www.elsevier.com/locate/worlddev
World Development Vol. 29, No. 3, pp. 529±547, 2001
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0305-750X/01/$ - see front matter
PII: S0305-750X(00)00105-4
Nonfarm Employment and Poverty in
Rural El Salvador
PETER LANJOUW *
The World Bank, Washington, DC, USA
Summary. Ð This paper analyzes two complementary data sets to study poverty and the nonfarm
sector in rural El Salvador. We ®nd that rural poverty in El Salvador remains acute and
signi®cantly higher than in urban areas. While the rural poor are mainly agricultural laborers and
marginal farmers, some nonfarm activities are also of importance to the poor. In fact, nonfarm
activities in El Salvador account for a signi®cant share of rural employment and income for both
the poor and the nonpoor. The poor, on the one hand, are engaged in ``last resort'' nonfarm
activities that are not associated with high levels of labor productivity. The nonpoor, on the other,
are engaged in productive nonfarm activities which are likely to present a potent force for upward
mobility. Signi®cant correlates of these high-productivity occupations include education, infrastructure, location, and gender. While most of the analysis is at the household level, the data also
permit some focus on small-scale rural enterprise activities. It appears that in El Salvador very few
rural enterprises report utilizing formal credit in setting up their activities. In addition, a signi®cant
proportion of enterprises are engaged in subcontracting arrangements with some larger, often
urban-based, ®rm. Ó 2001 Elsevier Science Ltd. All rights reserved.
Key words Ð nonfarm employment, poverty, El Salvador
1. INTRODUCTION
Poverty is the subject of much discussion in
El Salvador. There is a sense that without
concerted attention to poverty issues, a full and
sustained transition from decades of violence
will remain elusive. It is also feared that equitymotivated programs, such as the agrarian
reforms, have not succeeded in eliminating
poverty altogetherÐeither because implementation has not been as e€ective as hoped, or
because at least part of the poverty problem is
linked to issues beyond access to land and
tenure security alone. There is a feeling that
more must be done.
While few dispute the importance of addressing poverty in El Salvador, there is considerable debate surrounding certain key
aspects of the poverty problem. There is, for
example, no universal consensus on how
widespread poverty is in El Salvador, where
poverty is concentrated, and which household
characteristics are most closely linked to poverty. Much of this debate is prompted by differences in methodological approaches, and by
shortcomings in available data sources. Nonetheless, there does appear to be broad agreement that poverty in rural areas deserves
529
particularly close attention. Most approaches
to the measurement of poverty tend to indicate
that the rural poverty problem is particularly
pressing.
In order to understand the causes of rural
poverty and to design policies that address
these, one must examine in some detail the
operation of the rural economy. It becomes
quickly apparent that the rural economy extends well beyond agriculture. The nonfarm
sector in rural areas is highly heterogeneousÐ
encompassing the full spectrum of economic
activities which occur in rural areas but which
are not directly associated with agricultureÐ
and can represent a very important part of the
rural economy in terms of incomes and employment generated. While this sector has not
received the same level of attention as the agriculture sector, there is a growing appreciation
of its potential in terms of both poverty
*I
am grateful for the comments and suggestions of
Alberto Valdes, Ramon Lopez, and three anonymous
referees. I remain responsible for all outstanding errors.
The views in this paper are my own and should not be
taken to re¯ect the views of the World Bank or any
aliated institution.
530
WORLD DEVELOPMENT
alleviation and growth more generally. Recent
analyses of the sources of growth in East Asia
have stressed the central role played by the
nonfarm sector in rural areas. 1 A question of
considerable interest is whether the nonfarm
sector can play a similar role in stimulating
rural growth in El Salvador, and Latin America
more broadly.
This paper examines data from two recent
household surveys in El Salvador to assess to
what extent the nonfarm sector might be able
to contribute to rural poverty alleviation. The
layout of the paper is as follows. The next
section brie¯y attempts to organize our thinking about poverty and the nonfarm sector. This
is followed in Section 3 with a discussion of
poverty in El Salvador, and some tentative estimates of poverty based on consumption expenditures. Section 4 introduces some
quantitative evidence on the size of the nonfarm sector in rural El Salvador, and the range
of activities that comprise this sector. This
section also considers what relationship exists
between poverty and nonfarm employment.
Section 5 turns to an examination of the factors
that appear to in¯uence the involvement of
rural households in the nonfarm sector and also
the earnings associated with those activities. In
Section 6 we summarize the main ®ndings and
o€er some remarks on policy implications of
the ®ndings.
Before proceeding, we make a brief note
about the data sources underlying the analysis.
Two sources of data are being used in this paper. The Encuesta de Hogares de Propositos
Multiples, 1994-III, (EHPM) is a nationally
representative household survey ®elded by the
Ministerio de Plani®cacion y Coordinacion del
Desarollo Economico y Social (MIPLAN) in El
Salvador. The EHPM is an annual survey,
®elded throughout the year in four ``waves.'' In
total, roughly 20,000 households are covered.
The analysis in this paper is based only on the
third wave of the 1994 survey (the waves are
designed to be amenable to self-standing analysis) covering 4,229 households in total (1,743
in rural areas) and ®elded during the period
July-September, 1994. This wave is somewhat
special in that it contained a detailed consumption module (for a subsample of households) which permits an analysis of poverty
based on consumption expenditure rather than
income (see below). 2
The second source of data comes from a rural household survey ®elded by Fundacion
Salvadore~
na Para el Desarollo Economico y
Social (FUSADES) in 1996. The survey covered a sample of about 630 rural households
from all regions of El Salvador, strati®ed on
households' characteristics according to their
main economic activities, i.e. the self-employed,
agricultural workers, and nonfarm workers.
The survey was designed to be representative of
the rural population at the 10% level of significance (Lopez, 1996). The FUSADES survey
obtained information on a wide range of demographic characteristics, location, and income variables. The level of detail in the
information collected has permitted the calculation of a comprehensive measure of income
which is less likely to su€er from important
omissions than is conventionally the case with
income surveys. The FUSADES survey provides a useful complement to the EHPM in that
while it is smaller in sample size it provides
greater detail about rural livelihoods and activities.
2. A QUICK OVERVIEW OF THE
POVERTY±NONFARM LINKS
Rural o€-farm employment has been traditionally seen as a low-productivity sector, producing low quality goods. The sector, in this
view, is expected to wither away as a country
develops and incomes rise. There is thus no
obvious rationale for governments to promote
the sector, nor be concerned about negative
repercussions on the rural nonagricultural sector arising from government policies directed at
other objectives.
In recent years, opinion has been swinging
away from this view, however, and there are a
number of arguments that suggest that neglect
of the sector is socially costly. For example, it
has been argued that the sector has a positive
role in absorbing a growing rural labor force, in
slowing rural-urban migration, in contributing
to national income growth and in promoting a
more equitable distribution of income. 3
Lanjouw and Lanjouw (forthcoming) indicate that while de®nitional and data-related
uncertainties remain, the rural nonagricultural
sector is both large and, on aggregate, has been
growing over time. Hazell and Haggblade
(1993) emphasize that when rural towns are
included in employment calculations, the share
of the rural labor force employed primarily in
nonagricultural activities rises sharply. They
calculate that in Latin America, 47% of the
labor force in rural settlements and rural towns
EL SALVADOR
is employed in nonfarm activities. This can be
compared to 28% when only rural settlements
are included. Hazell and Haggblade also highlight the importance of female participation in
nonagricultural activities: 79% of women in the
Latin American rural wage-labor force are estimated to be employed in nonagricultural activities.
Nonagricultural activities can be broadly divided into two groups of occupations: high labor productivity/high income activities and low
labor productivity activities that serve as a residual source of employmentÐa ``last-resort''
source of income (Lanjouw and Lanjouw,
forthcoming). These latter activities are common among the very poor, particularly among
women. Such employment may nevertheless be
very important from a social welfare perspective for the following reasons: (a) o€-farm employment income may serve to reduce
aggregate income inequality; (b) where there
exists seasonal or longer-term unemployment in
agriculture, households may bene®t even from
low nonagricultural earnings; and (c) for certain subgroups of the population who are unable to participate in the agricultural wage
labor market, notably women in many parts of
the developing world, nonagricultural incomes
o€er some means to economic security.
It is dicult to say whether nonfarm employment is income inequality increasing or
decreasing without information about what the
situation would have been in the absence of
such occupations. One important consideration
remains that although aggregate income inequality may widen as rural nonagricultural
incomes increase, this may occur alongside a
decline in absolute poverty (if, for example, all
households bene®t from o€-farm income, but
the rich bene®t proportionately more). 4
Empirical evidence in many countries supports the notion that agricultural wages are not
perfectly ¯exible, and that rural agricultural
labor markets are segmentedÐwith certain
subgroups of the population such as women
and children unable to obtain employment at
the market wage. Lanjouw (1995) found some
evidence that small farms in Ecuador obtained
higher yields than large farms. A possible explanation for this observation could be that
small farmers apply more labor per unit of land
than large farmers. 5 Family labor is applied
beyond the level where the marginal product of
labor is equal to the market wage, because for
at least some family members the market wage
is not the opportunity cost of labor. If indeed
531
agricultural wage employment is not an option
for certain family members, then rural nonagricultural employment opportunities, even if
they are not highly remunerative can make a
real di€erenceÐespecially for those households
which do not possess farm land.
