Migration, fixed costs, and location-specific amenities: a hazard analysis for a panel of males.

Kenny, L.W., and D.A. Denslow, Jr., 1980. "Compensation Differentials in Teachers' Salaries." Journal of Urban Economics 7:198-207.

Kiefer, N.M. 1988. "Economic Duration Data and Hazard Function." Journal of Economic Literature 26:649-79.

Lancaster, T. 1990. The Econometric Analysis of Transition Data. New York: Cambridge University Press.

Marglin, S.A. 1963. Approaches to Dynamic Investment Planning. Amsterdam: North-Holland Publishing.

Mincer, J. 1978. "Family Migration Decision." Journal of Political Economy 86:749-73.

Mueser, P.R., and P.E. Graves. 1995. "Examining the Role of Economic Opportunity and Amenities in Explaining Population Redistribution." Journal of Urban Economics 37:176-200.

Neal, D., and W.R. Johnson. 1996. "The Role of Pre-Market Factors in Black-White Wage Differences." Journal of Political Economy 104:869-95.

Pashigan, B.P. 1979. "Occupational Licensing and the Interstate Mobility of Professionals." Journal of Law and Economics 22:1-25.

Pindyck, R.1991. "Irreversibility, Uncertainty, and Investment." Journal of Economic Literature 29:1110-52.

Pissarides, C.A., and J. Wadsworth. 1989. "Unemployment and the Inter-Regional Mobility of Labor." Economic Journal 99:739-55.

Roback, J. 1982. "Wages, Rents, and the Quality of Life." Journal of Political Economy 90:1257-78.

--. 1988. "Wages, Rents, and Amenities: Differences among Workers and Regions." Economic Inquiry 26:23-44.

Rosen, S. 1986. "The Theory of Equalizing Differences." In O. Ashenfelter, and R. Layard, eds. Handbook of Labor Economics. Amsterdam: North-Holland, pp. 641-692.

Sandefur, G.D. 1985. "Variations in Interstate Migration of Men Across Early Stages of the Life Cycle." Demography 22:353-66.

Schlottmann, A.M., and H.W. Herzog, Jr. 1981. "Employment Status and the Decision to Migrate." Review of Economic Statistics 63:590-8.

Schultz, T.W. 1961. "Investment in Human Capital." American Economic Review 51:1-7.

--. 1975. "The Value of the Ability to Deal with Disequilibria." Journal of Economic Literature 13:827-46.

Schwartz, A. 1976. "Migration, Age, and Education." Journal of Political Economy 84:701-19.

Tokle, J.G., and W.E. Huffman. 1991. "Local Economic Conditions and Wage Labor Decisions of Farm and Rural Nonfarm Couples." American Journal of Agricultural Economics 73:652-70.

Topel, R.H. 1986. "Local Labor Markets." Journal of Political Economy 94:S111-43.

U.S. Department of Commerce, Bureau of the Census. 1968-1993 Statistical Abstract of the United States. Washington DC.

Wooldridge, J.M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: The MIT Press.

(1) See Marglin (1963) and Harberger (1972) for an early discussion of the timing issue in investment projects. The growing literature on "real options" has rediscovered this principle (Pindyck 1991).

(2) Roback (1988) also discusses conditions where similar migration outcomes occur.

(3) When the annual rate of interstate migration is very low, as it is in the PSID, a discrete probit model of interstate migration based on a comparison of indirect utility values at the origin and destination has little explanatory power, unless variables that are jointly determined with migration are used as regressors.

(4) The hazard function H(t) is obtained from the survival function as H(t) = - dlnS(t)/dt, so there is a sign reversal of coefficients in going between the survival and hazard functions. The Weibull distribution is monotone (constant, increasing, or decreasing) in duration, but employment duration is generally nonmonotonic concave in duration (Lancaster 1990, p. 9: Gritz 1993). We permit this pattern in resident duration by adopting a mixture distribution--a Weibull and a gamma (Heckman and Singer 1985: Greene 2003, pp. 790-798). An alternative distribution with this pattern is log-logistic (Kiefer 1988; Lancaster 1990: Greene 2003, p. 794). All are distributions for a nonnegative random variable.

(5) Heterogeneity arises when a population of resident spells has potentially different distributions of duration after controlling for the effects of observables. The gamma distribution is frequently used for representing the distribution of v associated with unobserved heterogeneity. Chamberlain (1985) and Heckman and Singer (1985) have shown that failure to include individual heterogeneity, when it is present, causes significant bias in the estimated coefficients of the regressors in the hazard function. Han and Hausman (1990) have shown that a parametric gamma distribution of unobserved heterogeneity leads easily to estimable models and is not unduly restrictive.

(6) Individuals were selected into the panel in 1968 or added as new adult males because they either grew into adult males or joined the household of an original 1968 data panelist. Individuals who leave an original PSID household are followed and reinterviewed for as long as they can be located. Among 915 adult males in the PSID SRC sample in 1968, 56 (6.12%) died, 144 (15.74%) were lost, and 132 (14.4%) retired during 1968-87. Individuals who refused to participate in the survey at any time were deleted from the sample. Individuals who were classified as "lost" consist of those who joined the army (30.5%), and those who moved out from the United States and/or were really lost (69.5%). We did not adjust for attrition or reweight.

(7) We do not define migration as moves across standard metropolitan statistical areas (SMSAs), because the boundaries of SMSAs change over time and, furthermore, they are not exhaustive in their area coverage. Roback (1982, 1988), however, uses SMSA labor market unit.

(8) We acknowledge that U.S. states vary considerably in geographical area (e.g., across states like Colorado, Washington, California, New York, Florida, Texas) population and labor force size (e.g., as between Rhode Island and California), and number of cities.

(9) Given that we take the male's age as fixed at the beginning of a resident spell, the effect of the time dependence hazard rate is both the result of the individual aging and of pure time dependence. For making policy recommendations, this is the outcome that we want.

(10) Marriages vary in length, but we see no a priori reason for the decision on interstate migration to be made jointly with marital status.

(11) An alternative would be to define unemployment as an indicator of an individual being unemployed at the beginning of a spell, but this seemed too narrow. Our study is not a study primarily about unemployment/employment.

(12) The time trend also captures the impact of omitted variables that are correlated with time. This will improve the properties of the estimated coefficients in the wage equation (10).

(13) The author does include heating-degree days but not other indicators of the climate.

(14) An alternative route for simplifying the interstate migration problem is to assume that an equilibrium sorting process exists across k states such that the distribution of wage rates across states is perfectly correlated. Individuals would then be assumed to move to the state where their real wage is highest, and the same random variable (education) would determine an individual's potential earnings in each region. Educational attainment is then a matching mechanism. However, matching is likely to be on more than one attribute, which makes this alternative quite unattractive.

(15) Note that we do not claim to know exactly where an interstate migrant moves. We do this in an attempt to get free of the problem that choice of a new location is endogenous.

(16) Since a male's actual earnings at the origin reflect idiosyncratic factors and the predicted wage at a new location reflects expected compensation, we have captured two types of important factors that can be expected to affect a migration decision. For example, a male who is experiencing a large negative idiosyncratic shock to his wage at the origin will be more likely to migrate than other similar males. Our empirical measure of [DELTA]W is a proxy for the true measure and reduces potential endogeneity and selectivity problems, but we do not claim to have captured the effects of all relevant factors.

(17) Since wage rates are expressed in constant purchasing power, ignoring discounting is of little consequence over the study period.

(18) A copy of the methodology used for treating left-censored spells is available from the authors upon request.

The authors are C.F. Curtiss Distinguished Professor of Agriculture and Professor of Economics, Iowa State University, Ames, IA, and Economist and Ph.D. graduate of Iowa State University working in Washington DC, respectively. Table 1. The Distribution of the Number of Interstate Moves by Adult Working-Age Males, Twenty-Year Period, Starting in 1968

(Cumulative) Number of Moves Frequency Percent Frequency Percent 0 722 78.9 722 78.9 1 97 10.6 819 89.5 2 59 6.4 878 96.0 3 23 2.5 901 98.5 4 9 1.0 910 99.5 5 3 0.3 913 99.8 6 2 0.2 915 100.0 Source: PSID files. Table 2. Variable Names and Sample Means of Variables for the Male Hazard Function of Interstate Migration

Sample Means Symbol Variable Description W([ht.sub.is]) The average real hourly wage over the resident