The human capital approach to migration
Economists have long been interested in migration because it is a mechanism that in principle should have an equilibrating effect in the labor market. While early analyses of migration emphasized these aggregate effects, more recent work has focused on migration behavior at the individual level.
The human capital approach to migration (going back to the early 1960s and Sjaastad) treats migration as an investment increasing the productivity of human resources. Investment constitutes a sacrifice of current consumption in exchange for (presumably higher) future consumption. Investment activity thus entails both costs and returns, and in principle the desirability of the activity can be measured by the rate of return to resources allocated to the investment, or the present value of the stream of net returns.
DaVanzo's equation (1) nicely represents the basic microeconomic model of migration decision making:
PVij equals the present value of the net gain of moving from place i to place j; Ujt equals the expected utility or real income at place j in period t; Uit equals the expected utility or real income at place i in period t; Cijt equals the cost incurred in period t of moving from i to j; r equals the discount rate (0 <= r <=1; according to DaVanzo r is normally between .05 and .15); and T equals the expected length of remaining lifetime.
This approach assumes that both objective and subjective variables can be translated into utility terms. Further, note that Uit effectively represents the opportunity cost of migrating. Given an individual's current place of residence (i), migration will occur if there is some destination (j) for which PVij > 0. If there is more than one such destination, migration will be to the location (j) that has the highest value of PVij.
Ujt - Uit represents the gain in utility in period t from moving. This gain can be broken down into two components: a portion attributable to the change in the migrant's real earnings or real income stream, (Yjt/Pjt) - (Yit/Pit), where Y/P represents money income relative to prices or real income; and a portion representing the change in utility due to the difference in locational preferences between the origin and the destination, Ujtlp - Uitlp.
Likewise, the cost in period t of migration from i to j, Cijt, may be viewed as representing three elements: the direct money cost of moving, the opportunity cost of moving (forgone earnings), and the psychic cost of moving (loss of utility from leaving behind friends and family). As DaVanzo notes, empirical work typically treats costs as occurring at the time of the move; later costs, such as subsequent forgone earnings, are treated as negative benefits.
Empirical implementation of this conceptual approach raises a host of issues, some of which will be examined in detail below. At this point, however, it is useful to highlight some immediate implications of the model that help account for observed regularities in migration behavior in a wide variety of social and economic settings.
First, the model's emphasis on migration as an investment yielding returns throughout the migrant's lifetime helps account for the markedly higher propensity of the young to migrate. Young migrants have a longer period during which they can recoup the costs of investing in migration, and this results in a greater present value of returns which in turn should increase the probability of migration.
Psychic costs of migration are also likely to be smaller for young people, who have not had the same opportunities (time) as their older counterparts to develop strong ties to their present location. With fewer belongings and smaller families, the young also have lower direct costs of moving (a relatively minor factor).
More generally, the lifetime perspective inherent in the model emphasizes the importance of long-term considerations. Consideration of negative transitory effects of migration and of complementary occupational training investments reinforces the desirability of taking a long view in studying migration behavior. We'll see later in this section that taking a long view is particularly helpful in understanding rural-urban migration in the Third World (Todaro model).
At the same time, it should also be clear that short-run considerations may be relevant as well. At the individual level, the greater propensity of the unemployed to migrate presumably reflects lower opportunity cost of migration, as well as lower opportunity cost of investing in information relevant to the migration decision. At the aggregate level, there are fluctuations in migration flows to various places (e.g., Australia, U.S. historically) that reflect short-run fluctuations in economic activity (recession vs. prosperity).
Empirical issue: Measuring the earnings gain to migration
Empirical implementation of human-capital microeconomic models of migration frequently raises a number of methodological issues. Most of these issues arise when one attempts to operationalize specific concepts in the general model. DaVanzo provides a good overview of many of these issues.
Here, we will focus on one such issue: measuring the expected/actual gain in real earnings from migration. Examining this issue will open up some additional considerations.
Conceptually, measurement of the gain in real earnings from migration requires that one compare the lifetime (expected) earnings stream that would exist if the individual were to migrate with the lifetime earnings stream that would exist in the absence of migration. Clearly, both earnings streams cannot be observed: if the individual migrates, we see the former stream but not the latter; while in the absence of migration the former stream is unobservable. (Cf., projection problem, partially mitigated by discounting.)
Imagine, then, that we are examining the decision to migrate. We can look at an individual's earnings in the origin and make some assumptions about how those earnings might change over time. To apply DaVanzo's equation, though, we need to estimate potential earnings of prospective migrants in the destination.
Clearly, in the presence of heterogeneity of labor, our estimate of potential earnings for any individual should take into consideration salient characteristics, such as schooling, age or experience, and perhaps occupation. In effect, we can imagine imputing the earnings in the destination of potential migrants, based on the earnings of comparable nonmigrants in the destination (comparable vis-à-vis possessing the same bundle of income-producing attributes).
Two distinct aspects of migration make this sort of imputation tricky. First, migration frequently represents one of several investments occurring simultaneously: change of location is often accompanied by a change in occupation, which normally entails some investment in on-the-job training; and these occupational changes are often the outcome of job search by the migrant in the new location.
Hence, since investment tends to twist age-earnings profiles (i.e., lower them at the time of investment but make them steeper later), the existence of multiple investments occurring during and soon after migration can account for what Sjaastad described as the "negative transitory effect" of migration.
This implies that studies of the actual earnings gains from migration must allow sufficient time for the returns to these multiple investments to emerge. In addition, the fact that migrants frequently are making multiple investments means that it will be hard to find truly comparable nonmigrants, even controlling for age, experience, etc.
A second aspect of migration that makes imputation complicated is evident once we allow for location-specific capital (assets), a concept that DaVanzo uses at several points to help account for observed migration behavior. A particularly relevant type of location-specific capital is firm-specific human capital. This may be defined as skills and knowledge of a worker that increase the worker's productivity in a particular firm but are of no value outside the firm.
Theoretical analyses from labor economics suggest that both the costs of and returns to investments in firm-specific capital will be shared by the worker and the firm. This, in turn, means that with the exception of a new employee, a portion of a worker's earnings will constitute a return to firm-specific training. The magnitude of this return will generally be positively related to the worker's tenure on the job.
By definition, firm-specific human capital depreciates 100 percent when a worker leaves a firm. Hence, apart from migration resulting from an intrafirm transfer, migration will entail a capital loss for workers.
Further, the earnings of "comparable" workers in the destination will reflect returns to the firm-specific capital that they have accumulated. But these returns are not immediately available to the migrant -- i.e., there is not true comparability.
To enhance comparability, one could estimate wage equations along the lines suggested by DaVanzo, controlling for individual characteristics like schooling, age, occupation, etc., and also controlling for job tenure (e.g., WAGE = a0 + a1*EDUC + a2*AGE + a3*SKILL + a4*TENURE). Then wages for potential migrants could be imputed, given their characteristics, by estimating predicted earnings in the new location assuming zero tenure.
There is one further problem, however, and it is one we've encountered earlier. The sample of migrants that we observe is not a random sample of individuals; it is a self-selected sample in which -- if we believe the economic approach to migration -- the selection rule is presumably based on expectations about the gains to migration.
That is, we have a potential sample selection bias problem. Among individuals in a given location with similar personal characteristics, some people are movers and others are stayers. This distinction presumably reflects unobserved differences between the two groups that influence the expected gains to migration. Again, then, we have a "comparability" problem.
Failure to take this aspect into consideration means that we may get biased estimates of the prospective gains to migration. However, if the unobserved differences are correlated with observable variables (such as schooling or age), then the statistical techniques for dealing with selectivity bias that were mentioned earlier will be effective in dealing with this problem. As DaVanzo notes, application of these techniques suggests that empirical estimates of the (potential) gain to migration that fail to take selectivity bias into account often result in considerable bias, and thereby fail to provide appropriate tests of the conceptual model.
The issue of finding "comparable" counterparts is clearly a tricky one. Note, however, that the availability of longitudinal surveys provides an alternative method for doing so. That is, whatever unobserved characteristics (e.g., ability, motivation) cause a migrant's earnings to differ from those of "otherwise similar" nonmigrants presumably were operative prior to the move (as well as after). Hence, comparison of the migrant's earnings before and after the move provides a reasonable approximation to the earnings gain from migration that does not suffer from selection bias (cf., Van Adams' findings re training effects).
Such a comparison must allow for the "negative transitory" effects of migration. This means that migrants ideally should be followed for a number of years after migrating. Further, since premigration earnings may vary from year to year, these also ideally should be observed for multiple years. In addition, as emphasized in the following section, family considerations also should be taken into account.
Family considerations: The economics of family migration
To this point, we've looked at migration in the context of an individual's decision to migrate based on the costs and returns to that individual. Frequently, however, migration takes place in a family context, and this means that we need to think about the economics of family migration. Reflecting the household decision-making framework that has served as the underlying conceptual framework in the course so far, the household -- rather than the individual -- is viewed as the relevant unit for analysis.
In this context, then, a family member's migration decision depends on the expected costs and returns to the entire family, rather than just to the individual family member. A number of empirical findings, including many reported by DaVanzo, may be readily interpreted in this context (cf., lower likelihood of migration when wife is employed, or when wife's job tenure is greater).
While the impact of migration on total family earnings is most relevant from this perspective, there is clear evidence that married women who migrate can often be described as tied movers. That is, particularly in the relatively short term, migrant husbands are likely to experience gains in earnings while migrant wives experience losses (but with the gains exceeding the losses).
In earlier studies, the losses of earnings for wives were closely related to reduced employment outside the home following a move (still true now?). Evidence also suggests that over time the earnings gain to migration rises for both husbands and wives. (This negative transitory effect is also evident in studies of the earnings of immigrants to the U.S.)
Family considerations may also influence migratory behavior, even when the entire family does not migrate. Two examples cited by DaVanzo are particularly relevant in developing nations: seasonal migration and "portfolio diversification."
Seasonal migration often takes place in agricultural settings. This migration may be rural-->rural, in response to attractive temporary employment opportunities elsewhere (as with peak labor demand during planting and harvesting of commercial or export crops, as in Kenya or the Sudan), or it may be rural-->urban, reflecting periods of slack demand for agricultural labor (as occurs in the dry season; cf., intrahousehold substitution of labor in Kenya).
Portfolio diversification refers to the phenomenon whereby a family finances the investment in migration for a family member who is likely to benefit from moving (permanently or temporarily) and who then provides remittances back to the family. These remittances constitute a return to the family's investment in migration. In some cases, these remittances may be quite substantial: for Haiti, it's been estimated that they amount to more than five percent of GNP (cf., Kuwait; note also the extension of the portfolio diversification notion to include fertility).
Rural-urban migration in the Third World: The Todaro model
Rural-urban migration has been a significant force contributing to the growth of urban centers, both historically in the West (especially during the 19th century) and more recently (during the past half century) in the Third World. In the economic development literature of the 1950s and 1960s this internal migration was regarded as a desirable process whereby surplus labor was gradually withdrawn from the rural sector to provide needed workers for urban industrial growth (cf., Lewis model).
However, observers of migration and urbanization in developing countries often argued that the rural-urban migration that was taking place was occurring despite high (and in some cases, increasing) levels of urban unemployment. Hence, they argued, this migration could not be accounted for by economic considerations (cf., socio-cultural explanations -- bright city lights).
Todaro's model is an attempt to provide an economic explanation for continued migration in the presence of substantial open unemployment. More specifically, his model links up rural outmigration, industrial sector wages, and urban unemployment.
By focusing on the expected earnings of migrants (where expected earnings reflect both the prevailing modern sector wage and the probability of having a modern sector job) and on the notion of migration as an investment, Todaro provided an economic explanation for rural outmigration in the face of high urban unemployment. That is, in Todaro's model rural-urban migration represents rational economic behavior from the point of view of the individual migrant.
*** Discussion of the details of the Todaro model -- see math handout #2 ("Todaro Model"). ***
The bottom line of the Todaro model, then, is that if modern sector wages are sufficiently high relative to earnings in agriculture, migration will be economically rational for an individual even when the probability of securing industrial employment in any period is quite low.
To the extent that a potential rural migrant has friends or family in the city, the psychic costs of migration are reduced, and the expected duration of search for modern sector employment (and hence the opportunity cost of migrating) may also be reduced. Likewise, if expected earnings in the informal sector are high relative to earnings in agriculture, the desirability of rural-urban migration is enhanced further.
While migration may be a rational investment for the individual, the persistence of high wages in the modern sector means that there will be a permanent high level of unemployment and underemployment in the urban Third World. (Compare this to the classical economic solution to the problem.)
As Todaro notes, given the levels of industrial and agricultural earnings, and of modern sector growth in output and productivity, the probability of finding a modern sector job serves as an equilibrating mechanism, influencing flows of rural migrants and hence urban unemployment rates.
The Todaro model has strong policy implications regarding development strategies. For example, efforts to reduce urban unemployment by stimulating job creation in the urban modern sector are likely to be stymied eventually, in that they will raise the likelihood of finding a good job and hence induce a greater flow of rural migrants. Todaro has argued far and wide that the best strategy to reduce urban problems in developing nations is to seriously promote rural development (economic opportunities plus amenities like health care and education).
Cole and Sanders (C & S) start from Todaro's basic human capital perspective on migration, but they focus on the informal sector (what they call the urban subsistence sector). Todaro treated the informal sector as essentially just a way station on the road to modern sector employment. C & S argue, however, that Todaro's approach fails to explain a good deal of rural-urban migration. In their view, much migration takes place explicitly with the informal sector, not the modern sector, as the intended destination. Hence, they see rural-urban migration as a dualistic phenomenon.
The informal sector is seen as having low capital-labor ratios and few if any formal human capital requirements (i.e., ease of entry characterizes the informal sector). This is in contrast to the urban modern sector, which has relatively high (and often rising) educational requirements.
Focusing on Mexico, C & S argue that a lot of rural-urban migration is of low-skilled, low-income individuals who are seeking jobs in the informal sector. Associated with this notion is the idea that migration is becoming increasingly less selective.
In their critique of Todaro's model, C & S argue that for plausible values of the modern sector rate of job growth (gamma) and the proportion of the urban labor force not employed in the modern sector ((N-Nu)/N), the probability of obtaining a modern sector job this period (pi) will be very low (see their Table 1). Consequently, assuming that modern sector earnings are twice as high as earnings in agriculture, too much time is required before the present value of expected urban earnings would equal the present value of expected rural earnings (see their Table 2).
Rural migration to the informal (U-S) sector is seen by C & S as being based on a comparison of the present value of the stream of earnings in agriculture with the present value of the stream of expected earnings in the informal sector. This is represented in an equation analogous to Todaro's equation 1. At the same time, Todaro's equation 1 represents migration to the urban modern sector. Two equations are needed in this dualistic framework.
Hence, the rural agricultural earnings stream (the R-S wage) is the benchmark for all rural migrants, but those with adequate human capital migrate to the modern (U-M) sector while others migrate to the informal sector. Those in the former group who do not find jobs in the modern sector are seen as accepting temporary unemployment rather than taking menial jobs. By contrast, C & S argue that all informal sector migrants will be able to find jobs.
Earnings in the modern sector are seen as institutionally determined, as in Todaro's model. In the informal sector, since there are no barriers to entry, earnings vary with labor supply and labor demand. If there is population pressure on the land (i.e., with MPL close to zero in agriculture, the R-S sector), labor supply to the informal (U-S) sector will be perfectly elastic. Labor demand will be based on the demand for output in the informal sector.
Output in the informal sector is in the form of labor services, and capital is negligible. Thus, Qus = F(Nus). "Exports" to the modern sector are an important part of informal sector output and income. These "exports" vary inversely with the wage in the informal sector, and positively with the price of manufactured substitutes for informal sector labor, the level of per capita income in the modern sector, and the population of the modern sector. Hence, growth in the modern sector will stimulate growth in the informal sector.
With informal sector labor and rural labor as close substitutes, so long as the marginal productivity of labor (MPL) in agriculture is close to zero, labor supply to the urban informal sector will be perfectly elastic at a wage just above the rural wage. C & S's basic equilibrium condition then -- under which there is no migration -- is that Wus = Wrs.
Population pressure leading to labor force growth in either rural agriculture or the urban informal sector will tend to reduce wages in both sectors. There will be a countervailing tendency via increased informal sector "exports" resulting from the lower U-S wage.
Overall, it's evident that wages and employment in the informal sector will be closely tied to what takes place in both other sectors. Growth of the subsistence population keeps downward pressure on the subsistence wage (à la Malthus), and rapid growth of the modern sector translated into increased demand for informal sector output.
Cole and Sanders see their model as a "useful complement to the Todaro theory of migration." They conclude their paper by arguing that much rural-urban migration -- including migration to the informal sector -- is desirable from a social point of view as well as from the perspective of the individual migrant. This contrasts sharply with Todaro's view, in which migration is undesirable from the social point of view despite being rational for the individual.
Todaro's comments on C & S ("seriously flawed") points out some problems and areas of disagreement. Todaro argues that their assumption (as part of their critique of the relevance of his model) that modern sector wages will be twice as high as agricultural wages is flawed. He contends instead that a ratio of 4:1 or 5:1 would be more accurate.
Further, he contends that effective discount rates are very low, due in part to negative real interest rates arising from rapid inflation and administered nominal rates (cf., effort to explain PV notion in Zaire). Taking both aspects into account yields equalizing periods of only a few (2-6) years. Thus, his view is that the C & S critique of his model is undermined by these considerations.
Finally, Todaro also defends the distinction between the private benefits of migration and the social benefits as being real and relevant. In this regard, he cites evidence from surveys of governments in developing countries suggesting that most of them are dissatisfied with the existing population distribution.