Youth Employment and Training Programs: The YEDPA Years (1985)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Youth Joblessness and Race: Evidence from the 1980 Census George Cave In 1983 the Census Bureau released microdata based on the "long- form" questionnaire completed by about one-fifth of the respondents in the 1980 Census. "Public-Use ~ ~ ~ ~ ~ ~~~ Microdata Sample C" identities the retype of area"--central city, urban fringe, rural, and so on--for a full 1 percent of the U.S. population (Bureau of the Census, 1983~. The large size of this data set enables researchers to study the impact of area type and many other factors on a multitude of individual variables measured by the questionnaire. In addition to the 1 percent population sample, a 0.1 percent subsample provides data on 226,947 individuals surveyed in the 1980 Census. This paper compares the data on unemployment and other labor force behavior reported for black youths with that reported for white youths. The key question addressed is, Do black youths face special problems in the labor market due to their race? A related question is whether correcting black and white youth labor force statistics for location, education, family income, and other factors tends to eliminate the racial differences. This paper, like most others in the empirical literature on youth unemployment, uses simple single-equation methods to correct for these factors. However, the results must be interpreted very carefully for several reasons. First, most coefficients estimated on data for individuals are subject to "ecological correlation bias" if the labor market characteristics of the respondents' local areas are missing from the data set. Second, the most common single-equation or system methods may not estimate structural coefficients for individuals' and employers' behavior. Third, single-equation methods introduce simultaneity bias if, for example, the probability of unemployment influences the probability of labor force participation. Finally, even system methods may ignore some simultaneity and overcorrect for factors other than race. To some extent, residential location within the local labor market, quantity and quality of education, family income, and so on are, like unemployment, partly the consequences of race in the labor market. To ignore the effect of race on these determinants of labor George Cave is on the staff of the Manpower Demonstration Research Corporation. 367

368 force status is to ignore the indirect labor market effects of race on unemployment. This paper is organized as follows. First, the data sets and the statistical methods used most extensively in this study are described. Next, a brief overview is given of the seasonal, cyclical, frictional, and structural components that complicate empirical analysis of the youth unemployment problem. This section also includes a survey of several earlier empirical analyses. The empirical analysis of the labor force status of out-of-school black and white teenagers included in the Census microdata follows. Then the very different behavior of those teenagers who were enrolled in school at any time during the two months before Census day is explored. The paper ends with a summary of the major findings. THE DATA The 0.1 percent subsample of "Public-Use Microdata Sample C" has 226,947 self-weighting observations on individuals. Of these, 8,653 are young men aged 16-19. Because only 1,190 of these young men are black, stratifying the sample by region, education, and other factors produces some data cells with no nonwhites. The problem becomes even more severe when students are excluded from this group; there remain 2,061 white males, but only 372 black males. Fortunately, it is easy to increase the number of nonwhites by a factor of 10 by using the full 1 percent sample of nonwhites. However, calculations based on stratified samples containing nonwhites from the full C sample but only those whites in the 0.1 percent subsample require special techniques. Heteroscedasticity could arise from the 10-fold greater chance a nonwhite respondent had to get into such stratified samples. Still, the huge Census microdata samples enable appropriately cautious researchers to home in on interesting subgroups in ways that smaller samples do not permit. The main dependent variables used here reflect labor force status during the week of the Census survey.) Unfortunately, questions that would have identified "discouraged" workers during the survey week were not asked. However, analyzing nonstudents separately picks up some part of the often-neglected behavior of those who are not in school yet are neither employed nor unemployed. . . This measure of labor force behavior is the most common. Alter- natives are available: the number of weeks spent in unemployment and in employment in 1979 are recorded for everyone 16 years and older in "Public Use Microdata Sample C." Survey-week labor force behavior is related systematically to weeks and spells of unemployment over the course of a year; see Betsey (1978) and Hanoch (1976~. Using survey- week behavior does not distinguish between the short-term and long-term unemployed.

369 Other variables available for all respondents include census region, type of area, householder status, age, race, marital status, disabil- ities, years of school completed, whether the respondent has ever worked, and income status. For those with at least some employment in 1979, earnings, usual weekly hours, and industry in which employed are also available. Unfortunately, crucial variables that are not available include actual hourly wages, the number of spells of unemployment in 1979, the number of job offers refused during job search, and eligibil- ity for unemployment compensation. METHODOLOGY Two basic methods are used in this analysis to compare black and white labor force behavior. Both attempt to explain dummy variables for employment, unemployment, and nonparticipation. When black and white samples are combined, and when race is one of the independent variables, the coefficient of race shows the increase in the probability of the behavior, conditional on the other independent variables, that can be attributed to being black. As in Freeman (1982), linear probability models (LPMs) are estimated because they explain quite simply some important relationships among the three dependent variables. However, because of well-known econometric difficulties with linear probabilities, logistic methods are used as well .2 Using both methods, linear and logistic, equations are estimated for two types of dependent variables, unconditional and conditional. Unemployment and employment equations are estimated both for the entire population and for labor force participants only. In these models, the coefficients for conditional employment and unemployment have the same magnitude, but different signs. The next section points out structural interpretations for the conditional equations; these reflect employer behavior and make the unconditional equations reduced forms confounding employer and individual structural coefficients. OVERVIEW OF THE YOUTH UNEMPLOYMENT PROBLEM The Many Faces of Unemployment Even though economists have produced a large literature on unemployment and take many separate approaches to the subject ,3 not 2See Nerlove and Press (1973:Ch. 2). The LPM predicts probabilities outside the unit interval, is subject to heteroscedasticity, and in general does not fit the statistical assumptions underlying least squares regression. 3 Two important strands of this literature are largely theoretical: macroeconomic general equilibrium and wage-search distributions. Hey

370 much fundamental progress has been made in explaining unemployment. An adequate economic explanation of unemployment would separate relevant factors reflecting the preferences of individuals for consumption of goods and uses of time from factors constraining individuals' choices about consumption and work. Moreover, such an explanation would systematize many of the stylized facts about unemployment. Further, such an explanation would yield empirically testable hypotheses for existing data about unemployment. One problem is that the same word, unemployment, is used to denote many very different phenomena. For a long time, empirical work on unemployment among individuals has tried to classify such unemployment as "seasonal," "cyclical," "frictional," or "structural," although it has been recognized that a given spell of unemployment for a given individual might be very difficult to categorize. Seasonal and Cyclical Unemployment One sort of seasonal unemployment is a characteristic of certain occupations, such as construction work. Workers committed to such occupations generally do not take other kinds of jobs during the off-season, perhaps because their wages reflect compensating differ- entials for the known risk of unemployment at certain times of the year. This sort of demand-side unemployment is unlikely to affect young people, who generally have not yet committed themselves to occupations. The failure of schools and colleges to stagger their vacation periods produces another kind of seasonal unemployment, 4 which can be attributed to the supply side of the youth labor market. A deluge of young people compete for relatively few jobs each summer. If the kinds of jobs young people take during their summer vacations paid lower wages, some have argued, 5 the problem would be smaller. Cyclical unemployment occurs less predictably and is tied to the business cycle and to cycles of product demand within industries. There has been a great deal of recent work on the nature of long-term contracts between firms and workers who are periodically laid off temporarily and then rehired. Feldstein {1976) estimated that 75 (1981) provides a survey. Some important articles in the empirical literature are cited in the next section. however, Clark and Summers (1982:209) cite gross flow evidence that demand for young workers, on the whole, adjusted remarkably well to increased supply during the summer over the years 1968-1976. They surmise from preliminary statistical work that federal Neighborhood Youth Corps and CETA programs may explain their surprising findings on this point. 5See Brown (1981) for a survey of many of the issues surrounding minimum wage differentials for youth.

371 percent of laid-off workers in manufacturing subsequently are rehired by the same employer. He cited 1975 evidence that 41 percent of unemployed men aged 25-64 who had been laid off had made no attempt to find jobs during the previous month. However, since young people tend not to have made solid commitments to particular firms or even to particular industries, temporary-layoff theories are less-convincing explanations for their unemployment than for the unemployment of older workers. The next subsection provides rough evidence that even cyclical unemployment tied to the business cycle is less important for younger people than for more-established workers. This lack of cyclical sensitivity is reassuring, since elsewhere in the paper I focus on cross-sectional data pertaining only to March 1980 and calendar 1979. Cyclical Sensitivity of Youth Unemployment Because I talk about a single cross-section of individuals in the remainder of the paper, I do not have much to say about cyclical influences on youth unemployment. How important are they? Persistent unemployment of at least 3.5 percent (measured as annual averages of monthly Current Population Survey estimates) has afflicted the U.S. civilian labor force since the mid-1950s. The aggregate rate fell from 5.5 percent in 1954 to 4.1 percent in 1956. Then it jumped to 6.8 percent with the 1958 recession. It fell again to 5.5 percent in 1959 and 1960 and then rose abruptly to 6.7 percent for 1961. With the exception of a slight faltering in 1963, it fell steadily from its 1961 level until it reached the post-1953 trough of 3.5 percent in 1969, a war year e Since then, as Table 1 shows, it rose in 1970 and 1971, fell in 1972 and 1973, rose through 1975 to a three-decade peak, and fell to the 1979 low preceding the most recent recession. Although the aggregate time series is sensitive enough to reveal broad trends, it masks a great deal of the labor market behavior that disaggregation reveals. In addition to aggregate unemployment rates, Table 1 shows unemployment rates for certain sex, age, and race groups, including groups of teenagers 16-19 years old. With the exception that teenage unemployment was below 12 percent for the three years preceding 1958, 1969 was the post-1953 trough for each of the disaggregated series as well as for the aggregate unemployment rate. But many of the disaggregated series have kept the same rank relative to each other ever since 1950. Moreover, some of the series are much more stable over time than others. Table 1 shows coefficients of variation (ratios of standard deviations to means) for the aggregate series and for 20 disaggregated groups for the period between 1972 and 1982. It also shows the rank of each coefficient of variation (c.v.) among the 21 coefficients reported. The aggregate series ranks 14th at 21.09 percent. All 7 groups of teenagers rank above the aggregate series; in fact, the teenage groups account for 4 of the top 4 groups and 7 of the top 11 groups. The coefficient for the top group, black women aged 16-19, is barely a third of the aggregate coefficient. At the other end of the

