Directory UMM :Data Elmu:jurnal:UVW:World Development:Vol29.Issue3.2001:
World Development Vol. 29, No. 3, pp. 561±572, 2001
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0305-750X/01/$ - see front matter
www.elsevier.com/locate/worlddev
PII: S0305-750X(00)00108-X
The Determinants of Nonfarm Activities
and Incomes of Rural Households in Mexico,
with Emphasis on Education
ANTONIO YUNEZ-NAUDE
El Colegio de M
exico, Mexico
and
J. EDWARD TAYLOR *
University of California at Davis, USA
Summary. Ð This paper presents the main results of a study of the eects of education (as well as
other household assets) on the choice of activities and incomes of rural Mexican households. Our
study examines the various income sources, as well as the education of the household's head and its
members. Implications are drawn for rural education and development policies to promote rural
nonfarm incomes and employment. Ó 2001 Elsevier Science Ltd. All rights reserved.
Key words Ð development, farm and nonfarm activities, subsistence and commercial crops,
education, migration, diversi®cation, selectivity
1. INTRODUCTION
The economic reforms applied by the governments of Latin America in the recent past
have highlighted the need to develop human
capital in the region. Education is crucial to
raising economic productivity and competitiveness and to combating poverty. The issue is
especially pertinent in Mexico, due to the
opening of trade with its neighbors to the north
as well as with other countries with strong
economies, and to its extensive poverty and
income inequality and its poor record in education when judged by international and regional standards (see Singh & Santiago, 1997).
Mexico's education problems are worst in its
rural areas where poverty is concentrated. In
their study on the determinants of poverty and
inequality in Latin America, Attanasio and
Szekely (1999) estimate that the rural sector
accounts for 12.2% of poverty and in Mexico,
for 30.2% (the highest share among the 14
countries included in the calculations). They
found that dierences in education (years of
schooling) accounted for 28.6% of Latin
American poverty; while the ®gures for Mexico
561
and Chile are the highest among the countries,
46.9% and 47.8%, respectively.
Moreover, a fundamental characteristic of
rural households is diversi®cation of income.
This is especially true in countries at an intermediate level of development such as Mexico,
where there are dual agricultural sectors. Rural
households in these situations continue to produce staples for home consumption and earn
incomes from other sources (such as production
of cash crops and nonfarm activities). This is
due to their poverty and risk aversion as well as
to missing or failed markets for staple foods,
factors, and credit. 1 Recent development literature tends to depict income diversi®cation
into nonfarm sources as favorable to development, and education as contributing to diversi®cation by rural households in developing
* We are grateful for funding from Ford, Hewlett, and
McKnight Foundations and comments from the guest
editors as well as three anonymous reviewers. We are
also thankful for the support we received from Eric van
Dusen, George Dyer, X
ochitl Ju
arez, Angel Pita, and
Luis Gabriel Rojas in this research.
562
WORLD DEVELOPMENT
countries (for reviews of literature and evidence
see Ellis, 1998; Reardon, 1997; Reardon &
Stamoulis, 1998).
Given that study of the role of education in
the rural economy is critical for economic development in Mexico and that fact has been
recognized by recent Mexican governments,
and given the importance of nonfarm income
diversi®cation to Mexican rural households, it
is surprising that there is a dearth of empirical
research on the returns to education for rural
households active in both the nonfarm and the
farm sectors. To contribute to ®lling this gap,
we undertook research on impacts of education
on rural Mexican household incomes and activity choice in the farm and nonfarm sectors.
We de®ne ``rural'' as population concentrations
of less than 5000. In a previous article we discussed in detail the methodology used in this
research (Taylor & Y
unez-Naude, 2000).
The present study extends the latter analysis
by presenting data and regressions concerning
the impacts of dierent levels of schooling (as
well as other household assets such as migration) on household participation in nonfarm
and farm activities and on incomes from those
activities. We disaggregate by education of
dierent household members, but our unit of
labor allocation and income analysis is the
household. This is because in poor rural communities of Mexico, it is at the household
rather than the individual level that decisions
are made concerning family labor allocation to
farming, nonfarm activities, schooling, and so
on. The data come from a survey of rural
households in eight rural areas of Mexico.
We proceed as follows. In Section 2, we
present our model and place our study in the
context of recent literature. In Section 3 we
describe the education levels and other socioeconomic characteristics of the sample households and their communities. In Section 4 we
present econometric results. Section 5 concludes the paper.
2. THE MODEL
The empirical evidence concerning returns to
education in rural areas does not support unambiguously the generally accepted argument
that education spurs development. For example, Phillips (1987) criticizes the conclusion of
Jamison and Lau (1982) that the results of 37
studies indicate that on average farm productivity increases 8.7% when farmers complete
four years of primary education. Phillips's critique was that many of the studies showed
nonsigni®cant and even negative impacts of
education on production and the net income of
certain crops. Even in recent evidence one ®nds
a mix of results, with some studies showing
positive and signi®cant returns to education
and others the contrary. 2
We hypothesize that researchers in this domain can err should they not take into account
the technological change and household income
diversi®cation that characterizes rural transformation in developing countries. That is, selectivity and activity choice is frequently
ignored in agricultural economics literature on
returns to education. 3 Studies that focus on
one crop or activity ignore the self-selection by
households into speci®c activities (or out of
them).
Moreover, it is common for extant studies to
only include the schooling of the household
head, ignoring the eects of education of other
household members on production or incomes
of the household. Yet current schooling as well
as accumulated years of schooling of the various household members can in¯uence household investment allocations to the various
activities, as well as returns to those investments per activity. To ignore the endogeneity of
activity selection creates selectivity bias and to
omit household education variables leads to
underidenti®cation bias in the estimated parameters.
From an analytical perspective, activity
choice by rural households is equivalent to
technology choice. Farm households can realize
the bene®ts of education by dropping an activity, for example, traditional agriculture, in
which the returns to education are low, and
taking up another activity, such as modern
agriculture or wage employment, where the
payo to education is higher.
In the regression speci®cation we take into
account two aspects of the peasant economy.
The ®rst is that household income is the sum of
net incomes coming from various activities,
with the possibility that the returns to schooling
dier over activities. The second is that most
households do not receive incomes from all the
activities. Thus, in the model we take into account that, due to diversi®cation, household
income from a particular activity depends on
whether the household participates in the activity and on the net income it reports receiving
from it conditioned on its participation. The
expected income from a given activity is the
MEXICO
product of the probability of participation and
the expected income, conditioned on participation. Potentially that probability as well as
the expected income are in¯uenced by education and other variables, that thus ®gure in our
model.
That many households do not earn any income from certain activities can create selectivity bias. For example, the households that
participate in migration may have a comparative advantage relative to others in this activity.
This means that only using the subsample that
participates in a given activity would produce
biased results. To avoid the latter, we include
data from all households surveyed. To correct
for selectivity bias, we use a Probit model in
which the dependent variable is a 0/1 variable
for participation, and the regressors are the
variables that aect net incomes to these activities. The coecients estimated with these
k 1; . . . ; K Probits are used to test for the
eects of schooling (and other variables) on
participation in the various activities. The results of the Probits are used to correct for selectivity bias in the estimation of the returns to
schooling and other variables in the activity
income equations.
The selection of households who participate
or do not participate in a particular activity is
not random. Thus, the returns to schooling
(and other variables) are estimated based on the
households who chose to participate in some of
the activities is not representative of returns to
education for all the households in the sample.
The procedure to correct for this potential selectivity bias is a generalization of the Lee twostage estimator proposed by Amemiya (see
Taylor & Y
unez-Naude, 2000) in which we
present a version of the formal model used in
the regressions the results of which are presented below. 4
3. HOUSEHOLD CHARACTERISTICS
The data come from a survey of 391 households (with 2,960 members) between 1992 and
1995. The data cover household characteristics,
assets, the elements to calculate net incomes
and labor time allocation to the principal economic activities, expenditures, commercial and
noncommercial transactions, as well as the location of household activities as to whether
they are local, regional, national, or international.
563
The households were selected at random in
eight villages in four dierent states of Mexico:
Coahuila, Jalisco, Michoac
an, and Puebla (see
Figure 1). The choice of these eight communities is motivated by the desire to re¯ect the
socioeconomic diversity of Mexican small villages and rural households in order to be representative in a general way of small rural
producers in Mexico.
Concordia, a village in the northern state of
Coahuila, produces maize for the market, and
its inhabitants are very involved in nonrural
local and regional labor markets; we will call
Concordia the ``wage employees community.''
The residents of El Chante, located in the
central state of Jalisco, are mainly engaged in
cash cropping (especially of sugar cane), but
they also produce maize for own consumption
and for the market. We call El Chante the
``commercial agricultural community.'' Erongarõcuaro, Napõzaro, Urichu, and Pu
acuaro
together form the Municipality of Erongarõcuaro in the central state of Michoac
an. The
households of this zone produce maize for own
consumption, livestock, undertake o-farm
activities, and a large proportion migrate to the
United States. We call these four villages the
``migrants community.'' Finally, Naupan and
Reyesoghpan are villages populated by indigenous populations in the Sierra Norte of the
Center-East state of Puebla. The households in
this area are very poor, migrate little to the
United States, produce maize for home consumption and several cash crops, chiles in
Naupan and coee in Reyesoghpan cafe. We
call this are the ``indigenous community.'' See
Table 1 for the list and characteristics of these
communities.
The eight rural towns are closely linked to
the nonrural economy, as 60% of the incomes
of their households come from nonfarm activities. Most of the latter are generated by wage
employment in local and regional labor markets, in villages, rural towns, and intermediate
cities.
The characteristics of the sample households
and villages re¯ect the situation in rural Mexico. The eight villages studied each have less
than 5,000 inhabitants, which is about the average in rural Mexico. Our sample includes
both small private farmers as well as ejidatarios,
farmers in the communal sector. 5 As in the rest
of rural Mexico, the survey households produce
staples (maize) and cash crops, earn nonfarm
wage incomes as an important part of their
incomes, and are involved in migration in
564
WORLD DEVELOPMENT
Figure 1. Survey villages in Coahuila, Jalisco, Michoacan and Puebla.
