University of Southern California Dissertations and Theses

Buscando la prosperidad: Migration as long -term investment in El Salvador

(USC Thesis Other)

Buscando la prosperidad: Migration as long -term investment in El Salvador

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content BUSCANDO LA PROSPERIDAD:
MIGRATION AS LONG-TERM INVESTMENT IN EL SALVADOR
by
Paul Alexander Rivera
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
December 2002
Copyright 2002 Paul Alexander Rivera
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3093812
UMI
UMI Microform 3093812
Copyright 2003 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90089-1695
This dissertation, written by
3W1
under the direction o f h dissertation committee, and
approved by all its members, has been presented to and
accepted by the Director o f Graduate and Professional
Programs, in partial fulfillment o f the requirements fo r the
degree o f
DOCTOR OF PHILOSOPHY
Director
Date
Dissertation Committee
Chair
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Dedication
To Daniela and Colleen
who make it all worthwhile
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgements
In the process of producing this thesis, I have accumulated a debt of gratitude which I
am unsure of being able to repay at all, much less in a timely fashion. With that in mind,
the following should not be thought of as an attempt to begin repaying that debt, but rather,
as a preliminary accounting of the sometimes heroic efforts made by the truly outstanding
people who have made this work possible and with whom I am honored to be associated.
First and foremost, I thank my parents, Paul and Lourdes Rivera, for their unflagging
love and support, for never asking me just how long I expected to be in school, for their
unwavering confidence in me.
This work was carried out under the scrutiny and encouragement of Prof. Jeffrey Nu
gent, Prof. Richard Easterlin and Prof. Nora Hamilton. Financial support was provided by
the University of Southern California Department of Economics. For sparking my interest
in migration and remittances, I wish to acknowledge Prof. Alejandra Cox-Edwards who,
in addition, supplied critical doses of enthusiasm at pivotal moments. The Rural House
hold Surveys of El Salvador utilized in the analyses were provided by the BASIS Research
Program at The Ohio State University. All data manipulation and analysis was performed
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
with Stata 7.0. The present document was generated using the KFgX Document Prepara
tion System.
Kevin Frand and Aimee Plourde supplied caffeine, refuge, supervision and technologi
cal facilities which made possible a great deal of forward progress on this project. Um forte
abrago para Vitoria Saddi pelo apoio e pelo animo que sempre me tern brindado. Hearty
thanks to Mehdi Farsi without whom I remain convinced that I would never have survived
my first year of doctoral studies. Thanks also to Dr. Patti Delaney who has always provided
relevant and useful aspects of the cultural and applied perspective.
Most of all, I wish to thank Daniela and Colleen. Dani, for the unconditional love and
for being a constant reminder that there is more to life than work. Colleen, for sticking
with me, not quitting and not letting me quit. I love you both!
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Contents
Dedication ii
Acknowledgements iii
List Of Tables vii
List Of Figures ix
Abstract x
1 Introduction 1
2 A Brief History of El Salvador 9
2.1 Economic H is to ry ............................................................................................... 12
2.1.1 O u tp u t...................................................................................................... 12
2.1.2 Saving, Investment and C red it.............................................................. 14
2.1.3 Prices, Interest and Exchange R a te s .................................................... 16
2.1.4 L a b o r ...................................................................................................... 18
2.2 Demography and Social H isto ry ......................................................................... 19
2.3 Migration and Remittances .............................................................................. 22
2.3.1 P a tte r n s .................................................................................................. 22
2.3.2 The Typical Migration Experience....................................................... 23
2.3.3 Remittance Channels, Frequency, and A m ounts................................ 26
2.3.4 Street Gangs in El S alv a d o r................................................................. 28
3 Previous Research on Migration and Remittances 31
3.1 Main P erspectives............................................................................................... 32
3.2 Other V ie w s ......................................................................................................... 37
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.3 Research on El S alvador.................................................................................... 39
3.4 Dependency and Dutch D is e a s e ...................................................................... 46
4 Models of Migration-as-Investment 50
4.1 Static M o d e l........................................................................................................ 54
4.1.1 Static Model without Separability....................................................... 58
4.2 Intertemporal Investment M o d e l...................................................................... 62
5 Investment in El Salvador 73
5.1 Data and M ethodology....................................................................................... 76
5.1.1 Investment O p tio n s ............................................................................... 85
5.1.2 Methodology ......................................................................................... 93
5.1.3 Explanatory V ariables............................................................................ 94
5.2 Investment: R e s u lts ............................................................................................. 104
5.2.1 D iscussion...................................................................................................107
5.3 Financing M igration..............................................................................................I ll
5.3.1 Data and Methodology ........................................................................... 113
5.3.2 R esults..........................................................................................................116
5.3.3 D iscussion...................................................................................................118
5.4 R em ittan ces...........................................................................................................120
6 Conclusions 126
Reference List 129
Appendix A
Logit Regressions: Investment C ategories...................................................................133
Appendix B
Logit Regressions: Human Capital Investment .........................................................145
Appendix C
Logit Regressions: Migrant F in an ce.............................................................................152
Appendix D
Logit Regressions: R em ittances....................................................................................158
v i
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix E
Bivariate Probit E stim ations...........................................................................................163
v ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List Of Tables
2.1 Gini coefficients, selected Latin American countries...................................... 11
5.1 Age Distribution by Sex, 1998 and 2000.......................................................... 78
5.2 Mean Years of Education by Sex and Age Category, 1998 and 2000. . . . 80
5.3 Income Categories by Sub-groups, Percent of Total....................................... 82
5.4 Income Categories by Income Quintiles, Percent of Quintile........................ 83
5.5 Summary Characteristics by Income Quintiles................................................ 84
5.6 Net Migration by Destination, 1998 and 2000................................................. 87
5.7 Summary Statistics, Rural Household Surveys................................................ 99
5.8 Predicted probabilities, Investment Categories................................................... 105
5.9 Predicted probabilities, Migrant Finance............................................................. 117
5.10 Predicted Probabilities, Remittances.................................................................... 123
A .l Logit Regression Results, Investment Categories...............................................134
B.l Logit Regression Results, Human Capital Investment.......................................146
C.l Logit Regression Results, Sources of Migrant Finance..................................... 153
D .l Logit Regression Results, Remittances................................................................ 159
E .l Bivariate Probit Regression Results, Investment Categories............................ 165
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List Of Figures
1.1 Remittances (% GDP), Selected Countries (WDI 2001)................................. 8
2.1 Absolute and per capita GDP growth (WDI 2001)........................................... 13
2.2 Agriculture and Services, Percent of GDP (WDI 2001).................................. 14
2.3 Gross Domestic and Gross National Saving, Current US$ (WDI 2001). . . 15
2.4 Total and Rural Poplulation Growth Rates (WDI 2001).................................. 20
2.5 Rural Poplulation, Percent of Total (WDI 2001)............................................... 21
2.6 Remittances, Percent of GDP (WDI 2001)........................................................ 28
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
In the present day when international migration and remittance flows are nearly univer
sal phenomena, detailed analyses of the associated motivations and effects are essential.
In many countries, the volume of out-migration has altered the social and cultural setting
on a national level, and the sum of remittance income may total over 10% of a country’s
gross domestic product. However, the incentives to migrate or send remittances have tra
ditionally been studied in isolation from each other and the competing alternatives faced
by economic actors.
This thesis seeks to bring both depth and breadth to migration and remittance analysis
by framing the migration decision as a form of investment in an intertemporal setting where
remittances represent the return. The household reaches an optimal decision by maximiz
ing the expected net gain from the potentially multiple investment options. The results
of the theoretical model are found to depend critically on the demographic and economic
characteristics of the household and the individuals that comprise it.
In its empirical application, the migration as investment approach translates to a com
parative analysis of participation in various forms of investment and the effects of various
demographic and economic variables. The data is further decomposed and the analyses
x
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
repeated for the various forms of migration finance, as well as household-level indicators
for remittances.
Data analyses are based on the Rural Household Surveys of El Salvador for 1998 and
2000. As a result of the large and persistent out-migration from El Salvador and the mas
sive volume of remittances, now totalling over $1 billion annually, the Salvadoran experi
ence provides an excellent case study for the migration as investment approach.
The results prove to support the migration as investment hypothesis for specific sub
groups of the sample population, particularly those households which borrowed to help
send young male migrants. However, on a broader scale, the traditional model of indi
vidualistic migration remains a valid interpretation of the available data. Perhaps more
importantly, the empirical results are suggestive of a pattern of migration and remittances
aimed at sponsoring further migration from the household.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 1
Introduction
Migration and remittances have become nearly universal phenomena, and careful analysis
of their effects is of crucial importance. As many as 10% of the citizens of some countries
reside outside the country of origin, and the volume of money they remit home regularly
equals 5-10% of gross domestic product. Such massive flows of people and resources have
dramatic and fundamental effects on the societies and economies of the countries involved,
resulting in changing gender roles, the sectoral reallocation of productive resources and
skewed exchange rates, among myriad other effects. Research on different aspects of these
phenomena, however, has led to paradoxical implications.
While there is little doubt that migration benefits the individual and that remittances
provide a tremendous boost to the receiving family, much is made of the negative effects
of migration and remittances. Brain drains, delays in the absorption of improved technol
ogy and the decay of traditional institutions and culture are blamed on migration. Similarly,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
research on remittances has generated a set of stylized facts that accuse remittances of gen
erating dependency on foreign income, fostering wasteful spending, deepening social in
equality, and stimulating a form of Dutch Disease. Indeed, taken together, the migration
and remittance phenomenon would appear to be a form of externality where the benefits
are localized and the true social costs are widely dispersed and not accounted for by in
dividual decision makers. The crux of the paradox, however, is that the counterfactual is
not explored. In the face of adverse socio-economic conditions such as civil war, falling
real wages or depressed export prices, it is difficult to believe that the absence of migration
could have generated improved conditions for a significant proportion of the population.
This signals an excessive narrowness in most existing research which fails to go beyond
the incentives to migrate or remit and ignores the disincentives to pursue other options.
The primary goal of this thesis is thus to bring a degree of cohesion to the study of
migration and remittances by incorporating three fundamental factors. First is the explicit
recognition of the bi-directional causal link between migration and remittances. While it
is clear that migration is the necessary first step in generating remittances, it is further hy
pothesized that the remittance incentive in turn motivates the migration decision. Second,
individual migration is framed as a household-level choice. Much of the remittance liter
ature views social and familial ties as critical incentives to remit and maintain remittance
behavior over time, but such factors have not been used to account for the initial migration
decision. A counterpoint to standard analyses is thus provided by instead assuming that the
household is the basic decision making unit. Third, households are postulated as forward
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
thinking such that migration is undertaken as a form of investment, and remittance income
represents the return on investment. Thus, the household’s decision process involves not
only a time dimension, but also a rational choice between competing investment alterna
tives. This aspect of the analysis is both crucial and innovative. The recognition of migra
tion as a form of investment is itself not new but has previously only been peripherally or
anecdotally addressed and not integrated in to the analytical framework. The overall result
of this thesis is a unified concept of migration and remittances that consistently explains
individual and household level behavior as well as the greater stylized facts of migration
and remittances.
The experience of El Salvador over the last thirty years motivates my particular interest
in migration and remittance research and is the focus of the empirical aspects of this the
sis. Migration and remittances have been at the core of nearly all socio-economic changes
in El Salvador since 1979 at the outbreak of civil war. It is estimated that 10-15% of all
native-born Salvadorans today live outside the country, and the volume of remittances to
El Salvador surpasses even total income from the country’s main export, coffee.
At the household level, a seminal study by Segundo Montes finds that remittance in
come may boost income by as much as 50%; however, rather than providing a source of
savings or investment capital, remittance income has typically supplemented current con
sumption. Instead of providing an engine for economic growth, remittances have served
as a safety net for families otherwise experiencing reductions in real wages and limited
job opportunities, especially during the war years. Further, the apparent success rate of
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
migrants as judged by the low rate of return migration and the ever-growing sum of remit
tances has only encouraged a persistent outflow despite the end of armed conflict in 1991.
The volume of Salvadoran out-migration has been enough to alter such fundamental as
pects of traditional society as the roles of women and the basic structure of economic activ
ity. Indeed, El Salvador has often been used as the case-in-point for the negative repercus
sions of migration and remittances as ongoing socio-economic changes are readily linked
to the very visible flows of people and money.
Within El Salvador, migration and remittances have garnered a great deal of atten
tion from policymakers, academia and the popular press. In its early stages, migration
was thought to ease the pressures of unemployment, and remittances were welcomed as
a source of investment capital and foreign exchange. However, as remittances ballooned
to significant proportions of gross domestic product and were clearly being used almost
entirely for consumption, the resulting growth of imports and the service sector at the ex
pense of agriculture seemed to sour the official stance on remittance income.
The academic community only fueled the burgeoning negativity with evidence of the
Dutch Disease as the massive influx of foreign exchange overvalues local currency to the
detriment of the export sector (agriculture, in this case) and promotes domestic inflation
as the relative prices of non-traded goods rise (BCR 1998, Boyce 1996, Melhado 1997,
Rivera Campos 1996).
Similarly, the Salvadoran media has focused on the hypothesized social effects of mi
gration and remittances. Stories regularly appear exposing the increased delinquency, gang
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
activity and drug use stemming from dependency on remittance income and the reduced
incentive to work.
But these stories demonstrate the exception and not the rule. Despite remittance in
come, nearly 30% of urban Salvadorans live in poverty, as do 60% of rural residents. Mi
gration and remittances have not provided the people of El Salvador with the means to
enjoy extravagant lifestyles or substantially increased amounts of leisure, but rather, with
a basic and accessible method to achieve some improvement in well-being over time. In
deed, the proper perspective is not to question the effects of migration and remittances, but
rather, to take a broad view of the options available to Salvadorans and examine the role
of migration within that portfolio.
The empirical aspect of this thesis attempts to carry out a more dynamic analysis in
which migration is a form of investment by taking advantage of a panel of over 4,000 in
dividuals comprising nearly 600 households from the Rural Household Survey of El Sal
vador between 1998 and 2000. The availability of such an extensive and thorough panel
dataset makes it possible to track household choices and changes over time and relate them
to the characteristics of the household. However, the analytical process utilized goes be
yond simply addressing migration and instead takes a broad view of investment as a choice
among several options, including migration. In this way, it is possible to discern not only
the characteristics of households investing in migration, but also the general profile of house
holds choosing other forms of investment or no investment at all. The Salvadoran data,
however, suggest a highly skewed investment portfolio in that migration as investment
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vastly surpasses any other form of investment, and the more intensive focus on migration
is therefore not unwarranted.
A decomposition of the methods of migrant finance gives an indication of both the de
gree of sacrifice households are willing to make in order to invest in migration and the
genre of financing options available. The results make clear that the relative ease of en
try into migration as investment is nearly as important as the potential remittances. In a
poor and highly credit constrained society, households are not always able to finance ma
jor investment projects entirely out of pocket but are even less often able to borrow; thus,
compared to investment in land where access to credit is crucial, migration as investment
offers considerably more flexibility in the degree of access and cost of entry. Further, as
measured by remittances, the rate of return on migration outpaces that of most other invest
ment options. Montes finds that most households recover in remittances the explicit cost
of their investment within the first 18 months and reap significant net benefits thereafter.
A particularly robust finding permeates the empirical analysis: the driving force of the
investment decision is the quality of the potential migrant. Households with working age
men are more likely to invest, particularly in migration, and the data makes clear the rea
sons. While the remittance rates of male migrants are not found to be significantly different
from those of females, working age male migrants are the most likely to have received no
financing from the sending household, implying lower explicit cost to the household and
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
thus a higher net rate of return. These findings hold true even when various aspects of ab
solute and relative household wealth, geography and household composition are accounted
for.
The strength of the empirical findings holds important implications for migration and
remittance-related policies in El Salvador. Migration will continue as long as the core of
the nation’s workforce remains constrained by the paucity and substantial roadblocks as
sociated with domestic investment. Exchange rate management and attempts to harness
remittance income will not solve the problems; Salvadoran families must instead have the
incentive to stay and focus their energy and resources domestically. Effectively, the issue is
then one of institutional reform, the creation of reliable infrastructure and the enactment of
sound economic policies for long term growth, both social and economic. While an analy
sis of whether the political will exists to embark upon that road to progress in El Salvador
is beyond the scope of this thesis, the results herein do suggest that less pro-active courses
of action will only encourage cascading migration and a remittance-driven economy.
The remainder of this thesis proceeds as follows. Chapter 2 provides a brief history
of the relevant social and economic facts of the Salvadoran experience, with particular at
tention to the latter half of the twentieth century and the patterns of migration and remit
tances. Chapter 3 summarizes the existing body of work on migration and remittances,
including micro- and macro-economic perspectives and work related to the particular case
of El Salvador. Chapter 4 proposes two formal models of migration and remittances. A
static model with exogenous remittances is elaborated first to show the approach implicitly
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
El Salvador
Jordan
Mexico
Morocco
Portugal
25
20
15
10
5
0
1970 1975 1980 1985 1990 1995 2000
Years
Figure 1.1: Remittances (% GDP), Selected Countries (WDI 2001).
taken in many existing studies and the inherent flaws in deducing causality in such rela
tions. A dynamic model is then developed where agricultural households choose among a
menu of investment options, and the conditions are derived under which a household will
optimally choose to invest in one, multiple or none of the available options. Each option,
including migration, is defined by a cost of entry and an expected future return, and it is
clear that both the costs and benefits of the investment alternatives are affected by the un
derlying characteristics of the household. The empirical analysis of Chapter 5 thus seeks
to elucidate the variations in household characteristics that underlie and motivate various
investment decisions. Chapter 6 offers a conclusion and avenues for further study.
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 2
A Brief History of El Salvador
El Salvador is a tiny country on the Pacific side of Central America characterised by vol
canic mountains, lush jungle and coffee fields (reference to map). Approximately the size
of Massachusetts in area, El Salvador is by far the smallest country in the region, and its
6.4 million inhabitants1 are relatively tightly packed at 304 persons per square kilometer.
Despite important reductions in the birth rate (show a table), the population continues to
grow at 2% per year. The economic history of El Salvador is characteristic of a small agri
cultural export economy. Heavy reliance on a small number of cash crops - coffee and, to
a lesser extent, cotton and sugar - generated a largely landless rural society with govern
mental and economic power concentrated in the hands of a disproportionately small group.
El Salvador faces all the problems of a developing country. Widespread poverty re
mains the greatest concern: as late as 1995, 34% of all Salvadorans and 55% of all rural
residents lived below a monthly $60 per capita poverty line, and improvements have come
slowly and irregularly. Despite increased literacy and primary school enrollment rates in
1For the sake of comparison, there are roughly 13 million people living in the city of Los Angeles.
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the last several years, absolute and relative educational attainment remain low, and the
country consequently suffers from a lack of skilled labor. Further, the excess of unskilled
labor has gradually polarized the society such that El Salvador ranks among the countries
with the highest levels of inequality (Table 2). Basic infrastructure such as roads and broad
access to electricity and potable water remains poor. Government corruption has histori
cally been rampant, and El Salvador still ranks among the countries with the highest levels
of corruption in business2. Indeed, taken together, it is no surprise that the civil war war
which began in 1979 was rooted in popular discontent over socio-economic stagnation and
the poor prospects for future improvement.
As a basis for subsequent analyses, the remainder of this chapter outlines the patterns
of the relevant Salvadoran economic and social trends. This includes tracking changes in
the structure of the economy, patterns of saving and investment, changes in price levels,
and labor market conditions. The make-up of the society has also evolved, necessitating
an account of the changing demographics, urbanization and the causes and effects of over
a decade of armed conflict. Finally, the particular emphasis of this thesis makes relevant a
discussion of the history of migration and remittances in El Salvador.
2 An index published by Transparency International (2001) computes a 0-10 corruption score where a
higher score implies a ‘cleaner’ country. El Salvador scored 3.6 as compared to a world average of 4.8 and
ranked better than Nicaragua and Honduras, but worse than Mexico, Colombia, Brazil and Peru.
1 0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Year Brazil Colombia Costa Rica El Sal. Honduras Mexico Peru
1970 55.8 52.2
45.5 59.4
1971 49.7 42.7
51.3
1972 63.5 51.4 48.7
58.5
1973
1974 48.4 49.0 45.2
1975 34.7 56.0
56.8
1976 53.0 50.8 45.0
1977 47.3 48.8 44.2 48.8
1978 50.8 52.2
1979 53.0 55.0 45.0
1980 53.9 48.3 43.1
1981 51.8 45.5 53.2
1982 48.8 56.0 41.1
1983 52.4 43.4 45.9
1984 58.0 44.5 45.6
1985 56.4 44.7 41.8
1986 58.2 44.0 55.0 43.0 37.2
1987 55.7
1988 56.8 47.7
1989 59.9 47.0 45.1 55.0 49.8
1990 60.6 44.8 57.0 43.8
1991 63.7 49.3 47.1 50.0 41.9
1992 56.1 49.2 53.8 51.5
1993 58.9 57.0 47.5 54.0
1994 53.7 50.3 51.9 43.4
1995 57.9 57.2 47.7 47.5 53.7
1996 58.1 57.1 47.0 52.3 53.4 52.0 46.2
Mean 54.7 51.3 45.3 47.8 54.0 49.9 48.3
Table 2.1: Gini coefficients, selected Latin American countries.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.1 Economic History
2.1.1 Output
Since I960, Salvadoran output growth has averaged 3% while growth in per capita output
has averaged 0.8% (see Figure 2.1). However, excluding the war period, absolute and per
capita growth average 5.3% and 2.5%, respectively. The persistent gap between growth
and per capita growth nonetheless indicates relatively high population growth rates (com
bine Tables 1 and 24). Growth spikes in the early 1960s were due to significant efforts at
import substitution; however, subsequently reduced growth rates throughout the 1970s re
flect the exhaustion of industry-driven growth. Output plummeted with the onset of war
in 1979. The massive impact on the well-being of the populace cannot be overstated; real
GDP per capita in 1982 was roughly equal to that in 1961. In the late 1980s, positive
growth resumed, driven primarily by foreign aid and remittances from abroad (Figure 2.6).
Pre-war growth rates have resumed in the post-war period due to reduced uncertainty, mod
erate institutional reform, and the persistence of remittance income. There is no doubt,
however, that the structure of the post-war Salvadoran economy differs widely from the
traditional pattern.
Before 1960, industrial output comprised only 10% of Salvadoran GDP, and an in
crease to 20-25% in the early 1960s was a direct result of import substitution policies.
Further, due to pro-industry credit policies and protectionist tariffs, early growth in the in
dustrial sector came at the expense of agriculture. The beginning of the armed conflict,
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
GDP Growth
Per Capita GDP Growth
-10
-15
1960 1965 1970 1975 1980 1985 1990 1995 2000
Years
Figure 2.1: Absolute and per capita GDP growth (W D I2001).
however, clearly marks a change in the structure of the Salvadoran economy (Figure 2.2).
The rapidly growing importance of the service sector at the direct expense of agriculture
signaled the end of the traditional agro-export economy. The war, which was carried out
primarily in rural areas, only hastened the decline of the agricultural sector which had been
underway for over two decades, while the rise of the service sector is attributed to the mas
sive flow of remittances from abroad. Indeed, the growth of the service sector at the ex
pense of agriculture in El Salvador is typically seen as a form of Dutch Disease generated
by remittance income3. Current agricultural output in El Salvador primarily consists of
food stuffs rather than cash crops.
3For a discussion of Dutch Disease, see Section 3.4.
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
Agriculture
Services
60
50
40
30
20
10
0
1960 1965 1970 1975 1980 1985 1990 1995 2000
Years
Figure 2.2: Agriculture and Services, Percent of GDP (WDI 2001).
2.1.2 Saving, Investment and Credit
The data presented in Figure 2.3 track gross domestic saving (GDS) since 1960. Despite
the sparse coverage of banking institutions, gross domestic saving averaged 15% in the
pre-war era with a slightly positive trend. The beginning of the war in 1979 is clearly
marked as a sharp decline in saving as Salvadorans attempt to maintain their living stan
dards despite real wage declines and rapidly rising domestic price levels (Table 8).
The onset of armed conflict, however, also initiates a persistent divergence between
gross domestic saving and gross national saving (GNS).4 By the mid-1980s, GNS rose to
levels equal to those in the pre-war era, while GDS fell to and today remains below 5% of
4A 11 available evidence suggests that, traditionally, saving within El Salvador (GDS) and saving by Sal
vadorans (GNS) were virtually indistinguishable.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1800
Gross Domestic Savings
Gross National Savings
1600
1400
1200
« 1000
&
a
o
§ 800
600
400
200
1960 1965 1970 1980 1975 1985 1990 1995 2000
Years
Figure 2.3: Gross Domestic and Gross National Saving, Current US$ (WDI 2001).
GDP. No other piece of existing literature takes note of this difference. Most previous work
instead focuses on the steep decline in GDS which is attributed to a reduced propensity to
save as a result of growing dependency on remittance income. The evidence in Figure 2.3,
however, suggests a different story: due to the bank nationalization of 1980 and the eco
nomic uncertainty generated by war, Salvadorans sought extra-national savings options5.
The banking sector in El Salvador has been traditionally typified by low profitability,
bad debt, little credit extension to small businesses and almost no presence in rural areas.
The Salvadoran credit and financial sectors have thus far served almost exclusively to pro
mote the interests of major agro-export and, later, industrial concerns. Further, extensive
5This simple observation provides great support for the migration-as-investment approach presented in
Section 4.2.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
corruption meant that many loans were given with no real expectation of repayment; thus,
many direct subsidies were given in the form of loans. As a result, the availability of credit
for small or medium enterprises was extremely scarce, often limited to loans from family
members or local money lenders.
Similarly, banking facilities historically have been unavailable to most Salvadorans
due to lack of coverage, especially in rural areas. Salvadoran savings thus tended to be
held in the form of cash, livestock, real estate or productive assets. Only in the last ten
years has it been the case that over half the Salvadoran adult population holds some form
of bank account.
In 1980, Salvadoran commercial banks were nationalized with the intention of gener
ating government revenues and promoting coffee and cotton exports. This policy had two
major effects. First, credit to small businesses and rural areas was further choked off as
credit was channeled almost exclusively toward the public sector and large agro-export.
Second, nationalization of the banking system stimulated capital outflow as domestic in
vestment options were reduced dramatically; this accounts for an important part of the fall
in GDS observed in 1979 and 1980.
2.1.3 Prices, Interest and Exchange Rates
Over the last forty years, variability has characterized the price level in El Salvador. Through
the end of the 1960s, inflation was nearly non-existent and only began to rise markedly in
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the 1970s as infrastructure development projects and protectionist policies spurred domes
tic prices. The supply shock generated by war maintained inflation levels over 10% during
the early 1980s. Beginning in 1985, rising remittance income and the consequent increase
in demand for non-traded goods caused inflation rates to rise to 15-30% annually during
the latter half of the 1980s6. In the post-war period, however, inflation rates have fallen be
low 5% as protectionist policies have been largely abandoned and moderate progress has
been made to advance the efficiency of Salvadoran production.
Exchange rates have occupied a crucial role in the Salvadoran economic picture. Since
the early 1960s, El Salvador has operated under a managed floating exchange rate regime
with periodic devaluations. The exchange rate with the U.S. dollar remained unchanged
at 2.5 colones/dollar until 1985; however, beginning in the mid-1970s, the black market
rate for exchange persistently diverged from the official rate which was not adjusted for
changes in trade policies.
However, El Salvador formally dollarized as of January 1, 2001 such that the U.S. dol
lar is now legal currency. This effectively eliminated the exchange rate between the colon
and the dollar7 and severely restricts the Salvadoran government’s ability to pursue mon
etary policy. The dollarization seems to have been a move to spur investment in El Sal
vador by reducing the exchange rate and inflation risks associated with dealing in colones.
To that end, dollarization appears to have been successful since lending rates have fallen
and foreign investment in El Salvador seems to have risen. The change was nonetheless
6See Section 3.4 on Dutch Disease.
rThe exchange rate was fixed at 8.75 colones/dollar.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
controversial since it was only approved by a simple majority of the legislative assembly,
rather than the constitutionally required 2/3 majority, and appeared to be a classic back
room deal that was only announced once the decision to dollarize was effectively made.
Further, the need to dollarize has been strongly questioned since El Salvador has experi
enced some of the lowest exchange rate volatility and inflation rates within the region over
the last five years. Despite some initial confusion and allegations of price gouging against
local merchants, over 60% of all transactions in El Salvador were conducted in dollars by
October, 2001.
2.1.4 Labor
The armed conflict which began in 1979 could be said to have had its roots in popular dis
content over wages, persistently widening inequality and few opportunities for individual
advancement. War, however, only worsened the situation. During the war years, political
instability generated contagion throughout the economy such that stagnation, if not regres
sion was evidenced in most major economic sectors. By 1983, unemployment had reached
25% - roughly equal to unemployment rates in the U.S. during the Great Depression - , and
real GDP had fallen by 25%. The paucity of employment (as well as the political situation)
provided great impetus for the massive emigration of Salvadorans.
Those who remained functioned in an environment of reduced labor demand, particu