In sum, the existing literature points to a
potentially strong relationship between the rural nonagricultural sector and rural poverty.
Because of market imperfections and distortions, nonfarm activities are likely to employ
labor beyond the point where the marginal
product of labor is equal to the prevailing average agricultural or urban wage. The wide
range of nonagricultural activities in terms of
labor productivity suggests that for some
households and individuals these activities
provide a last resort safety-net function, while
for others they o€er a genuine opportunity for
sustained upward mobility. In this paper we
attempt to shed some empirical light on at least
some aspects of the relationship between poverty and the nonagricultural sector in the context of rural El Salvador.
3. RURAL POVERTY IN EL SALVADOR
Poverty has been the focus of attention in
many studies in El Salvador (recent examples
include FUSADES, 1993, World Bank, 1994a
and MIPLAN, 1995a). But to date, there is no
clear consensus as to the magnitude and dimensions of the poverty problem in El Salvador. There are numerous methodological and
data-related issues that stand in the way of
precise, quanti®ed poverty rate calculations in
El Salvador: for the country as a whole, and
among various population subgroups. These
issues are brie¯y noted below, but space prevents discussing them in detail. 6
In El Salvador, as in many other Latin
American countries, poverty analysis has generally been carried out on the basis of income as
the household-level indicator of well-being.
This is in contrast with conventional practice in
other parts of the world, where consumption
expenditures are commonly taken as the welfare indicator. The principal reason for choosing consumption is that experience has shown
that these are measured with greater accuracy
than incomesÐparticularly for the poor, who
are most likely to consume a relatively narrow
range of goods and services. 7 Many of the
household surveys ®elded in Latin American
countries are modeled on labor force surveys
532
WORLD DEVELOPMENT
designed to measure household earnings. Such
surveys are generally weak in capturing incomes from nonwage labor sources (such as
remittances, transfers, and self-employment)
and from agricultural activities. These omissions may be particularly pertinent to the
measurement of rural poverty.
The next step in measuring poverty is to relate household-level welfare indicators to some
poverty thresholdÐthe poverty line. In constructing a poverty line, many assumptions are
generally required, and these can at times become quite contentious. Issues that must be
addressed, for example, relate to the nutritional
cut-o€ point to be applied, whether account is
to be taken of di€erent requirements between
adults and children or between males and females, and what kind of adjustment must be
made to allow for nonfood items in the basic
consumption basket. While various poverty
lines have been formulated for El Salvador,
their treatment of these issues has not always
been very clearly documented (see, IIES-UCA,
1993, and also World Bank, 1994a). Nor has
the treatment been the same across the di€erent
calculations.
Yet further issues relate to the partial geographic coverage of many household surveys in
El Salvador (missing certain regions, or focusing
only on urban areas) and the nonavailability of
up-to-date population expansion factors that
may lead to biased assessments of the distribution of poverty across population subgroups.
Finally, there is an important issue associated
with the nonavailability of a spatial cost-of-living index. Such an index adjusts for the fact that
to reach a given standard of living in Metropolitan San Salvador might cost quite a di€erent
amount from other urban areas or rural areas.
Combined, these factors issue a strong
warning against e€orts to provide detailed calculations of poverty rates in El Salvador. While
such calculations would undoubtedly be of
great value, one might wish to ®rst concentrate
on reaching agreement on the appropriate
methodology to apply and in securing the requisite data. At the same time, the lack of
quantitative poverty measures need not impede
unduly one's ability to focus on poverty-related
questions. If one is prepared to con®ne one's
remarks to broad comparisons of poverty
across population subgroups (say, between rural and urban areas) and to be satis®ed in
stating that poverty among one group is higher
or lower than among the other (without attempting to state by how much), then meth-
odological di€erences and uncertainties may be
less of a constraint.
It is in this spirit that we present in Table 1
some tentative estimates of the incidence of
poverty in El Salvador in 1994, based on the
consumption expenditure collected by the
EHPM. These ®gures should not be accepted
without question (just as all other attempts to
measure poverty in El Salvador are likely to be
contentious), because they embody a range of
strong (and controversial) assumptions. First, a
speci®c methodology was applied to estimate a
robust incidence of poverty for the two subsamples of the EHPM for which highly di€erence consumption modules were ®elded (for a
detailed exposition, see Lanjouw and Lanjouw,
1996). Second, it was assumed that the rural
cost of living was lower than in urban areas, in
proportion to the ratio of the MIPLAN
(1995b) urban poverty line to rural poverty
line. Third, the poverty line which was taken
was simply the one published in MIPLAN
(1995b) for urban areas without any attempt to
establish its validity. Fourth, we have made no
adjustments for di€erences in needs across
family members (via equivalence scales) and
make no allowances for economies of scale in
household consumption.
The purpose of the poverty estimates in Table 1 is to shed light on some of the broad
geographic patterns of poverty in El Salvador.
At the level of the country as a whole, poverty
is much higher in rural areas than in urban
areas. This seems to be driven in particular by
the relatively low level of poverty in Metropolitan San Salvador. 8 Across the other four
broad geographic regions of the country, the
evidence is less strong that there exists a clear
distinction between rural and urban areas. El
Salvador is unique in its de®nition of urban
areas in that it counts as urban all municipal
centers (cabeceras municipales) without taking
into account the actual population of those
centers. This is in contrast with many other
countries (although it is not unique in Latin
America, see Klein, 1993) and is likely to result
in higher estimates of the incidence of urban
poverty than would obtain if only large conurbations were designated as urban. There is
rather little evidence of a concentration of
poverty within a speci®c broad region (sample
size constraints prevent a further geographic
breakdown within broad regions). Based on the
``low'' poverty line, there is some indication
that rural poverty in the central regions (located more closely to San Salvador) is lower
EL SALVADOR
533
Table 1. The incidence of poverty in El Salvadora ;b
High poverty line
Incidence of poverty
Low poverty line
Persons poor
Incidence of poverty
Persons poor
West
Urban
Rural
All
0.68
0.75
0.72
312,343
449,050
761,393
0.28
0.38
0.33
127,241
223,726
350,967
Central 1
Urban
Rural
All
0.74
0.76
0.75
329,117
560,313
889,430
0.31
0.33
0.32
138,726
240,980
379,706
Central 2
Urban
Rural
All
0.70
0.79
0.76
131,827
237,550
369,377
0.36
0.32
0.34
66,983
96,925
163,908
East
Urban
Rural
All
0.67
0.79
0.74
287,074
513,079
800,153
0.30
0.38
0.35
129,960
247,231
377,191
Metropolitan
San Salvador
0.40
518,926
0.08
100,751
National
Urban
Rural
0.56
0.77
1,579,287
1,759,992
0.20
0.35
563,661
808,862
Total
0.66
3,339,279
0.27
1,372,523
a
Source: Encuesta de Hogares de Propositos Multiples, 1994-III.
b
(i) The Encuesta de Hogares de Propositos Multiples 1994-III yields potentially problematic consumption ®gures as
two sharply divergent consumption questionnaires were ®elded to two non-overlapping subsamples of total sample.
The methodology developed in Lanjouw and Lanjouw (1996) was implemented in order to ensure comparability. For
more details, see Lanjouw and Lanjouw (1996). (ii) The ``high'' poverty line refers to a monthly per capita expenditure ®gure of 667 Colones (approximately US$75), and the ``low'' poverty line (which can be interpreted as a
measure of ``extreme poverty'') refers to a monthly per capita expenditure ®gure of 334 Colones (approximately
US$36). These poverty lines have been calculated from the data based on the per capita cost of a basic food bundle
(for urban areas) as calculated by the Ministerio de Plani®cacion in San Salvador, and adding a nonfood expenditure
allowance in accordance with the average amount spent on nonfood items by households with food expenditures
equal in value to the MIPLAN food basket. The ``low'' poverty line is simply 50% of the ``high'' line, and corresponds loosely to the $1 per day poverty line that underpins global poverty comparisons. (iii) Rural expenditures
have been in¯ated by a factor of 60% to account for MIPLAN's calculation that the basic urban food basket costs
60% more than a basic rural food basket. (iv) The broad regions presented above break down into the following
departments: West (Santa Ana, Ahuachapan, Sonsonate); Central 1 (La Libertad, Rural San Salvador, Chalatenango, Cuscatlan); Central 2 (San Vicente, La Paz, Cabanas); East (Usulutan San Miguel, Morazan, La Union). (v)
Overall aggregate populations vary slightly across the high poverty line column and low poverty line column, because
of rounding in the reported headcount rates.
than in either the East or West. We shall see
below that the most productive nonfarm activities tend to be concentrated in the two
central regions of the country.
