Directory UMM :Data Elmu:jurnal:UVW:World Development:Vol29.Issue3.2001:
World Development Vol. 29, No. 3, pp. 455±465, 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)00106-6
Rural Nonfarm Employment and Income
Diversi®cation in Colombia
KLAUS DEININGER and PEDRO OLINTO *
The World Bank, Washington, DC, USA
Summary. Ð Using data for rural households from Colombia we ®nd that o-farm employment
contributes a signi®cant share (45% on average) to household income but that the importance of
o-farm income and returns to household labor vary over the income distribution. Analysis reveals
signi®cant gains from specialization butÐfor households able to specializeÐno systematic
dierences in returns to labor between farm and nonfarm sources. We conclude that, in Colombia,
there is no con¯ict between development of the farm and the nonfarm sector but that, to maximize
gains from nonfarm development and reduce the scope for undesirable distributional consequences,
policies enabling households to specialize might be called for. Ó 2001 Elsevier Science Ltd. All
rights reserved.
Key words Ð Colombia, o-farm employment, income distribution
1. INTRODUCTION
There is little doubt that the importance of
rural nonfarm employment, which in many
countries already constitutes an important
sector of the rural economy, will greatly
increase as agriculture becomes more and more
integrated into global markets and as the links
between urban and rural areas intensify. What
is less clear, however, is how these forces for
diversi®cation can best be harnessed for
nonfarm employment to act as a catalyst for a
broader and inclusive pattern of development.
From a policy perspective, it is of particular
interest to ®nd out whether the rural poor are
able to make optimum use of the opportunities
provided by nonfarm employment or whether
speci®c policy measures to assist them might be
needed.
In this paper, we use data from Colombia to
address this question. Descriptive statistics
illustrate both the importance of nonfarm
employment and broad patterns of participation in nonfarm opportunities across dierent
groups of the population. Nonfarm income
(wages from agricultural and nonagricultural
employment, pro®ts from nonagricultural
enterprises, nonearned income, and remittances) contribute on average 45% to household income. There is also a nonlinear
(U-shaped) relationship between the importance
of o-farm work, asset endowments, and total
455
household income. By comparison, specialization (in either farm or nonfarm activities)
increases linearly with income and assets.
The strong positive association between total
income and specialization suggests that even
though nonfarm employment contributes to
diversi®cation of income generating opportunities at the regional level, individual households may still be better o by relying only on
one main income source. More importantly, to
the extent that market failures and lack of
endowments prevent them from specializing,
government policies that improve the functioning of factor markets and that help households increase their endowments could play an
important role to maximize the gains associated
with the emergence of nonfarm employment
opportunities.
To explore this in more detail, we examine
the impact of specialization and household
labor supply as well as the determinants of
specialization. We ®nd that, indeed, specialization allows households to increase their level
* We
are deeply indebted to Elsa Albarracin, Juliana
Bottia, Diana Grudzynski, Absalon Machado, Manuel
Rojas, Hernando Urbina and Guilermo Otanez without
whose enthusiasm and support the data underlying this
analysis would never have been collected. We are also
grateful for useful comments from three anonymous
reviewers.
456
WORLD DEVELOPMENT
of welfare (as measured by expenditures) by
between 10% and 36%, everything else equal.
Why, then, do not all households choose to
specialize in one activity? We ®nd that imperfections in markets for credit and land, lack of
education, and inequalities in asset ownership
constitute important barriers to increased
specialization.
From a policy perspective this implies that in
situations such as Colombia, where education
and assets are distributed in an unequal
manner, the impact of increased nonfarm
employment opportunities will not be independent from households' and communities
initial endowments. Households with little
human or physical capital may be forced to rely
on nonfarm employment as a low return ``refuge'', comparable to semi-subsistence, with
little prospect for economic advancement. Only
if they own suciently high levels of assets or
are able to access to credit and land rental
markets will households be able to make full
use of the opportunities for specialization and
increased returns to labor provided by a more
diversi®ed rural nonfarm economy. The
welfare-enhancing impact of nonfarm employment opportunities will thus be maximized if
policies aimed at promoting the rural nonfarm
sector are complemented with measures to
improve the functioning of factor markets and
to increase households' opportunities to accumulate human and physical capital.
The paper is structured as follows. Section 2
provides descriptive evidence on the importance and incidence of nonfarm employment
across the income distribution and the country's dierent regions as well as a description of
the data sources underlying the study. Section 3
discusses the main econometric results, in
particular the impact of specialization on
household welfare and the determinants of
households' decision to specialize. Section 4
links the results to the broader discussion of
nonfarm employment in the literature and, on
this basis, derives a number of conclusions for
policy as well as research.
2. INCIDENCE AND CHARACTER OF
NONFARM EMPLOYMENT
In this section we describe the data underlying the analysis and discuss descriptive statistics
about the incidence and nature of nonfarm
employment in rural Colombia. We ®nd an Ushaped relationship between the share of o-
farm income and household assets or total
income level. This is consistent with the evidence from a number of other countries where
presence of entry barriers to high-return jobs in
the o-farm sector, together with a relatively
unequal distribution of farm assets and
malfunctioning land rental markets, force poor
people with scant asset endowments into lowpaying o-farm jobs and prevent them from
taking maximum advantage from the opportunities oered by nonfarm employment.
(a) Background and data sources
Thanks to a large amount of studies on the
nonfarm sector across all continents, the
importance of rural nonfarm employment is
now widely recognized. Country case studies
illustrate that the share of nonfarm income in
total household income ranges between about
30% and 40%Ðwith the highest shares (45%)
reported from Africa and the lowest ones (29%)
from South Asia (Reardon et al., 1998).
Although household-level evidence on the
evolution of nonfarm employment is limited, 1
the contribution of nonfarm income sources
and o-farm employment to the rural economy
has grown substantially during the last two
decades and is likely to continue doing so in
view of globalization, progressive insertion of
rural areas into the larger economy, and
increased access to public services (Berdegue,
Reardon, & Escobar, 2000).
While our understanding of the magnitude of
the rural nonfarm sector has greatly improved,
the contribution of this sector to household
welfare, and the distribution of the bene®ts
from o-farm employment across the population, are still imperfectly understood. Evidence
on whether nonfarm employment contributes
to a more equal distribution of income is
decidedly mixed (Reardon, Taylor, Stamoulis,
Lanjouw, & Balisacan, 2000). To advance on
this issue, it is necessary to explore not only
what determines participation in the rural
nonfarm economy, but also how such participation aects households' welfare. This is
crucial not only for academic reasons but, more
importantly, to allow governments to take
measures that will enable the poor to take
advantage of the opportunities inherent in the
growing importance of the nonfarm sector,
thus turning it into a catalytic force for rural
growth and sustainable reduction of poverty.
Colombia is of interest for this issue in a
number of respects. In marked dierence to
COLOMBIA
the ups and downs of other Latin American
countries, the country had, until the recent
upsurge in violence and macroeconomic
problems, been characterized by stable
economic growth. At the same time, it shares
with other Latin American economies a highly
unequal distribution of assets. 2 Maldistribution of assets is particularly acute in rural
areasÐdespite more than three decades of
land reform, land access is highly unequal with
the Gini coecient of land ownership in 1990
estimated to be 0.81 (World Bank, 1996).
Other assets are distributed in a slightly less
unequal fashion, with a Gini of 0.77. This is
relevant for the rural population as agriculture
is still the single most important sector in the
economy, generating a ®fth of total value
added, over a third of foreign exchange, and
more than 30% of total employment in the
economy.
Starting in the early 1990s, the country
implemented a far-reaching program of
adjustment (apertura) which, by turning away
from a paradigm of import-substituting industrialization, opened up the agricultural sector
to the forces of international competitiveness.
This led to large gains for producers who were
well connected to markets and able to adjust
quickly to the changed system of incentives. At
the same time, it tended to reinforce old
dichotomies in the distribution of assets as
small producers who were not able to shift out
of traditional commodities suered considerable losses. Migration, together with rapid
growth of the rural nonfarm sector enabled
rural dwellers to improve or at least stabilize
their income in the face of these external shocks
(Jaramillo, 1998).
To identify whether nonfarm employment
can, in addition to constituting a safety net,
also act as a catalyst for an inclusive pattern of
economic development in the rural sector, we
use data from a survey of about 1,000 rural
households that was undertaken by the
Departamento Nacional de Planeacion (DNP) in
collaboration with IICA and the World Bank.
The main purpose of the survey was to examine
factors aecting technical eciency of dierent
farm sizes, the functioning of rural factor
markets, and sources of income and employment of the rural population. It contains
comprehensive information on labor use,
general household characteristics, asset
endowments, migration, and access to government services which can provide a better
understanding of the rural nonfarm economy. 3
457
(b) Descriptive evidence
The survey data reveal that Colombia's rural
inhabitants draw incomes from a wide variety
of sources. As illustrated in column 1 of
Table 1, farm pro®ts made up 56% of total
income, complemented by wage income from
farm and nonfarm sources (30%), nonfarm
enterprise pro®ts and nonearned income
(12.5%), and migration remittances (2.5%). The
average household in the survey had slightly
less than ®ve members with the head having
completed 2.9 years of schooling, compared to
a mean of 3.9 for all household members over
the age of 15 years. Households' mean asset
endowment amounted to about 25 ha of land
and business assets (including machinery, livestock, vehicles, and nonfarm enterprise assets)
worth about US$4,500. In line with what is
known from other sources of information, our
data point toward an unequal distribution of
assets. About 13% of the sample have relatives
who migrated out and may have provided
remittance income.
Individuals' wage rates and thus opportunities
in the farm and nonfarm sector vary depending
on their level of education, physical location,
and type of work performed. To account for
this, we complement information on income by
source with data on the amount of hours worked
in dierent economic activities. This indicates
that 62% of the 70 weeks per year which the
typical households spent working was used in
agricultural activities, 25% in wage labor, and
10% in independent nonfarm enterprises.