larly in the traditional agricultural and industrial sectors. Further, short-term, non-contractual
and day labor became increasingly prevalent, and large numbers of workers poured into the
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
informal sector as employers shifted to much shorter planning horizons. Real wages also
experienced steady decline throughout the war period as unemployment climbed, inflation
eroded purchasing power and wage freezes were implemented as part of government fiscal
austerity measures.
The signing of the peace accords in 1991 brought both a modicum of political stabil
ity and a resurgence in labor demand. The structure of the Salvadoran economy in recent
years, however, has changed drastically from the traditional agricultural economy which
existed prior to the 1970s. The agricultural sector comprises a steadily declining propor
tion of total employment, while services and manufacturing have grown. Also, the lack of
physical capital, technological sophistication and technical expertise has become markedly
problematic in the new economy. This has generated a situation of high returns to skilled
labor and a very low marginal revenue product of unskilled labor. Thus, while still dealing
with chronic unemployment, urban and rural underemployment has also become an impor
tant problem. Further, Funkhouser’s (1999) evidence of a “brain drain” from El Salvador
implies that emigration has only exacerbated the lack of skilled labor.
2.2 Demography and Social History
Despite the myriad changes brought about by war and the moderate social progress of the
last decade, the Salvadoran population retains much of its traditional character: a large,
growing, and predominantly rural population constantly battling poverty and inequality.
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Pop. Growth Rate
Rural Pop. Growth Rate
3.5
2.5
0.5
- 0.5
1960 1965 1970 1975 1980 1990 1985 1995 2000
Years
Figure 2.4: Total and Rural Poplulation Growth Rates (WDI 2001).
Figure (2.4) tracks the growth rate of the population of El Salvador. El Salvador his
torically has been the most densely populated country in Central America with 304 per-
sons/sqkm. Population growth rates above 3% annually have been typical in Salvadoran
history, but these declined in the late 1960s as the economic focus shifted away from agri
culture to more urban industrial activities. As with most economic indicators, the begin
ning of armed conflict in 1979 had a pronounced effect on population growth. The sharp
decline in population growth from 1979 to 1986 is largely the result of killings by “death
squads” and massive out-migration. By 1989, ten percent of all native-born Salvadorans
were living abroad, most in the United States. The population growth rate has since stabi
lized at 2.2%, lower than in the pre-war era, but still high for the region.
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
80
60
40
20
0
1960 1965 1970 1975 1980 1985 1990 1995 2000
Years
Figure 2.5: Rural Poplulation, Percent of Total (WDI 2001).
More detailed trends in the demographic make-up of El Salvador are worth noting. Sal
vadoran data suggests an aging population such that population growth is primarily gen
erated by increased longevity, not the birth rate. Tracking the working age population by
gender as a proportion of the total population shows the slight dominance of women be
ginning in 1979. This divergence testifies to the fact that most migrants and victims of war
were men. Female labor force participation has nonetheless grown steadily from 17% of
the total labor force in 1960 to over 35% today. As the Salvadoran economy continues to
become increasingly service-oriented and less agricultural, women will continue to gain
importance as labor force participants8.
8Indeed, the area of greatest job creation in El Salvador over the last ten years has been in textile-related
maquila industries where women comprise an average of 75% of the workforce.
21
Rural Pop.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The Salvadoran populace also has become increasingly urban; however, it is key to
note that, despite persistent urbanization, 53% of the population today remains rural. Fig
ure 2.5 suggests an important transition period for El Salvador: while facing the problems
of major urban centers, the needs of rural areas remain.
2.3 Migration and Remittances
Since this work addresses in particular the effects of international migration and remittance
income, the relevant histories of these phenomena merit some discussion. This section pro
vides a brief overview of migration patterns from El Salvador, a discussion of the typical
migration experience, and the characteristics of the typical migrant. Also, the key aspects
of remittance income are outlined: remittance channels, frequency, amounts, and uses of
remittance income.
2.3.1 Patterns
Throughout most of its history, Salvadoran international migration has been confined to
other Central American countries, especially Nicaragua and Honduras. The pressures of a
large and rapidly growing population in a very small country pushed many Salvadorans
across the borders of more sparsely populated neighboring countries where they estab
lished themselves as squatters. Estimates suggest that, by 1969, 300,000 Salvadorans were
practicing subsistence farming in rural Honduras. This migratory pattern continued until
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1969 when war broke out between El Salvador and Honduras9 largely over the perceived
invasion of Salvadoran squatters into Honduran territory. The subsequent expulsion of the
Salvadoran squatters marks a change in the primary destination of Salvadoran migrants.
After World War II, the flow of migrants to extra-regional locations - in particular, the
United States and Mexico - accelerated, but, in the wake of the war with Honduras, the
numbers of north-bound migrants grew dramatically. Even with the escape valve which
migration to Honduras had provided, land distribution in El Salvador remained highly un
equal, and there was little hope of accomodating the returning migrants. Many thus came
to the United States where Salvadoran enclaves were established, primarily in Southern
California and Washington, D.C. This shift in the migration pattern from El Salvador to
the United States is also associated with a change in the demographic characteristics of
the migrants. While pre-1969 migrants to the United States tended to be highly skilled or
professionals, migration since the 1970s primarily has consisted of unskilled workers.
2.3.2 The Typical Migration Experience
By all accounts, most migrants from El Salvador experience some combination of emo
tional, physical and financial trauma in order to arrive in the United States. There is no
question that the difficulties of the journey are due almost entirely to the illicit nature of
such migration. Aside from the large expenditure typically required to migrate, it is not
uncommon for migrants to be robbed, beaten, raped or killed en route to the United States,
9This event was popularly known as the Soccer War since fighting erupted following a match between
the two countries.
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and the diligence of border agents in both El Salvador and the United States often neces
sitates numerous attempts before the migrant actually arrives at the desired destination. It
is thus widely acknowledged that the prevalence of male migrants, especially as first-time
migrants in a household, may be in part associated with a higher probability of successful
arrival10.
It is certainly the case that the method of transport differs according to the wealth of
the migrant’s household. Those with greater wealth and education often obtain legal entry
to the United States as tourists or students and subsequently allow the temporary visas to
lapse. These migrants typically arrive by airplane and pay $500-$ 1,200 for the round-trip
flight. Although most arrive with no intention of returning to El Salvador in the foresee
able future, all such migrants purchase round-trip tickets to avoid arousing the suspicion
of customs agents. It should be noted that even temporary non-immigrant entry visas to
the United States require a potentially lengthy application process and are often denied for
lack of sufficient proof that the applicant will leave the United States in accordance with
the terms of the student or visitor visa11.
Ironically, less affluent migrants typically pay a higher price to migrate to the United
States. Many Salvadorans arrive in the U.S. via the services of a Salvadoran coyote, or
smuggler, most of whom charge approximately $5,000 per person, almost all of which is
10See Section 5.3 for empirical evidence in support of this view.
11 Specifically, in denying temporary visas, consular officials typically cite Section 214(b) of the Immi
gration and Naturalization Act which states that applicants must demonstrate “sufficient ties” to the home
country to remove any suspicion that the individual may intend to violate the terms of the visa. The decision
is left to the discretion of the consular official and the definition of what constitutes sufficient ties is neither
standardized nor codified.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
paid up front. By all accounts, the migration experiences of those who use the coyotes are
by far the most dangerous and the least likely to succeed. The journey itself is risky since it
requires sneaking out of El Salvador, completely traversing Mexico and stealthily crossing
the border with the U.S. Many migrants report having spent several days in the bed of a
truck, crammed against other people and under fruit or bales of hay, while others cover
most of the distance traveling by night on foot. En route, migrants may encounter piracy,
often at the hand of the coyote, and are left without money or their meager possessions.
The Sonoran desert in northern Mexico, the area where most illegal border crossings into
the U.S. occur, also proves to be a major obstacle where many hopeful migrants have met
their ends attempting to cross the vast sandy expanse of the desert. The services of the
coyote are clearly not guaranteed, and, having made the difficult journey at tremendous
expense, many migrants are thwarted by the efforts of the U.S. Border Patrol and deported
back to El Salvador12.
Regardless of the method of transport, Salvadoran migrants typically find themselves
employed in low skill jobs since their illegal status generally negates the possibility of ob
taining most high skill, or at least well paid, employment. It is thus not uncommon to find
Salvadoran migrants, educated and trained in a profession, performing tasks vastly incon
sistent with their level of training. Nonetheless, the growing Salvadoran communities in
the United States - particularly in Los Angeles and Washington, D.C. - have begun to co
alesce into social networks that help newcomers obtain jobs, housing and medical care.
12Although the probability of being caught by border agents is non-trivial, the enormous and growing
number of illegal immigrants to the U.S. suggests that the threat posed by the border agents is not an effective
deterrent.
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
While most of these networks function informally and are often part of the migrant’s ex
tended family, hometown associations and immigrant help centers have emerged which
attempt to establish the Salvadoran presence in a recognized and formal manner.
Certainly, legal immigration to the United States from El Salvador does occur, how
ever, the quota system for visa approval utilized by the U.S. Department of State allocates
to El Salvador only a fraction of the total number of visas desired. Over the last ten years,
the United States has issued an average of approximately 5,000 immigrant visas to Sal
vadorans each year, including visas for the immediate family members of U.S. citizens
which are not subject to the visa lottery; the amount is clearly minuscule as compared to
the over 600,000 Salvadorans currently residing in the United States. Also, the cost of le
gal immigration to the United States is not trivial, currently adding to $325 per applicant
(regardless of age) plus any costs incurred to obtain the required documentary evidence.
2.3.3 Remittance Channels, Frequency, and Amounts
As soon as their income flow allows, most Salvadoran migrants begin sending remittances
to their families in El Salvador. Evidence from the Encuesta de Hogares (national house
hold survey) and the Rural Household Surveys of El Salvador suggest that the standard
remittance amount is $100 per month, and that the regularity of remittance flows to El
Salvador do not reflect appreciable changes in response to macroeconomic conditions or
changes in the price levels, interest rates or exchange rates between the two countries.
2 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The vast majority of this massive flow of money - expected to reach nearly $2 billion in
2002 - does not enter the formal banking system. At present, over 85% of all remittances
are estimated to be sent via private couriers; indeed, the large out-migration of Salvadorans
has spawned a substantial private courier industry shuttling goods and money between the
U.S. and El Salvador. More formal remittance channels do exist. The Banco Agricola
de El Salvador does have a branch in Los Angeles where individuals can deposit money
to be retrieved by family members in El Salvador. The Salvadoran government has long
encouraged such methods since inclusion in the formal banking system would not only
facilitate remittance tracking, but could also help stimulate investment in El Salvador by
boosting the level of deposits available for loan. Although formal remittance methods have
gained popularity, access and coordination issues have hampered widespread use.
As Salvadoran hometown associations have grown, particularly in Los Angeles, some
efforts have been made to promote community remittances rather than family remittances.
These funds serve to help provide Salvadoran communities public goods and assistance
that are not only not provided as part of the Salvadoran infrastructure, but also unlikely to
be financed by the community itself. These often include electricity and water resources,
school houses and medical supplies. Although these organizations have met with success,
obtaining funds is a constant challenge since many migrants are not easily swayed to re
channel their remittance allocation away from family members and toward public goods.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.12
Remittances .....■ *....
0.1
0.08
0.06
0.04
0.02
1985 1975 1980 1990 1995 2000
Years
Figure 2.6: Remittances, Percent of GDP (WDI 2001).
2.3.4 Street Gangs in El Salvador
The growing street gang problem in El Salvador is also traced to migration and remittances,
although the connection is somewhat indirect1 3 During the early part of the massive exo
dus of Salvadorans in the 1980s, many brought their families, including small children. Al
though technically illegal aliens, many of these children - now adults - have little memory
of and no allegiance to El Salvador. Former guerrillas who had relocated to Los Angeles
sought out these youths, particularly those from the most economically challenged homes,
and organized into street gangs. In Los Angeles, most Salvadoran gang members belong
13The information in this section is based on an excellent treatise of the origins and effects of gangs in El
Salvador by Smutt and Miranda (1998) titled Elfenom eno de las pandillas en El Salvador. Noguera (1999)
provides a more general treatment of gangs in the Caribbean and Central America.
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to either the 18th Street gang (mixed Mexican-American/Chicano and Central American)
or Mara Salvatrucha (almost entirely Salvadoran). The Mara Salvatrucha gang has ex
panded nationally with organizations in Washington, D.C., Chicago and even rural parts
of Tennessee and Kentucky. These groups are known for dealing in illicit drugs and small
weapons, and quickly resort to violence to protect or expand their territory.
In the early 1990s, the United States implemented a policy of deporting illegal aliens
with criminal records, and this included many members of Salvadoran street gangs. Once
returned to El Salvador, many of these individuals were out of place, having lived most
of their lives in the United States, and naturally gravitated toward each other to quickly
re-form their gangs. Further, these groups found a large pool of willing converts among
Salvadoran youths frustrated with poverty and the lack of prospects, and envious of the
stylish and expensive clothing and jewelery flaunted by their newly returned compatriots.
El Salvador has since become an important crack cocaine trafficking center, and the
trade in small firearms has increased dramatically. More significantly, these groups are
now responsible for 10% of all homicides in El Salvador, and it is estimated that street
gangs alone are responsible for the killing of approximately 720 Salvadorans every year
(Smutt and Miranda 1998, p. 2)14. The presence of street gangs has fundamentally altered
the character of Salvadoran communities across the country. Despite historically rampant
14Smutt and Miranda (1998) report a homicide rate in El Salvador of 12 killings per 100,000 people at
tributable solely to gangs. By comparison, the general homicide rate in the United States is 8 killings per
100,000 people.
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
poverty and political issues which have divided Salvadorans along ideological lines, con
cern over personal safety within communities is a relatively new concept, and the peace
time presence of heavily armed patrols in parts of San Salvador is shocking to many.
The link between the gang problem and migration is indirect but clear. In the eyes of
many, continued migration further erodes Salvadoran communities by promoting a ‘some
thing for nothing’ culture which places little value on the home community. Remittances
only exacerbate the situation by removing the historically strong Salvadoran work ethic
from youths who instead remain idle and learn to value the prestige and luxuries associated
with membership in a street gang. Indeed, the proliferation of street gangs in El Salvador is
among the most visible and compelling critiques of Salvadoran investment in international
migration.
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 3
Previous Research on Migration and Remittances
The tremendous global incidence of migration and remittances has spawned a vast and
growing literature on various aspects of these phenomena and their effects. This chapter
presents two aspects of the existing body of household-level or microeconomic research.
First, in order to lay the foundation for the approach adopted in this work, efforts to ex
plain migration and remittances are discussed with particular emphasis on work attempt
ing to bridge the gap between migration studies and remittance analysis. A second section
provides a summary of existing work on the experience of El Salvador. The social and
macro-economic impacts of migration and remittances are also considered, focusing on
work addressing remittances as Dutch Disease and the adverse effects of large exogenous
income flows.
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.1 Main Perspectives
The new economics of labor migration1 5 has made great inroads into the understanding of
migration processes. In general terms, the key advance has been the shift in focus away
from the individual and toward the household as the key decision making unit in the mi
gration process. Earlier work typically proceeded along neo-classical lines, citing wage
gaps between countries and low migration costs (both implicit and explicit) as the moti
vating factors for individual or household migration (Isaac 1947, Lewis 1954, Ranis and
Fei 1961, Harris and Todaro 1970, Todaro 1976, Kindelberger 1978; also see Davis 1988
for a historical perspective of early migration studies).
While such factors were, and continue to be, extremely important, theoretical perspec
tives eventually caught up to the empirical reality that migrants belong to a broader social
context. In particular, the pivotal, if not primary, role of the non-migrating family mem
bers in the migration decision came to the forefront of social science research on the topic.
Work spearheaded by Stark, Massey and Taylor (Stark and Bloom 1985, Stark and Tay
lor 1991, Massey et al. 1993) typically characterizes migration as a “calculated strategy”
(Stark and Bloom 1985, p. 175) whereby remittances form part of an “intertemporal con
tractual arrangement between the migrant and the family” (ibid.). This conceptual con
nection between strategic migration and remittances allowed great progress. For example,
the existence of remittances lends credence to the notion of migration as a risk sharing and
income diversification strategy among household members. Particularly in the context of
15The nomenclature persists despite its twenty year history. See Stark and Bloom (1985) for a brief but
thorough review of seminal work.
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
an LDC where labor opportunities may be scarce and highly volatile, income from abroad
may provide a feasible way to smooth adverse income shocks. Among other things, this
helps explain why poor households are at least as likely as the wealthy to send a migrant
abroad, as opposed to individualistic models which postulate that the wealthy would be
more likely to undertake risky activities like migration.
Interdisciplinary research has further elaborated the motivations of the household by
incorporating anthropological perspectives and emphasizing the placement of the house
hold within the community or reference group (Shankman 1993). This led Stark to develop
the concept of relative deprivation as a motivation for the household to send a migrant
abroad. While a household expects to gain in absolute terms from remittance income16, it
may also stand to significantly augment its status with respect to its reference group, thus
reducing its level of “deprivation” relative to the wider community. This again clarifies the
strong propensity of migrants to come from households in the lower income deciles. Also,
relative deprivation has important implications for income inequality within and between
communities which are further discussed in Section 2.3.
Other research has sought to rigorously address the negative assessments of migration
and remittances. One of the strongest critiques of remittance income is that it is primarily
used for immediate consumption and fosters dependency rather than going to productive
investment. Indeed, many studies find that approximately 90% of remittance income is
16The concept of ‘gain’ here should be treated delicately, as most authors have, to account for the migrant’s
foregone opportunities in the sending country; in other words, gain is actually net gain. Also, non-economic
research often emphasizes the mitigating effects of psychic or emotional disutility associated with the sepa
ration of family members.
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
used to supplement consumption. However, many such findings are not based on analy
ses employing a control group, so nothing is known of the behavior of non-migrants or
how migrants would have fared in the absence of migration and remittances. This signals
a new genre of research question which, rather than addressing the consumption habits
of migrants, seeks to elucidate the pivotal differences between migrant and non-migrant
households.
Adams (1991) partially fills this void using empirical survey data from Egypt which in
cludes households both with and without migration experience, decomposed into income
quintiles. The income ranking is crucial since Adams finds that middle income households,
as opposed to poor or wealthy, represent the majority of the migrant pool. The results of
his analyses suggest that poor households have a very high propensity to consume remit
tance income, but that middle income households are more likely to invest, particularly in
housing. While Adams further solidifies the finding that large remittance inflows are as
sociated with rising real estate prices, he also provides evidence to suggest that remittance
income is not always squandered and instead is used to provide more long-term benefits
to the household, given the proper conditions.
The primary critique of Adams’ work is his sample selection. Households that had a
migrant abroad at the time of the survey were excluded from the analysis. The focus was
instead on return migrants which could directly give information regarding migration ex
penditures and remittance practices. However, this narrows the applicability of Adams’
results to temporary migrations, and the reasons for the migrant’s return are not provided.
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
This generates a problem in that remittances from temporary migration are also viewed
as temporary by the household, and it is therefore more likely that the household may al
locate such income to specific investment projects, such as land purchases. It may even
be the case that the migrant’s original objective was to raise capital for further investment
such that the failure to use remittances to that end would entail negative repercussions for
the non-migrating members of the household. Otherwise stated, if migration is viewed as
permanent, the non-migrating household may have less incentive to use remittances to
ward productive ends. The exclusion of permanent migrants from Adams’ sample would
thus seem to bias his results in favor of the desired conclusion.
The recognition of migration as a form of saving has been much more prevalent in other
social science literature, primarily anthropology and sociology. The differential emphases
of these disciplines, as opposed to economics, channel research questions in different di
rections, but a common analytical perspective is slowly emerging that views existing re
search as excessively focused on narrow aspects of migration and remittances, and thus,
unable to provide logically consistent explanations. The key to more recent work, includ
ing this thesis, is to view migration and remittances as integrated processes.
Cohen (2001) provides one of the more notable examples from the anthropological per
spective. He suggests that most work has focused on the socio-economic costs of migra
tion, such as dependency and the breakdown of traditional societies, or on the potential
benefits of remittances as a resource for infrastructure development, community revival
and economic growth. Instead, Cohen focuses on what he terms the transnational approach
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
which “defines the outcomes of migration and remittance use as rooted in a series of in
terdependencies” (p. 955) encompassing economic and cultural factors. Further, Cohen
postulates that the basic migration decision emerges at the household level and is taken
“with the goal of satisfying the various and changing needs of the household” (p. 955).
While the connection in Cohen’s work to the concept of migration as investment is
strong, his work requires expansion. In particular, the mechanism by which different cul
tural and economic factors affect each other are only loosely defined. While the model is
suggestive of the directions labor and financial flows may be expected to take, little rea
soning is given to explain the motivations households may have for choosing various paths
beyond the incentives provided by macroeconomic forces. Further, the empirical findings
upon which the model is based rely on a sample of only 54 households. While the sample
is certainly representative of the tiny village of Santa Ana del Valle in rural Oaxaca, Mex
ico, there is no reason to believe that the migration patterns of those households are not
rooted in highly site-specific conditions. Cohen does obtain a rough panel of data on those
54 households by extracting their migration histories since the 1950s and draws connec
tions between migration patterns and contemporaneous macroeconomic conditions. Un
fortunately, Cohen does not include any other life cycle information or critical measures of
intra-household variation which could provide more complete and endogenous analyses.
The present thesis could easily be seen as an expansion of Cohen (2001). The formal
model of Chapter 4 provides a richer framework for conceptualizing the flows of.people
and resources and characterizing investment decisions. Also, the empirical analysis of
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 5 is based on a sample of nearly 600 households, allowing far greater variability
in socio-economic factors and geographical settings than is possible in Cohen’s dataset.
3.2 Other Views
While migration-oriented studies have benefited greatly from the adoption of broader the
oretical perspectives, remittance studies have, until recently, tended to remain somewhat
narrow and individualistic. This may, in part, be due to the early conceptualization of re
mittance studies as a corollary of migration studies. Analyses thus focused on the migrant
and his1 7 decision to send remittances, the remittance amount and the remittance channel.
Among economists, these included characterizing optimal remittances from the migrant’s
viewpoint in a manner not unlike models of altruistic behavior18. Socio-anthropological
work implicitly acknowledged the role of the household by framing the remittance deci
sion on the part of the migrant as a function of familial and geographic distance from the
home community and the credibility of social sanctions arising from the failure to send re
mittances. Both approaches suggested that remittance flows would dwindle over time but
remained optimistic that, given proper incentives, remittance income could play an impor
tant role in the economic development of the home country.
The remittance issue thus became considerably more interesting when empirical evi
dence began to emerge highlighting the negative aspects of remittance flows on receiving
17While the migration of women and entire households has been significant, the typical migrant is young
and male. For a nice treatment of the role of women in migration, see Pedraza (1991).
18Funkhouser (1995) is among the best in this genre and involves a comparative study of remittances to
El Salvador and Nicaragua.
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
communities. These critiques have indeed become the stylized facts associated with large
remittance flows: a tremendous propensity to consume remittance income19, a commensu
rate lack of saving or investment, and reduced labor force participation by recipients. Such
effects are evidenced in virtually all countries or regions where out-migration has been sig
nificant: Western Samoa, India, East Africa, the Middle East, the Philippines, Mexico and
Central America.
Anthropological and interdisciplinary efforts have identified other potential reasons for
negativity aimed at migration and remittances. Massive movements of people clearly alter
the social structure of the household and the community. While some scholars champion
these changes, citing the increasingly dominant and pro-active roles of women in labor-
sending countries (since most migrants are men), others blame increased drug use and van
dalism among local youth on the dissolution of traditional family structures. Remittances,
these scholars further suggest, only worsen the situation. The steady flow of income from
abroad instills a “something for nothing” attitude among the younger generations who de
velop no pride in their local communities and feel that any opportunities for a better qual
ity of life lie elsewhere20. Several extreme cases have been documented in Mexico and El
Salvador where the self-enforcing cycle of migration and remittances generated a cascade
such that formerly vibrant villages remain all but deserted today.
19This remains true even when consumption is defined to exclude expenditures on health and education.
20The following quote from a Peace Corps worker in rural El Salvador appears in Garcia (1996): “Perhaps
the most important transformation is seen in the youth. These days they only wear T-shirts brought from the
United States, tennis shoes and fashionable jeans. They speak typical urban American slang and only study
to be able to go [to the United States] someday. Nobody wants to do anything in the village anymore” (p.
12, my translation).
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Other work by Stark addresses the effects of remittance income on inequality within
and between communities. This topic is clearly complementary to his earlier work on rela
tive deprivation. When households care about both their relative and absolute well-being,
migration will be undertaken by households in the lower income deciles. Remittances will
thus tend to reduce income inequality within the community as the relatively poor augment
household income via remittances. Conversely, Stark finds that remittance income tends
to exacerbate income inequality in those areas where the concept of relative deprivation
fails to hold21.
3.3 Research on El Salvador
The massive emigration from El Salvador during the 1980s and the myriad changes it has
imposed on the economy and social structure of the country have been detailed in two
works by Segundo Montes (1987, 1990). For these studies, Montes undertook extensive
surveys of 2,000 heads of household from four specific socio-economic groupings: urban,
urban poor, urban marginal and rural. The surveys cover a wide array of topics including
information on household characteristics, migration, income, remittances, employment,
consumption and saving patterns.
An interesting aspect of the migration patterns is that the head of household seldom
migrates (Montes 1990, p. 51-52). This makes sense given that the typical migrant is a
member of the urban poor: these households depend almost exclusively on the income of
21 For a more macro-oriented treatment of the same topic, see Davies and Wooton (1992).
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the head of household for subsistence. Further, the expense of migration - especially since
much of it is done illegally - normally renders it impossible for more than one member of
the family to migrate at a time, thus migration of the entire household is also very rare.
Montes finds that most migrants, and therefore sources of remittances, are members of the
extended family, sons, nephews, etc.
Although the presentation of Montes’ work is primarily descriptive and atheoretical,
these are seminal works in the analysis of both the civil war on the Salvadoran economy
and society and in the effects of Salvadoran remittances. The results of these studies de
scribe changing sectoral employment, dramatic changes in the roles of women and income
variations across geographic regions and social categories. More importantly, however,
Montes discusses migration and remittances not in the context of understanding economic
refugees, but rather from the perspective of rational households sending family members
abroad as legitimate and lucrative investments. Indeed, Montes finds that many families
went into debt and sold assets in order to raise the funds needed to send a migrant and that
remittances were largely used to re-establish pre-migration consumption levels (1990, pg.
121). It is worth noting that no other study has empirically or analytically made use of
Montes’ conception of migration and remittances as household investment.
This perspective, however, is illuminating. For the typical household, sending a fam
ily member abroad is an extremely large sacrifice. But, Montes finds that, via remittances,
those costs are usually fully recovered within 14 months, and any additional remittances
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are thus net gains. Few other options available to Salvadoran households could he ex
pected to yield such rapid and substantial returns. For the urban poor, income from re
mittances more than doubles the amount of income they would otherwise receive. Further,
Montes finds that nutrition, child morbidity and elementary education rates for remittance-
receiving households are substantially higher than for households without remittance in
come. Montes’ analyses, however, are not intertemporal, and any pre-existing differences
between these groups of households are unknown. It may then be the case that receiving
remittances serves as a proxy to signal some other pivotal difference.
A great deal of attention has instead been focused on Montes’ finding that over 80%
of remittance income is used for current consumption. Montes does, however, include ex
penditures on health, nutrition and education in his definition of consumption. This finding
is nonetheless the source of discussions regarding a Dutch Disease effect originating from
remittances. Montes’ work has inspired several other studies in the areas of migration and
remittances.
The emigration from El Salvador throughout the 1980s had a wide array of effects on
the Salvadoran economy and society. Among them is the “brain drain” which Funkhouser
(1999) explores in an empirical exercise. The work actually encompasses five Central
American countries (Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua), but with
the exception of Costa Rica, the findings are similar, as are the histories of civil unrest, mi
gration and remittances. As in his earlier work, Funkhouser utilizes both source country
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and U.S. data to generate his conclusions. Utilizing household surveys from each coun
try for (the year closest to) 1990, Funkhouser reveals interesting patterns. In particular,
younger cohorts tend to be larger and more educated than older ones, but data from the
1990 U.S. Census suggest that the average migrant is substantially more educated than the
home country average. The net effect is slower growth of skilled labor and the maintenance
of wage premia to skilled labor in the sending country.
With respect to remittances, Funkhouser tests the hypothesis that more educated emi
grants would tend to remit greater quantities with greater frequency. The idea behind such
a test is to assess the degree to which a migrant effectively compensates the family via
remittances for his absence and the expense of having sent him to the U.S. Funkhouser’s
analysis, however, rejects this hypothesis: the education of the emigrant does not signifi
cantly affect the amount or frequency of remittances.
The primary critique of Funkhouser’s work is his sample selection of individuals aged