4. THE NONFARM SECTOR IN RURAL
EL SALVADOR
A study of rural nonfarm employment in
Latin America based on census data suggests
that in 1975 roughly 20% of the economically
active population in rural El Salvador was
employed in the nonfarm sector (Klein, 1993).
This can be compared with ®gures of 16%,
18% and 40% for Honduras, Guatemala and
Costa Rica, respectively, in the early to mid1970s. More recent census ®gures for El Salvador have not been calculated, no doubt for
data-related reasons associated with the political and military instability of the intervening
period.
534
WORLD DEVELOPMENT
In Tables 2 and 3 we present estimates based
on the EHPM household survey. In 1994,
36.4% of the economically active rural population was employed in the nonfarm sector,
nearly twice as many as in the mid-1970s (Table
2). The range of activities in which the rural
population is engaged includes both manufacturing and services. Nearly 30% of all rural
nonfarm employment is engaged in some form
of manufacturing activity (combining textiles
and carpentry with the generic manufacturing
entry). Commerce represents nearly an additional 25%, construction about 13%, domestic
service just over 10%, and transport nearly 6%.
The remaining activities are largely various
service sector activities.
As a proportion of the economically active
population, women are far more likely to be
active in the nonfarm market than men (Table
2). Seventy-two precent of economically active
women are employed in the nonfarm sector,
compared to just about 25% of men. But in the
EHPM survey, only about 22.5% of all women
of working age are counted among the economically active population (compared to
73.6% of men). 9 The activities in which women
are heavily engaged include ®rst of all commerce, and are followed by manufacturing and
domestic service. For men, commerce and domestic service are of far less signi®cance, but
construction, manufacturing and also transport
are important sectors.
There is geographic variation in El Salvador in the signi®cance of the rural nonfarm
sector (Table 3). In the Central 1 region
(which includes the departments of Chalatenango, La Libertad, San Salvador, and
Cuscatlan) nearly 50% of the economically
active population is employed in the nonfarm
sector. This contrasts with only 23.2% in the
East (Usulutan, San Miguel, Morazan, and
La Union). The spectrum of activities across
regions is fairly uniform in terms of shares of
employment. For example, while manufacturing appears to be relatively more important in the West than in the other regions:
about 30% of all nonfarm employment occurs
in the textile, carpentry and manufacturing
sectors in the West (Ahuachapan, Santa Ana
and Sonsonate), it is as high as 24±26% in the
other regions.
In Table 4 we observe that involvement by
households in the nonfarm sector is broadly
correlated with lower rates of poverty. 10 The
highest incidence of rural poverty in the EHPM
survey is observed among households that engage in both agricultural labor and farming. In
fact, agricultural labor appears to be particularly closely linked to rural poverty in that of
eight possible household economic status cate-
Table 2. Nonfarm activities in rural El Salvador (percentage of persons aged above 10 years engaged in remunerated
labor)a
Percentage of populationb with primary occupation
Fishing
Manufacture
Textiles/garments
Wood/straw/leatherware
Utilities
Construction
Commerce
Restaurant/hotel
Transport
Finance
Administration
Teaching
Health
Domestic service
Other service
Total nonfarm
Farming
Agricultural labor
No. of observations
a
Male
0.7
4.2
1.0
1.4
0.3
6.1
2.1
0.2
2.8
0.1
0.2
0.4
0.0
0.9
4.3
24.7
54.4
20.8
2,230
(2.8)c
(17.0)
(4.0)
(5.7)
(1.2)
(24.7)
(8.5)
(0.8)
(11.3)
(0.4)
(0.8)
(1.6)
(0.0)
(3.6)
(17.4)
(100.0)
Female
0.2
10.0
8.0
3.7
0.0
0.4
28.1
1.8
0.0
0.5
0.2
1.6
1.3
12.4
4.1
72.3
11.4
16.1
685
(0.3)
(13.8)
(11.1)
(5.1)
(0.0)
(0.6)
(38.9)
(2.5)
(0.0)
(0.7)
(0.3)
(2.2)
(1.7)
(17.2)
(5.7)
(100.0)
Source: Republica de El Salvador: Encuesta de Hogares de Propositos Multiples, 1994-III.
Column percentages provided in brackets.
c
Primary employment refers to self-reported principal activity.
b
Total
0.6
5.6
2.7
1.9
0.3
4.7
8.4
0.6
2.1
0.2
0.2
0.7
0.4
3.7
4.3
36.4
44.0
19.6
2,915
(1.6)
(15.4)
(7.4)
(5.2)
(0.8)
(12.9)
(23.1)
(1.6)
(5.8)
(0.5)
(0.5)
(1.9)
(1.1)
(10.2)
(11.8)
(100.0)
EL SALVADOR
535
Table 3. Nonfarm activities in rural el salvador (% of persons aged above 10 years engaged in remunerated labor)a
Percentage of population with primary occupation
Fishing
Manufacture
Textiles/garments
Wood/straw/leatherware
Utilities
Construction
Commerce
Restaurant/hotel
Transport
Finance
Administration
Teaching
Health
Domestic service
Other services
Total nonfarm
Farming
Agricultural labor
No. of observations
a
b
West
0.0
6.3
2.0
2.3
0.4
4.1
7.1
0.1
2.2
0.2
0.5
0.8
0.5
3.0
4.1
33.6
43.7
22.7
802
(0.0)b
(18.8)
(6.0)
(6.8)
(1.2)
(12.2)
(21.2)
(0.3)
(6.5)
(0.6)
(1.5)
(2.4)
(1.5)
(8.9)
(12.2)
(100.0)
Central 1
0.9
6.5
3.8
3.1
0.4
6.5
9.9
1.3
2.8
0.2
0.1
0.4
0.5
5.8
6.3
48.5
33.1
18.3
830
(1.9)
(13.4)
(7.8)
(6.4)
(0.8)
(13.4)
(20.4)
(2.7)
(5.8)
(0.4)
(0.2)
(0.8)
(1.0)
(12.0)
(13.0)
(100.0)
Central 2
1.6
5.3
3.3
0.5
0.2
4.9
9.7
0.6
0.6
0.3
0.1
0.9
0.0
2.4
3.3
33.7
50.1
16.1
635
(4.7)
(15.7)
(10.0)
(1.5)
(0.6)
(14.5)
(28.8)
(1.8)
(1.8)
(0.9)
(0.3)
(2.7)
(0.0)
(7.1)
(9.8)
(100.0)
East
0.3 (1.3)
3.5(15.1)
1.5 (6.5)
0.6 (2.6)
0.0 (0.0)
2.8 (12.1)
7.1 (30.6)
0.0 (0.0)
1.9 (8.2)
0.2 (0.9)
0.1 (0.4)
0.9 (3.9)
0.2 (0.9)
2.0 (8.6)
2.1 (9.1)
23.2 (100.0)
56.6
20.2
648
Source: Republica de El Salvador: Encuesta de Hogares de Propositos Multiples, 1994-III.
Column percentages provided in brackets.
gories, the three associated with highest incidence of poverty include agricultural labor
amongst the household economic activities.
Such a pattern has also been noted in the
context of rural north India where agricultural
labor is widely viewed with distaste, in which
households participate only when faced with
acute hardship and no alternative sources of
income (Dreze et al., 1992). For this reason the
likelihood that agricultural labor households
are poor is signi®cantly higher than for many
other household types. Such a perspective
might also apply in rural El Salvador. Of
course, there are di€erent types of agricultural
labor. The category applied in Table 4 encompasses both casual daily wage labor and longterm, permanent employment on a farm,
plantation, or ranch. As a result, not all agricultural labor is likely to be unattractive as a
source of income. This is possibly re¯ected in
Table 4 in that households which engage in
both agricultural labor as well as nonfarm labor run only about an average ``risk'' of poverty (35%Ðsee Table 1).