Mexico and to the US, as well as in nonfarm
self employment (in manufactures and services,
including commerce). The average size of the
maize farms in Mexico is about 2.5 ha, a bit
above the average of 1.9 ha in our sample. 6
The average schooling of the sample household members (4.36 years) is close to the national rural average (4.3 years) The share of
individuals without education is also similar
(22% in rural Mexico vs 25% in the sample) as
is the share with only primary education, up to
six years of schooling (53.6% for rural Mexico
vs 55.5% for our sample), and with secondary
education, up to 12 years of schooling. The
only notable dierence between our sample and
national rural averages pertains to those with
higher education, more than 12 years of
schooling, where the national rural average is
2% while the sample's is 1%. The discrepancy
may be due to the villages in the sample being
small and lacking schools beyond the 12th
year level (see Bracho & Zamudio, 1994, and
Table 2).
4. DETERMINANTS OF RURAL
NONFARM INCOMES AND
PARTICIPATION
The dependent variables used in the regressions are the net incomes for each of six activities commonly undertaken by rural households
in Mexico: production of staple crops; production of cash crops and livestock; nonfarm
self-employment; wage employment in local
and regional labor markets; migratory wage
employment in Mexico and in the United
States. 7
The explanatory variables include: years of
schooling of the head of household and the
MEXICO
565
Table 1. Activity income levels and shares (1994 pesos and shares)a
All
communities
Commercial farming communityb
Wage employees
communityc
Migrants
communityd
Indigenous
communitye
Incime per c
apita
activitiesf
Commercial farming
Subsistence staples
Nonfarmg
International
remittances
National remittances
2,066
6,596
2,715
1,660
1,220
9.6
5.5
64.9
16.2
42.9
1.3
46.7
0.0
4.0
0.0
7.4
3.7
9.1
Total
100.0
100.0
100.0
100.0
100.0
23.4
4.8
58.7
9.1
52.0
2.6
42.7
2.7
0.5
7.2
81.9
3.0
a
Source: Authors' estimates.
El Chante, Jalisco.
c
Concordia, Coahuila.
d
Erongaricuaro, Napizaro, Puacuaro and Uricho, Michoacan.
e
Naupan and Reyesoghpan, Puebla.
f
Constant pesos of 1994.
g
Includes local and regional wage employment (the bulk of nonfarm income), commerce, and manufacture of
construction materials and handicrafts.
b
Table 2. Sample statisticsa
Dependent variables
(for all households)
Incomes
Total
Agriculture
Commercial crops and livestock
Staples
Nonfarm
Wage employment
Migration remittances
International
National
Community subsample
Commercial farming community
Wage employees community
Migrants community
Indigenous Community
Median
S.D.
15,866
20,187
3,712
758
3,599
5,716
11,811
1,779
12,032
9,376
1,446
636
4,893
2,110
12.5%
15.3%
50.0%
22.2%
Independent variables
(per household)
Education
Median
From 1 to 3 years
From 4 to 6 years
From 7 to 9 years
More than 9 years
Of the household head
Number of migrants
To the US
To elsewhere in Mexico
Size of the household
Age of the household
head
Cultivated landb
Livestockc
Median
S.D.
4.36
1.74
2.36
1.17
0.45
4.01
2.14
1.76
2.17
1.54
0.90
3.88
1.14
0.80
7.31
41.68
2.03
1.40
3.59
16.77
3.11
4,344
6.45
13,020
a
Source: Authors' calculations.
Constant 1994 pesos.
c
Hectares.
b
household members; experience of the head of
household (ageless schooling less ®ve); national
and international migration network of the
household, a variable that re¯ects accumulated
capital from migration, de®ned by the number
of immediate family members who are migrants
who were in migration at the beginning of the
survey year; family assets (farm size and value
of livestock holdings); other characteristics of
the household that can in¯uence the decision to
participate in various activities and thus that
determine incomes (family size and age of the
household head).
(a) Determinants of participation
The results of the Probit regressions are
presented in Table 3, which shows percentage
changes in probabilities of participation associated with a unit change in the explanatory
variables. A key pattern that emerges is the
positive relation between primary education
566
WORLD DEVELOPMENT
Table 3. Probit regressions for participation in activitiesa; b;c
Equations/activities
Variable
Farm sector
From 1 to 3 years of
schooling
From 4 to 6 years
schooling
From 7 to 9 years of
schooling
More than 9 years
of schooling
Education of the
household head
Experience of the
household head
Family members in
the US
Family members
elsewhere in Mexico
Landholdings
Family size
Value of Livestock
holdings
Wage employees
communityd
Indigenous community
Commercial farming community
Constant
Staples
Cash crops
and
livestock
5.62
(1.74)
3,84
(1.31)
)0.48
(0.15)
)0.16
(0.34)
)1.53
(1.15)
0.46
(1.30, 1.11)
2.55
(1.07)
)3.0
(1.05)
5.96
(5.74)
)2,45
(1.21)
0.42
(1.18)
)18.15
(1.81)
30.67
(3.84)
)21.65
(1.79)
)16.11
(0.97)
)0.16
(0.05)
)0.82
(0.26)
)0.99
(0.30)
0.35
(0.08)
0.36
(0.27)
0.38
(0.30,0.23)
1.64
(0.74)
5)1.54
(0.54)
3.42
(3.07)
1.40
(0.57)
3.860.39
(4.28)
9.43
(0.93)
18.64
(5.94)
5.10
(0.37)
)13.38
(1.34)
Nonfarm
self-employment
Wage
employment
4.00
(1.45)
4.76
(1.95)
5.44
(1.99)
)0.76
(0.17)
)0.35
(0.29)
0.55
(1.17, 0.76)
)7.27
(4.02)
)2.22
(0.93)
0.01
(0.01)
)1.09
(0.56)
0.3
(1.35)
)34.06
(4.32)
)2.78
(0.31)
)2.42
(0.12)
)13.93
(1.50)
2.72
(0.94)
6.03
(2.47)
7.32
(2.54)
5.24
(1.22)
)0.84
(0.65)
)0.15
(2.08 , 1.62 )
)4.26
(2.32)
)2.41
(0.99)
)1.8
(1.97)
)1.2
(0.59)
)0.62
(0.91)
25.99
(2.51)
19.62
(2.27)
)17.7
(1.59)
7.1
(2.06)
Migration
International
National
)4.00
(1.07)
3.41
(1.00)
3.34
(0.15)
)0.82
(0.91)
1.04
(0.15)
0.02
(1.96 , 1.57 )
13.18
(6.26)
)1.7
(0.62)
1.01
(1.14)
0.14
(0.05)
0.68
(0.14)
)18.71
(1.52)
)12.69
(0.02)
)14.38
(1.03)
0.09
(2.10)
4.12
(1.28)
1.66
(0.54)
)0.42
(0.13)
)2.09
(0.42)
)1.53
(0.94)
)0.54
(0.62, 0.87)
)4.01
(1.59)
16.89
(7.26)
)2.5
(2.02)
0.96
(0.41)
(1.58)
27.83
(2.76)
15.47
(1.54)
)25.57
(1.32)
)9.38
(1.63)
a
The numbers in the table are percentages changes in the estimated probabilities associated with a unit change in the
explanatory variable, evaluated at the medians of the rest of the variables in the Probit. For the regional dummy
variables, the table reports the dierence in the estimated probabilities between D 1 and D 0, evaluated based on
the medians of the rest of the variables with the exception of the other regional dummy variables, which are set at 0.
b
Notes: The t ratios are between parentheses and those appearing below the experience variable correspond to the
variable itself and to the squared term, respectively.
c
Source: Authors' calculations.
d
Fixed eects. The migrants community is used as the default case. That applies to the other tables as well.
*
Signi®cance level 0.10.
**
Signi®cance level 0.05.
(from one to six years of schooling) and secondary education (seven to nine years of
schooling) and the likelihood of participation
both in nonfarm self-employment and wage
employment. The only exception is that one
additional member with incomplete primary
education (one to three years of schooling) is
associated with a signi®cant and positive
probability of participation in basic staples
production (5.62%, with a t 1:74). This is
because those members of school age who do
not ®nish primary school (with at most three
years of schooling) have no better alternative
than to engage in a traditional activity such as
maize production.
Another result that is noteworthy is the
positive association between incomplete primary education (from one to three years of
schooling), complete primary education (four
to six years), complete secondary education
(seven to nine years of schooling), and the
likelihood that the household participates in
MEXICO
nonfarm activities (4% with t 1:45, 4.8%,
with t 1:95, and 5.4% with t 1:99, respectively). To interpret the ®ndings, note that in
small villages in Mexico, most of the manufactures sector is constituted by production of
simple construction materials and handicrafts;
the other main nonfarm activity is commerce
(undertaken by adults). Keeping in mind these
facts, the ®ndings can be explained as follows.
It is common for girls and young women to
produce handicrafts, and the young who have
not ®nished primary schooling who have no
better alternatives who work in these activities.
A similar argument can be made regarding
brick-making, as it is common for boys and
young men who have not completed primary
education to help the adults in this activity.
An additional member of the family with
complete primary education or complete secondary education is associated with a positive
and signi®cant probability of participating in
the wage-labor market (6% and 7.3%, respectively). An additional member with more than
nine years of schooling has a positive eect
(5.2%) on the probability of participation in the
wage-labor market, although the eect is not
signi®cant at the 10% level (t 1:22). These
results agree with those of Evans and Ngau
(1991) for a rural zone in Kenya, as their regressions show that more household education
(beyond ®ve years of education) means greater
nonfarm income.
An increase of one year of education of the
household head does not have signi®cant eects
on the participation of the household in any of
the activities. This result varies from that of
Evans and Ngau, and may be due to the fact
that the household heads are relatively old, as
many of the young have migrated out of the
small villages.