The information on credit and savings
provided by the survey points toward limited
access to ®nancial infrastructure and a reluctance to use credit, rather than household
savings, to ®nance investment. One-quarter of
households had pre-existing savings, but only
15% used credit. Of these, 10% was from the
formal and about 5% from the informal
sectorÐwhich comprised traders (3.5%) and
informal lenders (2.6%). Half of the households
did not use credit because of high rates or
complicated documentation, while another 14%
reported not to need credit and 7% did not have
collateral. Access to free technical assistance
was, with 33%, fairly widespread. 4
(c) Nonfarm employment, asset ownership, and
specialization
The literature has long emphasized the relative importance of ``pull'' and ``push'' factors as
458
WORLD DEVELOPMENT
Table 1. Descriptive statistics by quintile of the per capita expenditure distribution
Quintiles of per capita expenditure
Total
1
2
3
4
163.01
1254.01
39.90%
40.59%
15.42%
4.09%
284.44
1979.03
49.28%
36.16%
11.74%
2.82%
405.21
2264.34
59.09%
27.06%
11.54%
2.30%
587.26
2974.33
65.33%
24.83%
8.11%
1.73%
1226.19
4167.17
55.73%
27.51%
15.73%
1.03%
4.71
2.89
3.96
12.47%
6.20
2.12
3.27
16.74%
5.36
2.80
3.78
15.35%
4.52
2.96
4.04
11.16%
4.20
3.09
4.13
11.16%
3.28
3.47
4.56
7.91%
24.55
4447.97
234.28
51.53%
36.51
69.23
43.75
6.84
748.73
117.74
38.60%
17.21
72.87
39.95
17.49
3657.34
178.65
48.37%
26.32
75.19
40.29
20.17
3240.90
199.78
49.77%
32.21
70.31
46.54
31.70
5947.15
275.99
56.74%
44.33
67.10
46.89
46.53
8645.74
399.25
64.19%
68.65
60.70
45.08
17.86
8.40
7.62
25.46
6.62
7.47
24.31
8.04
10.59
16.39
8.41
7.38
14.17
9.22
6.04
8.97
9.70
6.65
0.88
0.89
0.73
1.07
0.93
0.79
26.79%
15.12%
10.60%
11.16%
8.84%
6.98%
17.21%
11.63%
10.23%
25.12%
11.16%
6.98%
33.02%
23.02%
16.28%
47.44%
20.93%
12.56%
3.53%
2.60%
0.93%
2.79%
2.79%
0.47%
1.40%
3.72%
6.51%
2.79%
6.05%
3.26%
14.05%
29.30%
21.21%
7.35%
4.74%
10.23%
39.07%
15.81%
10.70%
6.05%
8.84%
31.16%
24.65%
9.30%
5.58%
14.88%
33.95%
19.53%
9.77%
5.12%
18.60%
23.72%
17.21%
4.19%
2.33%
17.67%
18.60%
28.84%
2.79%
4.65%
33.21%
13.701
4.55%
31.63%
12.912
4.52%
28.84%
13.367
5.03%
28.84%
12.145
4.95%
42.33%
13.797
4.74%
34.42%
16.282
3.44%
Income and expenditure structure
Per capita expenditure
533.22
Total income
2527.78
of which farm pro®ts
55.96%
of which wage income
29.49%
enterprise pro®ts/non-earned income
12.53%
of which remittances
2.02%
Household characteristics
Number of household members
Head's education
Mean education (members >15)
Have migrants in the household
Asset ownership and labor supply
Area of land owned (ha)
Business assets (in US$)
Household assets
Level of specialization
Notional wage rate (US$ per week)
Total weeks worked
Weeks self-employed in
agriculture
Weeks spent on wage labor
Of which nonfarm
Weeks spent in nonfarm
enterprises
Weeks searching for employment
Credit and savings
Have savings account
Had used credit
Through formal ®nancial
institutions
Through traders
Through informal lenders
Reasons for non-use of credit
Not needed
Documentation too dicult
Rates too high
Do not have collateral
Other reasons
Infrastructure and services
Received technical assistance for free
Distance to infrastructure
Municipio severely aected by
violence
an inducement for households to turn to
nonfarm employment. Households are thought
of being pushed to engage in nonfarm
employment because of imperfections in intertemporal and factor markets and/or entry
barriers to high return activities. Pull factors
that would attract households to nonfarm
employment include: (i) higher income generated in nonfarm activities (wage and nonwage
5
employment); (ii) potentially lower risk; and
(iii) greater social status attributed to nonfarm
activities. Push factors are commonly thought
to include (i) lack of access to productive
resources (e.g., land) to expand farm output
because of unequal distribution and malfunctioning land rental markets; (ii) the need to rely
on costly mechanisms of diversi®cation and
self-insurance to ex ante mitigate risks in an
COLOMBIA
environment where intertemporal markets for
credit and insurance do not function well; and
(iii) entry barriers such as minimum requirements of human or physical capital that prevent
the poor from entering high-return activities.
The way in which push and pull factors
interact with a region's agro-ecological
endowment has given rise to a number of
speci®c patterns that relate the amount of
nonfarm income to overall household wealth or
total income. Many African countries with a
relatively egalitarian distribution of land assets,
an underdeveloped farm labor market, and a
predominance of a traditional production
technology that relies on inputs of family labor,
display a strong positive relationship between
the share of nonfarm income and total wealth
levels (Reardon, 1997). Similar phenomena are
reported from many agricultural regions of
China where an egalitarian distribution of land
translates into great equality of opportunity in
the sense of ensuring a basic level of income
and nutrition. Households with higher levels of
human capital tend to augment this with
employment in local Township and Village
Enterprises and income from temporary
migration (Zhao, 1999; Rozelle, Taylor, &
DeBrauw, 1999; Hare, 1999).
By contrast, many case studies from Latin
American countries, and from other parts of
Asia, ®nd a U-shaped relationship whereby
low-income households are often the ones who
obtain the highest share of their income from
(low-paying) nonfarm employment (see Reardon et al., 2000; Feldman & Leones, 1998;
Garcia & Alderman, 1993 and Adams, 1994,
for example). This phenomenon, under which
low-income and high-income households both
engage in nonfarm employment but house in
quite dierent types of occupations will, in
addition to the contemporaneous income
distribution, also aect the longer-term evolution of the rural economy. The reason is that
nonfarm income generally provides an important source for agricultural investment (Ilahi,
1999; Taylor & Yunez-Naude, 1999, De Janvry,
Gordillo de Anda, & Sadoulet, 1997). In such a
situation, poor households who do not have a
suciently large agricultural resource base and
have limited access to credit markets, and who
lack skills, access to social networks, and ``migration capital'' may well be caught in poverty
traps from which there is little escape. As a
consequence, the emergence of nonfarm
employment may give rise to increased
concentration of wealth and dierentiation of
459
the rural society with associated social tensions,
con¯ict, and violence (Andre & Platteau, 1998;
Francis & Hoddinott, 1993).
To examine the relevance of these factors for
the case of Colombia, we disaggregate the
statistics presented previously by quintile of the
per capita expenditure distribution (Table 1,
columns 2±6). In addition to con®rming that
income varies considerably across household
groups, doing so points to a strong positive
association between the level of income and the
extent of specialization in either the farm or the
nonfarm sector. Table 1 illustrates the Ushaped relationship between the share of
nonfarm income and asset endowments or total
income: The poorest quintile obtains 60% of
their income from nonfarm sources, a share
that declines to 35% for the fourth quintile and
then increases again to 45% for the top quintile. 5
In addition, and contrary to what one might
expect, there are no huge dierences in the
relative importance of enterprise pro®ts and
non-earned income between the top and the
bottom quintileÐin fact both groups obtain
about 15% of their income from these sources
(Table 1). The contribution of migrant earnings
to total household income decreases linearly
over the income distribution, from about 4%
for the bottom to about 1% for the top quintile.
This suggests that, contrary to situations where
(international) migration functions as a source
of funds for investment and a means of capital
accumulation, the amount of return ¯ows in
most of rural Colombia is of minor importance.
Moving from the composition of income to
household assets points toward a strong positive relationship between the amount of assets
owned and the level of specialization, de®ned as
the share of households in the group who spend
all their time in only one activity (i.e. either
farming, running a nonfarm enterprise, or wage
work). The share of specialized households
increases from 39% in the lowest quintile to
64% in the top quintile (Table 1). In terms of
the earlier discussion, this suggests that there
are either considerable entry barriers to higher
paying jobs or that imperfect insurance markets
prevent poor households from engaging (and
specializing) in high-return activities. 6
The potential quantitative importance of
these constraints is demonstrated by a
comparison of total labor supply and wages
received over the income distribution. Households in the top quintile work almost 20% less
than households in the bottom quintile,
460
WORLD DEVELOPMENT
implying that their higher level of income is
based on higher returns to labor and other
assets. Computation of a notional ``wage rate''
by dividing total income by the number of
weeks worked illustrates these dierencesÐ
while the poor receive on average US$17 per
week worked, the rich receive four times as
much, i.e. more than US$68. Descriptive analysis cannot distinguish between returns to labor
and other assets but given the magnitude of the
dierence, it would be of considerable interest
to ®nd out whether it can be explained solely in
terms of asset endowments or whether there are
additional gains from specialization and/or
from work in the nonfarm sector. Examining
this in more detail is the topic of the next
section.
3. THE IMPACT OF NONFARM
EMPLOYMENT
In this section, we aim to assess the impact of
nonfarm activity on household welfare. Based
on the descriptive statistics presented earlier, we
test two hypotheses. First, we surmise that
specialization, rather than the choice of sector
(farm or o-farm) has a major impact on the
returns households are able to obtain for their
labor. Second, we believe that, due to pervasive
imperfections in the functioning of land, labor,
and credit markets, households' endowments
have a strong impact on whether or not they
are able to specialize. Con®rmation of this
hypothesis would imply that, in addition to
augmenting households' endowments of
human capital and other assets, policies to
improve the functioning of rural factor markets
can go a long way to help harness the bene®cial
potential of growing specialization, either in
farm or nonfarm activities.
(a) Does nonfarm employment increase returns
to labor?
To explore returns to labor as well as other
household assets and factors of production, we
regress total household expenditure (as a proxy
for permanent income) on the household's total
labor supply to the market. 7 To identify the
impact of specialization on returns to labor, we
interact labor supply with a dummy variable
that equals one if the household specializes (i.e.,
supply labor to only one type of activity) and
zero otherwise. 8 We also include ownership of
productive assets (self-reported land values, the
value of business assets and farm machinery,
and the value of livestock). Coecients on
these variables measure returns to labor and
other household assets. Furthermore, access to
formal savings and the number of relatives
living abroad are included as two characteristics that are likely to increase households'
ability to draw on resources that would allow
to smooth consumption and overcome entry
barriers to or the high risk associated with
entry into pro®table nonfarm opportunities.
Labor supply and the specialization dummy
are clearly endogenous, i.e. correlated to
unobserved household characteristics such as
entrepreneurial drive etc. which, even though
they also have a direct impact on household
income, are omitted from the regression. As a
consequence, ordinary least squares (OLS)
would yield biased estimates of the relevant
coecients and it is necessary to use instrumental variable methods to identify the relationship in question. Given the panel structure
of the data, we use household-level changes for
the variables of interest (changes in family
labor supplied, changes in the livestock herd,
changes in the age structure of the household,
changes in the value of machinery stock,
changes from specialization to multiple activities) over 1997±99 as instruments for labor
supplied and the dummy for specialization in
1999. 9 Main results of instrumental variables
estimation of annual household expenditure
equation are summarized in Table 2. We
discuss these ®ndings below.
Specialization signi®cantly increases returns
to labor: According to the regression estimates,
households that adopt multiple income-generating strategies obtain a relatively low return to
their labor. By contrast, adopting a specialized
strategy more than doubles returns to labor.
This large and statistically signi®cant dierence
suggests that there are indeed formidable
barriers preventing low-income households
from adopting ``pure'' strategies. These barriers
are likely to include lumpiness of assets,
imperfect credit markets, and limited options
for diversi®cation and self-insurance. Rural
households who, for one of these reasons, are
unable to specialize, use nonfarm employment
very much as a ``refuge of poverty'' (Berdegue
et al., 2000), similar to low return subsistence
agriculture. Note that, from a quantitative
point of view, these dierences are quite
signi®cantÐthe regression estimates indicate
that, depending on the region, shift from
pluriactivity to specialization alone, with
COLOMBIA
461
Table 2. Instrumental variable estimation of annual household expenditure equationa; b
Explanatory variables
Labor supplied by the household
Specialization dummy labor supplied
Specialization dummy education labor
supplied
Agricultural specialization dummy
labor supplied
Value of non-agricultural business assets
($US)
Value of agricultural machinery/
equipment ($US)
Value of land and livestock owned by
household ($US)
Land owned, squared ($US)
Dummy for positive savings at the
beginning of year
Dummy for relatives in other states or
abroad
Constant
Number of observations
Adj. R2
(1)
(2)
2.786
(1.526)
6.141
(1.308)
3.167
(1.516)
(3)
2.847
(1.621)
6.371
(2.456)
1.601
(0.352)
0.049
(0.012)
0.063
(0.024)
0.011
(0.002)
)4.34e)9
()6.89e)10)
530.813
(108.872)
29.178
(98.970)
1128.428
(174.092)
808
0.38
0.050
(0.012)
0.059
(0.024)
0.008
(0.001)
)4.23e)90
()7.55e)10)
456.629
(108.381)
54.309
(97.555)
1111.452
(173.831)
808
0.39
)2.461
(22.222)
0.051
(0.019)
0.060
(0.032)
0.012
(0.002)
)4.38e)9
()8.02 e)10)
529.161
(109.955)
26.266
(102.464)
1143.667
(221.996)
808
0.38
a
Robust standard errors in parentheses.
Note: Regional dummies included but not reported.
*
Signi®cant at 10% level.
**
Signi®cant at 5% level.
***
Signi®cant at 1% level.
b
everything else constant, will increase households' welfare (as measured by expenditure) by
between 10% and 36%.
Education enhances returns to specialization:
Although specialization alone can be shown to
have signi®cant bene®ts, the returns to
specializing may depend on the households'
educational attainment. To test for this possibility, we repeat the above regression but
interact specialized labor supply with the level
of education. Results (column 2 of Table 2)
indicate that higher levels of education lead to a
signi®cant increase in returns to specialization.
According to the regression estimates, an
additional year of education increases income
for specialized households by between 3.4%
and 12%. For a household with seven (rather
than the median three) years of education,
specialization could thus lead to an increase in
expenditure of between 25% and 70%,
depending on the region. This provides strong
support for the notion that bene®ts from
expansion of nonfarm employment opportunities will be highest if this is combined with
policies to increase the formation of human
capital.
Returns to specialized labor are equalized
between farm and nonfarm employment: A
second question of interest is whether gains
from specialization are sector-speci®c, i.e.,
whether returns to labor for households who
are specialized dier signi®cantly depending on
whether or not they work in the farm or the
nonfarm sector. To test for this, we run a
similar instrumental variable regression that
includes an interaction between labor supplied
and a dummy variable for specialization in
agricultural activities. The estimated coecient
for this variable is not signi®cantly dierent
from zero at conventional levels, allowing us to
reject the notion that returns to specialization
are higher in nonfarm activities than they are in
farming. In other words, while household
endowments aect the expected returns,
households who specialize decide rationally
whether to allocate their labor to farming or
nonfarm activity. The policy conclusion is that
there are few barriers to entry into the nonfarm
462
WORLD DEVELOPMENT
sector other than those that aect specialization
in more general terms.
Returns to assets vary by type: The regression
also provides an estimate of the returns to the
various types of assets held by households in
the sample. We ®nd that returns to non-land
assets are quite high, ranging between 6.3% in
the case of farm machinery and 5% for
nonagricultural enterprise assets. Compared to
these assets, land and livestock (which are
highly correlated) seem to be highly overvalued; the coecient on the value of land
assets (self-reported, and including improvements) plus livestock indicates that US$1
invested in these two yields a return of only
1.15%. 10 The negative coecient on the square
of this variable indicates that, in addition, these
returns decrease with farm size.
There are three possible explanations for
such a low return to land and livestock. First,
there is likely to be some measurement error.
The stream of bene®ts normally derived from
land includes housing. No value for housing is
imputed, however, on the income/expenditure
side of the survey, implying that the regression
coecient will suer from downward bias.
Second, land may be held for speculative
purposes, implying that landowners would be
willing to accept a relatively low concurrent
yield on their investment in return for expected
appreciation of the land in the future. Finally,
violence, external shocks, and the threat of
losing property rights, may prevent landowners
from making economically optimal use of their
land. Indeed, there is evidence from the survey
that land is left uncultivated. In addition, it is
quite likely that the threat of losing property
rights or provoking invasion if land is rented
out prevents owners from supplying land to the
rental market. This is the case even though
renting out land could be bene®cial to landowners and the rural landless because renting
could yield higher returns than what is obtained
through self-cultivation and at the same time
allow poor households with a precarious
resource base to increase the returns to their
labor. Measures that would help activate land
rental markets may thus bene®t all parts of the
rural population.
Access to ®nancial infrastructure carries large
bene®ts: Access to low-cost means of saving
increases households' ability to self-insure and
diversify risks. Given the high costs of rural
®nancial intermediation, self-®nancing of
investments is generally also less costly than use
of formal credit. 11 Thus, in an environment
characterized by imperfections in the markets
for credit and insurance, one would expect
access to savings to perform an important
function. Indeed, the regressions show that,
other things equal, households who had savings
at the beginning of the year had income levels
signi®cantly higher than that of those who did
not have access to savings. It would be of
interest to ®nd out whether, as has been
observed in the literature, possession of savings
is related to previous exposure to the nonfarm
sector. Unfortunately no information on this is
available from our survey.
Migrant remittances do not perform an
important function: Contrary to what has been
observed in other countries where migrant
remittances provide an important safety net
and a source of funds for agricultural investment that allows migrants to increase their
agricultural productivity (Mochebelele &
Winter-Nelson, 2000), having relatives in other
departments or abroad does not have a
perceptible impact on household welfare in
ColombiaÐthe coecient is positive but not
signi®cantly dierent from zero. One possible
explanation is that migrants cut the social ties
with their communities of origin. Alternatively,
and similar to households who pursue diversi®ed strategies of pluriactivity locally, migrants'
inability to enter the market for higher-paying
jobs in the location of destination may force
them to pursue low-return activities even in
other localities which makes it dicult to
generate large surpluses that can be re-invested
in the local economy.
(b) Determinants of specialization
Our analysis thus far indicates that, even
though returns to labor do not vary signi®cantly between households engaging in farmand nonfarm employment, specialization
greatly increases household welfare. Adoption
of diversi®ed strategies due to market imperfections would not only reduce household
welfare but also total production. Any move
that could help make markets function better
(and thereby increase the level of specialization)
would thus be Pareto-improving (Newbery &
Stiglitz, 1981). Identi®cation of factors that
prevent specialization at the household level
and of measures to help households overcome
obstacles to specialization would thus be of
great interest and policy relevance.
To do so, and to test empirically the extent to
which household endowments aect labor
COLOMBIA
Table 3. Probit regression for households' specializationa; b
Number of adults (16 years and
older) in the household
Number of children (15 and
younger) in household
Head's years of education (years)
)0.08125
(0.02601)
0.04496
(0.02390)
0.03315
(0.01674)
0.03366
Value of agricultural machinery
(0.01757)
(1000 US $)
0.00401
Value of land and livestock (1000
(0.00159)
US $)
0.01090
Value of land and livestock squared
(0.00000)
(1,000,000 US $)
Value of non-ag business assets
0.00000
(1000 US $)
(0.00001)
Household members living abroad
0.06468
for more than 2 years
(0.10123)
Constant
)0.07875
(0.15825)
No. of observations
1075
0.0781
Pseudo-R2
Log likelihood
)686.451
a
Robust standard errors in parentheses.
Note: Regional dummies included but note reported.
*
Signi®cant at 5%.
**
Signi®cant at 5%.
***
Signi®cant at 10%.
b
supply decisions, we run a Probit equation for
specialization at the household level. As has
been stated repeatedly in the literature, if all
markets were perfect, household characteristics
and endowments should not have any impact
on labor supply decisions (e.g., Udry, 1997).
The ®nding that households' composition, asset
endowments, and educational status have a
signi®cant impact on their patterns of factor
use, including whether they will specialize, thus
con®rms that rural factor markets in Colombia
suer from considerable imperfections.
Combining measures to promote nonfarm
employment with those aimed at improving the
functioning of markets could be doubly bene®cial. Key results are displayed in Table 3 and
discussed below.
Asset ownership promotes specialization: The
coecient for ownership of land and livestock
(which, as noted earlier, are highly correlated)
is highly signi®cant and positive, suggesting
thatÐeither by increasing the scope for selfinsurance or by allowing to overcome entry
barriersÐhigher levels of land and livestock
ownership signi®cantly reduce households'
propensity to engage in and draw income from
a multitude of employment sources. 12 Finding
463
mechanisms, such as provision of ®nancial
infrastructure that would allow small-scale
savings could, by reducing the need for socially
inecient diversi®cation, be associated with an
increase in overall welfare in rural areas.
Large households are more likely to adopt
diversi®ed strategies: The fact that, for any
given level of asset endowment, households
with a larger number of adults are also more
likely to adopt multiple income-generation
strategies (Table 3) suggests that, in addition to
markets for credit, markets for land and labor
also suer from considerable imperfections.
Instead of specializing in one main activity and
adjusting to variations in household size (which
may be life-cycle related) through the land
rental market, large households appear to be
forced to adopt multiple income-generating
strategies, even if this is not in line with the
specialized skills they possess. On the other
hand, contrary to a priori expectations, we are
unable to ascertain dierences in the coecients
on the number of household members below
and above the age of 35 and therefore only
report the total number of adults in the
household.
Education is an important determinant of
specialization: More educated households are
less likely to adopt multiple strategies of
income generation. This is likely to re¯ect the
co-existence of low-paying ``menial'' jobs with
little human capital requirements side-by-side
with activities that are characterized by high
entry barriers (such as possession of a minimum level of human capital). Overcoming
these entry barriers is a sunk investment.
Unless they are forced to do so, households
who successfully managed to overcome these
barriers will not diversify into areas that have
lower returns, thus explaining the positive and
highly signi®cant coecient on this variable.
4. CONCLUSION AND POLICY
IMPLICATIONS
In addition to con®rming the importance of
nonfarm activities as a source of income and
employment, our data also support the
hypothesis that, in view of the relatively
unequal distribution of assets and land, ofarm employment in Colombia falls into two
quite distinct categories. A signi®cant share of
poor households engages in a combination of
wage labor in jobs with low entry requirements
plus self-employment in ``marginal'' on-farm or
464
WORLD DEVELOPMENT
informal sector activities, neither of which
provide the returns required to sustain signi®cant investment and oer prospect for longerterm accumulation. At the same time, nonfarm
employment oers increased opportunities for
enhanced specialization which increase the
welfare and the capacity to invest of households
who are able to overcome the associated entry
barriers, thereby providing the basis for longerterm development of the rural sector.
Our analysis suggests that, in addition to
creating the pre-conditions for vigorous
growth of the nonfarm sector, government can
help to maximize the private and social bene®ts from such growth through three steps,
namely by (a) improving the functioning of
land, insurance, and credit markets; (b)
investing in human capital; and (c) taking
steps to help improve the asset endowments of
the poor. By enabling households to specialize
and make full use of the opportunities inherent in the development of a nonfarm sector,
doing so will increase individual as well as
social welfare. The Asian example where, in an
environment with relatively egalitarian distribution of income, well-functioning factor
markets, and a strong emphasis on educational expansion, rural nonfarm employment
has led to a spurt of broad-based development
and rapid income increases for all rural
inhabitants (Hayami, 1995) suggests that such
a strategy could provide large bene®ts not
only to rural dwellers but to the economy as a
whole.
NOTES
1. In India, an initial increase in easy entry o-farm
jobs which are relatively low-paying gives way to the
expansion of better-paying o-farm opportunities which
are created in response to the demand for nonfarm
products and services (Lanjouw & Stern, 1993). In the
Philippines it is found that the expansion of employment
opportunities outside of the farming sector precipitates
an increase in the returns to human capital through
migration which gives rise to a successive shift away
from farming toward nonfarm employment (Estudillo &
Otsuka, 1998). Household censuses in Latin America
also show a secular increase in the importance of rural
nonfarm employment (Klein, 1992). Of course, globalization can, in certain cases, also reduce the extent of
rural nonfarm employment.
2. Lack of access to assets has been identi®ed as a
major cause of poverty in Colombia (Leibovich &
Nunez, 1999).
3. The sample was strati®ed into 11 agro-ecological
zones. In each of the zones, 10 municipalities and within
these municipalities 10 households were selected
randomly. All households were surveyed two times,
once in 1997 and then again in 1999. Due to attrition
and the inability to visit a number of localities because of
violence, the sample in the second round was reduced
from 1,075 to 808.
4. A very limited number of households (2%) use paid
technical assistance.
5. The presence of a U-shaped relationship is
con®rmed by regression analysis (not reported).
6. Households appear to be willing to accept a lower
return on labor as a ``risk premium'' in return for the
risk diversi®cation advantages associated with the
adoption of reliance on a multiplicity of income sources.
7. Per capita expenditure is usually thought of being a
better proxy to permanent per capita income since it
captures household's ability to smooth consumption.
8. In the second regression reported in Table 2, we
include a further interaction between specialization
and the amount of labor supplied to the nonfarm
sector to test whether there is a statistically signi®cant
dierence between returns to labor obtained by households who specialize in on-farm and on-farm activities,
respectively.
9. Details for this approach of using ®rst dierenced
variables as instruments for endogenous level variables
are given in Hausmann and Taylor Edward (1981).
10. Since the high correlation of land and livestock
(with a correlation coecient of more than 0.6) resulted
in instability of the coecients, we added them
together.
11. This is con®rmed by the fact that, as discussed
earlier, high cost of credit constitutes a powerful
deterrent to the use of credit. Valentine (1993) and
Reardon and Taylor (1996) also ®nd that nonfarm
income allows households to draw on non-covariate
streams of income, thus increasing their ability to deal
with and recover from shocks that may otherwise have
disastrous consequences.
COLOMBIA
12. The negative sign of the squared term points to
the presence of decreasing marginal impact of such
asset-ownership on the propensity to specialize. By
465
comparison, the coecient on machinery is signi®cant
at the 10% level and enterprise assets is not signi®cant.
REFERENCES
Adams, R. H. J. (1994). Nonfarm income and inequality
in rural Pakistan: a decomposition analysis. Journal
of Development Studies, 31(1), 110±133.
Andre, C., & Platteau, J.-P. (1998). Land relations under
unbearable stress: Rwanda caught in the Malthusian
trap. Journal of Economic Behavior and Organization,
34(1), 1±47.
Berdegue, J. A., Reardon, T., & Escobar, G. (2000).
Empleo e ingreso rurales no agricolas en America
Latina y el Caribe. Paper presented at the conference
Development of the Rural Economy and Poverty
Reduction in Latin America and the Caribbean, New
Orleans, Louisiana, March 24.
De Janvry, A., Gordillo de Anda, G., & Sadoulet, E.
(1997). Mexico's second Agrarian reform: household
and community responses, 1990±94. La Jolla: Center
for US±Mexican Studies, University of California at
San Diego.
Estudillo, J. W., & Otsuka, K. (1998). Green revolution,
human capital and o-farm employment: changing
sources of income among farm households in central
Luzon, 1966±94. Economic Development and Cultural
Change, 47(3), 497±523.
Feldman, S., & Leones, J. P. (1998). Nonfarm activity
and rural household income: evidence from Philippine microdata. Economic Development and Cultural
Change, 46(4), 789±806.
Francis, E., & Hoddinott, J. (1993). Migration and dierentiation in western Kenya: a tale of two sub-locations. Journal of Development Studies, 30(1), 115±145.
Garcia, M., & Alderman, H. (1993). Food security and
health security: explaining the levels of nutritional
status in Pakistan. Economic Development and
Cultural Change, 42(3), 485±507.
Hare, D. (1999). Push versus pull factors in migration
out¯ows and returns: determinants of migration
status and spell duration among China's rural population. Journal of Development Studies, 35(3), 45±72.
Hausmann, J. A., & Taylor, W. (1981). Panel data and
unobservable individual eects. Econometrica, 49,
1377±1399.
Hayami, Y. (Ed.) (1998). Towards the rural-based
development of commerce and industry. Selected
experiences from East Asia. EDI Learning Resource
Series. Washington DC: EDI.
Ilahi, N. (1999). Return migration and cccupational
change. Review of Development Economics, 3(2), 170±
186.
Jaramillo, C. F. (1998). Liberalization crisis and change in
Colombian agriculture. Boulder: Westview Press.
Klein, E. (1992). El empleo rural no agricola en
America Latina, documento de trabajo, No. 364,
Santiago de Chile, Programa Regional del Empleo
para America Latina y el Caribe (PREALC),
August.
Lanjouw, P., & Stern, N. (1993). Markets, opportunities
and changes in inequality in Palanpur 1957±1984. In
A. Braverman, K. Ho, & J. Stiglitz, The economics
of rural organization: Theory, practice and policy.
New York: Oxford University Press.
Leibovich, J., & Nunez, J. (1999). Activos y recursos de
la poblacion pobre en Colombia. El Trimestre
Economico, 66(3), 501±551.
Mochebelele, M. T., & Winter-Nelson, A. (2000).
Migrant labor and farm technical eciency in
Lesotho. World Development, 28(1), 143±153.
Newbery, D., & Stiglitz, J. (1981). The theory of
commodity price stabilization: A study in the economics of risk. Oxford: Clarendon Press.
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±748.
Reardon, T., & Taylor, J. E. (1996). Agroclimatic
shock income inequality and poverty: evidence
from Burkina Faso. World Development, 24(5),
901±914.
Reardon, T. et al. (1998). Rural non-farm income in
developing countries. In FAO, The state of food and
agriculture. Rome: FAO.
Reardon, T., Taylor, J. E., Stamoulis, K., Lanjouw, P.,
& Balisacan, A. (2000). Eects of nonfarm employment on rural income inequality in developing
countries: an investment perspective. Journal of
Agricultural Economics, 51(2), 266±288.
Rozelle, S., Taylor, J. E., & DeBrauw, A. (1999).
Migration remittances and agricultural productivity
in China. American Economic Review, 89(2), 287±
291.
Taylor, J. E., Yunez-Naude, A. (1999). Education,
migration and productivity: An analytic approach
and evidence from rural Mexico. Paris: OECD
Development Centre Studies.
Udry, C. (1997). Recent advances in empirical microeconomic research in poor countries. Mimeo, North
western University.
Valentine, T. R. (1993). Drought transfer entitlements
and income distribution: the Botswana experience.
World Development, 21(1), 109±126.
World Bank (1996). Colombia poverty assessment.
Washington, DC: World Bank.
Zhao, Y. (1999). Labor migration and earnings dierences: the case of rural China. Economic Development and Cultural Change, 47(4), 767±782.
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)00106-6
Rural Nonfarm Employment and Income
Diversi®cation in Colombia
KLAUS DEININGER and PEDRO OLINTO *
The World Bank, Washington, DC, USA
Summary. Ð Using data for rural households from Colombia we ®nd that o-farm employment
contributes a signi®cant share (45% on average) to household income but that the importance of
o-farm income and returns to household labor vary over the income distribution. Analysis reveals
signi®cant gains from specialization butÐfor households able to specializeÐno systematic
dierences in returns to labor between farm and nonfarm sources. We conclude that, in Colombia,
there is no con¯ict between development of the farm and the nonfarm sector but that, to maximize
gains from nonfarm development and reduce the scope for undesirable distributional consequences,
policies enabling households to specialize might be called for. Ó 2001 Elsevier Science Ltd. All
rights reserved.
Key words Ð Colombia, o-farm employment, income distribution
1. INTRODUCTION
There is little doubt that the importance of
rural nonfarm employment, which in many
countries already constitutes an important
sector of the rural economy, will greatly
increase as agriculture becomes more and more
integrated into global markets and as the links
between urban and rural areas intensify. What
is less clear, however, is how these forces for
diversi®cation can best be harnessed for
nonfarm employment to act as a catalyst for a
broader and inclusive pattern of development.
From a policy perspective, it is of particular
interest to ®nd out whether the rural poor are
able to make optimum use of the opportunities
provided by nonfarm employment or whether
speci®c policy measures to assist them might be
needed.
In this paper, we use data from Colombia to
address this question. Descriptive statistics
illustrate both the importance of nonfarm
employment and broad patterns of participation in nonfarm opportunities across dierent
groups of the population. Nonfarm income
(wages from agricultural and nonagricultural
employment, pro®ts from nonagricultural
enterprises, nonearned income, and remittances) contribute on average 45% to household income. There is also a nonlinear
(U-shaped) relationship between the importance
of o-farm work, asset endowments, and total
455
household income. By comparison, specialization (in either farm or nonfarm activities)
increases linearly with income and assets.
The strong positive association between total
income and specialization suggests that even
though nonfarm employment contributes to
diversi®cation of income generating opportunities at the regional level, individual households may still be better o by relying only on
one main income source. More importantly, to
the extent that market failures and lack of
endowments prevent them from specializing,
government policies that improve the functioning of factor markets and that help households increase their endowments could play an
important role to maximize the gains associated
with the emergence of nonfarm employment
opportunities.
To explore this in more detail, we examine
the impact of specialization and household
labor supply as well as the determinants of
specialization. We ®nd that, indeed, specialization allows households to increase their level
* We
are deeply indebted to Elsa Albarracin, Juliana
Bottia, Diana Grudzynski, Absalon Machado, Manuel
Rojas, Hernando Urbina and Guilermo Otanez without
whose enthusiasm and support the data underlying this
analysis would never have been collected. We are also
grateful for useful comments from three anonymous
reviewers.
456
WORLD DEVELOPMENT
of welfare (as measured by expenditures) by
between 10% and 36%, everything else equal.
Why, then, do not all households choose to
specialize in one activity? We ®nd that imperfections in markets for credit and land, lack of
education, and inequalities in asset ownership
constitute important barriers to increased
specialization.
From a policy perspective this implies that in
situations such as Colombia, where education
and assets are distributed in an unequal
manner, the impact of increased nonfarm
employment opportunities will not be independent from households' and communities
initial endowments. Households with little
human or physical capital may be forced to rely
on nonfarm employment as a low return ``refuge'', comparable to semi-subsistence, with
little prospect for economic advancement. Only
if they own suciently high levels of assets or
are able to access to credit and land rental
markets will households be able to make full
use of the opportunities for specialization and
increased returns to labor provided by a more
diversi®ed rural nonfarm economy. The
welfare-enhancing impact of nonfarm employment opportunities will thus be maximized if
policies aimed at promoting the rural nonfarm
sector are complemented with measures to
improve the functioning of factor markets and
to increase households' opportunities to accumulate human and physical capital.
The paper is structured as follows. Section 2
provides descriptive evidence on the importance and incidence of nonfarm employment
across the income distribution and the country's dierent regions as well as a description of
the data sources underlying the study. Section 3
discusses the main econometric results, in
particular the impact of specialization on
household welfare and the determinants of
households' decision to specialize. Section 4
links the results to the broader discussion of
nonfarm employment in the literature and, on
this basis, derives a number of conclusions for
policy as well as research.
2. INCIDENCE AND CHARACTER OF
NONFARM EMPLOYMENT
In this section we describe the data underlying the analysis and discuss descriptive statistics
about the incidence and nature of nonfarm
employment in rural Colombia. We ®nd an Ushaped relationship between the share of o-
farm income and household assets or total
income level. This is consistent with the evidence from a number of other countries where
presence of entry barriers to high-return jobs in
the o-farm sector, together with a relatively
unequal distribution of farm assets and
malfunctioning land rental markets, force poor
people with scant asset endowments into lowpaying o-farm jobs and prevent them from
taking maximum advantage from the opportunities oered by nonfarm employment.
(a) Background and data sources
Thanks to a large amount of studies on the
nonfarm sector across all continents, the
importance of rural nonfarm employment is
now widely recognized. Country case studies
illustrate that the share of nonfarm income in
total household income ranges between about
30% and 40%Ðwith the highest shares (45%)
reported from Africa and the lowest ones (29%)
from South Asia (Reardon et al., 1998).
Although household-level evidence on the
evolution of nonfarm employment is limited, 1
the contribution of nonfarm income sources
and o-farm employment to the rural economy
has grown substantially during the last two
decades and is likely to continue doing so in
view of globalization, progressive insertion of
rural areas into the larger economy, and
increased access to public services (Berdegue,
Reardon, & Escobar, 2000).
While our understanding of the magnitude of
the rural nonfarm sector has greatly improved,
the contribution of this sector to household
welfare, and the distribution of the bene®ts
from o-farm employment across the population, are still imperfectly understood. Evidence
on whether nonfarm employment contributes
to a more equal distribution of income is
decidedly mixed (Reardon, Taylor, Stamoulis,
Lanjouw, & Balisacan, 2000). To advance on
this issue, it is necessary to explore not only
what determines participation in the rural
nonfarm economy, but also how such participation aects households' welfare. This is
crucial not only for academic reasons but, more
importantly, to allow governments to take
measures that will enable the poor to take
advantage of the opportunities inherent in the
growing importance of the nonfarm sector,
thus turning it into a catalytic force for rural
growth and sustainable reduction of poverty.
Colombia is of interest for this issue in a
number of respects. In marked dierence to
COLOMBIA
the ups and downs of other Latin American
countries, the country had, until the recent
upsurge in violence and macroeconomic
problems, been characterized by stable
economic growth. At the same time, it shares
with other Latin American economies a highly
unequal distribution of assets. 2 Maldistribution of assets is particularly acute in rural
areasÐdespite more than three decades of
land reform, land access is highly unequal with
the Gini coecient of land ownership in 1990
estimated to be 0.81 (World Bank, 1996).
Other assets are distributed in a slightly less
unequal fashion, with a Gini of 0.77. This is
relevant for the rural population as agriculture
is still the single most important sector in the
economy, generating a ®fth of total value
added, over a third of foreign exchange, and
more than 30% of total employment in the
economy.
Starting in the early 1990s, the country
implemented a far-reaching program of
adjustment (apertura) which, by turning away
from a paradigm of import-substituting industrialization, opened up the agricultural sector
to the forces of international competitiveness.
This led to large gains for producers who were
well connected to markets and able to adjust
quickly to the changed system of incentives. At
the same time, it tended to reinforce old
dichotomies in the distribution of assets as
small producers who were not able to shift out
of traditional commodities suered considerable losses. Migration, together with rapid
growth of the rural nonfarm sector enabled
rural dwellers to improve or at least stabilize
their income in the face of these external shocks
(Jaramillo, 1998).
To identify whether nonfarm employment
can, in addition to constituting a safety net,
also act as a catalyst for an inclusive pattern of
economic development in the rural sector, we
use data from a survey of about 1,000 rural
households that was undertaken by the
Departamento Nacional de Planeacion (DNP) in
collaboration with IICA and the World Bank.
The main purpose of the survey was to examine
factors aecting technical eciency of dierent
farm sizes, the functioning of rural factor
markets, and sources of income and employment of the rural population. It contains
comprehensive information on labor use,
general household characteristics, asset
endowments, migration, and access to government services which can provide a better
understanding of the rural nonfarm economy. 3
457
(b) Descriptive evidence
The survey data reveal that Colombia's rural
inhabitants draw incomes from a wide variety
of sources. As illustrated in column 1 of
Table 1, farm pro®ts made up 56% of total
income, complemented by wage income from
farm and nonfarm sources (30%), nonfarm
enterprise pro®ts and nonearned income
(12.5%), and migration remittances (2.5%). The
average household in the survey had slightly
less than ®ve members with the head having
completed 2.9 years of schooling, compared to
a mean of 3.9 for all household members over
the age of 15 years. Households' mean asset
endowment amounted to about 25 ha of land
and business assets (including machinery, livestock, vehicles, and nonfarm enterprise assets)
worth about US$4,500. In line with what is
known from other sources of information, our
data point toward an unequal distribution of
assets. About 13% of the sample have relatives
who migrated out and may have provided
remittance income.
Individuals' wage rates and thus opportunities
in the farm and nonfarm sector vary depending
on their level of education, physical location,
and type of work performed. To account for
this, we complement information on income by
source with data on the amount of hours worked
in dierent economic activities. This indicates
that 62% of the 70 weeks per year which the
typical households spent working was used in
agricultural activities, 25% in wage labor, and
10% in independent nonfarm enterprises.
The information on credit and savings
provided by the survey points toward limited
access to ®nancial infrastructure and a reluctance to use credit, rather than household
savings, to ®nance investment. One-quarter of
households had pre-existing savings, but only
15% used credit. Of these, 10% was from the
formal and about 5% from the informal
sectorÐwhich comprised traders (3.5%) and
informal lenders (2.6%). Half of the households
did not use credit because of high rates or
complicated documentation, while another 14%
reported not to need credit and 7% did not have
collateral. Access to free technical assistance
was, with 33%, fairly widespread. 4
(c) Nonfarm employment, asset ownership, and
specialization
The literature has long emphasized the relative importance of ``pull'' and ``push'' factors as
458
WORLD DEVELOPMENT
Table 1. Descriptive statistics by quintile of the per capita expenditure distribution
Quintiles of per capita expenditure
Total
1
2
3
4
163.01
1254.01
39.90%
40.59%
15.42%
4.09%
284.44
1979.03
49.28%
36.16%
11.74%
2.82%
405.21
2264.34
59.09%
27.06%
11.54%
2.30%
587.26
2974.33
65.33%
24.83%
8.11%
1.73%
1226.19
4167.17
55.73%
27.51%
15.73%
1.03%
4.71
2.89
3.96
12.47%
6.20
2.12
3.27
16.74%
5.36
2.80
3.78
15.35%
4.52
2.96
4.04
11.16%
4.20
3.09
4.13
11.16%
3.28
3.47
4.56
7.91%
24.55
4447.97
234.28
51.53%
36.51
69.23
43.75
6.84
748.73
117.74
38.60%
17.21
72.87
39.95
17.49
3657.34
178.65
48.37%
26.32
75.19
40.29
20.17
3240.90
199.78
49.77%
32.21
70.31
46.54
31.70
5947.15
275.99
56.74%
44.33
67.10
46.89
46.53
8645.74
399.25
64.19%
68.65
60.70
45.08
17.86
8.40
7.62
25.46
6.62
7.47
24.31
8.04
10.59
16.39
8.41
7.38
14.17
9.22
6.04
8.97
9.70
6.65
0.88
0.89
0.73
1.07
0.93
0.79
26.79%
15.12%
10.60%
11.16%
8.84%
6.98%
17.21%
11.63%
10.23%
25.12%
11.16%
6.98%
33.02%
23.02%
16.28%
47.44%
20.93%
12.56%
3.53%
2.60%
0.93%
2.79%
2.79%
0.47%
1.40%
3.72%
6.51%
2.79%
6.05%
3.26%
14.05%
29.30%
21.21%
7.35%
4.74%
10.23%
39.07%
15.81%
10.70%
6.05%
8.84%
31.16%
24.65%
9.30%
5.58%
14.88%
33.95%
19.53%
9.77%
5.12%
18.60%
23.72%
17.21%
4.19%
2.33%
17.67%
18.60%
28.84%
2.79%
4.65%
33.21%
13.701
4.55%
31.63%
12.912
4.52%
28.84%
13.367
5.03%
28.84%
12.145
4.95%
42.33%
13.797
4.74%
34.42%
16.282
3.44%
Income and expenditure structure
Per capita expenditure
533.22
Total income
2527.78
of which farm pro®ts
55.96%
of which wage income
29.49%
enterprise pro®ts/non-earned income
12.53%
of which remittances
2.02%
Household characteristics
Number of household members
Head's education
Mean education (members >15)
Have migrants in the household
Asset ownership and labor supply
Area of land owned (ha)
Business assets (in US$)
Household assets
Level of specialization
Notional wage rate (US$ per week)
Total weeks worked
Weeks self-employed in
agriculture
Weeks spent on wage labor
Of which nonfarm
Weeks spent in nonfarm
enterprises
Weeks searching for employment
Credit and savings
Have savings account
Had used credit
Through formal ®nancial
institutions
Through traders
Through informal lenders
Reasons for non-use of credit
Not needed
Documentation too dicult
Rates too high
Do not have collateral
Other reasons
Infrastructure and services
Received technical assistance for free
Distance to infrastructure
Municipio severely aected by
violence
an inducement for households to turn to
nonfarm employment. Households are thought
of being pushed to engage in nonfarm
employment because of imperfections in intertemporal and factor markets and/or entry
barriers to high return activities. Pull factors
that would attract households to nonfarm
employment include: (i) higher income generated in nonfarm activities (wage and nonwage
5
employment); (ii) potentially lower risk; and
(iii) greater social status attributed to nonfarm
activities. Push factors are commonly thought
to include (i) lack of access to productive
resources (e.g., land) to expand farm output
because of unequal distribution and malfunctioning land rental markets; (ii) the need to rely
on costly mechanisms of diversi®cation and
self-insurance to ex ante mitigate risks in an
COLOMBIA
environment where intertemporal markets for
credit and insurance do not function well; and
(iii) entry barriers such as minimum requirements of human or physical capital that prevent
the poor from entering high-return activities.
The way in which push and pull factors
interact with a region's agro-ecological
endowment has given rise to a number of
speci®c patterns that relate the amount of
nonfarm income to overall household wealth or
total income. Many African countries with a
relatively egalitarian distribution of land assets,
an underdeveloped farm labor market, and a
predominance of a traditional production
technology that relies on inputs of family labor,
display a strong positive relationship between
the share of nonfarm income and total wealth
levels (Reardon, 1997). Similar phenomena are
reported from many agricultural regions of
China where an egalitarian distribution of land
translates into great equality of opportunity in
the sense of ensuring a basic level of income
and nutrition. Households with higher levels of
human capital tend to augment this with
employment in local Township and Village
Enterprises and income from temporary
migration (Zhao, 1999; Rozelle, Taylor, &
DeBrauw, 1999; Hare, 1999).
By contrast, many case studies from Latin
American countries, and from other parts of
Asia, ®nd a U-shaped relationship whereby
low-income households are often the ones who
obtain the highest share of their income from
(low-paying) nonfarm employment (see Reardon et al., 2000; Feldman & Leones, 1998;
Garcia & Alderman, 1993 and Adams, 1994,
for example). This phenomenon, under which
low-income and high-income households both
engage in nonfarm employment but house in
quite dierent types of occupations will, in
addition to the contemporaneous income
distribution, also aect the longer-term evolution of the rural economy. The reason is that
nonfarm income generally provides an important source for agricultural investment (Ilahi,
1999; Taylor & Yunez-Naude, 1999, De Janvry,
Gordillo de Anda, & Sadoulet, 1997). In such a
situation, poor households who do not have a
suciently large agricultural resource base and
have limited access to credit markets, and who
lack skills, access to social networks, and ``migration capital'' may well be caught in poverty
traps from which there is little escape. As a
consequence, the emergence of nonfarm
employment may give rise to increased
concentration of wealth and dierentiation of
459
the rural society with associated social tensions,
con¯ict, and violence (Andre & Platteau, 1998;
Francis & Hoddinott, 1993).
To examine the relevance of these factors for
the case of Colombia, we disaggregate the
statistics presented previously by quintile of the
per capita expenditure distribution (Table 1,
columns 2±6). In addition to con®rming that
income varies considerably across household
groups, doing so points to a strong positive
association between the level of income and the
extent of specialization in either the farm or the
nonfarm sector. Table 1 illustrates the Ushaped relationship between the share of
nonfarm income and asset endowments or total
income: The poorest quintile obtains 60% of
their income from nonfarm sources, a share
that declines to 35% for the fourth quintile and
then increases again to 45% for the top quintile. 5
In addition, and contrary to what one might
expect, there are no huge dierences in the
relative importance of enterprise pro®ts and
non-earned income between the top and the
bottom quintileÐin fact both groups obtain
about 15% of their income from these sources
(Table 1). The contribution of migrant earnings
to total household income decreases linearly
over the income distribution, from about 4%
for the bottom to about 1% for the top quintile.
This suggests that, contrary to situations where
(international) migration functions as a source
of funds for investment and a means of capital
accumulation, the amount of return ¯ows in
most of rural Colombia is of minor importance.
Moving from the composition of income to
household assets points toward a strong positive relationship between the amount of assets
owned and the level of specialization, de®ned as
the share of households in the group who spend
all their time in only one activity (i.e. either
farming, running a nonfarm enterprise, or wage
work). The share of specialized households
increases from 39% in the lowest quintile to
64% in the top quintile (Table 1). In terms of
the earlier discussion, this suggests that there
are either considerable entry barriers to higher
paying jobs or that imperfect insurance markets
prevent poor households from engaging (and
specializing) in high-return activities. 6
The potential quantitative importance of
these constraints is demonstrated by a
comparison of total labor supply and wages
received over the income distribution. Households in the top quintile work almost 20% less
than households in the bottom quintile,
460
WORLD DEVELOPMENT
implying that their higher level of income is
based on higher returns to labor and other
assets. Computation of a notional ``wage rate''
by dividing total income by the number of
weeks worked illustrates these dierencesÐ
while the poor receive on average US$17 per
week worked, the rich receive four times as
much, i.e. more than US$68. Descriptive analysis cannot distinguish between returns to labor
and other assets but given the magnitude of the
dierence, it would be of considerable interest
to ®nd out whether it can be explained solely in
terms of asset endowments or whether there are
additional gains from specialization and/or
from work in the nonfarm sector. Examining
this in more detail is the topic of the next
section.
3. THE IMPACT OF NONFARM
EMPLOYMENT
In this section, we aim to assess the impact of
nonfarm activity on household welfare. Based
on the descriptive statistics presented earlier, we
test two hypotheses. First, we surmise that
specialization, rather than the choice of sector
(farm or o-farm) has a major impact on the
returns households are able to obtain for their
labor. Second, we believe that, due to pervasive
imperfections in the functioning of land, labor,
and credit markets, households' endowments
have a strong impact on whether or not they
are able to specialize. Con®rmation of this
hypothesis would imply that, in addition to
augmenting households' endowments of
human capital and other assets, policies to
improve the functioning of rural factor markets
can go a long way to help harness the bene®cial
potential of growing specialization, either in
farm or nonfarm activities.
(a) Does nonfarm employment increase returns
to labor?
To explore returns to labor as well as other
household assets and factors of production, we
regress total household expenditure (as a proxy
for permanent income) on the household's total
labor supply to the market. 7 To identify the
impact of specialization on returns to labor, we
interact labor supply with a dummy variable
that equals one if the household specializes (i.e.,
supply labor to only one type of activity) and
zero otherwise. 8 We also include ownership of
productive assets (self-reported land values, the
value of business assets and farm machinery,
and the value of livestock). Coecients on
these variables measure returns to labor and
other household assets. Furthermore, access to
formal savings and the number of relatives
living abroad are included as two characteristics that are likely to increase households'
ability to draw on resources that would allow
to smooth consumption and overcome entry
barriers to or the high risk associated with
entry into pro®table nonfarm opportunities.
Labor supply and the specialization dummy
are clearly endogenous, i.e. correlated to
unobserved household characteristics such as
entrepreneurial drive etc. which, even though
they also have a direct impact on household
income, are omitted from the regression. As a
consequence, ordinary least squares (OLS)
would yield biased estimates of the relevant
coecients and it is necessary to use instrumental variable methods to identify the relationship in question. Given the panel structure
of the data, we use household-level changes for
the variables of interest (changes in family
labor supplied, changes in the livestock herd,
changes in the age structure of the household,
changes in the value of machinery stock,
changes from specialization to multiple activities) over 1997±99 as instruments for labor
supplied and the dummy for specialization in
1999. 9 Main results of instrumental variables
estimation of annual household expenditure
equation are summarized in Table 2. We
discuss these ®ndings below.
Specialization signi®cantly increases returns
to labor: According to the regression estimates,
households that adopt multiple income-generating strategies obtain a relatively low return to
their labor. By contrast, adopting a specialized
strategy more than doubles returns to labor.
This large and statistically signi®cant dierence
suggests that there are indeed formidable
barriers preventing low-income households
from adopting ``pure'' strategies. These barriers
are likely to include lumpiness of assets,
imperfect credit markets, and limited options
for diversi®cation and self-insurance. Rural
households who, for one of these reasons, are
unable to specialize, use nonfarm employment
very much as a ``refuge of poverty'' (Berdegue
et al., 2000), similar to low return subsistence
agriculture. Note that, from a quantitative
point of view, these dierences are quite
signi®cantÐthe regression estimates indicate
that, depending on the region, shift from
pluriactivity to specialization alone, with
COLOMBIA
461
Table 2. Instrumental variable estimation of annual household expenditure equationa; b
Explanatory variables
Labor supplied by the household
Specialization dummy labor supplied
Specialization dummy education labor
supplied
Agricultural specialization dummy
labor supplied
Value of non-agricultural business assets
($US)
Value of agricultural machinery/
equipment ($US)
Value of land and livestock owned by
household ($US)
Land owned, squared ($US)
Dummy for positive savings at the
beginning of year
Dummy for relatives in other states or
abroad
Constant
Number of observations
Adj. R2
(1)
(2)
2.786
(1.526)
6.141
(1.308)
3.167
(1.516)
(3)
2.847
(1.621)
6.371
(2.456)
1.601
(0.352)
0.049
(0.012)
0.063
(0.024)
0.011
(0.002)
)4.34e)9
()6.89e)10)
530.813
(108.872)
29.178
(98.970)
1128.428
(174.092)
808
0.38
0.050
(0.012)
0.059
(0.024)
0.008
(0.001)
)4.23e)90
()7.55e)10)
456.629
(108.381)
54.309
(97.555)
1111.452
(173.831)
808
0.39
)2.461
(22.222)
0.051
(0.019)
0.060
(0.032)
0.012
(0.002)
)4.38e)9
()8.02 e)10)
529.161
(109.955)
26.266
(102.464)
1143.667
(221.996)
808
0.38
a
Robust standard errors in parentheses.
Note: Regional dummies included but not reported.
*
Signi®cant at 10% level.
**
Signi®cant at 5% level.
***
Signi®cant at 1% level.
b
everything else constant, will increase households' welfare (as measured by expenditure) by
between 10% and 36%.
Education enhances returns to specialization:
Although specialization alone can be shown to
have signi®cant bene®ts, the returns to
specializing may depend on the households'
educational attainment. To test for this possibility, we repeat the above regression but
interact specialized labor supply with the level
of education. Results (column 2 of Table 2)
indicate that higher levels of education lead to a
signi®cant increase in returns to specialization.
According to the regression estimates, an
additional year of education increases income
for specialized households by between 3.4%
and 12%. For a household with seven (rather
than the median three) years of education,
specialization could thus lead to an increase in
expenditure of between 25% and 70%,
depending on the region. This provides strong
support for the notion that bene®ts from
expansion of nonfarm employment opportunities will be highest if this is combined with
policies to increase the formation of human
capital.
Returns to specialized labor are equalized
between farm and nonfarm employment: A
second question of interest is whether gains
from specialization are sector-speci®c, i.e.,
whether returns to labor for households who
are specialized dier signi®cantly depending on
whether or not they work in the farm or the
nonfarm sector. To test for this, we run a
similar instrumental variable regression that
includes an interaction between labor supplied
and a dummy variable for specialization in
agricultural activities. The estimated coecient
for this variable is not signi®cantly dierent
from zero at conventional levels, allowing us to
reject the notion that returns to specialization
are higher in nonfarm activities than they are in
farming. In other words, while household
endowments aect the expected returns,
households who specialize decide rationally
whether to allocate their labor to farming or
nonfarm activity. The policy conclusion is that
there are few barriers to entry into the nonfarm
462
WORLD DEVELOPMENT
sector other than those that aect specialization
in more general terms.
Returns to assets vary by type: The regression
also provides an estimate of the returns to the
various types of assets held by households in
the sample. We ®nd that returns to non-land
assets are quite high, ranging between 6.3% in
the case of farm machinery and 5% for
nonagricultural enterprise assets. Compared to
these assets, land and livestock (which are
highly correlated) seem to be highly overvalued; the coecient on the value of land
assets (self-reported, and including improvements) plus livestock indicates that US$1
invested in these two yields a return of only
1.15%. 10 The negative coecient on the square
of this variable indicates that, in addition, these
returns decrease with farm size.
There are three possible explanations for
such a low return to land and livestock. First,
there is likely to be some measurement error.
The stream of bene®ts normally derived from
land includes housing. No value for housing is
imputed, however, on the income/expenditure
side of the survey, implying that the regression
coecient will suer from downward bias.
Second, land may be held for speculative
purposes, implying that landowners would be
willing to accept a relatively low concurrent
yield on their investment in return for expected
appreciation of the land in the future. Finally,
violence, external shocks, and the threat of
losing property rights, may prevent landowners
from making economically optimal use of their
land. Indeed, there is evidence from the survey
that land is left uncultivated. In addition, it is
quite likely that the threat of losing property
rights or provoking invasion if land is rented
out prevents owners from supplying land to the
rental market. This is the case even though
renting out land could be bene®cial to landowners and the rural landless because renting
could yield higher returns than what is obtained
through self-cultivation and at the same time
allow poor households with a precarious
resource base to increase the returns to their
labor. Measures that would help activate land
rental markets may thus bene®t all parts of the
rural population.
Access to ®nancial infrastructure carries large
bene®ts: Access to low-cost means of saving
increases households' ability to self-insure and
diversify risks. Given the high costs of rural
®nancial intermediation, self-®nancing of
investments is generally also less costly than use
of formal credit. 11 Thus, in an environment
characterized by imperfections in the markets
for credit and insurance, one would expect
access to savings to perform an important
function. Indeed, the regressions show that,
other things equal, households who had savings
at the beginning of the year had income levels
signi®cantly higher than that of those who did
not have access to savings. It would be of
interest to ®nd out whether, as has been
observed in the literature, possession of savings
is related to previous exposure to the nonfarm
sector. Unfortunately no information on this is
available from our survey.
Migrant remittances do not perform an
important function: Contrary to what has been
observed in other countries where migrant
remittances provide an important safety net
and a source of funds for agricultural investment that allows migrants to increase their
agricultural productivity (Mochebelele &
Winter-Nelson, 2000), having relatives in other
departments or abroad does not have a
perceptible impact on household welfare in
ColombiaÐthe coecient is positive but not
signi®cantly dierent from zero. One possible
explanation is that migrants cut the social ties
with their communities of origin. Alternatively,
and similar to households who pursue diversi®ed strategies of pluriactivity locally, migrants'
inability to enter the market for higher-paying
jobs in the location of destination may force
them to pursue low-return activities even in
other localities which makes it dicult to
generate large surpluses that can be re-invested
in the local economy.
(b) Determinants of specialization
Our analysis thus far indicates that, even
though returns to labor do not vary signi®cantly between households engaging in farmand nonfarm employment, specialization
greatly increases household welfare. Adoption
of diversi®ed strategies due to market imperfections would not only reduce household
welfare but also total production. Any move
that could help make markets function better
(and thereby increase the level of specialization)
would thus be Pareto-improving (Newbery &
Stiglitz, 1981). Identi®cation of factors that
prevent specialization at the household level
and of measures to help households overcome
obstacles to specialization would thus be of
great interest and policy relevance.
To do so, and to test empirically the extent to
which household endowments aect labor
COLOMBIA
Table 3. Probit regression for households' specializationa; b
Number of adults (16 years and
older) in the household
Number of children (15 and
younger) in household
Head's years of education (years)
)0.08125
(0.02601)
0.04496
(0.02390)
0.03315
(0.01674)
0.03366
Value of agricultural machinery
(0.01757)
(1000 US $)
0.00401
Value of land and livestock (1000
(0.00159)
US $)
0.01090
Value of land and livestock squared
(0.00000)
(1,000,000 US $)
Value of non-ag business assets
0.00000
(1000 US $)
(0.00001)
Household members living abroad
0.06468
for more than 2 years
(0.10123)
Constant
)0.07875
(0.15825)
No. of observations
1075
0.0781
Pseudo-R2
Log likelihood
)686.451
a
Robust standard errors in parentheses.
Note: Regional dummies included but note reported.
*
Signi®cant at 5%.
**
Signi®cant at 5%.
***
Signi®cant at 10%.
b
supply decisions, we run a Probit equation for
specialization at the household level. As has
been stated repeatedly in the literature, if all
markets were perfect, household characteristics
and endowments should not have any impact
on labor supply decisions (e.g., Udry, 1997).
The ®nding that households' composition, asset
endowments, and educational status have a
signi®cant impact on their patterns of factor
use, including whether they will specialize, thus
con®rms that rural factor markets in Colombia
suer from considerable imperfections.
Combining measures to promote nonfarm
employment with those aimed at improving the
functioning of markets could be doubly bene®cial. Key results are displayed in Table 3 and
discussed below.
Asset ownership promotes specialization: The
coecient for ownership of land and livestock
(which, as noted earlier, are highly correlated)
is highly signi®cant and positive, suggesting
thatÐeither by increasing the scope for selfinsurance or by allowing to overcome entry
barriersÐhigher levels of land and livestock
ownership signi®cantly reduce households'
propensity to engage in and draw income from
a multitude of employment sources. 12 Finding
463
mechanisms, such as provision of ®nancial
infrastructure that would allow small-scale
savings could, by reducing the need for socially
inecient diversi®cation, be associated with an
increase in overall welfare in rural areas.
Large households are more likely to adopt
diversi®ed strategies: The fact that, for any
given level of asset endowment, households
with a larger number of adults are also more
likely to adopt multiple income-generation
strategies (Table 3) suggests that, in addition to
markets for credit, markets for land and labor
also suer from considerable imperfections.
Instead of specializing in one main activity and
adjusting to variations in household size (which
may be life-cycle related) through the land
rental market, large households appear to be
forced to adopt multiple income-generating
strategies, even if this is not in line with the
specialized skills they possess. On the other
hand, contrary to a priori expectations, we are
unable to ascertain dierences in the coecients
on the number of household members below
and above the age of 35 and therefore only
report the total number of adults in the
household.
Education is an important determinant of
specialization: More educated households are
less likely to adopt multiple strategies of
income generation. This is likely to re¯ect the
co-existence of low-paying ``menial'' jobs with
little human capital requirements side-by-side
with activities that are characterized by high
entry barriers (such as possession of a minimum level of human capital). Overcoming
these entry barriers is a sunk investment.
Unless they are forced to do so, households
who successfully managed to overcome these
barriers will not diversify into areas that have
lower returns, thus explaining the positive and
highly signi®cant coecient on this variable.
4. CONCLUSION AND POLICY
IMPLICATIONS
In addition to con®rming the importance of
nonfarm activities as a source of income and
employment, our data also support the
hypothesis that, in view of the relatively
unequal distribution of assets and land, ofarm employment in Colombia falls into two
quite distinct categories. A signi®cant share of
poor households engages in a combination of
wage labor in jobs with low entry requirements
plus self-employment in ``marginal'' on-farm or
464
WORLD DEVELOPMENT
informal sector activities, neither of which
provide the returns required to sustain signi®cant investment and oer prospect for longerterm accumulation. At the same time, nonfarm
employment oers increased opportunities for
enhanced specialization which increase the
welfare and the capacity to invest of households
who are able to overcome the associated entry
barriers, thereby providing the basis for longerterm development of the rural sector.
Our analysis suggests that, in addition to
creating the pre-conditions for vigorous
growth of the nonfarm sector, government can
help to maximize the private and social bene®ts from such growth through three steps,
namely by (a) improving the functioning of
land, insurance, and credit markets; (b)
investing in human capital; and (c) taking
steps to help improve the asset endowments of
the poor. By enabling households to specialize
and make full use of the opportunities inherent in the development of a nonfarm sector,
doing so will increase individual as well as
social welfare. The Asian example where, in an
environment with relatively egalitarian distribution of income, well-functioning factor
markets, and a strong emphasis on educational expansion, rural nonfarm employment
has led to a spurt of broad-based development
and rapid income increases for all rural
inhabitants (Hayami, 1995) suggests that such
a strategy could provide large bene®ts not
only to rural dwellers but to the economy as a
whole.
NOTES
1. In India, an initial increase in easy entry o-farm
jobs which are relatively low-paying gives way to the
expansion of better-paying o-farm opportunities which
are created in response to the demand for nonfarm
products and services (Lanjouw & Stern, 1993). In the
Philippines it is found that the expansion of employment
opportunities outside of the farming sector precipitates
an increase in the returns to human capital through
migration which gives rise to a successive shift away
from farming toward nonfarm employment (Estudillo &
Otsuka, 1998). Household censuses in Latin America
also show a secular increase in the importance of rural
nonfarm employment (Klein, 1992). Of course, globalization can, in certain cases, also reduce the extent of
rural nonfarm employment.
2. Lack of access to assets has been identi®ed as a
major cause of poverty in Colombia (Leibovich &
Nunez, 1999).
3. The sample was strati®ed into 11 agro-ecological
zones. In each of the zones, 10 municipalities and within
these municipalities 10 households were selected
randomly. All households were surveyed two times,
once in 1997 and then again in 1999. Due to attrition
and the inability to visit a number of localities because of
violence, the sample in the second round was reduced
from 1,075 to 808.
4. A very limited number of households (2%) use paid
technical assistance.
5. The presence of a U-shaped relationship is
con®rmed by regression analysis (not reported).
6. Households appear to be willing to accept a lower
return on labor as a ``risk premium'' in return for the
risk diversi®cation advantages associated with the
adoption of reliance on a multiplicity of income sources.
7. Per capita expenditure is usually thought of being a
better proxy to permanent per capita income since it
captures household's ability to smooth consumption.
8. In the second regression reported in Table 2, we
include a further interaction between specialization
and the amount of labor supplied to the nonfarm
sector to test whether there is a statistically signi®cant
dierence between returns to labor obtained by households who specialize in on-farm and on-farm activities,
respectively.
9. Details for this approach of using ®rst dierenced
variables as instruments for endogenous level variables
are given in Hausmann and Taylor Edward (1981).
10. Since the high correlation of land and livestock
(with a correlation coecient of more than 0.6) resulted
in instability of the coecients, we added them
together.
11. This is con®rmed by the fact that, as discussed
earlier, high cost of credit constitutes a powerful
deterrent to the use of credit. Valentine (1993) and
Reardon and Taylor (1996) also ®nd that nonfarm
income allows households to draw on non-covariate
streams of income, thus increasing their ability to deal
with and recover from shocks that may otherwise have
disastrous consequences.
COLOMBIA
12. The negative sign of the squared term points to
the presence of decreasing marginal impact of such
asset-ownership on the propensity to specialize. By
465
comparison, the coecient on machinery is signi®cant
at the 10% level and enterprise assets is not signi®cant.
REFERENCES
Adams, R. H. J. (1994). Nonfarm income and inequality
in rural Pakistan: a decomposition analysis. Journal
of Development Studies, 31(1), 110±133.
Andre, C., & Platteau, J.-P. (1998). Land relations under
unbearable stress: Rwanda caught in the Malthusian
trap. Journal of Economic Behavior and Organization,
34(1), 1±47.
Berdegue, J. A., Reardon, T., & Escobar, G. (2000).
Empleo e ingreso rurales no agricolas en America
Latina y el Caribe. Paper presented at the conference
Development of the Rural Economy and Poverty
Reduction in Latin America and the Caribbean, New
Orleans, Louisiana, March 24.
De Janvry, A., Gordillo de Anda, G., & Sadoulet, E.
(1997). Mexico's second Agrarian reform: household
and community responses, 1990±94. La Jolla: Center
for US±Mexican Studies, University of California at
San Diego.
Estudillo, J. W., & Otsuka, K. (1998). Green revolution,
human capital and o-farm employment: changing
sources of income among farm households in central
Luzon, 1966±94. Economic Development and Cultural
Change, 47(3), 497±523.
Feldman, S., & Leones, J. P. (1998). Nonfarm activity
and rural household income: evidence from Philippine microdata. Economic Development and Cultural
Change, 46(4), 789±806.
Francis, E., & Hoddinott, J. (1993). Migration and dierentiation in western Kenya: a tale of two sub-locations. Journal of Development Studies, 30(1), 115±145.
Garcia, M., & Alderman, H. (1993). Food security and
health security: explaining the levels of nutritional
status in Pakistan. Economic Development and
Cultural Change, 42(3), 485±507.
Hare, D. (1999). Push versus pull factors in migration
out¯ows and returns: determinants of migration
status and spell duration among China's rural population. Journal of Development Studies, 35(3), 45±72.
Hausmann, J. A., & Taylor, W. (1981). Panel data and
unobservable individual eects. Econometrica, 49,
1377±1399.
Hayami, Y. (Ed.) (1998). Towards the rural-based
development of commerce and industry. Selected
experiences from East Asia. EDI Learning Resource
Series. Washington DC: EDI.
Ilahi, N. (1999). Return migration and cccupational
change. Review of Development Economics, 3(2), 170±
186.
Jaramillo, C. F. (1998). Liberalization crisis and change in
Colombian agriculture. Boulder: Westview Press.
Klein, E. (1992). El empleo rural no agricola en
America Latina, documento de trabajo, No. 364,
Santiago de Chile, Programa Regional del Empleo
para America Latina y el Caribe (PREALC),
August.
Lanjouw, P., & Stern, N. (1993). Markets, opportunities
and changes in inequality in Palanpur 1957±1984. In
A. Braverman, K. Ho, & J. Stiglitz, The economics
of rural organization: Theory, practice and policy.
New York: Oxford University Press.
Leibovich, J., & Nunez, J. (1999). Activos y recursos de
la poblacion pobre en Colombia. El Trimestre
Economico, 66(3), 501±551.
Mochebelele, M. T., & Winter-Nelson, A. (2000).
Migrant labor and farm technical eciency in
Lesotho. World Development, 28(1), 143±153.
Newbery, D., & Stiglitz, J. (1981). The theory of
commodity price stabilization: A study in the economics of risk. Oxford: Clarendon Press.
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±748.
Reardon, T., & Taylor, J. E. (1996). Agroclimatic
shock income inequality and poverty: evidence
from Burkina Faso. World Development, 24(5),
901±914.
Reardon, T. et al. (1998). Rural non-farm income in
developing countries. In FAO, The state of food and
agriculture. Rome: FAO.
Reardon, T., Taylor, J. E., Stamoulis, K., Lanjouw, P.,
& Balisacan, A. (2000). Eects of nonfarm employment on rural income inequality in developing
countries: an investment perspective. Journal of
Agricultural Economics, 51(2), 266±288.
Rozelle, S., Taylor, J. E., & DeBrauw, A. (1999).
Migration remittances and agricultural productivity
in China. American Economic Review, 89(2), 287±
291.
Taylor, J. E., Yunez-Naude, A. (1999). Education,
migration and productivity: An analytic approach
and evidence from rural Mexico. Paris: OECD
Development Centre Studies.
Udry, C. (1997). Recent advances in empirical microeconomic research in poor countries. Mimeo, North
western University.
Valentine, T. R. (1993). Drought transfer entitlements
and income distribution: the Botswana experience.
World Development, 21(1), 109±126.
World Bank (1996). Colombia poverty assessment.
Washington, DC: World Bank.
Zhao, Y. (1999). Labor migration and earnings dierences: the case of rural China. Economic Development and Cultural Change, 47(4), 767±782.