20-64 years in order to “focus on persons that had completed their education”. Although
Salvadoran 20 year olds average 7 years of education, the average educational achievement
in Central America is still only 5.5 years; by truncating the sample at 20 years, the analysis
excludes a large and dynamic pool of potential workers and migrants. In particular, this
excludes a large group of less educated migrants who may remit significantly less than
those included in the sample.
Other work by Edward Funkhouser focuses on comparative studies of El Salvador and
Nicaragua. These countries experienced similar demographic changes during the 1980s,
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
but the socio-economic impacts have differed substantially. Funkhouser [1992] focuses
on the generally negative impact of remittances on household labor decisions. This result
is important as it provides a reasonable explanation for what has been called “widespread
laziness” by many Salvadorans; remittances have merely raised the reservation wages of
many workers. Funkhouser [1995] compares the remittance patterns of Salvadorans against
those of Nicaraguans and finds that Salvadoran families receive nearly twice as much in
remittances.
Within El Salvador itself, remittance issues have received some, albeit insufficient, at
tention. Work by Gabriel Siri (1996) titled “The Productive Use of Family Remittances
in El Salvador”2 2 is the most notable example. The title is, however, somewhat mislead
ing since the actual productive use of remittances in El Salvador seems impressive only in
the extent to which it is lacking23. Indeed, a better title might have been “How to Improve
the Productive Use of Remittances in El Salvador.” Further, Siri’s work provides only a
general treatment of the issue and recommendations for improvement. By no means is
this meant to say that Siri’s recommendations are without merit. Siri emphasizes the role
of institutional reform in facilitating the productive use of remittances: expansion of the
banking system, especially in rural areas; the establishment of U.S. branches of Salvado
ran banks to more efficiently channel remittances into the banking system; the expansion
of microfinance programs to ease (rural) credit constraints (despite the fact that programs
22My translation. Original title is “El Uso Productivo de las Remesas Familiares en El Salvador.”
23Siri finds, as have virtually all other researchers, that over 90% of remittance income in El Salvador is
spent on immediate consumption, even excluding expenditures on education and health from the definition
of consumption (p. 11).
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to date have met with little success); the development of community remittance programs
to direct remittance income toward the provision of public goods.
At best, Siri puts forth a solid plan for long term reform, but it is clear that the mobi
lization of such plans requires the coordination of manpower, resources and political will
to an extent which is not currently feasible in El Salvador. At worst, despite the supposed
microeconomic focus of the paper, Siri completely ignores the incentives and opportuni
ties faced by households seeking the most efficient allocation of their scarce resources. In
other words, Siri’s approach is one of ‘build it, and they will come,’ disregarding the real
ities of poverty, labor markets and shifts in the sectoral focus of productive activity.
Siri [1996] has expanded on Montes’ work by decomposing in greater detail the ulti
mate uses of remittance income. However, taking a cue from the Dutch Disease literature,
Siri also seeks to explain why so little remittance income is saved or invested. He points
primarily to two key factors. First is the simple urgency of more pressing current needs:
the vast majority of remittance income is spent on the acquisition of necessities like food,
housing and medical care.
Second, Siri highlights the lack of proper mechanisms for channeling remittance in
come into productive uses. For example, almost all remittance income is transmitted from
the United States via private couriers, directly to the receiving family. Ideally, suggests
Siri, Salvadorans in the U.S. would be able to deposit their remittance in a bank account
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
at a branch of a Salvadoran bank in the U.S. In this way, remittance income would im
mediately enter the banking system, rendering possible the expansion of domestic credit,
investment, etc.
It is worth noting that most of the discussion to date regarding the remittance situa
tion in El Salvador has revolved around this issue of creating more efficient ‘remittance
channels.’ However, the success of the few programs that have been undertaken has been
limited due in large part to the fact that the problem is not entirely domestic, but rather,
involves the coordination and mobilization of both institutions and Salvadorans abroad.
Despite strongly held opinions and much anecdotal evidence, surprisingly little serious
empirical work exists addressing the effects of remittances on labor force participation.
Among the most carefully done work is Funkhouser (1992) which also studies the case of
El Salvador. The fundamental finding in his work is that remittances have a strong negative
impact on labor force participation. He is, however, careful to note that emigration did
serve as an effective tool to relieve unemployment pressures during the war period, but
that, as the Salvadoran economy grows and recovers from the effects of civil war24, the
effects of the “brain drain” will become increasingly pronounced.
There are, however, several problems with Funkhouser’s work which were, at the time,
largely beyond his control. First, Funkhouser utilizes the Encuesta de Hogares for 1985, a
time when the armed conflict was ongoing. My personal communications with officials at
24.. .and, more recently, natural disasters...
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Digestyc (the organization in charge of the survey) suggests that such data have a gener
ally low level of reliability, especially in the rural areas. Second, at that time, the Encuesta
did not include questions on remittances, so Funkhouser’s remittance data is extrapolated
from Montes’ 1987 survey and applied to data from the national survey; the constructed
variable is thus appropriately labeled ‘predicted remittances.’ Third, while Funkhouser ac
counts for differential effects by gender and includes dummies for each departamento he
does not capture the more subtle labor force effects of demographic change due to migra
tion. In particular, the growing prevalence of female-headed households and the effects of
remittances on subgroups (other than gender) are left out. Fourth, Funkhouser’s estima
tions are probit regressions which capture the effects of a binary definition of labor force
participation. It seems, however, that remittance income might have more nuanced effects;
specifically, remittances might not affect the individual’s decision to participate in the labor
market so much as the degree to which the individual participates.
The present work seeks in no small way to expand on Funkhouser’s work by accounting
more specifically for the effects of remittances on various subgroups using cleaner data and
interpreting the results through a migration as investment approach.
3.4 Dependency and Dutch Disease
“Dutch Disease” is the more picturesque nomenclature ascribed to a particular aspect of
booming sector analysis. Pioneering work in this area has been provided by Corden and
Neary [1982] and Corden [1984],
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The primary goal of booming sector analysis is to understand the macroeconomic ef
fects resulting from large unexpected income flows in a particular sector of the economy.
The classic applications of these methods focused on export booms of extractive resources
such as petroleum or, in the case of the Dutch, natural gas.
The basic setup is as follows. The model economy consists of three sectors: booming,
traded and non-traded. The price of non- traded goods is determined endogenously, the
price of traded goods is determined on world markets, and the determination of booming
sector (export) prices depends on the degree of market power the country has over that
good. These models typically deal in real quantities such that the price of exports in terms
of traded goods represents the terms of trade, and the price of non-traded in terms of traded
goods is the real exchange rate. Different models posit varying degrees of factor mobility
intersectorally and internationally as appropriate.
The inflow of resources from the booming sector has two major effects: a spending
effect and a resource movement effect [9]. The spending effect is the natural result of in
creased income if both traded and non-traded goods are normal. The outward shift of the
demand curve for non-traded goods increases the relative price, and productive resources
(labor, in the short run) are drawn out of the other sectors. This resource movement is
added to that which occurs initially as a result of the boom. The result is that while out
put and income have increased in both the boom and non-traded goods sectors, the traded
goods sector finds itself adversely affected by an outflow of resources. The deterioration
of the traded goods sector as a result of the boom is by definition the Dutch Disease.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The only other study which seems to have made a genuine attempt to model remittances
as the source of Dutch Disease is Rivera Campos [1996] which also looks at the case of
El Salvador. While the study postulates a seemingly standard model and reaches some
interesting conclusions, closer inspection reveals major flaws.
Rivera Campos puts forth a model where a representative agent maximizes discounted
lifetime utility subject to period budget constraints which include remittances. Standard
Euler equations are derived and used to reach conclusions about the effects of price changes
and the sources of Dutch Disease. However, Rivera Campos’ results are obtained using
two extremely restrictive assumptions: agents are not allowed to save, and there is no in
vestment in the economy. He proposes a national accounts identity which divides total pro
duction exclusively between the production of exports and domestic goods; investment is
not accounted for. Similarly, it is required that in each period total resources (income plus
remittances) be spent on the consumption of traded or non-traded goods; agents are explic
itly prevented from saving.
In terms of basic economics, the model is flawed. The key point in utilizing dynamic
tools is to understand how agents plan for the future and behave over time. However, the
relevance of such discussion hinges fundamentally on the ability of agents to spread re
sources over their planning horizons. Typically, this is modeled as some form of saving,
be it via goods, money, bonds, securities, etc. However, by disallowing saving and in
vestment, Rivera Campos unintentionally transforms a dynamic problem into a static one.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Thus, results like Euler equations, which describe a link between optimal consumption al
locations over time, can no longer be said to be valid. It is clear that with no possibility of
saving or investment, the mechanisms do not exist for those links to operate.
Rivera Campos effectively makes no progress in understanding remittances as the source
of Dutch Disease. By constraining agents to consume all their resources each period, con
sumption must have a one-to-one relationship with changes in income. Thus a positive
shock to income from, for example, remittances is necessarily fully reflected in consump
tion, regardless of the income elasticity of consumption. In a more flexibly formulated
economy, consumers might choose to smooth consumption such that it reflects an expec
tation over lifetime income rather than period income.
The work presented in the following chapters attempts instead to postulate a more com
plete and less restrictive model which leads to qualitatively similar results.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 4
Models of Migration-as-Investment
Models framing migration as a form of investment have been relatively few in the context
of agricultural households. At the level of the individual, a vast body of research exists
identifying the costs and benefits of migration, particularly focusing on the importance of
the wage gap between the host and sending countries, as well as the demographic charac
teristics of the individual (or sample population). While it is clear that such migration is
undertaken as form of investment by the individual, the decision problem is fairly straight
forward since the costs and benefits of the investment options are completely internalized
by the decision maker.
However, framing individual migration as a household-level investment decision ef
fectively requires the ability to enact an arrangement between the sending household and
the migrant despite physical separation and a potentially prolonged gap between the initial
investment and the flow of returns. Further, empirical testing of investment decisions at the
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
household level requires data with an intertemporal dimension that could begin to disen-
tagle the costs and benefits of the investment decision, and in some way allow distinction
between cross-section and intertemporal effects.
Indeed, the relative lack of dynamic models of household investment in migration may
be due, in part, to the largely cross-sectional basis of most empirical research on migration
and remittances. Cross-sectional data typically shows strong correlations between remit
tance income and other economic factors: reduced labor force participation, reduced sav
ings rates, elevated levels of current consumption25. A sizable literature has thus devel
oped addressing the ‘dependency effects’ of remittance income and the degree to which
remittances erode local communities by fostering inequality and removing the incentives
to invest at home. But drawing such conclusions of causality from cross-sectional relation
ships ignores the past decisions and sacrifices of the household. What have these families
given up in order to be able to receive remittances? More importantly, are there no better
options than migration?
Further, from a theoretical perspective, the static and cross-sectional approach rele
gates remittances to a random event beyond any control of the household, and this is cer
tainly not the case. The assumption of causality in cross-sectional relationships implies
that there exists no time dimension in the household’s decision process; instead, house
holds are assumed to optimize period-by-period and never look forward. Remittances are
25The relationship between remittances and current consumption remains robust even when the consump
tion of durables, health and education are accounted for.
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
thus a ‘free lunch’ to the household since there is no perceived trade-off to receive the re
mittances, and there exists no current mechanism to alter future remittances. But few re
searchers would agree that decision makers are so powerless or myopic.
Institutional economics, anthropology and sociology emphasize the importance of com
munities, social ties and kin networks in the enforcement and maintenance of efficient be
havioral patterns. There is growing and compelling evidence to support the idea that small
and tight-knit groups may have effective mechanisms to enact plans with temporally and
geographically separated dimensions. A reasonable model of migration must therefore in
clude several factors. If households can plan for the future, then time must play a role in
optimal household decisions. But time is only relevant if the household can allocate in
come between periods, so the agricultural household should have access to a menu (albeit
potentially short) of investment or saving options, including migration. Also, it is clear
that in order to receive remittances - the return from migration - the household must incur
costs. These include both explicit expenditures to send a migrant abroad and the opportu
nity cost of lost wages or farm labor. While many variants of agricultural household mod
els exist, no model has yet attempted to explain migration as part of a household’s dynamic
investment options.
A gap thus exists to reconcile the ability to act dynamically while explaining cross-
sectional evidence, and to that end, this chapter elaborates two models. The first is a rela
tively simple static model where remittances are exogenous, and it is shown that the typical
cross-sectional empirical result is the natural outcome of framing remittances as a static
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and exogenous event. The static model is important to the extent that it formally states the
classic arguments against remittance income and provides a benchmark for further com
parison. A second model proposes the migration-as-investment approach. Households are
long-lived and plan for the future by allocating current income to either land purchases or a
migration-as-investment scheme. Two key complications are added to better emulate the
actual situations of many developing countries. First, households are credit constrained
such that investment projects must be financed out of current income or wealth. Second,
both land and migration investments require entry costs. For land, these are postulated as
a form of transaction costs which should be thought of as a measure of the degree to which
access to land markets is constrained. For migration, there is assumed to be a household-
specific investment level below which the investment should be expected to fail and no fu
ture remittances would be forthcoming. The degree to which this participation constraint
binds will be a key factor in any household’s decisions, and the underlying empirical char
acteristics which determine a household’s investment ‘fitness’ are examined in Chapter 5.
As such, the model derives the conditions under which different households may find it
optimal to invest in land, migration, both or neither.
It should be noted that these models are aimed at small agricultural households and are
not intended to be descriptive of the minority of powerful and wealthy landowners which
can be found in any country, developing or developed. These models instead attempt to
describe the actions of the vast majority of small producers for whom income, credit and
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
access constraints should be expected to be real and often binding, and for whom the out
comes of different investment scenarios imply substantial quality of life differences.
4.1 Static Model
Consider a variant of the agricultural household model. There exists a traded good M for
which there are many close substitutes on international markets; its price q is thus the in
ternational competitive price and is exogenous to domestic firms and consumers. Addi
tionally, there exists a non-traded good C which is produced by households for domestic
consumption; its price p is determined in a competitive domestic market and taken as ex
ogenous by individual firms and households. Household production takes place according
to a strictly concave production function F(K, LF) where K represents the household’s
existing capital stock and LF is the labor input. The household may also earn income by
supplying LM hours to the labor market at an exogenous wage rate w. Total remittance
income is represented by p. The household budget constraint can thus be summarized as:
pC + qM = PF{K, Lf ) + wLm + p (4.1)
As both consumers and producers, the optimal labor input for household production
may be dependent on the household’s consumption preferences. This complication is cir
cumvented here by assuming complete labor markets, thus allowing separability between
production and consumption decisions. In particular, separability implies that the profit
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
maximizing level of variable input usage does not depend on the household’s preferences
for consumption. Practically, this implies a sequence of events whereby households first
maximize farm profit and subsequently take those decisions as fixed in the utility maxi
mization problem. Let LF represent the optimal level of farm labor input obtained when
the household maximizes farm profits such that Y = max[pF(K, l F) — wlJ].
Since the assumption of separability may not realistically describe the prevailing con
ditions in many developing (or developed) countries, a static model without separability is
elaborated in Appendix A. Separability is undone by assuming a binding labor market par
ticipation constraint that forces the optimal farm labor input usage decision to be directly
linked to the household’s optimal consumption decision. The results of the model without
separability are nonetheless similar to those derived below, but apply to a narrower range
of values implied by the conditions of optimality
In the present scenario, remittances p are likely to be composed of an expectation and
a shock such that p — E(p) + R. The formulation of expectations of remittance amounts
is unique to each household and is not explicitly modeled here. Keeping in mind that the
expectation may be zero, households effectively treat the difference between actual and
expected remittances as an income shock. Negative shocks are possible if actual remit
tances fall below the household’s expectation. For simplicity of exposition, it is assumed
that E{p) — 0 such that p in equation (4.1) may be replaced by R.
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Households also face a time constraint. The household’s total time resource T must
be allocated between time spent working on the family farm LF, time spent on the labor
market LM, and leisure L.
T = Lf + Lm + L (4.2)
Incorporating the above constraints and assumptions, the household’s consumption bud
get constraint can be restated.
Equation (4.3) is also known as the full-income budget constraint since it incorporates
maximum profit, the full value of the time endowment and explicitly states leisure as an
expenditure.
The household derives utility from the consumption of traded goods, non-traded goods
and leisure according to a strictly concave utility function. The household’s problem is
then to:
pC + qM + wL — Y + wT -f R (4.3)
m axu(C, M, L) (4.4)
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
subject to (4.3). The optimization problem yields standard first order conditions. Let A
represent the LaGrange multiplier and u,- — dufdi, i = C, M, L.
uc T Ap — 0 (4.5)
— u m + Ag = 0 (4.6)
(4.7)
The optimal values of C, M and L can be obtained by combining (4.3) and (4.5)-(4.7).
The objective of the static model is to show that postulating remittances as an exoge
nous event leads directly to reduced labor force participation and increased current con
sumption. This is most easily seen by solving the utility maximization problem in terms
of L and examining its reaction to changes in R. Performing this exercise yields:
Equation (4.8) is unambiguously positive if traded and non-traded goods are both normal.
Since leisure is here the antithesis of labor force participation, remittance income has a
clear negative effect on labor force participation26. This result represents the crux of the
standard approach to remittances and their effects at the household level. The relaxation of
the budget constraint provided by remittance income leads directly to increased consump
tion spending and reduced participation in labor markets.
26Given the assumption of separability, the effect of remittance income is focused on labor market partic
ipation, but not farm labor.
dL!dR w + pdC/dL + qdM/dL
(4.8)
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The static model yields no particular insight into the effects of remittance income on
household behavior. Indeed, the static model shows only that the consumption of normal
goods increases with income. Its purpose is merely to state formally the framework used
for most remittance analysis. As a counterpoint, a dynamic framework is elaborated below
whereby migration and remittances are endogenized and presented as one of potentially
several options in the household’s investment portfolio.
4.1.1 Static Model without Separability
The static model presented in Section 4.1 assumed that labor markets were complete. This
allowed a separability result which simplified solving for optimal values of the control vari
ables by dividing the optimization problem into a two-stage process. In the first stage, the
household optimized its farm labor input and took that decision as given in the utility max
imization problem, the second stage. In the utility maximization stage, the optimal farm
labor is then a constraint rather than a choice variable, thus reducing the dimensionality
of the optimization problems. Most importantly, the complete markets assumption makes
it such that the agricultural household’s profit maximizing choices as a producer are inde
pendent of the household’s utility maximizing consumption choices.
However, especially in the context of a developing country, the assumption of com
plete labor markets may not be realistic. The purpose of this section is thus to provide a
static model of optimal choice in the presence of exogenous remittances while relaxing the
assumption of complete labor markets. The essence of the result in Section 4.1 is modified
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
only in that the standard result holds true so long as the marginal value of leisure is greater
than the marginal revenue product of farm labor.
The only technical modification necessary to remove the separability assumption is to
impose a binding labor market constraint:
Lm < L (4.9)
As before, the household’s time endowment is allocated between time spent working on
the farm (L-{), time spent in the labor market (Lm) and leisure (L): T = V + Lm + L.
Combining inequality 4.9 with the time constraint, the household’s budget constraint may
be written as:
pC + qM = pF(K , T — L — L) + wL + R , (4.10)
and the household’s problem is then to max u(C, M, L) subject to equation 4.10. The first
order conditions are as follows.
u\C) - p\ = 0 (4.11)
u'(M)-qX = 0 (4.12)
u\L) - Pf(L)X = 0 (4.13)
Note th a t/(x ) = dF(x)/dx.
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To find an optimal solution, the first order conditions are manipulated to obtain C and
M in terms of L.
(4.14)
(4-15)
These equations can be decomposed to more easily understand the curves they represent.
Let g() = u/-1(), and note that the argument of g() is the price-weighted ratio of the
marginal utility of leisure to the marginal product of farm labor. It is assumed that the
values of these functions are inversely related such that a low marginal product of labor
implies a large farm labor input and, consequently, a high marginal utility of leisure due to
its relative scarcity. Letting h(L) = n'(L)/f(L), it is then the case that dh(L)/dL < 0,
implying that the value of the function h(L) increases as leisure decreases. Further, the
properties of the standard utility and production functions imply that d2h(L)/dL2 > 0
such that h(L) is strictly convex and positive, and asymptotically approaches infinity as L
approaches zero, and vice versa. The optimal values of C and M may thus be re-written
as:
C ~ g(h(L)) (4.16)
M — g{q/p h(L)) (4.17)
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Since g() decreases in its arguments, it must be the case that dg(L)/dL > 0. Note also
that the price ratio q /p serves as a proportionality parameter in determining optimal shares.
Equations 4.10, 4.16 and 4.17 are then combined to implicitly define the optimal level of
leisure.
pg(L) + qg(L] p, q) = pF(K, T — L — L) + wL + R (4.18)
The desired result of this exercise is obtained by total differentiation of equation 4.18
to obtain dL/dR.
d L / d R = + ( 4 ' m
A positive value of dL/dR states that the optimal consumption of leisure increases as re
mittance income increases, implying, as in Section 4.1, that labor force participation falls
as remittance income increases. Note, however, that in the present example with incom
plete labor markets, the reduction in labor force participation is limited to the range of val
ues where + q— ^ > pf(L). This condition is violated only when, as a result
of large leisure consumption, agricultural production falls to the point where the marginal
revenue product of farm labor outweighs the marginal value of leisure. Given that the labor
market constraint encourages a large farm labor input, and thus a low marginal product of
farm labor, it seems unlikely that, in practice, increases in remittance income would also
increase labor force participation.
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 Intertemporal Investment Model
This variant of the agricultural household model seeks to bring out the investment deci
sions of households facing multiple investment options. The inclusion of a time dimension
allows households to optimally allocate resources across periods and make optimal deci
sions by weighing the costs of investment against the household-specific expected gains.
The framework established allows two key questions to be addressed. Under what con
ditions do farm households favor investment in migration over land? What role do the
characteristics of the household play in deriving optimal decisions?
The household’s objective is to maximize the discounted sum of finite lifetime utility
from the consumption of good c. Period utility u(c) is assumed to be concave, and /? rep
resents the household’s subjective discount factor.
T
max ^2 (4.20)
t= o
The timing of events within each period is as follows. Households begin each period
with the amount of land carried over from the previous period, and are assumed to realize
any remittance income at the beginning of each period. Given the amount of land and the
exogenously determined market wage rate, the household chooses the proportion of house
hold labor to be utilized on the farm. Household income is then determined by the sum of
production, remittances and labor market income earned by the proportion of household
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
labor not employed on the farm. Income is then divided among consumption, land invest
ment and migration investment. Any investment in land occurs at the end of the period,
and, added to existing landholdings, comprises the land endowment for the next period.
Within any period, the household’s expenditure constraint can thus be summarized as:
ct + g(Mt) + St < + (4.21)
At t — 0, the household is endowed with an amount of land k0 > 0. It is assumed that
land does not depreciate, thus total landholdings in any period are the sum of all previous
additions to land:
t - i
kt — kt— i + Mt— i — ko + 22 Mj (4.22)
j-o
The migration return (remittance) function £() is assumed to be strictly concave in its
arguments, suggesting diminishing returns from incremental migration investment. While
the speed of decay of incremental returns is not specifically parameterized, the investment
function should be expected to plateau quickly. The nature of the migration investment is
that returns depend heavily on the characteristics of the migrant(s) as summarized in the
parameter vector a. For empirical purposes27, these include the age, gender and education
level of the migrant (or potential migrant) as well as other characteristics which may in
fluence the migrant’s incentive to remit: marital status, children and relation to the head of
27See Chapter 5.
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
household. This vector should be thought of as an indicator of migrant quality such that
its components are employed by the household to determine the likelihood of successful
migration. The vector a is then a shift parameter in the return function which affects the
total potential return from migration, but not necessarily the rate of decay of the remittance
return function. The vector z is meant to account for factors exogenous to the household
which may impact the value of returns, in particular, the exchange rate between sending
and receiving countries and the size of the wage gap between the sending and host country.
The function g() represents the cost of investment in land. It is assumed that any posi
tive investment in land requires transaction costs beyond the price of land, and that house
holds not investing in land incur no such costs.
g{Mt) =
Mt + 8{Mt) if Mt > 0
(4.23)
0 if Mt = 0
The shape of the transaction cost function S() is not explicitly specified since it would seem
likely to vary by location and situation. The best case scenario - essentially, the absence
of informational asymmetry - would imply zero transaction costs or transaction costs un
affected by the scale of the transaction. The polar opposite would suggest a prohibitively
steep cost function such that most land would be highly concentrated in the hands of a
small percentage of landownders. Intermediate scenarios are also possible. If, for exam
ple, wealth as proxied by existing landholdings is indicative of credit worthiness, then it
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
may be the case that small farmers with low amounts of collateral may have greater diffi
culty obtaining additional land28; for those households, the transaction costs may increase
with the size of the desired plot of land: S'(M) > 0. Indeed, it may be more correct to
define transaction costs as an increasing function of the relative addition to existing land
holdings: Mt/kf This would imply, for example, that a small farmer would expect to face
greater transaction costs than a large farmer in order to purchase the same plot of land since
it would represent a greater proportional addition to the small farmer’s landholdings. The
transaction cost function would then be increasing in the size of the plot purchased with
the amount of existing wealth (landholdings) serving as a shift parameter.
Nonetheless, the function S{) is assumed to be an increasing function of only the ab
solute size of the plot purchased for two reasons. First, in anticipation of the application
to the case of El Salvador, the large proportion of small landowners and the relatively nar
row range of plot sizes29 suggests that the overall effect of incorporating kt would be small.
Second, the addition of kt would add complexity to the model without greatly enriching
the result. Although the level of productive assets kt is fixed within any particular period,
in the intertemporal setting, forward-looking households may view it as a control variable
such that, for example, households may be willing to make initial land purchases with high
28The relatively high concentration of land ownership in Salvadoran agriculture and the large number of
small family farms suggest that this intermediate case may be most appropriate to the Salvadoran case.
29According to the Rural Household Surveys of El Salvador, the average farmer possesses 2.3 hectares of
land, and over 85% of all farms are within one hectare of the average.
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
transaction costs knowing that future additions to landholdings could be made at a rela
tively lower cost. However, this would create an additional element to be optimized in-
tertemporally, thus clouding the interpretation of optimality conditions while not changing
the basic fact that the transaction costs reduce the incentive to invest in land.
The migration investment also requires a feasibility constraint. In particular, there ex
ists a household-specific investment level S below which future returns on migration will
be zero. This effectively defines for each household the minimum cost of entry into mi
gration as investment. The household-specificity of S . is key; in particular, the minimum
investment level for any household will depend on the vector a. It must be the case that
the extent of the household’s willingness to invest in migration is intimately linked to the
expected minimum investment cost and the expectation of future gain. The vector a thus
contains the set of information which the household would employ to both identify the
member(s) of the household most suitable for migration and the expected rate of return on
the migration investment. In the theoretical sense, a then serves as the household’s mea
sure of migrant quality; however, in empirical work, the latent true migrant quality is only
proxied by individual demographic data. In general, it could then be said that households
with relatively high quality migrants may be able to enter into a migration as investment
scheme at a lower cost than households with migrants of a poorer quality and may also
enjoy greater rates of return. In terms of the minimum cost of entry, high quality migrants
6 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
may be able to migrate successfully despite a low explicit investment by the sending house
hold.
St > S(a) (4.24)
The household’s labor endowment is normalized to 1 and allocated to farm labor and
wage labor in proportions t and 1 — t, respectively. This inelastic labor supply by house
hold members is counter to much of the cross-sectional empirical evidence on the labor
force participation effects of remittances on receiving households. Indeed, there is sub
stantial cross-sectional evidence suggesting a negative correlation between remittance in
come and labor force participation. However, the definitions of labor force participation
have not always been applied consistently, and individuals which exit the labor market in
favor of home production are sometimes counted as having reduced their labor force par
ticipation. This suggests in particular the possible failure of survey data to capture fully
optimal changes in household time allocations as opposed to changes in the degree of la
bor force participation. As a counterpoint, this model instead allows households to shift
their time allocations between farm labor and market labor while retaining the assumption
of full participation, regardless of remittance income.
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The LaGrangian equation can be formed by combining the household objective func
tion (4.20) with equations (4.21), (4.22), (4.23) and inequality (4.24).
t j3tu{ct' ) -f- Af[/(&t_i -f + (1 — h)wt
£ = E
t=0 + £{St-i; a , z ) - c t - Mt - S(Mt) - St) + As(St - S)
(4.25)
Maximization of equation (4.25) yields fairly simple first order conditions.
f3tu(ct) — A t = 0
/'(/<) - wt = 0
-At(l + S'(Mt)) + Xt+if(k t+1) — 0
— A t -f A f+i^ (5 i) + A s = 0
A sSt = 0
(4-26)
(4.27)
(4.28)
(4.29)
(4.30)
Note that equation (4.30) emerges as part of the Kuhn-Tucker conditions for optimality
with inequality constraints. Since it is ex ante uncertain whether the optimal level of mi
gration investment will be greater than S. for any particular household, failure to include
equation (4.30) leaves open the theoretical possibility that a seemingly optimal solution,
and, in particular, a seemingly optimal level of migration investment, may lead to wasted
funds if the level of investment is below S. If the value of As may be interpreted as the
shadow value of relaxing constraint (4.24), then only households for which the constraint
binds at an optimum will have a positive vaue of A s. If the constraint binds, the household
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
always sets 5 = 0, and equation (4.30) holds. The opposite is also true: if the migration
participation constraint does not bind at an optimum, there will be a positive level of mi
gration investment. It must then be the case that there is no value in relaxing the constraint,
thus A s = 0 and equation (4.30) again holds. The Kuhn-Tucker condition therefore pro
vides a convenient mechanism to ensure that the solution to the household problem yields
a true optimum. Further, the particular mechanism by which equation (4.30) holds for dif
ferent types of households will play a key role in determining an individual household’s
investment path.
For a household that chooses to invest in both land and migration, it must be the case
that equation (4.24) does not bind and that Xs = 0. This allows the standard intuitive result
to obtain: at an optimum, the rate of return on all investments is equalized.
t - i ( q \ _ f ( k t + M t )
e ( S , ) - T-Ts'Im.) ( 4 ' 3 I >
Note that the transactions costs associated with land purchases tend to bias the investment
portfolio toward migration investment by discounting the rate of return on land. At a prac
tical level, wealthier households with relatively large existing landholdings would be ex
pected to typify those investing in both land and migration. Fundamentally, most house
holds are unlikely to be able to afford to viably undertake both investments at once. Also,
households with existing farms may have an advantage over newcomers insofar as agri
cultural practices and farm management such that existing and experienced farmers may
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
be better able to extract the full agricultural potential of additional plots of land. Newcom
ers, on the other hand, may find agriculture to be a highly risky venture, and the additional
transaction costs involved are likely to discourage many from farming.
The more common scenario is then likely to be migration with no investment in land:
households which optimally set Mt = 0, St > 0 and find A .s = 0. In this case, the transition
equation for consumption can be described by a modified Euler equation:
/?V (ct) = £'(St)/T+V ( c m ) (4.32)
Households which choose to not invest at all simply consume the value of their pro
duction and market wages each period. The model suggests several reasons why a house
hold may choose to not invest. For example, a household without a potential migrant (null
a vector) will not be able to participate in migration; alternatively, a household may find
that the characteristics of its potential migrant would necessitate a minimum level of in
vestment S. too large to be feasible as part of an optimal plan. Similarly, households may
shy away from investment in land due to low expected rates of return which are further
diminished by the initial transaction costs.
The implications of the model are basic and significant in terms of policy recommen
dations and the conclusions which may be drawn regarding the motivations for migration
and remittances and their subsequent effects. For example, the model suggests that gov
ernment advertising campaigns encouraging families to save and invest their remittance
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
income are destined to fail. Families with remittance income have already invested in mi
gration, and the remittances are their return. Use of that return to augment consumption
should not be unexpected, especially with the knowledge and expectation of continuing
remittance income. And while it may be the case that remittances foster a form of depen
dency on foreign income, such accusations ignore the previous sacrifices and expenses of
the household. Fundamentally, options and opportunity are the driving forces. The migra
tion and remittance scheme will persist so long as households perceive their investment
options to be limited and constrained and their best opportunity for advancement to be out
side the country.
The theoretical framework presented in this chapter suggests that households optimize
their consumption, saving and investment decisions with respect to their current and ex
pected future income streams. While the model captures the idea of simultaneity in the
decision making process, as well as the sort of economic trade-offs the household faces, it
must still be the case that much of the subtlety and crucial aspects of inter-household varia
tion are subsumed by the theoretical framework. For example, the model makes clear that
the parameter vector a, which summarizes a household’s migrant quality, is likely to be
pivotal in any particular household’s decision to invest. A household with very high mi
grant quality may find that its expected return from migration far outweighs its expected
return from any other form of investment; further, high migrant quality may also reduce a
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
household’s cost of entry into migration as investment, thus easing the participation con
straint. A more thorough characterization of these household-specific factors and their real
importance in household decisions is crucial, but goes beyond the scope of the model.
Fortunately, empirical findings can serve to verify and enrich theoretical results. In
particular, the empirical findings in subsequent chapters serve to elucidate the relationships
which models can only peripherally address. What factors affect a household’s willingness
and ability to invest at all? Given the decision to invest, what affects the household’s choice
of investment mechanism? This involves not only choosing migration or land investment,
but also the method of financing: Do households borrow in order to send a migrant? Do
they draw down their stock of assets? Finally, if migration is an investment, then remit
tances are the return. To what degree to factors beyond migrant quality affect the return to
migration?
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 5
Investment in El Salvador
The Salvadoran experience provides an exceptional opportunity to study migration-as -
investment. Beginning in 1979, as a result of civil war, a massive exodus began30, and the
outflow persists until today, albeit at a slower pace. At present, approximately 10% of all
Salvadorans live outside of El Salvador, and most emigrants regularly send money to non
migrating family members in El Salvador. Despite the end of armed conflict in 1991, return
migration has been almost zero, and remittances have only grown. Taken together, the
volume of these remittances totals nearly 11% of Salvadoran gross domestic product and
surpasses the aggregate value of El Salvador’s annual coffee exports. While remittances
have received a great deal of attention at the macroeconomic level, comparatively little has
been done to explore the microeconomic aspects of remittances, and even less work has
been conducted to address the fundamental source of remittances - migration. This signals
a failure not only recognize the causal link between migration and remittances, but also a
failure to link household incentives to national consequences.
30See Section 2.2.
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
While the model of Chapter 4 provides a conceptual framework for the analysis of mi
gration and, more generally, investment in the context of an agricultural household, this
chapter focuses on the empirical characterization of the factors which determine the ulti
mate investment choices of the household by separately addressing four inter-related is
sues. The first refers to a household’s investment ‘fitness,’ a term adapted from evolution
ary biology and meant to be indicative of the household’s ability to invest successfully31.
Given that not all, and scarcely the majority, of all Salvadoran households invest, along
what lines can investors be separated from non-investors? The answer is somewhat para
doxical: while the age and gender of potential migrants is extremely important, most other
wealth, geographic or household composition indicators seem to have little or no effect.
The second issue is then to more finely characterize households which choose par
ticular forms of investment. The disaggregation of investment types reveals an invest
ment portfolio for Salvadoran households disproportionately focused on migration. At
over 30%, the proportion of households investing in migration is approximately twice that
of all other investment options combined. Again, the choice is driven by the characteristics
of the migrant: households with working age men are significantly more likely to invest
in migration.
Third, the theoretical model suggests that households should incur some cost to enter
into migration as investment, so the various methods of migrant finance are studied, includ
ing dissaving, borrowing and selling assets. It is not surprising, however, that these prove
31Reproductive fitness in evolutionary biology refers to an adaptation or behavior which augments indi
vidual probabilities for successful reproduction of viable offspring.
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
difficult to characterize, especially in the context of small agricultural households. These
households typically have few assets to sell, and access to credit channels is limited. This
manifests in the data as an only weakly systematic association between household char
acteristics and the type of financing. Indeed, the most significant result is that households
with working age men - the most likely to send a migrant - are less likely to engage in
costly finance options. This again relates to the migrant quality vector a in the model and
should not be interpreted to mean that households incur no costs to send a migrant, but
rather, that higher quality migrants impose lower explicit costs on the sending households.
Having discussed the costs of investment, the fourth issue is to discuss the benefits,
remittances. In particular, what characteristics of the household affect the likelihood of
receiving remittance? While female-headed and landowning households are found to be
more likely to receive remittances, the vast majority of the explanatory power lies in the
migrant quality indicators.
The remainder of this chapter consists of an expose of the data and methodology uti
lized and a detailed analysis of the results of the above-mentioned aspects of migration as
investment. The conclusions to be drawn from the results are simple and powerful. Sal
vadoran households with young men have tremendous incentive to invest in migration and
little incentive to invest within the country. The key to minimizing the ill effects of migra
tion and remittances thus cannot lie in exchange rate management nor in advertising cam
paigns urging Salvadorans to save their money; instead, steps should be taken to improve
the rate of return on domestic investment options and reduce the costs of entry to them.
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
While this implies the need for substantial institutional reform, policy makers should oth
erwise expect little change in the current investment patterns of Salvadorans.
5.1 Data and Methodology
This study seeks to investigate various aspects of the investment decisions of Salvadoran
households. Approaching these issues from the perspective of migration-as-investment re
quires a dataset that accounts for changes in household composition over time, in addition
to economic and demographic information. While a nationally representative longitudinal
dataset for El Salvador does not exist, the Rural Household Surveys of El Salvador com
piled by the Ohio State University Rural Finance Program via the BASIS Project (Broaden
ing Access and Strengthening Input Market Systems) provides a strong dataset for address
ing migration-as-investment. The surveys were administered in 1996, 1998 and 2000 to
738,623 and 696 rural households, respectively, in each year. The surveys include a strong
panel aspect: 668 households were surveyed in at least two years of the sample, and 470
households are present in all three surveys. It is worth noting that the Rural Household
Surveys do not seek to be nationally representative, but rather, seek to extract the char
acteristics of the rural population of El Salvador. Households were thus categorized as
landowners, agricultural workers, non-agricultural workers and mixed income workers32.
32 T w o different surveys were administered depending on the landowning status of the household. Non
landowning households were ex p o st identified as agricultural workers if at least 67% of household income
was derived from agricultural activities, and non-agricultural workers are those with no more than 33% of
household income from agricultural activities. Mixed income workers are those with 34-66% of household
income from agriculture.
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Landowners, agricultural workers and non-agricultural workers each represent about 30%
of each year’s survey, while mixed income accounts for the remaining 10%.
Unfortunately, there is a significant lack of consistency between the datasets such that
data available in one survey may not be available in subsequent surveys or may only be
available in a highly indirect and often incomparable fashion. For example, the 1998 and
2000 surveys contain detailed information about each piece of land a household owns or
works, the crops grown on those lands and the inputs used. In contrast, the 1996 data shows
only the household’s primary plot of land and the primary crop grown on it. The result
is that agricultural income and expenditure data for 1996 is only loosely comparable to
subsequent years. Similar problems exist for the household’s pool of migrants, the demo
graphic characteristics of migrants and the financing methods used to send migrants, all
crucial concepts in the analyses below. Finally, the Rural Household surveys assign each
member of the household an alphabetical code which identifies the individual throughout
the questionnaire. Combined with the household-specific code, the alphabetical code is the
only definitive way of identifying a specific individual within or between surveys. How
ever, the alphabetical identifiers used in 1996 do not correspond in subsequent surveys to
the same individuals within a given household, and it is thus not possible to reliably track
individuals over time using the 1996 data. For these reasons, the 1996 survey is excluded
from the analysis.
The bulk of the results presented below are thus based on the 593 households that were
surveyed in both 1998 and 2000. While this selective sample reduces the degree to which
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1998 N 0-12 13-20 21-34 35-50 51-64 65+
Male 2,027 16.96 10.92 7.81 5.94 4.78 2.77
Female 2,095 18.00 10.48 9.02 6.99 3.95 2.38
Total 4,122 34.96 21.40 16.84 12.93 8.73 5.14
2000 N 0-12 13-20 21-34 35-50 51-64 65+
Male 2,143 16.01 10.13 8.05 6.68 5.32 3.83
Female 2,141 15.97 9.99 9.50 7.82 4.32 2.38
Total 4,284 31.98 20.12 17.55 14.50 9.64 6.21
Table 5.1: Age Distribution by Sex, 1998 and 2000.
the data may be considered nationally representative, the fundamental characteristics of
the Rural Household Surveys are maintained. Summary statistics were generated for the
entire sample and for the selected subsample with no appreciable changes. In particular,
the proportions of households designated as landowners, agricultural workers and non-
agricultural workers were not altered, and basic data such as household size, education
levels, household income and migration experience remained unchanged.
While the Rural Household Surveys focus primarily on the socio-economic character
istics of the household, there are also detailed sections with demographic and labor force
information on each individual within the household. Table 5.1 shows the distribution of
the sample population by age categories for each year. The differences between 1998 and
2000 are characteristic of the demographic trends within El Salvador as a whole. The
addition of 162 individuals during the period is reflective of the nationwide 2% popula
tion growth rate, and the growing percentages of older adults as compared to children and
young adults also reflects the overall aging of the Salvadoran population.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Educational attainment in El Salvador has historically been low, and Table 5.2 suggests
that, despite tremendous efforts to expand educational coverage, there remains substantial
room for improvement. The 5.3 average years of education for Salvadorans in 1998 be
tween the ages of 13 and 34 is more than double the educational attainment of the older
cohorts. The positive trend continues in 2000 as the average for 13 to 34 year olds rises
an additional 7.5%. Although narrowing, there is also a persistent gender gap in the ed
ucational attainment of Salvadorans. At the primary and secondary levels, females have
slightly higher mean levels of education, but the fact that very few Salvadoran women pur
sue college-level education shifts the gap in favor of males for the population over 20 years
of age. Enrollment rates suggest a further narrowing of the gender gap in the future, par
ticularly at the secondary level. While school enrollment rates for males between the ages
of 13 and 20 remained constant at 46% during the period 1998 to 2000, females achieved
a nearly 4.5% increase from 44.5% in 1998. The variance in both educational attainment
and school enrollment by sex is also telling: in both cases, the variance of males is approxi
mately half that of females. This suggests that while mean enrollment rates and educational
attainment for males is driven by the cohort as a whole, the growth in averages for females
is motivated by very high achievement by a select few.
It is worth noting that, despite the positive trends, the persistently low mean education
levels in El Salvador are indicative of a largely unskilled labor force. Nonetheless, educa
tional attainment should not be the only measure of a population’s ability to progress, and
it may be more relevant to discuss the dispersion of more tangible skills. Adult literacy
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1998 0-12 13-20 21-34 35-50 51-64 65+
Male 1.04 5.55 5.35 3.22 2.03 1.37
Female 1.06 5.80 4.72 2.44 1.66 1.05
Total 1.05 5.67 5.02 2.79 1.86 1.22
2000 0-12 13-20 21-34 35-50 51-64 65+
Male 1.15 5.68 6.05 3.80 2.24 1.19
Female 1.37 5.87 5.43 2.74 1.44 1.28
Total 1.26 5.77 5.72 3.23 1.86 1.22
Table 5.2: Mean Years of Education by Sex and Age Category, 1998 and 2000.
in El Salvador is approximately 80% and nearly 90% among 13 to 20 year olds. While
the Rural Household Surveys do not provide data on numeracy, documentation of the Sal
vadoran expansion in primary education emphasizes the focus on basic arithmetic skills
in addition to literacy, and the level of improvement in numeracy is expected to be on par
with literacy.
The migration as investment approach focuses on decisions made at the household level,
so it becomes necessary to go beyond the experiences of individuals and organize the dataset
to form households. The final household-level dataset thus essentially contains four types
of variables: aggregates of individual data, averages of individual data, totals of household
data, and binary variables created from either individual or household data. For example,
information such as the household’s total labor market income is generated as the sum of
the labor market income of all members of the household. Most household level demo
graphic information is instead derived from an average of individual data: age, education,
dependency ratio, etc. Other information is originally provided in the data at the household
level: agricultural income and expenditures, land ownership status, investment activities
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(including migration). Finally, several variables are created to signal particular character
istics of the household such as a female-headed household, participation in various types of
investments, methods of investment financing and whether any member of the household
receives remittances. Unless otherwise stated, any further data descriptions or analyses
will be based on the household-level data.
The dataset categorizes households as landowners, agricultural workers, non-agricultural
workers and mixed workers based on the sources of household income. As such distinc
tions are likely to be indicative of other differences between these groups, Table 5.3 dis
plays the variation in household income categories by various subgroups. To bring out
geographical differences, households were divided along East-West lines33. In general,
the Western half of the country, which includes the capital, has more fertile land, is more
urbanized and has historically been less affected by natural disasters34. The more favor
able conditions in the West have generated economic differences between the regions. For
example, the 16.9% lesser incidence of landownership in the West is reflective of the lati-
fundio system - large plantations - that have dominated in El Salvador, particularly in the
Western coffee growing regions. Small family farms are thus more common in the East.
Similarly, the 12.4% disparity in non-agricultural workers reflects the lesser degree of ur
banization in the East and thus the lesser demand for non-agricultural employment.
33The states labeled as East are Cabanas, San Vicente, Usulutan, San Miguel, Morazan and La Union. The
Western states are Ahuachapan, Santa Ana, Sonsonate, Chalatenango, La Libertad, San Salvador, Cuscatlan
and La Paz.
34This is not entirely true. While the San Salvador region has been hit by several large and destructive
earthquakes, the Western region has been less frequently and less severely damaged by floods and hurricanes.
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Category All West East Fem. HOH
Landown 31.2 24.9 41.8 29.9
Ag. Work. 26.3 26.8 25.5 19.5
Non-Ag. 33.7 38.3 25.9 43.4
Mixed 8.8 9.9 6.8 7.3
N 1186 746 440 164
Table 5.3: Income Categories by Sub-groups, Percent of Total.
Especially as a result of the out-migration of the 1980s, gender issues have generated
a great deal of interest in El Salvador. The vast majority of migrants in the initial exodus
were male, and combined with the war-related casualties, the proportion of female-headed
households has grown to nearly 15%. In terms of income sources, 43.4% of female-headed
households are categorized as non-agricultural, as opposed to 33.7% of the sample popula
tion. While the land ownership status of female-headed households does not differ greatly
from the average, they are nearly 7% less likely to be agricultural workers. This pattern is
expected since most agricultural workers are men, and there is some anecdotal evidence to
suggest that female-headed households have a tendency to undertake small manufacturing
or service enterprises.
In the spirit of Adams (1991) the data is decomposed by income quintiles. This brings
out differences in investment patterns and socio-economic characteristics in addition to
providing a method to account for non-linearity in income data. Households were catego
rized into quintiles on the basis of per capita household income by year, thus allowing the
tracking of social mobility between time periods. A portion of the Rural Household Sur
veys asks heads of household about non-labor market and non-agricultural income sources
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Category
Ql
Q 2 Q 3 Q 4 <25
Landown 42.0 33.2 32.6 27.7 20.3
Ag. Work. 21.8 32.4 28.8 29.8 18.6
Non-Ag. 26.5 25.6 28.8 31.1 56.8
Mixed 9.7 8.8 9.7 11.3 4.2
N 119 119 118 119 118
Table 5.4: Income Categories by Income Quintiles, Percent of Quintile.
including income earned from the sale of items manufactured in the household, services
provided by the household and income from the rental of real estate, vehicles or equipment.
Per capita household income is thus calculated as the sum of agricultural income, labor
market income, rental income, store or household production income and remittances, di
vided by household size.
Table 5.4 details the sources of household income by quintiles35. The most striking
finding is the relative lack of agricultural landowners in the upper quintile. This finding
that 56.8% of upper quintile households are labeled as non-agricultural is not surprising
since the members of such households are typically more educated and typically employed
in business or government. On the other hand, 42.0% of lower quintile households are
agricultural landowners; while this may seem paradoxical, it should be noted that lower
quintile households typically possess the smallest and least profitable farms.
Table 5.5 details other socio-economic information by income quintiles. The smaller
average household size of 4.7 persons in upper quintile households translates to a 24.3%
dependency ratio as compared to the 31.8% average dependency ratio among the lower
35Note that the income quintiles are labeled from Q1-Q5 in ascending order.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Category
Qi
< 22 Q3 Q4 Q5
HH Size (pers) 5.8 7.0 6.4 6.1 4.8
Dep. Ratio (%) 31.6 31.0 30.9 33.7 24.3
HOH Educ. (yrs) 2.3 2.2 2.4 2.8 4.5
Fem. HOH (%) 15.5 16.0 15.7 10.9 11.0
East (%) 46.6 41.6 38.6 34.5 24.2
N 119 119 118 119 118
Table 5.5: Summary Characteristics by Income Quintiles.
four quintiles. A similarly significant difference exists in the educational attainment of the
head of household in the upper quintile as compared to the lower four quintiles. At an
average of 4.5 years, upper quintile heads of household nearly double the 2.4 year average
educational attainment of the lower four quintiles.
Anecdotal evidence exists suggesting that the gender of the head of household may
play a key role in determining household behavior and that, in particular, female-headed
households may be more likely to pursue migration. Table 5.3 confirms that female-headed
households receive a disproportionate amount of income from non-agricultural sources,
but Table 5.5 suggests that female-headed households are nearly 44% more likely to come
from the lower income quintiles. It is therefore ex ante unclear whether any behavioral
differences typically attributed to female-headed households are actually characteristic of
the lower income quintiles generally.
Table 5.5 again highlights geographical differences within El Salvador. Of all lower
quintile households, 46.6% are located in the Eastern region, and the higher quintiles have
a markedly diminished prevalence in the Eastern region such that only 24.4% of upper
quintile households are found in the East.
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The Rural Household Surveys of El Salvador unfortunately lack consumption data. An
ideal dataset would allow expenditure comparisons among various sub-groups, especially
among investors and non-investors. Nonetheless, the Rural Household Surveys provide
detailed information regarding potential household investment activity. Household pur
chases of land and farm equipment are known, as are new business ventures and whether
any member of the household has a savings account. Further, the surveys allow the ex
traction of the household’s investment in both intra- and international migration. It is thus
possible to address thoroughly household investment practices and analyze trends and ten
dencies by various characteristics of the household and its members.
5.1.1 Investment Options
Before deciding on a specific investment mechanism, the household must first determine
whether its situation is amenable to any form of investment at all. While such determina
tions are likely in reality to occur simultaneously with the household’s greater optimization
problem, the structure of the available data allows a relatively thorough characterization
of Salvadoran households that invest and of the degree to which specific factors influence
their decisions. The Rural Household Survey provides information on seven household
investment mechanisms: migration, land purchase, purchase of farm equipment, opening
a store or shop, opening a savings account, having children and obtaining education.
The migration option can be subdivided into two primary categories: international mi
gration and internal migration. For both the 1998 and 2000 datasets the total number of
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
each household’s migrants is computed by destination. Any particular household was then
considered to have invested in international migration if the total number of international
migrants in 2000 exceeded the total number in 1998, and an indicator variable was created
to identify such households. An identical procedure was followed to identify households
which invested in internal migration between 1998 and 2000.
Table 5.6 shows the distribution of Salvadoran migrants in 1998 and 2000. Specifi
cally, the net total number of migrants in each destination is shown by year. The United
States receives 94% of Salvadoran international migrants, and of internal migrants, over
60% resided in San Salvador in 1998. However, in 2000, the number of migrants in San
Salvador dropped by half, and the number of migrants to the United States nearly doubled.
The striking shift of migrants out of the San Salvador metropolitan area may be partially
attributed to the effects of Hurricane Mitch which struck El Salvador, primarily in the East,
during October, 1998 as migrants returned home to help family members recover and re
cuperate. Despite media reports describing such return migration, it is not supported by
the data since none of the reported return migrants in 2000 came from San Salvador. In
deed, it seems that the effect of Hurricane Mitch may have been to push more households
toward international migration. As crops were flooded and equipment destroyed, many
farm households likely found themselves unable to rebuild their farms, thus leaving little
choice beyond migration. Alternatively, some households may have found that, although
rebuilding may have been feasible, their investment funds were better spent on migration.
8 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1998 2000
Destination N % N %
San Salvador 380 29.87 191 11.07
El Sal. Other 177 13.92 203 11.76
U.S. 666 52.36 1,257 72.83
Other Cent. Am. 25 1.97 37 2.14
Other 24 1.89 38 2.20
Total 1,272 1,726
Table 5.6: Net Migration by Destination, 1998 and 2000.
In either case, lower entry costs or potentially higher expected returns from migration may
have been motivating factors.
Other exogenous factors are also likely to have stimulated the shift toward international
migration. Figure 2.1 shows that the Salvadoran economy experienced a slowdown during
the late 1990s which was associated with greater job scarcity and slow wage growth. At
the same time, the U.S. economy was experiencing a prolonged boom, and the differential
opportunities are likely to have attracted many Salvadoran migrants.
Although it is valid to consider internal migration as a form of household investment,
the costs and benefits associated with internal migration differ dramatically as compared
to international migration. The cost differences are clear. El Salvador is a small country,
and San Salvador lies in close proximity to the geographic center; the transportation cost is
thus necessarily minuscule as compared to international migration. Also, internal migrants
do not face the language or cultural barriers, or imminent threat of deportation36 associated
36The vast majority of Salvadoran migration to the United States occurs illegally.
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
with international migration. The potential benefits of internal migration are also consider
ably lower, especially for low skilled workers. While there exists a persistent urban wage
gap in El Salvador, it seems to be driven by skilled wages, and despite the growth of the
maquila system, the supply of low skilled workers continues to outpace demand. Thus
internal migration may frequently occur for reasons only loosely tied to investment: to re
allocate surplus farm labor, marriage, etc. One important consideration, however, is that
individuals seeking higher education may migrate to the capital. The data suggest that only
5% of internal migrants were studying in 1998 or 2000, so internal migration for human
capital investment seems rather uncommon.
Investment in land between 1998 and 2000 is derived from the 2000 Rural Household
Survey. Respondents are asked about plots of land ever acquired or divested by the house
hold (via purchase, sale, inheritance, squatting, etc.) and the year in which any such trans
actions took place. An indicator variable was thus generated to identify households which
purchased land in 1998,1999 or 2000. Calculated as such, it is found that 3.2% of house
holds invested in land during the period in question. Investigation of simple correlations
among the data suggests that these households share few systematic relations in terms of
wealth, income or household composition, and that key variations are likely to emerge by
generating more subtle interactions.
Rather than purchasing land, households may choose to augment their stock of pro
ductive farm capital as a form of investment. This option is included primarily to take ad
vantage of the features of the dataset and to more completely characterize the investment
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
options of the household. Agricultural production in El Salvador, however, is dispropor
tionately labor intensive, and the general paucity of farm machinery is reflected in the data.
For both 1998 and 2000, the household’s ownership of several types of farm machinery is
calculated: irrigation equipment, a mill, a seeder, a tractor and a truck. Quantities were
again compared by year to determine investment for each household. A striking feature
of the data is how few farms use such equipment. In 1998, only 4% of farm households
made use of a plow, and less than 2% utilized any of the other types of farm equipment.
Also, with the exception of mills, ownership of these types of farm implements declined
during the period. Closer inspection of the data reveals that nearly all of the major farm
equipment was owned by the three largest family farms, and most households instead rely
on handheld tools: pick-axes, wheelbarrows, pump-and-bucket watering. Taken together,
less than 1 % of households invested in farm equipment. Due to the disproportionate degree
of variation in potential regressors as compared to investment in farm equipment, directly
testing the incidence of such investment seems inappropriate37.
Anecdotal evidence suggests that households sometimes use remittance income to open
a shop or start of small business. While the data does not make a direct link between family
businesses and remittance income, it is possible to trace household investment in a small
business. The set of questions asked in the Rural Household Surveys provides two meth
ods to assess whether a household owns a family business. First, households were asked
directly whether they owned a store or family business. As before, values for 1998 were
37Indeed, applying the methodology described below on the indicator for investment in farm equipment
yields a pseudo-R squared of 2% with no variables showing statistical significance.
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
compared to 2000, and households were counted as having invested in a family business
if they owned a business in 2000 but did not in 199838. Second, the surveys ask detailed
questions regarding the employment history of each member of the household, including a
breakdown of time spent “on the family land or at home.” If any member of the household
reported having spent any time working at a family store or creating some product or ser
vice as a business, the household was counted as owning a family business. Both methods
yielded the same results.
Households may instead elect to postpone consumption by simply putting money into
a savings account. Historically, the use of banking services in El Salvador has been low,
especially in rural areas, due primarily to a lack of access, but also to a modicum of popular
mistrust of financial institutions. However, in the post-war era of institutional reform, bank
liberalizations and expanded coverage there have been concerted efforts by the Salvado
ran government to educate and encourage people to begin a savings plan. These efforts
have met with some success; in 1975, less than 10% of Salvadorans held their savings at
formal banking institutions, whereas the 2000 Rural Household Survey shows that 18.8%
of rural households held savings accounts, an increase of 4.7% from 1998. Unfortunately,
the data does not provide information on the volume of money put into savings, so while
it is possible to determine which households opened savings accounts between 1998 and
2000, it is not possible to ascertain which households intensified or diminished their saving
38The survey does not ask how many businesses the household owns, thus it is not possible to ascertain
whether a business-owning household in 1998 invested in an additional business.
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
practices. This renders ambiguous the interpretation of the savings indicator as participa
tion in investment; further, the moderately strong correlation of savings with other forms
of investment - particularly migration and a family business - may indicate that savings
accounts merely serve as temporary storage for funds with other ultimate uses. For these
reasons, savings is also excluded from subsequent analyses.
Particularly in the context of agricultural households, the investment and old-age se
curity motives for having children may be relevant. This may be especially true given the
labor-intensive nature of agricultural production in El Salvador. Investment in children
by a particular household was determined using the age profiles of household members in
2000. A household was designated as having invested in children if there were any children
aged two years or less at the time of the 2000 survey. Due to occasional inconsistencies in
the original data, the generated binary variable was cross-checked against the age profile
of household members in 1998 particularly to ensure that children identified as new addi
tions to the household in 2000 were indeed not present in 1998. Using this method, it is
found that 16.8% of households chose to have children during the period.
Finally, households may seek to invest in human capital via education. The Rural House
hold survey contains information on the school enrollment status of each member of the
household and the years of schooling completed. A household was counted as having in
vested in human capital if any member of the household was enrolled in school between
1998 and 2000. Measured as such, 86.3% of the households in the sample invested in hu
man capital. Despite a massive and largely successful effort to expand primary education
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
enrollment in El Salvador, continuation to higher levels of education remains low, so the
analysis of investment in human capital should take into account the level of education
sought. Households are then categorized as having invested in specific levels of human
capital according to the highest grade completed by the most educated household member
enrolled in school between 1998 and 2000. The education groupings are 1) elementary
(1-6 years completed), 2) junior high (7-9 years), 3) high school (10-12 years) and 4) post
secondary (more than 12 years). The frequency distributions of these groups suggests that
56.2% ofhouseholds invested in elementary education, 18.0% in junior high, 9.1% in high
school and 3.0% in post-secondary. The large differences between each level of human
capital investment suggest that it is more appropriate to analyze level-specific investment
rather than the inclusive measure. Further, due to the multiplicity of categories, the re
gression results and analysis of the human capital indicators are presented in Section 5.2.4
separately from the other forms of investment.
Manipulation of the data as stated above yields indicator variables for each form of in
vestment. Incorporating all variables, a broad indicator variable is created to denote house
holds which undertook any form of investment at all between 1998 and 2000. This vari
able provides a point of departure for empirical analysis: What factors influence whether
the household chooses to invest at all? Perhaps more importantly, do these factors manifest
differential effects on the various individual investment types?
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.1.2 Methodology
The empirical methodology focuses on a comparative approach and a broad-based treat
ment of investment. A core set of regressors is chosen to highlight the variation in four
general aspects of the household: the characteristics of the household and its decision mak
ers, the characteristics of potential migrants, measures of household income, and measures
of household wealth. This set of regressors is utilized to explain the overall likelihood of
investment as well as the disaggregated individual measures of investment. This allows a
comparative interpretation across the spectrum of investment options which constitutes an
important improvement over previous studies. Rather than focusing exclusively on inter
national migration, the present framework provides a richer setup whereby the motivations
to participate in certain forms of investment represent disincentives for other investment
types. For example, anecdotal evidence and previous empirical work often suggests that
migration may be especially common among lower income groups, but it is not known
whether the investment profile for lower income households is otherwise substantially dif
ferent. Similarly, female-headed households are also believed to invest disproportionately
in migration, but their other investment activities are not discussed; further, empirical anal
yses do not typically go so far as to distinguish the pure effect of being a female-headed
household from other possibly correlated fundamental factors. The present approach in
stead incorporates relevant social, demographic and economic characteristics, as well as
more subtle interactions into a framework that allows meaningful comparison across in
vestment types.
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In order to properly extract the profile of investing households during the period, a care
ful data arrangement is necessary. In particular, while the investment status of a house
hold is determined at the end of the period, the relevant factors influencing that decision
must have been those which existed ex ante. Thus, for the purpose of testing, the house
hold’s socio-demographic and economic data are taken from the 1998 Rural Household
survey, whereas household investment information is extracted from the 2000 survey as
described above. With the dataset thus constructed, logit regressions are performed ap
plying the same set of regressors to the constructed binary investment variable, as well
as the six individual indicator variables for each type of investment. The results of these
regressions are then suggestive of the direction and strength of the effects of explanatory
variables on the probability of household participation in various investment opportunities.
5.1.3 Explanatory Variables
As suggested by the model in Chapter 4, key differences in the composition and economic
status of the household should be the critical determinants of household investment and the
form it takes. The significant variation in the types of investments, suggests that the ex
planatory variables should also manifest substantially different effects by investment type.
Since comparison is a key aspect of the analysis, any of the explanatory variables, or their
interactions, described in this section which are expected to have an a priori effect on a
particular form of investment are equally applied to all investment types.
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A fundamental aspect of the migration-as-investment approach is the crucial role of
the household in the decision process. This does not suggest, however, that all individu
als within the household have equally strong voices, and for the purposes of this study, it
is assumed that the characteristics of the head of household are those which are the most
influential in the decision process. The included variables are those most likely to impact
the investment decisions of the household: the age, sex and education level of the head of
household. Older heads of household may perceive a shorter time horizon and may tend
to reduce investment activities, whereas younger heads of household may seek to build
wealth and consequently exhibit more intensive investment activity. Also, to account for
non-linearities in the age relationship, the square of the age of the head of household is
included among the explanatory variables. While not an explicit life-cycle measure, the
squared age brings out differences at the extremes of the distribution which may be masked
by the overall pattern.
The sex of the head of household is likely to affect the probability of participation in
specific investments, although there is no a priori reason to believe that the effect on the
overall probability of investment is significant. Specifically, as suggested above, female
headed households are less likely to invest in land purchases and should similarly be ex
pected to exhibit lower fertility.
The dependency ratio is calculated as the percentage of the household comprised of
individuals less than 15 years of age and over 65. These age breaks, particularly the lower,
are imposed by the data, but it is clear that in several cases individuals noted as dependents
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are significant contributors to the household. The potential effect of the dependency ratio
on household investment is ambiguous. While the presence of dependents certainly gives
incentive to invest, meeting the daily needs of dependents may not afford the household
the ability to spare funds for investment.
The indicator for households in the Eastern region is also included. As indicated above,
Eastern households tend to migrate more, but generally invest less. The inclusion of this
variable provides a way to verify whether the east indicator is itself important or is only a
proxy for other forms of household variation.
The model of Chapter 4 suggests that the profile of the constituents of the household ex
erts great influence on the household investment decision. Households with working-age
and educated members should experience both cost and efficiency advantages over older
or dependent-heavy households such that a wider array of investments would be feasible
and potentially more lucrative. This seems especially true in the case of migration where,
once the initial investment has occurred, the rate of return to the household depends almost
entirely on the migrant and is largely beyond the control of the originating household. For
simplicity and to expand on the migration example, the present strategy identifies for each
household the single non-head of household individual most likely to contribute to the suc
cess of the household’s investments and includes the demographic characteristics of that
person as explanatory variables. For lack of a better term, these individuals shall be re
ferred to as “investment participants.”
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The pool of investment participants in 1998 is generated via a selection algorithm. To
identify the most likely investment participant, a systematic selection process is applied
to the members of each household on the basis of age and relation to the head of house
hold. It is assumed for simplicity that only individuals from three relation categories may
be counted as investment participants and are selected in a specific order: 1) children of
the head of household or spouse, 2) siblings of the head of household or spouse, 3) the
spouse of the head of household. The selection order is based on the three highest mi
grant frequencies from the combined 1996 and 1998 data. Other relatives in the household
(cousins, nephews, grandparents, etc.) are not counted as investment participants since
neither the data nor any a priori information allows a consistent ranking of individuals be
yond the three categories listed above. Children under 15 years of age are eliminated as
well, and the only adults over 50 counted as investment participants are spouses of the head
of household. The role of the head of household in the success of household investments is
surely important; however, the characteristics of the head of household are included for all
households, and the head of household is eliminated from the investment participant selec
tion algorithm to avoid both logical and econometric problems of redundancy. Further, the
strategy behind the identification of the investment participant is to glimpse at the composi
tion of the household beyond the head of household; a selection algorithm which included
the head of household would thus inherently defeat the explicit purpose. It is particularly
significant to note that, among the 187 households that invested in migration during the
period, in only 26 cases did the selection algorithm fail to identify the actual migrant.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Having selected the pool of investment participants, the relevant demographic char
acteristics are extracted: age, education and sex. Working-age individuals should be ex
pected to correlate strongly with the overall probability of investment. The simple cor
relation coefficient of 0.189 between the education of the investment participant and the
probability of international migration provides further support for Funkhouser’s evidence
of a migration brain drain out of El Salvador. More significantly, the 0.757 correlation
between the indicator for a male investment participant and international migration is in
dicative of both the historical trend of Salvadoran migration and a strong expected result
in the regressions presented below.
The number of international migrants reported by the household in 1998 is also in
cluded as an explanatory variable. A negative correlation between the number of migrants
and future household investment in migration would suggest a form of diminishing returns
to migration investment. A positive correlation would instead imply that the cost of migra
tion is reduced or the probability of successful migration is substantially improved when
the household has previous migration experience. In particular, it may be the case that
growing social and kin networks in the host country encourage further out-migration.
Since Table 5.3 suggests some variability between the sources of household income
and other aspects of the household, indicators for agricultural landowners and agricultural
workers are included. In addition, several interactions are included to account for potential
correlations in the data. The agricultural worker indicator is interacted with the female
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Variable Units All All
199?
Intl. Mig. No Intl. Mig. 2000 West East Male HOH Fem. HOH
Household
Age HOH yrs. 50.09 49.15 53.26 48.27 51.04 49.23 51.55 50.08 50.20
Educ. HOH yrs. 2.82 2.74 2.39 2.81 2.90 3.30 2.00 2.91 2.21
Educ. HH yrs. 2.57 1.82 1.81 1.82 3.31 2.79 2.18 2.51 2.90
HH Size per s. 6.03 6.10 5.82 6.16 5.96 5.98 6.11 6.03 5.99
Fem. HOH pet. 13.83 12.82 19.05 11.48 14.84 15.42 11.14
Depend. Ratio pet. 17.80 4.37 3.52 4.55 31.24 17.31 18.64 17.88 17.37
Eastern HH pet. 37.10 37.27 57.14 32.99 36.93 38.26 29.88
Literacy pet. 64.05 63.42 65.97 62.88 64.68 67.78 57.74 63.99 64.45
Relative Income 1.00 1.11 1.00 1.14 0.89 1.15 0.74 1.02 0.86
Primary HH Income Sources
Ag. Landowner pet. 31.20 31.20 42.86 28.69 31.20 24.93 41.82 31.41 29.88
Ag. Worker pet. 26.31 26.31 15.24 28.69 26.31 26.81 25.45 27.40 19.51
Non-Ag. pet. 33.73 33.73 35.24 33.40 33.73 38.34 25.91 32.19 43.29
Mixed pet. 8.77 8.77 6.67 9.22 8.77 9.92 6.82 9.00 7.32
Table 5.7: Summary Statistics, Rural Household Surveys.
vo
VO
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Variable Units All All
1998
Intl. Mig. No Intl. Mig. 2000 West East Male HOH Fem. HOH
Migrant and Remittances
Mig. Age yrs. 31.42 25.48 24.85 25.93 32.91 32.66 29.89 31.32 31.85
Mig. Educ. yrs. 4.48 3.81 4.50 3.39 4.65 4.95 3.89 4.45 4.61
Mig. Male pet. 25.55 20.24 84.76 6.35 30.86 20.78 33.64 23.78 36.59
Mig. W/Children pet. 20.57 20.57 29.52 18.65 20.57 18.77 23.64 19.86 25.00
Mig. Married pet. 12.48 12.48 15.24 11.89 12.48 9.92 16.82 12.33 13.41
Mig. Elasped Time yrs. 2.95 1.97 7.47 0.79 3.94 2.74 3.32 2.84 3.65
Num. Migs. per s. 0.43 0.23 1.00 0.06 0.62 0.30 0.64 0.41 0.53
HH Remittances pet. 14.04 4.13 13.81 2.05 23.95 11.19 18.86 12.67 22.56
Investments
Mig. Intl.
Mig. El Sal.
pet.
pet.
24.62
8.35
17.71
9.78 7.62 10.25
31.53
6.91
18.90
7.64
34.32
9.55
22.99
8.32
34.76
8.54
Land Purchas pet. 3.12 3.04 2.86 3.07 3.20 2.82 3.64 3.23 2.44
Open Store pet. 17.37 15.35 23.81 13.52 19.39 19.03 14.55 16.83 20.73
N 1186 593 105 488 593 746 440 1022 164
Table 5.7 continued.
o
o
head of household indicator to correct for the negative correlation. Also, since a non
zero value for the landowner indicator implies past investment in land, an interaction term
between the landowner indicator and the land size variable (discussed below) brings out
whether large landowning households are more likely to invest - in land or otherwise. To
account for strong correlations in the data, an interaction between the age of the head of
household and the agricultural landowner indicator is added. While the simple correla
tion coefficient of 0.263 could be attributed to a longer time span of wealth accumulation
by older heads of household, another explanation seems more likely. Since the agricul
tural owner, worker, etc. categories are assigned within the data according to the relative
shares of income sources within the household, the positive correlation between age and
the agricultural landowner category is more likely to represent older heads of household
which have remained strictly as farmers and have not chosen - or have not been able to -
diversify their sources of income.
The Rural Household surveys allow the household’s financial status to be discussed
as both income and wealth. As previously stated, households were ranked into quintiles
by per capita household income, and these income quintile indicators are utilized to de
scribe the effects of income on the probability of investment. This method is used in par
ticular as an alternative to log-transformed per capita household income in order to more
pointedly bring out the differences in income among Salvadoran households. The specific
advantage of this method is that it allows direct comparison of the investment patterns of
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
distinct groups of households rather than only being able to suggest the strength and direc
tion of the influence of income on investment. Further, since the coefficients of the logit
regressions used in the analysis are not conducive, for example, to use in the calculation of
the income elasticity of investment, the categorization of households into income quintiles
results in no loss of analytical power or flexibility. Finally, as a check, the regressions pre
sented below were also run using log-transformed income and resulted in generally weak
coefficients of the appropriate sign with no appreciable change in the overall power of the
model.
As a small test of the relative deprivation motivation for investment, a relative income
variable is generated and included in the regressions. Among the basic household data are
the departamento (state) and canton (town) where the household resides. Two relative in
come variables were created as the percentage of per capita household income relative to
mean per capita household income within the departamento or canton. These variables
bring out a local aspect of household influences which goes beyond the income categories
and explicitly ackowledges the peer-group or community effects on investment activity
and choice. The relative income variable calculated by departamento should be expected
to have greater explanatory power since the relatively small number of observations for
cantones is likely to create irregular variation in the corresponding relative income calcu
lation.
The size of the household’s land holdings is used as the primary measure of household
wealth. The potential for land to serve as collateral in credit markets suggests that greater
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
land holdings may facilitate access to credit and prompt such households to further partic
ipate in credit-intensive investments relative to households with little or no land. The size
of the land holdings is also included, as stated above, for use in interaction terms with the
agricultural landowner indicator. A second measure of household wealth is an indicator
for whether any member of the household had a savings account in 1998. Unfortunately,
the data contain neither account balances nor the frequency of transactions into or out of
the account, and as such, provides a questionable measure of household wealth. Nonethe
less, having a savings account is indicative of a previous postponement of consumption
and, perhaps, the intention to use those funds for other investment purposes39.
5.2 Investment: Results
Estimation results are summarized in Table 5.8 which provides predicted participation prob
abilities for each type of investment and various key subgroups of the sample40. The pre
dictions are generated by applying the mean values of the explanatory variables to the co
efficient vector and individually replacing selected binary variables with 0 or 1 (as appro
priate) to highlight the effect of the particular variable41. The inner product of the data
39This variable is expected to be most useful in regressions extracting the determinants of migrant financ
ing, discussed in Section 5.3.
40Complete regression results are provided in Appendix A.
41 For multi-category variables, such as income, the procedure requires selecting with 1 the desired cate
gory and setting to 0 all other categories. Note that setting all income categories to 0 implicitly selects Income
3, the omitted category.
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and coefficient vectors is then applied to the logit conversion procedure to obtain predicted
probabilities42.
As a benchmark, probabilities are first calculated utilizing the mean values of all the ex
planatory variables, and it is found that, on average, Salvadoran households have a 53.9%
probability of engaging in some form of investment. The bulk of investment activity is
concentrated in elementary education and international migration, while investment in post
secondary education, land and internal migration have the lowest likelihoods.
Isolating sub-groups of the population brings out clear differences in estimated partici
pation probabilities. Households with male potential migrants are particularly noteworthy.
Such households have a nearly 80% likelihood of investing, and a 45.6% likelihood of in
vesting in international migration, 3.6 times the average estimated probability. It is also
42Letting X ' (3 represent the inner product of the data and coefficient vectors, logit probabilities are esti-
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
worth noting that, compared to the average probability of 7.4%, households with male po
tential migrants are nearly twice as likely to open a store or shop.
Low income households also exhibit notable predicted investment patterns. The 14.6%
greater than average probability of investment for low income households appears to be
motivated by international migration where low income households are 4.7 times more
likely than average to invest. While all households have a 50% probability of investing
in elementary education, the estimated probability rises to 74.7% for low income house
holds. Indeed, at all education levels through secondary, low income households have
above-average estimated participation probabilities. The estimated probability of having
children is also 4.7% higher for low income households as compared to the 6.6% average.
Three other aspects of the results related to income categories merit mention. First,
households in the fourth income quintile exhibit less than a 3% probability of investment
in international migration. This is less than one tenth of the predicted participation rates
for the other income categories. While it is tempting to attribute this seemingly anoma
lous result to problems of sample size or sample selection, it should be noted that other
predicted investment participation rates for Income 4 households are not markedly differ
ent from other income categories. Second, upper income households are twice as likely to
open a store or shop as the 7.4% average. Third, upper income households are predicted to
have an 18.4% probability of investing in high school education, over 5 times the average.
This persistence in education investment carries over to the post-secondary level where
106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
upper income households exhibit a 2.6% predicted probability of participation, whereas
the average is nearly zero43.
Highlighting other aspects of the household brings out more moderate degrees of varia
tion. Female-headed households are nearly 4% more likely than average to open a store or
shop, but nearly 3% less likely than average to invest in international migration. Also, not
surprisingly, female-headed households have only a 1% estimated probability of investing
in additional children. Interestingly, female-headed households have, at 2.7%, the high
est predicted post-secondary education investment probabilities among the groups high
lighted. Eastern households do not exhibit predicted investment patterns that are remark
ably different from the average.
In general, the differences in the magnitudes of the estimated participation probabili
ties for the various types of investment are striking. The growing accessibility and rela
tively low cost of early education make it the most common expected form of investment.
However, especially for certain sub-groups, international migration is a close second, thus
emphasizing its importance in the Salvadoran household investment portfolio. Equally no
table is the lack of predicted investment in land or post-secondary education.
5.2.1 Discussion
Taken together, the broad picture of Salvadoran investment reveals strong patterns. While
simple frequencies show that international migration and elementary education are the most
43The average predicted probability of investing in post-secondary education is 0.0002%.
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
widespread forms of investment, the regression analyses reveal that participation in differ
ent forms of investment is affected by specific characteristics of the household. The most
significant finding is that the presence of young men provides powerful incentive for the
household to invest in international migration. This finding is not only robust to the inclu
sion of controls for the demographic makeup and economic position of the household, but
transcends them such that the characteristics of the potential migrant provide nearly all the
explanatory power in the model.
The results also support the implications of the model in Chapter 4 to the extent that it
was hypothesized that the migrant quality indicators would play a crucial role in determin
ing household investment decisions. In particular, it can be inferred that investing in the
migration of young males can be expected to yield a higher net rate of return than other
forms of investment due either to higher expected gross return through remittances or a
lower participation constraint. Note that, in this case, the nature of the expectation on re
turns is important. Since the very existence of an entry constraint suggests that returns
are conditional on entry, it must be the case that the ex ante expected value of remittance
income is higher for households with young male potential migrants. This is to be dis
tinguished from an ex post expectation of remittances given that migration has occurred.
This means that male and female migrants may exhibit no significant difference in their
actual remittance patterns, but that male migrants may nonetheless be favored due to a pre
conceived expectation of a higher probability of successful migration at similar or lower
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
entry costs. These issues are further elaborated in Sections 5.3 and 5.4 which respectively
address the determinants of migrant finance and remittances.
In several cases, the interpretation of the regression results differs from the findings of
other studies. For example, female-headed households are found to be less likely to in
vest in international migration, whereas earlier case studies suggest the importance of the
migration option to many such households. The results of the present analysis, however,
provide a richer picture by employing a larger and more diverse sample than most case
studies and including a more complete set of controls. The results of the analysis then sug
gests that the indicator for female-headed households functioned in other studies as a proxy
for key aspects of the household, such as lower household income and higher dependency
ratios, which, when controlled for, allow the true effect of the female-headed household
variable to emerge. Similarly, the relative income variable reveals a strong positive asso
ciation with international migration, implying that households identified as relatively high
income within their immediate geographical area are more likely to invest. However, work
by Stark and Massey instead suggests that relatively low income households may be more
likely to invest - particularly in migration - in order to boost their relative standing in the
community as well as their absolute living level. Again, the more complete set of con
trols and the inclusion of key interaction terms more accurately portray the relations in the
data. In particular, the standard relative deprivation effect is verified for the upper income
quintile, suggesting that relative status considerations motivate high income households
to invest in migration; however, isolating that group, the overall effect of relative income
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
suggests that those most able to invest are also more likely to do so. Indeed, the results
suggest that relative deprivation may play a larger role in other forms of investment, such
as land or store ownership.
Fundamental differences are also found to exist between international and internal mi
gration. Other than the age of the potential migrant, the two forms of migration share no
other statistically significant relations to the set of explanatory variables. Further, many
variables exert opposite influences on internal versus international migration. For exam
ple, wealthier households (measured by either land size or the education of the head of
household) tend to migrate internally and less frequently invest in international migration.
Conversely, dependent-heavy households prefer international migration, as do households
with greater numbers of migrants. Also, male potential migrants are not significantly more
likely to journey to the capital, whereas the male indicator yields a very strong result for
international migration. Taken together, these differences highlight the dissimilar motiva
tions behind each type of migration. While the process of international migration is riskier
and more costly than internal migration, the potential returns are also substantially higher.
Combined with the large explanatory power provided by the characteristics of the poten
tial migrant, the results provide strong support for the idea that international migration is
undertaken as a form of household investment with forethought as to which member of
the household could provide the greatest return at the lowest cost. On the other hand, the
investment motivations behind internal migration are less distinct. Other than age, it is
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
not clear that the characteristics of the potential migrant affect the decision to migrate in
ternally; for example, there is no indication that more educated individuals may seek to
migrate to the capital to take advantage of job opportunities or pursue additional educa
tion. Further, the comparatively low cost of internal migration does not seem to play an
important role since low income households - in both absolute and relative terms - are less
likely to migrate internally. Indeed, there seems to be compelling evidence to suggest that
internal migration may occur for reasons only loosely linked to investment purposes.
The subsequent sections of this chapter focus on a more detailed decomposition of in
ternational migration as investment. Since the results presented thus far integrate elements
of both the costs and benefits of the various forms of investment, it must be the case that
subtle yet important facets of the migration decision are cached. The data do not provide
actual migration cost information, so Section 5.3 offers an indirect analysis of cost-related
aspects of migration by studying the sources of migrant finance. Section 5.4 subsequently
addresses remittances, the benefits of migration.
5.3 Financing Migration
Salvadoran newspaper reports suggest that some hopeful migrants often pay up to $5,000
for the service of being smuggled into the United States. Many others instead find their
way to the Mexican border with the United States and pay smugglers as little as $100 to
be taken across. This points not only to the variety of practical forms which investment
in migration can take, but also to a need to decompose the sources of investment finance
111
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
at the household level. While the analysis in Section 5.2 provides a broad overview of
the household investment decision, the next steps require a more focused approach on the
specific costs and benefits of investment44. Specifically, is investment financed by different
means, and to what degree are the factors that affect investment choice the same as those
which affect the financing option?
The Rural Household surveys unfortunately lack most investment cost and financing
data. Startup costs for new shops are not known, and households do not report the explicit
cost of sending a migrant. Indeed, only two pieces of investment cost or financing infor
mation are reported. First, households which purchased land do report the price paid, but
the sources of finance are not listed. Also, in terms of empirical verifiability, the number
of households investing in land during the period in question is sufficiently small so as
to generate very limited confidence levels in any statistical analyses. Second, households
which invested in international migration are asked whether the household provided any
assistance to the migrant and, if so, from what source45. This information provides the
basis for the analysis in this section: how do Salvadoran households finance international
migration, and what factors influence the type of financing chosen?
In general, the results of this analysis enrich the scenario presented in Section 5.2 by
exposing the nuances of the broad investment decision. While the characteristics of the
migrant remain important determinants of the method of finance, income differences, as
44The benefits of investment in migration - remittances - are discussed in Section 5.4.
45Neither the amount of investment nor the relevant interest rate are provided.
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
expected, play a markedly more significant role than was apparent in the selection of in
vestment type.
5.3.1 Data and Methodology
Respondents in the 2000 Rural Household survey were asked if any international migrant
sent since 1998 received ‘help’ from the non-migrating household in order to migrate. Al
though the type of help is not specified to be monetary, the subsequent question asks those
who responded affirmatively to the first question to name the sources of funds for the help
provided. However, even though the questionnaire allows respondents to name multiple
sources, only one source is provided in the data, so it must be assumed that the source listed
indicates only the primary source. The questionnaire provides twelve different response
categories for the sources of migrant finance, but, for the purpose of this analysis, the cat
egories are combined into four groups, and binary indicator variables are constructed for
each group. This is done to avoid conceptual redundancies in similar categories and to
elevate the number of observations in each category to statistically meaningful levels.
These four indicator variables comprise the dependent variables in the regressions be
low and are constructed as follows. The first is an indicator for households which bor
rowed to finance migration. Since migration is not acknowledged as a legitimate form of
investment by formal lenders, borrowers obtain informal loans from local money lenders,
employers, family members or even from the coyote (smuggler) himself. The data unfor
tunately provide neither the amount of the loan received nor the interest rate at which it
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
was to be repaid. Only thirty of the 187 households which invested in migration borrowed
in order to do so.
A second indicator groups households which received funds for migration from extra
household family members residing either within El Salvador or internationally. S uch funds
are assumed to be strictly transfers or remittances and not loans, and as such, only repre
sent help or sacrifice on the part of the non-migrating household in the opportunity cost of
not having used the funds for alternative purposes. Of the 57 such households, 41 received
their migration funds from abroad. It should be noted that there is no indication that the
benefactor living abroad and providing the funds for additional migration is considered a
member of the household. Demographic information (such as the relation to the head of
household) is not provided, and it is similarly unknown whether the household helped fi
nance the migrant already abroad. While it is thus tempting to interpret the evidence of
migrant financing from abroad as a source of intra-household migration cascade whereby
existing migrants encourage and facilitate further migration, such a conclusion is not di
rectly supported by the available evidence. Nonetheless, remittance income earmarked to
finance migration is suggestive of a potentially important investment-related use of income
which is typically thought to be used almost exclusively for immediate consumption.
The third indicator flags households which depleted their wealth by withdrawing sav
ings or selling assets to finance migration. Ideally, the data would allow separate analyses
of those who withdrew savings from those who sold assets; however, only 12 households
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
indicate having sold assets and were thus added to the most conceptually proximal cate
gory to avoid problems of sample size. Nonetheless, the information from those house
holds does suggest that this variable may reflect the combination of two fundamentally
different sample populations since 75% of the households which sold assets are in the two
lowest income quintiles as opposed to 44.7% of households which depleted savings.
A fourth variable serves as an indicator for households which sent a migrant interna
tionally but did not provide any financing. It is assumed that such migrants were self-
financed and that their migration did not impose an explicit cost on the remaining house
hold.
The empirical methodology is an extension of Section 5.2. Each of the four migrant
finance indicators are used as dependent variables in logit regressions utilizing the core set
of regressors described in Section 5.1.3. This again allows a comparative analysis which
enriches the results of Section 5.2 by bringing out the subtleties which may be masked by
the broad investment decision. Since the sample population is limited to those households
which invested in migration, conditional probability estimates similar to those presented
in Appendix E are not required, and it is, in this case, appropriate to simply focus on the
choice of migrant finance.
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.3.2 Results
The key results of the migrant finance regressions are presented in Table 5.9 as estimated
probabilities by subgroups of the population46. In general, the results are suggestive of
potentially dichotomous interpretations. First, the importance of extra-household family
financing for migration points to a long-term investment pattern centered on migration. Al
though the data cannot provide direct evidence, the role of extra-household (and primarily
international) family financing makes it likely that early investment in migration may be
undertaken with the expectation of generating funds for further migration from the house
hold. Second, the strong tendency for certain subgroups to not finance the migrants poten
tially detracts from the household investment concept. Although the absence of explicit
financing from the sending household is not at all inconsistent with the model of Chapter
4, it does lend credence to an individualistic model of investment in migration whereby
migrants leave of their own accord to further their individual goals.
On average, there is only an 8.7% probability that a migrant would be directly financed
by the sending household via dissaving or the sale of assets. By comparison, lower income
households are nearly 6% less likely than average to directly finance the migrant, whereas
upper income households and agricultural landowners are especially likely to directly fi
nance migration. Eastern and female-headed households also exhibit a below-average like
lihood of dissaving to finance migration.
46See Section 5.2 for details of the estimation procedure.
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Categories
HH
Save/Sell
Extra-HH
Family Borrow None
All Means 8.70 32.96 1.78 13.52
Male Mig. 8.46 24.67 1.69 28.62
East 7.34 42.56 2.58 8.05
Female HOH 6.14 27.22 1.92 18.22
Ag. Landowner 11.77 35.67 1.99 6.47
Ag. Workers 4.78 38.59 1.93 18.19
Income 1 (low) 2.56 31.96 0.31 81.53
Income 2 4.47 20.46 1.17 70.65
Income 3 3.69 35.64 7.22 22.33
Income 4 8.15 51.87 2.30 16.37
Income 5 (high) 8.88 48.35 0.00 38.52
Table 5.9: Predicted probabilities, Migrant Finance.
Nearly one-third of all migrants are expected to receive funding from extra-household
family members. This source of funding is not only the most common, but also proves to
be the least variable across the subgroups identified. Perhaps surprisingly, migrants from
upper income households exhibit nearly a 50% probability of receiving extra-household
funds for migration, as opposed to lower income (Income 2) households which are 12.5%
below the average. Eastern migrants are also 9.5% more likely than average to receive
funding from outside the household.
Borrowing is by far the least likely investment option with an expected average prob
ability of less than 2%. The pattern by income level is noteworthy: neither low nor high
income households borrow to finance migration, but median income households display,
at 7.2%, the greatest likelihood of borrowing.
117
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The profile of households expected to not finance migration at all exhibits the greatest
variability and is, for certain groups, the most common option. While the average pre
dicted probability of not financing migration is only 13.5%, it is found that low income
households have over an 80% likelihood of not financing. It should also be noted that male
migrants are over twice as likely than average to not be financed, while Eastern households
and agricultural landowners are more likely to provide some form of financing.
5.3.3 Discussion
Isolating the finance aspect of the investment in migration allows more subtle analysis.
In particular, the results suggest that international migration is undertaken by two main
groups of households. The first is comprised of wealthy and relatively educated house
holds whose migrants are predominantly male and well-educated. These households do
not go into debt to finance the migrant and rarely dissave; instead, the migrant tends to be
self-financed. This may indicate that the concept of migration as investment is less appli
cable to wealthy households which may have greater access to a wider array of investment
options. Evidenced migration from this group of households might thus be better framed
as a form of individual investment rather than a household-level decision.
The second group of households investing in international migration instead seem to
represent a dynamic and motivated element of the lower income quintiles. These are house
holds which had previously made relatively large investments in the human capital of the
young men sent as migrants and for whom the absence of the migrant from the household is
118
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
likely to impose a substantial opportunity cost in terms of foregone income. This provides
some support for a form of brain drain documented in a case study by Lungo et al. (1997)
where teenagers seek to further their education - emphasizing english language studies -
with the explicit purpose of improving the likelihood of successful migration in the fu
ture. Recall that the human capital regressions also showed a strong tendency for lower
income households to invest in post-secondary education. Unfortunately, the data do not
allow long-term panel analyses which could track a possible correlation between early in
vestment in human capital and subsequent investment in migration which would serve to
verify whether the contemporaneous association between human capital investment and
migration could be extrapolated to imply a time trend as well. Such a result would be in
dicative of an investment portfolio with a large temporal separation between the initial in
vestment cost and the eventual realization of returns. As an additional element of risk, the
borrowing constraint appears to be binding for these households, and the migrants clearly
tend to be self-financed.
In general, the results point to the importance of the characteristics of the potential mi
grant in the household investment decision, and it seems clear that households select mi
grants by reconciling cost considerations with the probability of successful migration. The
remaining question is to ask whether the investment paid off. What factors affect the prob
ability if receiving remittances? Are households which borrowed or sold assets to finance
the migrant more likely to receive remittance income?
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.4 Remittances
Although remittances are the most visible and frequently discussed aspect of Salvadoran
investment in migration, they represent only the final segment in the investment process.
Indeed, the results presented thus far have exposed the subtleties of the household invest
ment decision and the interplay of household characteristics which determine the ultimate
decision. In a sense, study of the determinants of observed remittances are then ex post
analyses whose results are suggestive of ex ante expectations but not directly part of the
actual decision taken. Nonetheless, the return on investment provided by remittances must
have been at the core of the initial investment decision, and the decomposition of the rel
evant effects provides at least a preliminary glimpse at the success of the household’s in
vestment.
The regressions in this section explore the relationship between the core set of explana
tory variables and the probability of receiving remittances. The dependent variable simply
indicates whether or not the household receives remittances from a migrant living abroad,
and for the purposes of testing, the regression is performed on two subsets of the sample
population. The first is similar to the investment analyses where the remittance indicator
in 2000 is regressed against 1998 data for all 593 households in the sample. This acknowl
edges that, despite no investment in migration between 1998 and 2000, many households
invested previously and may receive remittances. Since households may have multiple
migrants, the remittance histories of all migrants in the household are combined such that
the dependent variable indicates whether the household receives remittances from abroad
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
from any migrant. Second, the regression methodology is applied to the sub-group of 187
households which did invest in migration. Note that in this regression the dependent vari
able indicates whether the particular migrant sent since 1998 remits to the household. This
specifically means that a household which receives remittances from a migrant sent before
1998, but not from the migrant sent since 1998, the household, for the purpose of this re
gression, is labeled as not receiving remittances. Note that if the household reported send
ing multiple migrants between 1998 and 2000, the remittance activity is combined such
that the dependent variable indicates remittances from any new migrants47.
Conceptually, there is no reason to believe that for these households the remittance sta
tus in 2000 is indicative of the actual or long term probability of receiving remittances. In
particular, the return on migration should be expected to involve a time-to-build aspect,
and, other than Montes’ (1989) study, there is no evidence to suggest that two years is an
appropriate amount of time within which to expect returns48. Nonetheless, there is a sub
stantial degree of consistency between the two regressions, suggesting that the remittance
experiences of more recent investors do not differ dramatically from those which invested
previously.
These regressions also include indicators for the method of finance used to send the mi
grant. It would perhaps be most correct to estimate these regressions as joint conditional
probabilities49 which would allow, for example, analysis of the probability of receiving
470nly nine households reported sending multiple migrants.
48Montes provides several examples which indicate that Salvadoran households may, via remittances, re
cover their initial investment costs within the first 16 months after migration.
49See Appendix E for an example of joint conditional probability estimates applied to investment choice.
121
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
remittances conditional on the probability of employing a particular method of migrant fi
nance. Such regressions were run, but are not reported here; the multiplicity of methods
of finance requires a separate joint conditional probability estimate for each type yielding
copious amounts of output which, in the end, do not provide any insight or interpretations
that cannot be derived by the simple addition of indicator variables in the original regres
sion.
For the sake of exposition, Table 5.10 presents the predicted probabilities of receiv
ing remittances by subgroups of the sample population. It should be noted that the sub
stantially different magnitudes of the predicted probabilities between subsamples of the
population is due to conceptually different situations being analyzed. The full sample re
gression tests the relationship between the characteristics of the household in 1998 and the
household’s remittance status in 2000; however, many households may have invested in
migration well prior to 1998, and there is not necessarily any reason to believe that the
1998 household indicators would bear strong relations to the probability of receiving re
mittances in 2000. The finding that on average only 4% of households would be expected
to receive remittances points to the relatively low credibility of the results of that regres
sion.
The results from the sub-sample of households which invested in migration since 1998
are thus considerably more relevant since the explanatory variables relate directly to the
time at which the investment decision was made. Also, by selecting households which are
known to have invested, the results indicate only the probability of receiving remittances
122
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Categories
Full
Sample
Migrant
since 1998
All Means 3.97 36.59
Male Mig. 4.51 28.05
East 4.29 41.04
Female HOH 2.91 21.90
Ag. Landowner 7.05 56.68
Ag. Workers 1.57 13.19
Income 1 (low) 11.13 32.77
Income 2 10.40 48.83
Income 3 5.74 24.51
Income 4 2.24 40.97
Income 5 (high) 9.93 35.84
HH Borrow 51.37 69.67
Extra-HH Family 59.63 87.87
No Help 3.74 36.42
N 593 187
Table 5.10: Predicted Probabilities, Remittances.
without simultaneously needing to account for the conditional probability of an initial de
cision to invest at all.
The predicted probabilities suggest that on average over one-third of all households
with recent migrants would be expected to receive remittances. Several deviations from
the average are worth noting. In particular, households with a male migrant exhibit, at
28%, a lower than average probability of receiving remittances. Nearly 57% of agricultural
landowner households are expected to receive remittances, as opposed to only 13.2% of
agricultural worker households. The variation by income groups is relatively small, but
skewed toward the lower income groups such that Income 2 households have nearly a 50%
likelihood of receiving remittances.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
These results suggest several interpretations. The lower probability of receiving remit
tances from male migrants, despite the higher incidence of male migrants, suggests that
households may focus their migration investments in males due to a lower ex ante cost
rather than an expectation of higher future gross return. However, it was also found that
male migrants are significantly less likely to have received no funding to migrate; com
bined with the finding of lower remittance probabilities for male migrants, it is also the
case that the individualistic model of migration cannot be rejected for households with
male migrants. The general consistency across income categories further solidifies virtu
ally all empirical research on El Salvador which finds that households across the Salvado
ran socio-economic spectrum receive remittances.
The effects of the migrant finance variables are, however, the most striking. Even in
the full sample regression, these variables stand out since (other than general demographic
data) they provide the only information related to the particular time when the migration
investment was made. For households with recent migrants, having borrowed to finance
migration boosts the probability of receiving remittances to 69.7%, and households whose
migrant was financed by extra-household family members exhibit an 87.9% likelihood.
Households which did not finance the migrant, comprising the majority of the sub-sample,
are indistinguishable from the mean.
The predicted probabilities by method of finance suggest, as expected, that the migra
tion as investment approach is particularly relevant to those households which borrowed to
finance migration. Taking the risk of obtaining a loan for the purpose of migration signals
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the degree to which the expectation of future net gain outweighs that of the other possible
investment options. Also, there is every reason to believe that the social pressure on the
migrant may give particular incentive to remit when the household incurred debt to finance
migration.
The relationships between the other methods of finance and the probabilities of recev-
ing remittances, however, lend less support to the migration as investment approach. In
particular, the finding that extra-household family financing virtually assures remittances
seems to indicate a household approach which employs migration and remittances as a
stepwise methodology for the eventual migration of the entire (or a large proportion of)
the household. While this pattern conceptually falls within the cadre of the migration as
investment approach, the direct interpretation in this case of remittances as return on in
vestment seems questionable.
Indeed, taken together, the remittance regressions suggest that the migration as invest
ment approach is appropriate for specific segments of the population, but it appears difficult
to reject the individualistic model of migration for a significant proportion of the popula
tion. Specifically, the estimated probabilities imply that on average over 60% of house
holds which sent a migrant should not expect to receive remittances. However, a small
qualification should be inserted to acknowledge that these data observations are made only
two years from the time of investment in migration, and that the proportion of households
receiving remittances may, at least for a time, increase as the migrant gains more experi
ence in the host country.
125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 6
Conclusions
The goal of this thesis has been to provide a broad yet more specific treatment of migra
tion processes than has been available in previous work. By framing the option to migrate
in the context of an investment decision, the actions of economic agents may be viewed
as rational and optimal given their demographic and economic situations, as well as the
prevailing macroeconomic and institutional conditions. As part of this framework, pos
tulating the migration option as one of several possible types of investment is essential.
Rather than studying migration in isolation, the investment decision has been addressed
generally and in a way that allows meaningful comparison of the differential motivations
behind various forms of investment.
This work also makes a contribution to the study of Salvadoran investment and migra
tion. It is neither unexpected nor new to find that working age men are the most likely
candidates for migration. However, the interaction of the characteristics of the household
with the various forms of investment is unique; this approach removes migration and re
mittance analysis from the analytical vacuum and places it in a more general investment
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
context which simultaneously allows a more detailed decomposition of the motivations
households face. Further, the isolation of the forms of migrant finance and the probabil
ity of receiving remittance income facilitates the identification of the aspects of household
variation which are essential at each step in the decision.
The results of the empirical analyses help identify the degree to which the migration
as investment approach is both useful and relevant while adding detail and richness to the
picture of Salvadoran migration and remittance patterns. While the available data appear
to support the migration as investment approach in the case of El Salvador, the interpre
tations are not unambiguous, and it is thus not possible to reject entirely the traditional
individualistic model of migration.
In terms of the Salvadoran situation, the empirical results are suggestive of the genre of
policies which may curb out-migration and modify the uses of remittance income. Provid
ing incentives for young men to remain in the country is key. Although primary education
is rapidly approaching universal coverage and the Salvadoran university system has been
fortified, vocational training continues to lag; in the absence of low skill employment or
opportunities for vocational advancement, many young men choose the exit option. Also,
the ability to undertake more traditional forms of investment should be facilitated. This in
volves, in particular, deeper credit markets, as well as expanded efforts to make the avail
able opportunities known to potential farmers and entrepreneurs.
127
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The future possibilities for expanding the work begun in this thesis are very broad. In
tegrative approaches, such as migration as investment, which can incorporate the mutiplic-
ity of options faced by economic agents add both depth and breadth to traditional analyses.
Particularly within the sphere of international migration and remittances, I would seek to
further apply these methods to other countries or regions with migratory patterns differ
ent from those of El Salvador. Specifically, countries with longer histories of sustained
migration and a stronger element of return migration would be conducive to such anal
yses. Cross-country comparative studies are also somewhat lacking in the existing body
of literature. For example, Funkhouser (1995) finds that, despite similar patterns of out
migration, Salvadoran migrants remit over twice as much money per capita compared to
Nicaraguan migrants, but little analysis is provided. The integrated empirical methodol
ogy utilized here would serve to identify the relevant characteristics of the households and
individuals which may be indicative of differential behavioral patterns.
128
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reference List
[1] Richard H. Adams. The economic uses and impact of international remittances in
rural egypt. Economic Development and Cultural Change, 39:695-722, July 1991.
[2] Richard H. Adams. Remittances, investment, and rural asset accumulation in P ak
istan. Economic Development and Cultural Change, 47:155-174, October 1998.
[3] Inter-American Development Bank. Economic and Social Database. Internet:
http://www.iadb.org/int/sta/English/StaWeb/, Accessed March 2, 2000, 1998.
[4] The World Bank. World development indicators cd-rom. 2001.
[5] James K. Boyce. Economic Policy for Building Peace: the Lessons of El Salvador.
Lynne Rienner Publishers, Boulder. CO, 1996.
[6] Carlos Briones. Economfa informal en el gran san Salvador. In Perez Sainz and
R. Menjfvar, editors, Informalidad Urbana en Centroamerica: Entre la Acumulacion
y la Subsistencia. FLASCO/Editorial Nueva Sociedad, 1991.
[7] Carlos Briones and Katharine Andrade Eekhoff. Participacion en los Mercados Lab-
orales de los Residentes en las Areas Rurales: Limitaciones y Desafws. Number 2 in
Documento de Investigation BASIS. BASIS/FUSADES, San Salvador, El Salvador,
2000.
[8] Jeffrey H. Cohen. Transnational migration in rural oaxaca, mexico: Dependency,
development, and the household. American Anthropologist, 103, December 2001.
[9] W. Max Corden and Peter J. Neary. Booming sector and de-industrialisation in a
small open economy. The Economic Journal, 92:825-848, December 1982.
[10] Alejandra Cox-Edwards and Manuelita Ureta. Income transfers and childrens’
schooling: Evidence from El Salvador. October 1999.
[11] James Davies and Ian Wooton. Income inequality and international migration. The
Economic Journal, 102, July 1992.
[12] Republica de El Salvador. Encuesta de hogares de propositos multiples 1995. 1997.
129
with permission of the copyright owner. Further reproduction prohibited without permission.
[13] Banco Central de Reserva. Informe Mensual de la Situacion Economica de El
Salvador, Abril. Banco Central de Reserva, Departamento de Cuentas Macroe
conomicas, San Salvador, El Salvador, 1998.
[14] Dennis de Tray. Herramientas basicas para el analisis de una bonanza petrolera.
Planeacion y Desarrollo: Lecciones de Indonesia, (XXV), May 1994.
[15] Slobodan Djajic. International migration, remittances and welfare in a dependent
economy. Journal of Development Economics, 21:229— 234, 1986.
[16] Sebastian Edwards. Real Exchange Rates, Devaluation, and Adjustment: Exchange
Rate Policy in Developing Countries. The MIT Press, Cambridge, MA, 1989.
[17] I. Elbadawi and R. de Rezende Rocha. Determinants of Expatriate Workers’ Remit
tances in North Africa and Europe. Number 1038 in Policy Research Working Paper.
The World Bank, 1992.
[18] Edward Funkhouser. Remittances from international migration: A comparison of El
Salvador and Nicaragua. The Review of Economics and Statistics, 57(1): 137-146,
1995.
[19] Edward Funkhouser. Brain drain to the united states from central america. October
1999.
[20] Ernesto Gaba. Criterios para Evaluar el Tipo de Cambio de las Economias Cen-
troamericanas. Centro de Estudios Monetarios Latinoamericanos, Mexico, D.F.,
1990.
[21] Arnold Harberger. El manejo de los ingresos fiscales producidos por la bonanza.
Planeacion y Desarrollo, XXV, May 1994.
[22] Lawrence Hinkle and Peter J. Monteil. Exchange Rate Misalignment: Concepts and
Measurement for Developing Countries. Oxford University Press, New York, 1999.
[23] Julius Isaac. Economics of Migration. Oxford University Press, New York, 1947.
[24] Ronulfo Jimenez. El entorno macroeconomico y financiero de calpiA. 1999.
[25] Charles P. Kindelberger. Migration, growth and development. OECD Observer, 93,
July 1978.
[26] Anabella Larde de Palomo. Entorno de los mercados financieros rurales en el Sal
vador. 1999.
[27] Jennifer Lauby and Oded Stark. Individual migration as a family strategy: Young
women in the philippines. Population Studies, 42:473-86, 1988.
130
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[28] W. Arthur Lewis. Economic development with unlimited supplies of labor. The
Manchester School of Economic and Social Studies, 22, 1954.
[29] Lindsay Lowell and Rodolfo de la Garza. The Developmental Role of Remittances
in U.S. Latino Communities and in Latin American Countries. Tomas Rivera Policy
Institute, Claremont, California, 2000.
[30] Douglas Massey, Joaquin Arango, Graeme Hugo, Ali Kouaouci, Adela Pellegrino,
and J .Edward Taylor. Theories of international migration: A review and appraisal.
Population and Development Review, 19, 3 1993.
[31] Oscar Melhado. El Salvador: retos economicos de fin de siglo. UCA Editores, San
Salvador, El Salvador, 1997.
[32] Deborah Waller Meyers. Migrant remittances to latin america: Reviewing the liter
ature. May 1998.
[33] Segundo Montes. El Salvador 1987: Salvadorehos Refugiados en los Estados
Unidos. Universidad Centroamericana, San Salvador, El Salvador, 1987.
[34] Segundo Montes. Las Remesas que Envian los Salvadorehos de Estados Unidos:
Consecuencias Sociales y Economicas. Universidad Centroamericana, San Sal
vador, El Salvador, 1990.
[35] David Moran. Factores de expulsion y de atraccion en la migration interna de el
Salvador, 1978-1991. 1999.
[36] United Nations. 1998.
[37] Sergio Navajas and Claudio Gonzalez-Vega. Innovative approaches to rural lending:
Financiera calpia in el Salvador. 2000.
[38] Pedro Noguera. Exporting the undesirable: An analysis of the factors influencing the
deportation of immigrants from the united states and their impact on Caribbean and
central american societies. Wadebagai, a Journal o f the Caribbean and its Diaspora,
2, 1 1999.
[39] Embassy of the United States. Emigrants and Remittances-A Vital Factor in El Sal
vador’ s Economy. Internet: http://www.usinfo.org.sv/econremit.htm, Accessed: Au
gust 28, 1999, November 1998.
[40] Silvia Pedraza. Women and migration: The social consequences of gender. Annual
Review of Sociology, 17, 1991.
[41] Wim Pelupessy. Politicos agraria en El Salvador, 1960-1990. Editorial Universi-
taria Centroamericana, San Jose, Costa Rica, 1998.
131
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[42] Rafael Antonio Pleitez Chavez. El desarrollo de las microfinanzas rurales en el Sal
vador: La tecnologfa de credito de la financiera calpia. 1999.
[43] Gustav Ranis and J.C.H. Fei. A theory of economic development. American Eco
nomic Review, 51, 1961.
[44] R. Rivera Campos. Remesas familiares en el Salvador: Coyuntura y crecimiento.
1996.
[45] Sharon Stanton Russell. Remittances from international migration: A review in per
spective. World Development, 14, 6 1986.
[46] Mark Schreiner, Claudio Gonzalez-Vega, Margarita Beneke de Sanfeliu, and Mauri-
cio Shi. Notes on basic features of the rural borrowers of financiera calpia in el Sal
vador. 1999.
[47] Paul Shankman. The samoan exodus. In Victoria S. Lockwood, Thomas G. Hard
ing, and Ben J. Wallace, editors, Contemporary Pacific Societies. Prentice Hall, En
glewood Cliffs, NJ, 1993.
[48] Gabriel Siri. Uso productivo de las remesas familiares en el Salvador. 1996.
[49] Marcela Smutt and Jenny Lissette Miranda. Elfenomeno de las pandillas en El Sal
vador. Serie Adolescencia. UNICEF, FLASCO ProgramaEl Salvador, San Salvador,
El Salvador, 1998.
[50] Oded Stark and David E. Bloom. The new economics of labor migration. The Amer
ican Economic Review, 75, 2 1985.
[51] Oded Stark, J. Edward Taylor Taylor, and Shlomo Yitzhaki. Remittances and in
equality. The Economic Journal, 96, 1986.
[52] Edward Taylor and T.J. Wyatt. The shadow value of migrant remittances, income
and inequality in a household-farm economy. The Journal of Development Studies,
32:899-912, August 1996.
[53] Michael P. Todaro. Internal Migration in Developing Countries. International Labor
Office, Geneva, 1976.
[54] James Wilkie, Eduardo Aleman, and Jose Guadalupe Ortega. Statistical abstract of
latin america. 36, 2000.
[55] Makoto Yano and Jeffrey B. Nugent. Aid, nontraded goods, and the transfer paradox
in small countries. American Economic Review, 89:431-449, June 1999.
132
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix A
Logit Regressions: Investment Categories
The results of the logit regressions presented in Table (A .l) prove to be generally as ex
pected. Overall, it is clear that the characteristics of the household and potential migrants
are the motivating factors behind migration as investment, whereas other forms of invest
ment tend to be affected in more subtle ways by the interactions of key variables, imply
ing substantially different influences on various sub-groups of the sample population. Fur
ther, the explanatory power of the model varies widely between the various measures of
investment. While two-thirds of the variation in investment in international migration is
explained by the model, less than 10% of the variation in investment in a store or shop is
explained. The discussion below presents a comparative analysis of the regression results
broken down by category of explanatory variable.
133
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Invest Internatl. Internal Land Store Child
Intercept 4.132 -2.496 -18.985 -1.069 -4.118 1.624
Household
(2.757) (-1.066) (-3.579) (-0.350) (-1.857) (0.893)
HOH Age -0.178 -0.102 0.480 -0.101 0.094 -0.060
(-3.214) (-1.247) (2.700) (-0.919) (1.153) (-0.751)
HOH Age Sq. 0.001 0.001 -0.004 0.001 -0.001 -0.001
(2.415) (0.873) (-2.573) (0.589) (-1.246) (-0.570)
HOH Educ. 0.081 -0.133 0.587 -0.688 -0.154 0.099
(0.579) (-0.590) (1.602) (-1.962) (-0.742) (0.563)
Fern. HOH -0.128 -0.123 -1.993 0.367 0.535 -2.082
(-0.273) (-0.170) (-1.974) (0.374) (0.800) (-2.251)
Depend. Ratio 5.870 3.468 -4.337 -2.390 5.467 -1.626
(1.586) (0.517) (-0.301) (-0.205) (1.351) (-0.420)
East -0.278 0.053 0.448 0.112 -0.392 -0.094
(-0.925) (0.102) (0.823) (0.162) (-0.890) (-0.264)
Ag. Landowner 0.340 -0.408 0.416 -0.524 0.256 1.076
(0.909) (-0.659) (0.629) (-0.518) (0.497) (2.205)
Ag. Worker 0.162 0.180 0.132 -0.946 0.241 0.422
(0.515) (0.317) (0.217) (-1.033) (0.534) (1-174)
Health Shock -0.138 0.193 -0.288 0.633 -0.045 -0.142
(-0.540) (0.441) (-0.581) (1.100) (-0.128) (-0.451)
Table A .l: Logit Regression Results, Investment Categories.
The Household
A key feature of the present work is the focus on the household as a decision making
unit, and it should thus be expected that variation in household characteristics should man
ifest notable differences in investment patterns. These include the primary demographic
descriptors of the head of household (age, sex and education level) as well as the depen
dency ratio and a broad geographic location variable.
The characteristics of the household as proxied by the head of household have signifi
cant selected effects across the spectrum of investment options. In the overall investment
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Invest Internatl. Internal Land Store Child
Migrant
Mig. Age 0.074 0.231 0.205 0.123 -0.006 -0.008
(3.117) (6.679) (4.046) (1.969) (-0.152) (-0.214)
Mig. Age Sq. -0.001 -0.003 -0.003 -0.002 0.000 0.000
(-2.005) (-5.280) (-3.075) (-1.699) (-0.187) (0.183)
Mig. Educ 0.164 0.354 0.161 0.288 -0.233 -0.407
(1.099) (1.409) (0.793) (0.965) (-1.076) (-1.189)
Mig. Kids 0.788 1.342 0.013 -1.542 -0.472 1.019
(2.004) (3.048) (0.029) (-1.677) (-0.902) (1.959)
Mig. Male 1.646 2.400 0.229 -1.085 1.186 -0.428
(4.657) (5.930) (0.498) (-1.336) (2.142) (-0.826)
Mig. Married 0.415 1.133 -0.139 1.699 -0.044 -0.062
(0.874) (2.051) (-0.276) (1.793) (-0.075) (-0.116)
Num. Migs. 0.299 0.269 -0.331 1.533 -0.284 0.539
(1.422) (0.946) (-1.089) (1.323) (-0.661) (1.232)
Remittances -1.373 -1.290 1.887 -1.531
(-1.637) (-1.030) (1.562) (-0.973)
Income/Wealth
Income 1 (low) 0.222 0.874 -0.425 0.958 -0.473 0.547
(0.449) (1.061) (-0.505) (0.723) (-0.721) (0.935)
Income 2 -0.037 -0.231 -0.319 0.853 -0.605 0.167
(-0.094) (-0.340) (-0.455) (0.768) (-1.105) (0.353)
Income 4 -0.980 -3.199 0.495 0.543 -0.248 -0.108
(-2.238) (-3.897) (0.630) (0.417) (-0.439) (-0.210)
Income 5 (high) -0.101 1.452 -0.004 1.783 0.687 0.294
(-0.125) (0.926) (-0.002) (0.859) (0.585) (0.274)
Relative Inc. 0.412 2.002 -0.425 -0.275 -0.246 0.263
(0.814) (2.365) (-0.485) (-0.191) (-0.365) (0.447)
Land Size -0.328 -0.218 0.486 -0.871 -0.693 0.505
(-0.876) (-0.404) (0.612) (-0.503) (-1.299) (0.830)
Savings 0.379 1.384 -0.595 -2.476 -0.140 0.507
(0.723) (1.705) (-0.654) (-1.122) (-0.188) (0.705)
Table A. 1 continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Invest Internatl. Internal Land Store Child
Interactions
HOH Age * -0.100 - 0.111 0.029 0.051 -0.093 0.057
Depend. Ratio (-1.221) (-0.733) (0.100) (0.186) (-0.949) (0.555)
HOH Age * -0.002 0.003 -0.012 0.015 0.003 -0.003
HOH Educ. (-0.638) (0.708) (-1.650) (2.171) (0.624) (-0.700)
HOH Age * 0.005 0.005 -0.008 0.006 0.010 -0.008
Land Size (0.948) (0.646) (-0.690) (0.267) (1.315) (-0.635)
HOH Age * 0.000 0.000 0.000 -0.007 0.000 -0.001
Mig. Age (-0.284) (0.774) (-0.576) (-1.018) (0.433) (-1.039)
HOH Educ. * -0.032 -0.068 -0.011 -0.006 0.020 0.069
Mig. Educ. (-1.341) (-1.456) (-0.242) (-0.116) (0.517) (1.589)
Ag. Landowner * 0.015 -0.203 -0.028 0.635 0.024 -0.010
Land Size (0.103) (-1.082) (-0.070) (0.634) (0.112) (-0.036)
Ag. Landowner * 0.450 1.851 -1.426 0.456 0.472 -0.722
East (0.870) (2.112) (-1.624) (0.355) ■ (0.682) (-1.137)
Ag. Worker * 0.536 0.945 2.155 1.136
Fem. HOH (0.696) (0.666) (1.535) (1.379)
Relative Inc. * -0.393 -2.144 0.383 0.072 -0.429 -0.304
Income 5 (-0.731) (-2.374) (0.411) (0.048) (-0.575) (-0.452)
Relative Inc. * 0.042 0.098 -0.009 -0.468 0.093 -0.099
Land Size (0.772) (1.247) (-0.190) (-0.393) (1.614) (-0.459)
Relative Inc. * -0.170 -0.362 0.667 0.084 -0.276 0.383
Fem. HOH (-0.752) (-0.979) (1.390) (0.180) (-0.480) (0.999)
Pseudo-R Sq. 0.336 0.694 0.360 0.192 0.089 0.279
Table A. 1 continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
category, the age and squared age terms suggest that younger households seeking to build
wealth and older households, possibly saving for retirement or bequest, exhibit the most
intensive saving, while heads of household near the mean of 50 years of age seem to find
their funds occupied by more current expenses. Older heads of household are significantly
associated with migration to San Salvador but show a negative tendency to invest in inter
national migration. As previously stated, the negative association with international mi
gration may result from an excessive time-to-build requirement for remittances that older
households may not be willing to accept. Along the same lines, the strong positive rela
tion to internal migration may be indicative of both the lower risk and time lag for returns
which may prove more appealing to older households.
To the extent which the education level of the head of household may be a proxy for
household wealth, the results are less defined than might have been expected, although the
direction of effects are generally as hypothesized. More educated heads of household are
overall more likely to invest, particularly in internal migration, but there are negative as
sociation with land purchases, store ownership and international migration. Overall, the
patterns suggests that more educated heads of household may perceive their opportunities
for advancement to be outside the rural setting. The interaction of the age and education of
the head of household finds the effect to be particularly strong for younger and relatively
educated heads of household potentially seeking to take advantage of the urban educated
137
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
wage gap in El Salvador. Since the head of household is almost never the migrant, the posi
tive association between the education level of the potential migrant and internal migration
enhances the argument.
Although female-headed households are generally less likely to invest, they do ex
hibit a mild preference for land purchases and opening a store. The lack of significance
in the store option in particular suggests that case studies reporting substantial differences
in the investment patterns of women in favor of household manufacturing may be over
stated. Also, it is not surprising to find that female-headed households invest significantly
less frequently in children. More importantly, the regressions show that female-headed
households are negatively associated with migration, both international and internal. This
goes against most other studies which suggest that the accessibility of the migration option
makes it particularly appealing to female-headed households which typically have lesser
income and wealth. The interaction of female-headed households with relative income
instead suggests that higher income female-headed households are likely to migrate inter
nally, while any female-headed household investing in international migration is likely to
come from lower income groups. This implies that for some female-headed households,
the constraints to international migration are binding and tend to channel such households
into small manufacturing or service-related investments.
138
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The direction of influence of the dependency ratio on the various forms of investment is
again as expected despite the lack of statistical significance. Higher proportions of depen
dents reduce the incentives to invest in land and additional children, but increase the like
lihood of opening a store or shop. This positive association to household manufacturing or
service provision is expected since that genre of investment is typically one toward which
individuals categorized as dependents due to age are likely to, in reality, make important
contributions in labor time. The dependency ratio also has a weakly positive relation to
international migration which may be consistent with the perception of international mi
gration as a more long-term form of investment that may be particularly appealing when
the beneficiaries of the potential remittances are still young.
The location of the household as indicated by the east variable suggests that differences
in investment probabilities based only on geography are small. The reduced likelihood
of Eastern households to open a shop may be indicative of the lesser demand for home-
produced goods and services in the more rural Eastern setting. Conversely, the weakly
positive association with land purchases points to the stronger agricultural emphasis in the
east. The most notable effect of the Eastern indicator lies in its interaction with the agri
cultural landowner indicator and brings out the strong tendency for Eastern landowners to
invest in international migration50.
The health shock variable proves to be of unexpectedly low significance. Health shocks
to any member of the household can be costly both in terms of explicit expenditures and
50This interaction is discussed further in Section 5.2.3.
139
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
lost income, and the ex ante expectation is that at least some form of investment option
would be strongly negatively correlated with the presence of shocks. However, the shock
variable itself provides only a limited degree of information since neither the length nor
severity of the health shock are known.
The Investment Participant
Isolating the characteristics of the individual within the household most likely to con
tribute to investment success reveals the core of the motivations to invest. The most strik
ing results by far in the investment regressions are those related to the characteristics of in
vestment participant and provide strong support for a crucial finding of the model of Chap
ter 4: the forms of intra-household variation which affect entry costs and the probabilities
of successful investment are key determinants in the investment decision. In the case of El
Salvador, it seems clear that the presence of educated, working-age, male investment par
ticipants within the household greatly enhances the likelihood of investment, particularly
in international migration.
The coefficients for the age and squared age of the investment participant suggest that
investment participants near the mean of 30 years of age have positive effects on the like
lihood of all forms of investment, except shop ownership. Further, the age of the invest
ment participant contributes the greatest degree of explanatory power in the migration re
gressions. The strength and direction of investment participant age remain robust to the
inclusion of the interaction between the age of the head of household and the age of the
investment participant which are positively correlated in the data.
140
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The coefficients for the male investment participant indicator are nearly as strong. As
expected, households with available men are much more likely to invest in international
migration. Although still positive, the coefficient in the internal migration regression is
not statistically significant. Also, the male investment participant indicator has important
explanatory power in the probability of investment in a store or shop. The negative coeffi
cient in the land purchase regression is unusual, but not unexpected since, taken together,
it seems to suggest that the international migration option not only tends to reduce the at
tractiveness of investment in land, but actively provides disincentives to do so.
The education level of the investment participant is positively associated with inter
national and internal migration, and land purchases, but the statistical significance is low.
This provides only weak support for Funkhouser’s (1999) brain drain from El Salvador as a
result of international migration since neither the coefficients nor the significance levels of
the investment participant’s education level differ substantially between the international
migration and land purchase regressions. Indeed, the negative and weakly significant coef
ficient for the interaction of the education of the head of household and investment partic
ipant suggests that more educated investment participants from relatively high-education
households do not migrate.
Taken together, the results for the characteristics of the investment participant indicate
that young Salvadoran men with some education are expected to contribute most to the
household via international migration. These effects transcend not only the other demo
graphic characteristics of the household, but also measures of income and wealth.
141
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The coefficients for the number of migrants outside the household also suggest an inter
esting story. The positive coefficient in the international migration regression implies that
there is some tendency for migrants to attract additional migrants from within the house
hold, but the lack of statistical significance suggests that, on a broad scale, the effect is
weak. Instead, the slight significance of the number of migrants in the land purchase re
gression may imply that the number of migrants is a proxy for remittance income which
the household utilizes to purchase land.
Income and Wealth
The most direct measures of the household’s ability to invest - income and wealth -
generally prove to be of limited significance, although the directions of influence do pro
vide some insight. The regression coefficients for the income quintiles suggest that while
international migrants do come from all income categories, internal migrants are much
more likely to be from upper income households. Households near median per capita in
come are the least likely to purchase land, but are also the most likely to open a store or
shop.
The relative income variable yields much more interesting results. This variable mea
sures per capita household income relative to mean per capita household income in the
household’s departamento and is meant to capture the status motivations behind invest
ments. In particular, work by Stark suggests that, beyond the singular quest for economic
gain, relatively poor households may more actively seek to invest in migration in order to
improve their relative standing in the community in addition to their absolute well-being.
142
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, the positive and statistically significant coefficient in the international migration
regression indicates the exact opposite in the Salvadoran case. Not only is it the case that
high income households (quintile 5) are more likely to migrate internationally, but within
departamentos, regardless of absolute income, relatively high-income households have a
higher probability of investing in international migration. Although the interaction of the
relative income variable with the upper income quintile indicator suggests that the highest
income households are significantly less likely to migrate internationally, it is significant
that the effect of the relative income variable alone remains robust to the inclusion of the
interaction.
The size of the household’s land holdings is included as a measure of wealth, but the
regression results seem to be more indicative of the degree to which the family remains
tied to agriculture. Households with greater land holdings are less likely to invest in in
ternational migration or a shop, but do show a slight tendency for internal migration. Not
surprisingly, larger landowners seem less inclined to further invest in land, but somewhat
more likely to have children, both potentially indicative of the diminishing returns to land
and the labor-intensive nature of Salvadoran agriculture.
143
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix B
Logit Regressions: Human Capital Investment
Table (B) presents the results of the human capital logit regressions. As a result of the
expansion of primary education in El Salvador, most households invested in primary edu
cation, and the majority of those who did not either had no children or no school age chil
dren. The variability in the characteristics of the household is thus much greater than the
elementary school indicator, and the result is a pseudo-R squared of only 13.7%. At the
higher levels, enrollment drops sharply, and the explanatory power of the model rises to
nearly 60% as household income and wealth variables become increasingly significant.
Differing importantly from the other forms of investment, the regressors in the human
capital regressions show statistical significance in nearly all categories, suggesting, espe
cially at the higher levels, a complicated decision which accounts for income as well as
the demographic indicators for the household. Further, the schooling decision must rec
oncile the household’s ability to invest in human capital with the student’s ability to make
the investment lucrative in the future. Ideally, the data would then include reliable student
quality indicators that would be suggestive of the expected future rate of return on human
144
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Elem. Jr. High High Schl. Post-Sec.
Intercept -3.735 -10.932 -20.036 -63.506
Household
(-3.010) (-4.579) (-4.399) (-3.215)
HOH Age 0.198 0.362 0.523 1.368
(4.315) (4.099) (3.387) (1.984)
HOH Age Sq. -0.002 -0.003 -0.004 -0.012
(-4.760) (-4.182) (-3.207) (-1.953)
HOH Educ. 0.057 0.013 0.616 1.798
(0.518) (0.075) (2.457) (2.098)
Fern. HOH -0.717 1.278 0.882 11.206
(-1.737) (2.448) (1.318) (2.028)
Depend. Ratio -4.044 -2.517 -18.864 -30.091
(-1.450) (-0.447) (-1.303) (-0.958)
East -0.292 -0.089 0.052 0.637
(-1.169) (-0.272) (0.100) (0.492)
Ag. Landowner -0.429 -0.549 0.136 1.183
(-1.300) (-1.341) (0.229) (0.664)
Ag. Worker -0.571 -0.252 -0.036 -0.403
(-2.149) (-0.711) (-0.061) (-0.205)
Health Shock 0.046 -0.499 -0.459 0.721
(0.209) (-1.725) (-1.098) (0.616)
Table B.l: Logit Regression Results, Human Capital Investment.
capital investment. The closest approximation in the data to such an indicator is the indi
vidual’s completed years of schooling. As a measure of the probability of future success,
however, the total years of schooling may be questionable since it is not necessarily related
to the latent quality of the individual, and it is also not clear that marginal additions to an
individual’s human capital greatly affect the expected future rate of return to the house
hold’s investment. Further, using the years of schooling as an indicator for future success
biases against younger students who simply have not yet accumulated as much educational
experience.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Elem. Jr. High High Schl. Post-Sec.
Migrant
Mig. Age -0.017 0.012 -0.015 0.272
(-0.801) (0.455) (-0.418) (1.329)
Mig. Age Sq. 0.000 0.000 0.000 -0.005
(1.008) (-0.352) (0.085) (-1.031)
Mig. Educ 0.325 -0.314 0.340 -0.040
(2.493) (-1.649) (1.875) (-0.077)
Mig. Kids -0.328 -0.669 0.180 -5.268
(-1.016) (-1.624) (0.355) (-1.710)
Mig. Male -0.019 0.159 0.704 -3.131
(-0.061) (0.407) (1.341) (-1.325)
Mig. Married -0.140 0.401 0.179 -0.377
(-0.399) (0.896) (0.319) (-0.154)
Num. Migs. -0.247 0.565 -1.255 1.003
(-1.240) (1.806) (-1.728) (1.445)
Remittances -1.116 0.200 1.028 0.078
(-1.448) (0.158) (0.799) (0.036)
Income/Wealth
Income 1 (low) 0.936 0.651 1.016 10.305
(2.160) (1.268) (1.206) (1.857)
Income 2 0.269 -0.091 0.984 15.059
(0.803) (-0.210) (1.357) (4.679)
Income 4 -0.511 -1.020 0.527 16.353
(-1.413) (-2.151) (0.692)
Income 5 (high) -0.655 2.027 2.894 22.792
(-0.952) (2.366) (2.431) (5.118)
Relative Inc. 0.188 0.975 0.773 1.895
(0.437) (1.885) (1.058) (0.714)
Land Size 0.248 -0.516 1.577 0.849
(0.618) (-0.402) (1.727) (0.802)
Savings 0.574 0.287 -0.059 -0.919
(1.213) (0.518) (-0.081) (-0.501)
Table B .l continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable Elem. Jr. High High Schl. Post-Sec.
Interactions
HOH Age * 0.141 0.034 0.288 0.505
Depend. Ratio (2.120) (0.267) (1.028) (0.862)
HOH Age * 0.000 0.002 -0.006 -0.029
HOH Educ. (0.026) (0.543) (-1.293) (-1.768)
HOH Age * -0.006 -0.002 -0.023 -0.004
Land Size (-1.033) (-0.100) (-1.683) (-0.264)
HOH Age * 0.000 -0.001 0.000 0.000
Mig. Age (-0.775) (-1.767) (0.449) (-0.294)
HOH Educ. * -0.052 0.061 -0.055 0.051
Mig. Educ. (-2.524) (2.066) (-1.708) (0.787)
Ag. Landowner * -0.072 0.239 -0.351 -0.230
Land Size (-0.386) (0.417) (-1.214) (-0.587)
Ag. Landowner * 0.225 1.195 1.297 1.318
East (0.510) (2.163) (1.646) (0.612)
Ag. Worker * 0.357 0.007 0.353
Fem. HOH (0.561) (0.009) (0.326)
Relative Inc. * -0.221 -1.432 -1.545 -3.224
Income 5 (-0.478) (-2.534) (-1.955) (-1.210)
Relative Inc. * 0.006 0.120 0.037 -0.241
Land Size (0.161) (1.903) (0.537) (-1.019)
Relative Inc. * 0.448 -1.015 -0.149 -17.286
Fem. HOH (1.947) (-2.110) (-0.376) (-1.634)
Pseudo-R Sq. 0.141 0.149 0.242 0.623
Table B .l continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The completed years of schooling are then only evidence of past performance and could
be used, along with the demographic and economic profile of the household, to estimate the
probability of continued household investment in education using, for example, a Cox pro
portional hazards model. Such an analysis is not performed here for two reasons. First, a
detailed analysis of the proportional hazard of continued investment in education in El Sal
vador and, in particular, its relation to remittances received by the household, is performed
in Cox-Edwards and Ureta (1999). Second, the goal of the present analysis is comparison
of the effects of a common set of explanatory variables on the probability of participation
in various forms of investment. To that extent, continued use of binary dependent variables
for household investment in each level of education seems appropriate.
The binary dependent variables utilized in this portion of the analysis are four indi
cators for the highest level of education invested in by the household. As described pre
viously, these are assigned based on the number of years of schooling completed by the
most educated member of the household enrolled in school at some point in time between
1998 and 2000. The categories are 1) elementary (1-6 years completed), 2) junior high
(7-9 years), 3) high school (10-12 years) and 4) post-secondary (more than 12 years). A
potential weakness of the method utilized to construct the dependent variables is that it ex
plicitly prevents overlapping to indicate households that may be investing in human capi
tal at multiple levels by, for example, having multiple children in school at different levels.
148
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, counting all the education levels at which a household invested would tend to ho
mogenize the regression results by effectively adding multiple instances of the same house
hold’s characteristics within or between regressions. This would re-weight, and therefore
bias, the data in favor of the characteristics of households with many children. Restricting
the dependent variable to the highest level of education sought instead focuses the analysis
by more sharply bringing out the differences in households choosing to invest in various
levels of education.
The effects of the characteristics of the household on the probability of human capi
tal investment are strong. Households headed by educated females near the mean age of
50 are very likely to invest in education, particularly at the secondary and post-secondary
levels. Female-headed households are seen to be more likely to invest in human capital at
all levels, with the exception of elementary education which is consistent with the fact that
female-headed households are less likely to have young children. The education level of
the head of household has a strong positive association to the probability of investment in
secondary and post-secondary education. This relation is expected to the extent that more
educated heads of household may impose higher academic expectations on the other mem
bers of the household. Also, if the education level of the head of household may be viewed
as a proxy for wealth, wealthier households may be better able to afford postponing income
earning activities by some of its members in order to reap greater rewards later.
149
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Agricultural worker households are less likely to invest in human capital, particularly
at the elementary level. This is consistent with the general pattern of lower educational at
tainment, deeper poverty and greater isolation from schooling facilities among this portion
of the sample population. Elevated dependency ratios are negatively associated with hu
man capital investment at any level, although statistical significance is low. Nonetheless,
this further solidifies the pattern for dependent-heavy households which suggests that most
of their funds tend to be used for more immediate consumption purposes.
Explanatory variables related to income provide a mixed picture. Households in the
lowest income quintile are found to be significantly more likely to invest in education at
both the elementary and post-secondary levels, whereas middle quintile households are
the least likely to invest in human capital at all. Upper quintile households are more likely
to invest in all levels of education above elementary51, with a particularly strong effect at
the post-secondary level. It is important to note that the degree of statistical significance
of the income variables in the human capital regressions is a clear departure from the re
sults of the other investment regressions. While income is merely suggestive in all other
investment regressions, the probability of human capital investment depends heavily on
the income position of the household. Further, the pattern of human capital investment is
not monotonically skewed in favor of the wealthy, but rather, indicates active investment
in human capital by substantial portions of lower income households as well.
51The negative association between upper quintile households and the probability of investing in elemen
tary education may be due to the positive correlation between the age of the head of household and the upper
quintile indicator.
150
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix C
Logit Regressions: Migrant Finance
Migrant finance regression results are provided in Table C. The majority of household in
dicators seem to have limited explanatory power in the migrant finance regressions and
are instead primarily indicative of directions of influence. The key exception is the ed
ucation level of the head of household. More educated heads of household are not only
significantly less likely to draw down their stock of assets to finance migration, but are
also significantly more likely to not finance the migrant at all. However, the interaction
of the education levels of the migrant with the head of household shows that households
with equally educated migrants are, on the basis of the education of the head of household,
significantly more likely to finance migration out of the stock of assets.
Other aspects of the household provide suggestive but less powerful results. Heads of
household near the mean age of 50 years do not tend to dissave or sell assets and instead
tend to obtain funds from extra-household family members. Female-headed households
are not found to differ importantly in their sources of migrant finance, although they do
exhibit a slight tendency to not finance the migrant at all. It is nonetheless noteworthy that
151
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable
HH
Borrow
HH
Save/Sell
Extra-HH
Family No Help
Intercept 0.895 4.819 -8.057 -9.480
Household
(0.112) (1.186) (-1.818) (-1.602)
HOH Age 0.205 -0.117 0.205 0.099
(0.765) (-1.085) (1.715) (0.655)
HOH Age Sq. -0.004 0.001 -0.001 0.000
(-1.647) (0.793) (-1.407) (-0.187)
HOH Educ. -0.493 -0.662 0.323 0.606
(-1.352) (-2.021) (1.368) (2.046)
Fem. HOH -0.462 -0.390 -0.256 0.408
(-0.453) (-0.588) (-0.380) (0.545)
Depend. Ratio 2.472 2.201 3.652 -1.582
(0.502) (0.869) (1.349) (-0.497)
East 1.146 -0.309 0.695 -0.921
(1.330) (-0.518) (1.377) (-1.552)
Ag. Landowner 1.233 0.040 0.185 -1.094
(0.983) (0.056) (0.270) (-1.330)
Ag. Worker 0.667 -0.944 0.361 0.045
(0.741) (-1.404) (0.643) (0.068)
Health Shock 0.221 0.597 -0.689 -0.307
(0.320) (1.314) (-1.502) (-0.592)
Table C .l: Logit Regression Results, Sources of Migrant Finance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable
HH
Borrow
HH
Save/Sell
Extra-HH
Family No Help
Migrant
Mig. Age -0.158 -0.038 0.082 -0.045
(-1.008) (-0.453) (0.952) (-0.428)
Mig. Age Sq. 0.000 0.000 0.000 0.001
(-0.087) (-0.007) (-0.263) (1.154)
Mig. Educ 0.092 -0.122 -0.038 0.174
(0.759) (-1.463) (-0.470) (2.049)
Mig. Kids 2.106 -0.527 -0.616 0.447
(2.587) (-1.096) (-1.420) (0.862)
Mig. Male 0.023 -0.159 -0.764 1.428
(0.027) (-0.291) (-1.572) (2.173)
Mig. Married 1.079 -0.224 -0.295 -0.077
(1.427) (-0.437) (-0.604) (-0.148)
Num. Migs. -0.346 0.058 0.149 -0.045
(-1.493) (0.419) (1.222) (-0.322)
Income/Wealth
Income 1 (low) -3.323 -0.497 -0.338 2.724
(-1.739) (-0.471) (-0.372) (2.423)
Income 2 -1.807 0.167 -0.881 2.048
(-1.365) (0.209) (-1.180) (2.244)
Income 4 -1.520 0.758 0.714 -0.443
(-0.712) (0.584) (0.672) (-0.354)
Income 5 (high) -22.958 -0.018 0.515 2.827
Relative Inc. -3.107
(-0.010)
-1.924
(0.343)
-0.228
(1.586)
3.486
(-0.848) (-0.964) (-0.144) (1.860)
Land Size -0.610 -1.056 0.401 0.460
(-1.209) (-1.440) (1.007) (1.023)
Savings 0.979 0.116 -0.443 -0.127
(1.065) (0.187) (-0.700) (-0.193)
Table C.l continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable
HH
Borrow
HH
Save/Sell
Extra-HH
Family No Help
Interactions
HOH Age * -0.132 -0.039 -0.052 0.037
Depend. Ratio (-1.249) (-1.052) (-1.204) (0.783)
HOH Age * 0.011 0.005 -0.003 -0.012
HOH Educ. (1.241) (0.798) (-0.619) (-1.929)
HOH Age * 0.014 0.013 -0.010 -0.007
Land Size (1.247) (1.322) (-1.188) (-0.806)
HOH Age * 0.002 0.001 -0.001 0.000
Mig. Age (0.981) (0.966) (-0.840) (-0.059)
HOH Educ. * -0.244 0.142 0.099 0.069
Mig. Educ. (-0.754) (0.384) (0.471) (0.226)
Ag. Landowner * 2.181 -0.308 -0.629 0.123
Land Size (1.521) (-0.348) (-0.763) (0.128)
Ag. Landowner * 2.257 0.455
East (1.504) (0.336)
Ag. Worker * -0.014 0.048 -0.017 -0.008
Fem. HOH (-0.681) (2.352) (-0.881) (-0.454)
Relative Inc. * 3.705 1.675 -0.213 -3.454
Income 5 (1.003) (0.818) (-0.131) (-1.810)
Relative Inc. * -0.173 0.255 0.005 -0.175
Land Size (-0.494) (1.122) (0.061) (-0.900)
Relative Inc. * -0.426 0.286 -0.388 0.025
Fem. HOH (-0.229) (0.239) (-0.391) (0.019)
Pseudo-R Sq. 0.382 0.192 0.135 0.233
Table C.l continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
female-headed households do not appear to have lesser access to credit for the purpose
of migration. Although eastern households invest in migration less frequently than west
ern households, those who do have a higher tendency to provide support for the migrant,
often derived from extra-household family members. Categorizing households by the rel
ative sources of income shows that agricultural workers may use their assets as borrow
ing collateral rather than a direct source of liquid funds to finance migration. Agricultural
landowners instead draw from their stock of assets to finance the migrant.
The relation of the characteristics of the migrants to the method of finance makes it
clear that cost minimization plays a key role in the decision to send a migrant. The most
significant result suggests a strong tendency for households with educated, male and (to a
lesser degree) young migrants to not provide any financing at all. As noted previously, this
effect is reduced for more educated migrants with educated heads of household. Further,
the migrant effects are mirrored somewhat in the other forms of financing since households
with educated male migrants are less likely to borrow or draw down assets to finance mi
gration. The negative relation of migrant age to the probability of borrowing may be in
dicative of households viewing migration as a long-term investment and thus exhibiting a
greater propensity to borrow when the flow of returns is more likely to extend well into the
future.
Two notable results stand out in the relations between income and the sources of mi
grant finance. First, households in the two lower income quintiles have a strong tendency
to not finance the migrant. This is not only expected but also highlights a key motivation
155
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
for investing in migration among the lower income groups. Given the resources available
to most lower income households, the existence of an investment option with low entry
cost and potentially high returns holds great attraction. It is worth noting at this point that
neither the quality of the migration experience nor the perceived cost of migration are ho
mogeneous across income groups52; thus it may be the case that the household specific cost
of migration relative to household income may be less variable than the regression results
suggest.
The second noteworthy income-related finding is that upper income households do not
borrow to finance migration. While this is the strongest result in this set of regressions,
the other forms of migrant finance do not mirror the result, and the data therefore do not
otherwise indicate a preferred method of migrant finance for upper income households.
52See Section 2.3.2.
156
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix D
Logit Regressions: Remittances
The results of these regressions are presented in Table (D. 1) and are almost entirely as ex
pected. The characteristics of the migrant dominate in significance, followed closely by
the methods of migrant finance. Other demographic or economic aspects of the household
are suggestive but not powerful. Overall, the full-sample regression explains nearly 60%
of the variation in the remittance indicator, and 25% of the variation is explained for the
187 household sub-sample of recent investors.
The characteristics of the household appear to be of limited significance in determin
ing the probability of receiving remittances. In general, younger and more educated heads
of household have a greater probability of receiving remittances, while the probability for
female-headed households is lower. Although previous studies suggest that female-headed
households may be more likely to receive remittances, the opposite result here is expected
since the results of Section 5.2 found that female-headed households are also less likely to
invest in international migration. One unexpected result is that agricultural worker house
holds are significantly less likely to receive remittances despite the higher probabilities of
157
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Full Migrant Full Migrant
Variable Sample since 1998 Variable Sample since 1998
Intercept -6.262
(-2.689)
-1.675
(-0.579)
Household Migrant
HOH Age -0.052 -0.063 Mig. Age 0.377 0.173
(-0.723) (-0.736) (7.989) (3.149)
HOH Age Sq. 0.001 0.001 Mig. Age Sq. -0.005 -0.003
(1.132) (1.010) (-7.133) (-2.992)
HOH Educ. 0.271 0.171 Mig. Educ. 0.598 0.488
(1.290) (0.739) (2.493) (1.693)
Fem. HOH -0.372 -0.828 Mig. Kids -0.356 -0.313
(-0.585) (-0.999) (-0.924) (-0.656)
Depend. Ratio 1.379 3.440 Mig. Male 0.191 -0.567
(0.287) (0.580) (0.484) (-1.050)
East 0.126 0.299 Mig. Married -0.167 -0.353
(0.289) (0.534) (-0.408) (-0.736)
Ag. Landowner 0.459 0.595 Mig. Time 0.055 0.001
(0.763) (0.730) (0.424) (0.004)
Ag. Worker -1.104 -1.558 Mig. Time Sq. 0.001 0.001
(-2.264) (-2.616) (0.150) (0.099)
Health Shock -0.478 -0.575 Num. Migs. 0.081 0.252
(-1.217) (-1.239) (0.322) (0.709)
Finance
HH Borrow 4.317
(2.738)
2.842
(1.694)
Extra-HH Family 4.652
(2.362)
3.990
(1.693)
No Help 1.014
(1.031)
1.453
(0.989)
Table D .l: Logit Regression Results, Remittances.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Variable
Full
Sample
Migrant
since 1998 Variable
Full
Sample
Migrant
since 1998
Income/Wealth Interactions
Income 1 (low) 0.721 0.406 HOH Age * -0.014 -0.077
(0.985) (0.496) Depend. Ratio (-0.126) (-0.506)
Income 2 0.645 1.078 HOH Age * -0.007 -0.007
(1.108) (1.531) HOH Educ. (-1.592) (-1.361)
Income 4 -0.977 0.760 HOH Age * -0.008 -0.006
(-1.446) (0.761) Land Size (-1.521) (-0.738)
Income 5 (high) 1.326 0.528 HOH Age * -0.002 -0.001
(0.936) (0.312) Mig. Age (-2.283) (-1.714)
Relative Inc. 0.833 -0.011 HOH Educ. * -0.191 -0.275
(1.077) (-0.012) Mig. Educ. (-1.200) (-0.822)
Land Size 0.754 0.695 Ag. Landowner * -0.439 -0.691
(1.890) (1.180) Land Size (-0.588) (-0.727)
Savings -1.423 -1.531 Ag. Landowner * 1.475
(-2.065) (-1.915) East (1.068)
Ag. Worker * -0.079 -0.060
Fem. HOH (-1.960) (-1.298)
Relative Inc. * -1.259 -0.152
Income 5 (-1.562) (-0.157)
Relative Inc. * 0.032 0.017
Land Size (0.702) (0.309)
Relative Inc. * 0.024 0.609
Fem. HOH (0.063) (0.701)
Pseudo-R Sq. 0.598 0.257
Table D .l continued.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
investing in migration and financing migration through borrowing. This suggests a partic
ularly low net payoff to such households.
The profile of the potential migrant, however, proves extremely relevant in determining
the probability of receiving remittances. Educated migrants near the mean age of 30 years
are by far the most likely to remit. However, male migrants are not significantly more
likely to remit despite the overwhelming prevalence of male migrants. This lends credence
to the notion that male migrants are preferred for the lower initial cost of migration, as
opposed to an expectation of higher gross return.
Income and wealth variables are not found to bear strong relationships with the prob
ability of receiving remittances. The data suggest that households at the extremes of the
income distribution are more likely to receive remittances in contrast to middle income
households; again, this finding is consistent with the overall pattern of investment in mi
gration. An interesting and moderately significant result merges from the interaction be
tween relative income and the upper income quintile indicator which suggests that upper
quintile households at the lower end of the cutoff are more likely to receive remittances.
Also, household wealth as measured by the size of land holdings (or proxied by the educa
tion level of the head of household) suggests that wealthier households may be more likely
to receive remittances, although the relation is not statistically strong.
The structure of the regressions differs slightly from those of the previous sections
in that indicators for the type of migrant financing are included as explanatory variables.
These variables are important since they are indicative of the relation between the extent
160
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of sacrifice by the household and the ultimate payoff. As expected, these variables prove
significant. In particular, households which either received funds from extra-household
family or borrowed to send the migrant have a significantly higher probability of receiv
ing remittances. This provides some evidence in support of the importance of social net
works in the enforcement of migration as investment, which involves both temporal and
geographic separation.
Taken together, the pattern of remittances suggests that the rate of return to the migra
tion investment is not necessarily homogeneous. Relatively low income households with
young and educated male migrants appear to be able to successfully undertake migration
at a low entry cost and are more likely to receive remittances. However, other households
seem to have over-estimated their expectation of returns since their is evidence that par
ticular subgroups of the sample borrowed to finance migration, but are not receiving any
returns53. Also, the relative deprivation motivation for migration and remittances seems
to hold true only for upper income households such that the term seems inappropriate and
may be better referred to as the status motivation.
53However, as stated previously, there is no reason to believe that even those households will not eventu
ally receive remittances.
161
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix E
Bivariate Probit Estimations
The empirical methodology used in Section 5.2 estimates logit regressions for each type
of investment as well as a broad composite regression which describes the probability of
participating in any form of investment at all. The approach assumes for simplicity that
the results of the regressions are independent of each other such that, in particular, there
is no correlation between the error terms of the estimated models. However, a greater de
gree of technical and conceptual sophistication is possible. Specifically, it may be more
accurate to frame the empirical methodology as one which seeks to estimate the probabil
ity of participating in a particular type of investment, given that the household chooses to
invest at all. This effectively acknowledges that investment in, say, international migration
is conditional on the household’s decision to invest at all.
Rather than separately estimating the investment decision and the investment type, the
technique is then to estimate a bivariate, or joint conditional, model. For example, the
probability of investing in international migration is jointly estimated with the probabil
ity of investing at all and explicitly accounts for the conditional relation by incorporating
162
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the correlation between the error terms of the separate estimations. In these techniques,
the probit model is preferred over the logit since the assumption of normality in the probit
model simplifies joint estimation as compared to the binomial distribution utilized in logit
regressions.
163
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Variable Invest
Internatl.
Migration Invest
Internal
Migration Invest
Purchase
Land Invest Store
Intercept 2.160 -0.986 2.362 -9.915 2.609 -0.421 2.785 -2.320
Household
(2.382) (-0.790) (2.681) (-3.942) (2.984) (-0.303) (3.290) (-2.115)
HOH Age -0.098 -0.051 -0.101 0.256 - 0.111 -0.066 -0.113 0.049
(-2.935) (-1.107) (-3.123) (3.022) (-3.428) (-1.287) (-3.645) (1.213)
HOH Age Sq. 0.001 0.000 0.001 -0.002 0.001 0.000 0.001 0.000
(2.196) (0.775) (2.342) (-2.894) (2.637) (1.003) (2.705) (-1.278)
HOHEduc. 0.070 -0.028 0.066 0.333 0.036 -0.288 -0.004 -0.041
(0.812) (-0.239) (0.783) (1.729) (0.420) (-1.778) (-0.047) (-0.420)
Fem. HOH -0.070 -0.100 -0.115 -1.013 -0.008 0.254 -0.207 0.215
(-0.258) (-0.313) (-0.412) (-2.052) (-0.028) (0.518) (-0.793) (0.652)
Depend. Ratio 3.287 0.577 3.280 -1.588 3.021 -1.131 2.716 3.436
(1.508) (0.169) (1.531) (-0.256) (1.461) (-0.283) (1.287) (1.543)
East -0.209 -0.053 -0.124 0.582 -0.202 0.242 -0.095 -0.163
(-1.165) (-0.220) (-0.709) (2.089) (-1.155) (0.676) (-0.563) (-0.733)
Ag. Landowner 0.190 -0.133 0.219 0.376 0.200 -0.202 0.224 0.123
(0.892) (-0.515) (1.022) (1.105) (0.931) (-0.444) (1.081) (0.511)
Ag. Worker 0.101 -0.119 0.051 0.182 0.066 -0.431 0.076 0.110
(0.549) (-0.476) (0.279) (0.610) (0.367) (-1.000) (0.440) (0.537)
Health Shock -0.055 0.118 -0.077 -0.142 -0.100 0.339 0.027 -0.036
(-0.368) (0.620) (-0.509) (-0.587) (-0.661) (1.245) (0.189) (-0.203)
Table E.l: Bivariate Probit Regression Results, Investment Categories.
ON
45-
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Variable Invest
Internatl.
Migration Invest
Internal
Migration Invest
Purchase
Land Invest Store
Migrant
Mig. Age 0.039 0.107 0.041 0.099 0.041 0.061 0.047 -0.001
(2.583) (5.997) (2.980) (4.661) (2.954) (1.961) (3.396) (-0.032)
Mig. Age Sq. -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 0.000
(-1.869) (-4.072) (-1.817) (-3.397) (-1.851) (-1.755) (-2.208) (-0.086)
Mig. Educ 0.094 0.127 0.070 0.133 0.074 0.103 0.073 -0.165
(1.092) (1.178) (0.833) (1.219) (0.868) (0.674) (0.920) (-1.536)
Mig. Kids 0.560 0.738 0.547 0.045 0.514 -0.650 0.535 -0.242
(2.388) (3.119) (2.386) (0.186) (2.276) (-1.511) (2.415) (-0.909)
Mig. Male 1.141 1.332 1.001 -0.025 0.976 -0.568 0.923 0.486
(5.186) (6.183) (4.787) (-0.103) (4.719) (-1.480) (4.414) (1.943)
Mig. Married 0.212 0.565 0.164 -0.081 0.102 0.633 0.099 -0.066
(0.773) (2.095) (0.620) (-0.299) (0.389) (1.474) (0.385) (-0.230)
Num. Migs. 0.211 0.142 0.189 -0.240 0.158 0.816 0.178 -0.078
(1.625) (0.850) (1.487) (-1.362) (1.401) (1.419) (1.488) (-0.373)
Table E.l continued.
as
Lr\
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Internatl. Internal Purchase
Variable Invest Migration Invest Migration Invest Land Invest Store
Income/Wealth
Income 1 (low) 0.199 0.134 0.118 -0.336 0.162 0.419 0.021 -0.211
(0.695) (0.393) (0.414) (-0.757) (0.570) (0.712) (0.077) (-0.645)
Income 2 0.017 -0.153 -0.041 -0.358 0.019 0.337 -0.045 -0.314
(0.074) (-0.506) (-0.180) (-0.956) (0.082) (0.703) (-0.204) (-1.201)
Income 4 -0.590 -1.104 -0.564 0.121 -0.493 0.283 -0.493 -0.275
(-2.381) (-3.448) (-2.350) (0.322) (-2.065) (0.486) (-2.123) (-0.997)
Income 5 (high) 0.017 0.411 -0.024 0.098 -0.081 0.666 -0.174 0.233
(0.035) (0.518) (-0.051) (0.121) (-0.172) (0.688) (-0.394) (0.469)
Relative Inc. 0.300 0.639 0.218 -0.241 0.193 -0.146 0.066 -0.073
(0.960) (1.669) (0.768) (-0.519) (0.690) (-0.240) (0.239) (-0.226)
Land Size -0.136 -0.046 -0.164 0.140 -0.173 -0.313 -0.215 -0.445
(-0.677) (-0.197) (-0.752) (0.320) (-0.816) (-0.581) (-0.982) (-1.514)
Savings 0.134 0.349 0.190 -0.274 0.183 -0.948 0.186 0.066
(0.410) (0.884) (0.617) (-0.561) (0.602) (-0.979) (0.665) (0.202)
Table E. 1 continued.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Variable Invest
Internatl.
Migration Invest
Internal
Migration Invest
Purchase
Land Invest Store
Interactions
HOH Age * -0.054 -0.014 -0.053 0.005 -0.048 0.028 -0.037 -0.070
Depend. Ratio (-1.097) (-0.182) (-1.109) (0.043) (-1.053) (0.324) (-0.816) (-1.300)
HOH Age * -0.002 0.000 -0.002 -0.007 -0.001 0.006 0.000 0.001
HOH Educ. (-0.877) (0.105) (-0.866) (-1.739) (-0.486) (1.979) (0.031) (0.272)
HOH Age * 0.002 0.002 0.003 -0.003 0.003 0.004 0.003 0.007
Land Size (0.773) (0.589) (0.812) (-0.466) (0.882) (0.509) (0.965) (1.721)
HOH Age * 0.000 0.000 0.000 0.000 0.000 -0.004 0.000 0.000
Mig. Age (-1.281) (-0.483) (-1.117) (0.757) (-1.037) (-1.258) (-1.063) (-0.294)
HOH Educ. * -0.026 -0.119 -0.006 0.045 -0.013 0.215 0.025 0.004
Mig. Educ. (-0.318) (-1.414) (-0.071) (0.165) (-0.149) (0.533) (0.282) (0.024)
Ag. Landowner * 0.296 0.735 0.216 -0.912 0.269 0.090 0.201 0.250
Land Size (0.983) (1.974) (0.717) (-2.036) (0.901) (0.152) (0.692) (0.731)
Ag. Landowner * 0.257 0.648 0.328 1.203 0.243 -5.512 0.545 0.548
East (0.579) (0.954) (0.751) (1.782) (0.547) (0.000) (1.320) (1.280)
Ag. Worker * -0.019 -0.020 -0.016 -0.009 -0.016 0.002 -0.016 0.019
Fem. HOH (-1.378) (-1.059) (-1.170) (-0.411) (-1.166) (0.075) (-1.188) (0.971)
Relative Inc. * -0.324 -0.703 -0.227 0.243 -0.188 0.157 -0.054 -0.181
Income 5 (-0.976) (-1.687) (-0.742) (0.493) (-0.625) (0.235) (-0.182) (-0.527)
Relative Inc. * 0.035 0.041 0.027 -0.016 0.036 -0.316 0.017 0.049
Land Size (1.073) (1.247) (0.836) (-0.554) (1.061) (-0.690) (0.615) (1.831)
Relative Inc. * -0.082 -0.061 -0.085 0.291 - 0.111 -0.051 -0.079 -0.032
Fem. HOH (-0.629) (-0.436) (-0.623) (1.279) (-0.816) (-0.199) (-0.590) (-0.125)
Table E.l continued.
ON

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Business group affiliation in India

PDF

Insurance mechanisms, forest clearance, and the effect of government policies in rural economies

PDF

Health interventions, health status, and labor market outcomes in Indonesia: Three essays

PDF

Life-cycle demographic decisions and the economic well-being of single parents in the United States, 1964-1989

PDF

Weight variations among preschoolers: An analysis of evidence from rural Tamilnadu

PDF

Essays on agrarian institutions

PDF

Income inequality and economic growth: A theoretical and empirical analysis

PDF

Changing role of the household as an income equalizer: An empirical study of Kuznets hypothesis on the case of Taiwan

PDF

Infrastructure investment and regional productivity of the manufacturing industries in Mexico

PDF

Dual labor markets, public debt management, and exchange rate movements

PDF

The economics of dowry

PDF

Before and after reform: The effects of structural adjustment on women's employment in Latin America

PDF

Context is everything: An empirical study of institutions, instability and economic growth

PDF

Complex economic growth

PDF

By sea: The port nexus in the global commodity network (the case of the West Coast ports)

PDF

Knowledge and use of the contraceptive effect of breastfeeding: Evidence from Malaysia and Guatemala

PDF

An economic analysis of cooperative group behavior: Theory and evidence

PDF

Patronage, dependence and debt in peasant agriculture

PDF

A comparative examination of the impact of business -government relations on labor market reform in Egypt and Mexico, 1975--1995

PDF

Essays in environmental policy

Asset Metadata

Creator Rivera, Paul Alexander (author)

Core Title Buscando la prosperidad: Migration as long -term investment in El Salvador

School Graduate School

Degree Doctor of Philosophy

Degree Program Economics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag economics, general,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Nugent, Jeffrey B. (committee chair), Easterlin, Richard (committee member), Hamilton, Nora (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-262878

Unique identifier UC11334760

Identifier 3093812.pdf (filename),usctheses-c16-262878 (legacy record id)

Legacy Identifier 3093812.pdf

Dmrecord 262878

Document Type Dissertation

Rights Rivera, Paul Alexander

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA