Migration and the Sending Economy: A Disaggregated Rural Economy Wide Analysis

Transcription

1 Department of Agricultural and Resource Economics University of California, Davis Migration and te Sending Economy: A Disaggregated Rural Economy Wide Analysis by J. Edward Taylor and George Dyer Working Paper No November by J. Edward Taylor and George Dyer All Rigts Reserved. Readers may make verbatim copies of tis document for non-commercial purposes by any means, provided tat tis copyrigt notice appears on all suc copies. Giannini Foundation of Agricultural Economics

2 Migration and te Sending Economy: A Disaggregated Rural Economy Wide Analysis J. Edward Taylor Department of Agricultural and Resource Economics University of California, Davis George Dyer El Colegio de Mexico and te University of California, Davis Abstract Most economic researc on migration impacts in source economies focuses on te ouseolds tat send migrants and receive remittances, ignoring linkages tat transmit migration s influences to oters in local and regional economies. Tis paper offers an alternative, disaggregated economy wide perspective on migration and its impacts. It presents and illustrates a metodology to understand not only migration s effects on migrant-sending ouseolds, but also te ways in wic tese ouseolds transmit influences of migration to oters in te source economy, via local market linkages. Data from te 2003 Mexico National Rural Houseold Survey are used to calibrate a series of interacting rural ouseold models nested witin a general equilibrium model of te wole rural economy. Tis modeling approac combines te strengts of micro models focusing on rural ouseolds wit economy wide models, wic igligt economic linkages among economic actors but traditionally ave been implemented at an aggregate (national or multi-national) level. It explicitly takes into account te market structures tat govern economic interactions and promote or retard te spread of migration effects witin sending economies. Simulations reveal tat te impacts of international migration and remittances on sending areas may be positive or negative and depend critically on te ways in wic local markets transmit impacts among ouseolds. University of California, Davis, Department of Agricultural and Resource Economics Working Paper. We are greatly indebted to te William and Flora Hewlett Foundation and Te World Bank for supporting various aspects of tis researc and field work. Taylor is Professor and member of te Giannini Foundation of Agricultural Researc, and Dyer currently is a visiting researcer at te University of California, Davis.

3 Migration and te Sending Economy: A Disaggregated Rural Economy Wide Analysis Rural out-migration is at te eart of te economic development process (Lewis), and migration is te principal mecanism troug wic ouseolds in less developed countries (LDCs) become inserted into te international economy. A significant social science literature addresses interactions between migration and development in migrantsource areas (Massey, et al., 1998; Taylor and Martin, 2001). Te centerpiece of economic researc in tis area is te new economics of labor migration (NELM), wic conceptualizes te migrant as part of a larger social group, usually te family or ouseold. Te NELM provides a useful analytical starting point to understand ow migration and remittances transform te economies of ouseolds wit migrants in ways tat were overlooked by previous migration researc (see work by various autors in Stark, 1991). Te basic premise of tis paper is tat we need to go beyond a focus on migrant ouseolds in order to understand te complex ways in wic migration resapes migrant-sending economies. Houseolds, like migrants, are part of larger social groups, e.g., communities, regions, and nations. Troug teir market interactions, migrant ouseolds transmit te impacts of migration to oters witin tese groups. Non-migrant ouseolds become affected by migration troug teir interactions wit migrant ouseolds or wit ouseolds tat interact wit migrant ouseolds, etc. Because of tis, a ouseold does not necessarily ave to ave a migrant in order to be affected by 1

4 migration. In fact, it is possible tat most of migration s impacts on sending economies are found outside te ouseolds tat send te migrants and receive te remittances. Tis paper proposes a disaggregated economy-wide modeling approac to explore te direct and indirect influences of migration in migrant-sending regions and ten uses tis approac to study te impacts of migration under alternative market scenarios. Te model is calibrated using data from a nationally representative survey of ouseolds in rural Mexico. Our findings illustrate ow survey data and new modeling tecniques can permit te analysis of migration impacts previously overlooked by migration researc. Tey also reveal te sensitivity of international migration s impacts to local market conditions and differences in te effects of migration and remittances in te sort and long runs. In general, market integration in migrant-sending areas mutes and dissipates te impacts, bot positive and negative, of migration and remittances. In one scenario tat we explore, te rural economy is caracterized as a collection of interacting agents like tose caracterized in te staple agricultural ouseold model of Sing, Squire and Strauss (1986). All markets, including tose for labor and capital, are assumed to be perfect, and all ouseolds are price takers. In tis arguably unrealistic market environment, negative impacts of migration are minimal, and positive impacts are limited to remittances contribution to ouseold budgets, tat is, a direct transfer effect. 2

