the
Philippines. Two villages, called CLi and CL2, are in the Central
Luzon plain, called the "rice bowl" of the Philippines, and the other
three villages, called P1, P2, and P3, are on the island of Panay (see fig.
1). Villages CL1 and P1 are fully irrigated by well-maintained gravity
irrigation systems and are representative of favorable production envi-
ronments in Central Luzon and Panay Island, respectively. The CL2
and P2 villages are characterized by shallow, favorable rain-fed condi-
tions commonly found in the country. Village P3 is also rain-fed but
is located in the most unfavorable mountainous environment, which is
prone to drought. After conducting a census of households, a stratified
random sample of farm and landless households was selected ac-
cording to farm size and tenure status. Detailed information on tech-
nology adoption, input use, yields, prices, tenure patterns, assets, and
K. Otsuka, V. Cordova, and C. David 721
PHILIPPINES
II
CI
P I
FIG. 1.-Location of sample villages
income from all sources was collected in two rounds of surveys cov-
ering the dry and wet seasons of 1985.
Table 1 shows adoption patterns of MVs and other technology
indicators across the survey villages. High yielding modern rice variet-
ies were fully adopted in CLI, PI, and CL2, whereas the traditional
varieties (TVs) were planted in the hilly part of P2 and the mountainous
portion of P3 during the wet season. In the irrigated villages, more
than two rice crops were grown. Double cropping of rice by using
TABLE 1
RATIO OF IRRIGATED AREA, ADOPTION OF MODERN VARIETIES, RICE CROPPING INTENSITY,
AND PADDY YIELDS IN SURVEY VILLAGES, 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
CLi P1 CL2 P2 RAIN-FED, P3
No. of sample farm households 92 37 54 63 50

Ratio of irrigated area (%) 100 100 16 0 0
Adoption of MVs* (% area) 100 100 100 79 59
Rice cropping intensity 200 243 114 131 125
Yield (tons/ha)t 4.7 3.6 3.4 2.9 1.9
* Refers to wet season only.
t Weighted average of wet and dry seasons (weights are ratios of planted areas).
irrigation pumps was practiced by a few farms in CL2. With shorter

growth duration of MVs and a more even rainfall pattern in Panay
Island, two crops of rice were grown under rain-fed conditions in some
parts of P2 and P3. Reflecting the differential adoption of MVs and
different production environments, average yield per hectare was
higher in the irrigated than in the rain-fed villages, particularly in the
unfavorable rain-fed village.
According to table 2, farm size is largest in CLI, the most favor-
able village in rice production, and smallest in P3, the village with the
most unfavorable rice growing environment. The ratio of agricultural
landless households is highest in P1 and lowest in P3. The landless are
geographically more mobile than farmers, and many of the landless in
the favorable villages are migrants from less favorable areas, who
sought better labor employment opportunities. As a result, man-land
ratio, which is defined as the ratio of village population to cultivated
area, tends to be larger in the more favorable villages within Central
Luzon and Panay Island.
Before the land reform program was initiated in 1972, the majority
of farmers were tenants, particularly in the Central Luzon villages,
where large rice haciendas consisting of hundreds of hectares of rice
lands existed. Almost without exception, tenants used to receive half
of the rice output after deducting the harvesters' share and costs of
purchased inputs supplied by landlords. The share of output that ac-
crued to tenants amounted to about one-third of gross output. Under
the land reform program, share tenants are supposed to be converted
either to leaseholders if the landlord owns less than 7 hectares of land
or to amortizing owners if the landlord owns more than 7 hectares.5
The Certificate of Land Transfer (CLT) was issued to amortizing own-
ers. It promises them the right to purchase the land by paying amortiza-
TABLE 2
FARM SIZE, RATIO OF LANDLESS HOUSEHOLDS, MAN-LAND RATIO, AND
TENURE PATTERNS IN SURVEY VILLAGES, 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
CL 1 P1 CL2 P2 RAIN-FED, P3
Farm size (ha) 2.1 1.1 1.8 1.5 .9

Ratio of landless household (%) 22 41 28 16 15
Man-land ratio (no./ha) 2.5 4.4 2.1 2.9 3.1
Tenure (% of area):
Owner 9 27 18 47 33
Leasehold/CLT* 80 38 76 32 8
Share tenancy 6 34 0 17 58
Pawn 6 2 6 5 1
* Certificate of Land Transfer.
tion fees over 15 years to the Land Bank. Both leasehold rents and
annual amortization fees, if paid equally for 15 years, were fixed at
about 25% of yield for 3 normal crop years preceding 1972. Since then,
yields have nearly doubled in villages where MVs have been adopted,
increasing the divergence between the rental value of the land and
the fixed leasehold rents and amortization fees prescribed by law
over time.
Adoption of modern rice technology seems to have raised the
share tenants' interest in acquiring the status of land reform beneficia-
ries.6 In fact, share tenants had been largely converted to leaseholders
and CLT holders in our survey villages except in P1, where yield gain
was achieved before 1972, and in P3, where yield has been stagnant
due to the adverse production environments. The extent to which the
land reform program helped redistribute the income gains from the
modern technology that accrued to land is a major issue to be ad-
dressed in later sections.
The introduction of MVs increases not only returns to land but
also returns to labor by increasing labor requirements for crop care,
harvesting, and threshing and by increasing profitability of dry season
cropping.7 The spread of MVs, however, was followed by adoption of
labor-saving technologies, such as use of tractors and threshers, and
of direct seeding, which saves labor for seedbed preparation and trans-
planting. While these labor-savings technologies undoubtedly reduced
the demand for labor, a recent technology adoption study by David
and Otsuka questions the widely held view that the adoption of MVs
induced the adoption of labor-saving technologies.8
As shown in table 3, total labor use per hectare was not necessar-
ily higher in the more favorable villages because of the offsetting ef-
TABLE 3
LABOR USE (Man-days/Ha) AND ADOPTION OF LABOR-SAVING TECHNOLOGY (%) IN
SURVEY VILLAGES, 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
CL1 P1 CL2 P2 RAIN-FED, P3
Land preparation 14 13 16 17 30
(% hand tractor)* (92) (85) (76) (32) (0)
Crop establishment 23 5 17 5 22
(% area with direct seeding) (57) (91) (48) (87) (47)
Care of crop 8 27 8 8 11
Harvesting and threshing 35 32 36 30 27
(% area with mechanical threshers) (100) (100) (99) (99) (86)
Others 2 4 2 2 12
Total labor 83 81 79 62 91
(% hired labor) (66) (80) (53) (64) (37)
NOTE.-Weighted average of dry and wet seasons (weights are ratios of planted
areas).
* In terms of percentage of area.
fects of MV technology and labor-saving technologies on total labor

use. It is, however, clear that the use of hired labor tends to be larger
in the more favorable villages, because MV technology increases the
peak season demand for labor. Since hired laborers are mostly agricul-
tural landless laborers, who belong to the poorest segment of the vil-
lage society, MV technology seems to improve income distribution
through its effect on the hired labor demand.
III. Factor Income in Rice Production

