NIH Public Access
Author Manuscript
Land use policy. Author manuscript; available in PMC 2015 January 01.
NIH-PA Author Manuscript
Published in final edited form as:
Land use policy. 2014 January 1; 36: . doi:10.1016/j.landusepol.2013.07.006.
Consequences of Out-Migration for Land Use in Rural Ecuador
Clark L. Gray1,* and Richard E. Bilsborrow1
1University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA
Abstract
NIH-PA Author Manuscript
In rural Ecuador and elsewhere in Latin America, the departure of migrants and the receipt of
migrant remittances have led to declining rural populations and increasing cash incomes. It is
commonly assumed that these processes will lead to agricultural abandonment and the regrowth of
native vegetation, thus undermining traditional livelihoods and providing a boon for biodiversity
conservation. However, an increasing number of household-level studies have found mixed and
complex effects of out-migration and remittances on agriculture. We advance this literature by
using household survey data and satellite imagery from three study areas in rural Ecuador to
investigate the effects of migration and remittances on agricultural land use. Multivariate methods
are used to disaggregate the effects of migration and remittances, to account for other influences
on land use and to correct for the potential endogeneity of migration and remittances. Contrary to
common assumptions but consistent with previous studies, we find that migrant departure has a
positive effect on agricultural activities that is offset by migrant remittances. These results suggest
that rural out-migration alone is not likely to lead to a forest transition in the study areas.
Keywords
land use; agriculture; forest transition; migration; remittances
1. Introduction
NIH-PA Author Manuscript
Rural out-migration, or the departure of people from rural areas, is a key transformative
process in agricultural regions of the developing world. In Latin America, out-migration and
fertility decline have led to low or negative population growth rates in many rural areas,
some of which have also received substantial remittances from international migrants (Carr
et al., 2009; UN 2010). Out-migration thus has important implications for both the incomes
and labor resources of agricultural households and communities, but the net effect of these
processes on livelihoods and land use change are unclear. Many authors have argued that
out-migration and remittances undermine traditional agricultural activities and lead to land
abandonment and the regrowth of native vegetation (Aide and Grau 2004; Hecht 2010).
These changes could potentially contribute to “forest transitions” that protect biodiversity in
these areas (Rudel et al., 2005).
© 2013 Elsevier Ltd. All rights reserved.
*
Corresponding author: Clark L. Gray, Assistant Professor, Department of Geography, CB 3220, University of North Carolina at
Chapel Hill Chapel Hill, NC 27599, USA, http://geography.unc.edu/people/faculty-1/clark-gray, cgray@email.unc.edu,
001-919-962-3876.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Gray and Bilsborrow
Page 2
NIH-PA Author Manuscript
However, there are good theoretical and empirical reasons to be skeptical of this argument.
From a theoretical perspective, this view does not account for the fact that many rural
households allocate their labor to other economic activities beyond agriculture (Ellis 2000),
and that remittances can be invested in labor-saving agricultural inputs (Rozelle et al.,
1999). From an empirical perspective, several recent studies using econometric and spatial
approaches, as well as traditional ethnographic approaches, have found mixed or weak
effects of migration and remittances on agriculture in origin areas. Complementing the indepth but small-scale understanding of processes that come from ethnographic studies (e.g.,
Jokisch 2002), econometric approaches offer the opportunity to clearly distinguish between
the effects of migration and remittances while at the same time explicitly accounting for the
selectivity of migration (Rozelle et al., 1999). Studies using spatial and remote-sensing
methods provide the ability to view the landscape as a whole, and thus to derive conclusions
about the overall environmental impacts of migration (Müller et al., 2009). Up to now very
few studies have combined these approaches in a single study area (but see Rudel et al.,
2002).
NIH-PA Author Manuscript
We investigate these issues in the context of rural Ecuador, a highly biodiverse country that
is an important origin area of international migrants. We use a novel approach that combines
a household-level econometric analysis of land use with a community-level remote-sensing
analysis of vegetation change. These analyses draw on data from an original household and
community survey conducted in three study areas in rural Ecuador, as well as from a linked
spatial analysis of satellite imagery and other data sources. In the household-level analysis,
we use data from 440 households to model the effects of migration and remittances on
agricultural activities while accounting for other influences as well as the potential for
endogeneity. In the community-level analysis we use data from 80 communities to model
the effects of outmigration on changes in vegetation greenness while accounting for other
social and biophysical characteristics of communities. Together the two analyses provide
insights into both the economic and ecological effects of migration, with results that
challenge common assumptions about these processes. This work considerably expands our
previous study on this topic (Gray 2009a), and complements recent work by ourselves and
others on other aspects of the migrationenvironment nexus (Carr 2008; Gray 2009b; Massey
et al. 2010; Fussell et al. 2010; Dillon et al. 2011; Gray and Mueller 2012; López-Carr 2012;
Gray and Bilsborrow 2013).
2. Literature Review
2.1. Theoretical perspectives
NIH-PA Author Manuscript
In the rural developing world, households commonly draw on a variety of assets to invest in
a diverse portfolio of livelihood activities, including subsistence agriculture, cash cropping,
wage labor and migration (Ellis 2000). Migration is often part of a household strategy to
diversify into a new income source in the form of migrant remittances (Stark and Bloom
1985), though participation in migration is often constrained by lack of access to financial
capital and migrant networks (Massey and Espinosa 1997). The participation of individuals
in migration and agricultural activities is also highly selective, and in the case of rural Latin
America, men are often more likely to both participate in agricultural labor and to become
international migrants (Katz 2003). In this context, the departure of a household member
reduces household labor supply, which will reduce labor inputs to agriculture and other
activities unless there is a compensating increase in effort by the remaining household
members. If the migrant sends remittances, these could be used directly for household
consumption, substituting for agricultural production, and/or for hiring labor to compensate
for the absent household member. Remittances could also be invested in the intensification
of land use and the purchase of agricultural inputs such as fertilizer or pesticides. The
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 3
outcome in particular cases will depend on the risks and opportunities represented by
various livelihood strategies, among other factors.
NIH-PA Author Manuscript
Despite these many options available to rural households, many authors in environmental
studies and development studies have argued that out-migration and remittances will lead to
agricultural decline and disintensification (e.g., a reduction in agricultural intensity). In
environmental studies, this idea is part of “forest transition theory”, in which rural outmigration is viewed as a key mechanism leading to land abandonment and the subsequent
regrowth of native vegetation (Rudel et al., 2005). In the development studies literature, outmigration and remittances are often seen to undermine traditional livelihood activities, such
as subsistence agriculture, as part of a “migrant syndrome” (Reichert 1981; Jones 2009).
These perspectives do not account for the diversified nature of rural livelihoods, the
significant adaptability of rural households in the face of demographic and economic change
(Bilsborrow 1987; Netting 1993), and the numerous alternative pathways of adaptation of
described above. Both the forest transition and migrant syndrome frameworks have been
widely criticized as overly simplistic (Taylor et al., 1996; Perz 2007; García-Barrios et al.,
2009; Robson and Berkes 2011), but the view that agricultural decline and environmental
restoration follow out-migration remains common (Aide and Grau 2004; Kauppi et al.,
2006; Meyerson et al., 2007; Hecht 2010).
