Printer Friendly

Favorable self-selection and the internal migration of young white males in the United States.

This study offers an alternative empirical technique to test whether the favorable self-selection hypothesis applies to internal migrants in the United States. Our empirical specification attempts to determine if prospective migrants possess unobserved traits such as higher ability or motivation which influence their earnings potential relative to nonmigrants. Using NLSY data for 1985 through 1991, we find some support for the favorable self-selection hypothesis for white males who move from one SMSA to another. Prior to their move, prospective migrants enjoy a consistent advantage in annual wage and salary income relative to nonmigrants with similar earnings-related characteristics.

I. Introduction

Internal migration contributes significantly to relative population changes across various regions in the United States. For example, in 1984 39,380,000 persons changed residences in the United States; of these, 16.4 percent moved across state lines (Long 1988, p. 49). Following the early work of Schultz (1961), Becker (1964), and others, economists typically view the migration decision as a form of investment in human capital. Thus, an individual's decision to migrate is based on a rational consideration of the relevant costs and benefits which accrue from a change in location. In a broader context, rational individuals adjust their portfolio of human capital investments (which includes activities such as schooling, training, and health care) so as to maximize their overall lifetime returns. This study explores whether individuals who choose to migrate possess unobservable traits (such as higher motivation or ability) which somehow differentiate them from nonmigrants.

An individual must consider numerous factors in deciding whether or not to migrate from one location to another. First, there are differences in characteristics between the destination and origin locations: employment opportunities, wage levels, living costs, or availability of public services. Other differences exist in location-specific amenities such as climate, crime rates, or pollution. Second, life-cycle factors such as age, marital status, presence and age of children, education, and acquired labor market skills will affect how individuals evaluate the differences between destination and origin. Third, the net costs of moving must be considered. Such costs include direct outlays for relocation, lost earnings, and "psychic" costs (for example, leaving family and friends).

There is an extensive literature on the economics of internal migration (for a recent summary see Greenwood 1985). Numerous authors have assessed the link between net migration flows and differences in location-specific characteristics (Greenwood, 1985; Hunt, 1993; Treyz et al., 1993). Others have examined the influence of personal characteristics on the propensity to migrate (Graves and Linneman, 1979). Recent work by Borjas, Bronars, and Trejo (1992a, 1992b) has concentrated on the internal migration of young workers in the U.S. Their 1992a study focuses on the labor market assimilation of young workers in their new location. They find that, after initially earning less than comparable natives (non-migrants), young migrants will experience rapid increases in their wages. Borjas, Bronars, and Trejo estimate that the wages of migrants will equal natives after six years in the new location (1992a, p. 171). Their results also indicate that the initial wage disparity between natives and migrants is dependent upon the distance moved and the economic conditions prevailing in the new location (1992a, p. 173). Comparing the assimilation process of internal migrants with that of U.S. immigrants, they find that internal migrants assimilate more quickly than immigrants (1992a, pp. 174-75).

Recognizing that migrants are not randomly drawn from the population, Borjas Bronars, and Trejo (1992b) apply Roy's self-selection model to internal migration. They find that regional differences in skill returns will motivate young workers to migrate to areas where skill returns are commensurate with their skill endowments (1992b, pp. 160-61, 184). In other words, highly skilled workers will be attracted to areas which offer a higher return to skill and vice versa (1992b, pp. 161, 184).

With respect to international migration, Chiswick (1978) and others have investigated the hypothesis that immigrants to the U.S. have a greater innate ability and/or are more highly motivated than their native-born counterparts. Higher ability and motivation results in substantial increases over time in the earnings potential of immigrants. This "favorable self-selection hypothesis" suggests that immigrant earnings (while starting from a deficit) will eventually surpass those of native born workers (Chiswick 1978, pp. 919-20).