How the returns from rice production are distributed among factor
inputs and factor owners depends on relative factor prices and factor-
using or -saving characteristics of technology. Since rice production is
a major source of household income in our study villages, changes in
factor payments and factor shares in rice production significantly affect
the overall distribution of household income.
Factor Price Differential

Table 4 compares paddy prices, fertilizer prices, wage rates in selected
farm tasks, and custom rates of carabao (draft animal), tractor, and
thresher operations across villages during the 1985 wet season. As may
be expected, there are no significant differences in fertilizer and paddy
prices among the survey villages. The difference in wage rates across
villages varies by task. Yet, it is remarkable that wage rates are largely
similar in all villages, despite the wide locational difference and diverse
production environments across the study villages. This is consistent
TABLE 4
OF OUTPUT AND FACTORPRICESAMONGSURVEYVILLAGES,
COMPARISON
1985 WETSEASON
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
Paddy price (P/kg) 2.78 2.86 2.78 2.97 2.79

Nitrogen price (P/kg) 11.3 11.5 11.2 11.6 12.0
Wage rates (P/day):
Land preparation* 29 29 28 25 24
Repair of dikes 33 26 27 21 28
Transplanting 18 20 19 19 20
Fertilizer application 30 36 28 33 n.a.
Harvestingt 59 66 49 56 48
Custom rates:
Carabao (P/day) 54 49 53 47 37
Tractor (P/ha) 1,030 947 1,082 995 n.a.
Thresher (P/ton) 101 96 99 100 99
NOTE.-n.a. means not applicable.

*
Wage payment to operator of carabao.
t Imputed daily earning under output-sharing contract. Dry season data are shown
because harvesting wage rates were unusually low in villages CLi and CL2 due to severe
typhoon damage on yield.
with the hypothesis that differential adoption of modern rice technol-

ogy induced permanent migration from unfavorable to favorable pro-
duction environments so as to equalize the wage rates in different
areas.
Custom rates of carabao operation, however, tend to be lower in
the Panay villages than in the Central Luzon villages where availability
of pasture land is limited. On the other hand, the custom rates of hand
tractor and thresher operations are largely equalized. Such regional
difference in capital costs of carabao and tractor operations may ex-
plain, at least partly, why hand tractors are widely adopted in the
Central Luzon villages, even though wage rates are not significantly
higher (see table 3).
Factor Payments and Factor Shares

Under competitive factor market equilibrium, relative factor shares
can be shown to be determined by the relative factor prices and the
factor-using bias of technology.9 Factor market equilibrium cannot be
assumed in our study because the land markets are distorted by the
land reform program. However, if the production function is subject
to constant returns to scale, the competitive factor payment to land
can be estimated as a residual, after deducting from the value of pro-
duction the actual and imputed costs of current inputs, capital, and
TABLE 5
FACTOR PAYMENTS (P1,000) AND FACTOR SHARES (%) IN
RICE PRODUCTIONPER HECTARE, 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
Gross value of output 14.0 11.1 10.1 9.5 5.4

(100) (100) (100) (100) (100)
Current inputs 3.6 2.6 2.9 1.7 1.2
(26) (23) (29) (18) (22)
Capital 2.1 1.6 1.3 .8 .7
(15) (14) (13) (8) (13)
Owned* .6 .2 .4 .2 .3
Hired 1.5 1.4 .9 .6 .4
Labor 3.6 3.3 2.6 2.5 2.5
(26) (30) (26) (26) (46)
Familyt 1.5 1.0 1.3 .8 1.4
Hired 2.1 2.3 1.3 1.7 1.1
Residual 4.7 3.6 3.3 4.5 1.0
(34) (32) (33) (47) (19)
Leasehold rentt 1.7 2.2 1.5 1.5 1.3
Surplus? 3.0 1.4 1.8 3.0 - .3
NOTE.-Weighted average of dry and wet seasons (weights are the ratios of planted
areas). Figures in parentheses are the factor shares.
* Imputed cost using average machinery rentals in each village.
t Imputed labor cost using average daily earnings under piece rate contracts in
various tasks.
t Average leasehold rent.
? Residual minus leasehold rent.
labor."' In order to estimate the residual, the costs of family and ex-
change labor were imputed by applying the appropriate market wage
rates for different tasks and the costs of family-owned capital such as
carabaos, tractors, and threshers by the use of prevailing custom rates.
Table 5 shows the average factor payments and relative factor
shares per hectare per season estimated as average of wet and dry
season values using the ratios of cultivated areas in the two seasons
as weights. It is clear that the gross value of output per hectare is
largest in the most favorable village, CL1, and lowest in the most
unfavorable village, P3. Accordingly, the absolute factor payments in
each category of inputs tend to be higher in villages with better produc-
tion environments, which suggests that MV adoption in the favorable
areas results in higher returns to all factor inputs. On the other hand,
the relative labor share in P3, where the modern rice technology has
the least impact, is much higher than in other villages, which seems
to suggest that the modern technology as a whole, including biological
and mechanical innovations in combination with better production en-
vironments, has a labor-saving effect.
While the factor shares are concerned with the contribution of

factor inputs to production, the household income distribution is more
closely related to actual payments to owners of factor inputs. Thus,
the factor payments to labor and capital are divided into those accruing
to hired and family labor and those to rented and farmers' owned
capital, respectively, whereas the factor payments to land are classified
into actual leasehold rent and surplus to farm operators. The latter
distinction is important since actual land rents diverge from the rental
value of land where land reform is effectively implemented. In fact,
the estimated surplus is particularly large in CL1, CL2, and P2, where
share tenants have been largely converted to land reform beneficiaries.
It is clear that due to land reform, substantial amounts of income are
being redistributed from landlords to tenants in these villages.
The income accruing to hired labor is greater in the irrigated vil-
lages than in the rain-fed villages but remains, at most, 20% of gross
revenue. In contrast, the income share of land as measured by the
residual accounts for as much as 30%-50% of gross revenue. The only
exception is in P3, where the production environments are particularly
harsh due to mountainous topography and lack of irrigation.
It is evident from the above analysis that employment opportuni-
ties for hired labor in rice production are relatively limited. The resid-
ual returns to land and the surplus are the major sources of income
from rice by farm operators. As a result, the total income share of a
farm operator, defined as the sum of incomes that accrue to capital,
family labor, and land or tenancy right, is far greater than that of hired
labor, despite the lower factor share of family labor compared to hired
labor in rice cultivation. This clearly indicates that farmers are much
better-off than landless workers, unless lucrative job opportunities out-
side rice production are available.
IV. Determinants of Household Income

