Академический Документы
Профессиональный Документы
Культура Документы
1043±1056, 2001
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
www.elsevier.com/locate/worlddev 0305-750X/01/$ - see front matter
PII: S0305-750X(01)00018-3
and
BENJAMIN DAVIS *
Food and Agriculture Organization, Rome, Italy
Summary. Ð Cash transfer programs create multiplier eects when recipients put the money they
receive to work to generate further incomes. When this is the case, the ultimate income eects are
multiples of the amounts transferred. We analyze the PROCAMPO program in Mexico that was
introduced to compensate farmers for the anticipated negative eect of NAFTA on the prices of
basic crops. The transfer rules and the timing of the panel data collected allow unique control of
biases in this impact analysis. We ®nd that the multiplier among ejido sector recipients is in the
range 1.5±2.6. Multipliers are higher for households with medium and large farms, low numbers of
adults in the household, nonindigenous backgrounds, and located in the Center and Gulf regions.
High multipliers re¯ect income opportunities that had remained unrealized due to liquidity
constraints that are relaxed by the transfers. Opportunities come from the asset endowments that
these households received through the land reform, particularly irrigated land, and they are
enhanced by access to technical assistance. Ó 2001 Elsevier Science Ltd. All rights reserved.
by the targeted households (Cox & Jimenez, We calculate the magnitude of the short-run
1992; Cox, Eser, & Jimenez, 1998). These pro- income multipliers created by these transfers.
grams can also have positive indirect eects We also identify under what conditions and for
when the money transferred is used to acquire what types of households were these multipliers
productive inputs, creating short-run multiplier the largest. This, in turn, provides guidelines
eects, and to invest in productive assets, gen- for the management of transfer programs to
erating long-run multipliers on the cash re- rural poor households in order to maximize
ceived. multiplier eects.
To our knowledge, there are no studies of the Assessing the impact of a program is usually
positive indirect eects of cash transfer pro- plagued with the diculty of controlling for
grams via income generation by putting the biases arising from unobserved individual or
cash transferred to work. Yet, if the household regional characteristics that are correlated with
is liquidity constrained and hence has under- program placement and program participation.
employed and ill-allocated productive assets In the case of PROCAMPO, systematic na-
relative to an unconstrained situation, the cash tional coverage eliminates the standard pro-
transfer should generate bene®ts at least similar gram placement bias. Participation of eligible
to a credit program, and expectedly higher households is almost universal. But, eligibility
since there is no risk of failure to repay. The itself and the amount that households receive
money transferred can be used to purchase are explicitly related to their cropping patterns.
current inputs or to invest in physical and hu- There is no doubt that unobserved household
man capital. Whether income eects can be attributes aect both the cropping pattern, and
observed in the short-term depends on the hence the PROCAMPO transfers, and the
maturation time of the expenses. Expenses on outcome of interest, namely household income.
inputs for agricultural, commercial, or micro- Yet, we are able to exploit a truly exceptional
manufacturing activities will have eects visible situation where the basis for PROCAMPO
in the short-run, as opposed to investments in transfers are the 1993 cropping patterns, while
equipment and especially in human capital that the program itself started only in the Fall of
require longer maturation periods. 1994, and households in the panel were sur-
In this paper, we analyze the direct and in- veyed in the Springs of 1994 and 1997. As the
direct income eects of cash transfers to Mex- 1994 survey took place prior to the PRO-
ican farm households in the context of CAMPO transfers, but after the rules were set,
PROCAMPO (Program for Direct Assistance a household ®xed-eects estimation circum-
in Agriculture). This program was introduced vents the problem of household unobservables
to compensate for the anticipated negative that could bias the estimated impact of the
price eects of trade liberalization on basic PROCAMPO program. Variability in the
crops. We focus on the ejido sector, a large amount received by the dierent households
sector of generally impoverished households then allows to estimate the marginal eect of
that received access to land through the exten- one unit of transfer.
sive land reform program that ran during 1930±
92. These households are in the unusual posi-
tion of being endowed with productive assets 2. THE PROCAMPO PROGRAM
while at the same time severely starved for ac-
cess to credit due to the incomplete nature of As a consequence of NAFTA, trade liberal-
property rights in the ejido that prevents them ization for basic crops competitive with United
from using the land as collateral to access States and Canadian exports was anticipated to
credit. In this context, cash transfer programs create a sharp decline in domestic prices for
can be expected to have particularly large Mexican producers as prices for these crops
multiplier eects on income. In addition, they had been supported above border prices. The
can be expected to lead to visible labor reallo- PROCAMPO program was introduced when
cation eects toward the activities where the NAFTA started (Winter of 1994) as a com-
cash received is used if household labor was pensatory income transfer targeted to the pro-
partially allocated as a strategy to overcome ducers of these crops. The objectives were
credit market failures. Finally, cash transfers political (to manage the political acceptability
can be used to invest in productive assets and of the free trade agreement among farmers),
hence increase the future income position of economic (to provide farmers with liquidity to
recipients. adjust production to the new set of relative
CASH TRANSFER PROGRAMS 1045
prices), and social (to prevent an increase in out over the remaining ®ve years of the pro-
already extensive levels of poverty among gram. But, the real value of payments was not
smallholders and a rapid process of outmigra- fully maintained as it was left to erode from
tion to the cities and the border in the North). US$102 to US$68 per hectare during 1994±97.