5 In a more realistic scenario, rural wages are endogenous and affected by migrationinduced sifts in te rural labor supply. Upward pressure on rural wages creates negative lost labor effects of migration on rural production activities, particularly tose most intensive in labor. Te existence of rural nontradables ads a new set of potential migration impacts, as canges in supply and demand alter te prices of tese goods and services. Migrant remittances potentially generate bot sort and long run production and income effects. In te sort run, remittance-induced canges in demand may influence prices of nontradables. Tis transmits te effect of remittance transfers to production, as iger prices stimulate production of nontradables (possibly at te expense of tradables) and nominal (but not necessarily real) incomes in rural ouseolds increase. If remittances are not sufficient to counterbalance negative lost-labor effects of migration on production, rural general-equilibrium price effects will tend to reinforce negative lost labor effects on production. In te medium to long run, income gains from migration may loosen liquidity constraints on investing in production activities if rural capital markets are imperfect tat is, if ouseolds ave to self-finance teir investments. In te inter-temporal version of our model, investments in one period are permitted to increase capital stocks in future periods. Tis usually (but not always) stimulates production and incomes, as posited by NELM researc (e.g., Stark, 1991). 3

6 Remittances represent a direct effect of migration on rural ouseold incomes. Wen a policy or market sock, e.g., currency devaluation or an increase in destination wages, stimulates remittances, te direct effect on incomes in migrant-sending areas is positive. Indirect effects include general-equilibrium wage and price effects, dynamic investment effects, and te migration response to bot te initial sock and its wage, price and investment ramifications. Our findings suggest tat tese indirect effects are likely to be substantially larger tan te direct effects on wic researcers and policy makers tend to focus. Te rest of tis paper is organized as follows. Part 1 briefly summarizes te new economics of migration perspective, and it motivates and proposes te use of a disaggregated economy wide modeling approac to explore te likely impacts of migration in rural economies. Part 2 presents te model. In Part 3 te model is used to simulate te effects of canges in international and internal migration under alternative rural market scenarios. Conclusions appear in Part 4. 4

7 1 Migration and te Transformation of Migrant-Source Economies Te Transformation of Migrant Houseold Economies Typically, migrants do not sever teir ties wit te source ouseold after tey migrate. Family members wo remain beind (often, parents and siblings) may reorganize bot teir consumption and production activities in response to te migrant s departure, and migrants (often, cildren) typically sare part of teir earnings wit teir ouseold of origin, troug remittances. Continuing interactions between migrants and rural ouseolds suggest tat ouseold-based approaces are more appropriate tan individual-level models of migration decisions. Indeed, in individual-focused migration models, tere is no rationale for migrants to sare teir earnings wit te place of origin; we are left wit te puzzle of wy geograpically extended families are prevalent in LDCs but less so in ig-income countries (Rosenzweig, 1988) and few insigts into te likely impacts of migration and remittances on source economies. In wat as become known as te new economics of labor migration (NELM; see Stark, 1991 and Stark and Bloom, 1985) view, migration decisions are not entirely te domain of individuals. Tey take place witin a larger context typically te ouseold, wic potentially consists of individuals wit diverse preferences and differential access to income and tey are influenced by tis social milieu. Te perspective tat migration decisions are not taken by isolated actors but by larger units of related people, typically by ouseolds or families, is a trademark of te NELM. So is te contention tat people act 5

8 collectively not only to maximize income, but also to loosen constraints on investment tat are created by a variety of market failures, including missing or incomplete capital, insurance, and labor markets. Tat is, migrants play te role of financial intermediaries in imperfect-market contexts, wic caracterize most of te world s rural economies A number of studies find evidence in support of te basic tenets of te NELM (Lucas, 1987; Taylor, 1992; Rozelle et al., 1999, Taylor et al., 2003). A study of migration from ouseolds in rural Cin found tat, in almost all production activities, te loss of labor to migration reduced net income. However, migrants generated remittances, and remittances, in turn, significantly increased net incomes in bot farm and self-employment activities. Houseold-focused studies provide an incomplete picture of migration-development interactions, overlooking influences of migration beyond te migrant-sending and remittance-receiving ouseold. Beyond Migrant Houseolds: Te Transformation of Migrant-Source Economies Nearly all economic researc on migration focuses on ouseolds. However, ouseolds are part of larger economies, of communities, regions, and nations. Migrant ouseolds interactions witin tese economies project migration s impacts to oter ouseolds. For example, if a migrant ouseold uses remittances to finance a new project in te village, it may demand labor from anoter village ouseold. If te migrant-sending ouseold ad not participated in migration, it would not invest in te 6

9 project, and te oter ouseold would not ave a market for its labor (or oter goods or services) in te village. Investing may not be limited to te migrant ouseold, if tere is some kind of local credit market (formal or informal) to cannel savings among ouseolds. For example, a ouseold tat does not ave a migrant in te United States could borrow from a ouseold tat obtained capital from migrant remittances. Many or possibly most of te impacts of migrant remittances on migrant-sending economies may in fact lie outside te ouseolds tat send migrants and receive remittances; demand linkages create income multipliers and transmit impacts of remittances from remittance-receiving ouseolds to oters in te local economy, and ultimately, to regional commercial centers. Tese income growt linkages, in turn, sape future migration. Migration may unleas a variety of oter general-equilibrium effects on migrantsending economies. For example, if te loss of labor to migration drives up local wages, rural economies may restructure temselves around labor scarcity, sifting to less laborintensive (and more capital-intensive) activities and production tecnologies. Access to markets sapes te interactions, direct or indirect, between migrant and oter ouseolds in source economies. Access to a foreign or domestic migrant labor market is a prerequisite for migration to appen in te first place. Te existence of local markets creates te possibility for linkages to transmit impacts from ouseolds wit migrants to oters in te sending economy. Hig transaction costs do te opposite. For 7