Schooling does not have signi®cant eects on
the probability of migration to other parts of
Mexico or to the United States. The result for
national migration may be due to the stagnation of urban labor demand in Mexico in the
past decade. But our results concerning international migration vary from those of other
studies (see, for example, the Mexico±United
States Binational Migration Study, 1998).
Nevertheless, if we consider average years of
education of the members of the household, the
eect on the probability of migration to the
United States becomes positive and signi®cant
(see Taylor & Y
unez-Naude, 2000).
An additional year of experience of the
household head reduces the household's par-
567
ticipation in wage employment and spurs its
participation in migration; however, despite the
statistical signi®cance of the eect, the percentage eects are slight, 0.15% and 0.02%,
respectively.
With respect to the other family assets, the
results are as expected. For example, as in other
studies (Massey, Arango, Kouaouci, Pellegrinoy, & Taylor, 1999), we ®nd that family contacts strongly explain international migration.
An increase of one unit in family contacts in the
United States. reduces household participation
in local, regional, and national nonfarm selfemployment and wage employment. Moreover,
an additional hectare of land spurs signi®cantly
household participation in farming, and reduces participation in labor markets and migration to the rest of Mexico. An increase in
livestock holdings spurs participation in cash
cropping and in nonfarm activities.
Finally, taking into account that the ``migrants community'' is the reference point for
comparisons of ®xed eects, it is not surprising
that residence in the ``wage employees community'' increases by nearly 26% the probability (t 2:51) that the households participate in
local and regional labor markets, that the
probability of participation in the migration
labor market in the United States is negative
( 19%) and that of participation in migration
to the rest of Mexico is positive (28%).
(b) Determinants of income levels by activity
Table 4 shows the results of the regression of
income sources on family assets. We made the
calculations based on a system of net income
equations correcting for selectivity, and the
®gures in the table show the absolute eects of
a unit change in the explanatory variables on
incomes per activity, given the households
participation in the activity.
The estimated returns to basic education (one
to three years) are positive in income to maize
production incomes and remittance incomes
from migrants within Mexico. Controlling for
other variables, an additional member with
basic education of one to three years, relative to
one with no schooling is associated with an increase of 210 pesos in income from production
of basic staples (about US$ 70, with t 2:48).
An additional member with basic education of
one to three years increases by 96 pesos household income from national migration remittances. These results show high returns to basic
568
WORLD DEVELOPMENT
Table 4. Eects of education and other variables on net income per activity (in 1994 pesos)a ;b
Equations/activities
Variable
Farm sector
From 1 to 3 years of
schooling
From 4 to 6 years of
schooling
From 7 to 9 years of
schooling
More than 9 years of
schooling
Education of the
household head
Experience of the
household head
Family members in
the US
Family members in
the rest of Mexico
Landholdings
Family size
Value of livestock
holdings (1000)
Wage employees
community
Indigenous
community
Commercial farming
community
Inverse Mills Ratio
Constant
R2d
Nonfarm
self-employment
Wage-employment
Staples
Cash crops
and livestock
210
(2.48)
199
(2.57)
156
(1.89)
260
(2.12)
9
(0.24)
)32
(1.09)
)
)303
(0.65)
)251
(0.59)
)183
(0.40)
794
(1.12)
372
(1.86)
217
(1.37)
)
486
(0.86)
126
(0.24)
695
(1.25)
)667
(0.81)
227
(0.93)
)130
(0.67)
)
)
)
)
4.4
(0.31)
)127
(2.07)
)
538
(5.36)
305
(0.90)
0
(3.36)
)1,481
(0.91)
3,833
(2.62)
14,384
(7.50)
1,543
(2.06)
)8,196
(1.95)
0.44
)
285
(0.97)
)356
(1.34)
213
(0.62)
826
(6.21)
898
(1.18)
0.18
Migration
International
National
)200
(0.52)
)438
(1.20)
27
(0.07)
2,334
(4.22)
579
(3.50)
)9
(0.06)
)
)38
(0.26)
351
(2.92)
)72
(0.43)
)116
(0.42)
50
(0.56)
)10
(0.13)
863
96
(1.49)
)45
(0.88)
9
(0.13)
)228
(1.87)
7
(0.18)
22
(0.68)
)
)
(6.27)
595
)
(7.55)
292
(0.71)
)
458
(1.65)
)
)
)
)
)
)
)2,559
(1.30)
)1,813
(1.02)
5,002
(2.20)
5,806
(7.37)
2,782
(0.54)
0.19
2,420
(1.83)
)1,834
(1.53)
)226
(0.15)
5,398
(9.69)
1,355
(0.39)
0.40
)344c
(0.60)
)
1,158
(3.66)
358
(1.21)
)116
(0.32)
1,193
(7.64)
)375
(0.47)
0.29
)1,556
(1.79)
2,283
(5.58)
)200
(0.11)
0.28
a
The procedure corrected for selectivity bias in the systems of equations, using Lee's extension of Amemiya estimation method. The ®gures in the table are absolute changes in incomes by activity. The t ratios are between parentheses.
b
Source: Authors' estimates.
c
Includes the indigenous community, whose members do not migrate to the United States.
d
The system R2 0:89, the v2 709:17, the degrees of freedom are 74 and the size of the sample is 328.
*
Signi®cance levels 0.10.
**
Signi®cance levels 0.05.
education in traditional farming and in withinMexico migration.
As with primary education, an additional
member with higher primary education (four to
six years) has positive and signi®cant eects on
household income from production of basic
staples (199 pesos with t 2:57). The dierence
between the two strata of primary education is
that for the higher stratum, an additional
member increases household incomes by the
channel of international migration remittances
(351 pesos, t 2:92) but not of within-Mexico
migration.
Secondary education (from seven to nine
years of schooling) and preparatory, technical,
or secondary education (more than nine years
of schooling) also bring substantial gains in
incomes from basic staples, which also includes
commercial production of these crops. An additional member with more than nine years of
MEXICO
schooling produces a large increase in the salary income of the rural household (of 2,334
pesos or almost US$780, with a t 4:22).
By contrast with the results for years of
schooling of household members, an additional
year of education of the household head does
not produce an increase in basic staples incomes. Nevertheless, and similar to the results
for secondary education and above, one additional year of schooling for the household head
increases incomes from wage employment (559
pesos or US$186, with a t 3:5). Moreover,
increasing the schooling of the household head
generates income gains in commercial agriculture and livestock husbandry (372 pesos with
t 1:86).
An additional year of experience of the
household head only generates gains in incomes
from commercial agriculture and livestock
husbandry (217 pesos with a t 1:37).
Beyond the positive eect of an additional
member migrating to the United States or the
rest of Mexico (863 pesos with t 6:27 and 595
pesos, with t 7:55, respectively), other family
assets have positive eects on incomes. That is
the case with the estimated eect of an additional hectare of landholdings on incomes
coming from commercial agriculture and livestock husbandry, and an additional family
member on wage employment income. By
contrast, the impact of an additional hectare
does not have an eect on incomes from production of staples; note that increasing family
size by one memberÐwithout adding educationÐactually decreases farm income from
staples.
These latter results require further discussion.
The average farm size of farm households in
small villages is quite small (around two hectares) and typically only a part of this land is
dedicated to maize for home consumption. As
farm size increases, the likelihood increases that
the farm produces commercial crops and livestock (see Table 3) and farm incomes from
these sources increase. As for staples, even
though more land increases the probability of
producing these crops, incomes from these
crops do not rise. The latter might be because
maize is raised in small ®elds and usually for
home consumption, and the net income from
this activity is very low or sometimes even
negative. That may also explain why income
from production of staples declines as family
size increases (remember that Table 3 shows
that an additional family member does not increase the likelihood that the household pro-
569
duces staples). That is, given land constraints
and the use maize for home consumption, an
additional family member reduces income from
production of staples. Although family size
does not aect the probability of household
participation in wage employment, it is probable that an additional family member raises
family labor time in the labor market. That is
re¯ected in the positive impact of family size in
wage income, in turn implying that family labor
is diverted from production of basic grains into
wage employment.
The results arising from comparison of
communities are as expected. For example, and
recalling that the ``migrant community'' is the
reference point for comparisons, it is not surprising that income received from nonmigratory labor markets by the ``wage employees
community'' households are much higher than
those received by households in the ``migrants
community.''
The Inverse Mills Ratio (IMR) is signi®cant
for all the activities. This indicates the importance of self-selection of households in participation in a given activity and in determining
their incomes in that activity. The latter is due
to the fact that the variables that aect choice
of activity, via the selectivity variables, also
aect household incomes from speci®c activities.
(c) Returns to education in total incomes of
households
We ran an additional regression using ordinary least squares (OLS) of the Mincerian type
in order to complete our study of returns to
education, as well as to compare our results
with the only other study of which we are aware
concerning the eects of education in rural areas of Mexico. The dependent variable is the
logarithm of household total net income, and
the explanatory variables are the same as we
used in the regressions in Table 5.
Controlling for the other variables, the estimated eects of primary and secondary education are not statistically signi®cant. By
contrast, the number of family members with
more than nine years of schooling is associated
with a substantial increase in total household
income (11.8% or US$670 in 1994). The same
occurs in the case of household head education,
a one-year increase in which increases household income by 7.4% (around US$400). Besides
schooling, other assets increase household income; these include the number of family
570
WORLD DEVELOPMENT
Table 5. Total income results for the OLS regressions
Variable
The estimated eect on the log of total net income
(Mincerian form)
From 1 to 3 years of schooling
From 4 to 6 years of schooling
From 7 to 9 years of schooling
More than 9 years of schooling
Education of the household head
Experience of the household head
Experience squared
Family members in the US
Family members in the rest of Mexico
Value of livestock holdings (1000)
Landholdings
Family size
Wage employees community
Indigenous community
Cash crop community
Constant
Sample size
R2
*
0.057
0.014
0.054
0.118
0.074
0.016
)0.027
0.106
)0.133
0.207
)0.003
0.052
0.360
)0.151
0.459
7.720
391
0.252
Signi®cance level of 0.10.