372 a, o C9 a, C~ o o Q . _ C: 4- a) o Q a, S: - o  o a' ~5 m Y C c' 0 ~) ~ a) 1 ~; E ~ ~ ~7 1 - O C:5- 1 =3 0 O ~ ~ 1 4~ a' ~ cs a) 1 - E~ ~= O (O- 1 33 0 I ~ (, a) ~D-CS a) 1 E ~ ~ 7: O ~- I 3 0 Y ~ 1 C) ~ ~o ~ a) 1~ O ~ ~ 0~ c~ 1· · -~ (~ 1ON 0:) ~N 0 1 O ~ 1 ~D~ ~ 1 - a) ~ ~ ~ 1 - 1 3 0 1 1 ~ <) ~ a) 1 a) ~ c: ~ 1 - 1 0 1 Y ~ON 1 (: a) ~ 1 ~ E - 04 1 =3 Y C: ~ 1 c) a) 0 ~ a) 1 ~ EC~ ~= - O tT5- 1 tn 3 0 a) ~ aN 1 a) ~ 1 ·- E I I _ 0< 1 33 a) ~ ~ 1 ~ 0 0 ~ a' 1 ·- Em == I "C O ~S- 1 33 0 1 Y ~ 1 C) ~ ~ 1 t~ a) 1 1  - ~ ~ 1 ~ ~ 1 Y ~ 1 C) ~ o~ a) 1 ~ a) (\J ~ ~ 1 _= ~- 1 0 1 0 0N 1 ~ ~ ~ 1 ·- O 1 1 =~0 1 3 a) ~ 1 ~ co~ a) 1 ._ ~ ~ ~ ~ 1 - 1 3 0 1 ~ I a) ~ 1 E O vC) 1 3- 1 0 0 T:) a) 1 E~l ~= I O tT5- 1 3 0 1 a~ 1 ~ ~ 1 a,) 1 1 '~ ~) 1 1 ~ , c o~ a) I a) ~ ~ =, ~ ~5- 1 O I 0N 1 1 1 1 <) 1 1 ~ 1 o~ a) 1 ~i c: 7D 1 ~5- 1 0 1 1 1 a' 1 ~ a) 4~ 1 <t ~ ~ 1 =~) 1 ~ ~N ~ ~ . . . . . - 0 ~ i- ~< ) h~ ~ v0 0 ~ \0 ~ . . .. . . ~0~0~030 L~ ~ - )N J L~ . . .. . . J~ J - C~ ~ 0~ O N . . .. · . J J - ~<  h~ ~ JO k0 J . . .. · . 0~0 - 3 ~C~ J J J O \0 ~(~ 1-') . . .. . . ~0 0N ~ N N O h~ J ~ C7. . . .· · . J ~ J l-~S) ~ C~ C~ ~h~ 0 (~1 . . .. . . J J h~1-\.0 ~D 1-0 ~ ~ N . . .. . . ~ ~ ~ ~ 0N ~) N ~)~ ~ r~ O O Jh~ J . . .. . - ~ ~ ~N ~ O N ~ h~~) ~ O . . .. . - JN ~0 ~:) O L~ C~ J . . .. . . ~0 ~ J -~ \0~ ~- . . .. . . ~ L~\OC~ 0 0 J ~O J O . . .. . . 1¢ J L{\ 0 ~ 1- C~)~ N . . .. . . O ON ~N~ O ~) 00 a~ N . . .· · . J ~ ro~D h~ LS~ ~J h^\ O0\ O 0 J ~ 0 . . .. . . . . . ~) J ~0 0x C~ ~ ~O ~O ~ L~ 0N h~ L~ 0N . . .. . . J ~ J ~ <) ~ <) ~ \0h~ 1- 0 . . .. . . h~ J L~o:) ~ ~  - - r~ - ~ ~a~ ~ ~ ~ ~ r - \0< 1 · · 1 h - \ I- 1 1 1 1 0N Y) 1 · · 1 0 1 1 1 N 0 J ~ J I ... . - 1 ~ ~ ~ ~ ~ I 1 1 1-~ 1 · · 1 h~ O I ~J 1 1 1 L~ J ~h~ C~ I ... . - 1 J J~ <0 1 1 1 N ~J(~ I . . . · - 1 L~ \0 ~ CSx I 1 1 0~0 N~ ... . - 1 0~5, N ~ I J ~ c~ =r J I 1 1 J-= 1 · · 1 t~ h~ I ~ ~ 1 1 1 \0 0 1 · · 1 (> 1 1 1 1 ~J O \0 0~ ~ 1 ... . - 1 LS~ h~ ~h~ I- 1 1 ~N ~- ... . - 1 ~S:) J ~ O 0 1  ~) ~ J J I l l ~) ~ J h~ 0 1 . . . · - 1 ~N - ) - I ~ ~ 1 1 =N (:~ - I ... . - 1 ~ c~ \0 1-~ 1 · ~N I 1 1 \0 ~ \0 0 1 ... . - 1 ~) IS ~h~ ~ I 1 1 JN O~ I . .. . . , C5$ ~ I '- ~N I l l 01~-= 0 - 1 . . . . · I \0 L]> ~ \0 0 1 l l 0~ - ~J . . .. . , h~ ~ 0O J I NN I 1 l ~J 0N ~ 0 1 .. · · · ~ J J ~ ~ 0 l l ~N I · · 1 0N ~ 1 N I 1 l 03 ~ h~ \0 1 . . . · - 1 h~ J ~<) 0 1 1 0 0 ~\0 ~- 1 ... . - 1 1- ~ Ct~ 1 1 0a~c ~m - ~= 00 0N G~ 0N . N h~ <)h~J 00N0 ...  J ~0 00a~ ... <)'~h~ h~cr ... 1- ~J 0 . ~0 N .. ~h~ N N .. N N . N a: o o J L~ r~ o 0 h~ N 0 ~0 ~0 o o 0 ~0 r~ J o c~ r~ c~ 0 N h~ ~0 0 0 h~ 0 . ~0 N J . 0N . N ~0 o N o  L~ J o 0 r~ N ~0 0 ~0 0\ N N ~0 1 1 ~ -- O > ._ a) 4~ ~ ~ 4~ o 0 o o 0 o J ~0 0 N o: LS) o 0 o o 0\ L`> ~0 o 0 ~0 N 0N o N > C) `' O Q J O 4 E 4 Q a' C] o 4- ~n  - LS) N . _ 4- - m .O o o o o :8 to :t en . - ~n Q J O a' m =` o g J w0 4 Q a) cn O 4_ .. ~ LL 00 C: ~ O ~

373 scale, groups containing older men account for all of the bottom six ranks. Older women tend to rank just below the groups of teenagers. Thus, from a disaggregation of time series using coefficients of variation to index instability, it can be seen that older men's unemployment is most susceptible to macroeconomic forces exerted on all demographic groups over time, while teenage unemployment is influenced least by such forces. Frictional Wage-Search Unemployment Frictional wage-search unemployment results from the dynamics of labor markets. People move in and out of the labor market as they age, as their skills change, as wage levels rise and fall, and as family needs and financial fortunes vary. 6 Jobs are created as firms are established, as plants are opened in new areas, and as older workers are forced to retire.7 But it may take a few weeks for would-be workers and firms with vacancies to search out and find each other. A job applicant might not take the first offer of wages and working conditions, and a firm might not be willing to meet the first appli- cant's wage bid. Such "search" unemployment might affect young people disproportionately more than adults, B because they are making gradual transitions from full-time schooling to full-time labor force partici- pation. Young people experiment with industries and occupations before making lifetime commitments. Sometimes they have parents to support extended periods away from both school and the labor force, and some- times they might misreport such nonparticipation as unemployment. According to "search theories" of frictional unemployment, 9 heterogeneity among individuals and among firms leads firms to search for workers and individuals to search for vacancies. The latter type of search has virtually been identified with unemployment by many labor economists, usually under the restrictive assumption that the utility function governing individual behavior is defined over discounted future wages, net of search costs but ignoring foregone leisure. 6When movement into the labor force exceeds steady state levels, structural unemployment may arise in addition to frictional unemployment. It may take some time for employers to adjust their hiring and wage policies to the increased supply of potential employees. 7Destruction of jobs by the same kinds of processes may lead to structural unemployment for established workers at the same time that it creates frictional unemployment for new entrants into the same or another job market. but see note 15 below. See Hey (1981:Ch. 5) for an accessible survey of this literature.

374 According to these theories, every worker is unemployed because he turns down all proffered vacancies until he has been offered his acceptance wage. The acceptance wage is the wage such that there is no marginal gain in expected utility from continuing to search. Though firms search for workers, if only in the sense that they do not necessarily hire the first applicant for a given vacancy, most search theories tend to ignore this phenomenon and attribute unemployment solely to workers' searching for firms. The testable assertion here-- that the unemployed have refused actual job offers--has not been pitted against empirical evidence very often. But when it has been tested, the idea that most of the unemployed have refused wage offers has not fared well.~° Structural Unemployment In contrast to voluntary, frictional unemployment is the notion of involuntary, structural unemployment, which is defined by Killingsworth (1978:22) as "joblessness--usually long-term--which results from basic changes in the economic structure: new technology, the decline of some industries and the growth of new ones, geographic relocation of industries, permanent changes in consumer tastes, changes in labor force characteristics, and so on." For unskilled workers, among whom are most young people, legal minimum wages or high union wage scales may be an important barrier to Thor example, Rosenfeld (1977) has reported empirical work on the 3,238 out of 4,668 unemployed in the May 1976 CPS who answered supplementary questions on their job-search behavior. The high nonresponse rate and low potential for disaggregation indicate a need for more special surveys of this kind. Yet the implications of the respondents' answers for the validity of search unemployment theories seem clear. Since only 32 percent were on layoff, and more than 81 percent of laid-off workers reported some effort to find an interim job, only 6 percent of the unemployment could have been seasonal or cyclical in the sense used in this paper. Search unemployment due to high acceptance wages seemed less than universal: 22 percent stated willingness to accept jobs paying less than the federal minimum wage, then $2.30 per hour; another 33 percent were willing to take a wage between the minimum and S2.99; and only 22 percent reported an acceptance wage of $4.00 or more. Finally, only 10 percent of those who were unemployed four weeks or more and who had contacted at least one employer reported having refused any job offers. i~Killingsworth (1978) recently retold the postwar history of this old idea.

375 employment. 12 In an important theoretical paper, Weiss (1980) shows that even in the absence of legal minimum wages or union pay scales, firms may find it optimal to set wages fairly high and to refuse to employ members of certain demographic groups, even though they are willing to work for lower wages. At the risk of excluding some dislocated workers from the definition, 3 the structurally unemployed may be thought of as those who have searched for jobs but found no employers willing to hire them. This simple definition maximizes the contrast between frictional unemployment and structural unemployment, while remaining consistent with neoclassical labor economics. The frictionally unemployed will join the ranks of the employed as soon as they lower their acceptance wages. But lowering their acceptance wages will not help the struc- turally unemployed find worker 4 they have not refused any wage offers. There is some empirical evidence that the notion that the unemployed are refusing wage offers is especially inappropriate for young men.~5 If this is true, then empirical models of search t2See Demsetz (1961~. But a legal minimum wage or union wage scale need not be binding constraints on employer behavior if there are other, higher, wage rigidities. i3Lucas (1978) seems to. We might conceivably argue that the laid-off skilled steelworker in Pittsburgh who won't sell his house and take a minimum-wage job is voluntarily unemployed, but we cannot argue as easily that an unskilled teenager who cannot get that same job is unemployed voluntarily. ~ 4 Indeed, in Weiss's (1980) model it is precisely the positive relationship between the acceptance wage and expected productivity that causes the unemployment of workers with low acceptance wages. Compare Lucas (1978:354~: "The unemployed worker at any time can always find some job at once. . . e However miserable one's current work options, one can always choose to accept them." Stephenson (1976) analyzed 281 respondents of the 300 unemployed males aged 18-21 with 8-12 years of education who sought full-time jobs in November 1971 at the Indianapolis state employment service office. He states (on p. 110~: "Nearly 90 percent of both white and black youths, when describing the search before their last job, said they took their first offer. In contrast to the search literature which usually implies a choice among several offers, the central search problem of young men may be to find a single offer." Cave (1983) exploits this insight in modeling unemployment among unskilled workers. Of course, because of its self-selection, Stephenson's sample may not be representative of all unemployed youth in Indianapolis at the time.