The notion that internal migrants may also possess greater ability or motivation than nonmigrants is well-established (see Marshall 1920, pp. 199-200). However, direct empirical tests of this proposition are generally absent from the literature. This study investigates whether favorable self-selection is present in the internal migration of young white males in the United States. We use an empirical technique which tests whether those individuals who choose to migrate are more motivated and/or able than those who choose to remain in the same location. The logical basis for our empirical test is that, prior to changing locations, prospective migrants should be more successful in their labor market activities than nonmigrants.

Section II presents a basic human capital model of migration which incorporates the possibility that migrants have higher ability and/or motivation than nonmigrants. Section II also formulates an empirical model (adapted from the literature on union wage effects) that assesses the link between wages and the subsequent decision to move. Section Ill discusses our data and presents our empirical findings, while Section IV summarizes our results.

II. Model Specifications

Following the standard human capital formulation, we assume that an individual's decision to migrate is based on three general factors: 1) the expected wage differential between destination and origin; 2) the probability of employment in destination and origin; and 3) net migration costs: direct and indirect.(1) More formally, we can specify the present value of the migration decision as,

[Mathematical Expression Omitted]

where, [P.sub.D] = probability of employment at expected wage in destination [P.sub.O] = probability of employment at expected wage in origin [W.sub.D] = expected wage in destination [W.sub.O] = expected wage in origin

r = appropriate discount factor

t = time index (in years, post-migration)

C = net costs (direct and indirect) of migration.

If indirect costs are proportional to foregone earnings (in origin), the net migration costs (C) can be expressed as

C = [delta] [W.sub.o] + DC,

where 8 is a positive constant and DC denotes direct costs (see Chiswick 1978, p. 900).

If we assume an infinite post-migration time horizon and constant employment probabilities and wages, Equation (1) can be written in continuous form as,

[Mathematical Expression Omitted]

Setting Equation (2) equal to zero, we can solve for the internal rate of return to migration,

[Mathematical Expression Omitted]

For a wealth-maximizing individual, the choice to migrate occurs if the net present value [Equation (2)] is positive, or, if the rate of return [Equation (3)] exceeds the opportunity cost of funds. Empirically, the specifications in (2) and (3) suggest that an individual's likelihood to migrate is directly related to [P.sub.D], [W.sub.D], and inversely related to [P.sub.o], [W.sub.o], [delta], and DC.

Following Chiswick (1978), we can modify the migration decision process to account for potential differences in an individual's labor market ability and/or motivation. If an individual possesses greater labor market ability or motivation, then his or her earnings potential should be higher in both the origin and destination. Let [lambda] represent the percentage addition to earnings for the "superior" (self-selected) potential migrant. If we assume no change in direct costs for the self-selected potential migrant, a modified rate of return can be expressed as:

[Mathematical Expression Omitted]

Thus the rate of return to migration (and therefore the chances of migration) tends to be higher for those individuals with greater ability and motivation - that is, for those who are "favorably self selected."(2)

It is reasonable to assume that motivation and labor market ability are positively correlated with earnings. If motivation and ability are also positively correlated with the propensity to migrate - as stated in the favorable self-selection hypothesis - then more productive workers are more likely to move, holding other things constant, than workers with equal observable traits. Potential migrants should therefore have higher earnings potential than nonmigrants with comparable characteristics. This possibility forms the basis of our empirical test for the presence of self selection.

The empirical techniques developed by Mincer (1983) and Mellow (1981) to measure the wage effects of unionization can be adapted to the issue of migrant self selection. In essence, a longitudinal wage model can be specified which compares the wages of prospective migrants and nonmigrants. The basic wage structure of migrants and nonmigrants can be represented by the following human capital wage level equation:

(5) In [W.sub.t], = a + [bX.sub.1], + cMOVE + [e.sub.t]


In [W.sub.t] = log of annual earnings in period t

[X.sub.t] = vector of personal characteristics that are statistically

linked to earnings: such characteristics include education,

experience, occupation, industry, marital status, and health

(see Table Al)

MOVE = 1 if a worker resides in a different location from

period t to t + I

[e.sub.t] = error term N(o,u 2).