Signi®cance level of 0.05.
**
members in the United States (10.6%) and the
value of livestock holdings (20.7%). Finally,
the experience of the household head has the
highest eect in the ``commercial agriculture
community'' and the ``wage employees community.''
Our results concerning the returns to
schooling of household heads are very similar
to those estimated by Singh and Santiago
(1997) in their study on returns to education in
farm activities in a rural region of Mexico.
According to their calculations, the returns are
between 9% (for the head of household) and 8%
(for the wife).
5. CONCLUSIONS
Our study shows that education and years
of schooling aect activity choice of rural
households. Similar to other analyses in
Mexico, our results support the argument that
the returns to education in rural incomes are
high and statistically signi®cant, independent
of the level of schooling. Our study allows
elaboration on the details of schooling eects
as follows.
Primary, secondary, and preparatory education have positive eects on income from basic
grains for those who produce them, indicating
positive returns to schooling in Mexican traditional agriculture. Poverty and persistent market failures make it so that small farmers
continue to produce maize and beans for home
consumption and to manage risk, incorporating new knowledge through skills acquired in
school. At higher levels of schooling, it is to be
expected that returns to education for staples
production will be higher in commercial farming as compared to subsistence farming. Our
analysis indeed shows that post-primary
schooling has positive and signi®cant impacts
on commercial farm incomes.
The nonsigni®cant returns shown by primary
and secondary education on family nonfarm
enterprise incomes in the local area or in the
region may be due to two factors: (a) the content of public education (universal and standardized over the entire country) does not
provide young farmers with the necessary
knowledge to incorporate improved technologies in production of cash crops, livestock, and
nonfarm products and services; and (b) that
that level of education does not lead to increased wages. The former is the argument
made by Reardon and Schejtman (1999) and
the second is made by Attanasio and Szekely
(1999). According to the latter, the returns to
education in Mexican wage incomes only begin
to grow when they go beyond secondary education. Their estimates indicate that, during
1986±96, the returns to primary and secondary
education in wage employment actually diminished, and that the returns to advanced
schooling (beyond nine years of schooling) rose
considerably during the same period (from
MEXICO
190% to 243%). This result is similar to ours, in
that we ®nd that returns to advanced schooling
are high for wage employment.
The latter result, combined with a similar
result concerning the positive eects of education of the household head on wage income,
suggests that the establishment of manufactures
®rms in rural areas will increase household incomes. At the same time, the high returns to
preparatory and technical education in local
and regional labor markets indicates that this
type of education increases labor productivity
in the nonrural economy of villages and small
cities. This implies that development policies in
Mexico should not only focus on primary
education (as has been the case to present) but
also strive to increase the access of the rural
young to schooling beyond the nine-year level.
In general, and despite the insistence by
Mexican politicians of the importance of
education, eorts in that direction are still insucient. For example, faced with the recommendation of UNESCO that investment in
571
education in developing countries should constitute at least 8% of GNP, until 1998 the
highest percentage achieved by Mexico was only
5% (Romero Hicks, 2000; Singh & Santiago,
1997). Within this gloomy panorama, the rural
areas are in the worst situation, as there one
®nds the highest rates of illiteracy, the lowest
average education, the least years of schooling
and the least access to post-primary education
(Bracho, 1998).
Mexico should make a priority of rural education, not only with more investment, but
also by making its present investment more
eective and ecient. Our results suggest that
rural education needs to incorporate the
teaching of practical and technical skills that
are needed in the rural context, to make rural
activities more productive. In that sense, the
federalization of education put in place in 1992
may provide an incentive to state and municipal authorities to include in their education
plans the perspective of rural reality so that
rural education performs better.
NOTES
1. In Mexico, most entrepreneurs live in urban centers.
2. The studies that report positive impacts of education
include those of Yang (1997) for agreggate agricultural
value added in China, and Jollie (1996) for incomes in
Ghana, and Jacoby (1991) for own-farm and livestock
husbandry incomes in Peru, and Singh and Santiago
(1997) for farming activities in Mexico. Examples of
studies that report negative or nonsigni®cant impacts of
education include Adams (1995) for aggregate gross
value added in the production of wheat, sugar cane, and
rice in Pakistan; that of Rosegrant and Evenson (1992),
which estimated total factor productivity in India, and
that of Adams (1993) for total household incomes
excluding migration remittances in Egypt. See details in
Taylor and Y
unez-Naude (2000).
3. An exception is the study of Jollie cited above.
There is another group of studies that treat the theme of
portfolio diversifcation in rural households, but their
objective is not to evaluate the eects of education on
rural productivity and incomes. We refer to those that
analyze the factors that determine diversi®cation (among
which education) as well as the eects of diversi®cation
(see below and the review in Ellis, 1998).
4. It should also be mentioned that the data also make
it possible to run regressions of the determinants of
diversi®cation of sources of incomes by type of household, as is done in Leones and Feldman (1998) for the
Philippines and other authors for various regions of
developing countries (see a review in Ellis, 1998). The
data base is available from the network PRECESAM,
whose website is http://www.colmex.mx/centros/cee/
precesam/precesam.htm
5. We do not make a distinction between them
because, with the exception of land property rights,
the characteristics of ejido producers are very similar
to private smallholders. Moreover, with the ejido
reform of 1991 the land property rights of ejido
producers are being transformed, as the reformed
made possible sale and purchase of ejidal land, and
with that change, the characteristics of ejido producers
are converging yet further with those of private
producers.
6. The dierence is expected because our sample only
includes small producers and in Mexico there are also
medium farmers that produce basic grains (see, for
example, Hernandez Estrada, 2000 and Y
unez-Naude &
Guevara, 1998).
7. The net income from each farm sector activity
includes the implicit income earned from subsistence
production imputed at local prices.
572
WORLD DEVELOPMENT
REFERENCES
Attanasio, O., & Szekely, M. (1999). Introducci
on: La
pobreza en America Latina. Analisis basado en los
activos. El Trimestre Economico, LXVI (3), 263,
317±364.
Adams, R. H. (1993). The economic and demographic
determinants of international migration in rural
Egypt. Journal of Development Studies, 30(1),
146±167.
Adams, R. H. (1995). Agricultural income, cash crops,
and inequality in rural Pakistan. Economic Development and Cultural Change, 43(3), 467±491.
Bracho, T. (1998). Mexico. Per®l educativo de sus
adultos y tendencias de escolarizaci
on de sus nisimnos. Unprocessed, Colegio de Mexico.
Bracho, T., & Zamudio, A. (1994). Los rendimientos
economicos de la escolaridad en Mexico, 1989.
Economõa Mexicana, 3(2), 345±377.
Ellis, F. (1998). Household strategies and rural livelihood diversi®cation. Journal of Development Studies,
35(1), 1±38.
Evans, H. E., & Ngau, P. (1991). Rural±urban relations,
household income diversi®cation and agricultural
productivity. Development and Change, 22, 519±545.
Jacoby, H. G. (1991). Productivity of men and women
and the sexual division of labor in peasant agriculture of the Peruvian Sierra. Journal of Development
Economics, 37(1/2), 265±287.
Jamison, D. T., & Lau, L. J. (1982). Farmer education
and farm eciency. Baltimore, MD: Johns Hopkins
University Press.
Jollie, D. (1996) The impact of education in Rural
Ghana: Examining productivity and labor allocation
eects. Princeton University and World Bank, Unprocessed.
Hern
andez Estrada, M. I. (2000). Tipologõa de productores agropecuarios. In A. Y
unez-Naude (Ed.), Los
Peque~s Productores Rurales: Las Reformas y las
Opciones. Mexico: El Colegio de Mexico.
Leones, J. P., & Feldman, S. (1998). Nonfarm activity
and rural household income: evidence from Philippine microdata. Economic Development and Cultural
Change, 46(4), 789±806.
Massey, D. S., Arango, J., Kouaouci, A., Pellegrinoy,
A., & Taylor, J. E. (1999). Worlds in motion:
International migration at century's end. Oxford:
Oxford University Press.
Mexico±United States Binational Migration Study.
(1998). Migration between Mexico and the United
States. Austin, Texas: Morgan Printing.
Phillips, J. M. (1987). A comment on farmer education
and farm eciency: A survey. Economic Development
and Cultural Change, 35(3), 637±641.
Reardon, T. (1997). Using evidence of household income
diversi®cation to inform study of the rural nonfarm
labor market in Africa. World Development, 25(5),
735±747.
Reardon, T., Rello F., Schejtman. A., & Stamoulis, K.
(1998). Agroindustrialization in intermediate cities of
Latin America: Hypotheses regarding employment
eects on the rural poor. Paper presented at the
IFPRI Conference, Strategies for Stimulating
Growth of the Rural Nonfarm Economy in
Developing Countries, 17±21 May. Airlie House,
Virginia.
Reardon, T., & Schejtman, A. (1999). Los proyectos de
alivio de la pobreza rural desde la perspectiva del
empleo: una relectura del proyecto PROLESUR.
Unprocessed, United Nations Food and Agricultural
Organization, Reional Oce for Latin America and
the Caribbean, Santiago, Chile.
Romero Hicks, J. L. (2000). Descentralizaci
on educativa. V
ortice, 1, 149±154.
Rosegrant, M. W., & Evenson, R. E. (1992). Agricultural productivity and sources of growth in South
Asia. American Journal of Agricultural Economics,
74(3), 757±761.
Singh, R., & Santiago, M. (1997). Farm earnings,
educational attainment, and the role of public policy:
some evidence from Mexico. World Development,
25(12), 2143.
Taylor, J. E., & Y
unez-Naude, A. (2000). Selectivity and
the returns to schooling in a diversi®ed rural economy. American Journal of Agricultural Economics,
287±297.
Yang, D. T. (1997). Education and o-farm work.
Economic Development and Cultural Change, 45(3),
613±632.
Y
unez-Naude, A., Guevara, A. (1998). Evaluacion de los
programas de desarrollo regional en el plano comunitario. Reporte de Resultados. Mexico: El Colegio de
Mexico.