10 example, ig costs of migrating due to lack of information, etc. inibit some ouseolds from sending migrants abroad or to domestic urban centers. Hig costs of transacting in product markets limit te ability of ouseolds to market teir output or obtain inputs. Missing or incomplete credit markets prevent capital from being canneled to its most efficient uses in rural areas and constrain rural ouseolds to selffinance investments; e.g., in te NELM, te same ouseold tat does te migrating must also do te investing. Imperfections in rural commodity and factor markets may affect migrationdevelopment interactions negatively in some cases, positively in oters. For example, ig transaction costs in regional or national markets for goods or labor limit te possibilities for rural areas to benefit from regional trade integration, possibly intensifying migration pressures. However, tey also create local market linkages tat transmit migration s impacts bot positive and negative to oters in sending areas, namely, tose wo supply to te protected local market. For example, a migrationinduced sift in labor supply may drive up wages in migrant-sending villages, benefiting wage earners. Higer incomes in migrant ouseolds increase te demand for services, pusing up prices and creating benefits for suppliers of te service (to te detriment of consumers). If ig transaction costs isolate ouseolds from low prices in regional grain markets, grain farmers capture rents by producing for local demand. Migration, wic injects new income into te local economy troug remittances, may increase te demand for grain for uman consumption, feed for livestock, etc. Tis may create income gains for local grain farmers. Lack of access to regional credit markets may adversely affect 8

11 investment in rural areas wile creating rents for local moneylenders possibly including individuals wit migrant remittance savings. If migrants play te role of financial intermediaries wo obtain capital in te form of remittances for teir ouseolds, ten rural capital market imperfections may result in too muc migration. More generally, imperfections in rural markets for goods, factors, credit, insurance, etc., may intensify pressures for out-migration. Te importance of transaction costs and market structure is recognized by some micro-economic studies of ouseold-farm economies (de Janvry, Fafcamps, and Sadoulet, 1991; Strauss, 1986; Rozelle, Taylor, and de Brauw, 1999), wic explore ouseold beavior in te absence of selected markets. Recently, a nascent body of researc as begun to explore local impacts of policy canges using disaggregated or micro economy-wide modeling tecniques (Dyer, Boucer and Taylor, 2006; Taylor, Yúnez-Naude, and Dyer, 2005). Tese studies take into account te ways in wic agricultural ouseolds and firms interact in local markets, even wen ig transaction costs isolate tem from larger, regional and global markets. Understanding direct and indirect interactions between migration and development in migrant-source economies requires an approac tat goes beyond te traditional ouseold focus, elucidating market structures and te complex linkages tat connect ouseolds witin te regional economies of wic tey are part. Part 2 presents a disaggregated economy wide modeling approac designed for tis purpose. 9

12 2 Te Model Understanding te impacts of migration in migrant-sending economies requires understanding micro responses, te complex linkages among economic agents witin te economy, and te linkages between te rural economy and te outside world. Most economic researc on rural economies entails te microeconomic modeling of individual agents economic beavior, in most cases te beavior of agricultural ouseolds. However, tat approac misses te linkages among economic agents tat indirectly integrate rural people into global markets. Te modeling approac used in tis study nests individual micro models of four rural ouseold groups into a general equilibrium model of te wole rural economy. Te ouseolds in te model include: (1) landless rural ouseolds; (2) subsistence and infra-subsistence ouseold farms wit fewer tan 2 ectares of land, typical of smallscale, low productivity agriculture, frequently operating under marginal conditions and incomplete markets; (3) medium-older (2-5 ectares) commercial ouseold-farms, and (4) large-older (more tan 5 ectares) commercial ouseold-farms. Tis typology works well to describe te socioeconomic landscape of rural Mexico. Te Houseold Models Agricultural ouseold models are te building blocks of te disaggregated rural economy-wide model. In aggregate CGE models, ouseolds receive income by selling teir labor and oter factors to sared production activities, and tey spend tis income 10

13 on consumption goods. In our disaggregated model eac ouseold type as its own production tecnologies and expenditure demands. Te disaggregated model includes four different tecnologies to produce maize, from subsistence farmers using ox-andplow tecnology to relatively capital-intensive commercial farmers. Not only do tese ouseolds use different tecnologies to produce maize; tey also are engaged in oter production and labor market activities tat vary from one ouseold to anoter. Tey ave different access to domestic and/or international migration, reflecting findings of past migration researc tat access to migrant destinations is network-driven and ouseold-specific (e.g., see Massey, 1987; Taylor, 1987). Houseolds are assumed to maximize teir utility from consumption goods, bot ome-produced and purcased, subject to cas income, tecnologies, time, access to migration, and self-sufficiency constraints tat set consumption equal to production for subsistence maize ouseolds. Te solution yields a set of demands for labor and land inputs into eac activity, including migration, and consumption demands. For tradable goods and factors, prices are set by outside markets. For nontradables, prices are exogenous to individual ouseolds but determined by te interaction of supply and demand in rural markets. Endogenous rural prices result wen transaction costs are ig outside but not witin te rural economy. In our model tere are two suc prices: wages and prices of nontradable services. As in te textbook model (Sing, Squire and Strauss, 1986), land is a fixed input and tus implicitly as a sadow price tat varies across ouseolds. 11