NIH-PA Author Manuscript
2.2. Previous empirical studies
Previous studies of the consequences of migration for agriculture have employed qualitative,
econometric and spatial methods, and have variously found evidence of positive, negative
and zero net effects.
NIH-PA Author Manuscript
Numerous qualitative and small-scale studies have approached this issue through
ethnography and intensive observation of a small number of communities. Among these
studies, several have found evidence of disintensification of agriculture (Zimmerer 1993;
Preston et al 1997; Schmook and Radel 2008; Jones 2009; Qin 2010; Robson and Berkes
2011), while others have observed small or zero net effects (Black 1993; Klooster 2003) or
intensification through the investment of remittances (McKay 2005; Taylor et al., 2006; De
Haas 2006). Three studies have examined Ecuador specifically. Preston and Taveras (1980)
investigated agricultural change following outmigration in six rural communities in the
Ecuadorian highlands, and found that in most cases land belonging to internal migrants was
rented out or sold for further use instead of abandoned. Jokisch (2002) later examined two
communities in the central Ecuadorian highlands, and found that, despite rapid international
out-migration and large remittance inflows, smallholder agriculture continued mostly
unchanged. Finally, Rudel et al., (2002) similarly observed significant out-migration from a
rural community in the Ecuadorian Amazon without significant land use change.
A number of econometric studies have also investigated this issue, typically using crosssectional or longitudinal survey data in instrumental-variable regression models. Migration
and remittances are potentially endogenous because they are not randomly assigned but
instead reflect selection on both observed and unobserved characteristics. This issue
complicates simple comparisons of migrant-sending and non-migrant-sending households
because any differences between these households may be due to either the drivers or
consequences of migration. The instrumental variable approach directly addresses this
concern, but it requires the use of one or more variables, known as instruments, that affect
migration and/or remittances but do not affect the final outcome of interest (Wooldridge
2001). Rozelle et al., (1999) pioneered the use of this approach to study migration in a study
of smallholder agriculture in China, drawing on migrant networks and remittance norms as
instruments. This study revealed that out-migration and remittances had countervailing
negative and positive effects, respectively, on maize yields.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 4
NIH-PA Author Manuscript
Several other studies have since used similar approaches (e.g., Mendola 2008; Miluka et al.,
2010; Atamanov and Van den Berg 2012), including four from Latin America. Pfeiffer and
Taylor (2007) found that participation in cash cropping in Mexico declined with the
departure of international migrants, particularly for male migrants, but that growing staple
crops was unchanged. Damon (2010) similarly showed that area in cash crops in El Salvador
declined following international migration, but the area in staple crops unexpectedly
increased. In contrast, Vanwey et al., (2012) found in Amazonian Brazil that cash crops
increased with remittances and that migrant departure had no effect, though this study did
not account for endogeneity. Finally, in a study from Ecuador, Vasco (2011) showed that
fertilizer use and cattle ownership unexpectedly increased with migrant departure while
remittances had no effect. Thus, consistent with qualitative studies, econometric studies find
mixed and diverse effects from migration and remittances on household agricultural
activities.
NIH-PA Author Manuscript
Finally, a small number of studies have used spatial methods, including geographic
information systems and remote sensing, to evaluate the consequences of migration for
agriculture. In an early study that foreshadowed this approach, Rudel et al., (2000) linked
data on reforestation from 650 field plots in Puerto Rico to census data on demographic
change and other factors, revealing that reforestation increased as the local population
declined due to out-migration. This approach has since been extended to include satellite
data sources by multiple studies from Europe (Gellrich and Zimmermann 2007; Sikor et al.,
2009; Müller et al., 2009; Bauman et al., 2011) and two studies from Latin America. López
et al., (2006) found a positive correlation at the municipal scale between out-migration and
the expansion of shrublands in rural Mexico, and Hecht and Saachti (2007) similarly found a
positive correlation between the receipt of remittances and forest cover at the provincial
scale in El Salvador. However, neither of these two studies controlled for the effects of other
factors on land use change, so the results may not be robust to multivariate controls.
Our study draws upon the econometric and spatial approaches described above to investigate
the consequences of out-migration and remittances flows for agricultural land use in rural
Ecuador. A special aspect of this study is our use of independent data from both household
reports of land use and community-level measures of vegetation change in the same study
areas, allowing us to compare the results of these two distinct perspectives on land use
change. In both cases, we control for potential confounders, and in the household case we
attempt to explicitly account for the potential endogeneity of out-migration and remittances.
We also considerably expand upon a pilot study (Gray 2009a), which found mixed and weak
effects of migration on agriculture in a smaller Ecuadorian study area, but did not include
corrections for endogeneity or incorporate satellite data.
NIH-PA Author Manuscript
3. Data Collection
To investigate these relationships, we conducted a household survey in three study areas in
rural Ecuador, along with a parallel analysis of vegetation change using satellite imagery.
3.1. Study areas
Based on site visits, census data and environmental data sources, three study areas were
selected in the Andean highlands and transition zone to the Pacific coastal plain to capture a
variety of biophysical environments and patterns of out-migration (Figure 1). Each area
consists of 5–6 contiguous cantons (administrative units roughly equivalent to US counties),
which together contained 7% of Ecuador’s rural population in 2001. In the analyses
described below, we combine data from all three areas to provide additional statistical power
and to produce results generalizable across multiple agricultural contexts.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 5
NIH-PA Author Manuscript
The Santo Domingo study area is located in the transition zone between the Andes and the
Pacific coastal lowlands and encompasses a wide range of environments, from mountainous,
heavily forested areas in the east to flat and intensively cultivated areas in the west. The
climate is humid and tropical, key crops include heart of palm, cacao and plantains, and land
ownership is mixed between small farms and large ranches or plantations. Out-migration is
primarily to other coastal provinces.
The high-elevation Chimborazo/Cañar study area overlaps these two provinces, and includes
both high-elevation grasslands and densely-settled valleys with temperate climates.
Smallholder agriculture is the dominant land use in settled areas, and key crops include
maize, beans and potatoes. This area is part of Ecuador’s international out-migration
heartland. Migration to the United States became common in the 1990s and was later
superseded by a (nationwide) trend of migration to Spain beginning in the late 1990s
(Jokisch and Pribilsky 2002).
NIH-PA Author Manuscript
The third study area, in western Loja province, lies in the western Andean foothills but has
an unusually dry climate with recurrent droughts. Coffee-centered agroforestry, cattle
raising and maize-centered agriculture are the key agricultural activities, but population
densities and land use intensity are low relative to the other two study areas. This region is a
traditional sending region of internal migrants, to urban, coastal and Amazon destinations,
though international migration, primarily to Spain, also became common after 2000 (Jokisch
and Pribilsky 2002)
Results from the household survey, described below, provide additional insight into the
nature of the agricultural systems in the three study areas. The overall mean farm size is
approximately 6 hectares, with 10% rented in or loaned to the household. Farms in
Chimborazo/Cañar are smaller on average (4 hectares) while those in Santo Domingo are
much larger (16 hectares). Across all three areas, the largest proportion of land use is
allocated to pasture (37%), with annuals, perennials and forest each constituting about 20%,
and the small remainder in fallow and other uses. Annuals such as corn and beans are
produced primarily for subsistence, whereas cattle, wage labor, remittances, and perennial
crops are the main sources of cash income. Participation in agriculture is highly gendered,
with 80% of young men (ages 15–29) reporting participation in farm work compared to 44%
of young women.