376 unemployment 6 may not be appropriate for individual data on youth unemployment. Alternative models directly analogous to those used on aggregate date t 7 are proposed at the end of the next subsection. Implications for Empirical Work on Youth Unemployment Young people's behavior in labor markets is even more complex than the behavior of their elders. Unlike prime-age males, young people have nonparticipation as a real option, and they exercise it often. Whether they are in or out of the labor market, they must make another constrained choice their elders rarely face--whether to stay enrolled in school. Moreover, they work part-time rather than full-time more often than older people do. Empirical studies of youth labor markets must deal in some fashion with the joint determination of school enrollment, military status, labor force participation, hours worked per week, and wages. There is important simultaneity between par- ticipation and the chance of unemployment-if participation is chosen. There is also simultaneity between the number of years of education a person has and the chance he or she will find a place in the labor force. In addition to the simultaneity problems, there are problems of definition for the labor force variables. Several very different kinds of behavior are reported as the same empirical phenomenon, "youth unemployment." For someone who has quit school permanently and who cannot rely on family financial support, reported unemployment may reflect a chronic inability to find any hours of employment at any wage level. This kind of involuntary, structural unemployment may constitute what Conant (1961) called "social dynamite," and it has grave implications for adult poverty and crime. At the other end of the spectrum of interpretations of these statistics, reported unemployment in a particular week may reflect brief job search or normal experimentation with possible careers. For someone who has i6Since the search literature is mainly theoretical, there are few empirical search models to criticize. Kiefer and Neumann (1979, 1981) are careful to use data on permanently laid-off men for whom their sophisticated search model seems especially appropriate. i7For example, Fleisher and Rhodes (1976~. i8Recent empirical work, though not conclusive because of poor data, tends to make Conant's fears seem ill-founded. Freeman and Wise (1982) briefly survey work that, based on longitudinal data, finds no significant effect of employment history per se on later labor force behavior, once persistent individual skill and motivation differences have been controlled for. But Cave (1981) and Levy (1982) criticize these results as possibly reflecting unavoidable selection bias against the relatively small demographic group Conant worried about.

377 completed his or her education fairly recently, reported unemployment may reflect a single episode of leisurely job search or unrealistic wage expectations. Summer unemployment by those who have never left school ought to be treated as possibly the common experience of many young people. Someone who has worked long enough in the past to be eligible for state unemployment compensation may misreport actual nonparticipation as unemployment. A further problem of definition arises because many chronically jobless youths may not show up in unemployment statistics at all. Discouraged workers are counted among those who are not even part of the labor force in a particular week. They may have been unemployed for a long period in the recent past, they may be permanent school- leavers, and they may have low reservation wages, but unless they engage in specific search activities they are not counted as part of the unemployed labor force. In addition to being too inclusive an indicator of chronic joblessness, youth unemployment may be too exclusive. Another important statistical problem plaguing empirical work on youth labor force behavior has sometimes been called "ecological correlation bias" (see Freeman, 1982:115~.~9 Much of the empirical work on unemployment has used the Standard Metropolitan Statistical Area (SMSA), not the individual, as the unit of observation.20 The proportion of variation in labor force behavior across areas that is generally explained by SMSA regressions is much higher than the pro- portion of variation across individuals that is explained by regressions using individuals. Individual regressions may attribute to individual characteristics (such as education, race, and family income) explanatory power that really belongs to area variables (such as the density of employment opportunities) that are correlated with the individual characteristics. It is surprising that few studies of individual labor force behavior have made use of area information that may be available (albeit at great cost) even in microdata. 2 ~ i9Rosen (1984), reporting on data from a BLS cooperative federal-state statistical program, indicates that there is a great deal of variation in unemployment among local areas. In 1979, local unemployment rates ranged from 40 percent in Menominee County, Wisconsin, to less than 1 percent in Sioux County, Nebraska. 20Examples are Gilman (1965), Kalachek (1969), Fleisher and Rhodes (1976), and part of Freeman (1982~. These SMSA regressions generally explain a much higher proportion of variation in SMSA unemployment than is explained (typically well under 10 percent) in individual regressions. 2 ~Abowd and Killingsworth (1984) are an exception, although one might quibble with their choice of geographic area variables to match with individual data.

378 Increasing the level of aggregation by using area characteristics to eliminate a troublesome bias suggests making other analogies to labor force models for more aggregated data. By direct analogy to a recent analysis by Fleisher and Rhodes (1976) using SMSA data, a model to be estimated using microdata ought to have at least two simultaneous equations, one for an individual's probability of labor force partici- pation and one for his probability of unemployment conditional on labor force participation. In the participation equation, a coefficient on the unemployment probability would give the discouraged-worker effect, while a coefficient on an unemployment dummy variable for other members of the household would give the added-worker effect. If a separate conditional unemployment probability equation was estimated for each labor market, using only the young labor force participants in that labor market, the ecological correlation problem might be reduced. A Structural Model for the Youth Labor Market An alternative and even more direct approach to these problems of modeling the youth labor market is available. An appropriate, though likely quite expensive, empirical framework might generalize recent work by Heckman (1979) in the following way. Suppose that a microdata sample has i young people, and that each individual is revealed to live in one of j geographic areas, which may be considered separate labor markets. Let Mi, Ui, Ni, and Si be dummy variables for individual i for unconditional employment, unemployment, nonparticipation, and school enrollment, respectively. Let jail be the geographic area in which individual i lives. Let Zj~i' be a vector of characteristics (such as the proportion of working-age people who are young and the fraction of jobs that do not require much skill) associated with area jail. Let Xi be a vector of characteristics (such as number of years of schooling and area type within the geographic area) specific to individuals. Let whip be the wage for individual i. Then consider the system of equations, Pr(Ui=l; Ni=O) Pr(Ni=O) Pr (Si=O) _ Pui = Pui(Xui~ Fuji wi) _ PNi = PNi(XNi~ Wit PUi) ~ Psi = Psi(Xsi~ ZSj (i) ~ Wit PUi) Wi = wi(XWi, Zwj (i), PUi) This system of simultaneous equations uses some of the findings in the literature on the youth labor market to impose an empirically testable structure on microdata. 2 2 For example, the hypothesis of 2 2Most likely, the best data set for this purpose is the 1976 Survey of Income and Education. It is three times as large as the monthly CPS and has more wage and geographic data revealed for individuals than does the census microdata. Abowd and Killingsworth (1984) and Freeman (1982) chose it.

379 Fearn (1968) that unemployment may be strongly associated with school enrollment and that wages have a weak effect could be tested directly with the third equation; the second equation would pick up any discouraged-worker effect as a negative coefficient on unemployment; and so on. Much of the difficulty in empirical work on the youth labor market stems from fairly complex sample-selection problems, which the model presented here could capture. Wages are observed only for those who are employed; for everyone else, they must be imputed. Most important, conditional unemployment is observed only for those who participate in the labor force; a conditional probability of unemployment must be imputed to those who are out of the labor force. By straightforward extension of Heckman's work, a multivariate, normal-error structure for the four equations could accommodate this sample selection and thus _ _ _% a, . ~ ~ ~ =~ ~ _ ~ : by_ ~ ~ _ ~ by: A= ~ ~ ~ ~ ~ ~ ~ ~- ·: f ~ , ~ ~ a~u~= Baa On Fly w~ v~`e~w.=c '~;:~. To reduce the computational complexity and cost, the last two equations could be dropped and wages ignored, which would produce a microeconometric model more like the Fleisher-Rhodes model for SMSA data. NONSTUDENT TEENAGERS, AGED 16-19 Sample Selection and Descriptive Statistics Table 2 shows how two groups of black male teenagers were selected from the 1980 Census 1 percent Sample C, and how two groups of white male teenagers were chosen from the 0.1 percent C subsample. Table 3 shows how four similar groups of black and white female teenagers were chosen for analysis. Of 12,090 black males aged 16-19, 340 were inmates of institutions; 54 of 6,950 whites were inmates. Of the remaining 11,750 black and 6,896 white male teens, 3.7 percent and 2.4 percent, respectively, were in the armed forces. (The 8,239 black male students and 4,856 white male students will be discussed below.) There remain 3,082 black male civilian noninmate, nonstudents, 1,805 of whom are labor force participants. In addition, 1,875 white civilian noninmate, nonstudent male teens have been selected, 1,577 of whom are labor force participants. Almost 56 percent of the black male noninmate, non- students are located in the Census Bureau's "South" region; little more than 6 percent live in the "West." In contrast, the white nonstudents are distributed more evenly. As in the case of the blacks, the South has the largest share of the sample population and the West has the smallest, but the shares range only between 18 and 33 percent. Tables 4(a)-4(h) present descriptive statistics for the four black samples and the four white samples. The tables reveal some very striking gross racial differences in labor force behavior. First, 41.4 percent of the black males, but only 15.9 percent of the white males, are out of the labor force as well as out of school. Employment- population ratios are 68.6 percent for white males and 40.7 percent for black males. Among labor force participants, the white male unemployment rate is 18.5 percent, but the black male unemployment rate

380 _ _ _ ~_ __ o:) CC O O ~ ~ 0 ~C~J ~· 0 5 <) L~-\0 ~ C/)~ ~! ~ J L~ ~JC~J ~0 Ct/) L-\ ~0\ \0~ ~ ·a:~ ~o r ~h~ 5ro C~J O a ~ ·~ a: `0 J N ~-oO ~`0~) _ _~ _ _ __ __ \0 ( ~_ ~00 ~1\0 L ~- 3 _~ oo 0 3~0 ~L~ _ _ G`\~ ~-3 ~ ~ ~5 = ~_ _Cx 1- ~eJ~ 0 r~ ~ ~- _ ~-.__ __ - (\J CJ~ O ~ oo ~ ~C - ~n Q . _ C) . _ ~5 au C' o 11 o Q a S: a a, zo ~5 . _  . _ . _ . . o . _ a, CD - Q CO ~n UJ J a, ~'.__ _ __ - = C> - <~ _ J =~ ~i O L ~ (\J~ ~ ~\0V3 ~ _ __ __ ~J ~)J L~ _ ~ kD ~ L~ C~ - - C;~ 03 J ~ N co C7\ GN L - \ ~ <) o ~r~ oo ¢ ~:) ~O ~J ~J N - - _ _ ~ ~- - 02) 1- O ~ - _ L~ ~O r ~0 ~ ~_ \0 0 L~ r~ _ L~ ~_ L - \ ~ -  _ _ =< ~ 0O cr~ ~ :~ ~ . - - o_ \0 ~J ~J- _ _ _ _ L~ 0 ~ CM ~ L~ Z CN GN _ L~ _ _ L~5 ~ G ~ L,J -_ O ~oO Z L ~L~- ~) _ _ O _ ~C~ L ~_ ~I I llJ ~\0 ~_ LLJ a) a ~C) ~ 6 0 0 C ~0\ ~ ~r~ 1 ~1 ~ \0 0 ~O D ~f, ~1 ~J tn ~n C C C ~C ~ a) a, a' a' a) a, ~:5 _ _ ~. 4 ~n u ~v - ~C U' - C C: ~CC O O C ~O O O ~Z Z O o ~Z Z ·- C ·- C ~a) a) a) a' ~a' ~a) :5 Q) ~4 ~J 0) ~4 ~Ct ~ ·- ~E ~·- E E E ~a) c: c c 4 ~a) ~C: ~o, _ ·_ ·_ ._ ~_ ._ ._ ._ C C ~C ~0 0 0 - ~ u ~O O O C) cn cn G) Z Z Z ~, ~Z Z Z a' ~ ~ c: c ~a " ~ O ~ ~C ~ O C) ~c ~ c ~ ~c ~c a) ~ 0 ~ cn c ~ - ~ ~' t= L ~· - .L,·_ ._ ~ ~E ~._ I s ._ ._ ~- - C- - C) C - C - - :E ~ ~- ~·- ~·- ·- ~ ~._ ~._ ~._ ._ a: - C: ~ c; > ~s> > - ~e~ a ~> ~> > ·- C>_ ~ E O E - ~E O E · - _ · - ·- ~' ~ ~ _ Z ~'_ ~' ~ ~Z  J -C - ~C/0 c ~~ Y Y ~Y ~a ~a) c c, ~ ~ ~ ~, <( ~ U^i C; (,0 ~ <.r, c; V) ~C; <( ·- U ~- ~- ~·- - ~O C) - G) - C) - C;- - CD '- a) ~ C; c a'` ~ _ Z _ a: J ~ ~ Ja: ~ 3 ~3- 3- 3 3 <: · _ J IC 3 . ~ Q E V} - c a' - a' - E . _ o c V) 5) . . - C~ oo C~ E E o o 4 - e~ E - Q U) i : c o - . _ V, - - a, Q E tn - o . _ a) U) ~ . ._ Q D ~ - CL ~  2 D Q) ~ ac) Q - - o o ~ - ~ E ~ C m · 0 111 c ~ 4- _ . _ O ~n