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0305-750X/01/$ - see front matter
www.elsevier.com/locate/worlddev
PII: S0305-750X(00)00108-X
The Determinants of Nonfarm Activities
and Incomes of Rural Households in Mexico,
with Emphasis on Education
ANTONIO YUNEZ-NAUDE
El Colegio de M
exico, Mexico
and
J. EDWARD TAYLOR *
University of California at Davis, USA
Summary. Ð This paper presents the main results of a study of the eects of education (as well as
other household assets) on the choice of activities and incomes of rural Mexican households. Our
study examines the various income sources, as well as the education of the household's head and its
members. Implications are drawn for rural education and development policies to promote rural
nonfarm incomes and employment. Ó 2001 Elsevier Science Ltd. All rights reserved.
Key words Ð development, farm and nonfarm activities, subsistence and commercial crops,
education, migration, diversi®cation, selectivity
1. INTRODUCTION
The economic reforms applied by the governments of Latin America in the recent past
have highlighted the need to develop human
capital in the region. Education is crucial to
raising economic productivity and competitiveness and to combating poverty. The issue is
especially pertinent in Mexico, due to the
opening of trade with its neighbors to the north
as well as with other countries with strong
economies, and to its extensive poverty and
income inequality and its poor record in education when judged by international and regional standards (see Singh & Santiago, 1997).
Mexico's education problems are worst in its
rural areas where poverty is concentrated. In
their study on the determinants of poverty and
inequality in Latin America, Attanasio and
Szekely (1999) estimate that the rural sector
accounts for 12.2% of poverty and in Mexico,
for 30.2% (the highest share among the 14
countries included in the calculations). They
found that dierences in education (years of
schooling) accounted for 28.6% of Latin
American poverty; while the ®gures for Mexico
561
and Chile are the highest among the countries,
46.9% and 47.8%, respectively.
Moreover, a fundamental characteristic of
rural households is diversi®cation of income.
This is especially true in countries at an intermediate level of development such as Mexico,
where there are dual agricultural sectors. Rural
households in these situations continue to produce staples for home consumption and earn
incomes from other sources (such as production
of cash crops and nonfarm activities). This is
due to their poverty and risk aversion as well as
to missing or failed markets for staple foods,
factors, and credit. 1 Recent development literature tends to depict income diversi®cation
into nonfarm sources as favorable to development, and education as contributing to diversi®cation by rural households in developing
* We are grateful for funding from Ford, Hewlett, and
McKnight Foundations and comments from the guest
editors as well as three anonymous reviewers. We are
also thankful for the support we received from Eric van
Dusen, George Dyer, X
ochitl Ju
arez, Angel Pita, and
Luis Gabriel Rojas in this research.
562
WORLD DEVELOPMENT
countries (for reviews of literature and evidence
see Ellis, 1998; Reardon, 1997; Reardon &
Stamoulis, 1998).
Given that study of the role of education in
the rural economy is critical for economic development in Mexico and that fact has been
recognized by recent Mexican governments,
and given the importance of nonfarm income
diversi®cation to Mexican rural households, it
is surprising that there is a dearth of empirical
research on the returns to education for rural
households active in both the nonfarm and the
farm sectors. To contribute to ®lling this gap,
we undertook research on impacts of education
on rural Mexican household incomes and activity choice in the farm and nonfarm sectors.
We de®ne ``rural'' as population concentrations
of less than 5000. In a previous article we discussed in detail the methodology used in this
research (Taylor & Y
unez-Naude, 2000).
The present study extends the latter analysis
by presenting data and regressions concerning
the impacts of dierent levels of schooling (as
well as other household assets such as migration) on household participation in nonfarm
and farm activities and on incomes from those
activities. We disaggregate by education of
dierent household members, but our unit of
labor allocation and income analysis is the
household. This is because in poor rural communities of Mexico, it is at the household
rather than the individual level that decisions
are made concerning family labor allocation to
farming, nonfarm activities, schooling, and so
on. The data come from a survey of rural
households in eight rural areas of Mexico.
We proceed as follows. In Section 2, we
present our model and place our study in the
context of recent literature. In Section 3 we
describe the education levels and other socioeconomic characteristics of the sample households and their communities. In Section 4 we
present econometric results. Section 5 concludes the paper.
2. THE MODEL
The empirical evidence concerning returns to
education in rural areas does not support unambiguously the generally accepted argument
that education spurs development. For example, Phillips (1987) criticizes the conclusion of
Jamison and Lau (1982) that the results of 37
studies indicate that on average farm productivity increases 8.7% when farmers complete
four years of primary education. Phillips's critique was that many of the studies showed
nonsigni®cant and even negative impacts of
education on production and the net income of
certain crops. Even in recent evidence one ®nds
a mix of results, with some studies showing
positive and signi®cant returns to education
and others the contrary. 2
We hypothesize that researchers in this domain can err should they not take into account
the technological change and household income
diversi®cation that characterizes rural transformation in developing countries. That is, selectivity and activity choice is frequently
ignored in agricultural economics literature on
returns to education. 3 Studies that focus on
one crop or activity ignore the self-selection by
households into speci®c activities (or out of
them).
Moreover, it is common for extant studies to
only include the schooling of the household
head, ignoring the eects of education of other
household members on production or incomes
of the household. Yet current schooling as well
as accumulated years of schooling of the various household members can in¯uence household investment allocations to the various
activities, as well as returns to those investments per activity. To ignore the endogeneity of
activity selection creates selectivity bias and to
omit household education variables leads to
underidenti®cation bias in the estimated parameters.
From an analytical perspective, activity
choice by rural households is equivalent to
technology choice. Farm households can realize
the bene®ts of education by dropping an activity, for example, traditional agriculture, in
which the returns to education are low, and
taking up another activity, such as modern
agriculture or wage employment, where the
payo to education is higher.
In the regression speci®cation we take into
account two aspects of the peasant economy.
The ®rst is that household income is the sum of
net incomes coming from various activities,
with the possibility that the returns to schooling
dier over activities. The second is that most
households do not receive incomes from all the
activities. Thus, in the model we take into account that, due to diversi®cation, household
income from a particular activity depends on
whether the household participates in the activity and on the net income it reports receiving
from it conditioned on its participation. The
expected income from a given activity is the
MEXICO
product of the probability of participation and
the expected income, conditioned on participation. Potentially that probability as well as
the expected income are in¯uenced by education and other variables, that thus ®gure in our
model.
That many households do not earn any income from certain activities can create selectivity bias. For example, the households that
participate in migration may have a comparative advantage relative to others in this activity.
This means that only using the subsample that
participates in a given activity would produce
biased results. To avoid the latter, we include
data from all households surveyed. To correct
for selectivity bias, we use a Probit model in
which the dependent variable is a 0/1 variable
for participation, and the regressors are the
variables that aect net incomes to these activities. The coecients estimated with these
k 1; . . . ; K Probits are used to test for the
eects of schooling (and other variables) on
participation in the various activities. The results of the Probits are used to correct for selectivity bias in the estimation of the returns to
schooling and other variables in the activity
income equations.
The selection of households who participate
or do not participate in a particular activity is
not random. Thus, the returns to schooling
(and other variables) are estimated based on the
households who chose to participate in some of
the activities is not representative of returns to
education for all the households in the sample.
The procedure to correct for this potential selectivity bias is a generalization of the Lee twostage estimator proposed by Amemiya (see
Taylor & Y
unez-Naude, 2000) in which we
present a version of the formal model used in
the regressions the results of which are presented below. 4
3. HOUSEHOLD CHARACTERISTICS
The data come from a survey of 391 households (with 2,960 members) between 1992 and
1995. The data cover household characteristics,
assets, the elements to calculate net incomes
and labor time allocation to the principal economic activities, expenditures, commercial and
noncommercial transactions, as well as the location of household activities as to whether
they are local, regional, national, or international.
563
The households were selected at random in
eight villages in four dierent states of Mexico:
Coahuila, Jalisco, Michoac
an, and Puebla (see
Figure 1). The choice of these eight communities is motivated by the desire to re¯ect the
socioeconomic diversity of Mexican small villages and rural households in order to be representative in a general way of small rural
producers in Mexico.
Concordia, a village in the northern state of
Coahuila, produces maize for the market, and
its inhabitants are very involved in nonrural
local and regional labor markets; we will call
Concordia the ``wage employees community.''
The residents of El Chante, located in the
central state of Jalisco, are mainly engaged in
cash cropping (especially of sugar cane), but
they also produce maize for own consumption
and for the market. We call El Chante the
``commercial agricultural community.'' Erongarõcuaro, Napõzaro, Urichu, and Pu
acuaro
together form the Municipality of Erongarõcuaro in the central state of Michoac
an. The
households of this zone produce maize for own
consumption, livestock, undertake o-farm
activities, and a large proportion migrate to the
United States. We call these four villages the
``migrants community.'' Finally, Naupan and
Reyesoghpan are villages populated by indigenous populations in the Sierra Norte of the
Center-East state of Puebla. The households in
this area are very poor, migrate little to the
United States, produce maize for home consumption and several cash crops, chiles in
Naupan and coee in Reyesoghpan cafe. We
call this are the ``indigenous community.'' See
Table 1 for the list and characteristics of these
communities.
The eight rural towns are closely linked to
the nonrural economy, as 60% of the incomes
of their households come from nonfarm activities. Most of the latter are generated by wage
employment in local and regional labor markets, in villages, rural towns, and intermediate
cities.
The characteristics of the sample households
and villages re¯ect the situation in rural Mexico. The eight villages studied each have less
than 5,000 inhabitants, which is about the average in rural Mexico. Our sample includes
both small private farmers as well as ejidatarios,
farmers in the communal sector. 5 As in the rest
of rural Mexico, the survey households produce
staples (maize) and cash crops, earn nonfarm
wage incomes as an important part of their
incomes, and are involved in migration in
564
WORLD DEVELOPMENT
Figure 1. Survey villages in Coahuila, Jalisco, Michoacan and Puebla.