Nonfarm labor, we have already seen, is also
not homogeneous. In Table 4 one household
category engaging in nonfarm employment is
relatively highly exposed to poverty (households simultaneously engaging in farming, agricultural labor, and nonfarm labor). But
households reliant only on nonfarm labor are
signi®cantly less likely to be poor than all other
rural households. This observation illustrates
the important point raised in Section 2, namely
that the nonfarm sector typically comprises two
distinct sets of activities. On the one hand there
is a set of activities which are reasonably productive, relatively well-paid, and which have
the appealing feature of being comparatively
less exposed than agriculture to climatic variations and uncertainties. On the other hand,
there are a group of activities undertaken by
persons who are unable even to secure an agricultural laboring position; persons who are
perhaps old or disabled, or who may be prohibited by custom from participating in the
agricultural labor market (for example, women
and children). This second set of nonfarm jobs
plays a very di€erent role to the ®rst set. One
way to perceive these two is to regard the ®rst
set as a source of upward mobilityÐa route out
of poverty, and the second as a type of ``safety
net'' which helps to prevent poor persons from
falling into even greater destitution. 11
Both sets of nonfarm jobs have a very important role to play in reducing, or relieving,
poverty. But the types of policies that can be
pursued to help realize their potential are quite
di€erent. In the Indian state of Maharashtra,
the state government supports a large publicworks program o€ering nonfarm employment
at a wage below the prevailing agricultural
wage rate to anyone who presents himself at the
worksite (see Dreze, 1995 and Datt and Rav-
536
WORLD DEVELOPMENT
Table 4. Poverty and rural household activitiesa
b
Household characteristics
Agricultural labor and farming
Agricultural labor only
Agricultural labor farming and nonfarm employment
Farming only
Farming and nonfarm employment
Agricultural labor and nonfarm employment
Nonfarm employment only
Nonfarm income from non-wage sources
Percent of population
(%)
Incidence of extreme poverty
(%)c
5.0
9.6
2.7
26.1
19.9
9.1
26.1
1.6
54.7
48.7
43.9
41.5
35.9
35.2
20.3
16.3
a
Source: Encuesta de Hogares de Propositos Multiples, 1994-III.
Agricultural labor households are de®ned as such if at least one household member is employed as a salaried or
casual wage laborer in agriculture. Farming households refer to those households where at least one household
member is engaged in cultivation. Nonfarm households correspond to those households where at least one household
member is employed in a nonfarm occupation.
c
Extreme poverty is associated with per capita consumption levels falling below the ``low'' poverty line.
b
allion, 1993). This program provides nonfarm
employment to poor persons for whom agricultural wage employment is not an option
(perhaps because of cultural restrictions, or
because climatic conditions have sharply reduced demand for labor). It therefore exploits
the ``safety-net'' function of the nonfarm sector. The attraction of this approach is that the
poor are ``self-targeted,'' i.e., only those who
have no other viable employment alternative
present themselves at the worksite. Thereby the
costly administration of a government targeting
scheme is avoided. Policies aimed to expand
access of the poor to the high-income nonfarm
jobs are likely to look very di€erent. In these
cases, the emphasis is typically on removing
constraints and bottlenecks to such employment by providing training, supporting infrastructure, etc.
In Table 4, while we see evidence of both
types of nonfarm employment (i.e. associated
both with a higher risk of poverty and also with
a lower risk), the numerical importance of the
latter is far greater. Only 2.7% of the rural
population falls in the category of engaging
simultaneously in farming, agricultural labor,
and nonfarm labor. By contrast, nearly 30% of
the rural population is engaged only in nonfarm activities. And this segment is the least
poor of the entire rural population. It seems
possible therefore, that the rural nonfarm sector in El Salvador functions as a route out of
poverty as well as a safety net to those who
experience acute distress. 12
We turn next to an oft-overlooked segment
of the nonfarm sector. Table 5 examines the
FUSADES data to consider the role of small
enterprises in rural areas. While the EHPM
survey did not inquire speci®cally into household enterprises, the FUSADES rural survey
included a separate questionnaire on such activities. In Table 5 we see that of the 300
workers active in rural enterprises, about 40%
were family members. Over 50% of all rural
enterprises covered in the FUSADES survey
were home-based. Commerce was by far the
most common form of rural enterprise, although on average such enterprises were
smaller in terms of employment per ®rm than
pottery and brick-making enterprises. It is interesting to note that only about 5% of all rural
enterprises covered in the FUSADES survey
reported having received training.
Textile enterprises stand out among the more
common enterprises in that they draw particularly heavily on family labor and are most frequently home-based. Nearly a quarter of these
enterprises are engaged in a relationship with
some larger ®rm, in which they receive inputs
from the larger ®rm, assemble them, and then
re-sell their output to the same contractor. Such
subcontracting arrangements has been observed elsewhere in Latin America (see, for
example, Lanjouw, 1995), and are argued by
Hayami (1995) to have been common in rural
areas of East Asia during the earlier stages of
economic development. Hayami (1995) argues
that these arrangements are useful to both
parties in that they provide to the contractor
access to cheap labor, while the home-based
®rms are able to choose how and when to allocate their family labor, and do not have to
concern themselves with bringing the ®nal
goods to the market (which, in the case of
EL SALVADOR
537
Table 5. Rural enterprises in El Salvadora
Sector
Transport
Other services
Other industry
Repair shop
Restaurant/bar
Textiles
Wood/work
Food proc.
Pottery/bricks
Commerce
Total
a
Number
of ®rms
Number
of workers
Percentage
Percentage
family
home-based
members (%)
(%)
Percentage
with training
(%)
Percentage
supplying
contractor (%)
1
3
5
6
5
13
10
13
14
31
1
3
6
16
19
25
37
53
63
77
100
100
100
44
16
73
22
28
13
64
0
33
100
33
20
92
60
54
7
58
0
0
0
0
0
8
10
8
7
3
0
0
20
0
0
23
30
0
7
0
101
300
40
52
5
8
Source: Rural survey, FUSADES (1996).
clothing or shoes for example, might be very far
away).
The range of both nonfarm employment as
well as rural enterprise activities that are engaged in provides some clues as to the relationship between the nonagriculture and the
agriculture sector in rural areas. This broader
relationship has received considerable attention
in the literature. Mellor and Lele (1972), Mellor
(1976), and Johnston and Kilby (1975) have
argued that a virtuous cycle between agricultural intensi®cation and nonfarm activity can
emerge on the basis of production and consumption linkages. Production linkages
emerge, for example, when demand of agriculturalists for inputs such as plows and machinery repair stimulate nonfarm activity via
``backward'' linkages or where agricultural
goods require processing in spinning, milling,
or canning factories (``forward'' linkages).
Consumption linkages emerge as rising agricultural incomes feed primarily into increased
demands for goods and services produced in
nearby towns and villages. While it is dicult
to test the strength of such linkages with the
available data sources, the fact that a large
fraction of nonfarm activities center around
commerce, food processing, transport, and repair activities, suggests that these linkages are
certainly present in the El Salvador case. 13
It is also interesting to note the importance of
rural manufacturing and the existence of
sub-contracting arrangements between rural
home-based enterprises and large, perhaps urban-based, supplier companies. The existence
of such rural nonfarm activities can be very
important to the rural economy because they
introduce a source of rural income that is less
closely linked to agricultural ¯uctuations. This
is in contrast to the activities that are directly
linked to agricultural production and incomes.
In rural areas, insurance and credit markets
often do not operate well, or are missing altogether. This means that, in order to avoid
®nding themselves in a position where they
might need to take consumption loans, farmers'
production decisions are often aimed at cropping patterns which minimize the risk of harvest failure, but which have lower-value
expected yields (Murdoch, 1995). Access of
certain family members to noncyclical sources
of income from manufacturing activities might
help to encourage higher-value agricultural
production decisions.
5. CORRELATES OF NONFARM
EMPLOYMENT AND EARNINGS
We turn now to a closer examination of the
correlates of nonfarm employment in El Salvador by presenting, in Table 6, results from a
Probit model considering the likelihood of
nonfarm employment for the working-age rural
population. In light of the discussion in the
previous section we look not only at all nonfarm jobs together, but also distinguish between nonfarm jobs that can be considered as
``low-productivity'' jobs and those that are
``high-productivity.'' The distinction is based
on whether hourly earnings from these jobs are
lower, or higher, than average hourly earnings
from agricultural labor, respectively. Rather
than report the parameter estimates from the
Probit models, we report in Tables 6±8 the
marginal e€ects. 14
538
WORLD DEVELOPMENT
In the ®rst column of regression results in
Table 6, we ®nd that women are signi®cantly
less likely than men to ®nd employment in the
nonfarm sector (a probability 18 percentage
points lower than for a male, other variables at
their means). We shall return to this ®nding
below. As a person get older, he or she is signi®cantly less likely to be employed in the
nonfarm sector. Compared to the uneducated,
all persons who have been educated are significantly more likely to ®nd employment in the
nonfarm sector. There appears to be a
strengthening of the e€ect of education on the
probability of employment as education levels
improve.
Households with larger per capita landholdings are less likely to be employed in the nonfarm sector. This is not the case in all contexts,
because where highly desirable nonfarm jobs
are rationed, it is the wealthier households
(those with more land) who might be better
placed to secure such jobs. In El Salvador, while
some family members of the larger landowning
households are indeed likely to be working in
the nonfarm sector, they probably reside in San
Salvador and therefore do not feature in the
FUSADES sample. Cultivating households
(those reporting actual involvement in agriculture) are also less likely to have family members
employed in the nonfarm sector. For these
households, the ®rst claim on family members'
labor is apparently for assistance in the ®elds
rather than nonfarm sources of income.