Basically household income consists of the sum of the products of
factor prices (or returns) and the quantity of factor inputs supplied by
the household. Economic theory is useful in analyzing how technology
affects factor demand and factor prices." On the other hand, house-
hold income is determined not only by technology and factor prices
but also by the ownership of productive resources, such as land, hu-
man capital, and capital assets as well as by market and institutional
factors, such as access to nonfarm employment opportunities and land
tenure. In order to identify the impact of modern rice technology and
land reform on household income, in this section we estimate income
determination functions by income components.
Household Income by Source

Average annual incomes of farm households and agricultural landless
households by source and by village are shown in tables 6 and 7,
TABLE 6
AVERAGE ANNUAL INCOME OF FARM HOUSEHOLDS BY SOURCE (P1,000), 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
CLi P1 CL2 P2 RAIN-FED, P3
Rice production 23.5 11.6 10.5 11.6 2.5

Labor income* 8.5 3.6 4.2 3.5 2.4
Capital incomet 3.1 .8 1.4 .8 .4
Land incomet 11.9 7.2 4.9 7.3 - .4
Nonrice production 7.1 16.4 7.4 8.6 7.4
Farm? 2.6 5.9 4.1 3.2 3.8
Nonfarm 1.9 4.5 2.2 2.2 2.4
Remittances and pensions 2.6 6.0 1.1 3.2 1.2
Total income 30.6 29.4 17.9 20.2 9.9
Household size 6.0 6.3 5.0 5.5 6.1

(No. of working members)11 (3.4) (3.8) (2.6) (3.2) (3.3)
Income per person 5.1 4.7 3.6 3.7 1.6
(Per working member) (9.0) (7.7) (6.9) (6.3) (3.0)
* Includes
imputed family labor income and actual labor earnings outside own farm.
t Includes imputed return to owned machineries and carabao and actual rental
earnings.
t Residual for owner-cultivators and residual minus actual rent payments for
tenants.
? Includes nonrice crops, livestock and poultry raising, and fishing.
ii Over 15 and below 60 years old.
respectively. While our survey villages are essentially rice-dependent,

nonrice agricultural income and nonfarm income account for signifi-
cant shares of total income of these households.12 Rice income is de-
composed according to the earnings of the household's owned labor,
capital, and land or tenancy right from rice production. Income from
labor and capital includes not only imputed returns to the family labor
and owned capital used in its own farm but also actual earnings by
working as hired labor or by renting out capital equipment to other
rice farms. For owner-operators, income from land is the residual,
whereas for tenants and leaseholders it is the residual net of rent pay-
ments or amortization fees. It should be noted that the residual actually
reflects not only the returns to land but also returns to management
and errors in imputing income of family labor and family-owned capi-
tal. Other income consists of nonrice agricultural income, nonfarm
labor-earnings, pensions, and remittances.
Income of farm households from rice production is by far the
highest in the most favorable village, CLI (table 6). This village has
the largest average cultivation size (2.1 hectares) and the highest yield
and rice cropping intensity. In sharp contrast is the meager rice income
TABLE 7
AVERAGE ANNUAL INCOME OF AGRICULTURAL LANDLESS HOUSEHOLDS BY SOURCE
(PI,000), 1985
FAVORABLE
IRRIGATED RAIN-FED
UNFAVORABLE
Rice production* 6.3 1.2 3.2 2.7 2.9

Nonrice production 5.1 12.6 4.2 11.0 2.7
Farmt 1.9 1.2 1.0 1.0 .9
Nonfarmt 2.8 6.9 2.4 7.7 1.0
Remittances and pensions .4 4.6 .8 2.3 .8
Total income 11.4 13.9 7.4 13.7 5.6
Household size 5.4 5.3 4.8 4.5 4.0

(No. of working
members)? (3.4) (3.8) (2.6) (3.2) (3.3)
Income per person 2.1 2.6 1.5 3.0 1.4
(Per working member) (3.4) (3.7) (2.8) (4.3) (1.7)
No. of households 20 23 20 15 4
* Labor earnings from rice production and rental earnings of carabao.
t Includes labor earnings from nonrice crop production, livestock and poultry rais-
ing, and fishing.
$ Labor earnings from nonfarm employment.
? The number of household members aged 15-60.
in the unfavorable rain-fed village, P3, where cultivation size is small-

est (0.9 hectares), yield is lowest, and rice is generally produced only
once a year. Furthermore, because of the drought during the year of
the survey, the average land income is negative. Rice income in the
irrigated village of PI is not larger than in the favorable rain-fed villages
of CL2 and P2, primarily because of the smaller farm size. The rela-
tively small rice income in this village, however, is compensated for
by large income from nonrice sources. In the other villages, nonrice
incomes are relatively similar. As a result, the total income of farm
households, both on a per household and on a per person basis, is
highest in the irrigated villages, second highest in the favorable rain-fed
villages, and lowest in the unfavorable rain-fed village.
The average income of an agricultural landless household, which
originates mostly from labor earnings, is much lower than that of farm
households primarily because of the difference in land income in rice
production (cf. data in tables 6 and 7). As might be expected, landless
households depend relatively more on nonrice income than do farmers.
There is, however, a considerable variation in the level of nonrice
income among the five villages. The income from nonfarm work in P1
and P2 is significantly higher than in other villages. As a result, the
total landless household income in PI and P2 exceeds the total income

in CLI even though income from rice production in these villages is
only one-fifth to two-fifths of the corresponding income in CL1. This
may suggest that more favorable nonfarm employment opportunities
exist in Panay Island than in Central Luzon, which, together with the
smaller cultivation size in the former region, led landless workers to
allocate more time to nonfarm jobs. However, in P3, which is also
locted in Panay Island, income from nonfarm work is relatively small.
Therefore, it appears that outside job opportunities alone cannot ade-
quately explain the difference in nonfarm income among the five vil-
lages.
Determinants of Household Income by Source

Extending the standard agricultural household model,13we assume that
a household maximizes utility by allocating labor time of family mem-
bers to various farm tasks and nonfarm jobs, cultivable lands to the
production of various crops, and capital inputs to various farm activi-
ties, subject to budget and time constraints, ownership of land and
capital stocks, and the crop production functions. From the first-order
optimum conditions, we can derive labor time, land, and capital ser-
vice allocation functions to various tasks that commonly depend on a
set of factor prices, technology, personal characteristics of household
members, and the ownership of land and nonland resources. Factor
prices including land rent or residual return to land depend on technol-
ogy and personal characteristics (e.g., management ability) and, hence,
cannot be assumed exogenous in our context. Thus, it is desirable to
estimate the factor price and factor allocation functions simulta-
neously. Such procedure, however, cannot be applied to this study
because the quantities and prices of household-supplied factors for
wide ranges of activities cannot be accurately estimated, as is generally
the case in village studies. Moreover, factor prices are quite uniform
within each village and some of them are relatively similar across the
five villages surveyed. Therefore, in this study we estimate the reduced
form income determination functions without distinguishing factor
prices and quantities, which depend on technology, ownership of re-
sources, and other household characteristics.14
Specifically, we regress four income components (land, labor, cap-
ital income from rice production, and nonrice income) on technology
factors (i.e., adoption of MVs and production environments); farm
size and tenure; ownership of capital (i.e., tractors, threshers, and
carabaos); number of working members aged 15-60; proxies for the
quality of labor (i.e., schooling, age, and sex of household head); and
other factors represented by village dummy variables, which would
affect factor returns and allocations. However, it is very difficult to
identify the effects of MVs and production environment on household
income separately from our cross-sectional data, primarily because

MVs are planted throughout the favorable areas in the survey villages.
In the unfavorable areas, both MVs and TVs are planted, but these
are marginal areas of MV adoption so that there is no significant differ-
ence in yield and profitability between MVs and TVs. In this study,
technology factors are specified as interaction terms between adoption
rate of MVs and production environments; that is, the ratios of MV
adoption and irrigated area (MV x IRG) represent farming conditions
in CLI and PI, and between the ratios of MV adoption and favorable
rain-fed area (MV x RF) represent farming conditions in CL2 and
some parts of P2. The omitted technology variable that serves as con-
trol pertains to unfavorable rain-fed areas planted to TVs, as well as
MVs, which are mostly found in P3. Thus, the interaction variables
between MVs and production environment will capture the combined
effects of MV adoption and favorable production environment.15 Note
here that their estimated coefficients may be biased upward, to the
extent that irrigated and favorable rain-fed production environments
positively affect both factor returns and MV adoption. For estimation
of income determination functions, the log-linear form was adopted
except for the ratio and dummy variables."'
Regression results of farm household income determination func-
tions by income component shown in table 8 are remarkably consistent
with a priori expectations. The land income is significantly affected by
technology, farm size, and tenure. The estimated coefficients of MV
x IRG and MV x RF are 1.61 and .97, respectively, which implies
that returns in unfavorable rain-fed land amount to 20% and returns in
irrigated areas and favorable rain-fed areas planted to modern varieties
amount to 40%. The coefficient of farm size is significantly greater
than zero but not significantly less than unity, suggesting the absence
of scale economies in rice production. The significantly negative coef-
ficient of ratio of leasehold and CLT area to total farm area (designated
as LEASEHOLD in table 8) is simply due to rent payments to land-
lords. More interesting is the finding that in absolute terms the coeffi-
cient of the ratio of share tenancy area (designated as SHARE TEN-
ANCY) is significantly larger than the coefficient of leasehold tenancy;
the difference suggests that the land income under share tenancy is
only a half of that under leasehold tenancy.17 This finding supports our
hypothesis that the land reform program redistributed income from
landlords to former share tenants. Since landlords are not included in
our sample, an increase in the income of tenants, who are wealthier
than the landless, might have worsened the income distribution among
our sample households.
Adoption of MV appears to have no significant effect on the other
components of rice income except for the MV x IRG in labor income,
which reflects the impact of double cropping of rice that is pervasive
TABLE 8
ESTIMATION RESULTS OF FARM HOUSEHOLD INCOME DETERMINATION FUNCTIONS
RICE INCOME In
NONRICE
In Land In Labor In Capital INCOME
Intercept - .84 1.99 - .31 -3.41

(-.67) (3.57) (-.23) (-2.77)
MV x IRG 1.61** .49* .55 .61
(2.88) (1.92) (.94) (1.11)
MV x RF .97** .17 .15 .60*
(2.58) (1.27) (.47) (1.99)
In FARM SIZE .84** .42** .58** .17
(7.63) (8.71) (4.78) (1.57)
LEASEHOLD - .42** .02 .30 - .24
(-2.44) (.27) (1.44) (-1.27)
SHARE TENANCY - 1.16** .03 .09 -.47*
(-4.37) (.30) (.34) (-2.06)
In SCHOOLING -.02 -.07 .30** .32**
(-.15) (-1.34) (2.35) (2.71)
In AGE .04 -.26* -.14 1.12**
(.14) (-2.03) (-.44) (3.89)
In WORKERS .12 .36** .02 .20
(.90) (5.47) (.10) (1.38)
FEMALE -.18 - .28* .09 .20
(-.62) (-2.10) (.26) (.66)
TRACTOR .25 .33** 1.92** .18
(1.41) (3.75) (9.44) (.92)
THRESHER - .07 .16 .61** - .26
(-.33) (1.46) (2.52) (-1.09)
CARABAO .15 .04 .04 - .03
(1.00) (.62) (.25) (- .23)
CLI 1.04* .41 - .83 - .87
(1.88) (1.60) (-1.43) (-1.62)
.85 -.28 - 1.54** -.09
P1
(1.58) (-1.16) (-2.67) (-.18)
CL2 .74* .07 - .80** - .78**
(1.91) (.06) (-2.35) (-2.44)
P2 1.44** -.10 -.32 -.22
(4.39) (-.87) (-1.23) (-.90)
R2 .54 .68 .53 .22
F-value 15.36 36.97 16.65 4.99
NOTE.-Numbers in parentheses are t-values.

*
Significant at 5% level.
** Significant at 1% level.
under irrigated conditions. It is interesting to note, however, that MV

adoption in favorable rain-fed areas has a significant and positive effect
on the level of nonrice income. This is due to the fact that the shorter
growth duration MVs allowed other crops to be grown in rain-fed areas
where the rainfall pattern is more even.
As in the case of the residual income, farm size is also positively
correlated with labor and capital income from rice but not with nonrice
income. Since not only labor income from rice but also nonrice income
represent, essentially, returns to labor, many of the coefficients of the
characteristics of labor supply represented by the number of working
household members (designated as WORKERS) and of labor quality
represented by schooling, age, and gender of household head are sig-
nificant in these income functions. In particular, it is interesting to
observe that while the coefficient of schooling is not significant in
the regression equation for labor income from rice, it is an important
explanatory variable for nonrice income as well as capital income.
Ownership of tractors and threshers are also highly significant factors
explaining differences in capital income. Ownership of tractors also
has a positive and significant coefficient in the labor income regression,
because farmers typically operate their own tractors.
The estimated coefficients of village dummies are positive and
generally significant in the land income regression, which can be ex-
plained by the particularly unfavorable rice production environment in
P3. Also noteworthy are the nonsignificant effects of these dummies
in labor income and nonrice income regressions, which suggests the
absence of any particular locational advantages or disadvantages in
income earning opportunities from both rice and nonrice sources
among our sample villages. On the other hand, the negative coefficients
of village dummies in the capital regression seem to indicate the exis-
tence of active capital rental markets in some favorable villages, where
farmers generally rent out their tractors and threshers to nonvillagers.
With respect to the income of landless households (table 9), village
specific technology variables represented by the interaction terms MV
x IRG and MV x RF did not show any significant effects in both
equations (1) and (3). The average farm size in the village has a signifi-
cant and positive coefficient in rice income equation (1) but a negative
coefficient in the nonrice income equation (3). Alternatively, village
dummies are used in equations (2) and (4) to represent environmental,
technological, and socioeconomic characteristics of sample villages.
The estimated coefficients of these dummies, however, are not signifi-
cant except for CLi in equation (2), where the farm size is largest. As
in the case of farm households, village characteristics do not seem to
be critical determinants of labor income of landless households from
both rice and nonrice sources.
Income from rice is significantly affected by ownership of carabao,
which can be rented out, and by being a permanent laborer earning a
share of the crop or a fixed amount of rice per season under assured
employment conditions in rice farming throughout a crop season."8
Landless households with no permanent labor contract in rice farming
depend largely on other sources of income. More important, it is found
that schooling, the number of workers, and gender are important deter-
minants of differences in levels of nonrice income but not of rice in-
TABLE 9
FUNCTIONS
ESTIMATIONRESULTSOF LANDLESSHOUSEHOLDINCOMEDETERMINATION
In RICE INCOME In NONRICEINCOME
(1) (2) (3) (4)
Intercept .59 -1.01 -3.75 -2.09

(.35) (-.71) (1.53) (-1.02)
In SCHOOLING .09 .06 1.25** 1.28**
(.49) (.30) (4.73) (4.60)
In AGE .31 .32 .11 .11
(.89) (.90) (.23) (.22)
In WORKERS .18 .20 .61* .60*
(.78) (.84) (1.81) (1.77)
FEMALE - .34 - .42 .91* .97**
(-1.11) (-1.25) (2.24) (2.34)
CARABAO .66* .64* - .32 - .32
(2.22) (2.17) (- .72) (-.72)
PERMANENT LABORER .66* .69* - 1.15** - 1.18**
(2.30) (2.35) (-2.60) (-2.62)
Village-specific variables:
In FARM SIZE 1.86* ... -.84
(3.98) (-1.29)
MV x IRG -.02 . . .03
(-1.64) (1.58)
MV x RF -.02 ... .02
(-1.89) (1.37)
CL . .. 1.12* . .. .25
(2.18) (.32)
P1 - .04 . . .74
(-.09) (1.05)
CL2 ... .41 . . .35
(.88) (.48)
P2 ... .66 ... -.19
(1.41) (-.26)
R2 .430 .433 .457 .458
F-value 4.94 4.43 6.73 6.00
NoTE.-Numbers in parentheses are t-values.

*
Significant at 5% level.
** Significant at 1% level.
come. Thus, the difference in human capital seems important for ex-
plaining intervillage differences in labor income from nonrice sources.
V. Decomposition Analysis of Income Inequality

Existing studies of the impact of technical change on personal income
distribution simply compared measures of income inequality before
and after or between areas with and without technological change.19
Such analysis is at best suggestive because other factors besides tech-
nology also change over time or differ between areas. In this article,
the Gini coefficients for total and various components of income based
on actual income are compared with those based on counterfactual
income without MVs. The latter is derived by deducting the estimated

contribution of MVs to income from actual income, using variables in
household income determination functions.20 Since adoption of MVs
affects the various income components differently, a decomposition
analysis of the Gini measure of inequality is applied to quantify the
relative importance of the various income components in accounting
for the overall income inequality.21
The Gini decomposition formula is as follows:
G(y) = SiR(y, xi) G(xi), (1')
where G(y) = Gini ratio of total income; xi = income from ith source;
Si = average income share of ith source; R(y, xi) = rank correlation
ratio; and G(xi) = Gini ratio of ith income. The rank correlation ratio
is defined as
R(y, xi) = (2')

Cov[xi, r(y)]/Cov[xi, r(xi)],
where r(y) and r(xi) denote ranking of households in terms of y and xi,
respectively. It is clear that R is unity if
r(y) = r(xi). (3')
Otherwise, R is shown to be less than unity. In general, R is larger,

the larger the correlation between y and xi 22
Thus, the formula in equation (1') decomposes the Gini ratio of
income inequality into income shares, correlation effects, and the Gini
ratio for each factor component. It is important to note here that in
assessing the contribution of various income sources to overall income
inequality, we must consider not only the Gini ratio of component
income but also the income share and rank correlation.23 Using this
formula we can assess, for example, the contribution of the land in-
come distribution to the overall income inequality. The main desirable
feature of this formula is its clear link with the conventional economic
analysis. In particular, income shares are closely related with technol-
ogy bias in production and factor prices, which affect factor shares as
well as earner shares. Thus, the conventional factor share and earner
share analysis and the analysis of technology bias are directly relevant
for this decomposition analysis.24 Moreover, the rank correlation ratio
broadly indicates how various sources of income are substituted as a
result of time and land allocation decisions by agricultural households.
There are several limitations to our income distribution analysis.
First, the sample households did not include most of the landlords who
typically reside outside the village. Second, the dynamic impact of
TABLE 10
OVERALLGINI RATIOSAND CONTRIBUTIONS
BY INCOMECOMPONENT:COMPARISON
OF
ACTUAL CASE WITHCOUNTERFACTUAL CASE OF No MV ADOPTION
PERCENTAGECONTRIBUTION
BY
OVERALL RICE PRODUCTION*
GINI NONRICE
VILLAGES RATIO LAND LABOR CAPITAL INCOME
Actual:
CLI .42 42 23 16 19
P1 .46 19 7 4 70
CL2 .45 29 20 12 38
P2 .37 38 14 7 41
P3 .41 5 20 4 71
All .46 34 22 12 32
Counterfactual:
CLI .39 12 21 28 39
P1 .45 4 2 4 90
CL2 .38 15 28 19 37
P2 .35 15 26 11 48
P3 .34 1 32 6 61
All .42 10 22 19 49
* Numbers show the
percentage distribution of each income component to overall
Gini ratio, which is measured by the product of income share, rank correlation, and
component Gini ratio.
technical change on the interregional migration from unfavorable to

favorable areas, distribution of land ownership, and tenure status,
among others, cannot be identified. Third, we cannot identify the sepa-
rate effects of MV adoption and production environments, so that the
counterfactual analysis indicates not only the effect of MVs but also
the effect of production environments. Nonetheless, the estimation
results of the Gini coefficients by factor components based on actual
and counterfactual income provide some interesting insights arising
mainly from the effect of MV adoption on returns to land in favorable
areas.
Table 10 shows the computation results of actual and counterfac-
tual Gini ratios and the relative percentage contributions of factor com-
ponents measured by the product of income share, rank correlation
ratio, and component Gini ratio. In the actual case, the computed
overall Gini ratios are highest in P1 and CL2 and lowest in P2 and P3.
The most important income source contributing to the overall income
inequality in these villages is nonrice income. Because of the impor-
tance of nonrice income, which accounts for 40%-65% of total house-
hold income, its relative contribution to the overall Gini ratio amounts
to 38%-71%, even though the component Gini ratio and the rank corre-
lation ratios of nonrice income are low compared with other income
sources in these villages. In contrast, in CL1, the most favorable vil-

lage for rice production, the income share of nonrice income is as
low as 19% and the most important contributor to the overall income
inequality is the land income, whose component Gini ratio and rank
correlation ratio tend to be highest among four income sources not
only in this village but also in other villages. Overall, the importance
of labor and capital income from rice farming in the income distribution
is relatively minor. These observations suggest that given the unequal
distribution of ownership and/or access to land, the land income is a
significant income inequalizing factor, particularly in favorable rice
growing villages, whereas nonrice income is a more important factor
in villages where farm size is small (i.e., P1 and P3) and double crop-
ping of rice is not practiced because of the lack of irrigation (i.e., CL2,
P2, and P3). In the less favorable environment for rice production,
more resources are allocated to the production of nonrice crops and
nonfarm jobs. Such substitution relationships between income from
rice farming and nonrice income suggests that if there had been no
MVs, the contribution of nonrice income to the overall income inequal-
ity would have been much larger.
This is confirmed by the result of counterfactual analysis. As may
be expected, the overall Gini ratios are somewhat larger in the actual
than the counterfactual computation mainly because of the drastically
reduced contribution of land income to the overall income inequality
in the absence of MVs. However, because of the increased contribu-
tion of nonrice income to the overall Gini ratio, decreases in overall
income inequality due to the elimination of income effects of MV adop-
tion in favorable areas are much smaller than decreases in the contribu-
tion of land income. Such a change mainly reflects the greater relative
importance of nonrice income in the absence of MVs. As a result, the
decrease in the overall Gini ratio is relatively small not only in less
favorable villages but also in favorable villages, which seems to indi-
cate a relatively small effect of MV adoption on income inequality. It
should be noted that even this decrease could be an overestimate,
given the possible upward bias in the estimated coefficient of the MV
adoption variable mentioned in the previous section.
Our analysis does not take into account the dynamic impact of
MV adoption, such as interregional migration, which significantly
changes the composition of farm and landless households because mi-
grants are mostly landless laborers.25 In order to check the possible
bias of our analysis arising from the migration, we also computed the
Gini ratio and the income inequality contributions by factor component
using the pooled data of five villages. As table 10 shows, while the
Gini ratio based on all village data tends to be higher than the individual
village Gini ratios due to the diversity of natural and economic environ-
ments across villages, major qualitative results remain the same.
VI. Concluding Remarks

In this article, we have attempted to examine the impact of differential
MV adoption and land reform on functional and household income
distribution. Our analysis indicates that the ultimate income distribu-
tional consequences of differential adoption of MVs have not been
significantly adverse, in part because the inequitable effect of MV
adoption on regional income distribution in favorable areas was miti-
gated by the implementation of land reform and reallocation of re-
sources to nonrice production activities in unfavorable areas. Note
that our analysis is likely to overestimate the inequitable effect of the
differential adoption of MVs, because the positive effect of MVs on
wages through their direct and indirect effects on labor demand-
which is obviously equitable-is not taken into account. If such effects
could have been properly assessed, the inequitable impact of MV
adoption would have been shown to be more limited.
As far as the income of landless laborers (who belong to the poor-
est segment of the society) is concerned, there is no strong evidence
to show that the differential MV adoption itself has made them signifi-
cantly worse-off. Our analysis reveals the importance of human capital
in explaining the level of nonrice labor earnings of landless laborers.
This implies that investments in education of poor landless laborers is
an appropriate means to improve their well-being. Moreover, invest-
ment in agricultural research for the development of rice technology
suitable for the unfavorable production environments is likely to be of
secondary importance for reducing the regional income differential of
the poor. The poor are geographically mobile, and hence their relative
incomes are not significantly affected by differential MV adoption or
by the production environment, at least in the long run.
A more effective means for improving the welfare of the popula-
tion in the unfavorable areas is to invest in agricultural research for
crops more suitable to these environments, such as corn, root crops,
or tree crops. In addition, the government could contribute by in-
vesting in market infrastructure, by assisting farmers in lowering the
cost of production, and by facilitating interregional migration.
Notes
* We wish to thankLuisa Bambo,CristinaCarpio,DanielCeballo,Esther
Marciano,MilagrosObusan,Dolor Palis, and Sylvia Sardidofor their invalu-
able researchassistance. We have also benefitedfromthe helpfulcommentsof
RandolphBarker,Ernesto Bautista,Ian Coxhead,RobertE. Evenson, Yujiro
Hayami, ShigeruIshikawa, Justin Y. Lin, Konosuke Odaka, Jaime Quizon,
C. H. H. Rao, Nicholas Stern, GaneshThapa, and an anonymousreferee.
1. RandolphBarker and Robert W. Herdt, The Rice Economy of Asia
(Washington,D.C.: Resources for the Future, 1985).
2. See Cristina C. David and Keijiro Otsuka, "The Modern Seed-
FertilizerTechnologyand Adoptionof Labor-SavingTechnologies:The Philip-
pine Case," Australian Journal of Agricultural Economics 34 (August 1990):

132-46; Michael Lipton and Richard Longhurst, New Seeds and Poor People
(London: Unwin Hyman, 1989); and Vernon W. Ruttan, "The Green Revolu-
tion: Seven Generalizations," International Development Review 19 (Decem-
ber 1977): 16-23.
3. Keijiro Otsuka, Violeta G. Cordova, and Cristina C. David, "Modern
Rice Technology and Regional Wage Differentials in the Philippines," Agricul-
tural Economics 4 (1990): 297-314.
4. Keijiro Otsuka, "Determinants and Consequences of Land Reform
Implementation in the Philippines," Journal of Development Economics 35
(1991): 339-55.
5. See, e.g., Yujiro Hayami, M. Agnes Quisumbing, and Lourdes S.
Adriano, Toward an Alternative Land Reform Paradigm: A Philippine Per-
spective (Quezon City: Ateneo de Manila University Press, 1990).
6. See Otsuka.
7. Randolph Barker and Violeta G. Cordova, "Labor Utilization in Rice
Production," Economic Consequences of the New Rice Technology (Los
Bafios: International Rice Research Institute, 1978); and Barker and Herdt.
8. See David and Otsuka.
9. Toshihiko Kawagoe, Keijiro Otsuka, and Yujiro Hayami, "Induced
Bias of Technical Change in Agriculture: The United States and Japan, 1880-
1980," Journal of Political Economy 94 (June 1986): 523-44.
10. The assumption of constant returns to scale in this study is supported
by estimation results of the Cobb-Douglas production function (see Violeta G.
Cordova, "Rice Technology, Productivity, and Economic Returns between
Favorable and Unfavorable Villages in the Philippines" [International Rice
Research Institute, Agricultural Economics Department, Los Bafios, 1987]).
11. See particularly Robert E. Evenson, "Gains and Losses from Agricul-
tural Technology," Philippine Economic Journal 14 (November 1975): 363-79;
and Jamie M. Quizon and Hans P. Binswanger, "Income Distribution in
Agriculture," American Journal of Agricultural Economics 65 (August 1983):
526-38.
12. Nonfarm income is generally important in agricultural household in-
come in LDCs (see Dennis Anderson and Mark Leiserson, "Rural Nonfarm
Employment in Developing Countries," Economic Development and Cultural
Change 28 [January 1980]: 227-48).
13. See Inderjit Singh, Lyn Squire, and John Strauss. Agricultural House-
hold Models: Extensions, Applications, and Policy (Baltimore: Johns Hopkins
University Press, 1986).
14. For further details of theoretical considerations underlying the income
determination function approach, see Keijiro Otsuka and Cristina C. David,
"An Integrated Theoretical Framework for the Extensive and Intensive Sur-
vey Studies" (paper presented at the third workshop on Differential Impact
of Modern Rice Technology on Favorable and Unfavorable Production Envi-
ronments, Dhaka, Bangladesh, March 29-April 6, 1989).
15. It must be pointed out that these interaction variables do not simply
capture the impact of favorable agronomic environments relative to unfavor-
able rain-fed environment on the dependent variables. In fact, the productivi-
ties of TVs measured by yields per hectare were similar across production
environments, implying that productivities in favorable environments in-
creased only when MVs were introduced (see David and Otsuka [n. 2 above]).
Note that in order to isolate the impact of MVs, we also estimated the income
determination functions, which include MV adoption rate and/or irrigation
ratio without interactions. Although the qualitative results are essentially the
same, we chose the specification with interaction terms because of higher
significance and plausible magnitudes of the estimated coefficients of the tech-
nology variables.
16. Because of log-linear specification, some observations with negative
values of land income and zero labor earning from nonrice sources were dis-
carded. There was no significant difference in the results when the functions
were estimated in linear form with interaction terms. The log-linear form was
chosen mainly because of the robustness and the ease of interpretation of the
estimated coefficients.
17. According to Keijiro Otsuka and Yojiro Hayami's survey of tenancy
literature, "Theories of Share Tenancy: A Critical Survey," Economic Devel-
opment and Cultural Change 37 (October 1988): 31-68, income of share ten-
ants per unit of land tends to be lower than that of leasehold tenants by
20%-30%, presumably because of the larger production risk incurred by the
latter even in the absence of rent regulations.
18. For an analysis of permanent labor contracts in Central Luzon, see
Keijiro Otsuka, Hiroyuki Chuma, and Yujiro Hayami, "Permanent Labor
and Land Tenancy Contracts in Agrarian Economies: An Integrated Analy-
sis" (Tokyo Metropolitan University, Faculty of Economics, 1990, mimeo-
graphed).
19. See, e.g., V. T. Raju, "Impact of New Agricultural Technology on
Farm Income Distribution in West Godavari District, India," American Jour-
nal of Agricultural Economics 58 (May 1976): 346-50; and R. H. Goldman and
Lyn Squire, "Technical Change, Labor Use, and Income Distribution in the
Muda Irrigation Project," Economic Development and Cultural Change 30
(July 1982): 752-75. While these studies do not consider nonfarm income, its
importance in overall distribution of income is reported by Dennis L. Chinn,
"Rural Poverty and the Structure of Farm Household Income in Developing
Countries," Economic Development and Cultural Change 27 (January 1979):
283-301; and R. T. Shand, "Income Distribution in a Dynamic Rural Sector:
Some Evidence from Malaysia," Economic Development and Cultural Change
36 (October 1987): 35-50.
20. We also conducted a counterfactual Gini decomposition analysis of
the effect of land reform. The results, however, did not show an equitable
effect of land reform primarily because landlords are not included in our
sample.
21. The Gini decomposition formula was developed by J. C. H. Fei,
Gustav Ranis, and S. W. Y. Kuo, "Growth and Family Distribution of Income
by Factor Components," Quarterly Journal of Economics 92 (February 1978):
17-53; and Graham Pyatt, Chau-nam Chen, and J. C. H. Fei, "The Distribu-
tion of Income by Factor Components," Quarterly Journal of Economics 95
(November 1980): 451-73. The inequality contribution assigned to any income
source can vary depending on the choice of inequality measures as pointed
out by A. F. Shorrocks, "The Impact of Income Components on the Distribu-
tion of Family Income," Quarterly Journal of Economics 98 (May 1983):
310-26.
22. In the computation of G(y), households are ranked in accordance with
y, but in the case of G(xi) they are ranked in accordance with xi. In order to
adjust this difference, the rank correlation ratio appears in the formula. In
fact, R(y, xi) - G(xi) is equal to the pseudo-Gini coefficient G(xi) defined by
Fei, Ranis, and Kuo, which is obtained if we use the ranking of household in
accordance with total income y in the computation of factor Gini for xi.
23. To illustrate the problem of interpreting the Gini ratio of component

income, suppose that the land income of all farm households increased propor-
tionally. Obviously, income distribution becomes less equitable because the
income of landless households does not change. Nevertheless, the Lorenz
curve for the residual income remains unchanged and, hence, the Gini ratio
remains the same. In this case, however, the inequitable effects are reflected
in increases in the residual income share and the rank correlation ratio.
24. For the conventional factor share analysis, see C. G. Ranade and
Robert W. Herdt, "Shares of Farm Earnings from Rice Production," in Eco-
nomic Consequences of the New Rice Technology (Los Bafios: International
Rice Research Institute, 1978); and for the analysis of technology bias in agri-
culture, see Hans P. Binswanger, "The Measurement of Technical Change
Biases with Many Factors and Production," American Economic Review 64
(December 1974): 964-76; and Kawagoe, Otsuka, and Hayami (n. 9 above).
25. See Otsuka, Cordova, and David (n. 3 above) for statistical evidence
of the interregional migration in the Philippines.
American
Journal of Editedby RichardAdams and Steven Buccola
Agricultural OregonState University,Corvallis,Oregon
Economics Publishedby the AmericanAgriculturalEconomicsAssociation
May 1992
Articles:GordonC. Rausserand PinhasZusman,"PublicPolicy:Explanationand Constitutional
EricO'N.Fisherand Harryde Gorter,"TheInternational
Prescription"; Effectsof U.S. FarmSubsidies";
Alainde Janvryand ElizabethSadoulet,"AgriculturalTradeLiberalizationand the LowIncome
Countries:A GeneralEquilibrium-Multimarket Approach";S. Devadoss,"Market Interventions,
International
PriceStabilization,and WelfareImplications";
HowardD. Leathers,"TheMarketfor Land
and the Impactof FarmProgramson FarmNumbers";J. Stephen Clarkand CurtisE. Youngblood,
"EstimatingDualityModelswith Biased TechnicalChange";G. Scott Smithand MichaelE. Wetzstein,"A
StochasticAsset ReplacementModelfor RejuvenatedAssets";James A. Duffieldand RobertColtrane,
"Testingfor Disequilibrium
in the HiredFarmLaborMarket"; RichardB. Standifordand RichardE.
Howitt,"SolvingEmpiricalBioeconomicModels:A RangelandManagementApplication"; JulieA.
Caswelland DanielI. Padberg,"Towarda MoreComprehensiveTheoryof Food Labels";RogerA.
Dahlgranand Steven C. Blank,"Evaluating the Integrationof Contiguous,DiscontinuousMarkets";
plus otherarticles,comments, and book reviews.
Annualmembershipdues (includingJournal)$60; Annuallibrary
subscriptionrate$90; Individual
copies $18.00; ContactAAEA
Business Office,80 HeadyHall,IowaState Univ,Ames, IA50011.
Publishedin February,May,August,November,and December.

University of Chicago Press Is Collaborating With JSTOR To Digitize, Preserve and Extend Access To Economic Development and Cultural Change

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

University of Chicago Press Is Collaborating With JSTOR To Digitize, Preserve and Extend Access To Economic Development and Cultural Change

Uploaded by

Copyright:

Available Formats

Green Revolution, Land Reform, and Household Income Distribution in the Philippines

Author(s): Keijiro Otsuka, Violeta Cordova and Cristina C. David

You may need to log in to JSTOR to access the linked references.

Violeta Cordova and Cristina C. David

? 1992 by The University of Chicago. All rights reserved.

areas will be expected to widen returns to land from rice production

II. Technology Adoption and Land Tenure in Survey Villages

FIG. 1.-Location of sample villages

No. of sample farm households 92 37 54 63 50

irrigation pumps was practiced by a few farms in CL2. With shorter

Farm size (ha) 2.1 1.1 1.8 1.5 .9

fects of MV technology and labor-saving technologies on total labor

III. Factor Income in Rice Production

Factor Price Differential

Paddy price (P/kg) 2.78 2.86 2.78 2.97 2.79

NOTE.-n.a. means not applicable.

with the hypothesis that differential adoption of modern rice technol-

Factor Payments and Factor Shares

Gross value of output 14.0 11.1 10.1 9.5 5.4

While the factor shares are concerned with the contribution of

IV. Determinants of Household Income

Household Income by Source

Rice production 23.5 11.6 10.5 11.6 2.5

Total income 30.6 29.4 17.9 20.2 9.9

Household size 6.0 6.3 5.0 5.5 6.1

respectively. While our survey villages are essentially rice-dependent,

Rice production* 6.3 1.2 3.2 2.7 2.9

Total income 11.4 13.9 7.4 13.7 5.6

Household size 5.4 5.3 4.8 4.5 4.0

in the unfavorable rain-fed village, P3, where cultivation size is small-

total landless household income in PI and P2 exceeds the total income

Determinants of Household Income by Source

income separately from our cross-sectional data, primarily because

Intercept - .84 1.99 - .31 -3.41

NOTE.-Numbers in parentheses are t-values.

under irrigated conditions. It is interesting to note, however, that MV

In RICE INCOME In NONRICEINCOME

(1) (2) (3) (4)

Intercept .59 -1.01 -3.75 -2.09

NoTE.-Numbers in parentheses are t-values.

V. Decomposition Analysis of Income Inequality

income without MVs. The latter is derived by deducting the estimated

G(y) = SiR(y, xi) G(xi), (1')

R(y, xi) = (2')

r(y) = r(xi). (3')

Otherwise, R is shown to be less than unity. In general, R is larger,

technical change on the interregional migration from unfavorable to

sources in these villages. In contrast, in CL1, the most favorable vil-

VI. Concluding Remarks

pine Case," Australian Journal of Agricultural Economics 34 (August 1990):

23. To illustrate the problem of interpreting the Gini ratio of component

You might also like