Mexico and to the US, as well as in nonfarm
self employment (in manufactures and services,
including commerce). The average size of the
maize farms in Mexico is about 2.5 ha, a bit
above the average of 1.9 ha in our sample. 6
The average schooling of the sample household members (4.36 years) is close to the national rural average (4.3 years) The share of
individuals without education is also similar
(22% in rural Mexico vs 25% in the sample) as
is the share with only primary education, up to
six years of schooling (53.6% for rural Mexico
vs 55.5% for our sample), and with secondary
education, up to 12 years of schooling. The
only notable dierence between our sample and
national rural averages pertains to those with
higher education, more than 12 years of
schooling, where the national rural average is
2% while the sample's is 1%. The discrepancy
may be due to the villages in the sample being
small and lacking schools beyond the 12th
year level (see Bracho & Zamudio, 1994, and
Table 2).
4. DETERMINANTS OF RURAL
NONFARM INCOMES AND
PARTICIPATION
The dependent variables used in the regressions are the net incomes for each of six activities commonly undertaken by rural households
in Mexico: production of staple crops; production of cash crops and livestock; nonfarm
self-employment; wage employment in local
and regional labor markets; migratory wage
employment in Mexico and in the United
States. 7
The explanatory variables include: years of
schooling of the head of household and the
MEXICO
565
Table 1. Activity income levels and shares (1994 pesos and shares)a
All
communities
Commercial farming communityb
Wage employees
communityc
Migrants
communityd
Indigenous
communitye
Incime per c
apita
activitiesf
Commercial farming
Subsistence staples
Nonfarmg
International
remittances
National remittances
2,066
6,596
2,715
1,660
1,220
9.6
5.5
64.9
16.2
42.9
1.3
46.7
0.0
4.0
0.0
7.4
3.7
9.1
Total
100.0
100.0
100.0
100.0
100.0
23.4
4.8
58.7
9.1
52.0
2.6
42.7
2.7
0.5
7.2
81.9
3.0
a
Source: Authors' estimates.
El Chante, Jalisco.
c
Concordia, Coahuila.
d
Erongaricuaro, Napizaro, Puacuaro and Uricho, Michoacan.
e
Naupan and Reyesoghpan, Puebla.
f
Constant pesos of 1994.
g
Includes local and regional wage employment (the bulk of nonfarm income), commerce, and manufacture of
construction materials and handicrafts.
b
Table 2. Sample statisticsa
Dependent variables
(for all households)
Incomes
Total
Agriculture
Commercial crops and livestock
Staples
Nonfarm
Wage employment
Migration remittances
International
National
Community subsample
Commercial farming community
Wage employees community
Migrants community
Indigenous Community
Median
S.D.
15,866
20,187
3,712
758
3,599
5,716
11,811
1,779
12,032
9,376
1,446
636
4,893
2,110
12.5%
15.3%
50.0%
22.2%
Independent variables
(per household)
Education
Median
From 1 to 3 years
From 4 to 6 years
From 7 to 9 years
More than 9 years
Of the household head
Number of migrants
To the US
To elsewhere in Mexico
Size of the household
Age of the household
head
Cultivated landb
Livestockc
Median
S.D.
4.36
1.74
2.36
1.17
0.45
4.01
2.14
1.76
2.17
1.54
0.90
3.88
1.14
0.80
7.31
41.68
2.03
1.40
3.59
16.77
3.11
4,344
6.45
13,020
a
Source: Authors' calculations.
Constant 1994 pesos.
c
Hectares.
b
household members; experience of the head of
household (ageless schooling less ®ve); national
and international migration network of the
household, a variable that re¯ects accumulated
capital from migration, de®ned by the number
of immediate family members who are migrants
who were in migration at the beginning of the
survey year; family assets (farm size and value
of livestock holdings); other characteristics of
the household that can in¯uence the decision to
participate in various activities and thus that
determine incomes (family size and age of the
household head).
(a) Determinants of participation
The results of the Probit regressions are
presented in Table 3, which shows percentage
changes in probabilities of participation associated with a unit change in the explanatory
variables. A key pattern that emerges is the
positive relation between primary education
566
WORLD DEVELOPMENT
Table 3. Probit regressions for participation in activitiesa; b;c
Equations/activities
Variable
Farm sector
From 1 to 3 years of
schooling
From 4 to 6 years
schooling
From 7 to 9 years of
schooling
More than 9 years
of schooling
Education of the
household head
Experience of the
household head
Family members in
the US
Family members
elsewhere in Mexico
Landholdings
Family size
Value of Livestock
holdings
Wage employees
communityd
Indigenous community
Commercial farming community
Constant
Staples
Cash crops
and
livestock
5.62
(1.74)
3,84
(1.31)
)0.48
(0.15)
)0.16
(0.34)
)1.53
(1.15)
0.46
(1.30, 1.11)
2.55
(1.07)
)3.0
(1.05)
5.96
(5.74)
)2,45
(1.21)
0.42
(1.18)
)18.15
(1.81)
30.67
(3.84)
)21.65
(1.79)
)16.11
(0.97)
)0.16
(0.05)
)0.82
(0.26)
)0.99
(0.30)
0.35
(0.08)
0.36
(0.27)
0.38
(0.30,0.23)
1.64
(0.74)
5)1.54
(0.54)
3.42
(3.07)
1.40
(0.57)
3.860.39
(4.28)
9.43
(0.93)
18.64
(5.94)
5.10
(0.37)
)13.38
(1.34)
Nonfarm
self-employment
Wage
employment
4.00
(1.45)
4.76
(1.95)
5.44
(1.99)
)0.76
(0.17)
)0.35
(0.29)
0.55
(1.17, 0.76)
)7.27
(4.02)
)2.22
(0.93)
0.01
(0.01)
)1.09
(0.56)
0.3
(1.35)
)34.06
(4.32)
)2.78
(0.31)
)2.42
(0.12)
)13.93
(1.50)
2.72
(0.94)
6.03
(2.47)
7.32
(2.54)
5.24
(1.22)
)0.84
(0.65)
)0.15
(2.08 , 1.62 )
)4.26
(2.32)
)2.41
(0.99)
)1.8
(1.97)
)1.2
(0.59)
)0.62
(0.91)
25.99
(2.51)
19.62
(2.27)
)17.7
(1.59)
7.1
(2.06)
Migration
International
National
)4.00
(1.07)
3.41
(1.00)
3.34
(0.15)
)0.82
(0.91)
1.04
(0.15)
0.02
(1.96 , 1.57 )
13.18
(6.26)
)1.7
(0.62)
1.01
(1.14)
0.14
(0.05)
0.68
(0.14)
)18.71
(1.52)
)12.69
(0.02)
)14.38
(1.03)
0.09
(2.10)
4.12
(1.28)
1.66
(0.54)
)0.42
(0.13)
)2.09
(0.42)
)1.53
(0.94)
)0.54
(0.62, 0.87)
)4.01
(1.59)
16.89
(7.26)
)2.5
(2.02)
0.96
(0.41)
(1.58)
27.83
(2.76)
15.47
(1.54)
)25.57
(1.32)
)9.38
(1.63)
a
The numbers in the table are percentages changes in the estimated probabilities associated with a unit change in the
explanatory variable, evaluated at the medians of the rest of the variables in the Probit. For the regional dummy
variables, the table reports the dierence in the estimated probabilities between D 1 and D 0, evaluated based on
the medians of the rest of the variables with the exception of the other regional dummy variables, which are set at 0.
b
Notes: The t ratios are between parentheses and those appearing below the experience variable correspond to the
variable itself and to the squared term, respectively.
c
Source: Authors' calculations.
d
Fixed eects. The migrants community is used as the default case. That applies to the other tables as well.
*
Signi®cance level 0.10.
**
Signi®cance level 0.05.
(from one to six years of schooling) and secondary education (seven to nine years of
schooling) and the likelihood of participation
both in nonfarm self-employment and wage
employment. The only exception is that one
additional member with incomplete primary
education (one to three years of schooling) is
associated with a signi®cant and positive
probability of participation in basic staples
production (5.62%, with a t 1:74). This is
because those members of school age who do
not ®nish primary school (with at most three
years of schooling) have no better alternative
than to engage in a traditional activity such as
maize production.
Another result that is noteworthy is the
positive association between incomplete primary education (from one to three years of
schooling), complete primary education (four
to six years), complete secondary education
(seven to nine years of schooling), and the
likelihood that the household participates in
MEXICO
nonfarm activities (4% with t 1:45, 4.8%,
with t 1:95, and 5.4% with t 1:99, respectively). To interpret the ®ndings, note that in
small villages in Mexico, most of the manufactures sector is constituted by production of
simple construction materials and handicrafts;
the other main nonfarm activity is commerce
(undertaken by adults). Keeping in mind these
facts, the ®ndings can be explained as follows.
It is common for girls and young women to
produce handicrafts, and the young who have
not ®nished primary schooling who have no
better alternatives who work in these activities.
A similar argument can be made regarding
brick-making, as it is common for boys and
young men who have not completed primary
education to help the adults in this activity.
An additional member of the family with
complete primary education or complete secondary education is associated with a positive
and signi®cant probability of participating in
the wage-labor market (6% and 7.3%, respectively). An additional member with more than
nine years of schooling has a positive eect
(5.2%) on the probability of participation in the
wage-labor market, although the eect is not
signi®cant at the 10% level (t 1:22). These
results agree with those of Evans and Ngau
(1991) for a rural zone in Kenya, as their regressions show that more household education
(beyond ®ve years of education) means greater
nonfarm income.
An increase of one year of education of the
household head does not have signi®cant eects
on the participation of the household in any of
the activities. This result varies from that of
Evans and Ngau, and may be due to the fact
that the household heads are relatively old, as
many of the young have migrated out of the
small villages.