The proximity of a household to a paved
road signi®cantly improves the likelihood that
a family member will be engaged in nonfarm
employment. A similar, but insigni®cant, in¯uence is observed for distance to nearest secondary schoolÐintended to proxy distance to
the nearest town or settlement. Whether the
household has a power connection is strongly
signi®cant in increasing the likelihood that a
family member will be engaged in some form of
nonfarm employment. Certainly, home-based
activities such as tailoring, food preparation, or
carpentry are much more attractive if the
household is connected to the electricity grid.
In the FUSADES sample, dummy variables
for di€erent departments in El Salvador are
signi®cant only in the case of the department of
Sonsonate, La Libertad, San Salvador, La Paz,
and San Miguel. In these departments the
probability of nonfarm employment is signi®cantly higher than in the department of Morazan (in the east of the country). For all other
departments (with the exception of Ahuacha-
pan) the point estimate is positive, indicating a
particularly low probability of nonfarm employment in Morazan. These point estimates are
not, however, signi®cantly di€erent from zero.
In the second and third columns of Table 6
we consider the same speci®cation against a
binary dependent variable indicating, in turn,
whether the nonfarm job is a high-productivity
one or one which yields a return below the
average agricultural wage. The negative impact
of age is not signi®cant for high-productivity
jobs, while for low-productivity jobs, it seems
that household size is one factor increasing the
likelihood that a family member will seek an
outside job. For high-productivity jobs, the
e€ect of higher levels of education is particularly strong and positive, while for low-productivity jobs the education variables are all
insigni®cant. The per capita land variable and
cultivation dummy both remain negative and
signi®cant in the two respective cases. The access to infrastructure variables are broadly
similar, although in the high-productivity case,
the distance variables are not signi®cant, while
in low-productivity case, it is distance to the
nearest secondary school which is signi®cant
(once again indicating that nonfarm activities
appear to be more plentiful nearer to towns or
settlements). Connection to the electricity grid
is strongly signi®cant for both high and lowproductivity employment.
Of the regional dummies, Chalatenango and
San Salvador are the two departments in which
high-productivity jobs are concentrated. Relative to Morazan, low-productivity jobs are
more common in Sonsonate, Chalatenango, La
Libertad, San Salvador, Caba~
nas, Usulutan,
San Miguel, and La Union.
A somewhat puzzling ®nding from Table 6
was the observation that females were signi®cantly less likely to be employed in the nonfarm
sector than men. This ®nding is due partly to
the fact that in Table 6 the relevant domain was
taken to include all persons of working age in
rural areas. We have already noted that women
are far less likely to be ``economically active''
than men, as it is common practice to not include nonremunerated domestic activities
amongst ``economic'' activities. In Table 7 we
con®ne our attention to the ``economically active'' population, sticking with the FUSADES
data for the time being. Now, women are no
longer signi®cantly less likely to be employed in
nonfarm jobs. In fact, for low-productivity
jobs, women are signi®cantly more likely to be
employed in such occupations.
EL SALVADOR
539
Table 6. Probability of nonfarm employment as a primary occupationa
b
Probit model
Variable
Household
size
Female
Age (years)
Any nonfarm occupation
High-productivity job
Low-productivity job
Obs: 2738; at 1: 481; at 0: 2257
Obs: 2738; at 1: 331; at 0: 2407
Obs: 2738; at 1: 150; at 0: 2588
Prob valuec
Marginal
e€ect
Marginal
e€ect
)0.0004
0.8785
)0.003
0.1244
0.002
0.0790
)0.177
)0.001
0.0001
0.0137
)0.130
0.000
0.0001
0.3555
)0.028
)0.001
0.0001
0.0001
0.0002
0.0054
0.0001
0.0030
0.058
0.074
0.203
0.373
0.0003
0.0003
0.0001
0.0001
0.011
)0.001
)0.005
n/a
0.2287
0.9194
0.6280
)
)0.059
0.0001
)0.036
0.0017
)0.018
0.0359
)0.130
0.0001
)0.086
0.0001
)0.026
0.0016
)0.002
0.0979
)0.001
0.2067
)0.001
0.2976
)0.002
0.3041
)0.001
0.4751
)0.002
0.0080
0.044
0.0021
0.024
0.0359
0.012
0.0956
)0.011
0.058
0.152
0.008
0.127
0.247
0.065
0.094
0.086
0.032
0.073
0.116
0.085
0.176
0.7940
0.2218
0.0041
0.8722
0.0102
0.0001
0.2295
0.0744
0.1448
0.5800
0.1469
0.0226
0.0893
)0.023
)0.006
0.057
)0.032
0.055
0.114
0.033
0.065
)0.029
)0.010
0.022
0.059
0.009
0.121
0.3698
0.8521
0.1382
0.0061
0.1247
0.0061
0.4090
0.1067
0.4513
0.8242
0.5379
0.1187
0.7963
0.065
0.139
0.167
0.118
0.128
0.204
0.064
0.039
0.212
0.095
0.113
0.102
0.156
0.055
0.1894
0.2261
0.0102
0.0591
0.0258
0.0035
0.2259
0.4100
0.0068
0.1206
0.0558
0.0655
0.0148
Marginal e€ect
Education (highest level reached)
Primary
0.070
Middle school
0.066
High school
0.179
Tertiary level
0.273
Per capita
land
Cultivating
HH.
Distance to
road
Distance to
school
Electricity
connec.
Ahuachapan
Santa Ana
Sonsonate
Chalatenango
La Libertad
San Salvador
Cuscatlan
La Paz
Caba~
nas
San Vicente
Usulutan
San Miguel
La Union
Observed
probability
Predicted
probability
0.131
Log likelihood
Model
)1058.84
Constant
)1272.55
LR test
427.42
(model)
Degrees of
25
freedom
2
37.65
Critical
0.081
Prob. value
Prob. value
0.035
)833.34
)1009.49
352.30
)515.92
)581.47
131.10
25
25
37.65
37.65
a
Source: Rural survey, FUSADES (1996).
Domain: Entire rural population aged above 14.
c
A prob value of 0.05 indicates that with 95% con®dence one can reject the hypothesis that the parameter estimate is
zero.
b
In Table 8, we retain our focus on the economically active population but apply the signi®cantly larger EHPM data set to roughly the
same speci®cation. The only di€erence is that we
are unable to include the same infrastructure
access variables. In this data set, women are
540
WORLD DEVELOPMENT
Table 7. Probability of nonfarm employment as a primary occupationa
Any nonfarm occupation
b
Probit model
Obs: 1592; at 1: 481; at 0: 1111
Variable
Household
size
Female
Age (years)
Marginal e€ectc
Low-productivity job
Marginal
e€ect
Prob. value
Marginal
e€ect
Prob. value
0.003
0.6109
)0.004
0.2507
0.004
0.0397
0.031
)0.002
0.3025
0.0087
)0.031
0.001
0.1720
0.3593
0.045
)0.002
0.0025
0.0001
level reached)
0.101
0.097
0.0018
0.0147
0.083
0.107
0.0022
0.0020
0.013
)0.004
0.3841
0.8135
0.330
0.562
0.0001
0.0007
0.342
0.673
0.0001
0.0001
)0.007
n/a
0.6875
)0.052
0.0456
)0.032
0.1223
)0.014
0.3090
)0.336
0.0001
)0.235
0.0001
)0.070
0.0001
)0.003
0.1322
)0.002
0.2524
)0.001
0.4263
)0.002
0.5729
0.002
0.2675
)0.003
0.0148
0.094
0.0002
0.047
0.0220
0.022
0.0588
)0.065
0.007
0.132
)0.018
0.125
0.329
0.034
0.130
0.159
0.096
0.093
0.217
0.181
0.302
0.3839
0.9287
0.1048
0.8423
0.1149
0.0002
0.7013
0.1456
0.1079
0.3452
0.2751
0.0124
0.0396
)0.067
)0.053
0.027
)0.077
0.042
0.141
0.009
0.103
)0.048
0.002
0.021
0.116
0.029
0.208
0.2163
0.3270
0.6580
0.2014
0.4851
0.0418
0.8907
0.1559
0.4893
0.9827
0.7429
0.0947
0.6541
0.085
0.175
0.201
0.200
0.167
0.267
0.081
0.058
0.352
0.191
0.172
0.177
0.281
0.095
0.2917
0.0651
0.0390
0.0652
0.0640
0.0134
0.3355
0.4606
0.0068
0.0877
0.0802
0.0670
0.0130
Education (highest
Primary
Middle
school
High school
Tertiary level
Per capita
land
Cultivating
HH
Distance to
road
Distance to
school
Electricity
connec.
