The Paradox of Lessening Racial Inequality and Joblessness Among Black Youth: Enrollment, Enlistment, and Employment, 1964-1981*

THE PARADOX OF LESSENING RACIAL INEQUALITY AND JOBLESSNESS AMONG BLACK YOUTH: ENROLLMENT, ENLISTMENT, AND EMPLOYMENT, 1964-1981*

ROBERT D. MARE CHRISTOPHER WINSHIP University of Wisconsin-Madison Northwestern University and NORC

An important exception to improvementsin the relative socioeconomic status of blacks duringrecent decades is increased levels of joblessness among black youths relative to whites. Few proposed explanationsfor this trend reconcile worsening employment status for- black youths with improvementson other socioeconomic indicators. Threemechanisms that link reducedstatus differencesbetween the races in other spheres with increased disparity in employment are: (1) increased substitutionof schooling and militaryservice for employmentby young blacks; (2) reduced workexperience and disruptedemployment for young blacks at older ages as a result of later average ages leaving school and the armed forces; and (3) "creaming"from the civilian out-of-schoolpopulation of young blacks with above average employmentprospects as a result of higher school enrollmentand military enlistmentrates. Empiricalassessment of these argumentsshows that they account for a substantialpart of the growingracial employmentdifference among men aged 16 to 29. Although racial convergence on school enrollment and educational attainmenthas reduced other socioeconomic inequalities between the races, it has widenedthe employmentdifference.

One of the most important changes in 1977; Hill, 1978; Wilson, 1978; Collins, 1983), American society in recent decades has been statistical differences in socioeconomic welfare the lessening of socioeconomic differences between the races have unmistakably declined. between blacks and whites. Since World War The pattern of these changes suggests that II blacks and whites have converged on many these reductions in inequality may indeed per- indicators of socioeconomic status, such as sist inasmuch as the greatest convergence in grades of school completed, the proportion of indicators of educational and labor market suc- workers in managerial and professional occu- cess has occurred for young adults (e.g., pations, earnings, the quality of schools at- Welch, 1973; Smith and Welch, 1978; tended, economic returns to schooling, and Freeman, 1973; Farley, 1983). numbers of elected officials (e.g., Farley, 1983; A key exception to these trends, however, Freeman, 1973, 1976). Although views are has been divergence between the races in mixed on whether these changes indicate the levels of employment and unemployment for permanent break-up of historic patterns of ra- teenagers and young adults. Despite stable or cial inequality in the United States or are converging trends in race differences in job- mainly short-term results of exceptional eco- lessness for adult workers and otherwise salu- nomic growth and political effort (e.g., Farley, tary trends on other socioeconomic indicators 1983; Freeman, 1973; Butler and Heckman, for persons under thirty, race differences in proportions of the youth population employed and proportions of the youth labor force unem- * Direct all correspondence to: Robert D. Mare, ployed have grown. For example, in 1954 black Department of Sociology, University of Wisconsin, and white unemployment rates for 16 to 24 year Madison, WI 53706. olds were 15.8 and 9.9 percent respectively, a This research was supported by the National Sci- difference of 5.9 percentage points that grew to ence Foundation and the Wisconsin Center for Edu- 8.7 in 1960 and 12.0 in 1970. In 1980 the rates cation Research. Computations were performed at were 26.4 and 12.0 percent respectively, a dif- the University of Wisconsin Center for Demography ference of 14.4 percentage points. Similarly, and Ecology, supported by the National Institute of the percentage of 16 to 24 year olds who were Child Health and Human Development. We are employed declined for blacks from 47.2 in 1954 grateful to Wayne Bigelow, Ann Kremers, and War- to 40.6 in ren Kubitschek for research assistance, and to Zietta 1980, but increased for whites from Feris and Stuart Rakoff of the Department of De- 49.7 in 1954 to 62.0 in 1980 (U.S. Department fense and Jennifer Peck of the Bureau of the Census of Labor, 1982). These trends represent a dra- for supplying us with unpublished tabulations of the matic deterioration in the relative labor market armed forces. standing of black youths and raise doubts

