Black-White Earnings Inequality,Employment Rates, and IncarcerationBruce Western1Princeton UniversityBecky PettitUniversity of WashingtonNovemb er, 19991Direct all corresp ondence to Bruce Western, Department of So ciology, 2-N-2Green Hall, Princeton University, Princeton NJ 08544-1010. We thank Allan Beckand Darrell Gilliard at the Bureau of Justice Statistics for supplying unpublishedBJS data. Hank Farb er, Jerry Jacobs and Chris Winship made useful sugges-tions on an early draft. This pap er b enetted from seminar discussions at YaleUniversity, UniversityofPennsylvania, and Harvard University. This researchwassupp orted by a grant from the Princeton University Science Fund. The manuscriptwas prepared while the rst author was visiting at the Russell Sage Foundation,New York.AbstractThe high rate of joblessness among black men suggests that the earnings gapbetween blacks and whites may b e partly due to selective attrition from thework force. We use a predictiveBayesian metho d to estimate the black-whiteearnings dierential from 1982 to 1996, adjusting for the sample selectioneect of lab or inactivity. The analysis shows that among all working-agemen, lab or inactivity explains about 13 p ercent of the observed earningsdierential. Sample selectivity among young men reduces observed earningsinequality by ab out 20 p ercent, explaining more than half the rise in blackrelative earnings b etween 1985 and 1996. The sample selection eect is largefor young workers b ecause lab or inactivity among young black men has b eenb o osted by high incarceration rates.Diverging trends in relative earnings and employment oer a basic puz-zle for understanding the economic status of African American men. Theearnings of black men relative to whites grew rapidly from the 1940s. The1940 census shows that the mean weekly earnings of African American menwas just over 40 percent of the earnings of whites. By the 1980s, the rela-tive earnings of black men was around 70 p ercent (Smith and Welch 1989).Despite the app earance of dramatic improvement in the middle of the blackincome distribution, a large gap in employment rates op ened in the 1950s.Employed at almost the same rate in 1940, black men were nearly twice aslikely as whites to be jobless throughout most of the p ostwar p erio d. Mostrecently, during the 1980s and 1990s, inequality in lab or force participationincreased sharply among young workers due to the racially disparate impactof incarceration (Western and Pettit 1999).Because relative earnings and employmentmoved in sharply dierent di-rections, some researchers question whether earnings is a good measure ofblack economic progress. For example, Welch (1990, S42) asks if the \im-provement in black/white wage ratios is an illusion." Jaynes (1990) observesthat \the most imp ortant problem" for research on race relations \is to ex-plain why, if the market's relative valuation of black lab or has increased,black employment has b een declining." These comments reect a suspicionthat increased average earnings is an artifact of declining employment. Iflow earners are discouraged from seeking work or in prison, this pro duces asample selection bias. Observed average incomes increase, not b ecause of anymovement in the income distribution, but b ecause of selective attrition fromthe work force (Heckman 1989). Under these conditions, deep ening economicdisadvantage may b e misinterpreted as economic progress.Whether earnings or employment b etter describ es the economic p osition1of black men also reects dierent p ersp ectives on the U.S. lab or market.Structural forces provide a familiar account of racial dierences in jobless-ness. Declining employment among African Americans|and young blackmales in particular|is attributed to the decline of manufacturing in urbanareas and the spatial isolation of black neighb orho o ds (Wilson 1987; Masseyand Denton 1994). However, an institutional p ersp ective claims that schooldesegregation, improvements in scho ol quality, and civil rights policy raisedrelative earnings of black men (Burstein and Edwards 1994; Donohue andHeckman 1991; Card and Krueger 1992). Wilson (1978, 1987) adopts b othp ersp ectives, tracing the growth of the black middle class to institutionalchange in the 1960s, but linking p ersistent urban poverty to shrinking em-ploymentinmanufacturing industry.This pap er tries to reconcile the divergence of relative earnings and em-ployment of black men by joining insights from the institutional and struc-tural explanations. Following the structural p ersp ective, we treat low em-ployment rates among black men as a sample selection bias that inatesestimates of average earnings. From an institutional viewp oint, large in-creases in U.S. incarceration through the 1980s and 1990s severely reducedlab or force participation among young African American men. By 1996, in-carceration accounted for more than 30 p ercent of all joblessness among blackmales under 30, compared to just 10 percent for young white men. Becauseinmates tend to be unskilled with little earning power, the selection eectof incarceration on black earnings is large. Upward bias in average blackearnings, and downward bias in estimated inequality, has also increased withrising incarceration.To study the impact of incarceration and other joblessness on the relativeearnings of black men, we analyze data from the CurrentPopulation Survey2(CPS), correctional surveys by the Bureau of Justice Statistics (BJS), andadministrative data on incarceration rates. Combining these data yields ad-justments to conventional measures of relative earnings for the p erio d 1982{1996. Our main ndings show that (1) among working-age men, selectiveattrition from the work force explains more than 10 p ercent of the dier-ence between black and white earnings, and (2) the sample selection eectfor young workers explains more than half the rise in black relative earningsbetween 1985 and 1996.Recent Trends in Earnings and EmploymentTable 1 rep orts the mean of log weekly earnings for non-Hispanic blacks andwhites between 1982 and 1996 using the Merged Outgoing Rotation Grouples of the CPS. The data show two patterns. First, the earnings advantageof working-age whites changed little in the 10 years from 1985. White weeklyearnings exceed the earnings of blacks by ab out 30 p ercent. Second, amongyoung men, inequality increased until the mid-1980s and then declined