3.2. Sampling and interviews
NIH-PA Author Manuscript
In each of the three study areas, sample communities and households were selected using a
stratified, multi-stage cluster sample which included procedures to oversample both
migrantsending areas and households. Full details of the sampling methodology are
available in Gray and Bilsborrow (2013). Briefly, rural parishes (administrative units nested
within cantons) from the 17 study cantons were selected in the first stage with probabilities
of selection proportional to the rate of out-migration in 1996–2001, derived from the 2001
census. In the second stage, 1–3 census sectors were randomly sampled within the sample
parishes, reflecting differences in the population sizes of parishes across the three study
areas. Sample sectors were then visited to identify the constituent rural communities and to
list all resident households and migrants who had departed since January 1, 2000. Listed
households were classified into strata based on the destination of out-migrants, and a
stratified sample was drawn using a set of sampling rules that oversampled households with
migrants, and more so for those with rare migrant types. The final sample contained 29
parishes, 55 census sectors, 106 rural communities and 869 households, 97% of whom
provided a complete household interview.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 6
NIH-PA Author Manuscript
Fieldwork was carried out in June-December 2008 by a team of Ecuadorian interviewers.
Interviewers used a structured questionnaire to collect information on livelihood activities in
the past year and a variety of retrospective information dating back to 2000. Detailed data
was collected on livelihood activities in the past year, including production, use of inputs
and sales from agriculture as well as the value and frequency of migrant remittances
received. Plot-level information was also collected on land use and perceived land quality.
Additionally, retrospective information was collected back to the year 2000 using an annual
time step and a format that allowed comparisons across related topics. This included data on
the agricultural and other economic activities of household members, as well as the location/
residence of each current member and all adult out-migrants. Limited information was also
collected for the period prior to 2000, including the destinations of out-migrants from the
household prior to 2000.
Finally, a community questionnaire was implemented with community leaders to obtain
crosssectional and retrospective information about population size, migration, infrastructure
and economic activities.
3.3. Spatial analysis
NIH-PA Author Manuscript
Global Positioning System points were collected at community centers, the location of each
household, and a subset of agricultural plots. For the present analysis, community territories
were defined to encompass a 1 km radius from the community center, a radius consistent
with our field observations and the distribution of household plots. These territories were
used to extract values from two existing environmental datasets and four satellite images.
First, we used a Digital Elevation Model of Ecuador with 30 m resolution (Souris 2006) to
extract mean elevation, mean slope and the standard deviation of slope, a measure of terrain
unevenness. Second, we used the global WorldClim dataset to obtain historical measures of
annual precipitation and temperature as well as their seasonal variation. This dataset
contains interpolated climate information at a 1 km resolution, estimated as a historical
mean from 1950–2000 (Hijmans et al., 2005).
NIH-PA Author Manuscript
To measure vegetation change, we acquired four cloud-free Landsat ETM+ satellite images
with 30 meter resolution, two covering the Cañar/Chimborazo study area (from Nov 3, 2001
and Aug 31, 2007) and two covering the Loja study area (from Oct 31, 2000 and Sept 19,
2008). Adequate cloud-free imagery was not available for the Santo Domingo study area in
any year since 2000, necessitating the omission of that area from the community-level
analysis. Cloud-free data was available for 78 study communities. Two measures of
vegetation greenness, the Normalized Difference Vegetation Index (NDVI; Pettorelli et al.,
2005) and the Enhanced Vegetation Index (EVI; Huete et al., 2002), were then derived from
these images and extracted using 1 km buffers as described above. These indices are
transformations of the strength of reflectance in different spectral bands and measure the
extent of plant growth and photosynthetic activity, i.e. “greenness”. Both indices range from
−1 to 1, but here all values have been multiplied by 10 to facilitate interpretation of the
results. While these values cannot be directly interpreted in terms of land cover, a
supplementary land use classification of the 2007–2008 images revealed that greenness was
highest for annual crops, followed by coffee, forest, pasture and fallow.
4. Analysis
To investigate the consequences of migration and remittances for agriculture, we use the
datasets described above to conduct two sets of analyses, one at the household level using
survey-derived outcomes and the other at the community level using satellite-derived
outcomes. These two sets of analyses have different strengths and thus serve as a robustness
check on our findings.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 7
4.1. Household-level analysis
NIH-PA Author Manuscript
The first set of analyses examines the influence of migrant departure and receipt of
remittances on household agricultural activities in the 12 months prior to interview. We use
data from the 441 households who grew crops during this period and who were also resident
in the study community in the baseline year 2000. To avoid reverse causation between
migration and agriculture, we examine the effects of previous migration and remittances on
current agricultural outcomes while controlling for pre-migration characteristics, capitalizing
on the retrospective nature of the survey. The four outcomes are area planted in annual
crops, expenditures on modern agricultural inputs, days of labor hired, and total production
of annual crops. We focus on annual crops and inputs used on all plots in the past year
because the production of long-lived perennials such coffee and cacao, as well as livestock
such as cattle1, reflects decisions made many years prior.
NIH-PA Author Manuscript
Descriptive statistics for the outcomes are shown in Table 1. 83% of households grew
annual crops with a mean area of 1.1 hectares, 68% used modern inputs with a mean cost of
US$241, 35% hired agricultural laborers with a mean duration of 20 person-days, and the
average production of annuals was valued at US$598. 53% of households sold crops in the
past year. For crops consumed rather than sold, production was valued at the mean price
obtained by households who sold the crop. Maize was the most common primary annual
crop, followed by beans, potatoes and peanuts. Expenses on inputs went primarily towards
fertilizers and herbicides, followed by pesticides and improved seeds. Because the four
outcome variables are censored at zero and right-skewed, they were transformed by ln(y+1)
prior to analysis and we use regression models appropriate for censored outcomes as
described below.
To capture the effects of migration and remittances, four predictors were constructed: the
number of migrants sent by the household in the period 2000–2008, the numbers of male
and female migrants separately, and the value of monetary remittances received in the
previous 12 months (Table 1). Migrants were defined as individuals resident in the
household in 2000 who by 2008 lived outside of the canton. On average, households sent
less than one migrant (0.8) and received approximately US$290 in remittances. Among
migrants, 59% of were male and 32% went to international destinations. International
migrants, primarily to Spain and United States, remitted $1040 on average, representing
90% of all remittances received, versus $50 on average for internal migrants.
NIH-PA Author Manuscript
Fourteen control variables capturing pre-migration characteristics of the household and
community in the year 2000 are also included in the analysis to account for other potential
influences on land use. Selection of the control variables was motivated by the rural
livelihoods framework as described above (Ellis 2000). Household-level controls include
farm area, ownership of flat land, ownership of land with black soil, irrigation and/or
perennials, an index of non-agricultural wealth, the number of cattle owned, household size,
and the gender, age, and employment status of the household head. Values of farm area and
the number of cattle are right-skewed and were thus transformed by ln(x+1) prior to analysis
to reduce the influence of outlying values. The wealth index was created by using polychoric
principle components analysis (Kolenikov and Angeles 2009) to combine several
dichotomous and categorical measurements of housing quality and ownership of various
assets. Community-level control variables include the type of road surface and a climate
index measuring the extent to which the site had a tropical (versus temperate) climate. This
index was created by taking the first principal component of the climate variables derived
1For this reason we do not examine cattle ownership as an outcome in the household-level analysis. However it is worth noting that
cattle ownership did increase over the study period from 3.4 to 3.9 cattle per household, potentially leading to an increase in pasture
and affecting vegetation greenness, as we note below.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 8
from WorldClim as described above. Finally, largerscale contextual factors are accounted
for by canton-level fixed effects as described below.
NIH-PA Author Manuscript
To examine the influences of these factors on agriculture, we first estimate a series of tobit
models using various specifications of migration and remittances. Tobit models are an
extension of multiple regression that are appropriate for censored outcomes, i.e., outcomes
with a large proportion of zeroes. To account for the clustering of observations within
sample sectors and for larger-scale contextual effects, all standard errors have been adjusted
for clustering (Angeles et al., 2005) and all models include canton-level fixed effects, i.e., an
indicator variable for each canton. All models and descriptive statistics also incorporate
sampling weights, derived as the inverse of the probability of selection. We first estimate
three tobit models with the following specifications of migration and remittances: total
migrants only, total migrants along with the value of remittances, and, finally, migrants
separated by gender as well as the value of remittances. We present coefficients and
significance tests for these models, and later derive marginal effects of the predictors
conditional on the outcome being greater than zero. Outcome values are missing for three
households on the area and production of annuals, four households on input expenditures,
and two households on the person-days of hired labor, resulting in slightly different sample
sizes for the four models.
NIH-PA Author Manuscript
To examine whether these effects are due to endogeneity, we then re-estimate these models
as linear instrumental variable models. In this approach, migration and remittances are each
separately considered to be a function of the control variables plus a collection of
instrumental variables which are assumed to influence migration and remittances from
2000–2008 but not agricultural activities in 2008. Consistent with previous econometric
studies (e.g., Rozelle et al., 1999), instrumental variables included indicators for the detailed
age composition of the household in 2000, the number of internal and international migrants
sent by the household prior to 2000, and the propensities of internal and international
migration from the parish prior to 2000 (derived from census data). To allow convergence
for models with multiple endogenous variables, models were estimated as linear
instrumental variable models rather than tobit models, and thus account for endogeneity but
not censoring. All models reported here meet suggested thresholds for the strength,
significance and exogeneity of instruments, and all models also include adjustments for
clustering and canton-level fixed effects. Descriptive statistics for the instrumental variables
and detailed results for the instrumental variable models are available upon request.
4.2. Community-level analysis
NIH-PA Author Manuscript
The second set of analyses examines the influence of migration on vegetation change at the
community level. As described above, we focus on 78 study communities where cloud-free
imagery was available. Our outcomes are the change in two measures of vegetation
greenness, NDVI and EVI, from 2000–2001 to 2007–2008, multiplied by 10 for ease of
interpretation. A positive value indicates increasing greenness over time, and, as shown in
Table 2, both values are positive on average, consistent with an increasing trend of rainfall
over the same period. To examine the influences of migration on these outcomes, we
estimated multiple regression (OLS) models, which as described above included sampling
weights, corrections for clustering at the level of the sample sector, and fixed effects at the
level of the canton.
Consistent with the larger scale and remotely-sensed outcome of this analysis, a distinct set
of predictors was developed. A measure of migration and 16 control variables were
extracted from the community survey, the household survey, and the biophysical datasets
described above. Migration here is measured as the total number migrants from 2000–2008
divided by the adult population of the community in 2008, i.e., the ratio of migrants to nonLand use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 9
NIH-PA Author Manuscript
migrants, using data derived from the household listing. (A measure of total remittances was
not available at the community level.) Biophysical and social variables that are likely to
affect agriculture and potentially correlated with migration were selected as control
variables. Controls extracted from the biophysical datasets include mean annual
precipitation and temperature, seasonal variation in these values, mean terrain slope,
variation in slope, and baseline values of NDVI and EVI. Control variables derived from the
community survey include the occurrence of a drought from 2000–2008, mean farm size,
population size in 2000, frequency of transportation interruptions, and whether the following
crops were cultivated in 2000: maize, coffee, potatoes and peanuts. Finally, the mean level
of adult education in the community in 2000 was extracted from the household survey.
5. Results
5.1. Household-level results
The results of the household-level analysis are presented in Table 3. We first briefly describe
the effects of the control factors, focusing on Specification 1, before turning to the effects of
migration and remittances across all three specifications.
NIH-PA Author Manuscript
The control factors had jointly significant effects on all four outcomes that are broadly
consistent with expectations. The area planted in annuals increased with farm size and
access to a paved road and decreased with land in perennials and with irrigation, the latter
effect likely reflecting the intensive cultivation of a smaller area. Expenses on modern inputs
in turn increased with flat land, tropical climates, and nonagricultural employment and
wealth, both of which facilitate the purchase of inputs. Input expenses also decreased with
access to a paved road, perhaps due to lower prices in these areas. Days of labor hired-in to
the farm increased with farm size, irrigation, education of the head, nonagricultural wealth,
and the location on a gravel road, potentially reflecting lower labor costs in these semiremote areas. Finally, the production of annuals decreased with land in perennials, female
household headship, and tropical climates, and increased with the presence of a gravel road.
The canton indicators were also jointly highly significant in all models, indicating the
importance of accounting for large-scale contextual factors.
To test the effects of migration and remittances, three specifications are presented in order of
increasing complexity. Specification 1 includes only the total number of migrants, and
reveals only weak and non-significant effects on each of the four dependent variables,
results which are confirmed in the instrumental variable models. Thus while the area in
annuals marginally increased with the number of out-migrants (p = 0.09), this effect
becomes non-significant in the model that includes corrections for endogeneity.
NIH-PA Author Manuscript
Specification 2 adds the value of remittances as a predictor, and an intriguing countervailing
pattern on the area in annuals is revealed. The area planted in annuals increased with the
departure of migrants (p = 0.002) but decreased with the receipt of remittances (p = 0.001),
results confirmed by the instrumental variable approach. To more clearly present the
magnitude of these effects, we derive marginal effects of the predictors for non-zero
outcomes, simulate the effects for a household with 1 hectare of annual crops, and translate
the results back into the original non-logarithmic units. This simulation reveals that the
departure of one migrant would increase cropped area slightly to 1.12 hectares, with a 95%
confidence interval of 1.05–1.20 hectares. The departure of a single migrant accompanied by
$1000 in remittances (the average sent by an international migrant) would have a net
negative effect, decreasing the area in annuals to 0.75 hectares, with a confidence interval of
0.52–1.03 hectares. Finally, the effects of migration and remittances on the other agricultural
outcomes remain non-significant, results again confirmed by the instrumental variable
models.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 10
NIH-PA Author Manuscript
Specification 3 examines whether there is a difference in the effects depending on the
gender of the out-migrant. This reveals that the magnitude of the effect of migrant departure
on area in annuals in similar for both male and female migrants, but the effect is highly
significant for men (p = 0.009) and only marginally significant for women (p = 0.098), a
result that holds once endogeneity is corrected. The effects of migration and remittances on
the other three agricultural outcomes remain largely non-significant. While the number of
male out-migrants is positively associated with person-days of hired labor, this result is not
robust to corrections for endogeneity.