The coefficient on MOVE measures the difference in year t's earnings between workers who subsequently migrate (in period t + 1) and those who remain in the same location in both periods. If favorable self-selection exists, we would expect that prospective (ex ante) migrants have higher wages than nonmigrants. Thus, if workers have chosen to migrate by period t + 1, their current wages (period t) should be higher than comparable workers, ceteris paribus, due to unobserved factors such as higher ability and/or motivation. This implies that the coefficient on MOVE is greater than zero. Estimation of the parameter coefficients in (5) allows us to assess the effects of individual traits, including subsequent migration, on a worker's earnings.

An additional factor which may influence the wage model specified in equation (5) is the presence of location-specific amenities (or disamenities). The predominant recent trend in U.S. internal migration is to the Western and Southern regions and away from the North and East. Research by Greenwood, et al. (1991), Hunt (1993), and others, has emphasized the link between location specific amenities (such as crime rates, weather, congestion) and the net migration to or away from an area. Further research supports the notion that areas with low amenity endowments are likely to experience net out-migration, even though workers in these areas may receive higher compensating wages (Hunt 1993, pp. 342-43). Thus, some of the hypothesized wage premium for the favorably self-selected potential migrants may be due to compensating differences for living in a low amenity region.(3) We control for this possibility in two ways. First, regional dummy variables for a worker's residence in period t are added to the wage model in Equation (5). Given these dummy variables, we would expect the wages of individuals in low-amenity regions to reflect the higher compensating wages. Second, using data from Greenwood et al. (1991), we identify a worker's state of residence

III. Empirical Results

The wage specification in (5) requires a longitudinal data set, such as the National Longitudinal Survey of Youth (NLSY), which contains information on the personal characteristics of workers and detailed geographic information on a worker's residence over the relevant time period. The NLSY reports this type of information for individuals who were aged 14 to 22 at the time of the initial survey in 1979. Our base sample consists of white males who were native born, not enrolled in school, not in the active military, and were year round full time workers in both periods t, and t + 1. We restricted our sample to white males to avoid any race/gender biases in the wage structure. Since most moves are based on labor market or economic motives (Long 1988, p. 235), we chose year round full time workers because of their strong labor force attachment. An added benefit of the NLSY data is that young workers are more likely to change residence than older workers (Long 1988, p. 241).

The initial period (in other words, period t) for the wage structure in Equation (5) is the calendar year 1985. We then classify an individual as being a migrant or a nonmigrant based on his reported place of residence in the subsequent time period.

The basic geographic unit we use to measure a change in residence is the Standard Metropolitan Statistical Area (SMSA).(5) The detailed data available in the NLSY allow the MOVE variable in (5) to be modified to capture the effects of two types of residence changes for a worker from period t to period t + 1:

CHANGE SMSA = 1 if the worker resides in a different SMSA over the two periods. TO RURAL = I if the worker moves from an SMSA to a rural area over the two periods.

Positive estimated coefficients on CHANGE SMSA and TO RURAL support the notion that potential migrants are favorably self-selected.(6)

Table 1 presents the descriptive statistics for wages and personal characteristics for the overall sample of white males, and by migration category for the interval 1985 to 1991. The six year interval was initially chosen to obtain a relatively large sample of prospective migrants. Individuals who subsequently changed SMSAs by 1991 tended to have higher wages in 1985 than both nonmigrants and migrants to rural areas. SMSA migrants also had more schooling while nonmigrants had higher experience levels and were more likely to be married and report a work-limiting disability. However, only the differences in education and marital status are significant at conventional levels. Prospective movers to rural areas have lower wages, more schooling, and are more likely to be married than nonmigrants-though the differences are not statistically significant.