14 General-Equilibrium Effects General-equilibrium closure equations at te levels of te ouseold and rural economies determine te (net) marketed surplus of tradable commodities as te difference between supply and demand. Prices for tradables are exogenous (marketed surplus from bot individual ouseolds and te rural economy is endogenous). Tey are determined by markets outside te rural economy (e.g., world markets) or by policy. Prices of rural nontradables are endogenous. Tey satisfy local market-clearing conditions (rural marketed surplus is zero), and individual ouseolds are price takers witin te rural economy (ouseold marketed surplus is endogenous). It migt be argued tat tere is an active labor market in rural Mexico linked to internal and international migration. However, significant variation in te agricultural wage across te countryside suggests te existence of market imperfections generating local wages or at least wage rigidities. Access to migrant labor markets is not uniform; it is geograpically concentrated and saped by networks of family contacts at migrant destinations and oter local and ouseold-specific variables (Munsi, 2003). Daily agricultural worker wages in Mexico ranged from 50 to 140 pesos in summer Lacking a regional focus, we cannot explicitly model tese wage variations. Neverteless, some of our market scenarios reveal te sensitivity of migration impacts to assumptions concerning ow rural wages are determined. 1 Analysis of Mexico National Rural Houseold Survey (ENHRUM) data, described below. 12

15 Te solutions yield, for eac ouseold group, a set of core equations for outputs, input demands, migration, consumption demands, and eiter rural prices (for nontradables) or marketed surplus (for tradables). In te case of goods, marketed surplus is output minus ouseold consumption. In te case of labor, it is te ouseold s labor demand minus its labor supply, or net wage-labor supply. All of tese variables are functions of te ouseold caracteristics used to construct te four ouseold groups, as well as of exogenous market and policy variables (prices of tradables, te peso-dollar excange rate, and government transfers). Te base model reproduces a social accounting matrix (SAM) for eac ouseold group as well as for te wole rural economy. Once tese base models are created, tey are te starting point for eac of our policy experiments, reported below. Te equilibrium structure of te model insures tat, once te model adjusts to a policy or market sock, te new solution will also produce a consistent set of SAMs. Potential impacts of migration in te model are complex because of te interacting agricultural ouseolds, eac wit its own production tecnology, market access, and consumption demands. Tey reflect direct effects (i.e., te contribution of remittance transfers to ouseold incomes) as well as indirect effects tat result from te interactions among migration, remittances, wages and prices. Signing as well as quantifying te total impacts of migration and remittances on sending economies, terefore, cannot be done analytically; it requires a programming approac. A summary of equations in te model appears in te Appendix. 13

16 Data Te model was calibrated using data from te 2003 Mexico National Rural Houseold Survey (Encuesta Nacional a Hogares Rurales de México, or ENHRUM). Tis survey provided detailed data on assets, socio-demograpic caracteristics, production, income sources, and migration for a nationally representative sample of rural ouseolds surveyed in January and February INEGI, Mexico s national information and census office, designed te sampling frame to provide a statistically reliable caracterization of Mexico s population living in rural areas, tat is, in communities wit fewer tan 2,500 inabitants. For reasons of cost and feasibility, populations in amlets or disperse populations wit fewer tan 500 inabitants were not included in te survey. Te result is a sample tat is representative of more tan 80 percent of te population tat te Mexican government considers as rural. Detailed data on ouseold-farm production, wage work, and migration make it possible to estimate net activity incomes for eac ouseold in te sample, including from crop, livestock, commerce, nonagricultural activities (composed of andicrafts, village nonfarm enterprises, small-scale food processing, and various oter ome-based production activities), services, wage labor (agricultural and nonagricultural), and migration (internal and international), as well as from public transfers (PROCAMPO subsidies for basic grain producers and PROGRESA welfare payments). Tis list of incomes is exaustive; te sum of income from te twelve sources equals ouseold total net income. 14

17 Net incomes from production activities were calculated as te gross value of production (using observed local prices) minus purcased inputs. Gross income from livestock production was estimated as te cange in value of standing erds between te end and start of te survey year, plus sales of animals and animal products and ome consumption of ome-produced animals and animal products, minus livestock purcases and input costs (feed, medicines, and oter costs). Hired labor is rarely observed in livestock production. In te few cases were it was observed, it was included as a production cost wen estimating net family income. Incomes from nonagricultural production activities were estimated in a manner analogous to net crop income (as gross value of production minus purcased input costs). Calculated in tis way, net incomes represent total family value added in eac activity. Migrant remittances were summed across all remitters and, in te case of dollar-denominated remittances from te United States, transformed to pesos using te prevailing excange rate. Table 1 presents summary statistics for eac of te four ouseold groups in our model. Houseolds in rural Mexico on average ave just under 4 members, including 0.6 cildren 12 years or younger. Average scooling of ouseold eads is 4.5 years; owever, scooling of cildren is iger tan tis; tus, te average for all ouseold members is 5.5 years. Sixteen percent of rural ouseolds ad at least one family member (ead, spouse of ead, or cild of ead or spouse) working in te United States in Twenty six percent ad an internal migrant. Average landoldings were 4.8, ectares but wit a large standard deviation (25.1), reflecting a ig degree of landlessness. 15