5.2. Community-level results
NIH-PA Author Manuscript
The results of the community-level analysis are seen in Table 4. The control variables
generally had significant and similar effects on both measures of change in vegetation
greenness. Both measures increased with seasonal variation in precipitation, reports of past
drought, maize cultivation, and more frequent transportation interruptions, and decreased
with mean annual temperature, mean farm size, mean education and the cultivation of
potatoes. The change in NDVI also increased with the mean land slope and decreased with
the variation in slope and baseline NDVI. Together these results indicate that remote, midelevation sites with variable climates, small farms and low education experienced the most
greening, potentially reflecting recovery from past droughts and the expansion of agriculture
in poorer and remote areas. Finally, with respect to the key variable of interest, the effect of
the ratio of migrants to non-migrants is positive and significant for both change in NDVI (p
= 0.047) and change on EVI (p = 0.038). At the mean value of migration, this effect
accounts for 4.1% and 6.3% of the vegetation greening as measured by NDVI and EVI
respectively.
6. Discussion
NIH-PA Author Manuscript
Contrary to common assumptions, both the household and community-level results suggest
that migrant departure led to an expansion of cultivated area in rural Ecuador. At the
household level, migration and remittances appear to have countervailing effects, with the
area in annuals increasing with the number of migrants (particularly male migrants) and
declining with the value of remittances. Given that these changes in area occur without any
concurrent changes in inputs or production, the results suggest that households are
expanding the agricultural area while disintensifying production, i.e. lowering inputs and
production per unit area. This may reflect a strategy to maintain production by planting a
larger area but devoting less attention to laborintensive tasks such as weeding, a change
which could also increase vegetation greenness. The negative effect of remittances on
cultivated area suggests that remittances are substituting for agricultural production,
allowing a decrease in agricultural effort. This interpretation is consistent with the negative
but non-significant effects of remittances on the other three outcomes. For an international
migrant who sends the average value of remittances, the net effect is to slightly decrease the
area in annuals.
At the community level, out-migration had significant positive effects on both measures of
change in vegetation greenness. Given that annual crops have the highest levels of greenness
among regional land covers, this result suggests that migration led to an overall increase in
the area in annual crops, though it could also be partly accounted for by an increase in
irrigated pastures. Increasing greenness could also potentially indicate regrowth of native
vegetation on previously barren areas, but this interpretation is not consistent with the
household-level results described above or with our site visits to the study areas, where land
abandonment within 1 km of the community center was rarely observed.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 11
7. Conclusions
NIH-PA Author Manuscript
Drawing on original household survey data and a novel analysis combining econometric and
remote-sensing methods, we examine the effects of out-migration from rural areas and the
receipt of remittances on agricultural land use in rural Ecuador. Contrary to common
assumptions, we find evidence that rural out-migration leads to an expansion of agricultural
area, with remittances having a countervailing negative effect. These results are generally
consistent with the three most relevant previous studies, which found that the area planted in
staples increased with out-migration in El Salvador (Damon 2010) and that the effects of
male departure were greater than female departure in Mexico and southern Ecuador (Pfeiffer
and Taylor 2007; Gray 2009a). More broadly, the results are also consistent with the
literature as a whole in finding mixed, countervailing and relatively weak effects of
migration and remittances on agricultural activities (Jokisch 2002; Gray 2009a). Taken
together, these studies undermine common arguments that out-migration and receipt of
remittances will lead to agricultural abandonment and the regrowth of native vegetation in
the rural developing world (Aide and Grau 2004; Hecht 2010). Future studies should instead
recognize the significant flexibility of rural households in responding to demographic and
environmental change, and the important ways in which these household strategies mediate
population-environment relationships.
NIH-PA Author Manuscript
Future studies could also benefit from adapting and extending the methodological approach
described here. We show that combining econometric and GIScience approaches in the same
study area can contribute to better insights and more robust findings. Desirable extensions of
our approach for understanding migration-environment relationships include the
incorporation of panel survey data, the construction of a linked time-series of classified
remotely-sensed images, and linking households to land though the geolocation of
agricultural plots. In the context of human-environment research, the approach of combining
survey and remote sensing data has been pioneered in the study of frontier land use (Walsh
and Crews-Meyer 2002; Fox et al 2003).
We show here that these approaches are also relevant and complementary for the
investigation of migration-environment relationships, and indeed could also be productively
combined to study a wide variety of the other human-environment topics such as common
property resources, forest product collection, and the impacts of agricultural interventions.
Acknowledgments
NIH-PA Author Manuscript
Funding for this research was provided by the National Institutes of Health (HD052092, HD061752). We are
indebted to our Ecuadorian partners at CEPAR (Centro de Estudios de Poblacion y Desarrollo Social), as well as to
the team of field researchers and the participating communities. We thank Luis Vallejo and Brian Frizzelle for their
respective efforts on data cleaning and spatial analysis. Helpful comments on previous drafts were provided by
Anna Agbe-Davies, Jason Davis, Amanda Thompson, Colin West and Ken Young.
References
Aide T, Grau H. Globalization, migration, and Latin American ecosystems. Science. 2004; 305(5692):
1915–1916. [PubMed: 15448256]
Angeles G, Guilkey D, Mroz T. The impact of community-level variables on individual-level
outcomes: Theoretical results and applications. Sociological Methods and Research. 2005; 34(1):
76–121.
Atamanov A, Van den Berg M. Heterogeneous effects of international migration and remittances on
crop income: Evidence from the Kyrgyz Republic. World Development. 2012; 40(3):620–630.
Bilsborrow R. Population pressures and agricultural development in developing countries: A
conceptual framework and recent evidence. World Development. 1987; 15(2):1–18.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 12
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Black R. Migration, return, and agricultural development in the Serra Do Alvao, Northern Portugal.
Economic Development and Cultural Change. 1993; 41:563–585.
Carr D. Farm households and land use in a core conservation zone of the Maya Biosphere Reserve,
Guatemala. Human Ecology. 2008; 36(2):231–248.
Carr D, Lopez A, Bilsborrow R. The population, agriculture, and environment nexus in Latin America:
Country-level evidence from the latter half of the twentieth century. Population and Environment.
2009; 30(6):222–246.
Damon A. Agricultural Land Use and Asset Accumulation in Migrant Households: The Case of El
Salvador. Journal of Development Studies. 2010; 46(1):162–189.
De Haas H. Migration, remittances and regional development in Southern Morocco. Geoforum. 2006;
37(4):565–580.
Dillon A, Mueller V, Salau S. Migratory responses to agricultural risk in northern Nigeria. American
Journal of Agricultural Economics. 2011; 93(4):1048–1061.
Ellis, F. Rural Livelihoods and Diversity in Developing Countries. Oxford: Oxford University Press;
2000.
Fox, J.; Rindfuss, R.; Walsh, S.; Mishra, V. People and The Environment: Approaches for Linking
Household and Community Surveys to Remote Sensing and GIS. Boston: Kluwer Academic
Publishers; 2003.
Fussell E, Sastry N, VanLandingham M. Race, socioeconomic status, and return migration to New
Orleans after Hurricane Katrina. Population and Environment. 2010; 31(1–3):20–42. [PubMed:
20440381]
García-Barrios L, Galván-Miyoshi Y, Valdivieso-Pérez I, Masera O, Bocco G, Vandermeer J.
Neotropical forest conservation, agricultural intensification, and rural out-migration: The Mexican
experience. BioScience. 2009; 59(10):863–873.
Gellrich M, Zimmermann N. Investigating the regional-scale pattern of agricultural land abandonment
in the Swiss mountains: A spatial statistical modelling approach. Landscape and Urban Planning.
2007; 79(1):65–76.
Gray C. Rural out-migration and smallholder agriculture in the southern Ecuadorian Andes. Population
and Environment. 2009a; 30(4):193–217.
Gray C. Environment, land and rural out-migration in the southern Ecuadorian Andes. World
Development. 2009b; 37(2):457–468.
Gray C, Bilsborrow R. Environmental influences on human migration in rural Ecuador. 2013
Demography in press.
Gray C, Mueller V. Natural disasters and population mobility in Bangladesh. Proceedings of the
National Academy of Sciences. 2012; 109(16):6000–6005.
Hecht S. The new rurality: Globalization, peasants and the paradoxes of landscapes. Land Use Policy.
2010; 27(2):161–169.
Hecht S, Saatchi S. Globalization and forest resurgence: Changes in forest cover in El Salvador.
BioScience. 2007; 57(8):663–672.
Hijmans R, Cameron S, Parra J, Jones P, Jarvis A. Very high resolution interpolated climate surfaces
for global land areas. International Journal of Climatology. 2005; 25(15):1965–1978.
Huete A, Didan K, Miura T, Rodriguez E, Gao X, Ferreira L. Overview of the radiometric and
biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment.
2002; 83(1–2):195–213.
Jokisch B. Migration and agricultural change: The case of smallholder agriculture in highland
Ecuador. Human Ecology. 2002; 30:523–550.
Jokisch B, Pribilsky J. The panic to leave: Economic crisis and the ‘new emigration’ from Ecuador.
International Migration Review. 2002; 40:75–101.
Jones R. Migration permanence and village decline in Zacatecas: When you can't go home again. The
Professional Geographer. 2009; 61(3):382–399.
Katz, E. The changing role of women in the rural economies of Latin America. In: Davis, B., editor.
Current and emerging issues for economic analysis and policy research. Volume I: Latin America
and the Caribbean. Food and Agriculture Organization; 2003. p. 31-66.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 13
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Kauppi P, Ausubel J, Fang J, Mather A, Sedjo R, Waggoner P. Returning forests analyzed with the
forest identity. Proceedings of the National Academy of Sciences. 2006; 103(46):17574–17579.
Klooster D. Forest transitions in Mexico: Institutions and forests in a globalized countryside. The
Professional Geographer. 2003; 55(2):227–237.
Kolenikov S, Angeles G. Socioeconomic status measurement with discrete proxy variables: Is
principal component analysis a reliable answer? Review of Income and Wealth. 2009; 55(1):128–
165.
López E, Boccoa G, Mendoza M, Velázquez A, Aguirre-Rivera J. Peasant emigration and land-use
change at the watershed level: A GIS-based approach in Central Mexico. Agricultural Systems.
2006; 90:62–78.
López-Carr D. Agro-ecological drivers of rural out-migration to the Maya Biosphere Reserve,
Guatemala. Environmental Research Letters. 2012; 7:045603. [PubMed: 24069068]
Massey D, Axinn W, Ghimire D. Environmental change and out-migration: Evidence from Nepal.
Population and Environment. 2010; 32(2–3):109–136. [PubMed: 21350676]
Massey D, Espinosa K. What’s driving Mexico-U.S. migration? A theoretical, empirical, and policy
review. American Journal of Sociology. 1997; 102(4):939–999.
McKay D. Reading remittance landscapes: Female migration and agricultural transition in the
Philippines. Geografisk Tidsskrift-Danish Journal of Geography Volume. 2005; 105(1):89–99.
Mendola M. Migration and technological change in rural households: Complements or substitutes?
Journal of Development Economics. 2008; 85:150–175.
Meyerson F, Merino L, Durand J. Migration and environment in the context of globalization. Frontiers
in Ecology and the Environment. 2007; 5(4):182–190.
Miluka J, Carletto G, Davis B, Zezza A. The vanishing farms? The impact of international migration
on Albanian family farming. Journal of Development Studies. 2010; 46(1):140–161.
Müller D, Kuemmerle T, Rusu M, Griffiths P. Lost in transition: determinants of postsocialist cropland
abandonment in Romania. Journal of Land Use Science. 2009; 4(1):109–129.
Netting, R. Smallholders, householders: Farm families and the ecology of intensive, sustainable
agriculture. Stanford, CA: Stanford University Press; 1993.
Perz S. Grand theory and context-specificity in the study of forest dynamics: Forest transition theory
and other directions. The Professional Geographer. 2007; 59:105–114.
Pettorelli N, Vik J, Mysterud A, Gaillard J, Tucker C, Stenseth N. Using the satellite-derived NDVI to
assess ecological responses to environmental change. Trends in Ecology and Evolution. 2005;
20(9):503–510. [PubMed: 16701427]
Pfeiffer, L.; Taylor, J. Gender and the impacts of international migration: Evidence from rural Mexico.
In: Morrison, A.; Schiff, M.; Sjöblom, M., editors. The international migration of women.
Washington, D.C.: World Bank Publications; 2007. p. 99-123.
Preston D, Taveras G. Changes in land tenure and land distribution as a result of rural emigration in
Highland Ecuador. Tijdschrift voor Economische en Sociale Geografie. 1980; 71:98–107.
Preston D, Macklin M, Warburton J. Fewer people, less erosion: The twentieth century in southern
Bolivia. The Geographical Journal. 1997; 163:198–205.
Qin H. Rural-to-urban labor migration, household livelihoods, and the rural environment in Chongqing
Municipality, Southwest China. Human Ecology. 2010; 38(5):675–690.
Reichert J. The migrant syndrome: Seasonal US wage labor and rural development in central Mexico.
Human Organization. 1981; 40:56–66.
Robson J, Berkes F. Exploring some of the myths of land use change: Can rural to urban migration
drive declines in biodiversity? Global Environmental Change. 2011; 21(3):844–854.
Rozelle S, Taylor J, deBrauw A. Migration, remittances, and agricultural productivity in China. The
American Economic Review. 1999; 89(2):287–291.