Ahuachapan
Santa Ana
Sonsonate
Chalatenango
La Libertad
San Salvador
Cuscatlan
La Paz
Caba~
nas
San Vicente
Usulutan
San Miguel
La Union
Observed
probability
Predicted
probability
Prob. value
High-productivity job
Obs: 1592; at 1: 331; at 0: 1261 Obs: 1592; at 1: 150; at 0: 1442
0.244
Log likelihood
Model
)739.82
Constant
)975.36
LR test
471.08
(model)
Degrees of
25
freedom
2
37.65
Critical
0.155
0.055
)649.27
)813.80
329.06
)414.86
)497.02
164.32
25
25
37.65
37.65
a
Source: Rural survey, FUSADES (1996).
Domain: Rural population aged above 14 and engaged in remunerated work.
c
A prob. value of 0.05 indicates that with 95% con®dence one can reject the hypothesis that the parameter estimate is
zero.
b
signi®cantly more likely to be employed in the
nonfarm sector irrespective of whether the jobs
are high- or low-productivity ones. As in Table 7,
it appears that the greatest probability is women
being employed in the low-productivity occupations, controlling for all other characteristics.
EL SALVADOR
541
Table 8. Probability of nonfarm employment as a primary occupationa
b
Probit model
Any nonfarm occupation
Obs: 2914; at 1: 1035; at 0: 1879
Variable
Household
size
Female
Age (years)
Marginal e€ect
Low-productivity job
Marginal
e€ect
Prob value
Marginal
e€ect
Prob value
)0.006
0.0990
0.000
0.9737
)0.005
0.0508
0.495
0.001
0.0001
0.0300
0.073
0.001
0.0001
0.0190
0.367
0.000
0.0001
0.5350
0.0001
0.0001
0.111
0.477
0.0001
0.0001
0.012
0.008
0.3813
0.8133
0.0001
0.847
0.616
0.0001
0.0001
)0.114
)0.072
0.0292
0.0476
Education (highest level reached)
Primary
0.136
Middle
0.487
school
High school
0.569
Tertiary level
n/a
Per capita
land
Cultivating
HH
Ahuachapan
Santa Ana
Sonsonate
Chalatenango
La Libertad
San Salvador
Cuscatlan
La Paz
Caba~
nas
San Vicente
Usulutan
San Miguel
La Union
Observed
probability
Predicted
probability
Prob valuec
High-productivity job
Obs: 2914; at 1: 544; at 0: 2370 Obs: 2914; at 1: 491; at 0: 2423
)0.001
0.0070
)0.000
0.1848
)0.0002
.0858
)0.112
0.0001
)0.058
0.4363
)0.035
0.0439
0.140
0.203
0.339
0.313
0.0718
0.0071
0.0001
0.0001
0.060
0.072
0.168
0.185
0.3416
0.2431
0.0137
0.0112
0.090
0.129
0.198
0.138
0.1193
0.0258
0.0016
0.0293
0.266
0.492
0.263
0.347
0.017
0.207
0.053
0.154
0.197
0.351
0.0005
0.0001
0.0018
0.0001
0.8414
0.0096
0.5035
0.0564
0.0158
0.206
0.399
0.130
0.242
)0.002
0.126
0.027
0.030
0.056
0.187
0.0032
0.0001
0.0742
0.0007
0.9718
0.0697
0.6636
0.6416
0.3939
0.077
0.101
0.144
0.116
)0.000
0.088
0.034
0.123
0.132
0.168
0.1634
0.0766
0.0294
0.0436
0.9944
0.1361
0.5418
0.0478
0.0366
0.319
Log likelihood
Model
)1401.40
Constant
)1895.83
LR test
988.86
(model)
Degrees of
22
freedom
2
33.92
Critical
0.147
0.131
)1080.48
)1402.78
644.60
)1131.35
)1321.48
380.26
22
22
33.92
33.92
a
Source: Encuesta de Hogares de Propositos Multiples, MIPLAN, 1994-III.
Domain: Rural population aged above 14 and engaged in remunerated work.
c
A prob value of 0.05 indicates that with 95% con®dence one can reject the hypothesis that the parameter estimate is
zero.
b
Age is positively associated with nonfarm employment, particularly for the high-productivity
jobs. As in the previous tables, education is
positively associated with high-productivity
nonfarm jobs, but not with low-productivity
jobs. The e€ect of the land-ownership variable
and cultivating dummy is unchanged.
Among the regional dummies, Sonsonate,
Chalatenango, La Libertad, San Salvador, and
La Paz are strongly associated with high-productivity jobs (relative to Morazan) while lowproductivity jobs appear particularly numerous
in Santa Ana, Sonsonate, Cuscatlan, La Paz,
San Miguel, and La Union. The combination
542
WORLD DEVELOPMENT
of ®ndings from these regressions suggests that
nonfarm employment opportunities are clustered principally in the departments of Sonsonate, Chalatenango, La Libertad, San
Salvador, and La Paz. The more remote departments, particularly those in the northeast
and the far west appear to be less well-served. It
is interesting to note that in Table 8 more of the
regional dummies are statistically signi®cant
than in Table 7, and in absolute value the parameter estimates are generally higher. This
could be due to the smaller sample size of the
FUSADES survey, but is also likely to be the
consequence of being able to control better for
infrastructure variables in Table 7. The dummies in Table 8 are therefore likely to be
proxying, at least in part, for di€erential access
to infrastructure, such that regional di€erences
diminish once infrastructure is controlled for.
In Table 9 we turn to the correlates of earnings from nonfarm employment, basing our
analysis once again on the FUSADES data. We
present results for an OLS regression which
includes an adjustment for sample selection
(Heckman, 1979). Simply estimating the relationship between household characteristics and
earnings over the subsample of persons engaged in nonfarm activities could result in biased estimates. This would be the case if such
persons di€ered in some fundamental way from
the population at large, and if these di€erences
led to them obtaining di€erent returns from
their characteristics.
To control for this sample selection issue we
use the Mill's ratio method (Heckman, 1979).
We ®rst estimate a Probit model for the probability an individual is employed in the nonfarm sector. We use this to construct a new
regressorÐthe inverse Mill's ratioÐwhich we
then include (in addition to the other explanatory variables) in the estimation of the earnings.
Since the Mill's ratio is basically a function of
the exogenous variables in the Probit model,
this corrects for potential selectivity or nonrandom sampling bias. The procedure involves
the selection of an identifying variable: one that
is related to the decision to work in the nonfarm sector, but does not a€ect the expenditure
outcome. The identifying variable used here is
the percentage of the working population employed in the nonfarm sector in the department
in which the individual resides. 15 The notion
Table 9. Non-agricultural labor earningsa
With adjustment for sample selection
OLS Modelb
Prob. valuec
Variables
Estimate
Constant
Household size
Female
Age (years)
8.968
)0.006
)0.344
0.007
0.0001
0.6338
0.0001
0.0179
Education (highest level achieved)
Primary
Middle school
High school
Tertiary
0.174
0.370
0.596
1.160
0.1345
0.0018
0.0001
0.0003
Per capita land
Cultivating household (dummy)
Distance from paved road
Distance from secondary school
Electricity connection
West
Central 1
Central 2
Mills ratio (1,00,000)
Adjusted R2
Number of observations
0.086
)0.624
)0.003
0.005
)0.079
)0.166
)0.064
)0.148
)0.180
0.1515
481
0.4222
0.0001
0.6521
0.5629
0.3086
0.1100
0.5066
0.2413
0.9334
a
Source: Rural Survey, FUSADES (1996).
Dependent variable: (log) annual nonfarm labor income; domain: all persons aged 14 and above with nonfarm
employment.
c
A prob. value of 0.05 indicates that with 95% con®dence one can reject the hypothesis that the parameter estimate is
zero.
b
EL SALVADOR
here is that this variable should in¯uence the
probability of an individual's employment in
the sector (as it re¯ects the availability of
nonfarm jobs), but that conditional on employment it is not clear why this factor should
in¯uence earnings. In the event, our inverse
Mill's ratio term is not statistically signi®cant,
indicating that the sample-selection issue is of
relatively minor concern in our setting.
Females earn signi®cantly less than men. A
woman can expect to earn 29% less than a man
holding all other characteristics constant. 16
Earnings are sharply higher with higher education levels. While a person with primary
schooling would expect to earn about 19%
more than an uneducated person from nonfarm
work (although this is not statistically signi®cant), a person with middle schooling would
receive about 45% more, a person with highschool level education would receive 79% more,
and a person with education at the tertiary level
would receive about 223% more.
Although the coecient is not signi®cant, in
this regression, a person from a wealthier
household would expect to earn more from
nonfarm employment than a person from a
household with less per capita land. If the
household is a cultivating one, then a family
member employed in some nonfarm job would
be earning about 47% less than one from noncultivating household. While the infrastructure
variables in¯uenced signi®cantly whether a
person was likely to be employed in the nonfarm sector, these variables do not seem to in¯uence earnings signi®cantly. Earnings, given
employment in a nonfarm occupation, are also
not signi®cantly in¯uenced by geographic location.