381 ·<) L ~O· 3 ~)~ O f') ~)tl) J ~O 0 C~ 0 ~0 J - - ~ 0 3 ~ 0 u cr. ~ c~ - - ~ C' O = - cn ~ L - ` ~ _ - ~7 a) E 11 lc Q ._ C) . _ a, o 11 o Q au C53 C ~C ~5 zo . . o . _ C~ a) ~n a, Q E ~n  111 a, z · . ._ J C Y cI - y C: ~: J - C - E a' L~ a)a EE CC ._._ CC OO ZZ C  ~, ~- .L.~ C C: - CD U' C' J - - C 1- Cx J ~ o: ~ r~ - oc; Cx ~ G~ = 3 C L.~ 5 ~. _ 0 c: ~ J ~) _ - - _~ 0 - 0 - < r~ - - L~ CtN \0 L~ ~J O ~J - a) o LL 1 ~0 o r, ~J ~n C a' - U. C C z C a, ~n C o z a, E C ._ C o z - a" Z 0N ~ ~ _ - z C oo  . ~ 0 `0 ~r~ a-O - 0N ~0 a' o C ~L~ 1 o J ~n 4 C C a a) . _ cn ~n E C C a) 0 0 Z Z a) a 4 ~4 C ~ E E E C C C . _ . _ . _ C C C O O O Z Z Z . ~ Q C C~ a) Q ~' ,~C C o C tn - C~ 00 a, a' o 'I o 4- a) Q a) n j ' : C: o  4~ C . _ - C: Q E ~n 4 - o . _ ? . _ ~n c a, c 4 o ° a) Q ~n C a, a) Q c C C . . UJ O: a, ~ . O

382 O O ~ ~ ~ ~ ~ ~ J ~ oC) O -O (~ c() ~ J J C~ (X) <: ~(>J==J~loOL~vol- -ocoCyJL[>NI~~40~(~10 > 0 o: ~ 0N ~ ~ ~ ~ a~ LL] O ~ ~ _ ~ ~ (>i J J :: C) . . . . . . . . . · . 0 0 ~ ~ ~ 0 ~ ~ 0 0 C O ~ O ~ \0 o: ~ ~ ~ ~ ~ ~ O ~ ON ~ aN ~ ~ ~ ~ ~ 0N Z O ~ C\J (\J <) ~ L~J ~:) ~ (>J <: 0 ~ L~ ~ <) 0 o: L~ 2 J O -O ~ ~ ~ ~ ~ ~ ~ ~ ~ 0\ O '4 CO ~ \0 ~ O=t 04\D r~_ J== O > 0 ~ O (\J LS) O O N C\J ~J 0> · . . · · · · . . . - 00~0 ~ 000~00 ~ O O O O ~ O O O ~ O O C/) O O O ~ ~ O O ~ ~ O O O ~ O 0 0N 0 ~ ~ O O O O O O O C~ O ~ ~ ~ O O 1~1 0 0 0 ~ v0 0 0 0 0 0 0 O O ~ ~ O ~ O O O Lr ~ O O O O O 0 2 C O O i- hJ O O O O J ~ O C) O O °~) Ct) y........... C: =(\~-o3<OO3J=I-M:) - O oo ~ 0N U~ ~ O O LSNO ~ (X) ~ ~ ~-(\] ~1 _ CO  O ~ U~ ~ ~ \0 C~ - O O O CC) ~ ~ ~ O ~D ~ ~ ~O O ~ oD ~ <0 ~ ~ oO ~ O ( ) O o: O O C\J J 1~ J c~ C/~) O ~ ~ J O J ~ ~ oC) ~ N L3J Z O oD c~ O J (~ J O>`o J LrN 2 ~ O ~ J ~ o: ~ ~ ~ ~ ~ 0\ LL) O ~ co L~ ~ O O LS~ J ~ L~ · . . . . . · . . · O O O J O O O O O o: ~n m ~5 a, a, 25 ~n o z . _ . _ . _ - lL C: ~O - J ~C) tC m ~ c, z ~ <I: ~ Z ~ O I ~-O CD -~ hU ~ I ~ C:) ~ ~ Z <: 00000000000 00000000000 00000000000 00000000000 00000000000 00000000000 O O O O ~ O O O O O O CO O O O O L~ O O O O O L~ ·.··.·...·. N ~ ~ o:) ~ O o0 J J 1-O L~ ~ O ~ ~ ~ ~ ~ o~r~ - J ~ ~ ~ 1- J C . ~ a) n E 4J a a' Q a; 4- E . _ o  - v, c: . . oo a' a' o o a' E Q a) ~n · : c o .c (n O ~ o: ~ ~ ~ Cx ~ O \0 _ _ \= ~ ~ ~ ~ \= G~ J ~ u~ CM o: aN ~ ~ ~ ~ ~ ~ ~ ~ cr ~ _ ~ c~ ~ \0 -O ~J ~ ~ ~ ~ ~ N J u~ ~ >O~ JCUL~C;>C~r~<co t~ ~coa:_ L~ J= - C~ ........... O O ~ ~ J O O O O ~ O CO L~ O C~J cC O ~ O O <0 ~ <0 J C) ~ 0~0 ~ O ~ ~ ~ L5~ 0~) 0N Z O ~ ~ ~ ~ ~ ~ ~ a: J u~ <( 0 o: ~ (~ a~ ~ ~ ~ vo -O ~ ~ ~ a: ~ J oc) ~ ~ O O ~ O O J L5~ ~ ~J ~ ~ ~ <t 0 ~ a~ c~) ~ oc) ~ (\J ~ m) \0 > 0 ~ ~ ~ 0N 0 0 ~ CM ~ ~ . · · . . . . . . . - 00~0~ 000000  U) O O O O ~ O O O O O O U:) O O O O C~ O O O O O O 0 0 0 0 ax ~ 0 0 0 0 0 ~ O O O O ~ O O O O O O L~ O O O O ~ O O O O O O O O O 2 ~ O O O O O O C: O O O O o: O O O O O O h] O O O O ~ O O O O O O Ct) ~........... C ~ ~ ~ ~ J ~ ~ co ~ ~ o - O ~ ~ J J 0~(U 00 ~ ~ C) o:) ~1 C~:~ ~ ~0 (\J ~1 0 ~Z~ J _ ~ ~0 > oooomr~- - 0000~< O O O 0 0\ \0 ~ ~ ~ ~ \0 O O O O 0 \0 \0 ~ ~ \0 ~ \0 ct) 0 ~J ~ o:) \0 oo 0N ~ 0~) 0N (M l~l z 0 Lf~ 0 co ~ o: J ~ eC) 03 C:) ~ O L~ ~ ~ ~ C~ ~ ~ 00 L~ \0 L~ O ~ ~ ~ \0 0 ~ ~ \0 ~· - -. O O O CS~ O O O O 0 oo a a E ~ ~a) (D O O O O O O O O O O O 0 0 0 0 0 0 0 0 0 0 0 0 ~ooooooooooo ~0 0 0 0 0 0 0 0 0 0 0 ._ ~ooooooooooo ~ooooooooooo a' O) ~ G O O O ~ O O O O O O ~a) cn 0 0 0 0 ~ 0 0 0 0 0 0 ~. . . ~ ~-~00~03- . ~1-~ ~ J ~ 1-~ N o~ O _ ~G v0~ N N O I5 ~ a) ~ =' Q ~n ~ E 0 c ~Z c ~5 0 a) ' ~ a' `1, Q In O ~ Z ~ J : E l ~ ~ c: ca cr c m ~ c: z ~ ._ ~ ~ -=m O ~Z C) C) I <:-O ~ ° 1 ~- ~ L~J C: I l~ C: U) :E Z <: v' a) . . ~  Q E ~n a c) a) Q a, 4~ E c o c' . . oo a) c: a) E E o o a' E Q a' · : C] c o 4J ~7 . _ c tn - a) Q E 4~ o .o . - Q Q a) Q ~ E a' a -c ., 0 c ) a' Q m o .. LL c: cc c ~ ._ 0 0 cn ~

