Download Remittances' Impact on Return Migration: Longitudinal Analysis in Nang Rong, Thailand and more Study notes Literature in PDF only on Docsity!
INTRODUCTION
As early as 1885, when Ravenstein’s “laws of migration,” stated that each current of migration produces a compensating counter-current, return migration has been acknowledged as important to any thorough understanding of migration. Yet, for many years the view of migration as primarily a one-way phenomenon dominated research studies. More recently researchers have begun to utilize new data sources to empirically examine return migration, and theoretical links have been proposed between migrant remittance (money and goods sent by migrants to their home households) and a migrant’s intention to return home (Lucas and Stark 1985, Hoddinot 1994). However, although this theoretical link been between remittance and return migration has been suggested, various shortcomings in past research designs have left a dearth of high-quality empirical investigations into the subject matter. In this study we aim to fill this gap in the literature by examining the effect of remittance on the return of migrants to their households of origin by using longitudinal panel data from Nang Rong, a rural, agricultural district located in the Northeast Thailand. We take advantage of the richness of Nang Rong data to overcome a number of limitations of past studies by examining the phenomenon over time, using exp licit data on several varieties of both migrant-to-household and household-to-migrant remittance, and by using a longitudinal sample which avoids some of the problems associated with sample selectivity inherent in many existing studies. REVIEW OF THE LITERATURE The literature on return migration has tended to concentrate on both economic and non- economic factors effecting movement back to origin communities. Overall, studies of
return migration are mainly based on individual cost-benefit models that focus on successes and failures at destination as the main reasons for return. Borjas (1989), who infers out- migration from sample attrition in a longitudinal data set of foreign-born scientists and engineers in the United States, concludes that the least successful scientists and engineers are the most likely to leave the sample. However, his data is limited in the sense that he uses a highly selective sample, in which it is impossible to tell whether migrants indeed returned or if they simply migrated to another location. Further, even if they did return, migrants may have had erroneous information about economic opportunities in the destination prior to migration, so their return may not be a failure in the job market so much as a poor choice to start with. A follow- up work by Borjas and Bratsberg (1996), which incorporates variables ignored in Borjas (1989) original analysis (such as job markets and life-cycle plans) comes to a similar conclusion that return migration intensifies the selection that characterizes the original immigration flows. Research by Oropesa and Landale (2000) using data from both origin and destination communities finds that return migration to Puerto Rico from the U.S. mainland is associated with impoverishment, which further supports the skills bias argument. Nonetheless, the evidence for a skills bias in return migration is in no way unequivocal. For instance, Michael Piore’s (1979) book Birds of Passage , examines the role of immigrants in the U.S. labor market, and argues that unskilled migrants are more likely to stay in the U.S. to fill the less skilled jobs in the country’s dual labor market, while successful migrants are more inclined to return home abroad. Shumway and Hall (1996) show that return migrants do not appear to have lower earnings profiles, so return
organization, households are found in nearly all societies. They mediate between individuals and larger social structures (Boyd 1989, Goldscheider 1995), and serve as an important context for a variety of individual behaviors, including: marriage, migration, fertility, and mortality (Entwisle et al. 2003). In contemporary post-industrial societies, households play a crucial role as consumption units in the maintenance and support of their members. In agrarian settings, households play a significant role as production units, and farming is often organized around the household. Research in the New Economics of Migration (NEM) tradition sees migration and remittance as part of a household strategy to diversify risk in the face of incomplete or absent capital, futures, and securities markets (Stark 1991). Stark hypothesizes that migrants play the role of financial intermediaries, enabling rural households to overcome credit and risk constraints on their ability to achieve the transition from familial to commercial production. To overcome such constraints, one or more migrants are sent out to work to make money. Migrants remain a part of their origin household throughout the migration experience, and they remit a portion of their earnings, thereby relaxing the household’s credit constraints. Their return migration may thus represent success in fulfilling their responsibility to their home household. Indeed, past research in many geographic regions such as China (Hare 1999) and Mexico (Roberts and Morris 2004) has suggested that rural- to-urban migrants remit in order to maintain a high degree of attachment to their rural origin communities and households. Also, work by Ahlburg and Brown (1998), using data from a survey of Tongan and Samoan migrants in Sydney, concluded that those who plan to return home remit significantly more than those who do not plan to do so.
While the household’s gain from this household-migrant arrangement is obvious, it is less clear what the migrant gains from such a contract. Therefore, in this study we focus on the migrant’s motivation rather than the household’s motivation. To better understand the migrant’s motive, NEM theorists have developed a conceptual model of remittance, which views remittance as part of a mutually-beneficial, inter-temporal, self- enforcing, implicit contract between a migrant and a household (Lucas and Stark, 1985; Stark and Lucas, 1988). According to the model, this contract is motivated by either altruism or instrumental self- seeking, such as concern for inheritance or the right to return home ultimately in dignity. In this contract, migrant and household use remittance to better each other’s welfare in addition to using remittance instrumentally to pursue personal gains. Instrumental motivations are of three varieties: coinsurance, investment, and promise of bequest. Coinsurance, the first type of motive, occurs when a migrant and household take turns insuring each other from market fluctuations and risky ventures, such as when the household provides a safety net to insure the migrant against involuntary unemployment or when the migrant sends remittance to allow a household to invest in a relatively risky new production technology, such as a high- yield crop variety (Stark and Lucas, 1988). Investment, the second instrumental motive, occurs under circumstances when the household invests in a migrant, such as the financing of a migrant’s education with the anticipation of future returns from accruements to the migrant’s human capital endowments. Alternatively, migrants, in their absence, may be following an investment motive by sending remittance in order to safeguard assets or land in the origin community
It may be that both altruistic and self- interested motivations for remittance depend on whether individuals choose to remain a part of the household at origin. Indeed, the authors also find that remittances by heads of household are substantially and significantly greater than those by other migrants. Headship no doubt reflects strong sense of household membership and a responsibility to care for members of one’s own household. However, simply being a household member at a given point in time is not enough to ensure continued remittance. The authors’ further find that children of the head of household do not remit more than any other members of the household. The fact that the household head’s children do not send remittance may have to do with their membership in other households at the destination community. Thus, the authors conclude that those who continue to be identified as household members are very persistent remitters (Stark and Lucas 1988, Menjivar et al 1998). Hoddinot’s (1994) work in Kenya also shows evidence of both altruistic and bequest motives. However, his work does not rule out the possibility that these are independent motivations, which may be followed as separate strategies by different migrants. Hoddinot finds that the effect of acres of land per adult son is a positive and strongly significant predictor of remittance, which he interprets as an indication that wealthier parents, who can offer a greater reward for remittances, are better placed to extract a greater share of benefits of migration. Thus remittances are affected by the credibility of the parental threat to reduce future bequests. However, the author also finds that elderly widows, who are dependent on transfers for their livelihood, are more likely to receive remittance, a finding that’s more consistent with an altruistic motive.
A common weakness found in all of these studies is the inability of their cross- sectional designs to deal with the endogeneity of remittance with various independent variables. In the Botswana study, for instance, the data come from a single wave of the 1978-1979 National Migration Study (NMS) of Botswana. The problem is that, in dynamic settings, such as Botswana, as well as other geographical areas, one cannot rule out the possibility that past remittance sent with an altruistic intent have helped to raise current household income, which may change the migrant’s attitude to the family property. A similar problem exists in the Ahlburg and Brown (1998) study. Not only do these authors measure plans to move, rather than actual movement, but their research also has the shortcoming that plans to return may be related to the decision to remit at a given point in time. Hence, it is impossible to tell whether remittance affects return migration or vice versa. What is needed is a design that measures remittance prior to return, rather than one that measures return migration and remittance contemporaneously, as is done in most cross-sectional studies. This would argue in favor of data on migration and remittance collected over time, which we use in the present study. Given the empirical evidence and the theoretical model developed by NELM researchers, our research utilizes an explicit household perspective. More specifically, we view migration as part of a household strategy to diversify household income through remittance under an implicit contract with other non- migrant household members. We argue that migrants send remittance because of either altruistic or more self- interested intentions. The altruistic motivation suggests that the migrant feels the responsibility to care for family members, and the latter self- interested motives point to the desire to
implicit contractual obligations, and can return to reap the benefits of such an agreement. Second, relative to migrants who do not remit, all else equal, migrants who do remit are more likely to do so if the household owns fixed capital and land. This is expected because either these migrants desire to inherit property, or because they want to safeguard it during their absence. Having described a general theory and hypotheses, we now consider the distinctive characteristics of our study site and the ways in which contextual factors can mediate or intervene between remittance and return migration. In what follows we describe the research setting, the data, the operationalization of key measures, the methodology used in this endeavor, as well as the operationalization of control variables. SETTING Nang Rong district is located in Buriram province in the southern portion of Northeast Thailand. The district is in the rice-growing delta of the region along the main highway between Nakhon Ratchisima and the Cambodian border (Curran 1995). Although it has experienced rapid economic development in the last several decades, Nang Rong still remains a primarily rural region in which rain- fed patty rice cultivation is the primary economic activity. Nang Rong is similar to many parts of developing countries, in that asymmetric economic development between rural and urban areas stimulates regular flows of rural- to-urban migrants searching for employment opportunities. In the year 2000, 80 percent of Thailand’s population was still living in rural areas. During this time, however, agriculture only made up ten percent of the Gross Domestic Product (GDP), while services made up 50 percent (World Bank 2003).
For the past several decades, the vast majority of job growth was concentrated in urban areas (Curran 1996), and tremendous numbers of migrants flowed into cities to take the increasing job positions. In Thailand, Bangkok and the Central region are major migration destination areas, while the North and Northeast Regions are major sending areas (Jampaklay 2003). Migrants start to leave Nang Rong villages around the age of twelve, when compulsory education ends (Rindfuss et al. 2000), and frequently migration is carried out in conjunction with variations in agricultural labor dema nds. During the agricultural seasons when labor demand is low, migrants often flock to Bangkok in search of work, with flows being particularly heavy during the dry months of March and May (Pejaranonda et al. 1995). Many migrants travel back and forth between their place of origin and their place of destination, thereby keeping close connections with their households and communities. DATA Data comes from the second and third waves of the CEP-CPC longitudinal study of social change in Nang Rong, Thailand. The first wave of data was collected in 1984, when a household survey was administered to all village households in 51 villages. Follow-up waves of data collection occurred in 1994 and 2000, at which time a complete census was again conducted in each of the villages found in the original sample. Each household survey collects data on all permanent residents, as well as proxy reports on anyone who was away at the time of the survey. Data was also collected on migration, monetary and in-kind remittance, land, household assets, and so on.
people under 13 are children, and are probably migrating with their parents. Furthermore, people over 55 have probably ended their working life and are unlikely to change their residence. Second, we restrict our sample to only migrants who have been gone for at least one year prior to the 1994 survey. 2 This is done to ensure that migrants have had enough time to stabilize their economic situation so that they are able to send remittance. Furthermore, because our data on remittance is measured one year prior to the 1994 survey, we ensure that each migrant was at risk of exposure throughout the entire duration^3. We measure return migration as a dichotomous variable equal to one if the migrant returned, and equal to zero otherwise. Table 1 shows the frequency distribution of return migration for the entire sample, for migrants who sent monetary remittance, and for migrants who did not send any money. Results reveal that overall, just under one- fifth (18.47 percent) of all migrants eventually returned. Moreover, compared to non- remitting migrants, a slightly higher percentage of remitting migrants returned (14. versus 21.57 percent). Also, from figure 1, which shows the frequency distribution of return migration for the whole sample, migrants who remit, and those who do not remit, it is also evident that there are more return migrants among those who remit, compared to those who do not remit. Migrant remittance is the key independent variable of interest in this analysis. Like all independent variables in our analysis, remittance is measured in 1994. This design allows us to overcome the aforementioned shortcoming of many studies that measure remittance and migration contemporaneously.
We use data on several types of remittance as well as information of two directions of remittance flow. The data set includes not only migrant-to-household remittance, but also household-to-migrant remittance. The advantage of using two- directional data is that NEM theory has implications for both directions of support, and information might be lost if one direction is ignored. Further, remittance data is available not only for monetary remittance, but also for goods in-kind remittance. Having data on two types of remittance is theoretically interesting because data on goods-in-kind remittance may measure social ties between households and migrants that go beyond general monetary need, and may suggest an awareness of more specific needs of whoever is receiving the remittance. Goods- in-kind data come from a series of survey items that ask whether remittance in the form of clothes, food, household items, electrical appliances, or vehicles was sent in the twelve months prior to the 1994 survey. Goods- in-kind remittance (both household-to-migrant and migrant-to-household) is operationalized as a dichotomous variable equal to one if any of the above items was sent, and zero otherwise. Data on monetary remittance comes from a similar survey item about whether or not any money (measured in Thai Baht) was sent, and was also operationalized as a dichotomous variable. It can be seen from table 2, which shows descriptive statistics for all independent variables, that generally migrant-to-household remittance is more common than household-to-migrant remittance. Also, while just over half of all migrants sent money only about 40 percent of migrants sent goods-in-kind. Household-to- migrant remittance was far less common, with under fifteen percent of migrants receiving any goods, and
to accommodate the modeling of correlated data. GEE is a population average model. The GEE approach allows for covariances among clustered observations and has the advantage of not requiring parametric assumptions about the form of the covariance structures clustered observations (Hardin and Hilbe 2003). In the GEE approach, the β vector is estimated by solving the estimation
equation:
U ( β )= ∑ (∂∂ uβ^ i )'[ v j ( α )]−^1 ( yi − ui )= [0]px1 (2).
Where ui is the expectation of yi , which is linked to a linear combination of the covariates and the corresponding estimate through the logit function (Zhou et al.2003). Efficiency is gained by choosing a hypothesized structure to minimize the within-cluster correlation. We choose an exchangeable correlation structure for this research, which assumes that there is no specific order for each migrant in same household and they are equally correlated within each cluster, which is valid for migrants within households. Under the exchangeable correlation structure, α is a scalar that represents the pair correlation among different migrants within households. The estimated variance is robust for the clustered observations. Since previous research using the Nang Rong data set suggests that the cluster effect at village level is quite small (Piotrowski 2004), in our study, we ignore the cluster effect at the village level. OPERATIONALIZATION OF CONTROL VARIABLES In order to account for other variables that are theoretically related to return migration and remittance, we include a number of controls (all measured in 1994) into our model. Among the controls are measures of the demographic characteristics of migrants, the location of various family members of the migrant, and several household level variables.
In the interest of brevity, we limit detailed discussion of control variables to only those that are relatively less intuitive or more theoretically interesting. Demographic characteristics of the migrant include education, occupation, age, migration duration, gender, and marital status. Education is measured as a set of indicator variables for whether the migrant had less than a primary school education, only a primary school education, or more than a primary school education. From table 2, which shows descriptive statistics for all independent variables, it can be seen that about 35 percent of migrants have less than a primary school education, while nearly a fifth (about 18 percent) have only a primary school education. Thus, just under half of migrants (around 47 percent) have more than a primary school education. Occupation is also measured as a set of dummy variables, indicating whether a migrant works in agriculture, as a craftsman/worker/laborer, in a professional position, in a service occupation, or is unemployed^5. From table 2, it can be seen that about 43 percent of migrants are employed as a craftsman/worker/laborer, while eight percent are professionals, nine percent are employed in service, and five percent are unemployed. That leaves about 35 percent of migrants working in agriculture. Both education and occupation can be considered a migrant’s human capital endowments, and we have included these measures mainly in reaction to the debate about skills and return migration. If less skilled migrants are indeed more likely to return, it is expected that migrants with relatively less education and in relatively lower paying, less secure jobs (such as a agricultural job compared to a professional job) should be more likely to return. Thus, we expect that higher human capital endowments will make
location. Therefore, we measure marital status and the location of spouse together as a set of dummy variables^7. Table 2 shows that only two percent of married migrants have a spouse living in the origin household. Relatively more migrants, almost 36 percent, have a spouse that lives in the same migration destination, while four percent of spouses live in a different migration location. In twelve percent of cases, the migrant ’s spouse location could not be identified because of missing data^8. Also, two percent of spouses are divorced or widowed, while the majority of migrants (about 44 percent) are never- married. We expect that migrant’s decision to return will be strongly affected by the location of their spouse. Migrants who have a spouse living in the origin household should be the most likely to return, while those whose spouse is a migrant living in the destination community should be the least likely to return. This is reasonable because the latter migrants are probably living with a new household to which they may feel are more obligated than their former household. This effect may also be related to the location of the migrant’s children. Thus, we also include two indicator variables for whether any of the migrant’s children live in the household, or whether any of them are migrants. In addition to the location of the migrant’s spouse, we also examine whether the migrant’s parents still live in the origin househo ld. Research in Thailand suggests that adult Thai children provide old age security to their elderly parents (Knodel, et al. 1995, Knodel and Chayovan 1997). We measure parent location (including in- laws) using a series of dummy variables indicating whether: both parents are in the origin household, only the mother lives in the origin household, only the father lives in the origin household, or neither parent lives in the origin household. Table 2 shows that the most
common arrangement is for both parents to live in the household (this is true for 60 percent of the cases), while having only a mother live in the household (occurs 18 percent of cases) is almost four times more common that having only a father live in the household (occurs five percent of the time). In only 17 percent of cases does neither parent live in the origin household. Household variables include controls measuring the presence of agricultural equipment, a household wealth index, the amount of land owned, whether the household grew rice, and counts of the people living in the household who are of working age, of non-working age, or who are migrants. The presence of agricultural equipment is measured as an indicator variable equal to one if the household owns any of the following assets: a large tractor, a small tractor, a rice thresher, a water pump, or an electric generator. Agricultural equipment is specifically linked to farming. For instances, rice farmers in Nang Rong use small tractors for tilling fields, water pumps for irrigation, and rice threshers cutting rice stalks. In some cases, farmers rent agricultural equipment to other farmers as a way of making additional income. Table 2 shows that about one-fifth (18 percent) of households own some form of agricultural equipment. Following work by Filmer and Pritchett (2001), household wealth will initially be measured using an additive index, which results from a principal components analysis of a set of household assets (see appendix 1 for details)^9. After constructing the wealth index, each household will be grouped into one of three categories, based on its overall household wealth index score. Specifically, households in the bottom 33rd percentile will be considered to be at the bottom of the wealth distribution, those in the 34th^ to 79th