The program was designed as a 15 years tran- In spite of this, transfers still represent a very
sition toward free trade. signi®cant cash contribution to farm house-
Transfers are on a per hectare basis, decou- holds, particularly the poor, with the potential
pled from current land use and ®xed across the of not only adding importantly to their incomes
whole country. Hectares that qualify were but also aecting their behavior as farm pro-
planted in any of nine basic crops (corn, beans, ducers, workers, and entrepreneurs in other
rice, wheat, sorghum, barley, soybeans, cotton, income-generating activities.
and cardamom) in one of the three agricultural Current participation in PROCAMPO is
years preceding August 1993. Since there are limited to the households that were incorpo-
two agricultural cycles per year (Fall±Winter rated into the program when it was introduced
and Spring±Summer), payments are made twice in 1994. At that time, farmers had to show that
a year for the area that had been planted in the they had planted at least one of the nine staple
corresponding cycle. Payments are hence quite crops during the 1991±93 agricultural cycles.
dierent across households, but exogenous to Under PROCAMPO, eligible farmers must go
current behavior. The only restriction is that at each agricultural cycle to one of more than
land must currently be used in crops, livestock, 700 CADER (Centros de Apoyo al Desarrollo
or forestry, or be part of an approved envi- Rural) oces around the country and solicit
ronmental program as opposed to being left their PROCAMPO payments. The maximum
idle, with freedom to choose among these quantity of land for which they may receive
options. transfers is equal to or less that the area they
PROCAMPO is a cash transfer program of had registered in 1994. Payments are, in most
signi®cant magnitude. Since its inception in cases, received as checks distributed at the
1994, it has covered on average 14 million CADER oces. PROCAMPO quali®cation
hectares a year, including more than 95% of the certi®cates can also be used as collateral against
area that had been planted in corn, beans, which to borrow from commercial banks or
sorghum, and wheat. Payments are made to input retailers, giving bene®ciaries ¯exibility in
approximately three million producers a year, the timing when cash is available to them
for a total expenditure in 1998 of US$919 mil- against the cost of the interest charged.
lion (SAGAR, 1998). 1 The compensatory
payments are regressively distributed in the
farm sector as they are proportional to the area 3. THE EJIDO SECTOR AND THE DATA
that had been planted in these crops. The 45%
producers with farms smaller that ®ve hectares In this paper, we analyze the impact of the
thus receive only 10% of the total PRO- PROCAMPO program on households in the
CAMPO transfer (SAGAR, 1998). Transfers ejido sector. The ejido sector was the product of
are, however, progressively distributed on a per the sweeping land reform that followed the
hectare basis since they are uniform per hect- peasant-led revolution of 1910. It contains ap-
are, unrelated to the yields that were achieved proximately 60% of the Mexican rural popu-
and to whether households were selling basic lation, half of the country's agricultural land,
crops or not before NAFTA, and hence were to and half of its irrigated land (Lamartine Yates,
be negatively aected by the expected decline in 1981). In terms of social welfare, it is a major
prices or not. Transfers thus reach producers reservoir of rural poverty and an important
who had never bene®ted from pre-NAFTA source of migrants to the United States. This
price support programs due to lack of marketed sector has been aected by important reforms
surplus (Martinez, 1999). In 1997, transfers since 1990 (DeWalt & Rees, 1994). They in-
represented on average US$329 per recipient clude both global reforms aecting the context
and US$68 per hectare. This represents 46% of where ejidatario households operate (trade lib-
the gross maize income for a farmer obtaining eralization and NAFTA; generalized descaling
the average yield of 1.06 tons/ha and the aver- of subsidies) and reforms directly targeted at
age price of US$140 per ton observed in the the sector (introduction of individual property
ejido. These payments were to remain constant rights over land plots formerly in usufruct;
in real terms for the ®rst 10 years, then phased descaling of ocial credit, marketing, and
1046 WORLD DEVELOPMENT
technical assistance services provided to the households in the survey had received PRO-
ejido by specialized state agencies; devolution CAMPO transfers in 1997. Transfers reach
of control over ejido aairs to the community; equally small and large landholders, but are
and greater freedoms for individual ejidatarios received more by medium farmers and by
in making decisions about income strategies). households with more migration assets (de®ned
The data we use are derived from a nation- as the number of migrants to the United States
wide panel survey of ejido communities and among the brothers and sisters of the house-
ejidatario households within these communi- hold head). Dierences also emerge across re-
ties. The data were collected in 1994 by the gions. The lower shares of households receiving
Mexican Ministry of Agrarian Reform and the PROCAMPO in the North-Paci®c and in the
University of California at Berkeley (see de South are due to historical cropping patterns
Janvry et al., 1997) and in 1997 by the Mexican outside the nine basic crops covered by the
Ministry of Agrarian Reform and the World program.