Schooling does not have signi®cant eects on
the probability of migration to other parts of
Mexico or to the United States. The result for
national migration may be due to the stagnation of urban labor demand in Mexico in the
past decade. But our results concerning international migration vary from those of other
studies (see, for example, the Mexico±United
States Binational Migration Study, 1998).
Nevertheless, if we consider average years of
education of the members of the household, the
eect on the probability of migration to the
United States becomes positive and signi®cant
(see Taylor & Y
unez-Naude, 2000).
An additional year of experience of the
household head reduces the household's par-
567
ticipation in wage employment and spurs its
participation in migration; however, despite the
statistical signi®cance of the eect, the percentage eects are slight, 0.15% and 0.02%,
respectively.
With respect to the other family assets, the
results are as expected. For example, as in other
studies (Massey, Arango, Kouaouci, Pellegrinoy, & Taylor, 1999), we ®nd that family contacts strongly explain international migration.
An increase of one unit in family contacts in the
United States. reduces household participation
in local, regional, and national nonfarm selfemployment and wage employment. Moreover,
an additional hectare of land spurs signi®cantly
household participation in farming, and reduces participation in labor markets and migration to the rest of Mexico. An increase in
livestock holdings spurs participation in cash
cropping and in nonfarm activities.
Finally, taking into account that the ``migrants community'' is the reference point for
comparisons of ®xed eects, it is not surprising
that residence in the ``wage employees community'' increases by nearly 26% the probability (t 2:51) that the households participate in
local and regional labor markets, that the
probability of participation in the migration
labor market in the United States is negative
( 19%) and that of participation in migration
to the rest of Mexico is positive (28%).
(b) Determinants of income levels by activity
Table 4 shows the results of the regression of
income sources on family assets. We made the
calculations based on a system of net income
equations correcting for selectivity, and the
®gures in the table show the absolute eects of
a unit change in the explanatory variables on
incomes per activity, given the households
participation in the activity.
The estimated returns to basic education (one
to three years) are positive in income to maize
production incomes and remittance incomes
from migrants within Mexico. Controlling for
other variables, an additional member with
basic education of one to three years, relative to
one with no schooling is associated with an increase of 210 pesos in income from production
of basic staples (about US$ 70, with t 2:48).
An additional member with basic education of
one to three years increases by 96 pesos household income from national migration remittances. These results show high returns to basic
568
WORLD DEVELOPMENT
Table 4. Eects of education and other variables on net income per activity (in 1994 pesos)a ;b
Equations/activities
Variable
Farm sector
From 1 to 3 years of
schooling
From 4 to 6 years of
schooling
From 7 to 9 years of
schooling
More than 9 years of
schooling
Education of the
household head
Experience of the
household head
Family members in
the US
Family members in
the rest of Mexico
Landholdings
Family size
Value of livestock
holdings (1000)
Wage employees
community
Indigenous
community
Commercial farming
community
Inverse Mills Ratio
Constant
R2d
Nonfarm
self-employment
Wage-employment
Staples
Cash crops
and livestock
210
(2.48)
199
(2.57)
156
(1.89)
260
(2.12)
9
(0.24)
)32
(1.09)
)
)303
(0.65)
)251
(0.59)
)183
(0.40)
794
(1.12)
372
(1.86)
217
(1.37)
)
486
(0.86)
126
(0.24)
695
(1.25)
)667
(0.81)
227
(0.93)
)130
(0.67)
)
)
)
)
4.4
(0.31)
)127
(2.07)
)
538
(5.36)
305
(0.90)
0
(3.36)
)1,481
(0.91)
3,833
(2.62)
14,384
(7.50)
1,543
(2.06)
)8,196
(1.95)
0.44
)
285
(0.97)
)356
(1.34)
213
(0.62)
826
(6.21)
898
(1.18)
0.18
Migration
International
National
)200
(0.52)
)438
(1.20)
27
(0.07)
2,334
(4.22)
579
(3.50)
)9
(0.06)
)
)38
(0.26)
351
(2.92)
)72
(0.43)
)116
(0.42)
50
(0.56)
)10
(0.13)
863
96
(1.49)
)45
(0.88)
9
(0.13)
)228
(1.87)
7
(0.18)
22
(0.68)
)
)
(6.27)
595
)
(7.55)
292
(0.71)
)
458
(1.65)
)
)
)
)
)
)
)2,559
(1.30)
)1,813
(1.02)
5,002
(2.20)
5,806
(7.37)
2,782
(0.54)
0.19
2,420
(1.83)
)1,834
(1.53)
)226
(0.15)
5,398
(9.69)
1,355
(0.39)
0.40
)344c
(0.60)
)
1,158
(3.66)
358
(1.21)
)116
(0.32)
1,193
(7.64)
)375
(0.47)
0.29
)1,556
(1.79)
2,283
(5.58)
)200
(0.11)
0.28
a
The procedure corrected for selectivity bias in the systems of equations, using Lee's extension of Amemiya estimation method. The ®gures in the table are absolute changes in incomes by activity. The t ratios are between parentheses.
b
Source: Authors' estimates.
c
Includes the indigenous community, whose members do not migrate to the United States.
d
The system R2 0:89, the v2 709:17, the degrees of freedom are 74 and the size of the sample is 328.
*
Signi®cance levels 0.10.
**
Signi®cance levels 0.05.
education in traditional farming and in withinMexico migration.
As with primary education, an additional
member with higher primary education (four to
six years) has positive and signi®cant eects on
household income from production of basic
staples (199 pesos with t 2:57). The dierence
between the two strata of primary education is
that for the higher stratum, an additional
member increases household incomes by the
channel of international migration remittances
(351 pesos, t 2:92) but not of within-Mexico
migration.
Secondary education (from seven to nine
years of schooling) and preparatory, technical,
or secondary education (more than nine years
of schooling) also bring substantial gains in
incomes from basic staples, which also includes
commercial production of these crops. An additional member with more than nine years of
MEXICO
schooling produces a large increase in the salary income of the rural household (of 2,334
pesos or almost US$780, with a t 4:22).
By contrast with the results for years of
schooling of household members, an additional
year of education of the household head does
not produce an increase in basic staples incomes. Nevertheless, and similar to the results
for secondary education and above, one additional year of schooling for the household head
increases incomes from wage employment (559
pesos or US$186, with a t 3:5). Moreover,
increasing the schooling of the household head
generates income gains in commercial agriculture and livestock husbandry (372 pesos with
t 1:86).
An additional year of experience of the
household head only generates gains in incomes
from commercial agriculture and livestock
husbandry (217 pesos with a t 1:37).
Beyond the positive eect of an additional
member migrating to the United States or the
rest of Mexico (863 pesos with t 6:27 and 595
pesos, with t 7:55, respectively), other family
assets have positive eects on incomes. That is
the case with the estimated eect of an additional hectare of landholdings on incomes
coming from commercial agriculture and livestock husbandry, and an additional family
member on wage employment income. By
contrast, the impact of an additional hectare
does not have an eect on incomes from production of staples; note that increasing family
size by one memberÐwithout adding educationÐactually decreases farm income from
staples.
These latter results require further discussion.
The average farm size of farm households in
small villages is quite small (around two hectares) and typically only a part of this land is
dedicated to maize for home consumption. As
farm size increases, the likelihood increases that
the farm produces commercial crops and livestock (see Table 3) and farm incomes from
these sources increase. As for staples, even
though more land increases the probability of
producing these crops, incomes from these
crops do not rise. The latter might be because
maize is raised in small ®elds and usually for
home consumption, and the net income from
this activity is very low or sometimes even
negative. That may also explain why income
from production of staples declines as family
size increases (remember that Table 3 shows
that an additional family member does not increase the likelihood that the household pro-
569
duces staples). That is, given land constraints
and the use maize for home consumption, an
additional family member reduces income from
production of staples. Although family size
does not aect the probability of household
participation in wage employment, it is probable that an additional family member raises
family labor time in the labor market. That is
re¯ected in the positive impact of family size in
wage income, in turn implying that family labor
is diverted from production of basic grains into
wage employment.
The results arising from comparison of
communities are as expected. For example, and
recalling that the ``migrant community'' is the
reference point for comparisons, it is not surprising that income received from nonmigratory labor markets by the ``wage employees
community'' households are much higher than
those received by households in the ``migrants
community.''
The Inverse Mills Ratio (IMR) is signi®cant
for all the activities. This indicates the importance of self-selection of households in participation in a given activity and in determining
their incomes in that activity. The latter is due
to the fact that the variables that aect choice
of activity, via the selectivity variables, also
aect household incomes from speci®c activities.
(c) Returns to education in total incomes of
households
We ran an additional regression using ordinary least squares (OLS) of the Mincerian type
in order to complete our study of returns to
education, as well as to compare our results
with the only other study of which we are aware
concerning the eects of education in rural areas of Mexico. The dependent variable is the
logarithm of household total net income, and
the explanatory variables are the same as we
used in the regressions in Table 5.
Controlling for the other variables, the estimated eects of primary and secondary education are not statistically signi®cant. By
contrast, the number of family members with
more than nine years of schooling is associated
with a substantial increase in total household
income (11.8% or US$670 in 1994). The same
occurs in the case of household head education,
a one-year increase in which increases household income by 7.4% (around US$400). Besides
schooling, other assets increase household income; these include the number of family
570
WORLD DEVELOPMENT
Table 5. Total income results for the OLS regressions
Variable
The estimated eect on the log of total net income
(Mincerian form)
From 1 to 3 years of schooling
From 4 to 6 years of schooling
From 7 to 9 years of schooling
More than 9 years of schooling
Education of the household head
Experience of the household head
Experience squared
Family members in the US
Family members in the rest of Mexico
Value of livestock holdings (1000)
Landholdings
Family size
Wage employees community
Indigenous community
Cash crop community
Constant
Sample size
R2
*
0.057
0.014
0.054
0.118
0.074
0.016
)0.027
0.106
)0.133
0.207
)0.003
0.052
0.360
)0.151
0.459
7.720
391
0.252
Signi®cance level of 0.10.
Signi®cance level of 0.05.