Table 1 Means (Standard Deviations) by Migration Category for Selected Variables in 1985. Time Interval for Subsequent Move: 1985-1991
                                          Migrants from
                                            SMSA to

                   Overall                                Different
Variable           Sample     Nonmigrants   Rural Area     SMSA

ANNUAL EARNINGS    17601.72    17392.29     17595.21      18319.96
                   (8455.31)   (8592.78)    (7101.04)     (8722.01)
LOGWAGE                9.65        9.63         9.68          9.70
                      (0.54)      (0.56)       (0.51)        (0.51)
AGE                   23.93       23.90        23.91         24.02
                      (2.27)      (2.30)       (2.18)        (2.23)
EDUCATION             12.68       12.55        12.90         12.99
                      (1.99)      (1.94)       (2.14)        (2.07)
EXPERIENCE            74.48       75.19        74.87         71.83
                     (24.47)     (24.93)      (22.51)       (23.99)
DISAB                  0.02        0.02         0.02          0.01
                      (0.13)      (0.14)       (0.13)         0.09)
MARRIED                0.40        0.41         0.48          0.31
                      (0.49)      (0.49)       (0.50)        (0.47)
SOUTH                  0.28        0.24         0.55          0.26
                      (0.45)      (0.43)       (0.50)        (0.44)
MIDWEST                0.33        0.39         0.24          0.20
                      (0.47)      (0.49)       (0.43)        (0.40)
NORTHEAST              0.23        0.22         0.16          0.29
                      (0.42)      (0.41)       (0.37)        (0.46)
WEST                   0.17        0.16         0.05          0.25
                      (0.37)      (0.37)       (0.22)        (0.43)
LOWAMEN                0.55        0.57         0.34          0.58
                      (0.50)      (0.50)       (0.48)        (0.50)
Sample size          508         348           58           102

Table 2 presents three sets of OLS estimated coefficients for the wage level model in (5). The specification reported in Column 1 contains estimates for regional dummy variables (with the Northeast as the omitted region) as well as a dichotomous variable indicating if a worker's residence in 1985 was a low" amenity state (see Greenwood et al. 1991, pp. 1386-87). Columns 2 and 3 report the wage level estimates with the region and amenity effects separated. The positive and significant estimated coefficients on schooling, experience, and marital status are consistent with results reported in other wage studies.
Table 2
Estimated Coefficients for the 1985 Wage Level Equation
(Dependent Variable: Log of Annual Wage and Salary)

Variable         (1)              (2)           (3)

INTERCEPT       8.675(*)        8.630(*)      8.627(*)
              (38.793)        (40.033)      (41.221)
EDUCATION       0.0527(*)       0.0533(*)     0.0514(*)
               (3.894)         (3.889)       (3.821)
EXPERIENCE      0.0061(*)       0.0061(*)     0.0061(*)
               (6.535)         (6.522)       (6.584)
DISAB           0.1401          0.1454        0.1101
               (0.854)         (0.887)       (0.679)
MARRIED         0.1571(*)       0.1591(*)     0.1550(*)
               (3.332)         (3.381)       (3.392)
CHANGE SMSA     0.0963(*)       0.0936(*)     0.1062(*)
               (1.706)         (1.662)       (1.910)
TO RURAL        0.0066          0.0139        0.0218
               (0.093)         (0.197)       (0.314)
MIDWEST        -0.0890        - 0.0785
               (1.457)         (1.318)
WEST           -0.0822        - 0.0316
               (0.868)         (0.457)
SOUTH         - 0.0371          0.0005
               (0.471)         (0.008)
LOWAMEN       - 0.0510                     - 0.0334
               (0.780)                      (0.769)
R[sup.2]        0.278           0.277        0.284
n             508             508          508

Note: The absolute value of the t-statistic appears in parentheses.
dummy variables were estimated for a worker's industry and
occupation (see
Appendix Table A1)
(*) Significant at the 10 percent level.

The coefficient on CHANGE SMSA (Column 1) indicates that prospective migrants in 1985 earned 9.6 percent more than nonmigrants, ceteris paribus. Columns 2 and 3 report similar wage advantages for prospective migrants. This is consistent with the notion that unobserved factors such as ability or motivation can increase the observed wages of potential migrants relative to nonmigrants with similar character-istics. This finding supports the favorable self-selection hypothesis. For prospective migrants to rural areas, however, we find no evidence of favorable self selection.