Bank (see Cord et al., 1998). The data charac- The direct value of PROCAMPO transfers
terize resource use and income formation by represents on average almost 8% of 1997 in-
households. The sample with complete panel come for all households in the survey. While
information on income includes 958 house- PROCAMPO transfers show some regressivity
holds. with respect to farm size, relative importance of
The vast majority of ejido households have transfers is reversed in the other asset endow-
access to PROCAMPO and coverage is quite ments, representing a higher share of income
homogenous across categories of households. for the indigenous population than for nonin-
As the data in Table 1 show, over 86% of the digenous households, and for households with
low levels of labor, education, and migration States increased due to the double incentive of
assets. Note that these values underestimate the poor agriculture and labor market conditions
total eect of PROCAMPO on incomes as they in Mexico enhancing migration and a sharp
neglect the indirect eect of PROCAMPO depreciation of the real exchange rate with the
transfers achieved through the income multi- US dollar that drastically increased the pur-
pliers that we will analyze below. During the chasing power of dollar remittances.
period under study, total household income In what follows, we analyze the income
increased by 14%. Hence, direct PROCAMPO multiplier eects of the PROCAMPO transfers.
transfers represent more than 60% of the reg- We do this by identifying the determinants of
istered increase in income. PROCAMPO income change during 1994±97. We measure
transfers served as an important compensating the multiplier eect of PROCAMPO on total
mechanism for the larger land holders and for household income using several alternative
households with low labor, education, and econometric speci®cations. We then calculate
migration assets for whom the observed change this multiplier for speci®c subsets of the popu-
in income was less than the PROCAMPO lation and for each income source. Finally, we
transfer. Direct PROCAMPO transfers, how- track the origins of the PROCAMPO multi-
ever, fell short of compensating for the drop in plier in agriculture by analyzing how PRO-
income in the North-Paci®c region where agri- CAMPO has induced greater use of purchased
culture is more techni®ed and diversi®ed. chemical inputs.
The data in Table 2 show the structure of
household income by source for 1994 and 1997.
There are several remarkable facts to be noted.
One is that, even though all households are 4. ESTIMATING PROCAMPO INCOME
landed, the share of total income that derives MULTIPLIERS
from nonfarm activities is very high, and it rose
from 47% to 55% during the period analyzed. (a) The econometric procedure
This increase is in part due to the PROCAMPO
program, that did not exist in 1994, and in As for any program impact evaluation, it is
1997, provided on average 7.7% of total crucial to properly control for biases that could
household income and 14% of nonfarm in- come from endogenous participation in the
come. In the period, income from agriculture PROCAMPO program. In this case, not only is
declined due to adverse price incentives and PROCAMPO participation not random but,
wage income stagnated as unemployment in for a participant, the magnitude of the cash
Mexico rose with the aftermath of the peso transfer is directly determined by the house-
crisis. By contrast, self-employment income hold's historical behavior in its choice of
rose and remittance income from the United cropping patterns.
1048 WORLD DEVELOPMENT
Consider the following income equation that proportional to the area cultivated in nine basic
relates income y t in year t 1997 to the crops in 1993. The cropping pattern in 1993 is a
household's asset endowment and characteris- household decision taken jointly with all other
tics zt , the level of PROCAMPO transfer re- decisions that determine the income of the
ceived P t , the eect of unobservables /t , and a household. It is therefore function of the
random eect et : characteristics and the assets of the household
in 1993 (z93 ), and of the unobservables /93 .
97 97
yi97 z97 97 97
i b aPi /i ei ;
1 Hence, any correlation between unobservables
in 1993 and 1997 would create a selection bias
where b97 is the vector of marginal return to the in the estimation of the PROCAMPO multi-
assets in 1997 and a the PROCAMPO income plier in a simple cross-section analysis.
multiplier. Unobservables include both house- To eliminate this potential bias, we write the
hold characteristics such as managerial ability income equation in dierence using the panel
or land quality and external factors such as data for 1994 and 1997. Since there was no
local conditions or government programs other PROCAMPO program in 1994, dierencing
than PROCAMPO. The standard problem in gives:
estimating the PROCAMPO multiplier with 97 94 97
such an equation is the potential bias on a yi97 yi94 z97
i b z94 97
i b aPi
/i /94
i
created by a correlation between any unob-
e97
i e94
i :
2
servable and the PROCAMPO transfer P.
In the impact assessment literature, the po- The dierence in unobservables
/97
i /94
i
tential sources of correlation are classi®ed un- only captures unobserved events that have oc-
der headings referred to as program placement curred during 1994±97, while the PROCAMPO
and household selection (Pitt, Rosenzweig, & transfers are determined on the basis of the
Gibbons, 1993; Ravallion, 1999). A program information available in 1993. Can there be any
placement bias may occur if there is any sys- correlation between the dierence in unob-
tematic geographical bias in the way the servables and P?
