Occupational change and differing returns to migration by gender.
The internal migration of labor has long been recognized as not only an important means of redistributing labor to its most productive use but a means of increasing one's socioeconomic status. A key element in the linkage between socioeconomic opportunity and the labor market concept of economic efficiency is occupational structure. Structural unemployment exists due to mismatches between job vacancies and available applicants across geographical markets. Possible solutions include migration, occupational change, or both. Changing occupation is particularly salient for those making geographical moves. Data from the Panel Study of Income Dynamics (PSID) sample used in this study indicate that migrants are fifty percent more likely to change occupations than those remaining in the same geographic location.
Critical to both the decision to relocate and the decision to change occupation is the impact such choices would have on earnings. One problem with measuring the effects of occupational change on earnings is that full pecuniary compensation does not accrue immediately. In order to measure lagged gains in earnings, studies must investigate the effect of occupational change on earnings for both migrants and nonmigrants over a sufficient period of time.
OCCUPATIONAL AND GEOGRAPHIC MOBILITY
Labor mobility as studied in this paper can be seen as either a change in occupation or geographic location. Each such change may substantially influence ones earnings. Earnings may be affected positively as in the case of promotions, negatively, possibly resulting from a forced job change due to a layoff, or simply a decision to move on to something new. Thus, whether an occupational change will have a positive or negative effect cannot be posited a priori, but will depend on whether the change is viewed as an advancement.
Numerous studies have noted the relationship between geographic and occupational mobility, including survey articles on migration by Greenwood (1975), and Ritchey (1976). Fewer studies have examined the interaction of occupational change and migration (Shaw, 1991; Lichter et al., 1979; Schlottmann & Herzog, 1984; Wilson, 1985; Congdon, 1988). Such studies examine the effect of occupational mobility on the propensity to migrate, and the effect of migration on career mobility. Wilson contends that migration is positively related to career mobility.
MIGRATION OF MARRIED COUPLES
Recently, human capital studies of migration have analyzed the costs and benefits as they impact family units as opposed to individuals only. Emphasis of migration as a family decision rather than an individual's decision began in the 1970s (see DaVanzo, 1972; Kaluzny, 1975; Sandell, 1977). Mincer (1978) established the theoretical framework for family migration, hypothesizing that a married couple will relocate only if one spouse's financial gain from migrating more than offsets the earnings loss to the other. This implies that subsequent to a couple's migration the tied mover, often the wife, will work less and experience lower earnings. Working status of wives also influences the probability of migration. For example, Rives and West (1993) find that labor attachment of wives deter family migration.
Using data from the National Longitudinal Survey (NLS) Sample of Older Women, Lichter (1983) contends that while migration has a significantly negative effect on earnings of women in the short- term, this effect is minimal over time. This is consistent with an early study by Wertheimer (1970), who finds migrants (both men and women) accept a temporary decline in earnings following migration. More recently, Borjas et al. (1992) find the same conclusion using data from the National Longitudinal Survey of Youth.
Spitze (1984) uses the Young and Mature Women samples, of the NLS to study how migration affects both employment and earnings of married women. Spitze finds that effects of migration are negative for women's employment status, weeks worked, and earnings.
Data from the NLS are also used by Maxwell (1988) in studying the effect of changes in marital status on the returns to migration. Maxwell finds that changing marital status significantly influences the returns to migration for women. Maxwell shows that women who remain married not only suffer immediate losses but that these losses increase over time.
THE DATA AND ECONOMETRIC MODEL
Data used in this study are selected from the Panel Study of Income Dynamics (PSID) for the years 1981-87. The PSID data set is longitudinal, and contains microdata for a representative sample of individuals and their family units residing in the United States. The PSID's rules for tracking individuals and family units over time lead to accurate representation of the U.S. population both cross-sectionally and in terms of demographic change since 1968.
Variables in the data set include demographic and economic characteristics as well as geographical location of individuals for each year. Since pooling return migrants with other migrants may bias the results, only those making one move since 1980 are included in the sample. Students are also excluded from the sample.