In order to obtain a more complete picture of favorable self selection, we examined the wage effects of prospective migration for five other interval periods.

Table 3 reports the average earnings by subsequent migration category for these periods. As expected, prospective migrants to other SMSAS earn more than nonmigrants and migrants to rural areas. To assess the favorable self-selection hypothesis for these intervals, separate wage level regressions were also estimated. Table 4 reports the estimated coefficients for the migration variables, assuming the specification from Column 1 of Table 2. Although prospective migrants to
Table 3
Average 1985 Annual Wage for Migrants and Nonmigrants by Interval

                                   Migrants           Migrants
               Nonmigrants    (to different SMSA)   (to rural
Interval      Average Wage       Average Wage         Average Wage

1985-86         $17,560           $20,051             $20,002
Sample size         639                43                  12
1985-87         $17,635           $19,137             $16,077
Sample size         552                85                  27
1985-88         $17,585           $18,173             $15,698
Sample size         457                93                  66
1985-89         $17,815           $18,900             $15,775
Sample size         418               111                  72
1985-90         $17,619           $18,790             $15,587
Sample size         404               115                  78

Table 4
Migration Coefficients by Interval(a)
(absolute value of t-statistics in parentheses)

Interval      CHANGE SMSA     TO RURAL

1985-86         0.1054          0.1899
               (1.347)         (1.327)
1985-87         0.0611        - 0.1150
               (1.053)         (1.209)
1985-88         0.0545        - 0.0424
               (0.957)         (0.627)
1985-89         0.0709        - 0.0549
               (1.353)         (0.863)
1985-90         0.0866(*)       0.0716
               (1.664)         (1.176)

(a) These estimates are based on the specification in Column 1 of
Table 2.
(*) Significant at the 10 percent level.

other SMSAs earn consistently more than nonmigrants, the difference is significant only for the five-year interval 1985-1990.

IV. Conclusion

The question of whether individuals who migrate are randomly drawn from the population, or are somehow self-selected from a more-able and motivated group has received increasing attention in both the immigration as well as internal migration literature. This study offers an alternative empirical technique which attempts to test directly the favorable self-selection hypothesis with respect to internal migrants in the United States. Our empirical specification was based on the wage level models of Mellow (1981) and Mincer (1983) that examine if union workers possess certain (perhaps unobserved) traits which influence the union/nonunion wage differential. Using NLSY data for 1985 through 1991, we find some support for the favorable self-selection hypothesis for white males who move from one SMSA to another. In general, prospective migrants enjoy a wage advantage over nonmigrants with similar earnings-related characteristics.

Our results suggest that white males who subsequently change locations among SMSAs tend to have higher earnings potential than nonmigrants with similar observable traits. This enhanced earnings potential does not appear to be influenced by compensating differences due to regional variations in location-specific amenities. Our findings can also be interpreted as providing additional support to the results in Bodas, Bronars, and Trejo (1992b). Their study finds that migrants tend to self-select areas which offer better rewards for their particular talents. Even though prospective migrants already earn more than comparable nonmigrants, our observation of an ex post move may indicate that these workers have rationally chosen a new location which places a higher reward on both their observed and unobserved characteristics.

Table A1
Variable Definitions

LOGWAGE:       logarithm of 1985 annual wage and salary income

EDUCATION:     years of schooling completed by 1985
EXPERIENCE:    respondent's actual labor market experience as
               measured by the sum of an months worked
               from 1975 through 1985
DISAB:         = 1 if respondent has a work limiting disability;
               = 0 otherwise
MARRIED:       = 1 if respondent is married, with spouse present;
               = 0 otherwise
MIDWEST:       = 1 if respondent resides in the north central
               Census region in 1985; = 0 otherwise
WEST:          = 1 if respondent resides in the western Census
               region in 1985; = 0 otherwise
SOUTH:         = 1 if respondent resides in the southern Census
               region in 1985; = 0 otherwise
LOWAMEN:       = 1 if state in which respondent resides in
               1985 is a low amenity state; = 0 otherwise
               (see Greenwood et al. 1991)
Additional dummy variables were specified for the following:
Occupation:    managerial, professional, technical, sales,
               service, farming, craft, operatives, laborers