PROCAMPO program reaches the population. The timing of the decision ensures that
The survey data con®rm that this is not the PROCAMPO transfers cannot be in¯uenced by
case. All ejidos except 11 have been reached by these unobserved events. Yet the opposite may
PROCAMPO, and for eight among these 11 happen: as PROCAMPO transfers have been
ejidos, the reason for not receiving PRO- extended since 1994, some changes in house-
CAMPO transfers was noneligibility of the in- hold assets may result from past PROCAMPO
dividual households. Hence only three ejidos, transfers. For the analysis to be valid, all pro-
comprising 11 households (i.e., 1.2% of the ductive assets that could have been accumu-
sample), may have been aected by lack of lated with PROCAMPO transfers between
access due to unequal PROCAMPO reach. 1994 and 1997 must consequently be observed
A household selection bias occurs if the in- and not remain among the unobserved / fac-
dividual participation to the program or the tor. An important such asset particularly sen-
amount of transfer received is correlated to sitive to windfall income is livestock. It is
unobserved characteristics, be it through the therefore essential that livestock assets at the
explicit rules of the program itself or from self- beginning of the survey years be included in the
selection of households that do not participate set of asset variables z. Incomplete information
despite their eligibility. We can look at the on livestock raises some problems in that re-
reasons given by households for not partici- spect. We will return to this point after we
pating when asked by enumerators. This shows present the basic estimation results.
that 45% of the nonparticipants say they do not To summarize, we estimate the change in
qualify for PROCAMPO transfers, 10% that income equation:
the transaction is too cumbersome to be worth yi97 yi94 z97 97
z94 94 97
i b i b aPi gi ;
3
the cost, and 22% that they did not know about
the program. One clearly cannot assume that where g combines dierence in unobservables
lack of knowledge of the program or com- and error term, and is uncorrelated with z97 ; z94 ,
plaining about its functioning is not correlated and P 97 .
with determinants of income. This self-selection Note that as the environment and notably
is thus a potential source of bias. As for the prices have changed between the years of the
eligibility rule, PROCAMPO transfers are two surveys, we do not assume constant returns
CASH TRANSFER PROGRAMS 1049
to assets and let the coecients vary with the related, they are likely to be correlated across
year. For the household characteristics that are observations from the same geographical area.
invariant over time, only the dierence between The third is a more standard problem of het-
the two parameters b97 and b94 is estimated. eroskedasticity as the volatility of income is
For the characteristics that changed over the directly related to the agricultural income itself
period, in response to structural changes or to and hence likely to land assets. To address these
shocks, the estimation of two separate param- potential problems, we estimate the income
eters allows to distinguish the eect due to the equation model with Robust Regression and
change in the return to any asset z as captured Least Absolute Deviations (or median) esti-
by the dierence between the parameters and mators. The Robust Regression screens out or
the eect due to the change in asset position discounts outliers by weighting observations. It
(Oaxaca, 1994). is an iterative process in which the calculation
of weights is based on the absolute residuals of
(b) The overall PROCAMPO multiplier the previous iteration. The LAD estimator does
not assume any speci®c distribution of the re-
The set of z variables includes land (irrigated siduals g and gives consistent estimates even in
and rainfed land, pasture, and the household the presence of heteroscedasticity and nonin-
share of common property land), livestock, dependent residuals. LAD estimators are also
human capital assets (gender and age of less sensitive to outliers than ordinary least
household head, number and average educa- squares (OLS) because they minimize the de-
tion of adults), and social assets (Mexico and viations around the median rather than the
US migration assets, ethnicity, and access to square of the deviations around the mean.
technical assistance and to formal credit). While LAD estimators seem to perform well in
A household's migration assets characterize large samples, the standard deviations of the
both the historical migration and the current parameters are, however, usually large for
permanent migration of household members. small samples, which is our case. Hence, results
The construction of this variable is based on from the two estimation techniques have dis-
the information common to the two surveys. 2 tinct advantages and inconvenients, and should
Historical migration is measured by the num- be looked at as mutually reinforcing.
ber of household members who had migrated The results from the Robust Regression es-
earlier but had returned home at least two years timation of Eqn. (3), reported in Table 3, show
prior to the survey. The current permanent that household income is importantly deter-
migration is measured by the number of chil- mined by irrigated and rainfed land assets,
dren of the household head who are perma- number of adults, and access to technical as-
nently established away from home. Note that sistance, and in 1997 by US migration assets
no members of the household that are currently and adult education. Note that, as conditions
in temporary migration are included in these for agricultural production deteriorated in 1997,
assets. This is because current temporary mi- land assets have lost importance in income
gration is a household decision jointly taken determination compared to 1994. In contrast,
with all the other choices that contribute to the human capital assets and migration assets,
formation of income. which are both sources of o-farm income,
Among ejidatarios, access to technical assis- have gained in importance. Geographically, the
tance and to formal credit are essentially supply region that bene®ted most during the period is
determined and hence both considered exoge- the Gulf and the region that did worse is the
nous to the household decision. In addition to North-Paci®c.
these assets, regional eects are added to con- Access to cash transfers through PRO-
trol for geographical characteristics such as CAMPO creates positive externalities on in-
land quality, weather, and local level of eco- come change, with a one peso transfer inducing
nomic development. a direct increase of 1.97 pesos as estimated with
An important issue with agricultural house- Robust Regression. The corresponding 95%
hold income is its extreme volatility, due to con®dence interval is [1.5±2.6]. Hence, the
large ¯uctuations in weather conditions. This marginal income eect of a one peso income
creates several econometric problems. The ®rst transfer through PROCAMPO on bene®ciary
is the presence of a large number of observa- households is high. This is associated with
tions that are clear outliers. The second is the PROCAMPO helping relax the liquidity con-
fact that, since these ¯uctuations are weather straint on farm households.
1050 WORLD DEVELOPMENT
PROCAMPO transfer
PROCAMPO transfer (pesos) 947 2.1 0.00
Goodness-of-®t
Number of observations 956
F(30, 925) 10.10 0.00
*
Signi®cantly dierent from the 1994 parameter at 95%.
To check on the robustness of the estimation with a 95% con®dence interval of [0±4.5]. We
of PROCAMPO multipliers, we use several also use a robust regression where PRO-
alternative econometric speci®cations: median CAMPO is the only exogenous determinant of
regression and OLS. The ®rst yields a multiplier the observed change in household income,
of 2.20 with a 95% con®dence interval of yielding a multiplier of 1.82 with a 95% con®-
[1.3±3.1] and the second a multiplier of 2.24 dence interval of [1.3±2.3]. The remarkable
CASH TRANSFER PROGRAMS 1051
similarity of these values across econometric dowments, nor the best endowed households
estimators and speci®cations gives con®dence who may face less severe liquidity constraints,
in the robustness of the multiplier eects esti- would bene®t as much as a group intermediate
mated for the PROCAMPO transfers. between the two. This is exactly what the
PROCAMPO transfers can be used not only multiplier by farm size indicates (classes are
to acquire inputs, thus generating short-run de®ned on land use in 1994, as land use in 1997
multipliers, but also to invest in productive is endogenous): the multiplier is 0.24 on smaller
assets, including health, that generate long-run farms, 2.77 on medium farms, and 2.04 on the
multipliers. Observable assets such as livestock, larger farms. In terms of human capital assets,
that generate income in 1997, may have in part one sees that the multiplier is higher for
been acquired with PROCAMPO transfers households with a smaller labor force (2.75 vs.
from the current year. In this case, the esti- 0.93), since the liquidity constraint is more
mated equation would under-estimate the im- binding on them as they could engage less in
pact of PROCAMPO by not attributing to the compensatory activities that serve as sources of
program part of the role of these assets in in- liquidity. The multiplier is also higher for
come generation. We are consequently on the households with higher levels of education
safe side regarding a possible overestimation of (1.60 vs. 1.25), although the dierence in pa-
the multiplier. 3 rameters is not statistically signi®cant. It may
Finally, PROCAMPO can also produce in- very well be that the role of education is of little
terhousehold multiplier eects by raising ag- importance in the short-run when transfers are
gregate demand in the region, both for principally spent on acquiring additional
nontradables and for labor. This multiplier amounts of traditional farm inputs as opposed
should be larger where the aggregate regional to new technologies and the introduction of
transfer is higher. PROCAMPO multiplier ef- new activities that may follow in the longer run,
fects would thus be composed of two elements, giving greater value to the role of education in
one obtained through the transfer to the par- multiplying cash transfers. The PROCAMPO
ticular individual, which we have measured to eect is independent of the presence of migra-
this point, and another through a regional ef- tion assets indicating that households with re-
fect. We test this by including the aggregate mittances are not subject to liquidity
state-level transfer in the regression equation in constraints. Finally, the PROCAMPO income
Table 3. When we do this, the individual multiplier is lower for indigenous households
PROCAMPO multiplier is 2.05 with a 95% (0.19 vs. 2.27 for nonindigenous), and for
interval [1.5±2.6], while the state level PRO- households living in the North and North-Pa-
CAMPO multiplier is not signi®cant. This does ci®c (in both cases not signi®cantly dierent
not mean that neighborhood eects may not be from zero). These multipliers reveal the shadow
important, only that they do not apply at the income value of liquidity for the corresponding
geographical scale for which we have data, category of recipient households, re¯ecting in
namely the state. In any case, the individual each case the particular marginal cost of un-
multiplier we measure can only underestimate captured opportunities due to constraining li-
the total multiplier eect of PROCAMPO, quidity. Results thus reveal that the greatest
leaving us again on the safe side regarding absolute income payo (pesos of income per
possible overestimation. pesos of transfer) from relaxing liquidity con-
straints is among medium and large farmers,
(c) Which households are more eective in families with a small number of adults and no
generating income from PROCAMPO transfers migration assets, nonindigenous households,
and regionally the Gulf and the Center.
We can identify who in the heterogeneous
ejido population was able to derive greater ad-
vantage from the PROCAMPO transfers by 5. PROCAMPO TRANSFERS AND THE
comparing the income multiplier across subsets INTENSIFICATION OF AGRICULTURE
of the population. Results are presented in
Table 4. One expects that the multiplier should The above results on PROCAMPO multi-
be greater when a household has more assets pliers by farm size show that transfers can be
and when they are more underused due to productively used in agriculture. This is clearly
greater liquidity constraint. This suggests that con®rmed by farmers' responses to questions
neither the households with very low asset en- about use of their PROCAMPO receipts. In the
1052 WORLD DEVELOPMENT
survey, 70% of the households responded that however, that transfers are not sucient to in-
they use the PROCAMPO money to purchase duce changes in cropping patterns or in area
inputs. In a larger survey done by the Ministry planted. To con®rm this direct information, we
of Agriculture (SAGAR, 1998), 44% of the re- analyze in what follows the changes in agri-
spondents said that PROCAMPO transfers al- cultural income and in input use induced by
lowed them to increase their input purchases PROCAMPO transfers using the 1994 and
and another 17% to start using purchased in- 1997 panel data.
puts. This happened despite the fact that Results are presented in Table 5. Although
PROCAMPO transfers have often arrived late explaining the very volatile agricultural income
in the season (93% say that they arrive after the is dicult, results indicate a positive multiplier
promised date). While farmers would clearly eect in agriculture, with every peso of PRO-
bene®t more from receiving PROCAMPO at CAMPO transfer generating, at mean value,
the time they purchase their inputs, many of 0.33 pesos in agricultural income. As can be
them purchase inputs either by collateralizing seen from the interaction terms with land, this
their PROCAMPO rights, or directly by eect is obtained through ownership of
obtaining credit from suppliers on the basis of irrigated land. It is also largely in¯uenced by
the forthcoming transfers. Respondents say, the availability of complementary technical
CASH TRANSFER PROGRAMS 1053
Table 5. Eect of PROCAMPO on agricultural activities, partial results (partial results: coecients on assets not
reported)
Coecient P-value
Change in agricultural income (Robust regression)
PROCAMPO transfer 0.19 0.31
PROCAMPO*irrigated area 0.23 0.00
PROCAMPO*rainfed area )0.01 0.10
PROCAMPO*technical assistance 1.44 0.00
PROCAMPO*access to credit )0.99 0.00
PROCAMPO at mean value of exogenous variables 0.33 0.02
PROCAMPO at mean value, but no credit 0.51 0.00
PROCAMPO at mean value, but no technical assistance 0.23 0.10
Change in use of chemicals (Probit random eect)
PROCAMPO transfer 0.00016 0.02
Average value without PROCAMPOa 48.6%
Average value with PROCAMPOa 52.5%
Change in livestock income (Tobit random eect)
PROCAMPO transfer 0.28 0.07
a
Average expected probability calculated by sample enumeration.
assistance. Setting technical assistance to zero transfers to income maintenance is the sum of
would reduce the income multiplier from 0.33 the monetary value of the transfers (direct ef-
to 0.23. The interactive term with credit con- fects) and the income generated through indi-
®rms the role of PROCAMPO as a substitute rect eects measured by multipliers in excess of
for credit. Setting access to credit to zero one. We can use the estimated multipliers in
(rather than at the mean value of 0.18) would Table 4 to predict what would have been the
increase the multiplier from 0.33 to 0.51. change in income during 1994±97 had there be
Hence, it is those households who control more no PROCAMPO program. This is done in
irrigated land, have access to technical assis- Table 6. For all households, the observed in-
tance, and no access to credit that are able to come change is 14.2%. The direct cash transfer
take greater advantage of the cash transferred represents an increase of 8.7% over 1994 in-
in generating more income. Use of chemicals as come. With a multiplier of 2.06, the indirect
estimated by a random eect probit equation eect is a contribution to income of 9.3 % over
also shows a signi®cant positive response to 1994 income. Hence, the total PROCAMPO
PROCAMPO transfers. We do not have in- contribution is an increase in income of 18%
formation on the amount used in 1994 and over the 1994 level. Had there been no PRO-
hence cannot measure the eect reported in CAMPO program, household income would on
interviews of increased applications for those average have declined by 3.9%. Contrasting this
who already used some chemicals in 1994. result with the data in Table 1 shows the im-
Livestock income also responds to PRO- portance of accounting for the indirect income
CAMPO transfers, with every peso of transfer eects of cash transfers in assessing the impact
generating 0.28 pesos of livestock income. Note of such programs. In Table 1, it looked as
that this estimation does not include the use of though PROCAMPO transfers would not have
PROCAMPO transfers to purchase livestock. been necessary for income maintenance since
It therefore likely underestimates the long-term other sources of income would have sustained
eect of PROCAMPO on livestock income. an income increase of 5.4%. In fact, these other
incomes were themselves part and parcel of the
hidden multiplier eects of PROCAMPO
6. THE TRUE CONTRIBUTION OF transfers.
PROCAMPO TO INCOME If they had not put the cash transfers to
MAINTENANCE work, many categories of households would
have suered income declines in spite of the
As indicated in Section 3 of this paper, the appearance of rising income after discounting
true contribution of the PROCAMPO cash the direct transfers. This applies to medium
1054 WORLD DEVELOPMENT
farmers, nonindigenous households, and not what PROCAMPO is since the transfer is
households in the Center region. There are ®ve a single ®xed rate per hectare, the same for all
instances where transfers were not used to producers, including for producers who have
generate more income: smallholders, household never sold any products on the market and
with a large number of adults, indigenous are hence not aected by falling prices. If it
households, and households in the North and were a welfare program, it should be targeted
North-Paci®c, all of which were found to have at the poor, or at least at small producers.
income multipliers of less than one. In spite of This is not the case since transfers are pro-
this, however, all households were made better portional to area, and hence disproportion-
o by the cash transfers compared to the in- ately received by large landowners. If it were
come levels they would have achieved without an eciency-enhancing program, intent on
the transfers. 4 facilitating the adjustment of agriculture to
trade liberalization, it would need to be ac-
companied by technical assistance and by
7. CONCLUSION initiatives to set up institutions in support of
the competitiveness of agriculture. This is not
Based on what we have seen, how can we the case since the program is not bundled with
characterize what type of program PRO- other interventions. So, what type of program
CAMPO is relative to established categories of is it? Basically a hybrid between compensa-
development programs? If it were a program tion, welfare, and adjustment, de®ned in the
to compensate for trade liberalization, trans- context of a political window of opportunity
fers should be function of past production to compensate the losers from trade liberal-
levels (hence of yields) and, even better, of ization, but with a de®nite welfare twist to-
sales (hence of marketed surpluses). This is ward the rural poor.
CASH TRANSFER PROGRAMS 1055
We have seen that the indirect eects of cash troduction of new activities (see SAGAR, 1998,
transfer programs have received little attention. for a similar observation). In correspondence
Yet, we found that these eects can be highly with the new set of incentives introduced by
signi®cant and that they should consequently NAFTA, transfers should instead be used for
deserve full consideration in the design of such the modernization of agriculture and its diver-
programs. We showed, in particular, that si®cation toward high value activities with
PROCAMPO created large indirect eects comparative advantage. That this is not hap-
among ejidatarios through multiplication of the pening is not surprising. In the current context
liquidity received. The multiplier for all of declining institutional support to agriculture,
households is in the range 1.5±2.6. Multipliers only 18% of the ejidatarios have access to for-
are higher for households with medium and mal credit, 13% to Alianza para el Campo, the
large farms, low numbers of adults in the main public program in support of rural de-
household, no migration assets, nonindigenous velopment, and 7% to technical assistance.
backgrounds, and located in the Center and Multiplier eects could thus be signi®cantly
Gulf regions. Large multipliers re¯ect uncap- increased if PROCAMPO were accompanied
tured marginal income opportunities due to li- by a serious eort at institutional reconstruc-
quidity constraints that are relaxed by the tion and technological change in support of the
transfers. Opportunities come from the asset modernization and diversi®cation of ejido ag-
endowments that these households have, par- riculture.
ticularly irrigated land, and they are enhanced The second is that high multipliers show
by access to technical assistance. Liquidity capacity to borrow even at high interest rates.
constraints derive from incomplete property Willingness to pay for liquidity is somewhat
rights in the ejido sector, and from the current overrepresented by the magnitude of the
disarray of ®nancial institutions servicing agri- multipliers since they need to be discounted
culture following descaling of the agricultural for the risk of borrowing, which is not
development bank implied by structural ad- present with gifted money. The results show,
justment. Large multipliers thus re¯ect sizeable however, that there is a clear unmet need for
gaps between opportunities and constraints. liquidity that can be productively invested,
Households with migrants sending remittances and that ejidatarios can pay for this service at
and with higher levels of education may thus levels of interest rates that are quite compat-
have lower multipliers because they were able ible with current commercial rates. This
to work around the liquidity constraints more shows the high payo that exists from con-
eectively than other households. Households structing an alternative set of ®nancial insti-
with little land and with ethnic backgrounds tutions able to replace the parastatals that
may have lower access to liquidity, but also previously serviced the sector. If land is to
have lower opportunities to invest additional serve as collateral in accessing loans in these
cash received, again resulting in lower multi- ®nancial institutions, then the current titling
pliers. program should have a high payo. Titling
There are two policy implications that derive without access to credit will, however, not
from this analysis. change the current situation. PROCAMPO
The ®rst is that, if multiplier eects are im- multipliers help reveal the shadow value of
portant and policy responsive (as suggested by liquidity in every category of potential bor-
heterogeneity of multipliers across households), rowers. They consequently provide a metric to
then the PROCAMPO program would gain identify where the eective demand for ®-
from being managed as part of a comprehen- nancial services is the greatest.
sive eort to maximize these multipliers since The magnitude of the PROCAMPO multi-
the ultimate goal is to raise the income of tar- pliers should be taken as proof that the ejido
geted households. This can be done by intro- sector is not to be discounted as a lively sector
ducing complementary rural development for investment and growth. To prevent this
initiatives that increase opportunities to use the potential from going to waste, what urgently
transfers productively. We observed that addi- needs to be done is to complement transfers
tional liquidity serves principally to increase the with investment opportunities in new com-
use of current inputs. Transfers thus enhance modities and new technologies, and to recon-
the level of traditional activities. We found no struct the institutions that service the sector,
evidence of technological change or of the in- most particularly for accessing liquidity.
1056 WORLD DEVELOPMENT
NOTES
1. PROCAMPO is supported by a loan from the Inter- ing for the role of these assets in the estimation of
American Bank. income eects would overestimate the multiplier eect of
the current PROCAMPO transfers. If this process of
2. A more complete speci®cation of migration assets, accumulation continues in the future, however, the eect
including both family and social networks, was used and of the current transfer on the income of the next two
shown to be important in the decision of a household to years is likely to be the same as that of past transfers on
send migrants and hence in receiving remittances (Win- current income, canceling the overestimation eect of
ters, de Janvry, & Sadoulet, 2000). But lack of compa- past investments.
rable data on the extended family in 1994 and 1997
forced us to reduce the variable to the family members in 4. The only negative multiplier, for households in the
this analysis. North-Paci®c, is not signi®cantly dierent from zero.
REFERENCES
Behrman, J. (1996). The impact of health and nutrition Martinez, J.-C. (1999). Programa de apoyo al PRO-
on education. The World Bank Research Observer, CAMPO. Mimeo, Washington, DC: Inter-American
11(1), 23±27. Development Bank.
Cord, L. et al. (1998). Mexico ejido reform: avenues of Oaxaca, R. (1994). On discrimination and the decom-
adjustment ®ve years later. Latin America and position of wage dierentials. Journal of Economet-
Caribbean Region. Washington, DC: The World rics, 61(1), 5±21.
Bank. Pitt, M., Rosenzweig, M., & Gibbons, D. (1993). The
Cox, D., & Jimenez, E. (1992). Social security and determinants and consequences of the placement of
private transfers in a developing country: the case of programs in Indonesia. World Bank Economic Re-
Peru. The World Bank Economic Review, 6(1), 155± view, 7(3), 319±348.
169. Ravallion, M. (1999). The mystery of the vanishing
Cox, D., Eser, Z., & Jimenez, E. (1998). Motives for bene®ts: Ms speedy analyst's introduction to evalua-
private transfers over the life cycle: an analytical tion. Policy Research Working Paper No. 2153.
frame work and evidence from Peru. Journal of Washington, DC: The World Bank.
Development Economics, 55(1), 57±81. SAGAR (Ministry of Agriculture, Livestock, and Rural
de Janvry, A., Sadoulet, E., & Santos, B. (1995). Project Development). (1998). PROCAMPO, 1994±1998.
evaluation for sustainable rural development: Plan Claridades Agropecuarias, 64(December), 1±40.
Sierra in the Dominican Republic. Journal of Envi- Sahn, D., & Alderman, H. (1995). The eect of food
ronmental Economics and Management, 28(2), 135± subsidies on labor supply in Sri Lanka. In: D. van de
154. Walle, & K. Nead (Eds.), Public spending and the poor.
de Janvry, A., Gordillo, G., & Sadoulet, E. (1997). Baltimore: The Johns Hopkins University Press.
Mexico's second agrarian reform: Household and Subbarao, K. et al. (1997). Safety net programs and
community responses. La Jolla: Center for US± poverty reduction: Lessons from cross-country experi-
Mexican Studies, University of California at San ences. Washington, DC: The World Bank.
Diego. Subbarao, K. (1997). Public works as an anti-poverty
DeWalt, B., & Rees, M. (1994). The end of agrarian program: an overview of cross-country experiences.
reform in Mexico: Past lessons, future prospects. La American Journal of Agricultural Economics, 79(2),
Jolla: Center for US±Mexican Studies, University of 678±683.
California at San Diego. Winters, P., de Janvry, A., & Sadoulet, E. (2000). Family
Grosh, M. (1994). Administering targeted social pro- and community networks in Mexico±US migration.
grams in Latin America: from platitudes to practice. Journal of Human Resources, 36(1), 159±184.
Washington, DC: The World Bank. Wydick, B. (1999). The eect of microenterprise lending
LamartineYates, P. (1981). Mexico's agricultural di- on child schooling in Guatemala. Economic Devel-
lemma. Tucson: University of Arizona Press. opment and Cultural Change, 47(4), 853±869.