When comparing earnings of migrants to those of nonmigrants, the results may be biased. This is because those who migrate may tend to have relatively better economic opportunities elsewhere and relatively lower opportunities at their present location. Such self-selection bias will occur in these comparisons when unmeasured characteristics are related to both the migration decision and earnings. For example, unmeasured characteristics in the migration decision such as ambition and a tendency toward human capital investments are related to both the decision to migrate and earnings. Thus migration is not a random variable. As a result, there is an inherent bias in the effect of migration on earnings.
Migration studies that account for self-selection bias generally separate migrants and nonmigrants into two distinct samples in what has become known as a mover/stayer model (Robinson & Tomes, 1982). My study differs from the traditional mover/stayer model in that it estimates earnings of migrants and nonmigrants together as a pooled sample (see Barnow, Cain, & Goldberger, 1981; Amemiya, 1978; Heckman, 1978; Lee, 1979 for theoretical formulation of this method), while correcting for self- selection bias. This enables a direct estimate of the effect of migration on earnings. Moreover, it allows measurement of the interactive effect of migration and occupation variables.
Semi-log form is used in the estimation of earnings because earnings tend to be skewed to the right since earnings cannot take on a negative value. Consequently, the resulting estimated coefficients represent the rate of return to the various explanatory variables. The model is specified as:
(1) [MIG.sup.*] = f(TENURE, HOUSE, #CHILD, EDU, NCENTRAL, SOUTH, WEST, MIG1, MIG2, MIG3, UNR, CONSTANT)
(2) EARN = f(MIG, MIG1, MIG2, MIG3, EDU, TENURE, RACE, OCCUP, DOCC*MIG,
DOCC1*MIG1, DOCC2*MIG2, DOCC3*MIG3, DOCC4*MIG4, LAMBDA, CONSTANT)
(3) MIG = 1 if [MIG.sup.*] [greater than] 0 and MIG = 0 if [MIG.sup.*] [less than or equal to] 0
MIG = 1 if residence changed during the past year = 0 otherwise
[MIG.sup.*] = one's desire to migrate
TENURE = number of months working for current employer
HOUSE = 1 if homeowner = 0 otherwise
#CHILD = number of children under 18 in the household
EDU = formal years of education
UNR = unemployment rate in county of residence
MIG1 = 1 if residence changed one year ago = otherwise
MIG2 = 1 if residence changed two years ago = 0 otherwise
MIG3 = 1 if residence changed three years ago = 0 otherwise
RACE = 1 if white = 0 otherwise
OCCUP = 1 if occupation is professional or technical = 0 otherwise
NCENTRAL = residence in North Central census region
SOUTH = residence in South census region
WEST = residence in West census region
DOCC = changed occupation during 1986
DOCC1 = changed occupation during 1985
DOCC2 = changed occupation during 1984
DOCC3 = changed occupation during 1983
DOCC4 = changed occupation during 1982
Equation (1) models the probability of migration. Home ownership, job tenure, and number of children generally promote stability and consequently discourage migration, while education can be expected to increase mobility (Ritchey, 1976).
Explanatory variables included in equation (2) are standard regressors used in earnings functions along with lagged migration variables and lagged migration/occupational change interaction terms. Thus while controlling for other determinants of earnings as well as selectivity bias, the model will discern the influence of past migration and occupational change on earnings. This is done for single men and women separately and then for husbands and wives in order to measure the effect of marriage on the returns to migration by gender.
Equation (1) is computed first. From this estimation, an estimate is made of the unobserved correlation between earnings and migration (LAMBDA). This estimate is then used in equation 2 along with traditional human capital explanatory variables to estimate earnings. Thus consistent estimated coefficients in the earnings function will result. This alternative to the traditional mover/stayer model lends itself to measuring the direct effect of migration on earnings as well as the indirect effect occupational change has on returns to migration. For example estimated earnings for one who migrates and changes occupation in the current period, other factors held constant, can be measured as the sum of the estimated coefficients of MIG and DOCC.
In order to obtain an overview of the relationship between migration and occupational status, migration status is cross-tabulated with occupational classification and change in occupation in Table 1.
Migration propensities appear to be related to both occupation and occupational change. Those in professional occupations tend to be relatively mobile, as do those who change occupation. However this does not indicate whether the change was voluntary, i.e., whether migration resulted from the push of a lost job, or the pull of a job in a new occupation. As expected, single individuals have a higher propensity to migrate across occupations than those who are married.
Table 2 provides a descriptive look at earnings by gender, marital status and migration status. Married wage earners earn more than single wage earners. This [TABULAR DATA FOR TABLE 1 OMITTED] is expected since married wage earners tend to be older and have more labor market experience. It is interesting to see that while married migrants earn more than married nonmigrants, single migrants earn less than nonmigrants. This may be because couples tend to be less mobile and need more pecuniary compensation to do so. Of course, to gain better understanding of earnings resulting from migration we must model earnings while controlling for other personal characteristics.
Table 2. Mean Earnings Per Hour by Gender and Marital Status Single Married Migrants Male 11.05 17.10 Female 7.74 9.31 Nonmigrants Male 12.09 13.66 Female 8.44 8.48
Earnings functions controlling for individual characteristics are first estimated for single household heads (Table 3). Traditional explanatory variables are shown to contribute to earnings. Formal education, and experience gained with a firm both add to individual earnings for both men and women. The same is true for those who are white and employed in professional or technical occupations. Men who live in the West census region experience a boost in earnings.
Migration has a significant impact on earnings for both genders. While relocation, other things held constant, has a negative impact on the earnings of single women during the first year, there is no immediate impact on single men. However men who do migrate experience a decrease in earnings when migration is accompanied by a change in occupation, as illustrated by the significantly positive coefficient for DOCC*MIG.
Table 3. Estimation of Earning, Single Household Heads Male Female MIG -0.858 -2.136(**) MIG1 0.466(*) (*)0.003 MIG2 -0.043 0.414 MIG3 0.355 -0.138 EDU 0.061(**) (*)0.126(***) TENURE 0.002(***) 0.002(***) RACE 0.304(***) 0.043 OCCUP 0.095 0.149 DOCC*MIG -0.736(***) -0.012 DOCC1*MIG1 -0.327 -0.080 DOCC2*MIG2 -0.260 -0.179 DOCC3*MIG3 -0.248 -0.113 NCENTRAL 0.207 -0.038 SOUTH 0.009 0.040 WEST 0.254(*) 0.158 LAMBDA 0.653(**) 1.073(***) CONSTANT 1.046(***) 0.764 [R.sup.2] 0.300 0.288 Notes: *** Statistically significant at the one percent level. ** Statistically significant at the five percent level. * Statistically significant at the ten percent level.
Migration does have a positive influence on the earnings of single men during the first year subsequent to a move as given by the positive and significant estimated coefficient for MIG.
Another interesting discovery is the substantial influence of LAMDA in estimations of single men and women. This illustrates that when comparing migrant earnings to those who have not moved there will be an inherent upward bias in the earnings differential. This bias is even greater for single women than men.
The impact of migration is different for dual earning married couples, as given in Table 4. In the first year of a move husbands receive a boost in earnings from migration. For wives the direct effect of migration, MIG, does not yield a significant estimated coefficient. However, wives who change occupation do experience diminished earnings from migration as can be seen by the significance of DOCC*MIG. Significance of MIG1 for husbands indicates that they experience lower earnings for a year later, due to migration.
Other human capital variables prove to enhance earnings. Education, tenure with employer, and working in a professional occupation add to earning power. Living in the South and being nonwhite diminish earnings for husbands.
A number of socioeconomic findings have been made in this paper. First, a link between occupational change and migration was established. Those who change occupations tend to be more mobile, as do those in professional occupations in general.
Table 4. Estimation of Earnings Husbands Wives MIG 0.837(*) 0.239 MIG1 -0.227(*) 0.159 MIG2 0.102 0.010 MIG3 0.195 -0.054 EDU 0.077(***) 0.081(***) TENURE 0.002(***) 0.002(***) RACE 0.033 0.089(**) OCCUP 0.164(***) 0.235(***) DOCC*MIG -0.265 -0.582(***) DOCC1*MIG1 0.071 -0.157 DOCC2*MIG2 -0.107 -0.150 DOCC3*MIG3 -0.147 0.264 NCENTRAL -0.060 -0.064 SOUTH -0.129(*) -0.061 WEST -0.007 0.051 LAMBDA -0.290 -0.032 CONSTANT 8.441(*) 8.610(*) [R.sup.2] 0.333 0.338
Second, the direct effect of migration, along with the indirect effect of occupational change associated with migration, on earnings was measured form multiple periods subsequent to migration. Estimations showed that migration and occupational considerations do play a significant role in economic well-being by gender and marital status.
Third, an upward bias was discovered that exaggerates the financial returns to migration for single men and women. This bias is particularly strong for women.
Amemiya, T. (1978). The estimation of a simultaneous-equation Generalized Probit Model, Econometrica, 46, 1193-1205.
Barnow, B.S., Cain, G.C., & Goldberger, A.S. (1981). Issues in the analysis of selectivity bias. Institute for Research on Poverty Reprint Series, University of Wisconsin, Madison, 425.
Congdon, P. (1988). Heterogeneity and timing effects in occupational mobility in the irish republic. In M.D. Uncles, (Ed.) Longitudinal data analysis: Methods and application, London: Pion Ltd.
DaVanzo. J. (1972, September). Analytical framework for studying U.S. interregional migration. RAND corp. Research Report.
Greenwood, M.J. (1975). Research on internal migration in the United States: A survey, Journal of Economic Literature, 13(2), 397-433.
Heckman, J.J. (1978). Dummy endogenous variables in a simultaneous equation system, Econometrica, 46, 931-960.
Kaluzny, R.L. (1975). Determinants of household migration: A comparative study by race and poverty level, Review of Economics and Statistics, 57, 269-274.
Lee, L. (1979). Identification and estimation in binary choice models with limited (censored) dependent variables, Econometrica, 47, 977-996.
Lichter, D.T. (1983, December) Socioeconomic returns to migration among married women, Social Forces, 62(2), 487-503.
Lichter, D. T., Heston, T., & Fuguit, G. (1979). Trends in the selectivity of migration between metropolitan and nonmetropolitan areas: 1955-1975, Rural Sociology, 44, 645-666.
Maxwell, N. (1986). Economic returns to migration: Marital status and gender differences, Social Science Quarterly, 69, 108-121.
Mincer, J. (1978). Family migration decisions, Journal of Political Economy, 86, 749-773.
Ritchey, P.N. (1976). Explanations of migration, Annual Review of Sociology 2, 363-404.
Rives, J.M., & West, J.M. (1993, Spring). Wife's employment and worker relocation behavior, Journal of Socio-economics, 22(1), 13-22.
Robinson, C., & Tomes, N. (1982, August). Self-selection and interprovincial migration in Canada, Canadian Journal of Economics, 15(3), 474-502.
Sandell, S.H. (1977). Women and the economics of family migration, Review of Economics and Statistics, 59, 406-414.
Schlottmann, A., & Herzog, H. (1984). Career and geographic mobility interactions for the age selectivity of migration, Journal of Human Resources, 19(1), 72-86.
Shaw, K. (1991). The influence of human capital investment on migration and industry change, Journal of Regional Science, 31(4), 397-416.
Spitze, G. (1984). The effect of family migration on wives' employment: How long does it last? Social Science Quarterly, 65, 21-36.
Wertheimer, R.F. (1970). The monetary rewards of migration within the U.S., Washington, D.C.: The Urban Institute.
Wilson, F. (1985). Migration and occupational mobility: A research note, International Migration Review, 19, 278-292.
|Printer friendly Cite/link Email Feedback|
|Author:||Krieg, Randall G.|
|Publication:||The Journal of Socio-Economics|
|Date:||Dec 1, 1996|
|Previous Article:||Confucian ethics and economic development: a study of the adaptation of Confucian values to modern Japanese economic ideology and institutions.|
|Next Article:||The decision to quit smoking: theory and evidence.|