Rudel T, Coomes O, Moran E, Achard F, Angelsen A, Xu J, Lambin E. Forest transitions: Towards a
global understanding of land use change. Global Environmental Change Part A. 2005; 15:23–31.
Rudel T, Bates D, Machinguiashi R. A tropical forest transition? Agricultural change, out-migration,
and secondary forests in the Ecuadorian Amazon. Annals of the Association of American
Geographers. 2002; 92(1):87–102.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 14
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Rudel T, Perez-Lugo M, Zichal H. When fields revert to forest: Development and spontaneous
reforestation in post-war Puerto Rico. The Professional Geographer. 2000; 52(3):386–397.
Schmook B, Radel C. International labor migration from a tropical development frontier: Globalizing
households and an incipient forest transition. The Southern Yucatán case. Human Ecology. 2008;
36(6):891–908.
Sikor T, Müller D, Stahl J. Land fragmentation and cropland abandonment in Albania: Implications for
the roles of state and community in post-socialist land consolidation. World Development. 2009;
37(8):1411–1423.
Souris M. Ecuador 30m Resolution DEM. L'Institut de Recherche pour le Développement. 2006 http://
rslultra.star.ait.ac.th/~souris/ecuador.htm.
Stark O, Bloom D. The new economics of labor migration. American Economic Review. 1985;
75:173–178.
Taylor J, Arango J, Hugo G, Kouaouci A, Massey D, Pellegrino A. International migration and
community development. Population Index. 1996; 62:397–418.
Taylor M, Moran-Taylor M, Ruiz D. Land, ethnic, and gender change: Transnational migration and its
effects on Guatemalan lives and landscapes. Geoforum. 2006; 37:41–61.
United Nations. World Urbanization Prospects: The 2009 Revision. 2010
VanWey L, Guedes G, D’Antona Á. Out-migration and land-use change in agricultural frontiers:
insights from Altamira settlement project. Population and Environment. 2012; 34(1):44–68.
[PubMed: 23129878]
Vasco, C. The impact of international migration and remittances on agricultural production patterns,
labor relationships and entrepreneurship: The case of rural Ecuador. Kassel University Press;
2011.
Walsh, S.; Crews-Meyer, K. Linking people, place, and policy: A GIScience approach. Springer; 2002.
White, M.; Lindstrom, D. Internal migration. In: Poston, D.; Micklin, M., editors. Handbook of
Population. New York: Academic Kluwer Publishers; 2005.
Wooldridge, J. Econometric Analysis of Cross Section and Panel Data. MIT Press; 2001.
Zimmerer K. Soil erosion and labor shortages in the Andes with special reference to Bolivia, 1953–91:
Implications for ‘conservation-with-development'. World Development. 1993; 21:1659–1675.
NIH-PA Author Manuscript
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 15
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 1.
Map of Ecuador showing the three study areas.
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 16
Table 1
Mean values of outcomes and predictors by migrant-sending status for the household-level analysis.
NIH-PA Author Manuscript
Full
sample
No
migrants
Migrants
but no
remittances
Migrants
and
remittances
Grew annual crops, 2008 (0/1)
0.83
0.83
0.82
0.84
Area in annual crops, 2008 (ha)
1.13
1.00
1.46
1.03
Used modern inputs, 2008 (0/1)
0.68
0.63
0.80
0.63
Variable
Outcome variables
Expenditures on modern inputs, 2008 (US$)
241
304
234
99
Hired agricultural labor, 2008 (0/1)
0.35
0.32
0.41
0.36
Days of hired labor, 2008 (person-days)
20.1
13.4
38.6
11.9
Production of annual crops, 2008 (US$)
598
569
564
712
Number of migrants, 2000–2008, (#)
0.80
0.00
1.49
1.80
Number of male migrants, 2000–2008 (#)
0.47
0.00
0.84
1.10
Number of female migrants, 2000–2008 (#)
0.33
0.00
0.65
0.69
Value of remittances, 2008 (US$)
293
0
0
1357
Area of land owned, 2000 (ha)
5.14
3.55
6.07
7.71
Owned flat land, 2000 (0/1)
0.18
0.20
0.12
0.23
Owned land with black soil, 2000 (0/1)
0.39
0.42
0.28
0.47
Owned land with irrigation, 2000 (0/1)
0.23
0.19
0.20
0.34
Owned land with perennials, 2000 (0/1)
0.29
0.27
0.30
0.30
Number of cattle, 2000 (#)
3.44
2.86
2.79
5.63
Household size, 2000 (#)
5.20
4.23
6.17
6.26
Female head, 2000 (0/1)
0.25
0.30
0.11
0.29
Age of head, 2000 (years)
45.7
44.5
44.0
50.5
Education of head, 2000 (years)
4.60
4.35
5.71
3.76
Nonagricultural work by head, 2000 (0/1)
0.15
0.12
0.27
0.06
Migration and remittances
NIH-PA Author Manuscript
Control variables
NIH-PA Author Manuscript
1
Nonagricultural wealth score, 2000 (0–10)
3.16
3.18
3.32
2.92
Paved road, 2000 (0/1)
0.18
0.24
0.11
0.13
Gravel road, 2000 (0/1)
0.42
0.37
0.50
0.42
Dirt or no road, 2000 (0/1)1
0.40
0.39
0.39
0.45
Tropical climate index (0–10)
4.30
4.57
4.39
3.54
Nhouseholds
441
183
131
127
Reference category
Land use policy. Author manuscript; available in PMC 2015 January 01.
Gray and Bilsborrow
Page 17
Table 2
Mean values of outcomes and predictors by region for the community-level analysis.
NIH-PA Author Manuscript
Variable
Full
Cañar/
Chimborazo
Loja
Change in NDVI x10, 2000–08
2.56
3.93
0.65
Change in EVI x10, 2000–08
2.95
4.47
0.82
Ratio of migrants/nonmigrants, 2000–08 (%)
0.30
0.28
0.32
Mean annual precipitation (cm)
100
78
129
Outcomes
Predictors
CV of monthly precipitations (%)
6.97
5.14
9.51
Experienced drought, 2000–08 (0/1)
0.23
0.28
0.15
Mean annual temperature (°C)
16.8
13.3
21.7
SD of monthly temperatures (°C)
3.01
3.24
2.69
Mean land slope, 2000 (%)
16.9
15.9
18.3
SD of land slope, 2000 (%)
7.18
7.33
6.97
NIH-PA Author Manuscript
Mean farm size, 2008 (ha)
3.03
2.21
4.17
Maize cultivated, 2000 (0/1)
0.83
0.77
0.91
Coffee cultivated, 2000 (0/1)
0.18
0.00
0.42
Potatoes cultivated, 2000 (0/1)
0.44
0.75
0.00
Peanuts cultivated, 2000 (0/1)
0.27
0.00
0.64
Population size, 2000 (#)
302
457
87
Mean education, 2000 (years)
4.70
3.97
5.70
Rare transport interruptions, 2000–2008 (0/1)1
0.11
0.15
0.07
Occasional transport interruptions, 2000–08 (0/1)
0.32
0.20
0.49
Regular transport interruptions, 2000–08 (0/1)
0.36
0.41
0.29
Frequent transport interruptions, 2000–08 (0/1)
0.20
0.24
0.15
NDVI x10, 2000
0.77
−0.06
1.92
EVI x10, 2000
0.84
0.15
1.81
78
34
44
Ncommunities
NIH-PA Author Manuscript
1
Reference category
Land use policy. Author manuscript; available in PMC 2015 January 01.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Table 3
Specification 1: Migrants only
Specification 2: Migrants & remittances
Specification 3: Gender & remittances
Predictors
Ln(area in annuals)
Ln(input expenses)
Ln(days hired labor)
Ln(production of annuals)
0.03
0.32
−0.04
Ln(area in annuals)
Ln(input expenses)
Ln(days hired labor)
Ln(production of annuals)
0.12
0.36
0.02
Ln(area in annuals)
Ln(input expenses)
Ln(days hired labor)
Ln(production of annuals)
Migration and remittances
Number of migrants
+
0.04
**
0.09
Number of male migrants
Land use policy. Author manuscript; available in PMC 2015 January 01.
**
0.12
+
0.11
0.11
0.10
***
−0.08
−0.03
−0.05
***
0.35
0.10
Number of female migrants
0.08
Ln(value of remittances)
***
−0.08
***
0.35
−0.04
−0.03
−0.05
−0.04
*
0.50
Gray and Bilsborrow
Household-level tobit models of the effects of migration and remittances on agricultural outcomes.
−0.04
Control variables
Ln(area of land owned)
0.19
***
Owned flat land
0.21
Owned land with black soil
−0.08
Owned land with irrigation
−0.45
Ln(number of cattle)
+
1.28
−0.51
***
1.37
0.18
0.95
0.02
0.22
−0.73
−0.07
0.28
***
0.04
−0.60
0.15
−0.15
0.27
0.02
0.18
−0.49
**
−0.20
Owned land with perennials
0.36
**
1.55
0.41
−2.42
***
−0.20
−0.43
**
***
0.04
***
1.36
0.17
0.18
+
0.94
0.04
0.22
−0.50
−0.48
−0.73
−0.07
1.31
0.27
**
1.55
0.41
***
0.08
−0.57
0.17
−0.14
0.28
−2 38
**
−0.20
***
−0.43
0.03
+
1.31
−0.50
0.27
***
1 39
0.94
0.03
−0.41
−0.74
**
1 57
−0.58
−0.15
0.28
0.06
0.01
0.09
−0.19
0.06
0.02
0.09
−0.20
0.06
0.02
0.09
−0.19
−0.11
−0.50
0.48
−0.71**
−0.07
−0.41
0.51
−0.65*
−0.07
−0.41
0.51
Age of head
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Education of head
−0.02
0.00
−0.02
0.01
−0.02
Nonagricultural work by head
−0.03
−0.01
−0.07
−0.07
−0.07
Nonagricultural wealth score
0.01
0.06
0.01
0.06
0.01
Paved road
0.19
+
*
1.64
*
+
0.36
**
−1.45
Gravel road
0.07
−0.72
Tropical climate index
0.00
0.44*
Constant
0.31
−0.19
0.46***
2 53***
Sigma
Canton fixed effects (joint F)
10.9
***
***
19 3
**
0.27
−0.44
**
0.41
0.08
**
2.02
0.14
−7.24
***
2.38***
7.0
***
0.32
*
0.81
−0.27+
4.83
***
2.02***
***
3.1
0.17
+
*
−0.04
0.08
*
1.55
*
0.36
**
−1.48
0.06
−0.74
−0.02
0.40 +
0.38
+
0.45***
***
11.1
−0.06
2.52***
12.0
***
**
0.27
−0.48
0.41
**
0.09
2.03
**
0.13
***
0.31
*
0.80
−0.30+
***
−7.20
4.93
2.37***
2.01***
7.2
***
***
32
+
0.17
−0.04
*
0.08
*
1.55
*
0.36
**
−1.48
0.06
−0.74
−0.02
0.40 +
0.38
+
0.45***
11.4
***
***
0.08
Female head
0.08
0.40
0.17
Household size
−0.04
0.17
−0.06
2.52***
11.8
***
0.01
***
0.27
−0.49
0.40
**
0.13
1.99
**
0.14
−2 38
−0.65
0.01
−0.06
0.07
0.30
*
0.82
−0.30
−7 25
***
4 94
2.37***
2.01
***
7.3
*
0.00
+
***
***
3.5
***
1
Linear instrumental variable model
0.06
−0.47
0.06
−0.11
**
0.22
0.16
0.18
0.42
Page 18
Number of migrants
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Specification 1: Migrants only
Specification 2: Migrants & remittances
Specification 3: Gender & remittances
Predictors
Ln(area in annuals)
Ln(input expenses)
Ln(days hired labor)
Ln(production of annuals)
Ln(area in annuals)
Ln(input expenses)
Ln(days hired labor)
Ln(production of annuals)
0.22
Number of female migrants
1
Ln(production of annuals)
**
0.13
0.18
0.32
+
0.40
0.13
438
437
439
438
1.13
+
−0.12***
−0.44 +
−0.08
−0.38
−0.12***
−0.45 +
−0.08
−0.43
438
437
439
438
438
437
439
438
Land use policy. Author manuscript; available in PMC 2015 January 01.
Selected coefficients from a parallel set of linear instrumental variable models that correct for the potential endogeneity of migration and remittances.
+
*
Ln(days hired labor)
0.22
Ln(value of remittances)
Nhouseholds
Ln(input expenses)
Gray and Bilsborrow
Number of male migrants
Ln(area in annuals)
p<0.10
p<0.05
**
p<0.01
***
p<0.001
Page 19
Gray and Bilsborrow
Page 20
Table 4
Community-level OLS models of the effects of migration on vegetation greenness.
NIH-PA Author Manuscript
Predictor
Ratio of migrants/nonmigrants
Mean annual precipitation
0.35 *
0.63 *
0.00
0.01
0.28 ***
0.40 ***
Experienced drought
0.22 **
0.25 +
−0.22 ***
−0.34 ***
SD of monthly temperatures
−0.11
−0.05
Mean land slope
0.04 *
0.02
SD of land slope
−0.12 **
−0.06
Mean farm size
−0.03 **
−0.04 **
Maize cultivated
0.21 *
0.41 **
Coffee cultivated
NIH-PA Author Manuscript
0.10
0.11
Potatoes cultivated
−0.40 ***
−0.61 **
Peanuts cultivated
0.03
0.31
Ln(population size)
−0.01
−0.06
Mean education
−0.08 *
−0.13 *
Occasional transport interruption
0.26 *
0.32 *
Regular transport interruption
0.45 *
0.56 *
Frequent transport interruption
0.54 **
0.79 *
−0.35 ***
-
-
0.11
Canton fixed effects (joint F)
24.1 ***
20.9 ***
Constant
5.81 ***
6.56 ***
78
78
NDVI x10
EVI x10
Ncommunities
NIH-PA Author Manuscript
*
Change in EVI
CV of monthly precipitations
Mean annual temperature
+
Change in
NDVI
p<0.10
p<0.05
**
p<0.01
***
p<0.001
Land use policy. Author manuscript; available in PMC 2015 January 01.