Industry:      agriculture, mining, manufacturing, construction,
               transportation, wholesale, retail, financial
               services, other services, public administration


Becker, Gary. 1964. Human Capital. New York: National Bureau for Economic Research. Borjas, George J., Stephen G. Bronars, and Stephen J. Trejo. 1992a. "Assimilation and the Earnings of Young Internal Migrants." The Review of Economics and Statistics 74(1):170-75. _____. 1992b. "Self-selection and Internal Migration in the United States." Journal of Urban Economics 32:159-85. Chiswick, Barry R. 1978. "The Effect of Americanization on the Earnings of Foreign-born Men." Journal of Political Economy 86(5):897-921. Philip E. Graves, and Peter D. Linneman. 1979. "Household Migration: Theoretical and Empirical Results." Journal of Urban Economics 6:383-404. Greenwood, Michael J., Gary L. Hunt, Dan S. Rickman, and George I. Treyz. 1991. "Migration, Regional Equilibrium, and the Estimation of Compensating Differentials." American Economic Review 81:1382-90. Greenwood, Michael J. 1985. "Human Migration: Theory, Models and Empirical Studies." Journal of Regional Science 25(4):521-44. Hanushek, Eric A. 1981. "Alternative Models of Earnings Determination and Labor Market Structures." Journal of Human Resources 16(2):239-59. Hirsch, Barry T. 1978. "Predicting Earnings Distributions across Cities: the Human Capital Model vs. the National Distribution Hypothesis." Journal of Human Resources 13(3):366-84. Hunt, Gary L. 1993. "Equilibrium and Disequilibrium in Migration Modelling," Regional Studies 27(4):341-49. Long, Larry. 1988. Migration and Residential Mobility in the United States. New York: Russell Sage Foundation. Marshall, Alfred. 1920. Principles of Economics: An Introductory Volume 8th Edition. London: Macmillan and Company. Mellow, Wesley. 1981. "Unionism and Wage Rates: a Longitudinal Analysis." Review Economics and Statistics 63:43-52. Mincer, Jacob. 1983. "Union Effects: Wage Turnover, and Job Training." In Research in Labor Economics, Supplement 2. New York: JAI Press: 217-52. Schultz, Theodore W. 1961. "Investment in Human Capital." American Economic Review 51:1-17. Treyz, George I., Dan S. Rickman, Gary L. Hunt, and Michael J. Greenwood. 1993. "The Dynamics of U.S. Internal Migration." Review of Economics and Statistics 75(2):209-14.

(1.) Direct costs include pecuniary and nonpecuniary moving expenses while indirect costs consist primarily of foregone earnings incurred during transit. (2.) If direct costs are positively related to [W.sub.o] a higher rate-of-return for favorably self-selected migrants would exist as long as changes in the direct costs are proportionately less than changes in [W.sub.o] (Chiswick 1978, pp. 900-901). (3.) We are grateful to an anonymous referee for suggesting the possible effects of location-specific amenities on the wages of potential migrants. (4.) These variables attempt to capture the effect of variations in regional-specific amenities on workers' wages. See Hunt (1993) for a survey (5.) The SMSA was chosen because it approximates the notion of a local market (see Hirsch 1978, and Hanushek 1981). Thus, our notion of favorable self-selection is applied to workers who migrate from one labor market to another. Workers who initially reside in rural areas are excluded because a comparable definition of a local labor market is not readily available. (6.) This assumes that the excluded group is workers who reside in the same SMSA in both periods.
COPYRIGHT 1995 University of Wisconsin Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1995 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:includes appendix
Author:Gabriel, Paul E.; Schmitz, Susanne
Publication:Journal of Human Resources
Date:Jun 22, 1995
Previous Article:Health, income, and risk aversion: assessing some welfare costs of alcoholism and poor health.
Next Article:The effects of interrupted schooling on wages.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters