International Migration and emittances: Assessing the Impact on ural Households in El Salvador A DISSETATION SUBMITTED TO THE FACULTY OF THE GADUATE SCHOOL OF THE UNIVESITY OF MINNESOTA BY Amy Lynne Damon IN PATIAL FULFILLMENT OF THE EQUIEMENTS FO THE DEGEE OF DOCTO OF PHILOSOPHY Paul Glewwe, Adviser July 008
Amy Lynne Damon 008
Acknowledgements I would like to thank by adviser Paul Glewwe or his kind guidance and mentorship throughout my time at the University o Minnesota. His thorough comments, endless patience, and incredible diligence in getting drats back in a matter o days greatly acilitated the dissertation process. I would also like to thank my committee members Elizabeth Davis, Deborah Levison, obert King, and Benjamin Senauer or ielding questions throughout this process and improving the inal product. I have been lucky throughout my doctoral program to be inancially supported by a number o excellent ellowship programs. I would like to thank the Interdisciplinary Center or the Study o Global Change (ICGC) or inancial support during my irst year o graduate studies and or providing an outlet to exchange ideas across disciplines. Further, I would like to thank the Heug amily or supporting my last year o doctoral studies with a Heug-Harrison dissertation ellowship. This additional year o support undoubtedly improved the quality o this research and expanded its scope. In addition, I am grateul or pre-dissertation research and conerence grants rom the Center or International Food and Agricultural Policy, ICGC, and COGS. I would like to thank the BASIS program at Ohio State University or allowing me access to the data used in this dissertation, as well as Daniel Flores at DIGESyC in El Salvador who was instrumental in allowing me access to data used rom his oice. I would urther like to thank Febe Saravia, Sherry Stanley, and Nelson Escobar, who i
helped shape some o the nascent ideas relected in this dissertation with many long conversations about El Salvador. I would like to thank my riends, study partners, and uture colleagues in the department, Minh Wendt, Daisy Liu, Le Pham, Steve Dehmer, and my long time oicemate and riend Kenji Adachi, as well as many others. And lastly, thank you to my amily, Mom, Dad, Lisa, and Brad, and riends, Carmen, Nadine, Meagan, Kelli, Hadas, Doug, Zhao, Jonathan, who made chilly Minnesota a warm and riendly place to live. ii
Table o Contents Acknowledgements...ii Chapter. Introduction. Chapter. Data and Descriptive Statistics.. Chapter 3: Household Labor Allocation in emittance-eceiving Households: The Case o El Salvador...35 Chapter 4. Estimating Land Use and Crop Supply esponse to Migration and emittances in El Salvador..88 Chapter 5. International Migration and emittances: A Driver o Asset Accumulation and edistribution in ural El Salvador?...3 Chapter 6. Conclusions and Future esearch 35 iii
List o Figures Figure. Causal elationships Between Migration and Outcomes o Interest 9 Figure. emittances, Agriculture, and Industry share o GDP, 99 005... 0 Figure.3. Percent o Households that eceive emittances by Municipality, 004 0 Figure 4. GDP Shares o Crop Values, El Salvador 99 003... 08 Figure 4. Percent o Employment in Agriculture, El Salvador 994 004 08 iv
List o Tables Table. Salvadoran Population in USA and El Salvador.. 7 Table. Ten states with the largest oreign-born population rom El Salvador, 000.... 7 Table.3 emittances eceived rom Abroad in El Salvador, 99 005.. 8 Table. Attrition Probit... 8 Table. Household Migrant and emittance Status: 995, 997, 999, 00. 30 Table.3 Descriptive Statistics.. 3 Table.4 Income by Source, 996, 998, 000, 00 3 Table.5 Migrant Characteristics, El Salvador 00.. 3 Table.6 Comparison o Household Characteristics between Migrant and Non-Migrant Sending Panel Households 33 Table.7 Descriptive Statistics and Means Comparison... 34 Table 3. Description o Variables Used in Empirical Analysis... 78 Table 3. Probit Models Explaining Migration, El Salvador, 996, 998, 000, 00..... 79 Table 3.3 Cross-sectional Model o emittances (in US dollars), El Salvador, 00.... 80 Table 3.4 Panel Data Model o emittances (in US dollars), El Salvador, 996, 998, 000, 00... 8 Table 3.5 Fixed Eects Instrumental Variable esults or On-Farm Labor Hours in Levels, El Salvador, 996, 998, 000, 00.. 8 Table 3.6 First Dierenced Instrumental Variable esults or On-Farm Labor Hours in Levels, El Salvador, 996, 998, 000, 00 83 Table 3.7 Fixed Eects Instrumental Variable esults or O-Farm, Sel- Employment, and Housework Labor Hours in Levels, El Salvador, 996, 998, 000, 00... 84 Table 3.8 First Dierenced Instrumental Variable esults or O-Farm, Sel-Employment, and Housework Labor Hours in Levels, El Salvador, 996, 998, 000, 00... 85 Table 4. Description o Variables Used in Empirical Analysis... 09 Table 4. Instrumental Variable egression Explaining Agricultural evenue Coeicient o Variation, El Salvador, 996, 998, 000, 00.. 0 Table 4.3 Fixed-eects instrumental variable regression explaining land use by migration status, El Salvador, 996, 998, 000, 00... Table 4.4 Fixed-eects instrumental variable regression explaining land use by remittances received, El Salvador, 996, 998, 000, 00... Table 5. Description o Variables used in Empirical Analysis.. 6 Table 5. Land Area Held Explained by emittances, El Salvador 996, 998, 000, 00.... 7 Table 5.3 Land Area Held Explained by Migration Status o the Household, El Salvador 996, 998, 000, 00.. 8 Table 5.4 Land ented In in Hectares (emittances), El Salvador 996, v
998, 000, 00... 9 Table 5.5 Land ented In in Hectares (Migration), El Salvador 996, 998, 000, 00. 30 Table 5.6 ental Activity - ented Out (emittances), El Salvador 996, 998, 000, 00. 3 Table 5.7 ental Activity - ented Out (Migration), El Salvador 996, 998, 000, 00... 3 Table 5.8 Livestock Holdings (emittances), El Salvador 996, 998, 000, 00... 33 Table 5.9 Livestock Holdings (Migration), El Salvador 996, 998, 000, 00... 34 vi
Chapter. Introduction Since the end o the civil war in 99, El Salvador has become increasingly integrated into the international globalized economy. Aside rom high levels o oreign direct investment, and industrialization in general, international labor migration, and the money that lows rom migrants abroad to their amily members who remain in El Salvador (remittances), have become important actors in El Salvador s development process. During the civil war in the 980s, El Salvador experienced out-migration, capital light, and an inlow o remittances. During the period between 978 and 987, El Salvador reported a net emigration o more than 650,000 people (Funkhouser, 99). Many o these emigrants moved to the United States and settled in urban areas. These migrants were typically urban, working age, married men who were better educated than non-migrants (Funkhouser, 99). According to the U.S. census, over 800,000 Salvadorans live in the United States, or 3 percent o the Salvadoran population, see Table.. The migrants rom the 980s established important Salvadoran migrant networks between El Salvador and the United States, which partially explains the high proportion o the Salvadoran population that currently resides in the United States and the importance o migrant remittances to the contemporary Salvadoran economy. Migration and remittances are important to El Salvador s aggregate economy, but remittances and migration have an equally important impact on households economic outcomes. Households that send migrants to the United States oten depend on remittances as a primary source o income, making their household welare inextricably linked to the migration process. Migration and remittances are receiving increasing - -
attention in the economics literature given their importance in many developing countries. This growth in research spans many topic areas and relects the many questions yet to be answered about the impact o migration and remittances, such as: How do migration and remittances aect economic development? Do remittances compensate or the accompanying reduction in household labor? Do households invest remittances into productive activities and productive assets? Do more remittances today lead to more growth tomorrow? Do remittances lead to large reductions in rural poverty? Does sending a large portion o a country s working age population to a more productive economy in exchange or remittances, represent a viable strategy or long-term, sustainable economic development? All o these questions are part o the complicated story that weaves together migration, remittances, agricultural development, poverty reduction, population dynamics, and investment, which will in part determine the long-term welare and development prospects or Central America. To begin to understand these questions, improved knowledge about the impact o migration on household labor supply, agricultural production, and agricultural asset accumulation is required. This dissertation explicitly examines these relationships. Policy elevance o Migration and emittances Understanding the economics o migration is crucial to understanding changes in rural welare and development in El Salvador, and also has signiicant and timely policy implications. This dissertation contributes to the understanding o the economics o migration by providing a more complete body o knowledge on Salvadoran migrants to - -
the U.S., which can be used to make both immigration policy in the U.S. and Salvadoran rural development policy more eective. International migration rom El Salvador to the United States has grabbed the attention o development practitioners, politicians, and policy advocates in both the United States and El Salvador. Salvadoran policy makers promote migration because they see remittances as a way to break the cycle o persistent poverty in the region and generate oreign exchange. The U.S. immigration debate, which continues to be a topic o political interest, recognizes the importance o immigration as a source o low wage labor or U.S. companies. However, at the same time, U. S. policy makers are concerned about the negative implications o migration (i.e. illegal migration, labor exploitation, depression o low-skilled wages impacting U.S. workers, and criminal activities associated with illegal migration) and push or increasingly strict immigration enorcement and border controls to stem the low o undocumented immigrants across the Mexican-U.S. border. Both U.S. and Salvadoran policy makers need more inormation about the shortand long-term implications o their respective policies. Central American policy makers, or example, need better inormation about the possible long-term consequences o international migration on economic development. U.S. policy makers, on the other hand, need more inormation about how direct investment in rural communities in El Salvador (and other migrant sending countries in Central America) could potentially lessen the tide o immigrants rom the region. This dissertation will help ill this There are many examples o governmental promotion o migration. Central American governments continually lobby Washington DC or more permissive immigration policies. - 3 -
knowledge gap by providing detailed inormation about the impact o international migration on rural productive activities and household labor allocation in El Salvador. Conceptual Map o the Problem To visualize the relationships between poverty, migration, remittances, agricultural production, and household labor allocations consider the diagram presented in Figure.. This diagram creates a schematic o the problems and relationships investigated in this dissertation project. The analytical problem that is immediately apparent in this conceptual diagram is that this system is endogenous. It may be the case that the decision to migrate is made jointly with agricultural production and labor supply decisions (which are determinants o household poverty levels), and that, as shown in the diagram, agricultural production and investment inluence the decision to migrate. However, it is possible to identiy other exogenous causal actors in the migration process. Namely opportunities to migrate, such as inormal social networks existing in the U.S. measured by historical migration rom that town, as well as wage dierentials between the U.S. and El Salvador that induce migration. This dissertation examines the household decision making process to send a migrant (step and step ) as well as agricultural and household labor outcome o interest (step 3 to step 4 in Figure.). Speciically, migration impacts agricultural productivity by decreasing available amily labor and relaxing credit constraints. Economic theory would suggest that a decrease in available amily labor can be compensated by hiring wage labor. However, hiring wage labor assumes that labor markets are unctioning in the migrant originating communities, and that wage labor is a The problem o endogeneity will be more thoroughly addressed in subsequent chapters. - 4 -
substitute or amily labor. For example, i the primary agricultural decision maker o the household migrates, it is unlikely that hired unskilled labor will be a perect substitute or amily labor. The relationship between migration, remittances, agricultural investment and household labor supply is complicated by the act that once a migrant leaves the household, remittances rom that person typically increase initially, but decrease to zero at some point in the uture. emittances thereore may be only temporary relie rom poverty since lows oten decrease over time (Orozco, 004). Another dynamic component to consider is that migration leads to more migration because migration can develop social networks, which will acilitate more migration. The temporary nature o remittances impact and the establishment o migrant networks, in the long run, may increase the vulnerability o migrant sending households to shocks, because the labor supply o those households is reduced, and their supplementary source o insurance and credit through remittances decreases over time. However, i remittances are initially invested into productive activities that generate income over time, this vulnerability would be mitigated. Trends in Migration and emittances Salvadoran immigrants live throughout the United States, but over 50 percent o this population is concentrated in Texas and Caliornia, as shown in Table.. The Washington, D.C metro area including Virginia, Maryland, and the District o Columbia also has a large percentage o Salvadoran immigrants. Salvadorans in the United States remain connected to amily members and their communities in El Salvador through a - 5 -
number o mechanisms, the most important o which are the remittances they send to individuals or households in El Salvador 3. emittances rom migrants have increased dramatically since the Chapultepec Peace Accords were signed in 99, oicially ending the civil war. Salvadorans living abroad sent $US858 million in remittances to El Salvador in 99. 4 This number has steadily increased over the past 5 years, and in 006 El Salvador received $US3. billion in remittances. 5 Table.3 documents the remittance levels and growth o remittances between 99 and 006. Over this time period remittances increased by 58 percent. The GDP shares or remittances, agriculture, and industry are presented in Figure.. In 005, remittances accounted or about 7 percent o the GDP in El Salvador. This share o the GDP has remained airly stable over the past 5 years; however, it has grown signiicantly relative to agriculture s share o GDP. In 998, remittances overtook agriculture as a share o GDP. However, even through agriculture represents a decreasing portion o the Salvadoran economy, in 006 about 40 percent o the population was considered rural 6. The rural economy remains an important determinant o household welare in El Salvador (DIGESTyC, 006) 3 emittances may also be sent rom other areas such as the capitol city, San Salvador, to a rural household. In 00, 8 percent o remittance-receiving households received remittances rom domestic migrants. The model presented in Section 3 does not distinguish between domestic remittances in the descriptive statistics, however in the empirical analysis only international remittances are considered. 4 These remittance igures are a lower bound, since a signiicant proportion o remittance lows occur through inormal channels and such remittances are not accounted or in the oicial statistics. 5 emittance igures here are in nominal dollars, however given that the exchange rate was pegged to the dollar beore 00, and the economy was dollarized ater 00, it is expected that the nominal and real igures are airly close. 6 The term rural is not necessarily synonymous with agricultural, however agriculture remains the major economic activity in the rural areas. - 6 -
Figure.3 shows the geographic concentration o households that receive remittances. This igure shows that the eastern and northern areas o El Salvador are the most likely to receive remittances. These areas were largely rebel strongholds during the civil conlict throughout the 980s and were hardest hit by violence in the civil conlict. This intense violence created massive out-migration during this period and created a system o migrant networks rom these areas which reduced the cost o migration or uture migrants..0 LITEATUE EVIEW Migration and Development The Harris-Todaro approach While explanations o migration have a very long history in economics, the irst explicit exploration o migration s role in economic development was given by Todaro (969) and Harris and Todaro (970). Classic economic development theory considered labor migration out o agriculture to be a key component in economic development, both historically in industrialized countries and currently in developing countries (Bardhan and Udry, 999). It assumed that development requires the spatial transormation o society, rom a dispersed rural sector to concentrated urban sectors, in order to provide labor or an industrial economy (Bardhan and Udry, 999). The Harris-Todaro (HT) model assumes that potential migrants compare the expected utility o migrating to urban areas with the expected utility o remaining in the countryside to determine their migration decision (Todaro, 969; Harris and Todaro, 970). The HT model is a two sector (urban and rural) model in which rural to urban labor migration is a result o expected income dierences between the two sectors. Since - 7 -
the introduction o the HT model, many extensions have been developed to account or inter-sectoral labor migration (Amano, 983), intersectoral capital mobility (Neary, 98), endogenously determined urban wage rates (Calvo, 978 and Stiglitz, 975), and welare implications o the HT model (Bhagwati and Srinivasan, 974, 975; Corden, 974). While many o the underlying assumptions in this model are still widely accepted, it is also recognized that the migration process is not as clean as required or this model. Speciically, a major assumption o the HT model is that migrants maximize their individual utility by migrating to the labor market with the highest expected value o income. However, in the early 980s a strand o literature, reerred to as the New Economics o Labor Migration (NELM) was pioneered by Oded Stark (99) 7 and Stark and Bloom (985) which addressed the assumption that migration is an individualistic decision process. The NELM model considers the migration decision to be a joint decision between the migrant and the amily. The ew Economics o Labor Migration The NELM was an attempt to move the economics proession away rom the perception that labor migration was solely a response to spatial dierences in expected income as proposed by the HT model. Stark developed this new theory based on three main premises. First, he suggested that migration was not a result o individual optimizing behavior, but rather the rational behavior o a group, such as a amily. Given the collective optimization o household welare, consequences o migration such as remittances were calculated in the migration decision, not simply accidental byproducts. 7 This reerence is to a book by Oded Stark printed in 99, which is based on the NELM strand o literature he pioneered in the early 980s. - 8 -
The NELM also extends standard human capital theory, which posits that the perormance o migrants in the destination labor market is an outcome o skill levels and endowments. The NELM suggests that the preerences and constraints o the sending household are important actors in the determination o destination labor market perormance. Given that the decision to migrate is a amily decision, a cooperative arrangement must be made or migration and remitting to take place, thereore amilial considerations o intra-amilial trade in risk, coinsurance arrangements, devices to handle principal agent problems, moral hazard problems, and contract enorcement all inluence the migrant s perormance in the destination labor market (Stark, 99). Second, Stark contends that migration is not simply a response to labor market wage dierentials. ather, he suggests, people assess their relative wealth within a given reerence group and are induced to migrate i they are relatively worse o than their peers. This theory, known as the relative deprivation theory, implies that relative deprivation will be important actors in the decision to migrate (Stark 984, Stark & Taylor 99, Stark 99). Stark and Taylor (99 & 989) test the relative deprivation hypothesis in the case o Mexico, and ind supporting evidence that relative deprivation is a signiicant explanatory actor in international migration between Mexico and the U.S. Third, the NELM posits that migration is a unction o missing (or undeveloped) markets in a given area. One o the consequences o missing markets in developing countries is that rural households cannot access markets that would allow them to invest, diversiy, and beneit rom the processes o industrialization These households must - 9 -
instead migrate to capture these beneits 8. ozelle et al. (999) applied the NELM ramework to the case o China, and concluded that, in China, missing capital or insurance markets induced migration. This paper will be discussed in more detail in the subsection addressing agriculture and remittances. The NELM ramework has contributed to an improved understanding o the economics o migration, however there is some evidence that the NELM ramework is most applicable when migrants leave the household or a short period o time, speciically as a survival strategy, and return to the household relatively quickly (Sana and Massey, 005). The NELM its the case o Mexico, due to the relatively low cost o migration, and the historical short-term migration pattern. However, it has been shown to be less applicable in cases like El Salvador, Guatemala, or Nicaragua, where migration during the 980s was predominantly a result o civil conlict and household members generally migrated or indeinite periods o time (Sana and Massey, 005). As a result o a dierent historical trajectory, accompanied by a high cost o migration, household members rom Central America tend to leave or a longer period o time, with less chance o eventually returning to the household. Given this distinction, this thesis builds on the NELM, but does not necessarily ocus on testing it explicitly, as was done in other studies. Intra-household Decision Making about Migration The decision to migrate, according to the NELM, is a joint decision made within the household, and is likely made in light o labor market conditions, household 8 For example, a rural amily in the United States has access to beneits o the industrialization process through the stock market. In rural areas in most developing countries this type o mechanism does not exist (Stark, 99). - 0 -
characteristics, and characteristics o the potential migrant. In most analyses, household structure and remittances are assumed to be exogenously determined, or given. However, it is logical that households are simultaneously deciding earning and spending decisions along with who migrates. This implies that household composition is determined through a series o strategic decisions made by the household. Maitra and ay (003) ind some evidence o this trend in South Arica. They ind that households determine their structure based on pensions received rom the government. Maitra and ay provide a ramework or examining the process o household composition decision making and pension lows. This paper is relevant because it provides a ramework or the determination o who migrates based on a low o income, which could be analogous to the determination o who migrates based on expected remittances. The decision to undertake migration likely depends on the migrant s earning potential in the destination labor market and their probability o remitting. Briere et al. (00) explicitly consider the role o household composition on the motivation o migrants to remit. They hypothesize the migrants remit as a means o insurance, or an investment in a uture bequest, and ind that emale migrants in the U.S. oten ulill the role o insurer and only when the male is the only migrant or a given household do they send remittances as insurance. Further they ind that investment in bequests is done equally by males and emales, but only by international migrants. The out-migration o working age adults has important consequences or households let behind. For example, when rural households send all o their working age members, this depletes the direct care available or aging household members, or whom remittances oten cannot compensate or reduced direct care, given the lack o - -
medical acilities in these areas. Thereore, changes in household composition rom migration have ripple eects not limited to agricultural production. Understanding how and why households send migrants is an important intermediate step in understanding how remittances contribute to labor allocation decisions and agricultural development. For example, with increasing emigration o the working age population rom Central America, an increasing number o households have no or ew working age members. The household composition ater migration may result in ewer acres planted, less labor intensive crops and lower yields. Further, these households may be reluctant to invest in agriculture, as a result o other household demands on remittances such as health care and other basic needs. Given the importance o household composition in the story o agricultural productivity and household labor allocation, this dissertation explicitly considers household composition in its analysis. Migration, emittances, and Development International migration research has yet to reach a consensus on the impact o remittances on development and poverty. There is some evidence that remittances lessen the depth and severity o poverty, and households that receive remittances are less likely to ace extreme poverty (Adams Jr. and Page, 005; Taylor et al, 005). Supporting the thesis that remittances have a positive eect on economic development, Durand et al. (996) argue that the large portion o remittances spent on consumption items in Mexico has a signiicant multiplier as well as productive investment eect in the Mexican economy, and that these eects are oten omitted rom analyses considering remittances and development. - -
Other authors have questioned the positive role o remittances in development. There is some evidence that remittances have an inlationary price eect and hence reduce the purchasing power o non-recipients (Bracking, 003). Bracking, however, largely ignores the potential local positive wage eect rom out-migration. There is some evidence that remittance inlows increase demand in the local labor market, but only i the local economy transmits remittance eects rom remittance-receiving households to non-receiving households 9 (Taylor, 999). In the case o Mexico, given the existing credit and risk constraints, remittances were ound to have a high shadow value or arm households. However, the impact o remittances on arm-household income was ound to be dierent across households and dependent on a household s initial asset stock (Taylor and Wyatt, 996). While that study does use a two survey panel data set collected in 983 and 989, the sample size is only 55 households in a single region. This small regional sample prevents results rom being generalizable on a national level. elationships between migration, remittances, poverty and development have commanded increasing empirical attention in the economics literature in recent years. These once largely ignored inancial lows are now increasingly recognized as an important inluence on a variety o household outcomes and behaviors. With ew exceptions, this body o work has not integrated rigorous theoretical models to guide empirical analysis. There have been several studies testing the NELM ramework but more work on theoretical considerations is needed to integrate migration and remittances into the arm household model. That said, empirical evidence has provided some important general insights into the impact o remittances on household behavior. Adams 9 I the village economy is integrated regionally and labor and inputs can come rom neighboring towns, the positive labor eect is lost (Taylor, 999). - 3 -
and Page (005) suggest that remittances have a mitigating eect on poverty in general at the household level. Stark et al (986) report that remittances exacerbate inequality at the village level where there are ew international migrants, and have an equalizing impact on income distribution where there are large numbers o international migrants. Edwards and Ureta (003) investigate the inluence that remittances have on schooling or children let behind and ind that children in households that receive remittances are more likely to stay in school. Overall, the migration literature has evolved rom describing o rural to urban migration in the case o the Harris-Todaro model, to examining and testing the NELM speciically examining eects o migration and remittances. This dissertation also builds on the NELM theory by examining the eect o credit or insurance constraints on a household and how migration and remittances overcome these constraints to aect labor allocation or agricultural production activities. Literature relevant to each speciic question and this dissertation s contribution to that literature is presented in each chapter. esearch Question and Objectives My research questions examine which households migrate and the ways in which migration and remittances aect household labor supply, agricultural production, and agricultural assets in El Salvador. () Which households choose to migrate and what determines remittance amounts? Household with an international migrant are not randomly selected rom a given population. These households that engage in migration have certain characteristics, some observable and some unobservable, that provide incentives or opportunities or migration. - 4 -
Further, the migrants themselves have a set o characteristics that determine the amount o remittances that are sent back to the household. It is important to understand these characteristics determining both migration and remittances to better understand the outcome o migration. Further, investigating the determinants o migration and remittances allows or an investigation o appropriate instrumental variables to be used in explaining agricultural and labor outcomes. () How do remittances and migration aect the allocation o household labor supply? Household members make labor supply decisions or both on-arm and o-arm work allocations. These decisions are inluenced by relative wage rates between on-arm and o-arm work and other activities. The aect that remittances and migration have on household labor allocation depends on the separability that exists between production and consumption in the household. I the household operates in and environment o perectly unctioning markets then an increase in remittances will result solely in an increase in consumption. However, i a household is credit constrained then separability breaks down and an increase in remittances will change on and o arm labor supply allocations. (3) How are agricultural production, agricultural land use, and agricultural assets aected by the receipt o remittances? The NELM suggests that remittances can provide both relie rom credit constraints and substitute or missing insurance markets. As such, remittances, depending on the existing asset portolio, may inluence the level, type, and orm o - 5 -
agricultural production 0. The receipt o remittances could result in several possible scenarios. Two non-exhaustive scenarios presented as examples are: () The household receiving remittances could invest in improved production practices (i.e. increased inputs or appropriate technology), thereore increasing yields, or () the household could invest in cash crops or livestock. The research objectives o this thesis are the ollowing: (i) (ii) Determine the characteristics o migrant-sending households. Test how receipt o remittances aects the allocation o household labor supply. (iii) Determine how remittances contribute to agricultural production and investment. The rest o this dissertation is structured as ollows. The second chapter reviews the data that will be used and provides relevant descriptive statistics. The third chapter examines the characteristics o migrant sending households and determinants o remittances as well as the relationship between household labor allocations and migration and remittances, the ourth looks at agricultural production decisions and migration and remittances, and the ith explores the impact that migration and remittances have on agricultural asset accumulations. The inal chapter provides a summary and conclusions as well as prospects or urther research. 0 There is some overlap between this question and the previous labor allocation question, but the labor allocation question is more expansive and concerned with labor allocation decisions beyond agricultural labor allocation. - 6 -
Table. Salvadoran Population in USA and El Salvador Salvadoran Population 990 000 In El Salvador 5,0,76 6,76,037 In USA 565,08 87,336 As a % o Population in El Salvador % 3% Source: DIGESTyC, 006; U.S. Census Data, 000 Table. Ten states with the largest oreign-born population rom El Salvador, 000 Area umber Percent United States 87,336 00 Caliornia 359,673 44 Texas 0,59.4 New York 76,977 9.4 Virginia 54,704 6.7 Maryland 4,88 5. New Jersey 8,595 3.5 Florida 3,65.9 Massachusetts 8,606.3 the District o Columbia 5,886.9 Nevada,080.5 Source: Migration Inormation Source, 006-7 -
Table.3 emittances eceived rom Abroad in El Salvador, 99-005 Year Total emittances (US$ in millions) Percentage Change emittances as a % o GDP 99 790.0 4.9% 99 858.30 8.6% 4.4% 993 864.0 0.7%.5% 994 96.50.4%.9% 995,06.40 0.3%.% 996,086.50.4% 0.5% 997,99.50 0.4% 0.8% 998,338.30.6%.% 999,373.80.7%.0% 000,750.70 7.4% 3.3% 00,90.50 9.% 3.8% 00,935.0.3% 3.5% 003,05.30 8.8% 4.0% 004,547.60.0% 6.% 005,830.0.% 6.7% Source: Banco Central de eserva de El Salvador, (Central eserve Bank o El Salvador), 006-8 -
Figure. Causal relationships between migration and outcomes o interest Step Step Step 3 Step 4 US Salvadoran wage dierentials Migration opportunities, (eg. social networks, US immigration policy) Poverty Migration Changes in household composition emittances. Agricultural production and investment. Household labor supply 3. Other outcomes health, education etc. - 9 -
Figure. emittances, Agriculture, and Industry share o GDP, 99 005 30.00% 5.00% 0.00% 5.00% 0.00% 5.00% 0.00% 99 99 993 994 995 996 997 998 999 000 00 00 003 004 005 emittances Agriculture Industry and Manuacturing Source: Banco Central de eserva de El Salvador, (Central eserve Bank o El Salvador), 006 Figure.3. Percent o Households that eceive emittances by Municipality, 004 Source: EHPM 00 004, Chapter 5 UNDP Human Development eport El Salvador - 0 -
Chapter. Data and Descriptive Statistics. The data used or the empirical analysis in this study is a our year panel data set collected in El Salvador in 996, 998, 000, and 00. The Salvadoran Foundation or Economic and Social Development (FUSADES) along with the World Bank collected the irst National ural Household Survey in 996. The second and third rounds were implemented by FUSADES in cooperation with the ural Finance Program at The Ohio State University (OSU) under the Broadening Access and Strengthening Input Markets System (BASIS) program in El Salvador. Ater the completion o the BASIS program, OSU and FUSADES conducted a ourth round o data collection in 00 to complete the our year panel (Pleitez-Chavez, 004). The 996 survey was designed as a stratiied random sample that is representative o all rural areas in El Salvador. The initial sampling plan was based on the 99 labor orce census, which called or 9 land-using households and 436 rural worker households, constituting a stratiied sample o 68 households. However, to allow or more precise analysis on agricultural production activities, an additional 0 land using households were added. As such, the total sample size or the 996 survey was 738 households. These households represent all departments (states) in El Salvador. The questionnaire design was based on the World Bank s Living Standards Measurement Study (LSMS) surveys, with detailed modules on education, health, occupation, assets, income generating activities, credit transactions, land transactions, agricultural In Spanish: Fundación Salvadoreña para el Desarrollo Económico y Social Land-using households were deined as households owning or renting more than 0.35 hectares o land - -
production activities, household labor allocation, hired labor, and migration and remittances. The 998 sample resurveyed 494 households rom the original primary sample o 68. The attrition rate between 996 and 998 was 4 percent. Originally the survey team intended to interview all 68 households, but it was only able to re-interview 469. The panel was supplemented by re-interviewing 5 o the 0 additional land using households added to the irst round o the panel (Conning et al, 00). Given that the 996 panel was stratiied by departments, the 5 replacement households were chosen rom residents o the same department as those households they were replacing (Pleitez- Chavez, 004). In 000, the households were again re-interviewed. In the 000 round all 469 households interviewed in 998 were located and re-interviewed. This process was undertaken again in 00, at which time 45 households were located. This resulted in a panel data set that includes inormation on 45 households or all our years with a cumulative attrition rate o 8 percent between 996 and 00. The high attrition rate between the irst and second round is largely because the original survey was not intended to be the irst round o a panel survey, so data that would be useul or inding households in later years was not collected. Further, it is possible that some attrition may be due to migration itsel, which could introduce bias that is not detected in the attrition probit analysis. Given the attrition rate in the panel, Table. presents a series o probit estimates to investigate i there are any non-random characteristics o attritors. This table provides an attrition probit or the total attrition over the our year panel as well as an attrition probit between 996 and 998, 998 and 000, as well as between 000 and 00. - -
esults rom these models suggest that larger households are more likely to remain in the panel over the eight year time rame, while households with children and those in the department o Cuscatlán are less likely to remain in the panel. Individual year results largely support the overall results, except or the result in the 000 to 00 model that suggests that emale headed households are more likely to stay in the sample between these years. Generally, these results indicate that attrition rom the survey was not systematic and should not be a signiicant source o bias in the empirical analysis. Two other datasets that provide instrumental variables or estimation purposes are also used. Wage and unemployment rates in American cities are calculated rom the U.S. Current Population Survey using only non-resident Latino workers in the migrant s destination city. In addition, the Salvadoran national household survey collected by the Salvadoran National Census oice (DIGESTyC) or 996, 998, 000, and 00 is used to calculate the percent o households and individuals in each municipality 3 that received remittances in that year. Descriptive Statistics This section presents basic descriptive statistics relevant to the analyses presented in the ollowing chapters. Table. shows the number o households in each year o the panel, the number o households with international and national migrants 4, and the number o households that received remittances. esults rom Table. suggest that the number o migrants in the panel remains approximately stable (although because o 3 Where municipal estimates are unavailable, department level estimates are used. 4 All migrants reported by the household are considered migrants in this study, there is no cuto in terms o duration o migration in considering the person a migrant. - 3 -
reverse migration the individual household with migrants change over the sample years), with the exception o 998 where the number o migrants decrease rom 49 to 97 between 996 and 998. However, in percentage terms this drop only represents a 4 percent drop in the number o migrants because the total sample drops between 996 and 998. The number o households that received remittances increases rom 4 percent to 30 percent between 996 and 00. In Table.3 indicates that the number o migrants included in the sample increases rom 36 in 996 to 43 in 00 when using only households which remain in the survey all 4 years. Overall, the household size decreases, the average number o senior citizens per household increases, and the average number o children within each household increases. This is consistent with previous demographic studies, which suggest that Latin America is experiencing an increase in the proportion o its population who are senior citizens, partly due to out-migration (Brea, 003). The percentage o households with migrants increases over the eight year period rom 35 percent reporting at least one migrant in 996 to 43 percent in 00. The percentage o households that receive remittances increased rom 8 percent in 996 to 35 percent in 00. The level o remittances, conditional on receiving remittances, in 00 dollars, also increased rom $US,000 in 996 to $US,700 in 00. These descriptive statistics suggest that migration and remittances as an income earning strategy became increasingly popular in El Salvador between 996 and 00. Table.4 provides a detailed description o income and income sources or the our panel years. emittance income as a percentage o total household income increased rom 8 percent in 995 to 4 percent in 00. However, it is important to note that this - 4 -
includes households that do not receive remittances. The income share o remittances is higher when households that receive no remittances are excluded. Table.5 describes the migrant characteristics in 00. In terms o opportunities and wages in the labor market, results suggest that males on average earn more than emales. Males earn on average $US 378 per week and women earn $US 39 per week. Unemployment rates also dier slightly. Male migrants have an unemployment rate o 4.5 percent and emale migrants have an unemployment rate o 4.8 percent. More than hal o migrants are reported to be either the son or the daughter o the household head. Nearly 0 percent are the brothers o the household head and another 5 percent are sisters. Less than 5 percent report the household head as the migrant. The last 0 percent are other relations such as in-laws, grandchildren, mother/ather, etc. Tables.6 and.7 describe the data used in the empirical analysis and provide descriptive statistics comparing remittance-receiving households with non-remittance receiving households. Given the panel structure o the dataset, the data are described in each o the our years survey data was collected and are limited to the 45 households that remained in the panel or all our rounds. Variables in tables.6 and.7 are described by migrant and non-migrant household means and the total sample means and provided as well or a pooled sample o all household over all years. A simple t-test is conducted between non-migrant and migrant household means. Even though these t-tests ail to control or a number o potentially important inluencing actors, they provide useul inormation about the dierences between migrant and non-migrant households. Comparing amily characteristics in Table.6, we see that migrant households have larger households, more senior citizens, and are more likely to be emale headed. - 5 -
Interestingly, we do not observe a dierence in the education levels or number o children in the household between these two groups. It is not surprising that migrant households are larger and have more senior citizens, given that these households are more likely to be able to withstand a loss o amily labor and migrants are typically at the beginning o their working lives, leaving the older generation behind. Migrant households also have signiicantly lower dependency ratios 5, which is somewhat contrary to expectations, given it is generally the working age amily members who migrate. In terms o labor allocation, these simple comparisons suggest that migrant-sending households have lower o-arm work hours, or males, emales, and total. However, onarm work hours are not aected by migration according to these simple comparisons 6. Only emales have higher on-arm work hours in migrant-sending households but this dierence is only signiicant at the 0 percent level. We will see that these results are not completely consistent with empirical results presented in chapter 3. These results provide only preliminary evidence o the dierences between migrant-sending and non-migrantsending households given that these comparisons do not control or other actors that likely aect the outcomes o these comparisons. In terms o agricultural production, migrant households apparently engage in livestock production more readily than non-migrant households, as seen in Table.7. Migrant households exhibit a higher value o their livestock as well as higher value o 5 The dependency ratio is deined as the number o people below 6 and over 65 years o age, divided by those between 6 and 65 years o age. 6 Given that we are not controlling or other actors in these t-tests, since migrant households have larger land holdings, hours per hectare o land may be signiicantly dierent between migrant and non-migrant households. - 6 -
animal products derived rom this livestock. Interestingly, these two groups do not dier in the value o crops produced. It appears that migration may have little impact on input use at the arm level. The percentage o households that use improved seeds, inorganic ertilizer, and pesticides does not dier across migrant and non-migrant households. Migrant and non-migrant households may dier in their land holdings and the use o these holdings. Migrant households have a higher average number o hectares, at.5 or migrant households, and.3 or non-migrant households. Migrant households also have a higher average acreage in uncultivated land, rented out land, rented in land, and land dedicated to the homestead, as well as land in pasture. It seems that, on average, migrant households cultivate more land. The mean area dedicated to basic grains and other cash crops are signiicantly higher or migrant households, but the mean hectares dedicated to coee production do not vary signiicantly by migration status. While these descriptive tables oer a preview o the data and some inormation about relationships between certain variables they do not oer inormation about causality between migration and outcome variables. More in depth empirical work is presented in subsequent chapters to examine the impact o migration and remittances on household behavior and outcomes. - 7 -
Table. Attrition Probit Variable Total Sample 996-998 998-000 000-00 Household Size 996 0. *** 0.86 *** -0.4-0.364 ** Standard Error 0.04 0.043 0.094 0.54 Migrant-sending household 996-0.5-0.7-0.07-0.6 Standard Error 0.095 0.098 0.7 0.37 Senior Citizen in Household 996-0.055-0.04 0.088 0.344 Standard Error 0.0 0.4 0.54 0.364 Child in Household 996-0.59 *** -0.3 0.075 0.374 Standard Error 0.068 0.070 0.56 0.55 eceived emittances 996 0.000 0.000 0.000 0.000 Standard Error 0.000 0.000 0.000 0.000 Hours o on-arm work 996 0.000 0.000 0.000 0.000 Standard Error 0.000 0.000 0.000 0.000 Hours o o-arm work 996 0.000 0.000 0.000 0.000 Standard Error 0.000 0.000 0.000 0.000 Dependency atio 996 0.5 0.43 0.078-0.396 Standard Error 0.095 0.097 0.38 0.339 Livestock value 996 0.000 0.000 0.000 0.000 Standard Error 0.000 0.000 0.000 0.000 Land Area 996 (Hectares) -0.03-0.08-0.05-0.04 Standard Error 0.00 0.00 0.033 0.066 Female headed household 996-0.07 0.095-0.34.47 *** Standard Error 0.85 0.9 0.554 0.379 Santa Ana -0.03-0.36-0.79-0.596 Standard Error 0.34 0.4 0.450 0.54 Sonsonate -0.63-0.60 0.387 Standard Error 0.44 0.5 0.407 Chalatenango 0. 0.33 0.58-0.639 Standard Error 0.86 0.309 0.459 0.687 La Libertad -0.035-0.38 Standard Error 0.30 0.36 San Salvador 0.95 0.5-0.350-0.404 Standard Error 0.5 0.65 0.47 0.548 Cuscatlan -0.575 ** -0.647 ** -0.04-0.48 Standard Error 0.9 0.93 0.600 0.703 La Paz -0.3-0.50-0.833 Standard Error 0.7 0.77 0.775 Cabanas -0.40-0.336 0.35 0.463 Standard Error 0.87 0.94 0.50 0.565 San Vicente -0.67-0.8 0.5 Standard Error 0.30 0.37 0.647 Usulutan 0.084-0.6 Standard Error 0.49 0.54 San Miguel -0.007-0.05 0.073-0.498 Standard Error 0.37 0.46 0.4 0.55 Morazan -0.0-0.33 0.03 Standard Error 0.75 0.83 0.548 La Union -0.07-0.6-0.386-0.09 Standard Error 0.40 0.47 0.539 0.54 Constant -0.706-0.383 -. 0.345-8 -