No abstract

1970s and 1980s has been one of the most remarkable shifts in the structure of labor compensation in recent history-the last major shift in this distribution having occurred during the 1940s.I In the postwar period, earnings and income distributions have been regarded as stable bedrock-that is, until recently-with a common theme being the stability of those distributions in light of massive changes in labor markets and public spending on redistributive programs. Entire books were written on the subject.2 This view of long-term stability has been dramatically altered by an enormous, and still growing, series of investigations showing a widening of the income and wage distributions.3 As documented in an article by Frank Levy and Richard Murnane, the widening of the earnings distribution began gradually in the 1970s and accelerated in the 1980s.4 Earnings differentials by education and experience grew rapidly in the 1980s. A debate has now ensued concerning the causes of these trendswhether they are the result of general shifts in supply or demand caused by skill-biased technological change, shifts in the pattern of interna-1. See Williamson and Lindert (1980) and Goldin and Margo (1992). 2. See, for example, Reynolds and Smolensky (1976). 3. Two of the earliest studies were Dooley and Gottschalk (1984) and Lawrence (1984). 4. Levy and Murnane (1992). 217 218 Brookings Papers on Economic Activity, 2:1994 tional trade, changes in union rents or the minimum wage, or other factors-. Our study suggests that this literature has missed a critical aspect of the widening wage distribution: a growing instability in wages has caused a major part of the trend. The fact that annual cross-sectional "snapshots" of the work force over the 1970s and 1980s reveal an increase in the variance, and other measures of dispersion, of the wage distribution has been interpreted as reflecting solely an increase in the dispersion of average-or what we call "permanent"-wages. We show, instead, that part of the increase in the annual variance of wages for white men has resulted from an increase in the variance of short-term changes in earnings-or "transitory" movements in earnings. We show that an increase in earnings instability has effected a full one-third to one-half of the widely noted increase in the variance of earnings from the 1970s to the 1980s. The recognition of a major upturn in earnings instability prompts many further questions. Does it reflect a true increase in the instability of wage rates, or just an increase in employment turnover orjob mobility? Does it reflect an increase in "voluntary" or"involuntary" movements between jobs? Is it concentrated in particular sectors of the economy? Is it associated with aggregate shocks or sectoral shocks, or is it entirely idiosyncratic on an individual level? We shall provide partial and suggestive answers to these and other questions and present an exploratory investigation of causes. Despite the limited nature of our inquiry, it is unambiguously clear that this search for the causes of...

The authors would like to thank Zvi Griliches, the participants at the Econometrica/SSRI conference "Empirical Applications of Structural Models," held in Madison, Wisconsin, in May 1990, and the participants at several other conferences and university workshops for comments. Support from the Federal Reserve Bank of Minneapolis is also appreciated.

coefficients. We provide a statistical framework for conducting tests for attrition bias that draws a sharp distinction between selection on unobservables and on observables and that shows that weighted least squares can generate consistent parameter estimates when selection is based on observables, even when they are endogenous. Our empirical analysis shows that attrition is highly selective and is concentrated among lower socioeconomic status individuals. We also show that attrition is concentrated among those with more unstable earnings, marriage, and migration histories. Nevertheless, we find that these variables explain very little of the attrition in the sample, and that the selection that occurs is moderated by regression-to-the-mean effects from selection on transitory components that fade over time.Consequently, despite the large amount of attrition, we find no strong evidence that attrition has seriously distorted the representativeness of the PSID through 1989, and considerable evidence that its crosssectional representativeness has remained roughly intact.The increased availability of panel data from household surveys has been one of the most important developments in applied social science research in the last thirty years. Panel data have permitted social scientists to examine a wide range of issues that could not be addressed with cross-sectional data or even repeated cross sections.Nevertheless, the most potentially damaging and frequently-mentioned threat to the value of panel data is the presence of biasing attrition--that is, attrition that is selectively related to outcome variables of interest.In this paper we present the results of a study of attrition and its potential bias in one of the most well-known panel data sets, the Michigan Panel Study of Income Dynamics (PSID). The PSID has suffered a large volume of attrition since it began in 1968--almost 50 percent of initial sample members had attrited by 1989. We study the effect of attrition in the PSID on the means and variances of several important socioeconomic variables --such as individual earnings, educational level, marital status, and welfare participation--and on the coefficients of variables in regressions for these variables. We also examine whether the likelihood of attrition is related to past instability of such behaviors--earnings instability, propensities to migrate or to change marital status, and so on. A companion paper studies the effect of attrition on estimates of intergenerational relationships (Fitzgerald et al., 199733).An understanding of the statistical issues is important to understanding our approach. We provide a statistical framework for the analysis of attrition bias which shows that the common distinction between selection on unobservables and observables is critical to the development of tests for attrition bias and adjustments to eliminate it.However, we show that selection on observables is not the same as exogenous selection, for selection can be based on endogenous observables such as lagged dependent...