18 Landed ouseolds tend to be sligtly larger tan landless ouseolds (see bottom 4 panels of Table 1). Tey also tend to be younger and more educated. Te age of te ouseold ead ranges from 48.6 to 55.6 years in te tree landed groups, compared wit 45.3 in te landless group. Average scooling of ouseold eads is 3.6 to 4.0 years in te landed ouseolds compared wit 5.0 years in te landless. Migration propensities are sligtly lower in landless tan in landed ouseolds. Te percentage of ouseolds wit at least one migrant in te United States is 14.4% for landless and 16.2% for all ouseolds. Te percentage wit an internal migrant is 20.4% for landless and 25.7% for all ouseolds. Te wealt index, centered on zero for te full sample, is sligtly larger for landless ouseolds (0.43). Incomes in rural Mexico are igly diversified (see Table 2). Most rural ouseolds engage in crop production. However, te sare of net income from crop production in total income ranged from less tan 1% (for landless ouseolds) to 15.3% (for te largest landolding group). Livestock accounted for anoter 0.9% to 6.9% of total rural income. Most agriculture and livestock-producing ouseolds also received income from several oter sources, including wages, migrant remittances, public transfers, and non-agricultural production. By far te largest rural income sare was from wages (43.1% to 64.1%). Only 13% of wages were from agricultural work, reflecting rural ouseolds integration wit non-agricultural labor markets. Remittances from international migrants accounted for between 6.4% and 11.2% of rural income. Internal-migrant remittances accounted for 2.1% to 7.2% of income. Houseolds in rural Mexico received between 2.1% and 11.1% of teir income from government transfers. 16

19 Te largest sare went to te largeolder group, wo are major beneficiaries of Mexico s PROCAMPO program. Te largest sare accrued to te landless group, consisting mostly of PROGRESS welfare payments. 2 Rental income constituted between 3.4% and 4.4% of total income for te landed groups but was negligible for te landless group. From Table 2 it is evident tat income sources vary sarply across te four ouseold groups. Predictably, te sare of total income from crop production activities is igest for largeolder ouseolds. In contrast, te smallolder and landless ouseolds receive most of teir income from wage work, mostly off-farm. Te sare of bot international and internal migrant remittances is igest for te middle landed ouseold group. Tecnologies and expenditure patterns also vary across ouseold groups. Factor intensities are reflected in factor sares of value-added. In general, te largeolder ouseolds ave te igest capital value-added sares in staple and cascrop production. Tese ouseold expenditure patterns are critical in saping te size and direction of consumption and investment demand linkages in Mexico s rural economy. Large budget sares for locally produced goods create a potential for ouseold income canges to stimulate local production activities. For nontradables, local prices transmit canges in demand to production activities. For tradables, prices are determined in markets outside 2 Te PROCAMPO program replaced price supports wit direct subsidy payments to basic grain producers, to facilitate te transition from price supports to freer and more open international markets. PROGRESA gave needs-based payments to poor rural ouseolds. 17

20 te local economy. Tus, local demand does not affect production, but it does determine te size of te net surplus available to outside (e.g., urban or export) markets. Table 3 reports expenditure sares by ouseold group. Te budget sare of omeproduced crops is minimal for landless ouseolds (0.5%) and igest for te smallolder group (13.0%). Te sare for ome-produced livestock products ranges from 1.0% (landless) to 3.3% (large-older). To te extent tat landless ouseolds consume crop and livestock products, tey do so primarily by purcasing tese goods in stores inside (commerce) or outside te village. Investments in education, ousing, and oter goods constitute up to one-fourt of total expenditures in rural Mexican ouseolds. Te largest investments are on education, wic represents between 7.6% and 12.5% of total budgets of te four ouseold groups. Te largest education sare is in landless ouseolds, for wic uman capital is te key asset. Housing investments absorb between 2.7% and 4.8% of total expenditures, and oter investments (savings and production) constitute between 5% and 6% in landless and smallolder ouseolds and between 9% and 10% in te two largest landolder groups. All data were organized into a series of SAMs, one for eac rural ouseold group. Tese were ten integrated into a rural sector mega-sam. Te SAM provides te data input into te micro economy-wide, computable general equilibrium model. 18

21 Tradables, Nontradables, and Local Price Transmission Goods or factors are tradables if outside prices are transmitted perfectly into te economy tat is being modeled in te present case, te rural economy represented by our model. Perfect price transmission requires well-functioning rural markets wit low transaction costs. Te extent to wic outside prices are transmitted to rural ouseold groups is empirical, ranging from perfect to nil. Goods or factors tat are tradable for one ouseold group may be nontradable for anoter (see de Janvry, Fafcamps and Sadoulet, 1991). Wen a good or service is tradable for a given ouseold group, its price is determined eiter in rural markets or exogenously (e.g., in te world market). To date, tere as been little effort to test empirically for price transmission in rural economies (an exception is Rozelle, 2002). Tere is evidence, owever, tat imperfect price transmission results in endogenous rural prices as well as sadow prices of nontradables for some ouseold groups. For example, in Mexico, were te government supported prices for basic grains in te 1990s, ig transaction costs prevented most farmers from benefiting from tis opportunity to sell to te government (e.g., see Dyer, et al., 2006). Te extent to wic price supports may ave influenced local prices indirectly (e.g., troug traders) is not known. In our simulations we explore te sensitivity of impacts of international migration to assumptions regarding te structure of rural markets, specifically, ow prices of factors, goods and services are determined. Four alternative market scenarios are considered. In Scenario I, all prices, including te rural wage, are assumed to be 19

22 exogenous, determined outside te rural economy. Tis scenario represents a rural economy-wide analogue to te perfect-markets agricultural ouseold model tat is a staple of microeconomic researc (e.g., see Sing, Squire and Strauss, 1986). Scenario II assumes tat all output prices are given but tat te rural wage is endogenous. Te rural wage, wic is exogenous to eac ouseold, is determined endogenously by te interaction of te supply and demand of rural labor. In Scenario III, wages are exogenous but prices of rural services are determined endogenously, witin te rural economy. For tese nontradable rural services (from construction to aircuts), eac ouseold group is a price taker, but te rural market price of te service is determined endogenously, by te interaction of supply and demand. Scenario IV combines Scenarios II and III. In all scenarios land is a fixed factor for all ouseold groups. Most land in te surveyed ouseolds is ejido (reform-sector) land. Because land is treated as fixed, te rental rate is implicitly an endogenous sadow price tat varies from ouseold to ouseold in te model. Te structure of te model permits us to explore te impacts of migration on production, incomes, and trade in a rural economy wit diverse ouseold activities and tecnologies. Te four market scenarios make it possible to also explore te sensitivity of findings to market structures. Findings indicate tat market structures are critical in determining bot te sign and magnitude of international migration s impacts in sending economies. 20

23 Equations in te Model and Calibration Eac ouseold model nested witin te rural model as 5 components: ouseold-farm (1) production, (2) income and (3) expenditures; (4) a set of general equilibrium closure equations, wic insure tat rural markets clear or else te rural economy is a net buyer ( importer ) from or seller ( exporter of marketed surplus) to outside markets, and (5) a price-determination component. For rural tradables, prices are exogenous. For rural nontradables (labor and/or services in Scenarios II-IV), prices are endogenous equilibrium prices, at wic rural supply and demand are in balance. To operationalize our model, a ouseold version of a social accounting matrix (SAM) (Stone 1986; Pyatt and Round, 1979) was constructed from survey data for eac ouseold in te sample (te data are described in te next section). Data in te ouseold SAMs were used to calibrate te individual ouseold models. Eac ouseold model is in effect a CGE corresponding to te ouseold group. Its factor and consumption demands depend on tecnology and preferences. Cobb-Douglas production exponents in te ouseold-specific production functions were set equal to measured factor sares in value added, as implied by profit maximization. Tis requires putting values on family inputs and on output, wic is not always straigtforward in te imperfect market environments caracterizing LDC rural areas. Te Mexico survey obtained information on te quantities of all factor inputs (land, family labor, ired labor, animal power, tractors and oter pysical capital) in crop 21

24 and noncrop production, and also on prices wenever market excanges took place. Market prices were usually observed for intermediate inputs. Te cief exception was maize seed, wic usually was selected from te previous year s arvest. We valued seed and crop output at teir prevailing local market prices. Valuing labor and tractor services (most ouseolds did not own teir own tractor) is straigtforward, using market prices. Most families used teir own animal power in production; owever, in a few cases animal services were ired, and we used te observed prices to value all animal capital inputs. Subtracting te costs of intermediate inputs, ired labor, and mecanical and animal capital services from te gross value of production yields a residual representing family labor and land value-added. Ejido land could not be rented at te time of te survey; tis resulted in wat essentially was a missing land rental market, wic means tat land rents ad to be imputed to obtain land value-added. Family labor and land value-added sares were imputed econometrically, from regressions of total (land plus family labor) value-added from crops and livestock on land and family labor inputs. Houseold-farm savings and expenditures were estimated using a linear expenditure system (LES) approac wit no minimum required quantities (Deaton and Muellbauer, 1980). Tis specification implies tat preferences are described by a Cobb-Douglas utility function. Te parameters in te demand equations were set equal to measured budget sares for eac ouseold and eac good. 3 3 Te budget and factor sares for te subsistence good were obtained by valuing tis good at a sadow price equal to its observed per-unit cost of production; see Becker (1965). 22

25 Tese relatively simple specifications for te production and demand functions are sufficient to demonstrate te transmission of migration effects in rural economies. Tey imply simple linear demand functions for inputs and consumption and are easy to estimate compared wit more complex functions. Even wen individual ouseoldgroup responses are linear, aggregate outcomes of market socks are nonlinear, saped by ouseolds production and demand parameters and endogenous prices. 4 Many ouseolds received remittances from migrants in te United States or from internal migrants, mostly in nonfarm jobs. Aggregate CGE models assume tat migration occurs until wages at te migrant origin and destination equilibrate (or tat te disparity in wages remains te same before and after te policy socks tat te researcers simulate in teir models). From a micro ouseold-farm point of view, tis is te wrong migration condition to impose. Houseolds may allocate teir members labor eiter to local production (obtaining te value of te marginal contribution of te member s labor to local production) or else to migration (obtaining migrant remittances). We estimated remittance elasticities econometrically, from a regression of te log of ouseolds migrant remittance receipts (net of amortized migration costs) for ouseolds tat participated in migration on te log of teir labor allocated to migration, controlling for family migration networks and correcting for potential sample selectivity bias. Tese 4 We ave found te results of policy experiments using similar models to be robust wit respect to ow we specify functional forms, including more complex production and expenditure functions wit assumed elasticities. Tis is not surprising, inasmuc as te model is always estimated at te same point given by te survey data, and policy experiments involve marginal canges in exogenous variables. 23

26 elasticities were allowed to vary across ouseold groups in te village-town model, reflecting differences among ouseolds in access to foreign and domestic migrant labor markets. Te migration equilibrium condition in our model is tat te value of te marginal product of labor in rural production activities equals te marginal effect of migration on remittance income (and te rural wage). Equations in te village-town model are summarized in te Appendix. 3 Rural Economy-wide Impacts of Migration and Remittances Estimated influences of migration on local economies depend bot on te structure of markets, wic transmit influences among ouseolds, and on weter tese influences are considered in te sort or long run. We begin by simulating te sort-run effects of a 10% increase in te economic returns from internal and international migration on te rural Mexico economy under eac of te tree market scenarios. We ten explore te long-run effects of migration, using a multi-period model. Te difference between tese two sets of simulations is tat te second is designed to capture dynamic impacts of migration via investments, as posited by te new economics of migration. Most of te sort-run impacts involve adjustments of local production to lost migrant labor (in Scenarios II and IV), to canging output prices (in Scenarios III and IV), and to increases in ouseold income, troug remittances and te new demand for locally supplied goods and services tat tey stimulate (in all scenarios). 24

27 Rural Economy Wide Impacts of Migration in te Sort Run Tables 4 and 5 report te sort-run effects of a simulated 10% increase in te economic returns from Mexico-to-U.S. migration, as migt result from currency devaluation or an increase in wages or employment for migrants in te United States. Tis simulation was carried out by increasing te sift parameter in te foreign remittance functions by 10%. It produces tree sets of effects on te rural economy. Te first is a direct transfer effect on te ouseolds tat send migrants abroad: remittances by migrants already abroad increase by 10%. Second, tere is an allocation effect, as iger returns from international migration encourage ouseolds to sift more labor into international migration (and away from oter labor activities tat compete wit international migration). Tird, were permitted (in Scenarios II, III and IV), canges in labor availability and iger remittances influence wages and te prices of nontradables, and tis creates general-equilibrium effects on production, incomes, and investments by ouseolds tat do not send migrants to te United States. A large part of tese effects leak out of te rural economy, via linkages wit outside (regional) markets in Mexico. Altoug te initial impacts of te migration increase are in ouseolds wit migrants, te rural economy-wide model captures te transmission of tese impacts from migrant to non-migrant ouseolds in te local economy, as well as to te rest of Mexico, troug trade. Local demand linkages depend critically on te degree to wic te migrantsending ouseolds are integrated wit outside markets. Wit ig levels of outsidemarket integration, as in Scenario I, te loss of family labor to migration stimulates te 25

28 demand for labor substitutes, and remittances stimulate ouseold demand for market goods. However, rural wages and prices are not affected, because tey are assumed to be determined in outside markets. Te more closely integrated ouseolds are wit outside markets, te more te impacts of remittances are diffused to tose markets and te smaller te impacts on ouseold production. In te case of a micro-economic model of pure commercial agricultural ouseold tat is a price taker in all markets, as in Sing, Squire and Strauss (1986), migration and remittances do not ave any effect on production, by assumption. Scenario I depicts tis peraps extreme case. In scenario II, were te rural wage is endogenous, migration may influence production in all ouseolds via an increase in wages. Tis general-equilibrium effect is ignored by most microeconomic models of agricultural ouseolds. 5 In te presence of rural nontradables (e.g., services), migration and remittances may ave more extensive effects on production as well as on consumption, and te sign of tese effects is ambiguous. Tese effects are explored under Scenarios III and IV. In any ouseold tat does not ave access to perfect credit and insurance, migration also may ave important effects on production in te long run, via investments. Dynamic effects are explored in our inter-temporal simulations (see next section). Under Scenario I te effects of te 10% increase in returns to international migration are unambiguously positive for income and nil for production. Te direct transfer effect of remittances (tat is, ignoring te migration response) increases te average real income of landless ouseolds by 75.5 pesos, or 0.83%; of smallolder 5 One of te few exceptions is Braverman and Hammer (1984). 26

