Purchasing power

The idea that a difference in the purchasing power of money between places of origin and destination can give rise to return migration was originally modeled in Stark et al. (Reference Stark, Helmenstein and Yegorov1997). A rudimentary framework can capture the spirit of that model in a straightforward way.

Suppose that an individual chooses freely between two locations where he can live and work – in the destination (city) or in the place of origin (village) – and that early in life he prefers to migrate to the city than to stay in the village. The observation that quite often the duration of labor migration is shorter than the individual's working life implies city-to-village migration. To see this, we sketch in Appendix A a model of return migration that is not associated with reversal of the wage differential between the places of destination and origin.

As shown in Appendix A, an increase in purchasing power in the home village, and a rise in wages in the city shorten the period of migration. Conventional migration models might have led us to expect that higher wages at destination would increase the relative attraction of the destination, presumably prolonging the stay there. Because life is finite and savings confer higher satisfaction in the home village than in the city when spending them in the home village buys more, return migration is attractive. As regards the response to a rise in wages in the city, our model suggests the converse. An interesting implication of the model is the possibility that a migrant might return home even though the wages there are zero. Once again, the reasoning is that the consequent possibility to save more in the city when wages there increase, results in spending time in the home village having greater appeal because it allows for drawing on savings there to improve consumption.

Strategic accumulation of human capital

Røed and Stark (Reference Røed, Stark and Mundlak1998) develop a model in which work abroad enhances skills for work at home in a manner that work at home cannot. The very purpose of migration is to lift the earning prospects at home. (Røed and Stark construct a wage function that relates positively the earning possibilities at home to the number of periods spent working abroad.) While living at home, as such, confers utility that living abroad does not, skill enhancement abroad, leading to a higher wage at home, can encourage migration even if the wage differential between abroad and home is negligible or negative, and can attract return migration even if the wage differential between abroad and home is positive. The combination of acquisition abroad of skills valuable in the home country and the preference for living at home compels a migrant to weigh the acquisition of additional skills and the associated extended stay abroad against a longer spell at home during which the higher wage can be enjoyed at home. Even when the wages awaiting him at home continue to increase while he is abroad, a migrant may prefer to return home earlier. Røed and Stark cite evidence of migration from and within Scandinavia, which reveals that nearly 80% of Danes and Norwegians who moved from their country of origin in 1981 returned within nine years. For Sweden, the corresponding percentage was between 50% and 60%.

Occupational status and social prestige

Fan and Stark (Reference Fan and Stark2011) develop a theory of migration as a response to occupational stigma. Migration of this nature is usually by unskilled workers who are reluctant to perform degrading work at home, yet do not mind doing the very same work away from home. However, migration could also be undertaken in pursuit of occupational status by skilled workers. Many nationals from a number of developing countries such as India and China in recent times and Israel, Japan, and Korea in the 1960s and 1970s, who obtained their PhDs in the USA, returned home. While these individuals could have commanded good salaries in the USA, they nonetheless went back to receive considerably lower salaries at home. Drawing, with some modification, on the model developed by Fan and Stark, we can shed light on this phenomenon of return migration from a new angle. Fan and Stark assume that individuals derive utility from two sources: income and occupational status. Fan and Stark contend that an individual with a prestigious occupation – say, a university professor – enjoys occupational status. Moreover, the shorter the social distance between that individual and the people in his living environment, the higher his utility from occupational status. The esteem that the individual experiences is closely related to the social group to which he belongs, and to the psychological distance between himself and that group. In the USA, the social distance between, for example, an individual with a PhD who comes from Korea and the community he lives in is usually large. Consequently, he derives little status (utility) from being, say, a professor in the USA. In Korea, however, the social distance between himself and his community is short. People in his social group (family members, old classmates, and so on) bask in his glory. This is valuable to him; namely, he derives considerable utility from being a professor in his home country. Thus, individuals may choose to migrate back to their home country in order to draw away from a social group that is a large social distance from them and immerse themselves in a social group that is a small social distance from them. The model developed by Fan and Stark therefore implies that the migration of skilled workers can also be related to a desire for occupational status and social distance. This line of reasoning suggests a novel explanation for return migration from rich to poor countries.

If a migrant comes from a low-status family and if status at the place of origin matters to the migrant and his family, remittances can be utilized as a status-elevating device. This consideration could affect both the disposition and distribution of remittances. To the extent that elevated status hinges on a visible signal, a new house, for example, could well fit the bill. One reason why migrants’ remittances are so often channeled into improved housing is that a superior house is an unequivocally visible measure, or statement, of status-yielding success, suggesting that the incentive to migrate in the first place was to acquire higher status. Once a house is built, the need for remittances declines, the rationale for migration weakens, and return may occur.

Asymmetric information

Information is rarely symmetric: when meeting to conduct a transaction, agents do not know all the characteristics of fellow agents that affect the outcome of the transaction, and often they know even less about the characteristics of fellow agents when the latter come from abroad, from a different culture. In particular, employers at the place of destination have less information about the individual skill level of migrant workers than both employers at the place of origin and the workers themselves. These observations invite us to address the role of informational asymmetry in return migration. Here we show informally how, in the context of the labor market, asymmetric information can lead to return migration and actually cause additional dynamics in migration behavior. A formal analysis is presented in Stark (Reference Stark1995).

In the context of an employment relationship, asymmetric information applies to the skill level of (new) employees. When information about the individual skill levels of migrant workers is unknown to employers at the place of destination, all migrant workers receive a wage based on the average product of the group of migrants. This wage formation procedure encourages low-skill workers who, without this averaging, would have stayed behind to move together with high-skill workers. When informational symmetry is restored through monitoring and observation, and wages are adjusted accordingly, the low-skill workers return to their place of origin. The high-skill workers stay, and their wages rise. Because now the individually identified (high-skill) stayers and (low-skill) returnees are removed from the pooling process, migration as above is replicated by a yet higher-skilled subset of workers. By migrating, workers constituting the first wave of migration block the migration of workers constituting a higher-skilled wave: the low average wage paid to the workers who constitute the first wave deters migration by higher-skilled workers whose wage at home matches their true skill level, but whose wage at the place of destination would be penalized by being averaged down. And by subsequently exposing themselves to identification, the first-wave workers pave the way for the migration of higher-skilled workers. Thus, migration proceeds in waves, with each wave dividing into workers who stay as migrants and workers who return, and within each wave the returning migrants are the low-skilled workers.

Two examples further illustrate the interaction between the state of information and the incidence and pattern of return migration. Suppose, first, that workers can have one of two skill levels - low and high. The model presented in Stark (Reference Stark1995) offers the following predictions: migration is ex-post fully positively selective even though to begin with it is not; migration splits into workers who stay at the place of destination and workers who return; and the returning workers are low skilled. Suppose, next, that workers in a profession can have one of several skill levels rather than just one of two skill levels. A plausible implementation of the model of labor migration under asymmetric information generates the following predictions: migration is sequential (phased); and not all workers who end up as migrants move at the same time. Each wave of migration (of migrants) breaks down into workers who stay as migrants, and workers who return. Ravenstein's century-old “law of migration” (Reference Ravenstein1885, p. 199), stating that “each main current of migration produces a compensating counter-current” – often quoted but not generated analytically – turns out to be a derivative or a variant of the asymmetric information approach to migration. Within waves, the returning migrants are the low-skilled workers; thus, migration is ex-post positively selective within waves. When the migration process is completed, migration is mildly positively selective – the average skill of migrants is superior to the average skill of workers found at the place of origin – but not all migrants are of higher skill than all workers at the place of origin. (Only in the case of two types of workers does migration turn out to be ex-post fully positively selective.) Wave by wave, the average skill level of migrants rises. And wave by wave, the skill level of the returning migrants also rises.

Social comparisons: ordinal preferences

Here we highlight the role of distaste for falling behind others in the income hierarchy in driving migration and return migration.

How important is the desire for rank in explaining behavior? Brown et al. (Reference Brown, Gardner, Oswald and Qian2008), who study the relationship between pay and well-being, find that the wage rank of workers has a significant influence on the satisfaction that they derive from their wage payments. Powdthavee (Reference Powdthavee2009) even finds that an individual's position in the wealth hierarchy within his local community affects his perceived economic well-being more than a comparison of his income with the mean income of the community. Boyce et al. (Reference Boyce, Brown and Moore2010) argue that individuals gain utility (general life satisfaction) mainly from the rank (or position) of their income within a comparison group rather than from their absolute income per se. Indeed, writings both in economics and in sociology have long maintained that individuals have a strong preference for high (social) rank and are stressed when they have a low (social) rank. Smith has remarked that “the desire of … obtaining rank among our equals, is, perhaps, the strongest of all our desires” (Smith, Reference Smith1759, Part VI, Section I, Paragraph 4). Veblen (Reference Veblen1899) has shown that other people's higher pay can depress one's utility. Maslow (Reference Maslow1943) views status as a basic human need, and Huberman et al. (Reference Huberman, Loch and Önçüler2004, p. 103) infer from a study of five societies that “subjects valued status independently of any monetary consequence.” There is considerable evidence from research in modern economics to the effect that the desire to avoid low rank motivates workers to exert more effort (Neckermann & Frey, Reference Neckermann and Frey2008; Kuhnen & Tymula, Reference Kuhnen and Tymula2009; Duffy & Kornienko, Reference Duffy and Kornienko2010; Kosfeld & Neckermann, Reference Kosfeld and Neckermann2011), and students to perform better (Bandiera et al. Reference Bandiera, Larcinese and Rasul2009; Azmat & Iriberri, Reference Azmat and Iriberri2010). Contemporary evidence nicely supports Smith's assessment of the power of the incentive to escape low rank as distinct from the desire for the tangible benefits associated with high rank. We demonstrate the role of distaste for falling behind others in the income hierarchy in prompting migration and return migration with the help of an example. In the spirit of Smith (Reference Smith1759), we distinguish between absolute income (which we hold constant) and ordinal rank.

Let there be a population of n individuals, where n is a natural number. To ease representation, we assume that the income of individual i is i, i = 1, 2, …, n; namely, the individual's income is the individual's name. However, everything that follows goes through if instead we have (x ₁, x ₂, …, x _n) as an ordered vector of incomes of a given population of n individuals, namely when x _i is the income of the individual i, and x ₁ < x ₂ < … < x _n.

To begin with, all the individuals are in region A. Let (empty) region B come into being or become accessible such that migration between the two regions is possible, and is cost free. In all relevant respects, the two regions are identical. This implies that there is no reason arising from a difference in the regions’ attractiveness for an individual to prefer to locate in one region as opposed to another. The reason for migration is that an individual wants to be in the region that better aligns with his social preferences, which are for a higher rank in the income hierarchy. When in terms of the outcome of social comparisons the regions are equally attractive (a tie), an individual does not migrate. Once an individual is in a region, the region becomes instantly his exclusive sphere of comparison. However, an individual can relocate as many times as he wishes, at no cost to himself, in response to the actual distribution of the individuals between the two regions. Put differently, an individual bases his decision to migrate on the observed current state, without forming expectations as to how other individuals will behave simultaneously. For ease of exposition, we refer to the steps of the dynamics of migration as periods, with the initial period being referred to as zero.

In period 0, the n individuals are in region A. In this case, in period 1, we will have the following distribution of the population (individual n does not have an incentive to migrate, whereas all the other individuals move so as to gain a higher rank in region B):

But now each of the individuals n − 3, n − 4, …, 1 can obtain a higher rank (second) if he were to move back to region A. Thus, individuals n − 3, n − 4, …, 1 move to region A, and in period 2 the distribution of the individuals will be:

Once again, several individuals from region A (specifically, individuals n − 5, n − 6, …, 1) will have an incentive to move to region B. We see that by period $\displaystyle {n \over 2}$ if n is even or by period $\displaystyle {{n - 1} \over 2}$ if n is odd, all comings and goings will come to halt, and we will have the following steady-state distribution of the population:

such that what remains to be characterized is the location of the two lowest-income individuals; namely, the location of individuals 1 and 2. It turns out that the whereabouts of these two individuals depends on whether n is even or odd, and on whether when n is even whether n or n − 2 is a multiple of 4, and when n is odd whether n − 1 or n − 3 is a multiple of 4. Specifically, we have the following characterization. When n is even, 1 and 2 are in different regions: if n is a multiple of 4, then 1 is in region A, and 2 is in region B; if n − 2 is a multiple of 4, then 2 is in region A, and 1 is in region B. And when n is odd, then 1 and 2 are in the same region: if n − 1 is a multiple of 4, then 1 and 2 are in region A; if n − 3 is a multiple of 4, then 1 and 2 are in region B.

In this example, there is a series of return migrations prompted by a desire for higher rank. In each period until the steady state is reached, the migration of individuals with relatively high incomes inflicts a negative externality on individuals with relatively low incomes. To correct for that, the latter individuals migrate back.

Relative deprivation: cardinal preferences

Considerable empirical evidence finds that relative deprivation is a statistically significant explanatory variable of migration behavior. Stark and Taylor (Reference Stark and Taylor1991) show that relative deprivation increases the probability that household members will migrate from rural Mexico to the USA. More recently, Quinn (Reference Quinn2006) reports that relative deprivation is a significant motivating factor in domestic migration decisions in Mexico. Stark et al. (Reference Stark, Micevska and Mycielski2009) explore the relationship between aggregate relative poverty, which is functionally related to aggregate relative deprivation, and migration. Drawing on Polish regional data, they demonstrate that migration from a region is positively correlated with the aggregate relative poverty in the region. Czaika (Reference Czaika2011) finds that in India, relative deprivation is an important factor in deciding whether a household member should migrate, especially in the case of migration over a short distance. Basarir (Reference Basarir2012) reports that people in Indonesia are willing to bear a loss of absolute wealth if there is a relative wealth gain from migration. Jagger et al. (Reference Jagger, Gerald and Arinaitwe2012) demonstrate that relative deprivation is a significant explanatory variable of circular migration in Uganda. Drawing on data from the 2000 USA census, Flippen (Reference Flippen2013) shows that both blacks and whites who migrate from the north to the south generally have lower average absolute incomes than their stationary northern peers, yet in the wake of their migration, they enjoy significantly lower relative deprivation. Vernazza (Reference Vernazza2013) finds that, even though interstate migration in the USA confers substantial increases in absolute income, the trigger for migration is relative deprivation (low relative income), not low absolute income. Hyll and Schneider (Reference Hyll and Schneider2014) use a data set collected in the German Democratic Republic in 1990 to show that aversion to relative deprivation enhanced the propensity to migrate to western Germany. The very same motives that underlie migration in these studies can just as well underlie return migration, as demonstrated both in this section and in the preceding section.

We illustrate how sensing relative deprivation can prompt return migration with the help of a sharp example. Here we measure relative deprivation in a cardinal way. Use of such a measure differentiates the approach taken here from the approach taken in the preceding section where rank (an ordinal measure) was used. This difference is intentional, because it adds robustness: there are several ways in which falling behind others can be calculated, and in the preceding and current sections we incorporate two of these ways.

An individual whose income x is in region B, where the income of every other individual is also x. The individual can, alternatively, migrate to region A where the income awaiting him will be z > x, and where the income of the only other individual is also z. As assumed in the preceding section, the region where an individual is located constitutes the individual's sphere of social comparison. The individual likes absolute income and dislikes relative deprivation and assigns to these two terms in his utility function positive weights that sum up to one. Because the income in region A is higher than the income in region B and because no relative deprivation is experienced in region A, the individual will migrate from region B to region A. Let us assume now that the income of the other individual in region A increases from z to a higher income, while the income of the individual who migrated from region B to region A remains unchanged at z. Now, this latter individual senses relative deprivation in region A, which negatively affects his utility from staying in that region. It follows that if the individual's utility in region A is now lower than his utility in region B, the individual will prefer to migrate back to region B, even though this migration results in him having a lower income. The intuition is that if the weight that the individual assigns to relative deprivation is higher than a critical level, then low income with no relative deprivation is worth more to him than high income combined with relative deprivation. A formal representation of the reasoning in this subsection is presented in Appendix B.

Matching in the marriage market, and divorce

Men may migrate to accumulate wealth that will improve their prospects in the marriage market. Single men assign a higher value to social status than do single women (Huberman et al., Reference Huberman, Loch and Önçüler2004), taking the importance attached to low relative wealth as a measure of the importance attached to low status. This difference by gender can be explained by the fact that low status carries a harsher penalty for single men than for single women, which, in turn, arises from the fact that low status for single men translates into inferior outcomes in the marriage market: in selecting a partner, the social status of a man is more important to a woman than the social status of a woman is to a man (Kenrick et al., Reference Kenrick, Sadalla, Groth and Trost1990).

To understand why status matters to men more than it does to women, we invoke evolutionary, socio-biological reasoning, attributing gender-specific behaviors to different selective pressures faced by females and males.Footnote ² Male fitness is limited by access to fecund females, whereas female fitness is limited by physiological and energy constraints. Successful males can enhance their fitness by monopolizing the reproductive performance of several females, whereas the fitness of females cannot profit from multiple mates to the same extent. Females are, therefore, a ‘contested resource’ for which males compete.Footnote ³ This competition need not take the form of a direct contest for females. Instead, males compete for assets ranging from feeding territories and food to more intangible ‘resources’ like social status, which can be converted into a reproductive opportunity, whether because they are directly attractive to females or because they help defeat rival males. In short, status is a means of gaining a valuable resource via a better hierarchical position, and evolution has embedded this concern for status into individual preferences.

The received literature has long correlated high status with superior outcomes in the marriage market, and social status with relative wealth. We refer briefly to a number of studies that have modeled these links. With regard to status and the marriage market, Becker (Reference Becker1973) provides a theoretical foundation for the importance of status in the maximization of matching quality in the marriage market. Cole et al. (Reference Cole, Mailath and Postlewaite1992) develop a model in which “men and women who will match have preferences over the matches they will enter into. … Relative success in the matching process will be determined by agents’ status” (p. 1097). Cole et al. note that men differ in their wealth, and that women are characterized by varying degrees of quality, which, in turn, constitutes an argument in men's utility function. The model of Cole et al. suggests that, in equilibrium, women of higher quality choose richer men. This choice or preference intensifies men's distaste for having low relative wealth. Robson (Reference Robson1996) remarks: “Males obtain more offspring as a consequence of greater wealth both directly and because this attracts more mates. The second effect induces gambling driven by relative wealth …” (p. 190).

With regard to the conversion of relative wealth into (social) status, a natural starting point is Smith (Reference Smith1759), where we read that wealth accumulation yields social status and that status matters for individual welfare. Veblen (Reference Veblen1899) dwells at length on the notion that in modern Western societies the aspiration for high relative wealth is motivated by an underlying desire for social status. In his study of the origins of modern English society, Perkin (Reference Perkin1969, p. 85) comments that “the pursuit of wealth was the pursuit of social status.” Frank (Reference Frank1985) emphasizes the significance of relative wealth for the acquisition of social status. Robson (Reference Robson1992) develops a model of decision-making in which agents care not only about their wealth, but also about their relative position in the wealth distribution. Robson (p. 837) writes: “[O]rdinal rank in the wealth distribution enters von Neumann-Morgenstern utility as an argument in addition to wealth itself. Thus, higher wealth increases utility not only directly but also indirectly via higher status.” Futagami and Shibata (Reference Futagami and Shibata1998, p. 110) define a “person's relative wealth position in the society [as] status.” Pham (Reference Pham2005, p. 407) develops a model in which social status is “increasing with individual wealth and decreasing with the average wealth of the society.”

When men migrate in order to improve their wealth rank at the place of origin and marry better at home, their migration and return migration are components of a single strategy. For example, there is indirect evidence (Drinkwater et al., Reference Drinkwater, Eade and Garapich2009; UK Office for National Statistics, 2015) supporting the notion that Polish single men migrate to the UK temporarily to accumulate wealth in order strategically to improve their position (value) in the marriage market in Poland.

Under a variety of circumstances, divorce will propel migration and return migration, and not only to reduce the psychological costs associated with dissolution (‘to start a new life’) or to avoid making child-support payments. Often, location decisions are made jointly by the married couple so as to maximize joint wage income. If the inter-market variation in men's wages is larger than the inter-market variation in women's wages, and if women's wages are not strongly positively correlated with those of men, then when it comes to the choice of location, the husband's wage will weigh more heavily than the wife's wage in the couple's decision of where to locate in order to maximize their joint earnings. The terminology often used to describe this situation is that the wife is a tied migrant. Suppose, then, that a married couple migrated with the wife's wage at the place of destination being lower than her wage at the place of origin. Consequently, during the marriage, the labor market location of the husband aligns with his earning ability better than the labor market location of his wife aligns with her earning ability. In the event of divorce, the husband would be less likely to have a wage-based motivation to migrate back than the wife who, under these assumptions, has sacrificed her own wage prospects during her marriage. All other things being the same, under these circumstances of divorce, higher return migration rates by women than by men will be observed. To some extent, this prediction is confirmed by evidence presented three decades ago by Sander (Reference Sander1985); in the USA in 1970, farm and rural non-farm divorced women migrated more frequently than farm and rural non-farm divorced men.

Receipt of inheritance

A nice way to link migration, return migration, and inheritance is via remittances. In its choice of who migrates, a family may be affected by the prevailing rules of inheritance. Consider a setting in which inheritance is governed by primogeniture – the exclusive right of the eldest son to the family's farm (Stark, Reference Stark1999). As migrants, second and later children will not be motivated to remit by an aspiration to inherit. In contrast, as a migrant, the eldest son will be encouraged to remit if he knows that his transfers will be invested in the family's farm, and these transfers, in turn, will serve to boost the value of his inheritance. Once the son actually inherits the farm, he will have an incentive to return.

The institution of primogeniture can be linked with return migration in other ways too. As migrants, second and later children who do not expect to inherit will be motivated to boost their earnings by acquiring human capital as an alternative to the ‘land capital’ (the family's farm) that the eldest son will receive. To the extent that the acquired human capital is location specific, the opportunity cost of second and later children of returning home at the time of inheritance will be high, even if in the course of their lifetime the regime of primogeniture was replaced by equality of inheritance. Moreover, to the extent that investment in human capital bears fruit in the form of good returns on its gains away from home, return migration by those who acquired human capital will be less likely than by those who did not. Thus, in situations in which the eldest son as well as other children migrated, the likelihood of return migration by the eldest son will be higher.

Meeting a target

Suppose that the original aim of migration (and possibly of remitting) is to achieve a specific investment objective in the home village – say, setting the migrant's parents up as shopkeepers in order to secure for them a steady stream of income in the future, or to install a tube well, or to acquire a truck in order to take produce grown in the village to the nearby town. The reason for migration to the city or beyond is to acquire a specific physical capital. Once the acquisition is made and the objective is met, the migrant will return. Similar reasoning applies to paying for advanced schooling (college education) for children. Migrants from the Philippines told the author that they would end their work in a rich Gulf state and return home once their children graduated from college.Footnote ⁴

Shocks at home

One reason for a migrant's return is a severe adverse shock to the livelihood of his household back home. Even if continued work at the migration destination and the consequent transfer of remittances would help a stricken household, migrants and their households may consider it beneficial at a time of extreme misfortune (for example, a hurricane or an earthquake) for the migrant to return. The support that the migrant can provide to the household by his actual presence (for instance, when labor to repair serious damage cannot be hired locally) may exceed the benefits from transferring remittances. This could explain the seemingly paradoxical return migration to El Salvador in the wake of the devastating earthquake of 2001 (Halliday, Reference Halliday2006).

There are countries and regions that are prone to severe shocks: for example, flooding in Bangladesh during the monsoon season is a frequent occurrence. There are a variety of ways in which people can and do prepare to protect themselves from the effects of natural disasters, and one of them is obtaining on-site help from the migrant's family members. It appears that when natural disasters are mild, remittances serve as a relief; when the disasters are severe, return migration is the chosen response. This consideration yields yet another testable proposition related to the incidence and pattern of return migration.

Costly separation caused by migration

There is a form of migration where return migration is built in as a regular and repetitive feature. Under certain conditions, this form is preferable to alternative forms of migration.

Consider a family with one breadwinner, whom, in line with the ‘New Economics of Labor Migration’ (Stark, Reference Stark1993), we view as an agent who migrates on behalf of his family in order to maximize the well-being of the family. Suppose that the breadwinner has three migration options available to him: a sequence of episodes of seasonal migration; permanent migration by himself; and permanent migration together with his family. In comparison with the second and third options, an advantage of the first option can arise from the conjunction of two factors: the cost of separation from home, and the cost of living abroad. When the latter cost is relatively high (when it is expensive for the entire family to live abroad), migrating alone has an edge; it facilitates earning money in the host country and spending it in the home country (recall the subsection entitled ‘Purchasing power’). A drawback, however, of migrating alone is the suffering caused by separation from one's family. This drawback is particularly burdensome when it becomes increasingly difficult for the breadwinner and his family to sustain separation as its duration lengthens. In essence, the cost of separation is the sum of two costs: that which is caused by the breadwinner's suffering as a result of separation from his family, and that which is caused by the family's suffering as a result of separation from the breadwinner; an altruistic breadwinner takes into account the cost imposed by his absence on the members of his family who stay behind.

Kepinska and Stark (Reference Kepinska and Stark2013) documented the durability of seasonal migration from Poland to Germany, a phenomenon that has persisted for more than a century. They refer to the role of the tradition of engaging in seasonal migration as a force that helped invigorate the process and contributed to its sustainability even when, to different degrees and at different times, the process was interrupted by a shifting political, regulatory, and legal environment. (Evidence in support of the role of tradition is provided, among other things, by the continuation of the seasonal flow of migrants from once-border regions – which became internal regions following World War II – despite the fact that since the redrawing of the German-Polish border, proximity is no longer a factor encouraging repeated, short-term seasonal moves.) Stark and Fan (Reference Stark and Fan2007) sought to provide a rationale for the observed durability of this very phenomenon. They showed analytically how the cost of living differential and the cost of separation shape seasonal migration outcomes to a significant degree and, in particular, how balancing of the two costs invites a series of seasonal migrations, each of a relatively short duration. Placed in the context of the current paper, the ability to return constitutes a necessary condition for migration. Consequent testable predictions are then, other things being equal, that: heads of families with young children for whom the cost of separation will be high will be more inclined than other heads of families to opt for seasonal migration / be correspondingly more present in the group of return migrants; a narrowing of the difference in the cost of living between abroad and home will encourage substitution of permanent migration together with families for seasonal migration alone / result in a correspondingly reduced incidence of return migration; and so on.

Failure

Migrants will return to their place of origin when at the place of destination they fail to gain employment, they lose hope that their fortunes will turn around, or they are unable to sustain themselves any longer. Rough living is the mother of return.

Although ‘the economics of return migration as a response to failure’ has yet to be written, there are several ways in which behavior and choices can reduce the likelihood of return migration because of failure. One such way will be for the migrant's household to select as a migrant a member who will be strongly committed to search for gainful employment, say, and who will try hard to avoid failure. In the Philippines, this consideration has resulted in the choice of a daughter rather than a son as a family migrant, even when the earnings of male migrants are on average higher than the earnings of female migrants (Lauby & Stark, Reference Lauby and Stark1988). A lower probability of return migration caused by failure even when combined with low expected earnings is preferable to the household as compared to a higher probability of failure-generated return migration combined with high expected earnings.