Before ending this section, we consider
brie¯y some of the factors that may in¯uence
the establishment of rural nonfarm enterprises.
We focused on the 101 rural enterprises covered
in the FUSADES survey to inquire into their
access to infrastructure, as well as any diculties they report with infrastructure services. It is
often argued that the key to rural development
is not so much the provision of state-of-the-art
components of infrastructure as ensuring access
to a basic package which combines simple, or
even rudimentary, communications, transport,
power, and water services (World Bank,
1994b).
About two-thirds of rural enterprises have
access to electricity. In El Salvador, very few
households without connection appear to provide for their electricity requirements via the
543
purchase of generators. In the FUSADES survey, no rural enterprise without an electricity
connection reported such private provision of
electricity. Water provision was most common
via private sources such as a well or river, although nearly 47% of enterprises were served
by the public network. Of all rural enterprises,
nearly 19% reported shortages of water during
at least part of the year. Very few rural enterprises have a telephone connection, and 35% of
enterprises reported transport-related diculties associated with the poor state of road infrastructure. Thus, for most infrastructure
sectors a sizeable fraction of rural enterprises
report not having access to the services, or
having problems with the service. While the
situation in rural El Salvador is perhaps not as
critical as in other, larger and less densely
populated countries, it does appear that access
to rural infrastructure is far from perfect. Per
capita costs of rural infrastructure provision, in
a country such as El Salvador, are presumably
not nearly as high as they would be in other,
less densely populated, countries.
Table 10 indicates that the vast majority of
rural nonagriculture enterprises report obtaining their start-up capital from personal savings.
In fact, only 7% of enterprises were originally
®nanced through formal sector credit sources.
These ®ndings could imply that there are serious problems of credit supply in rural areas. It
is often suggested that imperfect information,
high transactions costs, and poorly delineated
property rights (in¯uencing the ability of borrowers to o€er land as collateral) make it dif®cult for ®nancial institutions to o€er credit to
small borrowers in rural areas. Another possibility, however, is that demand for formal
credit is low. It has been suggested, for example, that the poor functioning of rural ®nancial
markets in developing countries extends also to
a inadequate formal sector savings mobilization. In such circumstances rural households
may ®nd themselves unable to earn positive
returns by depositing or investing surpluses
with ®nancial intermediaries. This might leave
them with no alternative other than to invest
their savings in home-enterprise activities (Vijverberg, 1988, and Banerjee, 1996). Hence, the
very fact that formal credit institutions are
imperfect, might lead to more rather than less
investment in small home-based enterprises.
Reform of ®nancial markets in rural areas
should thus not only concentrate on facilitating
the ¯ow of funds for investment purposes but
also focus on mobilizing savings and o€ering
544
WORLD DEVELOPMENT
Table 10. Rural enterprises and start-up ®nancea
Principal source of start-up ®nance (percentage of ®rms)
Sector
Number of
®rms
Transport
Other services
Other industry
Repair shop
Restaurant/bar
Textiles
Wood/work
Food proc.
Pottery/bricks
Commerce
Total
a
Number of
workers
Personal
savings (%)
Friends and
relatives (%)
Informal
sources (%)
Formal
sources (%)
1
3
5
6
5
13
10
13
14
31
1
3
6
16
19
25
37
53
63
77
100
67
100
100
60
85
30
46
85
68
0
33
0
0
20
0
30
8
7
16
0
0
0
0
0
15
40
23
7
6
0
0
0
0
20
0
0
23
0
10
101
300
70
11
12
7
Source: Rural survey, FUSADES (1996).
real returns on such savings. The net impact of
such reforms on the number of rural nonfarm
®rms is not clear, but average productivity
levels would be expected to rise.
6. CONCLUSIONS
The preceding analysis has demonstrated
that the nonfarm sector in El Salvador, while
not the dominant sector in rural areas, is important in terms of both employment rates as
well as incomes. Roughly 36% of the economically active population in rural El Salvador is
employed in the nonfarm sector. Women are
particularly highly represented among the economically active population employed in this
sector; nearly three-quarters of all such women
are employed in the nonfarm sector. Of course,
the proportion of women in the conventionally
de®ned ``economically active'' population is
only about a third that of men.
The range of activities in which the rural
population is engaged extends from manufacturing to services. A substantial number of
nonfarm activities are linked to agriculture in
the sense that they feed into agricultural production, transport or transform agricultural
goods, or provide goods and services which
agricultural households are likely to be purchasing. There is also a sizeable manufacturing
component that is not obviously linked to agricultural production. Here, household ®rms
are engaged in subcontracting relationships
with large, probably urban-based, ®rms. Because these ®rms are not closely linked to the
agricultural sector, the incomes derived from
such activities may introduce an important
source of liquidity into rural areas which does
not covary with agricultural incomes, much the
same as remittance incomes from abroad.
The analysis has shown that rural poverty in
El Salvador is acute, with some evidence that
extreme poverty (de®ned in the context of this
study as the incidence of poverty associated
with the lower poverty line) is higher in the
relatively more remote eastern and western
regions of the country. The evidence suggests
that while the rural poor are highly represented among agricultural laborers and marginal farmers in particular, there is some
suggestion that at least some nonfarm activities are also of importance to the poor. These
activities are in the nature of ``last resort''
activities and are not associated with high
levels of labor productivty. Although it is
dicult to view this subsector of the nonfarm
sector as o€ering great prospects of lifting the
poor above the poverty line, it is important to
recognize that even these activities have a
considerable distributional impact: they prevent the poor from falling into even greater
destitution. Policy makers should be alert to
this function of these nonfarm activities, and
refrain from taking measures that would undermine this safety-net role.
The nonfarm sector could also present a
potent force for upward mobility. The least
poor in rural areas are households that are
heavily engaged in the nonfarm sector. An
important challenge is to increase access of the
poor to nonfarm activities that yield these high
and stable incomes. While data constraints
preclude making strong statements about cau-
EL SALVADOR
sality, the analysis in this paper points to a
number of areas to which further attention may
be directed when addressing the design of policies.
First, there is strong evidence to suggest that
higher income nonfarm jobs go to those with
higher levels of education. While it is not the
case that the uneducated are unable to secure
any nonfarm employment, the kind of jobs
obtained tend to be low-productivity activities,
which while no doubt of great value in relieving
poverty somewhat, do not, in all likelihood,
raise the standard of living of these households
markedly. As individuals become better educated, their earnings rise sharply. Having said
this, it is important to point out that at a given
level of education, women tend to earn signi®cantly less then men from nonfarm employment.
Infrastructure services appear to exercise
signi®cant in¯uence on the likelihood of ®nding nonfarm employment. Persons residing in
remote, inaccessible areas are far less likely to
®nd employment in the nonfarm sector.
Households connected to the electricity network are far more likely to have family
members engaged in the nonfarm sector. It is
not certain whether the electricity services
themselves stimulate nonfarm activities, or
whether connection to the electricity network
acts as a proxy for location near a large(ish)
conurbation, where nonfarm jobs are likely to
be relatively more common. Rural enterprises
in El Salvador are fairly poorly serviced by
infrastructure services. In the FUSADES rural
survey, only 7% of ®rms had a telephone
connection, more than one-third of all ®rms
had no access to electricity, more than a third
of all ®rms reported transport-related diculties, and more than half were reliant on their
own sources of water (and a sizeable fraction
reported periods of non-availability). While
the cost of infrastructure provision in rural
areas is higher than in urban areas, in El
Salvador the rural population is fairly dense
which would help to reduce per capita costs.
As argued for example in World Bank (1994b)
the focus of infrastructure providers should
perhaps be on a basic package of simple rural
infrastructure services rather than on the introduction of highest quality (but high-cost)
infrastructure.
Regionally, the distribution of nonfarm occupations is not uniform. The corridor of
Sonsonate, La Libertad, Chalatenango, San
Salvador, and La Paz appear to contain the
545
highest concentration of nonfarm activities,
including in particular the high-productivity
jobs which represent the greatest source of
upward mobility. The remote departments of
Morazan and Ahuachapan appear to be particularly poorly endowed with nonfarm activities. It is an important priority to try to
understand better what factors can help to
induce the emergence of more nonfarm activities in these relatively under-served departments.
A very low percentage of rural enterprises
report having obtained ®nancing from formal
sector credit sources in setting up their enterprises. The overwhelmingly dominant source of
®nance came from personal services. This implies, ®rst of all, that the poorest in rural areas
are not those most likely to set up rural enterprises (although, of course, they may ®nd employment in such ®rms). It might also tell us
something about the availability of rural savings facilities. The emphasis in rural ®nancial
reform is often on increasing the availability of
credit in rural areas. The foregoing comments
indicate that attention might also be usefully
focused on improving rural savings mobilization.
It is interesting to note that in rural El Salvador, very few enterprises report having bene®ted from special training in conducting their
a€airs. This is despite there being a substantial
experience in El Salvador with training and
special programs for microenterprises. It thus
seems important to consider focusing additional e€orts to provide such assistance to ®rms
in rural areas.
The observation that at least a fraction of
the rural nonfarm activity involves subcontracting arrangements between home-based
®rms and larger ®rms merits further study.
There is already a sizeable experience with
``Maquila'' industries in El Salvador. These are
generally large-scale undertakings in which an
industrial park is created, providing a whole
package of infrastructure and logistical services, and foreign investors are then invited to
locate their assembly plants there. The smaller,
more localized ``subcontracting'' model resembles this approach in spirit, although it is of
course at a much smaller scale. Important
outstanding questions surround just what
minimal package of infrastructure, training,
and ®nance is necessary to give further stimulus to this approach. Clearly, one unmistakable
prerequisite is a measure of security and stability in rural areas.
546
WORLD DEVELOPMENT
NOTES
1. Aoki et al. (1995, p. 40) argue that the East Asian
success in utilizing cheap labor in rural areas, in sectors
outside of traditional farming, was ``one of the most
important elements of East Asian development'' (see
also Hayami, 1995).
10. We emphasize here that we are not making any
statements about directions of causality. While nonfarm
activities could serve as a means to lift the poor above
the poverty line, it is also possible that one must ®rst
attain a certain degree of a‚uence before gaining access
to nonfarm jobs.
2. The EHPM contains household weights that allow
one to aggregate up to population totals. Although the
sample frame was based on the Census from the early
1970s, the household weights have reportedly been
adjusted with reference to the 1992 Labor Force Census,
so adding up to population should be valid (personal
communication from the Director of the Departamento
de Investigaciones Muestrales).
11. For further discussion of this point, see Lanjouw
and Lanjouw (forthcoming).
3. Lanjouw and Lanjouw (forthcoming) provide a
recent survey of the literature.
4. See Elbers and Lanjouw (2001) for evidence in
Ecuador that an expanding nonfarm sector reduces
absolute poverty while it increases inequality.
5. Lanjouw (1995) tries to control for land quality
di€erences by holding crops constant. In any case,
anecdotal evidence for Ecuador would suggest that large
farmers possess land of better quality than small
farmers.
12. Although it should be stressed, again, that we are
not able to dismiss the converse conjecture: that only
only the nonpoor are in some sense able to ``a€ord'' to
be represented in the nonfarm sector. Once again, care
must be taken not to infer causality from the simple
statistical associations presented here.
13. De Janvry and Sadoulet (1993) suggest that such
linkages might not be so important in Latin America as
a whole. With a highly skewed distribution of land and
income, a few landowners bene®t from the bulk of the
income e€ects of agricultural growth, and these landowners are often absentee and therefore do not demand
locally produced goods. This point indicates that one of
the additional bene®ts of land reform may be the
stimulus that is given to the local nonfarm economy.
7. In contrast, the poor often have myriad income
sources, as they try to bring together revenues from all
kinds of activities to be able to meet their subsistence
needs.
14. These indicate the change in the probability of
nonfarm employment associated with a marginal change
in given explanatory variable, when all explanatory
values are taken at their respective means. In the case of
a dummy explanatory variable the calculation is in terms
of the change in probability associated with the dummy
variable changing in value from zero to one.
8. Note that the cost-of-living adjustment mentioned
above, reduces the gap between Metropolitan San
Salvador and the rest of the country. Failure to make
such an adjustment would have resulted in an even wider
gap.
15. Using this as our identifying variable implies that
we cannot use the departmental regional dummies in the
second stage OLS regression. To capture spatial e€ects
in the second stage we thus include dummies at the
regional level (see Table 1).
9. ``Economically active'' is de®ned rather restrictively,
it refers to activities which receive some kind of direct
monetary remuneration, and thereby excludes domestic
activities within the household, or nonremunerated
activities on the family farm or enterprise.
16. A coecient c multiplying a dummy variable can
be interpreted as a percentage change in the endogenous
variable only as long as c is close to zero. For larger
values, in absolute terms, the percent change in the
endogenous variable is given by 100‰exp…c† 1Š.
6. See Ravallion (1994) for a good survey of these
issues.
REFERENCES
Aoki, M., Murdock, K., & Okuno-Fujiwara, M. (1995).
Beyond the East Asian miracle: Introducing the
market-enhancing view. In M. Aoki, H. Kim, & M.
Fujiwara, The role of government in East Asian
economic development: Comparative institutional
analysis. Manuscript, Economic Development Institute, the World Bank.
EL SALVADOR
Banerjee, A. (1996). Notes towards a theory of industrialization in the developing world. MIT: Unprocessed.
Datt, G., & Ravallion, M. (1993). Transfer bene®ts from
public works employment: bene®ts for rural India.
Economic Journal, 104(427), 1346±1369.
De Janvry, A., & Sadoulet, E. (1993). Rural development in Latin America: Relinking poverty reduction
to growth. In M. Lipton, & J. van der Gaag,
Including the poor (pp. 249±305). Washington, DC:
The World Bank.
Dreze, J. P. (1995). Famine prevention in India. In A. K.
Dreze, J. Sen, & A. Hussain (Eds.), The political
economy of hunger: Selected essays (pp. 69±177).
Oxford: Oxford University Press.
Dreze, J. P., Lanjouw, P., & Stern, N. H. (1992).
Economic mobility and agricultural labour in rural
India: A case study. Indian Economic Review (Special
number in Memory of Sukhamoy Chakravarti), Vp;
XXVII, 25±54.
Elbers, C., & Lanjouw, P. (2001). Intersectoral transfer,
growth, and inequality in rural Ecuador. World
Development, 29(3), 481±496.
FUSADES (1993). Empleo, ingreso y pobreza rural.
Boletin Economico y Social, 91.
Hayami, Y. (1995). In search of rural entrepreneurship in
Asia: Concept and approach. Paper presented at a
World Bank Conference on Rural Development in
Latin America, September.
Hazell, P., & Haggblade, S. (1993). Farm±nonfarm
growth linkages and the welfare of the poor. In M.
Liptan, & J. Van der Gaag (Eds.), Including the poor.
Washingtan, DC: The World Bank.
Heckman, J. (1979). Sample selection bias as a speci®cation error. Econometrica, 47, 153±161.
IIES-UCA (1993). La reduccion de la pobreza: Comentarios al informe de miplan. Unprocessed. IIES-UCA.
Johnston, B., & Kilby, P. (1975). Agriculture and
structural transformation: Economic strategies in late
developing countries. London: Oxford University
Press.
Klein, E. (1993). El empleo rural no-agricola en America
Latina. In CEPLAES, Latinoamerica Agraria Hacia
el Siglo XXI. Quito, Ecuador: CEPLAES.
547
Lanjouw, P. (1995). Ecuador, Pobreza rural. Cuestiones
Economica, Banco Central del Ecvador, No. 27,
December.
Lanjouw, J. O., & Lanjouw, P. (1996). Poverty comparisons with non-compatible data: theory and illustrations. Unprocessed. Policy Research Department, the
World Bank.
Lanjouw, J. O., & Lanjouw, P. (forthcoming). The rural
nonfarm sector: Issues and evidence from developing
countries. Agricultural Economics.
Lopez, R. (1996). Rural poverty in El Salvador: A
quantitative analysis. Unprocessed. Department of
Agricultural and Resource Economics, University of
Maryland at College Park.
Mellor, J., & Lele, U. (1972). Growth linkages of the
new food grain technologies. Indian Journal of
Agricultural Economics, 18(1), 35±55.
Mellor, J. (1976). The new economics of growth: A
strategy for india and the developing world. Ithaca,
NY: Cornell University Press.
MIPLAN (1995a). El Salvador: Evaluacion de los factores de persistencia de la pobreza en los hogares
pobres. Unprocessed. Ministerio de Plani®cacion,
San Salvador.
MIPLAN (1995b). Principales resultados: Encuesta de
hogares de propositos multiples. Unprocessed. MIPLAN and Ministerio de Relaciones Exteriores, San
Salvador.
Murdoch, J. (1995). Income smoothing and consumption smoothing. Journal of Economic Perspectives,
9(3), 103±114.
Ravallion, M. (1994). Poverty comparisons. Chur: Harwood Press.
Vijverberg, W. (1988). Pro®ts from self-employment.
LSMS Working Paper No. 43, the World Bank.
World Bank. (1994a). El Salvador: The challenge of
poverty alleviation. Report No. 12315-ES, the World
Bank.
World Bank (1994b). Infrastructure for development:
World development report 1994. New York: Oxford
University Press.