29 ouseolds, by pesos (1.22%);of medium olders, by pesos (2.06%), and of large olders, by 93.6 pesos (1.14%; see Column A in Table 4). Te migration effect is sligtly larger tan te direct transfer effect (Column B). It raises real incomes in te four groups by anoter 334 pesos (in small older ouseolds) to pesos (in medium older ouseolds). Te total direct transfer plus migration effects range from pesos (in landless ouseolds) to pesos (in medium older ouseolds). Higer income stimulates te demand for village-produced food (by 5.3%) and nonfood (4.9%) items, as well as for tradables purcased outside te village, primarily in nearby commercial centers (by 4.8%). It also increases rural ouseolds investments (by just under 5%), more tan alf (5.13%) are in education and ealt, 24.2% in ousing and 16.4% in livestock. Table 5 summarizes production effects under te four scenarios. Under te assumption tat all rural prices are fixed, tere are no linkage effects in Scenario I. Wit all rural prices fixed, no influences of migration or remittances are transmitted to production. Te absence of production effects under Scenario I in Table 5 illustrates tat te separability property of te staple agricultural ouseold model carries troug to te entire rural economy in an economy-wide model tat nests witin it a series of perfectmarkets ouseold models. Tis would be true even if some ouseolds were subsistence producers, as sown by Holden, et al. (1992). Matematically, wen all prices are given, none of te first-order conditions for profit maximization in rural production activities canges as a result of eiter te loss of labor to migration or 27

30 ouseolds receipt of remittances. Te only effects of international migration under tis scenario are on te income and consumption side. In Scenario II, international migration as linkage effects on production (Table 5), but tese are all negative because migration competes wit local production activities for labor, bidding up te rural wage. Te rural wage increases by 5.5%. Wit all output prices fixed, all goods and services are tradable by assumption under Scenario II. Tis means tat output prices do not increase to compensate for te rising cost of labor inputs, and production in all activities falls. 6 Te production activities most adversely affected are tose most intensive in family and ired labor. Crop and nonagricultural production contract by 6.5% and 3.7%, respectively. Livestock output falls only sligtly (by 0.5%), due to its limited use of labor. Te negative impacts of labor lost to migration on local production mirror findings from past econometric studies, wic focus on ouseolds and do not take into account general equilibrium effects (e.g., Lucas, 1987; Taylor, 1992 and Rozelle, Taylor and de Brauw, 1999). Tey reflect a rural economy-wide analogue of te Dutc disease, under wic production of tradables contracts as labor is canneled into an export activity, in te present case, migration. Te most labor-intensive activity competing wit international migration is internal migration, wic decreases by more tan 10%. Houseolds bring back internal migrants to migrate abroad as well as to work for iger wages at ome. 6 A negative effect of iger wages on production is not inevitable in a model wit diverse production activities. Houseolds could respond to iger wages by canneling more resources into activities tat are not labor intensive, and output in tose activities could increase. 28

31 Te contraction in production results in negative income effects for medium and large older ouseolds (see Table 4, Column D). Te general-equilibrium wage ( Linkage ) effect reduces tese ouseolds incomes by 96.6 and 40.2 pesos, respectively. However, iger wages exert a positive effect on te incomes of landless and small older ouseolds. For te latter, wage-linkage effects captured in Scenario II add and pesos of income, respectively. Te total effect of increasing returns from U.S. migration remains positive and large for all groups. In Scenario III, output prices of rural non-tradables cange in response to ouseold demand. Tis creates a positive stimulus for te activities tat produce tese nontradables (Table 5). However, it produces a negative effect on real incomes of rural ouseolds tat consume nontradables (Table 4, Column F). By maintaining te fixedwage assumption of Scenario I, tis simulation igligts te linkage effects of migration tat can result from te presence of non-tradable goods and services in sending economies. Te linkage effects under Scenario III in Table 5 are nil for crops and livestock, wic by assumption are tradables. However, nonagricultural production increases by 1.6%. As remittances increase ouseold incomes and te demand for normal goods, te prices of tese nontradables increase, stimulating production. However, in real terms, te income effects of te increased price of nontradables are all negative, ranging from 17.9 to 31.1 pesos. Te total income effect of increasing returns from international migration are still positive, but tey are smaller tan in te perfect markets scenario. 29

32 Confronted by iger prices for rural nontradables, ouseolds sift consumption to tradables, wic comprise te bulk of Outside Village Consumption in Table 5. Te latter increases by 6.4% in Scenario III compared wit 4.8% in Scenario I. Conversely, village non-food consumption, wic includes te nontradables, increases 2.8% in Scenario III and 4.9% in Scenario I. Te results of Scenario IV generally represent a mixture of tose from Scenarios II and III. Increasing returns from international migration produce negative output effects for tradables (crops and, to a smaller extent, livestock), as in Scenario II. However, tey ave an effect on nonagricultural production tat is positive, as in Scenario III, altoug smaller because of te negative wage effect. Te effect on total real income is iger tan in te perfect markets scenario for landless and small-older ouseolds (492.8 and pesos, respectively; see Table 4, Column I), reflecting te positive wage effect on tese ouseolds incomes. However, because of te iger cost of consuming nontradables, tis real income gain is smaller tan in te pure endogenous wage scenario (541.1 and pesos). Te reverse obtains for te medium and large older ouseolds. Teir total real income increases less in Scenario IV tan in te oter scenarios, reflecting te combined negative effects of iger labor costs and iger consumption prices. As migration increases, ouseolds reorient teir demand towards markets outside te village. 30