**
members in the United States (10.6%) and the
value of livestock holdings (20.7%). Finally,
the experience of the household head has the
highest eect in the ``commercial agriculture
community'' and the ``wage employees community.''
Our results concerning the returns to
schooling of household heads are very similar
to those estimated by Singh and Santiago
(1997) in their study on returns to education in
farm activities in a rural region of Mexico.
According to their calculations, the returns are
between 9% (for the head of household) and 8%
(for the wife).
5. CONCLUSIONS
Our study shows that education and years
of schooling aect activity choice of rural
households. Similar to other analyses in
Mexico, our results support the argument that
the returns to education in rural incomes are
high and statistically signi®cant, independent
of the level of schooling. Our study allows
elaboration on the details of schooling eects
as follows.
Primary, secondary, and preparatory education have positive eects on income from basic
grains for those who produce them, indicating
positive returns to schooling in Mexican traditional agriculture. Poverty and persistent market failures make it so that small farmers
continue to produce maize and beans for home
consumption and to manage risk, incorporating new knowledge through skills acquired in
school. At higher levels of schooling, it is to be
expected that returns to education for staples
production will be higher in commercial farming as compared to subsistence farming. Our
analysis indeed shows that post-primary
schooling has positive and signi®cant impacts
on commercial farm incomes.
The nonsigni®cant returns shown by primary
and secondary education on family nonfarm
enterprise incomes in the local area or in the
region may be due to two factors: (a) the content of public education (universal and standardized over the entire country) does not
provide young farmers with the necessary
knowledge to incorporate improved technologies in production of cash crops, livestock, and
nonfarm products and services; and (b) that
that level of education does not lead to increased wages. The former is the argument
made by Reardon and Schejtman (1999) and
the second is made by Attanasio and Szekely
(1999). According to the latter, the returns to
education in Mexican wage incomes only begin
to grow when they go beyond secondary education. Their estimates indicate that, during
1986±96, the returns to primary and secondary
education in wage employment actually diminished, and that the returns to advanced
schooling (beyond nine years of schooling) rose
considerably during the same period (from
MEXICO
190% to 243%). This result is similar to ours, in
that we ®nd that returns to advanced schooling
are high for wage employment.
The latter result, combined with a similar
result concerning the positive eects of education of the household head on wage income,
suggests that the establishment of manufactures
®rms in rural areas will increase household incomes. At the same time, the high returns to
preparatory and technical education in local
and regional labor markets indicates that this
type of education increases labor productivity
in the nonrural economy of villages and small
cities. This implies that development policies in
Mexico should not only focus on primary
education (as has been the case to present) but
also strive to increase the access of the rural
young to schooling beyond the nine-year level.
In general, and despite the insistence by
Mexican politicians of the importance of
education, eorts in that direction are still insucient. For example, faced with the recommendation of UNESCO that investment in
571
education in developing countries should constitute at least 8% of GNP, until 1998 the
highest percentage achieved by Mexico was only
5% (Romero Hicks, 2000; Singh & Santiago,
1997). Within this gloomy panorama, the rural
areas are in the worst situation, as there one
®nds the highest rates of illiteracy, the lowest
average education, the least years of schooling
and the least access to post-primary education
(Bracho, 1998).
Mexico should make a priority of rural education, not only with more investment, but
also by making its present investment more
eective and ecient. Our results suggest that
rural education needs to incorporate the
teaching of practical and technical skills that
are needed in the rural context, to make rural
activities more productive. In that sense, the
federalization of education put in place in 1992
may provide an incentive to state and municipal authorities to include in their education
plans the perspective of rural reality so that
rural education performs better.
NOTES
1. In Mexico, most entrepreneurs live in urban centers.
2. The studies that report positive impacts of education
include those of Yang (1997) for agreggate agricultural
value added in China, and Jollie (1996) for incomes in
Ghana, and Jacoby (1991) for own-farm and livestock
husbandry incomes in Peru, and Singh and Santiago
(1997) for farming activities in Mexico. Examples of
studies that report negative or nonsigni®cant impacts of
education include Adams (1995) for aggregate gross
value added in the production of wheat, sugar cane, and
rice in Pakistan; that of Rosegrant and Evenson (1992),
which estimated total factor productivity in India, and
that of Adams (1993) for total household incomes
excluding migration remittances in Egypt. See details in
Taylor and Y
unez-Naude (2000).
3. An exception is the study of Jollie cited above.
There is another group of studies that treat the theme of
portfolio diversifcation in rural households, but their
objective is not to evaluate the eects of education on
rural productivity and incomes. We refer to those that
analyze the factors that determine diversi®cation (among
which education) as well as the eects of diversi®cation
(see below and the review in Ellis, 1998).
4. It should also be mentioned that the data also make
it possible to run regressions of the determinants of
diversi®cation of sources of incomes by type of household, as is done in Leones and Feldman (1998) for the
Philippines and other authors for various regions of
developing countries (see a review in Ellis, 1998). The
data base is available from the network PRECESAM,
whose website is http://www.colmex.mx/centros/cee/
precesam/precesam.htm
5. We do not make a distinction between them
because, with the exception of land property rights,
the characteristics of ejido producers are very similar
to private smallholders. Moreover, with the ejido
reform of 1991 the land property rights of ejido
producers are being transformed, as the reformed
made possible sale and purchase of ejidal land, and
with that change, the characteristics of ejido producers
are converging yet further with those of private
producers.
6. The dierence is expected because our sample only
includes small producers and in Mexico there are also
medium farmers that produce basic grains (see, for
example, Hernandez Estrada, 2000 and Y
unez-Naude &
Guevara, 1998).
7. The net income from each farm sector activity
includes the implicit income earned from subsistence
production imputed at local prices.
572
WORLD DEVELOPMENT
REFERENCES
Attanasio, O., & Szekely, M. (1999). Introducci
on: La
pobreza en America Latina. Analisis basado en los
activos. El Trimestre Economico, LXVI (3), 263,
317±364.
Adams, R. H. (1993). The economic and demographic
determinants of international migration in rural
Egypt. Journal of Development Studies, 30(1),
146±167.
Adams, R. H. (1995). Agricultural income, cash crops,
and inequality in rural Pakistan. Economic Development and Cultural Change, 43(3), 467±491.
Bracho, T. (1998). Mexico. Per®l educativo de sus
adultos y tendencias de escolarizaci
on de sus nisimnos. Unprocessed, Colegio de Mexico.
Bracho, T., & Zamudio, A. (1994). Los rendimientos
economicos de la escolaridad en Mexico, 1989.
Economõa Mexicana, 3(2), 345±377.
Ellis, F. (1998). Household strategies and rural livelihood diversi®cation. Journal of Development Studies,
35(1), 1±38.
Evans, H. E., & Ngau, P. (1991). Rural±urban relations,
household income diversi®cation and agricultural
productivity. Development and Change, 22, 519±545.
Jacoby, H. G. (1991). Productivity of men and women
and the sexual division of labor in peasant agriculture of the Peruvian Sierra. Journal of Development
Economics, 37(1/2), 265±287.
Jamison, D. T., & Lau, L. J. (1982). Farmer education
and farm eciency. Baltimore, MD: Johns Hopkins
University Press.
Jollie, D. (1996) The impact of education in Rural
Ghana: Examining productivity and labor allocation
eects. Princeton University and World Bank, Unprocessed.
Hern
andez Estrada, M. I. (2000). Tipologõa de productores agropecuarios. In A. Y
unez-Naude (Ed.), Los
Peque~s Productores Rurales: Las Reformas y las
Opciones. Mexico: El Colegio de Mexico.
Leones, J. P., & Feldman, S. (1998). Nonfarm activity
and rural household income: evidence from Philippine microdata. Economic Development and Cultural
Change, 46(4), 789±806.
Massey, D. S., Arango, J., Kouaouci, A., Pellegrinoy,
A., & Taylor, J. E. (1999). Worlds in motion:
International migration at century's end. Oxford:
Oxford University Press.
Mexico±United States Binational Migration Study.
(1998). Migration between Mexico and the United
States. Austin, Texas: Morgan Printing.
Phillips, J. M. (1987). A comment on farmer education
and farm eciency: A survey. Economic Development
and Cultural Change, 35(3), 637±641.
Reardon, T. (1997). Using evidence of household income
diversi®cation to inform study of the rural nonfarm
labor market in Africa. World Development, 25(5),
735±747.
Reardon, T., Rello F., Schejtman. A., & Stamoulis, K.
(1998). Agroindustrialization in intermediate cities of
Latin America: Hypotheses regarding employment
eects on the rural poor. Paper presented at the
IFPRI Conference, Strategies for Stimulating
Growth of the Rural Nonfarm Economy in
Developing Countries, 17±21 May. Airlie House,
Virginia.
Reardon, T., & Schejtman, A. (1999). Los proyectos de
alivio de la pobreza rural desde la perspectiva del
empleo: una relectura del proyecto PROLESUR.
Unprocessed, United Nations Food and Agricultural
Organization, Reional Oce for Latin America and
the Caribbean, Santiago, Chile.
Romero Hicks, J. L. (2000). Descentralizaci
on educativa. V
ortice, 1, 149±154.
Rosegrant, M. W., & Evenson, R. E. (1992). Agricultural productivity and sources of growth in South
Asia. American Journal of Agricultural Economics,
74(3), 757±761.
Singh, R., & Santiago, M. (1997). Farm earnings,
educational attainment, and the role of public policy:
some evidence from Mexico. World Development,
25(12), 2143.
Taylor, J. E., & Y
unez-Naude, A. (2000). Selectivity and
the returns to schooling in a diversi®ed rural economy. American Journal of Agricultural Economics,
287±297.
Yang, D. T. (1997). Education and o-farm work.
Economic Development and Cultural Change, 45(3),
613±632.
Y
unez-Naude, A., Guevara, A. (1998). Evaluacion de los
programas de desarrollo regional en el plano comunitario. Reporte de Resultados. Mexico: El Colegio de
Mexico.