American Sociological Review 1984, Vol. 49 (February:39-55) 39 40 AMERICAN SOCIOLOGICAL REVIEW about future convergence in racial socioeco- have differed markedlyfor blacks and whites. nomic trends in the adult labor market when Most important,as a result of more favorable current youth cohorts reach maturity (e.g., family backgrounds, improved quality of Congressional Budget Office, 1982; Freeman schools, reduced discriminationby institutions and Wise, 1982). of higher education, and better labor market Despite the prominenceof these trends, they incentives, black school attendance has in- have not been satisfactorily explained. Many creased markedly over this period. Young proposed explanations focus on social and blacks now spend more of their time in school economic changes believed to hurt black and leave school at later ages than in the past. youths disproportionately:the spread of mini- For whites, school enrollment rates grew mum wage legislation;the rapid growth of the graduallyinto the 1960sbut have been stable or youth populationin the aftermathof the post- declining since then. Similarly, trends in mili- war baby boom; increased labor market com- tary service have differed over this period for petition among, women, immigrants, and blacks and whites. For all men in the period youth; unfavorable changes in the job com- since the mid-1950s the armed forces peaked position and physical location of industry; during the Vietnam mobilization and have shortfalls in aggregate demand for labor; and since contracted. The post-Vietnamera, how- reduced willingness of youths to take low- ever, has witnessed a reversal of historically status employment (e.g., Congressional Bud- higher rates of military enlistment for whites. get Office, 1982; Osterman, 1980). The im- Blacks are now disproportionatelyrepresented portance of these factors has not been deter- in the armed forces and among veterans. mined empirically,although their effects can in Changes in race differences in employment many cases be questioned on the groundsthat may result in part from changing race dif- they appearto alter the employmentchances of ferences in the structureand timing of young black and white youths alike and leave unaf- persons' movement from schooling and the fected the race difference(e.g., Mareand Win- armed forces to work. This article develops ship, 1979). More important, whatever the this conjecture by considering three quantitative impact of these factors, existing arguments:(1) that young blacks, much more accounts of trends in the youth labor force fail than their white counterparts,are increasingly to reconcile the deterioratingtrend in black substitutingschooling and military service for youth employment with progress on other so- work; (2) that delayed ages at leaving school cioeconomic indicators for blacks generally and increased representationin the recent vet- and recent black entrants to the labor force in eran populationhave reduced average years of particular. civilian work experience for young blacks; and This article considers explanations for the (3) that because schools and the militaryretain changingrelative employment status of black young persons with better than average em- and white youths that, at least in part, link this ployment prospects, rising black school en- changeto other trendsmore favorableto young rollmentand militaryenlistment have reduced blacks. It examines the implicationsof racial the average attractivenessto employers of the convergence in patterns of movement from relatively smaller out-of-school civilian youth schoolingand the militaryto work for trends in population that remains. We develop these relative levels of black and white youth em- arguments and assess them empirically using ployment. More specifically, it explores the data from the March CurrentPopulation Sur- effects of trends in school enrollmentand mil- veys (CPS) of 1964 through 1981. itary service for changes in proportionsof the As shown below, these mechanismsaccount races employed amongmen aged 16 to 29 from for a substantial part of the broadeningrace 1964 to 1981. difference in fractions of young persons who High levels of joblessness for young persons are employed. An important feature of our mainly result from their participationin activi- argumentis that it reconciles the salutarytrend ties that compete with work for their time (for toward educational equality between blacks example, schooling or military service); from and whites with broadening employment dif- differences between adults and younger job ferences for young men. It shows that past race seekers in educationalattainment, work expe- differences in youth employment were partly rience, and attractivenessto employers gener- concealed by race differencesin the transitions ally; and from difficulties youth experience in from schooling and the military. The elimina- moving from other activities to full-time par- tion of the latterdifferences reveals substantial ticipation in the labor force (Freeman and and persistent underlying racial inequality in Wise, 1982; Mare et al., forthcoming;Oster- the youth labor market. We illustrate, there- man, 1980).The postwar period has witnessed fore, the complexity of changes in racial significantchanges in the distributionand tim- stratificationby showingthat the eliminationof ing of activities for young men, trends which some inequalities reveals or induces others. RACIAL INEQUALITY AND JOBLESSNESS 41 RELATIONSHIPSAMONG of schools and the military account for ap- ENROLLMENT, ENLISTMENT, AND proximately 80 percent of the difference in EMPLOYMENTTRENDS proportionsemployed at age 30 and age 16 for Enrollment, Enlistment, and recent cohorts of American men (Mare et al., forthcoming). Intracohort Employment Growth In research reported elsewhere (Mare et al., forthcoming)we have examined the effects of Changing Competition of Schooling school enrollmentand militaryservice on em- and Military ployment growth during a cohort's young As noted above, age-specific rates of school adulthood(ages 16 to 29). Employmentproba- enrollmenthave increasedmarkedly for young bilities rise with age in a cohort as young men blacks, whereas they have increased more leave school and the armed forces and seek slowly or declined for whites. Because stu- civilianemployment. The mechanismsthrough dents work less than nonstudents, enrollment which employmentrises are threefold. (1) Stu- trends have significantlyreduced the propor- dents work less than nonstudentsas a result of tion of young blacks who are employed and the time limits that study places upon work, raised or held stable the proportionfor whites. shortagesof part-timejobs compatiblewith ac- The effects of enlistment on employment are ademic schedules, and their access to schol- largely definitional.If the armedforces are re- arships and parental support. Since the pro- garded as not employed, increasing participa- portion of a cohort enrolled in school declines tion in the armed forces by young black men with age, the proportionemployed rises as the relative to whites accounts in part for blacks' temporal and financial disincentives to work reduced relative levels of civilian employment. disappear.(2) Recent school leavers and veter- Conversely, if the armed forces are defined as ans experience high rates of joblessness be- employed (National Commission on Employ- cause they lack work experience and change ment and UnemploymentStatistics, 1979), the jobs often as they and their employers attempt worseningrelative employment status of young to find a satisfactorymatch (Lazear, 1977;Os- black men is overstated. terman, 1980).As a cohort ages, the proportion of men in the vulnerablestage of newly leaving school or the armed forces declines, thereby Changing Patterns of Movement from raising average employment probabilities. Schooling and the Armed Forces to Work (3) The age pattern of employmentfor out- As black school enrollmentrates have risen, at of-school civilians mirrors the timing of de- any age the proportionof young out-of-school parture from school and the armed forces of blacks who are "recent school leavers" has men whose employmentprospects are best. As increased. Whereas in the past typical black a cohort ages from its mid-teens onward, the 20 year olds, for example, were out of school average educational qualificationsof persons for several years, that age group now includes newly leaving school rise. Persons leaving many more persons who have been out of school later also tend to have more advanta- school less than two years. Conversely, if geous family backgrounds, higher levels of whites are not leaving school any later, they measuredability, and attributesgenerally more experience no reductions in age-specific aver- attractive to potential employers than their age levels of work experience and no increases counterparts who drop out earlier because in fraction of young men newly out of school. factors that bode well for labor marketsuccess New veterans experience similar or higher also bode well for academic success. Thus as a rates of joblessness comparedto recent school cohort ages, its out-of-school populationis in- leavers. Because young blacks are now over- creasinglymade up of individualswith the best represented in the- armed forces and among employmentprospects, thereby increasingthe young veterans, they experience declining rel- proportionof the cohort that is employed. With ative employment levels. regardto the armed forces, enlistees typically have average education, family background, and ability, but above average civilian em- Changing Selectivity of Schools ployment probabilitiesbecause unemployment and the Armed Forces is concentratedamong the least educated, least As enrollment and enlistment rates change able, and least advantaged.Paralleling the ef- over time, the proportionof the out-of-school fects of the timingof school attrition,the influx civilians who are employed may change in re- of relatively capable enlistees into the civilian sponse to the varyingdegree to which schools labor force in a cohort's late twenties raises and the armedforces "cream"young men with cohort employmentrates. Taken together, the above average employment prospects. For competitive, disruptive, and selective effects young blacks, increases in rates of school en- 42 AMERICAN SOCIOLOGICAL REVIEW rollment may have removed from the out-of- 1971 and of age by race by grades of school school populationyoung men who in past years completed for the period 1972-1981. They are would not have attended school but would deflated by the approximate CPS sampling have been relatively successful in securingem- fractions, althoughthey are based on a census ployment (in comparisonto men with charac- of the armed forces. teristics leading them to leave school early in To reduce the computational burden, the all years). For whites, in contrast,to the extent CPS data are grouped into a table with the that school enrollment rates have declined dimensions:(1) Employmentstatus (employed, somewhat in recent years for some age groups, not employed);(2) Schooling (less than 12, 12, employment among out-of-school men may more than 12 grades); (3) Age (7 two-year have increased because of the greater attrac- categories from 16 through 29); (4) Race tiveness to employers of those men who in (nonblack, black); (5) Veteran Status (nonvet- earlier years would have attended school. eran, veteran); (6) Enrollment/YearsOut of Similarly, as blacks have moved from being School Status (enrolled, first year out of underrepresentedto overrepresented in the school, second or third year out, fourth or military, the degree to which the military more year out); (7) Survey Year (1964, . . selects men with above average employment 1981). Enrollment status is ascertained from prospects has become greater for blacks than the "major activity" item in the March CPS, for whites. which identifiespersons spendingmost of their time at school.1 Years out of school is estimated from age and highest grade of school The Effects of Trends in completed.2Age is parameterizedas having a Grades of School Completed lineareffect withinage intervals(16-19, 20-23, The above arguments are not to gainsay the 24-29). positive effects of rising black educationalattainment on employment. Employment STATISTICALMETHODS chances vary directly with grades of school completed (e.g., Feldstein and Ellwood, 1982; The multivariate analyses include single- Nickell, 1979). That black employment has equationprobit models that predictwhether an fallen relative to that of whites despite racial individualis employed (dr) and two-equation convergence in grades of school completed probit models that jointly predict employment suggests that, in the absence of the lattertrend, for out-of-school men and whether or not an the black-white employment gap would have individualis in the not enrolled civilian popu- grown even more. Thus trends in grades of lation. In the single-equationmodel, an indi- school completed and the trends in school en- vidual's probability of employment is non- rollment rates that underlie them may have linearly related to the independent variables. offsetting effects on employmenttrends. For the ith individual,

DATA = 1) =cY1 1 exp ( ty) dty (1) p(dyi -00 We use public-use data files from the March add 2 Current Population Surveys (CPS) for the years 1964 through 1981 and unpublishedDe- I "MajorActivity" is coded on the CPS files for partmentof Defense (DOD) tabulationsfor the 1968-81, but not for 1964-67. For all 18 years, how- third quarterof each year. From the CPS, we ever, it is possible to identifypersons who were not select civiliannoninstitutional men aged 16-29, working because they were in school or who were a total of 260,840 observations.The numberof working part time because they were attending school. Classifying persons as enrolled who meet independent observations, however, is ap- either of these criteriaclosely approximatesdefining proximately50 percent of this because the ro- enrollmenton the basis of the MajorActivity. tation group structureof the CPS dictates that 2 Yearsout of school is approximatedas age minus one-half of its housing units are visited in the highest grade of school attended minus five. Thus, same monthone year later (U.S. Bureauof the age, gradesof schooling, and years out of school are Census, 1978). The CPS, moreover, is a mul- linearlydependent and their separateeffects on em- tistage, stratifiedcluster sample. Thus the as- ploymentare generallynot identified.In the analyses sumptionof simple randomsampling, made in presented here, identificationis achieved through the multivariateanalyses reportedhere, is not aggregationof categoriesof schoolingand years out of school, as described in the text. The estimated met. Because reported test statistics do not effects of these variables agree closely with cross- allow for the nonrandomnessof the CPS sam- sectional analyses of data that do not suffer from ples, the statistical significance of estimated linear dependence of the three variables because parametersis typically overstated. The DOD they contain an independentmeasure of time out of tables are of age by race for the period 1964- school (Mare et al., forthcoming). RACIAL INEQUALITY AND JOBLESSNESS 43

where cyi = 4,8kXik, Xik denotes the kth inde- represent race-specific employment proba- pendent variable (k= 1,. . . ,K), and Ak are the bilities as: probit coefficients. This model is applied to (D-1 (p[dy, = 1]) = f30 + y4di young men as a whole to estimate year and 81 enrollmenteffects on employmentand to out- + Y. 3tdtj of-school civilian men to assess the effects of t=65 additionalvariables. We reportnot only the Afk 81 + X ytdidti, in (1) but also the quantities ap(d= 1) ,which t=65 aXk where '1 is the standard normal distribution measure the effects of the kth independent function, di is a dummy variable that equals variable on the probability of employment one if the ith individual (i= 1, . . . ,N) is black evaluated at the sample proportionemployed. and zero otherwise, dti is a dummy variable The fk provide a suitable means of comparing that equals one if the ith individualis observed the effects of the independentvariables across in the tth year (t= 1965, . . . 1981) and zero age groups inasmuchas they are unaffectedby otherwise, and the f8s and vs are coefficients. To summarize changes in race-specific em- varyingproportions employed. The ap(d,= 1) ployment, we then specify year-to-year aXk changes as resulting from two components, linear trend and business cycle; that is, directlymeasure the impactof the independent variables on the employment probability. D-L (p[dyi = 1]) = 8o + yddi + A1Yi In the two-equation model, (1) for out-of- + PA2U school men is combinedwith a similarequation + yldiYi, (2) predictingwhether an individualis in the out- of-school civilian population estimated over where Yi denotes the year in which the ith the total population (e.g., Heckman, 1979; individual is observed (Y= 1964, . . . ,1981), Judge et al., 1980). This model allows for and Ui is the average unemploymentrate of common unmeasuredvariables (e.g., ability or men aged 25-54 in the year in which the ith attractivenessto employers)to affect the prob- individual is observed. The latter measure abilitiesof employmentand enrollmentand as- summarizesthe overall level of economic ac- sesses the degree of correlationbetween them. tivity in each year. Because the trend-cycle The model is tantamount to augmenting (1) model satisfactorily predicts employment with a latent variable that predicts enrollment trends, we adopt this parameterizationfor the and enlistment. If this latent variableis corre- balance of the analysis. lated with independent variables included in To assess the effects of enrollmenton em- (1), their coefficients will differ between the ployment, we augment(2) with a dummyvari- single- and two-equationmodels.3 able denotingwhether an individualis enrolled in school (or enlisted in the armed forces). In EMPIRICALSPECIFICATION addition, we consider whether the depressing effects of enrollmenton employmentdiffer by The analysis consists of two parts, each of race, with the business cycle, and over time, which is performed within the age groups and whether there are race-specific trend and 16-19, 20-23, and 24-29: (1) description of cyclical effects of enrollmenton employment. trends in racial employmentdifferences for all We investigate these effects by furtheradding young men and the effects of enrollmenttrends to equation (2) variables denoting the interac- on employmenttrends; and (2) analysis of the tions amongrace, enrollmentstatus, trend, and causes of employment trends and their race business cycle. differences for out-of-school civilian men. As discussed below, race differences in employment differ between students and nonstu- between All Men dents; employmenttrends also differ Trends in Employment for Young these two groups, but there is little evidence of To describe employment trends we consider more complex effects of race, enrollment several single-equationprobit models. We first status, trend, and cycle on employment.Using the resulting simplified model, we adjust em- I It would be preferableto regardenrollment, en- ployment probabilitiesfor the business cycle listment,employment, and joblessness as four sepa- and decompose change in employment from rate alternatives. Although probit models for four for changes in alternativesare available(e.g., Hausmanand Wise, 1964to 1981into components (1) 1978), reliable statistical software is only available enrollment rates; (2) employment trends for for three-choicemodels such as the one considered students; and (3) employment trends for non- here. students. 44 AMERICAN SOCIOLOGICAL REVIEW Employment Trends for Out-of-School Men ment for blacks may imply that the race difference on this latent variable in the out-of- To investigate trends for out-of-schoolcivilian increas- men we examine the effects of measured and school civilian populationhas grown After ingly unfavorableto blacks and thus contrib- unmeasured variables on employment. in em- examining the race-specific trend in employ- uted to the widening race difference ment for out-of-school civilians using a model ployment. of the same form as (2), we consider additional independentvariables in models of the form FINDINGS Trends in Employment, Enrollment, = fo + ?-1 (p[dy= 1]) + yodi f1Yi and Enlistment + f2Ui + y1diYi K Figure 1 presents age-race-specifictrends in + X OkXik, (3) proportionsof young men employed in March k=1 of 1964 through 1981. The individual points where Xik denotes the value on the kth inde- (letters) are the observed year-race-age- pendent variablefor the ith individual.In addi- specific employment proportions translated tion to race, trend, and cycle, the independent into the probit(z-score) scale. The lines are the variables are age, grades of schooling, years estimated employment proportions under out of school, veteran status, and veteran models that restrict year-to-year changes to status-age interaction.7 follow the overall business cycle and a linear Unmeasured Determinants of Em- time trend, and that also restrict race dif- ployability. We also seek to investigate ferences to follow a lineartrend. A comparison whether decreases in employment levels of of rows A and B of Table 1, which presentsthe blacks relative to whites are partlythe result of likelihood statistics for these models, suggests disproportionatedeclines, for the black out- that the data reject the trend and cycle restric- of-school civilian population, in men with the tions inasmuchas the chi-squarestatistics im- best employmentprospects. To investigatethis plied by the differencesof likelihoodsare large directly would require measures of ability, for the 21 degrees of freedom saved by the motivation, and other attributes that are not simpler model. Throughout our discussion, observed in CPS or other time-series data. however, we regard differences in log likeli- Thus we model the probabilitiesof school en- hoods of nested models only as descriptive rollment/enlistmentand employment as joint measures of relative fit, since the number of outcomes affected by common unmeasured individualobservations is so large and the ran- variables. The employmentmodel includes the variablesdiscussed above. The enrollment/enlistment model includes the effects of year (17 W dummy variables),race, race-yearinteraction, .88 - 8 8 age, race-age interaction,grades of schooling, .82 1 C and age-gradesof schooling interactionon the probabilityof being enrolled in school or en- ~76 24-29\ _ listing in the armedforces. Using the bivariate .73 probit model, we estimate these equations and .69 _ B .66 U

the z-transformed - _ _ the correlation between .62 probabilities(probits) of employment and enrollment/enlistment,controlling for the measured independentvariables. A positive correlation indicatesthat similarunmeasured factors 34 B ~~BB B

B affect remainingin school or enlisting and, if 31~~~ _ one is not in school or the military,being employed. Rising relative enrollmentand enlist- 24~~~~~~~ B B\

.34 ~ ~ ~ ~B ~ ~ \ - 4 We considered other factors, such as two-way 21 - white (W) B interactionsbetween race, age, grades of schooling, 18 _ 8-* Black (B) school, and time. Interactionscombin- years out of 16 ing race or time with other variables were nearly always statistically insignificant. Some other in- 6818 4 65 66 67 6s 7078 71 72 71 74 75 76 77 78 0 81 teractionswere significantbut did not affect the in- Year terpretationof the lower-order effects considered Figure 1. ProportionsEmployed by Age, Race, and here. We also examined other classificationsof the Year: Observed and Predictedby Linear independentvariables, but these did not affect our Trend and Business Cycle (Model B. conclusions. Table I) RACIAL INEQUALITY AND JOBLESSNESS 45

Table 1. Likelihood Statistics for Selected Models of Trends in Employmentof Young Men by Age and Population Age 16-19 20-24 24-29 Degrees Degrees Degrees of of of Model - 2La Freedom -2L Freedom -2L Freedom Total' A Year, Race, Year x Race 127409 1882 104344 2937 93290 2921 B Trend, Cycle, Race, Trend x Race 127499 1913 104694 2968 93342 2952 C Model B, Enrollment 119698 1912 77487 2967 68643 2951 D Model C, Trend x Enrollment 119454 1911 76637 2966 68123 2950 E Model D, Race x Enrollment 119436 1910 76623 2965 68118 2949 F Model E, Trend x Race x Enrollment 119432 1909 76622 2964 68118 2948 G Model F, Cycle x Enrollment 119414 1908 76548 2963 68097 2947 H Model G, Cycle x Race 119414 1907 76547 2962 68097 2946 I Model H, Cycle x Race x Enrollment 119414 1906 76546 2961 68096 2945 Not Enrolled Civilians a Year, Race, Year x Race 39833 1323 48521 2158 56168 1935 b Trend, Cycle, Race, Trend x Race 39902 1354 48639 2189 56262 1966 c Model b, Cycle x Race 39902 1353 48639 2188 56261 1965 a -2L denotes -2 times the natural logarithm of likelihood. Differences in -2L between nested models are distributed x2 (under the null hypothesis) with degrees of freedom equalling the difference of degrees of freedom for the two models. b Armed Forces are coded as enrolled and not employed. dom sampling assumption is violated for the the 1960s,but are now approximatelyequal for CPS data. Figure 1 shows that the trend-cycle the two groups despite the stagnantenrollment model closely traces observed employment growth during this period for all young men. proportions. Employment is stable or rising for young white men and falling for their black counter- Enrollment and Enlistment Effects parts. The downward employment trend for on Employment Trends young blacks is strongest for men aged 20 and The trend in race differences in employment above, whereas the white increases are re- shown in Figure 1 may not uniformlyapply to stricted to men aged 24 and below. As a result, students and nonstudents, nor may black and the largest increase in the race difference in white young men combine schooling and work employmentoccurs among 20 to 23 year olds. at similarrates (Coleman, 1974;Mare and Win- In the early 1960s, race differences in employ- ship, 1979), nor may employmentrates follow ment for young men were small-much smaller the business cycle in similarfashion for blacks than differences for mature workers. By the and whites. To examine relationships among late 1970s, however, these differences had enrollment,enlistment, and employmentmore dramaticallywidened for all young men. systematically,we consider probitmodels that Figures 2a and 2b report race-age-specific representdifferences in trend and cyclical em- proportionsof young men enlisted in the armed ployment patterns between races and enroll- forces and enrolled in school respectively for ment statuses. Rows A through I of Table 1 1964-81. For each age grouplarger proportions present likelihood statistics for these models. of white men than black men were in the armed Under simple random sampling, differences forces in 1964, whereas the opposite is true in between these statistics for nested models are 1981. The trend is particularly striking for distributedchi-square under the null hypothe- those 20 to 23 years old, for whom the race sis of no difference. difference in enlistment has ballooned to ap- Model C augments the trend-cycle model proximately 10 percentage points. For the portrayed in Figure 1 with a dummy variable youngerand older age groups, black enlistment takingthe value one if an individualis enrolled proportionsin 1981 are at or above the peak or enlisted and zero otherwise. The chi-square levels of the Vietnam mobilization, whereas statistic for the single degree of freedom used white proportions have declined. Similar for this effect is large, indicatinga much larger trends have occurred for school enrollment employmentprobability for out-of-schoolmen. rates. In each age group enrollment propor- Model D includes a separateemployment trend tions were lower for blacks than for whites in for studentsand nonstudents,again resulting in 46 AMERICAN SOCIOLOGICAL REVIEW

White 16-19 * _ . *20-23 30_30 - 24-29

Black

25 - 16-19

* .20-23

---24-29

220

115

.- I I I II I I I I l I I I I I I .. 1 964 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 Year Figure 2a. Percent in Militaryby Age, Race, and Year (Total Population) a substantialchange in the likelihood statistic Table 1 indicates, however, the chi-square for all age groups. Model E allows for the ef- statistics associated with these effects are neg- fects represented in Model D plus a varying ligible for all age groups of young men, a find- enrollment effect for blacks and whites. The ing consistent with other analyses of youth un- chi-square statistics for this effect are 18, 14, employment (Wachterand Kim, 1982). and 5 for the three age groups, suggestingrace We use Model G to decompose change in differences in the effect of student status on employmentbetween 1964and 1981into three employment.Model F representsseparate em- parts: (1) changingrates of enrollmentand en- ployment trends for the four race-enrollment listment; (2) changing proportions employed status groups. Relative to Model E, Model F among students; and (3) changingproportions representsa slightimprovement in fit for 16-19 employed among out-of-school civilians.5 year olds, but not for the older groups. Model G allows for the effects in Model F plus a varying effect of enrollment status over the I Componentsare adjustedto the averagebusiness business cycle. This latter effect may result cycle level for 1964-81 and are presented in the from differentialenrollment rates with overall probit scale. The race-specificcomponents are cal- employmentlevels or differentialimpact of re- culated accordingto the formula cessions on job prospects of students and nonstudents. The chi-square statistics for the in- Z81-Z64 = (P81P64) (z8l 64) clusion of this effect are substantialfor all age groups. Models H and I allow for varying effects of + ([1-P81]-[1-P64])( L)} race and of race within enrollment statuses over the business cycle. These effects might result if the employmentprobabilities of blacks + ( [Z1Z4] P81+P64 ) and whites respond differentially to market booms and busts, as might occur if the two groupswere nonrandomlyassigned to a "labor + ([Z n1Z]n (1-P81)+(l-P6) J) queue" (e.g., Thurow, 1975;Hodge, 1973).As RACIAL INEQUALITY AND JOBLESSNESS 47

55 White 50 16-19 20-23 45 24-29

40 Black 16-19

35 -* 20-23 0 ;, - - - 24-29 30

00 20 '._ '

15 N

5 ___------.--- '- -_ _- -_ - __

I I I I l I l III I I I I I I l 1964 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 8n 81 Year

Figure 2b. Percent Enrolled by Age, Race, and Year (Total Population)

(Coefficients for this model are availablefrom substantial proportions of the widening race the authors on request.) Table 2 displays the difference-approximately 20, 45, and 60 per- results. cent of the change for the age groups 16 to 19, The relative importance of changes for 20 to 23, 24 to 29 respectively. For men under whites and for blacks differs considerably 24, however, these components are made up among the age groups. For the oldest group, more of decliningwhite enrollmentand enlist- where the race difference has widened least, ment rates than rising black rates, a pattern the only change is deteriorationin black em- consistent with the trends in Figure2. Changes ployment. For 20 to 23 year olds, where the in black employmentnot accountedfor by en- largest widening has occurred, white employ- rollmenttrends are almost entirelythe result of ment increases and black decreases contribute declining employment among out-of-school equally to the growing spread. For teenagers, men. Changes for whites under 24 mainly re- increases in white employment are somewhat sult from employment increases among stu- larger than black decreases. dents. For whites 24 and over, employment Changes in rates of enrollmentaccount for changes are minimal. To summarize, from 1964 through 1981 blacks at any age have become-more likely (Kitagawa, 1955), where P64 and P81 are the year- than whites to be enrolled in school or enlisted specific proportions either enrolled in school or en- in the armed forces. Because of the strong listed in the armed forces, Z64and z81 are the year- negative association between enrollment and specific probits of the proportions employed ad- employment, these trends have contributedto justed to the average prime age male unemployment the growingwhite employmentadvantage over rate for 1964-1981, and the superscripts e and n denote the enrolled/enlisted and not enrolled/not en- blacks among young men. A substantialpro- listed populations respectively. The components for portion of change in the employment dif- race differences are the differences of their respec- ference, however, remainsunexplained by the tive race-specific components. direct substitutionof schooling or militaryser- 48 AMERICAN SOCIOLOGICAL REVIEW

Table 2. Components of Changea in Employment, 1964-1981, by Age and Raceb White-Black White Black Difference Percent Percent Percent of of of Components Change Change Change Change Change Change Age 16-19 Change in Enrollment Rate 0.0778 23.8 -0.0203 10.5 0.0981 18.9 Change in Employment Rate, Enrolled 0.2387 73.0 -0.0031 1.6 0.2418 46.5 Change in Employment Rate, Not Enrolled 0.0104 3.2 -0.1694 87.9 0.1798 34.6 Totald 0.3269 100.0 -0.1928 100.0 0.5197 100.0 Age 20-23 Change in Enrollment Rate 0.2308 59.6 -0.0984 27.9 0.3292 44.5 Change in Employment Rate, Enrolled 0.2072 53.5 0.0582 -16.5 0.1490 20.1 Change in Employment Rate, Not Enrolled -0.0506 -13.1 -0.3127 88.6 0.2621 35.4 Totald 0.3874 100.0 -0.3529 100.0 0.7403 100.0 Age 24-29 Change in Enrollment Rate 0.0769 -157.6 -0.1328 32.5 0.2097 58.3 Change in Employment Rate, Enrolled 0.0655 -134.2 0.0406 -9.9 0.0249 6.9 Change in Employment Rate, Not Enrolled -0.1912 391.8 -0.3163 77.4 0.1251 34.8 Totald -0.0488 100.0 -0.4085 100.0 0.3597 100.0 a Decomposition examines change between two three-year intervals, 1964-66 and 1979-81, in probit transformed probabilities of employment. b Total change is estimated from a model that adjusts employment to a common business cycle (aggregate unemployment) level for all years. c Members of the armed forces are defined as "enrolled" and not employed. d Components may not sum to 100.0 percent because of rounding. vice for work, especially amongthose 16 to 19 years old, for men who have been out of school and 20 to 23 years old. Insofaras this change is a longer period. Black out-of-school civilians the result of rising employment among white are less likely to be employed than their white students, it is beyond the scope of our analysis counterparts,controlling for other independent (see the concluding section). Change that re- variables. The coefficients for the race dif- sults from employment trends among out-of- ference are small (denoting approximatedif- school civilian men, however, may result from ferences of two to four percentage points in the additionalmechanisms linking enrollment/ employment),because they measure race dif- enlistment and employment trends discussed ferences at the point where the linear trend is above. equal to zero, that is, in 1963. The negative coefficients for the interaction between race indicate that the race difference in Determinants of Employment for and trend employment grew by approximately1.3, 0.6, Out-of-School Civilians and 0.2 percentagepoints per year for the age Model 1 of Table 3 includes the effects of mea- groups 16 to 19, 20 to 23, and 24 to 29 respec- sured independentvariables on the probability tively, once measuredfactors are controlled. of employmentestimated by a single-equation Model 2 incorporatesthe effects includedin probit model applied to each of the three age Model I as well as the effects of common un- groups. The results show that employment measureddeterminants of employmentand en- rises with age, albeit at a decreasingrate. Vet- rollment or enlistment in the armed forces.6 erans are less likely to be employed than non- For each age group, there is a substantialposi- veterans, but veteran employmentrises more tive correlation between the unmeasured rapidly with age, indicating that the veteran causes of employment and enrollment or en- disadvantagegradually disappears (Mare et al., forthcoming). Employment probabilities are 6 To save space, the selection equations for these typically higher for young men -with more models are not reported here. Estimates are avail- grades of schooling and, among those 16 to 23 able from the authors on request. RACIAL INEQUALITY AND JOBLESSNESS 49

Table 3. Effectsa of Independent Variables on Probability of Employment for Out-of-School Young Men, 1964-1981, by Age and Method of Estimation Single-Equation Two-Equation Model 1 Model 2 Coefficients Coefficients Independent Variables 3pIOX () 1 apIaX (*) S.E. (p8) S.E. (/3)

Age 16-19 Age 0.101 0.266 1.7 0.022 0.057 0.3 Veteran (vs. Nonveteran) -0.192 -0.504 -0.2 -0.249 -0.655 -0.2 Age x Veteran 0.004 0.012 0.1 0.007 0.017 0.1 12 Grades (vs. <12 Grades) 0.216 0.569 26.9 0.214 0.562 10.3 >12 Grades (vs. <12 Grades) 0.147 0.385 9.2 0.286 0.752 14.6 Out 2-3 Years (vs. Out 1 Year) 0.084 0.222 12.5 0.094 0.246 9.6 Out 4+ Years (vs. Out 1 Year) 0.095 0.250 9.4 0.105 0.275 7.7 Black (vs. Nonblack) -0.037 -0.098 -1.8 -0.117 -0.307 -4.7 Trendb -0.001 -0.002 -0.1 -0.037 -0.095 -3.3 Prime Age Unemployment Rate -0.045 -0.119 -10.9 -0.050 -0.131 -8.6 Black x Trendb -0.131 -0.344 -7.4 -0.084 -0.221 -2.9 pC 0.0 0.650 5.9 -2 Log Likelihood 144349d 144336 Degrees of Freedom 1905 1904

Age 20-23 Age 0.033 0.127 7.4 0.028 0.110 5.8 Veteran (vs. Nonveteran) -0.283 -1.102 -2.7 -0.326 -1.267 -3.0 Age x Veteran 0.010 0.037 2.0 0.012 0.044 2.3 12 Grades (vs. <12 Grades) 0.097 0.379 22.5 0.108 0.419 22.5 >12 Grades (vs. <12 Grades) 0.119 0.462 17.5 0.161 0.622 12.5 Out 2-3 Years (vs. Out 1 Year) 0.088 0.341 12.8 0.082 0.321 10.6 Out 4+ Years (vs. Out 1 Year) 0.098 0.383 12.0 0.095 0.369 10.7 Black (vs. Nonblack) -0.022 -0.087 -1.7 -0.032 -0.125 -2.3 Trendb -0.028 -0.107 -5.7 -0.029 -0.111 -5.7 Prime Age Unemployment Rate -0.034 -0.133 -14.4 -0.035 -0.136 -14.3 Black x Trendb -0.060 -0.235 -5.7 -0.050 -0.196 -4.3 pC 0.0 0.258 3.0 -2 Log Likelihood 119848d 119836 Degrees of Freedom 2960 2959

Age 24-29 Age 0.013 0.076 12.2 0.013 0.082 10.1 Veteran (vs. Nonveteran) -0.125 -0.765 -3.7 -0.149 -0.918 -3.7 Age x Veteran 0.004 0.025 3.2 0.004 0.030 3.3 12 Grades (vs. <12 Grades) 0.071 0.436 28.8 0.089 0.543 30.0 >12 Grades (vs. <12 Grades) 0.100 0.613 38.0 0.130 0.801 35.0 Out 2-3 Years (vs. Out 1 Year) 0.009 0.054 0.5 0.010 0.060 0.5 Out 4+ Years (vs. Out 1 Year) 0.008 0.048 0.5 0.004 0.031 0.3 Black (vs. Nonblack) -0.040 -0.242 -5.2 -0.052 -0.316 5.6 Trendb -0.037 -0.228 -13.3 -0.043 -0.263 -12.9 Prime Age Unemployment Rate -0.019 -0.118 -14.1 -0.023 -0.137 -13.6 Black x Trendb -0.021 -0.128 -3.4 -0.017 -0.100 -2.0 pC 0.0 0.550 -2 Log Likelihood 109475d 109453 Degrees of Freedom 2944 2943

a Effects are evaluated at the sample means of the dependent variable (for 16-19, p = .6213; for 20-23, p = .8258; for 24-29, p = .9089). b Linear trend multiplied by 10. c Disturbance correlation for equations predicting employment and enrollment or enlistment (for total population). d Log likelihood and degrees of freedom adjusted to be comparable to Model 2 on assumption of an identically specified enrollment equation and zero disturbance correlation. 50 AMERICAN SOCIOLOGICAL REVIEW listment. The differencein likelihoodstatistics ment prospects are relatively good. Employ- between Models 1 and 2 (under random sam- ment rates generally decline for whites, but pling a chi-squarestatistic with one degree of amongmen who are equivalenton unmeasured freedom)is substantial. This stronglysuggests factors they decline even more. Finally, the that students and enlistees would have above estimated race difference in the employment average prospects for employmentwere they trend is smallerin the two-equationmodel, in- to become out-of-school civilians. It also im- dicatingthat relative increases in employment plies that adverse employment trends may and enlistment for blacks have adversely af- partly result from enrollment and enlistment fected the relative employmentchances of the trends that exclude from the nonstudentcivil- remainingout-of-school black civilians. Once ian populationyoung men with good employ- unmeasureddeterminants of employabilityare ment chances. taken into account, the widening race dif- Several differences between the single- and ference is partly explained. two-equationresults reinforce this interpreta- tion of the correlation between unmeasured causes of employment and enrollmentor en- Decomposition of Change in Employment listment. First, the two-equation estimate of for Out-of-School Civilians the age effect for those 16 to 19 years old is Figures 3a-3c summarizethe trends in several much smaller than the single-equation esti- of the measured independent variables in- mate, indicatingthat age variationin employ- cluded in the equations reported in Table 3. ment is partly explained by the tendency for Figure 3a shows that the proportionsof young young men with the best employment pros- men who have completed at least high school pects to leave school at a later age than those grew markedlybetween 1964and 1981and that with poorer chances (Mare et al., forthcom- black proportions grew more rapidly than ing). Second, for all age groups, the two- those of whites, especially for the two older equation estimates of the race coefficient are groups. Figure 3b shows that young blacks notably larger than the single-equationesti- have an increasingly unfavorabledistribution mates. As noted above, this coefficient mea- of lengthof time since leaving school. For both sures the race differencein 1963.Differences in blacks and whites, rising ages of school de- employment proportionsbetween blacks and parture have reduced the proportions of the whites in 1963 that are adjustedfor measured age groups 16 to 19 and 20 to 23 who have been characteristicsalone understatethe difference out of school for more than a year. The trend, in employment between young blacks and however, has been much strongerfor blacks. whites who are equivalent on both measured Finally, Figure 3c shows that, whereas in the and unmeasured variables. In 1963, when 1960s whites were much more likely than white enrollment and enlistment rates ex- blacks to be veterans, the opposite is now the ceeded those of blacks, white out-of-school case. civilians were relatively less employable than their black counterpartsand have only moder- 0m5 ately higher employment. Once unmeasured 95_ characteristics are controlled, however, a 90_ much larger white advantageis revealed. 85 _ for the coefficient the linear , Third, trend, 75_ < which denotes the annual change in employ- 70 b Ae , a Y ment for whites, is larger in the two-equation models, especially for ages 16 to 19. This sug- va60_/ z gests that reduced enrollment and enlistment ff55 _i t_ rates for whites have raised the proportionof white out-of-school civilians whose employ- e45_ /,' ' / __ 80~~~~~~~~~~~~O2 7 An anomaly occurs in the estimated normal 35 3n i statistic for rho for the 24 to 29 age group. The statistic should be approximatelythe squareroot of of-Shoo Population 20 the likelihood ratio chi-squarefor the improvement ~~~~~~~~WhiteMlac 15 -- 16-19 16-19 in fit betweenModels 1 and 2, thatis, \/_ = 4.7. The * 20-23 -- 20-23 statistic, however, is only 1.5. This suggests an ill- 10 - - 24-29 - - 24-29- behaved likelihoodsurface for this age group,a possible consequence of the small numbersof students 1964 65 66 67 68 69 711 71 72 73 74 75 76 77 78 79 80 81 or membersof the armedforces. Explorationof the Year. likelihood surface failed to yield alternative esti- Figure3a. PercentCompleting 12 or MoreGrades of mates. Our specific estimates should be interpreted Schoolingby Age, Race, and Year (Out- cautiously. of-School Population) RACIAL INEQUALITY AND JOBLESSNESS 51

95 1\/.- w10

85 \61" \ = 02

X 0 20- \ 232/

10 --242 --2-95_

5 40 IC|L 12:_0__ _ ...

35 6 6 6 6 8 6 n 7 2 3 7 5 7 7 7 9 8 R

30 15~~~~~~~Ya

Fi2e3.5re tO to c o l2orM r er 966 66 86n7Z 1 7 2 t l 67 87 O8 2 y AgRcadYa1(u-fSho 0ei Poplaton Figure 3c ecetVtea y gRaendYa 20-23opultion ~ ~ ~ ~ ~ ~ OtofSho

Table 4 decomposes changes in race-specific (probit-transformed) employment proportions cent of the noncyclical black employment de- between 1964 and 1981 into parts associated cline for the 16 to 19, 20 to 23, and 24 to 29 age with trends in both measured and unmeasured groups respectively. Nonnegligible proportions variables. The decomposition is based on the of the decline for the two younger age groups coefficients of Model 2 in Table 3 and on the are also due to the reduced average time young means of the independent variables in the blacks have been out of school. The residual model averaged over the intervals 1964-66 and component of change is very large, but this 1979-81.8 For whites, the decomposition results in part from the substantial offsetting shows that the drop in employment is almost change in grades of schooling which has been a entirely due to the business cycle for the age force raising black employment. groups 16 to 19 and 20 to 23, whereas a sub- Although the race-specific decompositions stantial part of change is noncyclical and unex- provide meager support for the mechanisms of plained for those 24 to 29 years old. For blacks, change discussed in this paper, the decomposi- in contrast, most of the decline in employment tion of change in the race difference shows is not the result of cyclical change. The in- that the combined influences of change related creasingly unfavorable distribution for blacks to enrollment and enlistment on blacks and on unmeasured determinants of employment whites account for substantial portions of the accounts for approximately 10, 7, and 17 per- broadening race employment difference. As the final two columns of Table 4 show, the race difference on unmeasured de- 8 The components in Table 5 for measuredvari- changing ables are the differences between the means for terminants of employment accounts for 1979-81 and 1964-66 weighted by their respective roughly 40, 20, and 50 percent of change in the parametersin Model2 of Table4. The componentfor race difference in employment for the three age "Enrollmentand EnlistmentSelection" is groups, a result of the declining enrollment and enlistment rates of whites combined with stable or rising rates for blacks. Approximately 7 P O(L?C, O(Cs,) and 3 percent of the changing difference in

V1_p2 (K -6(c181 Kl- (c)] employment for those 16 to 19 and 20 to 23 years old respectively is accounted for by where p is the correlationbetween unmeasuredde- changing race differences in average time since terminantsof employmentand enrollmentor enlist- their twen- ment, c denotes predicted values in the selection leaving school. For young men in ties, approximately 10 percent of change in the equation, 0 and P are the standardnormal density and distributionfunctions respectively, and the sub- race difference is attributable to the growing scripts64 and 81 denote that the ratios are evaluated representation of blacks in the veteran popula- at their means for 1964-66 and 1979-81. tion. 52 AMERICAN SOCIOLOGICAL REVIEW

Table 4. Components of Changea in Employment (Probit), 1964-1981, Out-of-School Noninstitutional Men, by Age White Black Difference Percent Percent of of Percent Noncyclical Noncyclical of Component Change Change Change Change Change Change Age 16-19 Grades of Schooling 0.023 57.5 0.041 -10.2 -0.018 -4.1 Years Out of School -0.009 -22.5 -0.039 9.7 0.030 6.8 Age 0.002 5.0 0.000 -0.0 0.002 0.5 Veteran Status 0.003 7.5 -0.003 0.7 0.006 1.4 Age-Veteran Status Interaction 0.000 0.0 -0.000 0.0 0.000 0.0 Business Cycle -0.110 - -0.110 Enrollment and Enlistment Selection 0.129 322.5 -0.040 10.0 0.169 38.3 Residual -0.108 -270.0 -0.360 89.8 0.252 57.1 Totalb -0.070 100.0 -0.511 100.0 0.441 100.0 Age 20-23 Grades of Schooling 0.047 -123.7 0.092 -22.9 -0.045 -12.4 Years Out of School -0.002 5.3 -0.014 3.5 0.012 3.3 Age -0.003 9.8 -0.002 0.5 -0.001 -0.3 Veteran Status 0.034 -89.5 -0.004 1.0 0.038 10.5 Age-Veteran Status Interaction -0.000 0.0 -0.000 -0.0 0.000 0.0 Business Cycle -0.145 - -0.145 Enrollment and Enlistment Selection 0.047 -123.7 -0.029 7.2 0.076 20.9 Residual -0.161 423.7 -0.444 110.7 0.283 78.0 Totalb -0.183 100.0 -0.546 100.0 0.363 100.0 Age 24-29 Grades of Schooling 0.110 -64.3 0.175 -52.6 -0.065 -40.1 Years Out of School 0.000 0.0 0.000 0.0 0.000 0.0 Age 0.007 -4.1 -0.001 0.3 0.008 4.9 Veteran Status 0.021 -12.3 0.003 -0.9 0.018 11.1 Age-Veteran Status Interaction -0.005 2.9 0.002 -0.6 -0.007 -4.3 Business Cycle -0.126 - -0126 Enrollment and Enlistment Selection 0.025 -14.6 -0.059 17.7 0.084 51.9 Residual -0.329 192.4 -0.453 136.0 0.124 76.5 Totalb -0.297 100.0 -0.459 100.0 0.162 100.0 a Decomposition is based on probit coefficients for Model 2 in Table 4 and three-year average data for 1964-66 and 1979-81. b Components may not sum to reported totals because of rounding.

Taken together, the increasingrelative num- been no relative change in educationalattain- bers of blacks in the vulnerablenewly out-of- ment. By this measure, we account for ap- school or out-of-the-militarypopulations and proximately43, 22, and 24 percentof change in the changingracial distribution on unmeasured the race difference for the three age groups. determinantsof employment account for ap- proximately46, 35 and 59 percent of the widening white-black employment difference Accounting for Changing Race Differences among out-of-schoolcivilian men for the three in Employment: Summary age groups. Of course, the residual change in The decompositionsof Tables 2 and 4 can be the race difference remainslarge, again partly combined to yield overall estimates of the the result of offsetting improvements in the contribution to widening race differences in relativeeducational status of blacks that would youth employmentof increasedsubstitution of otherwise have reduced the race employment schooling and military service for work by difference. A more conservative estimate of young blacks, increasing relative numbers of our success in explaining the changing race black veterans and recent school leavers, who difference, 100 minus the residualpercentage, are especially vulnerable to joblessness, and indicates the amount of the change that these changingrace-specific composition of the out- mechanismswould have explainedif there had of-school civilian population on unmeasured RACIAL INEQUALITY AND JOBLESSNESS 53 factors affecting success in school, the mili- war and peace, the draft, the age structureof tary, and the labormarket. Table 5 summarizes the population, and macroeconomic condi- the decompositions. The first two rows of the tions. That young men make enrollment and table indicate the contributionof changingen- enlistment decisions while cognizant of their rollmentrates and changingpatterns of move- job prospects, or, more generally, that school- ment out of school and the armed forces to ing, work, and militarydecisions and opportu- changingrace differences in employment.The nities are determined jointly, are not fully third row indicates the percentage of the taken into account. Althoughour two-equation change that is offset by relative levels of estimates suggest that the most employable schoolingcompleted by blacks. The fourthand men substituteschooling or militaryservice for fifth rows denote the unexplainedportions of work, they may nonetheless reflect the effects change for out-of-school civilians and for stu- of a complex mixtureof incentives, opportuni- dents respectively. If we ignorethe dampening ties, and costs that structureyoung men's deci- effect of changes in relative levels of educa- sions about work, schooling, the military,and tional attainmenton the race difference, their other activities. results imply that the three mechanisms ac- With regardto the explanatorylimitations of count for approximately35 percent of change our analyses, a large part of the change in the in the race differencefor the age group 16 to 19, race employmentdifference from 1964to 1981 57 percent for the age group 20 to 23, 80 per- is unexplainedby the mechanismsconsidered cent for the age group 24 to 29, and 56 percent here. A majorsource of change is growingem- for the three age groups combined. If we take ployment among white teenage students. This account of the changes in educationalattain- change may result from reductions in obliga- ment and focus instead on the complementof tions to school and family, from changed eco- the residualchanges in Table 5, then the three nomic requirements of middle-class student mechanismsstill account for approximately34 lifestyles, or from real or perceived changes in percentof changefor the age group 16 to 19, 52 the long-runeconomic benefit to work experi- percent for the age group 20 to 23, 67 percent ence while in school (relative to attending for the age group24 to 29, and 50 percentfor all school without working) (Meyer and Wise, three groups. In sum, a substantialportion of 1982). For out-of-school young men, some of the increasing race difference in youth em- the explanations for youth labor force trends ployment can be accounted for by processes enumerated in the introductionmay account related to rising black participationin school for part of the unexplaineddecline in the rela- and the armed forces. tive employmentposition of young blacks. In- adequate demand for young workers induced changingskill composition of the econ- CONCLUSION by the omy, spreading minimum wage legislation, Despite our success in accountingfor changing large youth cohort sizes, and unsalutaryaggre- employmentdifferences between young blacks gate economic conditions may disproportion- and whites, it is importantto recognize the ately hurt young blacks if employers prefer to limitationsof our models and their explanatory hire white youths over blacks (e.g., Hodge, power. Our assessment of enrollment effects 1973; Thurow, 1975). on employmentfor the total populationand of Finally, our analyses do not apply to the selection effects on the employmentof out-of- substantial widening of the black-white em- school civilians assumes that enrollment and ployment difference occurring between 1940 enlistment trends are driven by exogenous and 1960,which appearsto resultnot only from variables, such as family backgroundtrends, enrollmenttrends, but also from the decline in

Table 5. Percentage Decomposition of Change in Race Differences in Employment, 1964-1981, by Age: Summary Age Component 16-19 20-23 24-29 16-29 Change in Enrollment Rate 18.9 44.5 58.3 39.3 Change in Employment Rate for Not Enrolled Change in Transition from School and Military to Work 16.3 12.2 22.1 15.8 Change in Educational Attainment for Out-of-School Civilians - 1.4 -4.4 -14.0 -4.7 Unexplained Change for Out-of-School Civilians 19.8 27.6 26.6 23.9 Change in Employment Rate for Enrolled 46.5 20.1 6.9 25.7 Totala 100.0 100.0 100.0 100.0 a Components may not sum to reported totals because of rounding. 54 AMERICAN SOCIOLOGICAL REVIEW

low-skill, especially agricultural, jobs (Fisher, equalities within the black population are likely 1973; Mare and Winship, 1979; Cogan, 1982). to grow. Despite their limitations, however, our analyses provide much support for our REFERENCES arguments. The growing race difference in em- Butler, Richard and James J. Heckman ployment is, in part, a consequence of other- 1977 "The government's impact on the labor wise salutary changes in the lives of young market status of black Americans: a critical blacks, especially increased school enrollment review." Working Paper No. 183, National and educational attainment. In this sense, Bureau of Economic Research, Stanford, broadening race differences in youth employ- California. ment are consistent with diminished dif- Cogan, John F. ferences on other indicators of socioeconomic 1982 "The decline in black teenage employment: 1950-70." American Economic Review achievement. This argument does not imply 72:621-38. that widening race differences in employment Coleman, James S. are an acceptable development or a "neces- 1974 Youth: Transition to Adulthood. Chicago: sary" byproduct of otherwise favorable socio- University of Chicago Press. economic change. Rather, the apparent ab- Collins, Sharon M. sence of broad race differences in youth em- 1983 "The making of the black middle class." ployment prior to the mid-1960s was an illusion Social Problems 30:369-82. created by enormous race differences in other Congressional Budget Office aspects of socioeconomic standing and activity 1982 Improving Youth Employment Prospects: Issues and Options. Washington, D.C.: patterns that concealed race differences in em- U.S. Congress. ployment. Before the recent relative rise in Farley, Reynolds black schooling and military service, young 1983 Catching Up: Recent Changes in the Social blacks had a "head start" in the labor market and Economic Status of Blacks. Cam- because their white counterparts elected to bridge: Harvard University Press. seek full-time employment at much later ages. Feldstein, Martin and David T. Ellwood Now that blacks and whites are more equal in 1982 "Teenage unemployment: what is the their timing of movement out of school and the problem?" Pp. 17-33 in Richard B. military, the true race difference in employ- Freeman and David A. Wise (eds.), The Youth Labor Market Problem: Its Nature, ment is revealed. Causes, and Consequences. Chicago: Uni- We cannot resolve whether recent changes versity of Chicago Press. in racial stratification are permanent or tran- Fisher, Alan A. sitory. Our results suggest, however, that 1973 "The problem of teenage unemployment." worsening labor force statistics for black Unpublished doctoral dissertation, De- youths do not denote increasing racial in- partment of Economics, University of equality, but rather persistent racial in- California, Berkeley. equalities previously hidden by race dif- Freeman, Richard B. ferences in other aspects of young adulthood. 1973 "Changes in the labor market for black Americans, 1948-1972." Brookings Papers In addition, the array of socioeconomic dif- on Economic Activity 1:67-120. ferences between the races that remain to be 1976 The Black Elite. New York: McGraw-Hill. overcome is larger than would be revealed by Freeman, Richard B. and David A. Wise socioeconomic and labor force data for the 1982 The Youth Labor Market Problem: Its Na- early 1960s. Although the race difference in ture, Causes and Consequences. Chicago: employment is not a "new" inequality, it per- University of Chicago Press. sists while other differences have gradually Hausman, Jerry A. and David A. Wise eroded. Political efforts to reduce discrimina- 1978 "A conditional probit model for qualitative tion in the workplace and to increase black choice: discrete decisions recognizing interdependence and heterogeneous prefer- schooling have apparently not extended to race ences." Econometrica 46:403-26. differences in school quality, family environ- Heckman, James J. ment, and the availability of jobs commensu- 1979 "Sample selection bias as a specification rate with the skills that young persons possess, error." Econometrica 47:153-62. all of which may contribute to persisting race Hill, Robert B. differences in joblessness. If joblessness in 1978 The Illusion of Black Progress. Washing- youth hurts later socioeconomic achievement ton, D.C.: National Urban League. (Feldstein and Ellwood, 1982), then persis- Hodge, Robert W. tently high youth unemployment for blacks will 1973 "Toward a theory of racial differences in Forces 52:16-30. full in the adult labor market. employment." Social prevent equality Judge, George G., William E. Griffiths, R. Carter Moreover, if good opportunities are now avail- Hill and T. Soung-Chao Lee able to blacks with high levels of education 1980 The Theory and Practice of Econometrics. (Freeman, 1976; Wilson, 1978), then in- New York: Wiley. RACIAL INEQUALITY AND JOBLESSNESS 55

Kitagawa,Evelyn M. ployment." Journal of Political Economy 1955 "Componentsof a differencebetween two 87:S117-S131. rates." Journalof the AmericanStatistical Osterman,Paul Association 50:1168-94. 1980 Getting Started:The Youth Labor Market. Lazear, Edward Cambridge:MIT Press. 1977 "Schoolingas a wage depressant."Journal Smith, James P. and Finis Welch of Human Resources 12:164-76. 1978 "Race differencesin earnings:a survey and Mare, Robert D. and ChristopherWinship new evidence." Report R2295-NSF, Rand 1979 "Changes in race differentials in youth Corporation,Santa Monica, California. labor force status." Pp. 1-82 in Fifth An- Thurow, Lester C. nual Report to the Presidentand the Con- 1975 GeneratingInequality: Mechanisms of Dis- gress of the National Commissionfor Em- tributionin the U.S. Economy. New York: ployment Policy, Expanding Employment Basic Books. Opportunities for Disadvantaged Youth: U.S. Bureau of the Census Sponsored Research. Washington, D.C.: 1978 The CurrentPopulation Survey: Design and National Commission for Employment Methodology. Washington, D.C.: U.S. Policy. Bureau of the Census. Mare, Robert D., ChristopherWinship and Warren U.S. Departmentof Labor N. Kubitschek 1982 Labor Force Statistics Derived from the forth- "The transitionfrom youth to adult:under- Current Population Survey: A Databook, com- standingthe age pattern of employment." Volume 1. Washington,D.C.: U.S. Gov- ing AmericanJournal of Sociology. ernment PrintingOffice. Kim Meyer, Robert H. and David A. Wise Wachter, Michael and Choongsoo 1982 1982 "Time series changes in youth jobless- "High school preparationand early labor B. force ness." Pp. 155-85 in Richard Freeman experience." Pp. 277-347 in Richard and David A. Wise (eds.), The YouthLabor B. Freemanand David A. Wise (eds.), The Market Problem: Its Nature, Causes, and Youth Labor MarketProblem: Its Nature, of and Consequences. Chicago: University Causes, Consequences. Chicago:Uni- Chicago Press. versity of Chicago Press. Welch, Finis National Commission on Employmentand Unem- 1973 "Black-white differences in returns to ployment Statistics schooling." American Economic Review 1979 Counting the Labor Force. Washington, 63:893-907. D.C.: U.S. GovernmentPrinting Office. Wilson, WilliamJ. Nickell, Stephen 1978 The Declining Significance of Race. 1979 "Educationand lifetime patternsof unem- Chicago: Universityof Chicago Press.

MANUSCRIPTS FOR THE ASA ROSE SOCIOLOGYSERIES

Manuscripts (100 to 300 typed pages) are solicited for publica- tion in the ASA Arnold and Caroline Rose Monograph Series. The Series welcomes a variety of types of sociological work- qualitative or quantitative empirical studies, and theoretical or methodological treatises. An author should submit three copies of a manuscript for consideration to the Series Editor, Professor Ernest Q. Campbell, Department of Sociology, Vanderbilt Uni- versity, Nashville, TN 37235.