383 Z - (\J L~ CN i- KD L~ ~ ~ ~ O J ·~ <) C~ C>-G>~:) - '= ~ L~ J ~ ~ aN ON ~o ~ ~ ~- ~ ~ O (>J C L-\ ~ J ON '( =' ~o (\J L(~\ ~ C: ~ L~ <) O - O O O ~ 1- 0 J L - \ O LS~, > =xD 1 - 0~0~) ~OJ LLI=Jo:JI-(>J~=JO: t~ . . . . . . . . . . ~ O ~ ~ ~ O C ~ 2 0 U) CO  CD C' m a' LO ~: Q . _ . _ C~ o 11 o a) S: a' S: aJ ~5 4 - z . _ . _ . _ - LL m  L-J C' L~ _ . ~ L~ - O O L~ ~) C~) 0 ) L~ ~ (O Z ~O - :) ~ 1 ~N ~: cn L~ > ~O '~0 o: - O ~ O ~ 0\ ~O ~ ~ k0 J C: 0 v0 ~ ~ O U) O (`J O CxJ N L~J Z O L~ \0 ~ 14- C) ~ O O ~ L~ ~ L^J O ~) \0 L~\ :! ...... ~ C O O LrN O ~Z ~ C: ~ O L4J C: Z ~- _ Z c) (_) T _ ~ ~ (_) ~ O ~ O ~ ~ CSx ~ ~ O ~ J O ~ J Lt~ :~ O ~ ~ 0~) 0 L~ ;~) 0N ~ ! ~ L~ ~ J ~ a~ ~) - ~ ~ ~ GN L - \ ~ ~ 1-1-~ 0~0 ~(~1 J- 0$ =" ~ := (~ J ·~ J ~ 1-O O (~1 (>J . . . . . . . . . , 0~0~000~0 ~0 C~ O O ~ O O t~ O O O O O - 00000 (~J 3 ~ O O O ~ O 2 O 3 O ~ O O C) O L~ J O O O O 1-  <0000 - . . . . . . C ~ O L~ J 02) \0 1- L~\ ~J L() ~ O (~1 o: N r ~ J oD O ~ ~ 1-<0 J <) ~ ~1 C: v0 C~ .r . .. , O O OOOOO OO C O O O O O O OOOOOCO O O O O O O C ~n a) _C . a) Q E a) a' Q . _ o ~n Q . . 00 a c E o o 4_ a) E Q a) ~n i · : C: C] o 4- C ~n a~ Q E cn o . _ CD . _ C, ~ Q ~ E  a, a~ v -' .,_ + a. o C~) ~ a' Q' Q m ° .. LU cc ° E cr: L ~ ~ 6-O (D O O L_ C ~ ~, ~ < (0 ~ O O O 2 O C O 0~00000 C L~ O O O O O 000000~ . . . . . . . L~ 00 ~ ) O ~ ~ \0 (J\ N ~ L~ (~1 L~ 0N 0 ) 0 C'J 02) c ~ Z ~! =' ~ 03) ~ C~ ~ L~\ ~ O O O GN ~ O ~ 1- L~ C~ 5N -:)~O-O(~)OC3:)O~)~) ~) ~ ~ ~ L~ ~N J ~ ~o 6 0 ~ ~0 ~ L~ ~ O ~ O ~- --O: C ~) KO l~ O l-O: > ~ (~) ~ (\d ~ O ~ ~D 0:) J L`J O ~ ~ J ~o co ~ ~ co o ) .......... 00~ 0~20~00 L~ O J ~ oC) 1-~ ~ J ~ <) (~) C \0 o ) M0 0 ~ L~ O <0 ~ ) Z O (41 (~-) ~ C~ ~ ~ (\J 1- 6 0 ~ (~ ~ ~ ~ ~ 2 ~ C~ - 2 L~ O O L~ (\J aN L~ L~ O L/~ ON C~ ~ () ) o ) o N <t ~ L~ 0N ~ O C~ (\J ~ Lf) (~1 > C ~ 0N ~ ~ O O N ~ 1- . . . . . . . . . . O C ~J O L - \ O O O O N co c 0 0 0 a~ 0 0 0 0 O 0 2 0 G~ C O ~ O O O O ~ J O O O O =0000~000= O O O O ~ O O O O O O O O ~J O O O O O O L~J O CD ~ . ~ ~r~ - 0 - ~ ~) L~ u7 Z _ ~ <: a' . _ C~  - :~ ~5 Q . _ C~ . _ a' C' o LL o Q ao a) 4- a, ~5 z . _ . _ . _ - - 111 m O O ~ O O O O C~ O O ~0 0 0 0 3 . . . . . . . . ~ oD J ~ 1-~ <) o: _ ~ <) ~ ~0 J O oO J ~ ~ ~L~ (\J L~ O~- `0L~ 0N > O cO C~l ~ J ~) N ~ N - O L~ ~ J N a~ J ~ J N ~3 ~ J J co J ~ O C~ ~J C) O N (M L~ O a~ O ~ r-J L~ ~ ~) O C~ ~ ) ~ J (`J I~J Z O J Ct, ~ ~ ~ ~ ~J L~ 0~) C<) C~ J O (\J (\J I=J O ~ CS~ X) ~ O ' ) . . . . . . . . . 0 cr, O O O O O O O O O OCOOOOOOO 000000200 O O O O O O O O O  000000000 ~ O C O O L~ O O O O ~ ~ 0 2 0 ~ 0 2 0 0 m0000~0000 . . . . . oo ~ ~ r~ o: ~ ~ S>~) O~D J O 0:) N L~ N ~1 ~ ~1 ~L~ (\J ~1- '~ N L~ h] Z ~ O - <: C: L ~Z -~ ~L~ ~ ~ ~ T ~ - C (~) - =) L.J (_) ~ L~ C ~ ~ ~ <t ~n - . _ _C . _ a) Q E ~n 4 a' a) Q Q) 4 - E . _ o U, Q . . 00 cr a a E E o c) o c E 4  Q a) j c o . _ C ~n - Q E 4 o . _ . _ Q CL ~ Q ~ ~. a' ~ C-) Q - _C 'I o m a) Q o .. a IIJ ~ ~_ ~ . _ O O C/) ~

384 Z o ~ o o o o ~ oo ~ ~ ~ -O ~ 1 J oO ~0 ~ O (U ~ 0~0~ 6 0 C:) ~ N ~J ~0 -O O U~ ~ ~ ~ > 0 Lt~ ~ 0N 0 N L,J O ~ 1-J- 0\ C) . . . . . . O O ~ O ~ O L~J ~ O J ~ CXJ 0~ - 0 C~J ~ (U-O oo 0\ 0\ U~ C~ J ~ eJ · · · . O O O O L3J O ~ ~ ~-~ \0 03~0 ~oJ 0\ OOO\0 - L~O=O~O\ - Z O oO 0 0` 0 0$ ~ ~ J <:: O ~ ~ ~ ~J <) 0) 0D LnJ LS -00-~J:,J~O>~O~ ~O - C^ - L~ - =oDO - - <1 0 C~J ~) J O O ~ J ~ J > 0 ~ ~ (U ~ ~ O CM (U (U . . . . . . . . . . . O O ~ O ~ O O O O O O J  y m CO a LL a) V) :] 4J z C~ . _ . _ . _ - a' - 111 m mooOo~oooooo ~noooomoooooo 0 0 0 0 0N 0 0 0 0 0 0 OOOOOJOOOOOO llJ O O O O J O O O O O LS~ ~ ~OOOOOOOOOOC~J C ~o0000~00000~ - L~0000~000000 ~ (y........... C/) CY ~ ~ O ~ ~ (\J L~ ~ 1-~ J - 0 ~ a~ c J J ~ O (~)=J - a`~J3~ o:)OO`C~ Z~ =~J J O _ >  O=~=J-= 000000-JN00 O~-~N ~ 00=JJ~-J~- O O~JOO-N~NO~ ZO--NO-N0NJ~ ~OJJ=~-J-O~- Lu 0 ~ a' ~ 0- · ~ -000~000000 ~ _ ~ 00000000000 O O O O O O O O O O O 00000000000 00000000000 00000000000 ~00000020000 ~OOOOJ00000 Cl)00000~00000- . . . . . . . . . . . --J--N~ -=O>-OJJ-ON 3JN==J-=O~ `~J .__- ~:t \0 C ~L~ ~ Z C: ~ O - ~ ~ C-) to LLI C~ Z -~ cn Z C~'=<-O ~ ~ c) ~ i,, tD ~ S Z C ~n . _ . a,  Q E ID 4. c a) Q 4 - E c o c . ~ Q . . - 00 a) E E o C: o 4 a' Ec Q a, ~n i · : C o 4 - C . _ S ~n - a, Q E ~n 4 - o C~ . _ a) ? .o Q ~. ~Q ~ E C ~ ~Q a~ v 4 - C O ~ ~Q m . . LL C:  o ~n o a, 4 - E c c z 0 o~ ox r~ 0 ~-~ ~ ~ ~ o o o ko ~ ~ ~ L~ ~ ~ ~ o -O ~ O ~) ~ N J ~ N O ~ (~ ~ N N ~-J ~- ~OJ~O~ -O ~) Lt~ 1-L~-N ~) N > 0 ac) J N \0 ~ ~ 1-~ ~ ~ LL] O N ~ J-J-J J J ~' a ........... 00~o~oooooo l~ 0 ~1~ 0 0 N O ~ ~ ~ N ~ O 0~ ~O CU ~ ~ ~ (U ZOL~r~\O - ~0U~ ~SON=a:) J - O~ ~. oo -O N N 1-~r~ O ~ ~ N N ~ 0~ J N al:~ ~ ~L O ~ O ~ ~ ~ N N J-J > 0 0 ~ ~ ~J ~J O N N N 0 · · . . . . . · · · . O O N O O O O O O O O o N 'a, 4- ._ a  a) C51 a a) 4 - ~n r" o z . _ ._ . _ C' - UJ m ~OQOO~OOOOOO cn 0 0 0 0 ox 0 0 0 0 0 0 O O O O ~ O O O O O O aoooo~oooooo ~ 0 0 0 0 ~ 0 0 0 0 0 ~ ~ ~oooo~ooooo~ C~ C:OOOOChOOOOO~O - ~OOOOJOOOOOO · ~ ~ ~ ~ t~ J ~ ~ N J - O ~ ~ N ~) N ~ ~ ~ ~ N C-) O~-~ J- ~ I~ ~ ~ \0 ~ Z _ N Gt) - \0 =) J N ~ N 0 `0 C ~O-~ ~J ~ N O ~-0 J ~O ~ (~ ~ N O ~ 0 N N ~ {J) 0 ~-~) 0 ~ ~ N {t:) ~ J J ~J O ~ ~ ~ ~ ~J N ~J--J CD O L^~) ~0~m ~0 00- Z O _ O O ~ 0~:) 0 0\ L~ N 1- ~O~-J00~0 LLJ O O O N N mOmLf~r~ · . -O-00~000000 _ _ ~ O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O 5- OoOO~OOOOO =0000~00000 O O O 0 0` 0 0 0 0 0 ·······.·~ ~ ~ ~ ~ ~ ~ ~ ~ N J r~ ~ ~ ~ ~ (~ ~ tS, ~ N 0\-~J.-~ ~-\0  _ , ~_ N Z ~O - c: m Z -~ ~n ~ C) C~) I ct-C C:) - =) LL) C: I 4~ a ~ ~: z ~ v) a) 4 - s a) Q E a) Q a) 4J E . _ o c y . . oo C) a E o o 4 - E 4 - Q a, ~n j · : C) c o c . _ S - Q E o . _ a) . _ D v, 3 C a, a, S 4 - o tD m o .. a, LL ~  (~ E C ~ . _ O o ~n c Q v' D :' v) 4 - c a, Q

385 ZO - ~O= o om~ ooo~ m~ -OOCOJ - CC) - O`OO ~O~=OL~0C~ ~OJOC)~O``OJ~JO ~ O I~ ~ 02) t~ 3 1~ O > 0 ~ ~ 0\ ~ ~J J 0` ~ O ~OJ`OJJ~) - =40 e OO~OCUOOOOO C] ~ ~OO~=O\CO~ 00D c) 00~0 ZO - ~= tO(U - - ~ 6 O-~CU I~~~J=-J ~ 0~ 0~r ~r ~OC0~0~' - ~0 ~O~O`JOO~IJ~J. > 0 ~ ~0 CM ~ ~ O ~J N ~0 ··.····.·~ OOCMO=OOOOO m CO CL ._ C) ._ a) o lL o Q J a' a  a, 4 - a, 4 - z CD ._ . _ .> - LU m ~n LLJ cn C~ ~ - LLJ ~n ~ - 0 C' Z ~n _ _ Q O 0 0\ 0 0 0 0 0 O 0 0\ 0 0 0 0 0 O O C~ O O O O O O 0 0N 0 0 0 0 0 O 0 0\ 0 0 ~ O O O O J O O O O O O O \0 0 0 0 0 O O ~ O O O O J J (~ ~ 0` 1~~ ~ t_ ~ U~ 0 ~ J. L~J ~1 0D C\l ~ 0D 0 0) ro 0\ L~ OCU-JOJ`U`00 OtU-00~0(~ ~-0 O~J - = J - ~J - - ~ O ~J :t 0\ u~ O N O 1-0 O~Ja~O~J~)NO~-C~J ZO~J~0CS`CUJ - <0~:~0-m~ -- L~o - ~_0~0O e~ ~ e ~ e -O-O ~ O O O 0 o0 _ ~_  O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O 00000000 0000~000 0000~000 ·····... O ~-J ~t ~) ~ 0` . ~O`~:~ - Jr~ ~J~ 00- ~)O _ Z ~: O - ~ <: ~ m C: Z -S ~ L~ Z ~ _` ~ <,-C) -~ Ld.~ C ~ ~ ~ ~ ~ ~ < ~n a) . _ . Q ~n 4 - C a, Q) Q a) 4 - CE . _ C C ~n ~o . . - CO C ,C o C: o C E 4 - ro ~n Q O' C: ._ ~n ~ C) a~ LL C O 4 - C C _ C~ : C) a a, ~ Q O  E rL CD ~ tD ~2 J o ._ CD a, a) ~, ~ .o - ~ 25 a -' ~ E o C ~Z ~Q a, c 4, . ~ O a' ~Q . ~ m ° ~ ·. G) Ul cc c m 0 'o ~ C ZO - ~ - O O ~D ~ ~ ~ ~ ~ ~ ~ 0\ -o~om`Oo==o``o~ ~O=~J=~0N ~OOJ-O--JO~ -ON~ON~N~ ~OJ=NN~J~JO ~O-~-J~J-0 C: . . . . . . . . . . OO-OJOOOOO _ ~O=N~-~0~0 OONO-~-J~N~ Z O ~ ~ ~ ~ ~ \0 ~ON-~---=N~ -OONN--N~O~ C~O - O-- 00~= ~O_-=N~N-~= >o_~_~_ON-~ . . · . . · . . · . 00_ O\DOOOOO U' O ~n 0 o  C) O L~ o C/) ~ O ~0 - m0 ~ ~ . cn ~ o~ O O . , _ _ Om_~-ON Z~ N N ~0\ U, ~ > - - C~ C~ ~n hJ 0 0\ 0 0 0 0 0 0 0\ 0 0 0 0 0 OJOOOOO NGOOOO mOOOOu~ 0 0 0 0 0 r~ O O O 0 0 NOOOO ····.. _~N~ os CO ~ _ J~0 ~0 ON-~-0 O-~ m~0= mC~O O-~O-~N~ --0 ~ ~) N ~ 0 O N ~ ~-~C) ~0 0 ~ ~ ZO~JO-J0 ~OmO~)O- N ~ ~ N L~IO~= N ~ N O ~ =) ~) _ · · · . . · · . . . -O-O 0\ 0 0 0 0 oO ~ _ _ 0000000000 O O O O O O O O O O 0000000000 0000000000 O O O O O O O O O O ~0000~00000 0000~00000 O O O O N O O O O N . . . . . · · . · - 0~ ---m~ - ~C~\O O-~-0~oO~J_~) J m_mr~ - CiJ J-~ -J. ~ J _ NN C~ Z ~ J O - ~:~ ~ :n Z -~ ~ Z~ ~ I ~-O`_ -~ l~ C~) I L~ C)  ~_ C . ~ Q E 4 - c a a' Q a c c o c cn Q . . - 00 a, E o o c a c 4 Q a, C] j · : C) c o c c cn C: a, Q E ~n ~o 4~ o . _ a) ~n ? . D a, ~Q ~ E C ~ () D a' ~ C .,_ O ~ ~Q m o .. a' C ) E ~ C _ ._ _ ~ o o ~ C

386 is 30.5 percent. 2 3 Among the young women, 55.4 percent of the blacks and 32.3 percent of the whites were out of the labor force as well as out of school. Among the rest of the young women, the unemployment rate was 32.2 percent for blacks and 13.5 percent for whites. As one might expect, considering the transition from school to work that they are making, nonstudent teenagers who participate in the labor force are a bit older and have completed a bit more education than the group as a whole, regardless of race or sex. But it is surprising that 1979 household income, excluding the teenager's 1979 earnings, is greater for the labor force participants than for the group as a whole. Among the young men, the effect is much larger among black teenagers than among the white group. Among the young women, the difference between means is slightly larger for whites. The fraction of black male nonstudents who are out of the labor force is 2.6 times the fraction of white males; this ratio is 1.7 for females. So one important question is, What accounts for the tremendous differences in labor force participation rates? In particular, what proportion of each racial group is made up of probable discouraged workers? Labor Force Status Without placing too restrictive a structure on the data, participa- tion, employment, and unemployment can be expected to be related to age, region, area type, years of education completed, a marital status dummy variable (FAM), disability, and household income net of the teenager's earnings. Region may to some extent reflect the structure of wages and job availability in local economies. Area type would capture some of these same forces but, unlike region, would be highly correlated with individual and family characteristics. Since young people generally make gradual transitions from school to work, years of education completed, especially those in excess of 12, ought to increase labor force participation and employment. The greater 2 3When these labor force and unemployment statistics for nonstudents are added to those reported in Table 8 for students, national unemployment rates comparable to those reported by the BLS emerge. The implied rates are 14.6 percent for white male teens and 26.9 percent for black male teens. The period census respondents were asked about overlaps with two BLS survey periods, those for March and April 1980. Unemployment rates, not seasonally adjusted, from Employment and Earnings, bracket neatly the white teenage rate in this paper. White male teenage unemployment as a percentage of the civilian noninstitutional labor force aged 16-19 was 14.1 percent for April and 14.7 percent for March. However, the analog here to the volatile "black and other" teenage unemployment rate reported by the BLS is not as close; those BLS rates were 27.7 percent for April and 32.3 percent for March.

387 financial responsibilities of married teenagers ought to increase their participation and make voluntary unemployment, but not involuntary unemployment, less likely for them. A disabled teenager ought to be less likely to participate in the work force and less likely to find employment when he does. Low household income ought to impel a teenager into the work force, but it might also reflect poor job opportunities for every member of the family, or serve as a proxy variable for poorer- quality schools. Exploratory regressions were run using these and other variables to explain unconditional and conditional labor force status in subsamples of young men of the same race. Repeated attempts to use three regional dummies (for North Central, West, and South) and three area type dummies (for rural, urban outside urban area, and central city) generally were unsuccessful for participation, employment, and unemployment equations in all subsamples. Only the dummy variables for central city area and for the South region consistently were significant; often, the South coefficient was large and extraordinarily significant. Thus, in the linear and logistic regressions reported here, central city and South are the only geographic dummy variables used. Intercepts pick up, along with other unidentified effects, the unidentified contributions of living outside the South and outside the central city. Gross Effect of Race on Unemployment In Tables 5(a)-5td), linear probability models of multiple choice are used to show quite clearly the gross racial differentials in the labor force behavior of young people. The table has four parts. In parts (a) and (b), all young men aged 16-19 who were not enrolled in school are included. In parts (b) and (d), only nonenrolled labor force participants are included. In each model, the intercept is simply the value for whites, while the coefficient on color gives the racial difference. For example, in "MODEL03" of part (a), just as in Table 4(b), precisely 15.8933 percent of male white teenagers are seen to be out of the labor force. The figure for blacks is 25.6908 percent higher, for a total black male percentage of 41.584. This figure is slightly different from the 41.4341 percent given in Table 4(a) because there are only one-tenth as many blacks as before. Because unemploy- ment, employment, and nonparticipation partition the sample in (a), the intercepts in the first three equations must sum to unity and the coefficients must sum to zero. 2 4 2 4Pindyck and Rubinfeld (1981:301-303) provide the trivial and tedious details. In essence, even when X and Y are dummy variables, LPM coefficient estimates are computed using the usual OLS formula, (X'X)-lX'Y. The first factor, the inverted cross-product matrix, contains totals of individuals of each race to be used as the denominators of the coefficient estimates. The second factor has the counts of labor force status by race to be used as numerators.

388 TABLE 5ta) Linear Probability l\Aodels Civilian, Nonstudent Male Teenagers MODEL: MODEL()1 SSE290.853302 F RATIO1.35 DFE2176 PROB>F0.2451 DEP VAR: U MSE0.133664 R-SQUARE0.0006 VARIABLEnrESTIMATEERROR T RATIOPROB>iTi 1N r LIlCEl'T1().1552()00.0C)8443197 18.38170.C)C)C)1 COIOR10.0263180.022637 1.16260.2451 MODEL: MODEL02 SSE476.853354 F RATIO95.48 DFE2176 PROB>F().00()1 DEP VAIN: M MSE0.219142 R-SQUARE0.042C) VARIABLEDFESTIMATEERROR T RATIOPROB>tTt INTEItCEP!10.6858670.010811 63.44210.0001 COLOR1-~).2832260.028985 -9.7716().0001 MODFl: MODFL03 SSE324.241827 F RATIO115.54 DFE2176 PROB>F0.C)001 DEP VAR: N MSE0.149008 R-SQUARE0.C)5C)4 VARIABLEDFESTIMATEERROR T RATIOPROB>iTI INTERCEPl1().1589330.008914653 17.82830.00()1 COLOR10.2569080.023901 10.74890.0001 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 19831: blacks, noninmate 1 percent sample; whites' noninmate 0.1 percent subsample. TABLE 5Ib) Linear Probability l\/odels: Civilian, N.onstudent Labor Force Pa rticipa,nts, Male Teenagers MODEL: blODEL21 SSE275.212078 F RATIO16.14 DFE1752 PROB>F0.0~)01 DEP VAR: U MSE0.157085 R-SQUARE0.0091 PARAMETERSTANDARD VARIABLEDFESTIMATEERROR T RATIOPROB>ITi INTERCEPT10.1845280.009980467 18.4889C).C)001 COLOll10.1262070.031418 4.01700.0001 MODEL: MODE L22 SSE275.212078 F RAN-IO16.14 DFE1752 PROB>F0.00C)1 DEP VAR: M MSE0.157085 R-SQUARE0.()091 PARAMETER VARIABLEDrESTIMATE INJERCEPi1(:).815472 COLOR1-0.126207 STANDARD ERROR T RATIOPROB>ITI 0.0099801~67 81.70680.00(11 0.031418 -4.0170O.0001 _____ _ _ ______ _ ____________________________ _____________________________ MODEL: MODEL31 SSE269.970506 F RATIO16.80 DFE1750 PROB>F0.0001 DEP VAR: U MSE0.154269 R-SQUARE0.0280 PARAMETERSTANDARD VARIABLE DFESTIMATEERROR T RATIOPROB>l r I INTERCEPT 10.2095420.012013 17.44300.0001 SO 1-(~.(1776520.021166 -3.66870.0C)03 COLOR 1(~).762G81().()47822 5.49290.0001 I N-rRAc r 1-t).1945700.063717 -3.05370.0023 MODEI: MODEL32 SSE 269.970506 F RA110 16.8~) DFE 1750 PROB>F 0.00C)1 DEP VAR: M MSE 0.154269 R-SQUARE 0.0280  VARIA£3lE INTERCEPT SO COLOR INTRACT t)' l PARAMETER STANDARD ESTIMATE ERROR (:).79C)458 0.012013 0.()77652 0.021166 -().262681 0.01~7822 ().19457() 0.C)63717 T RATIO PROB>ITI 65.80(94 0.0001 3.6687 ().()(~()3 -5.4929 0.C)001 3.0537 0.0023 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

389 TABLE 5(c} Linear Probability Models: Civilian, Nonstudent Female Teenagers MODEL: MODEL01 SSE 196.948170 DFE 2278 DEP VAR: U MSE 0.086457 F RATIO2.41 PROB>F0.1210 R-SQUARE0.0011 VARIABLE DFESTIMATEERROR T RATIOPROB>ITi INTERCEPT 10.0915680.006687799 13.69170.0001 COLOR 10.0265880.017143 1.55100.1210 MODEL: MODEL02 SSE538.071541 F RATIO121.12 DFE2278 PROB>F0.0001 DF:P VAR: M MSE0.236203 R-SQUARE0.0505 VARIABLE DFESTIMATEERROR INTERCEPT 10.5856180.011054 COLOR 1-0.3118430.028335 MODEL: MODEL03 SSE505.261297 DFE2278 DEP VAR: N MSE0.221800 VARIABLE DFESTIMATEERROR INTERCEPT 10.3228140.010712 COLOR 10.2852550.027458 T RATIOPROB>ITf 52.97700.0001 -11.00540.0001 F RATIO107.93 PROB>F0.0001 R-SQUARE0.0452 T RATIOPROB>ITI 30.13610.0001 10.38880.0001 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 5Id) Linear Probability Models: Civilian, Nonstudent Labor Force Participants, Female Teenagers FlODEL: MODEL21 SSE181.706169 F RATIO27.04 DFE1443 PROB>F0.0001 DFP VAR: U MSE0.125923 R-SQUARE0.0184 PARAMETERSTANDARD VARIABLE DFESTIMATEERROR T RATIOPROB>ITI I'4TERCEPT 10.1352180.009808031 13.78640.0001 COLOR 10.1662530.031970 5.20020.0001 tdODEL: MODEL22 SSE181.706169 F RATIO27.04 DFE1443 PROB>F0.0001 DEP VAR: M MSE0.125923 R-SQUARE0.0184 PARAMETERSTANDARD VARIABLE DFESTIMATEERROR T RATIOPROB>ITI INTERCEPT 10.8647820.009808031 88.17080.0001 COLOR 1-0.1662530.031970 -5.20020.0001 MODEL: MODEL31 SSE181.559482 F RATIO9.40 DFE1441 PROB>F0.0001 DEP VAR: U MSE0.125995 ~ R-SQUARE0.0192 PARAMETERSTANDARD VARIABLE DFESTIMATEERROR T RATIOPROB>ITI INTERCEPT 10.1353470.011872 11.40090.0001 SO 1 -0.000406997 0.021084 -0.0193 0.9846 COLOR 1 0.203636 0.047712 4.2680 0.0001  1NTRACT 1 -0.065849 0.064934 -1.0141 0.3107 [VlODEL: MODEL32 SSE 181.559482 F RATIO 9.40 DFE 1441 PROB>F 0.0001 DEP VAR: M MSE 0.125995 R-SQUARE 0.0192 PARAMETER STANDARD VARIABLE DF ESTIMATE ERROR T RATIO PROB>iTI INTERCEPT 1 0.864653 0.011872 72.8338 0.0001 SO 1 0.0004069971 0.021084 0.0193 0.9846 COLOR 1 -0.203636 0.047712 -4.2680 0.0001 1NTRACT 1 0.065849 0.064934 1.0141 0.3107 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

390 What should we use as our measure of youth unemployment? Should we use the unconditional ratio of the number of unemployed to the size of the cohort displayed in parts (a) and (c) of Table 5, or should we use from parts (b) and (d) the rate of unemployment conditional on partici- pation in the labor force? If we choose the first option, we need go no further in our analysis of youth unemployment. For both young men and young women, "MODEL01" shows that the race differential is small and insignificantly different from zero.25 In that case, our emphasis must be on the lower labor force participation rates and lower employment-to-population ratios of blacks. 2 6 Clearly, we must focus attention on conditional measures of unemployment if we are to say anything sensible about racial differ- ences in youth unemployment. The much lower labor force participation of blacks and their higher unemployment rates tend to cancel out each other completely in unconditional measures of unemployment. But "MODEL21" in Table 5(b) shows that the gross male racial differential is 12.6 percent and significant2 7 when we use a conditional measure of unemployment. Since 12.6 percent is less than the 18.5 percent unemployment rate for whites, however, among nonstudent male teenagers the widely believed, roughly two-to-one ratio of black unemployment to white unemployment was too pessimistic by one-third on census day in 1980; it was only 1.68. What happens to the gross differential when we correct for region? "MODEL31," Table 5tb), tells us that the ratio of black to white male youth unemployment rates was 2.25 outside the South but only 1.52 in the South. 2 ~ The corresponding ratios for women are 2.5 and 2.0, 2 sThe asymptotic t-statistic for a logit version of "MODEL01" is only 1.16; the LPM does not lead us astray here. 2 6Freeman (1982:116) gives as one of four "basic findings" the following: "Because determinants of youth unemployment often have the same directional impact on labor force participation rates as on employment, [they] have little effect, or occasionally a contradictory effect, on unemployment rates. This suggests that analyses focusing on unemployment can give misleading impressions about the determinants of the youth labor market position." Indeed, the advance title of the conference at which Freeman presented his important paper was not "The Youth Labor Market Problem," but rather "Youth Unemployment." The approach here permits a more direct attack. 2 7The asymptotic t-statistic from the logistic regression, 3.94, confirms the LPM result. 2 The LPM framework gives unemployment rates for white Southerners, black non-Southerners, and black Southerners as sums of the coefficients. For example, the rate for black Southerners is the sum of all four coefficients.

391 Table 5(d). These differences reflect larger regional differentials in unemployment rates for blacks than for whites, especially among young men. The unemployment rate for southern black males was only 42 percent of the rate for nonsouthern black males, while the geographic ratio of unemployment rates for white males was much higher, at almost 63 percent. From Table 4(a), we know that an astounding 56.8 percent of the black male, nonstudent labor force participants live in the South. Were this not so, the national ratio of male unemployment rates by race would be much higher than 1.68. Gross Effect of Race on Other Labor Force Status Variables Table 5 shows that what was said before about the linear probability model was correct; in any set of LPM regressions of dependent variables that partition the sample, the intercept coefficients must sum to unity and the slopes to zero. When, as in part (b), just two labor market states partition the sample, one of the two LPM regressions is redun- dant. In "MODEL22, " the intercept is the difference between unity and the intercept in "MODEL21. " This simply means that the rate of white employment, conditional on labor force participation, is one minus the white unemployment rate. The slope restriction simply ensures that the racial difference in conditional employment rates has the same absolute value as the racial difference in unemployment rates. In parts (a) and (c) of the table, the race coefficients for unconditional employment and nonparticipation show how the uncon- ditional unemployment ratios came to be nil. Among males, a 28.3 percent gross racial differential in employment-to-population ratios is offset almost exactly by a 25.7 percent racial differential in labor force participation. Among females, a 31.2 percent gross racial differential in employment-population ratios is offset by a 28.S percent racial differential in labor force participation. Structural Interpretation There are two strikingly different structural interpretations for the racial differential in labor force participation. Freeman (1982) and others have found little difference between the wages of employed black and white young people.29 Heckman (1974), implementing neo- classical labor force participation theory empirically, shows how a woman observed outside the labor force can be modeled as having an imputed market wage below her personal reservation wage. If Heckman's model is applied to the youth labor market, ignoring unemployment, then 2 9Freeman (1982:142) reports S1E log hourly earnings regressions showing only a 3 percent disadvantage for blacks aged 18-19 and blacks aged 20-24. For 16-17 year olds, he reports an actual wage advantage of 17 percent for blacks.

392 an explanation for lower black participation is higher black reservation wages. On the other hand, if the analogy to macroeconomic models posited above is maintained, the lower black participation is a discouraged-worker effect consistent with the higher black unemployment rate. The Effect of Additional Explanatory Variables For the reasons stated above, it might be unwise to place much faith in structural estimates of the effect of race on unemployment based on single-equation techniques, or even on system techniques, ignoring the complex sample selection used to generate data on youth unemployment, wages, and school enrollment. Thus far, the focus has been on merely measuring gross effects of race on unemployment. The estimates presented in this section should be considered much more tentative, because of the many sources of bias that have not been corrected here. Tables 6ta)-6Id) present estimates of conditional unemployment probabilities for nonstudents. In parts (a) and (c) of the table, the use of age as a regressor reduces the strong impact of years of education that is present in parts (b) and (d); in fact, age replaces education as the greatest reducer of chi-square. Southern location keeps the strong negative effect on the chance of male unemployment that it had in the LPM. The failure of household income to explain much variation in male unemployment is surprising if unemployment is voluntary search; perhaps replacing household income with a dummy variable for unemployment of the head of the household would produce stronger effect. Race has the largest effect of all the dummy variables and still seems to have a strong impact on unemployment. In parts (a) and (c) of Table 7, there is still no significant racial difference in unemployment ratios even after adjustment for other individual characteristics. Parts (b) and (d) of the table show how race affects nonparticipation when no discouraged-worker effects have been permitted to occur in the equation. Comparison of parts (b) and (d) of Table 7 shows how remarkably the effect of marital status (FAM) differs by sex. There is no independent effect of southern location on young women's participation, even though the effect is strong for young men. STUDENT TEENAGERS, AGED 16 - 19 Sample Selection and Descriptive Statistics Parts (a) through (d) of Table 8 contain descriptive statistics for 4,856 white male students from the 0.1 percent C subsample and for 8,239 black male students from the full 1 percent sample. Parts (e) through (h) give the same statistics for 4,811 white and 8,505 black female students. These statistics contrast sharply with those in Table 4(a)-4th) for nonstudents. A very high proportion of all groups was

393 TABLE 6(a) Conditional Unemployment Probabilities Using Age as a Regressor, Civil fan, Nonstudent La bor Force Participants, Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 1754 OBSERVATIONS 346 POSITIVES 1408 NEGATIVES ~ O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1742.02 CONVERGENCE OBTAINED IN 5 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.2112D-02. MODEL CHI-SQUARE= 97.40 WITH 10 D... F D=0.053 -2 LOG L= 1644.62 P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P D INTERCEPT 5.42764696 1.2824882317.91 0.0000 EDUCATN -0.12686428 0.0338479614.05 0.0002 0.008 AGE -0.29201459 0.0728478316.07 0.0001 0.009 HHINCOME 0.00169628 0.004474080.14 0.7046 0.000 FAM -0.29255623 0.245098341.42 0.2326 0.001 D1SAB1L 0.46270604 0.325922032.02 0.1557 0.001 SO -0.59775813 0.1548124414.91 0.0001 0.008 CC 0.25130318 0.152770402.71 0.1000 0.002 SOXCOL -0.64729610 0.397801032.65 0.1037 0.002 CCXCOL 0.19118747 0.397734260.23 0.6307 0.000 COLOR 0.95430253 0.361124886.98 0.0082 0.004 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 6tb} Conditional Unemployment Probabilities, Civilian, Nonstudent Labor Force Participants, Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 1754 OBSERVATIONS 346 POSITIVES 1408 NEGATIVES O OBSERVATIONS DELETED DUE TO MISSING VALUES LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT CONVERGENCE OBTAINED IN 5 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.9685D-03. MODEL CHI-SQUARE= 81.58 WITH 9 D.F Y= 1742.02 D=0.045. -2 LOG L= 1660.44. P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P D INTERCEPT 0.46768080 0.359909541.69 0.1938 EDUCATN -0.17088188 0.0318564028.77 0.00()0 0.016 HH1NCOME 0.00124161 0.004433050.08 0.7794 0.()00 FAM -0.38980929 0.243564762.56 0.1095 0.0()1 DISABIL 0.44866365 0.324832031.91 0.1672 0.001  SO -0.58287631 ().15ll0105214.32 0.00()2 0.()08 CC 0.21(~87438 ().151724891.93 0.1646 0.001 SOXCOL -0.64609606 0.396410102.66 0.1()31 0.002 CCXCOL 0.15225014 0.396093170.15 0.7007 O.f)OO COLOR 0.94986882 0.359367886.99 0.0082 0.004 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

394 TABLE 6(c) Conditional Unemployment Probabilities Using Age and Education as a Regressor, Civilian, Nonstudent Labor Force Participants, Female Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 1445 OBSERVATIONS 1227 U = 0 218 U = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -A LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1225.96 MODEL CHI-SQUARE= 82.93 WITH 10 D.F.(SCORE STAT.) P=0.0 CONVERGENCE OBTAINED IN 5 ITERATIONS.R= 0.209. MAX ABSOLUTE DERIVATIVE=0.5369D-06.-2 LOG L= 1152.29. MODEL CHI-SQUARE= 73.66 WITH 10 D.F.(-2 LOG L.R.) P=0.0 . VARIABLEBETA STD. ERROR CHI-SQUARE PR INTERCEPT 6.69912234 1.57608915 18.07 0.0000 EDUCATN -0.17411742 0.05434578 10.26 0.0014 -0.082 AGE -0.33559105 0.09334111 12.93 0.0003 -0.094 HH1NCOME -0.01201040 0.00595276 4.07 0.0436 -0.041 FAM -0.22592831 0.20554140 1.21 0.2717 0.000 D1SAB1L 0.62566258 0.40597028 2.38 0.1233 0.017 SO -0.07155320 0.18027037 0.16 0.6914 0.000 CC -0.18357430 0.19867950 0.85 0.3555 0.000 SOXCOL -0.23789505 0.44137867 0.29 0.5899 0.000 CCXCOL 0.41980877 0.45426632 0.85 0.3554 0.000 COLOR 0.90828910 0.42853113 4.49 0.0340 0.042 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 6(d) Conditional Unemployment Probabilities, Civilian, Nonstuclent Labor Force Participants, Female Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 1445 OBSERVATIONS 1227 U = 0 218 U = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1225.96 MODEL CHI-SQUARE= 70.27 WITH 9 D.F.(SCORE STAT.) P=0.0 CONVERGENCE OBTAINED lN 5 ITERATIONS.R= 0.187. MAX ABSOLUTE DERIVATIVE=0.2327D-06.-2 LOG L= 1165.10. MODEL CHI-SQUARE= 60.86 WITH 9 D.F.(-2 LOG L.R.) P=0.0 . VARIABLEBETA STD. ERROR CHI-SQUARE P R INTERCEPT 1.36485040 0.569151905.75 0.0165 EDUCATN -0.25780649 0.0491680127.49 0.0000 -0.144 HH1NCOME -0.01170492 0.005890193.95 0.0469 -0.040 FAM -0.27235321 0.203775611.79 0.1814 0.000 D1SAB1L 0.62538138 0.403537222.40 0.1212 0.018 SO -0.06252239 0.179112500.12 0.7270 0.000 CC -0.19235087 0.197657780.95 0.3305 0.000 SOXCOL -0.27959955 0.439686130.40 0.5248 0.000 CCXCOL 0.39196512 0.452613660.75 0.3865 0.000 COLOR 0.91665873 0.427214974.60 0.0319 0.046 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S.  Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

395 TABLE 7(a) Unconditional Unemployment Probabilities Using Age as a Regressor, Civilian, Nonstudent Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2178 OBSERVATIONS 346 POSITIVES 1832 NEGATIVES ~ O OBSERVATIONS DELETED DUE TO MISSING VALUES LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1906.96 CONVERGENCE OBTAINED IN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.1942D-07. MODEL CHI-SQUARE= 42.29 WITH 10 D.F D=0.019. -2 LOG L= 1864.66. P=O. 0000. VARIABLEBETA STD. ERROR CHI-SQUARE P D IN r ERCEPT0.43564000 1.16266490 0.14 EDUCATN-0.09511908 0.03154844 9.09 AGE-0.05546068 0.06679326 0.69 HH1NCOME0.00338062 0.00423961 0.64 FAM-0.15806311 0.24129147 0.43 D1SAB1L0.05311048 0.29838801 0.03 SO-0.62447856 0.15057693 17.20 CCO.17428858 0.14758136 1.39 SOXCOL-0.22117611 0.36407431 0.37 CCXCOL-0.19581342 0.36304163 0.29 COLOR0.40942448 0.32828724 1.56 7079 0026 0.004 4064 0.000 4252 0.000 5124 0.000 8587 0.000 0000 0.008 2376 0.001 5435 0.0(~0 5896 0.000 2123 0.001 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 7tb) Labor Force Participation Probabilities, Civilian, Nonstudent Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: N 2178 OBSERVATIONS 424 POSITIVES 1754 NEGATIVES O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 2147.20 CONVERGENCE OBTAINED lN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.1532D-04. MODEL CHI-SQUARE= 302.10 WITH 10 D.F D=0.122  -2 LOG L= 1845.10 P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P D INTERCEPT 10.66536953 1.08712728 96.25 . EDUCATN -0.07836863 0.03144011 6.21 0.0127 0.003 AGE -0.62317655 0.063846()2 95.27 . 0.042 Hti1NCOME -0.0()667815 0.00448927 2.21 0.1369 0.001 FAM -0.92074141 0.3()599906 9.05 0.0()26 0.0()4 D1SAB1L 0.92544267 0.25494550 13.18 0.0003 0.006 SO 0.28386695 0.13819125 14.22 0.01400 0.002 CC 0.1lI608299 0.15444()70 f).89 0.3442 ().0()0 SOXCOL -0.62563728 0.30290435 4.27 0.0389 0.002 CCXCOL 0.46067254 0.31294010 2.17 0.1410 0.001 COLOR 1.25767966 0.29242890 18.50 0.0000 0.008 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

396 TABLE 7(c) Unconditional Unemployment Probabilities Using Age as a Regressor, Civilian, Nonstudent Female Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2280 OBSERVATIONS 2062 U = 0 218 U = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1437.94 MODEL CHl-SQUARE= 21.24 WITH 10 D.F. CONVERGENCE OBTAINED IN 5 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.3232D-07. MODEL CHI-SQUARE= 22.01 WITH 10 D.F. (SCORE STAT.) P=0.0195. R= 0.037. -2 LOG L= 1415.93. (-2 LOG L.R.) P=0.0150. VARIABLE BETA STD. ERROR CHI-SQUARE P INTERCEPT 0.73630620 1.39571223 0.28 0.5978 EDUCATN -0.01062479 0.04807882 0.05 0.8251 0.000 AGE -0.13802017 0.08384371 2.71 0.0997 -0.022 HH1NCOME -0.00702891 0.00558243 1.59 0.2080 0.000 FAM -0.66409315 0.19371289 11.75 0.0006 -0.082 D1SAB1L 0.25378135 0.35259030 0.52 0.4717 0.000 SO -0.06763013 0.17256885 0.15 0.6951 0.000 CC -0.13627472 0.18998197 0.51 0.4732 0.000 SOXCOL 0.06433473 0.39554228 0.03 0.8708 0.000 CCXCOL 0.11657359 0.41129213 0.08 0.7768 0.000 COLOR 0.07286950 0.38902035 0.04 0.8514 0.000 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 7(d) Labor Force Participation Probabilities, Civilian, Nonstudent Female Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: N 2280 OBSERVATIONS 1445 N = 0 835 N = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 2995.54 MODEL CHI-SQUARE= 461.55 WITH 10 D.F. CONVERGENCE OBTAINED IN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.1291D-ll. MODEL CHI-SQUARE= 488.29 WITH 10 D.F. (SCORE STAT.) P=0.0 R= 0.395. -2 LOG L= 2507.26. (-2 LOG L.R.) P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P R INTERCEPT 8.82746514 0.9981828078.21 . EDUCATN -0.27468560 0.0345544063.19 0.0000 -0.143 AGE -0.36753557 0.0590064538.80 0.0000 -0.111 HH1NCOME -0.01075921 0.003970127.34 0.0067 -0.042  FAM 1.27876118 0.11323533127.53 . 0.205 D1SAB1L 0.58214365 0.265080054.82 0.0281 0.031 SO -0.01060688 0.113942060.01 0.9258 0.000 CC 0.02361572 0.125377070.04 0.8506 0.000 SOXCOL -0.27130198 0.274396920.98 0.3228 0.000 CCXCOL 0.25352076 0.282743320.80 0.3699 0.000 COLOR 1.43670015 0.2709279428.12 0.0000 0.093 - SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

m au ~7 en a) a) ~5 In . _ . _ . _ C' . . .8 In . _ to an a) . _ Q . _ C) a) m 397 ZO~0=N O=~-J-~-N<~> -O-=NJ-O-~- O 0 ~ ~ \0 ~ ~ ~ ~ o: a: O o: ~ (U ~ ~ ~ of ~ ~ ~ -OWNS >O - to - to) ~O~-JJO~=O A........... 00~ O - C0000 C) cn ~OO=O~J-~JJ OO~J0~= ZO-~JJ~N ~OJJ=~-O~ -O~-N~-N=~- O ~ 0N ~ O ~O~J-ON5~" >OO~N~OON~ . . . . . . . . . . . 00~0~00000 oo cn 0 0 0 0 ~ 0 0 0 0 0 0 en oocomoooOoo 00000~000000 OOOOOJOOOOOO Lit o o O O ~ O O O O O O O O O O ~ O O O O O CJ0000~00000 - m000~00000= cn ~ ........... ~ =~JO~O~= - O-O-~JO~N ON NJ< ~ Zip O=~ JUDO - ~\.0 O~ oo \0 6 ~N  ~n > c" o ~ ~ ~ ~ 3 ~ J ~ O ~ - O ~X) CSs J 1~ ~O~oO O ~ ~ ~ O ~O~aD~J=0~3 U) O N N ~J 3 ~ ~ N ~ ~ J h] Z O ~ 3 N J ~ ~ J `0 ~1 O ~ ~ ~ ~ O N ~ ~ J h] O O ~ ~ ~ O O U~ ~ ~ J ........... ~000~00000 - 000~000000 OOO~OOGOOO 0000~000000 0000~000000 O O O ~ O O O O O O 0000~000000 OOONOOCOOO 000~00000 . . . . . . . . . . ~ 3 O J ~ O N oo~0o: o~Ln~JO~ N J ~ ~(~J J <) ~ ~ ~) J J J~ J 03 ~J C~ C) C: 6 l) L~ ~r-0 c: L" (= ~ ~ Z 6 V) 4- S . ~ a, Q E 4 Q a, 4 -  E . _ o ~n C~ . . oo a) c E o o C~ E 4~ Q a, C] . ~n a) . ~ ~.= _ 4 - ~_ ~2 . _ s tn 2 ~ _ ~ : ~ a, a) a, _ ~n na ~ cn O ._ ~ ~._ c ._ a, ~.. _ C~ . _ :3 ~ .`n <1) ~= ~ ~ E ~ ~a, a) ~ ._ O_ Q , ._ 4- a O ~ a) C) ~Q a) . ~_ m 0 Q  .. a, 00 LL ~LL G c m 0 0 ~ C 2N=~N~ =~-ON~=J~ 2N-~=N~ OO-~J~NNOJ~ OOO=NOJ-~- O~J-~- ON~ ONMJ-O~JJJO . . . . . . . . . . . 00~ 00000~0" ~O~-J ~ ~ O ~ O N OONOJ-~N-- ZON~J-~=O-- 60J=~N-NJ=~N -Oo0~1~0~0~0~ N ~O=OJ~N~ 6 0 J vo ~ o 0 N O ~ J 0\ >OO=~JOONNN~ . . . . . . . . . . . 00~ 0000000 o cooooooooooooo O O O O L~ O O O O O O 00000~000000 OOOOOJOOOOOO ~OOOO~OOOOOu~ ~ 0000~00000 - c: 0 0 0 0 r~ ~OOOOOOOOOOJ u) ~ ........... ~ ==O~J~=J~m - O~O==N~ O=N~O~ ~J= ~ ZJ m~m ~ N~ - ~ J J O \0 u~ <: cn > _ O O ~ O N~- - O-N=~NO' O N N = N ~ ~ IS) O-= c) O o: ~ oO O ~ r~ ~ N 0` 0` ~ O J~N - ~J- LLI Z O ~ ~ L~Lt~) - ~ N J 0 6 0 ~ ~ N ~ o N ~ 0 \0 mOOJ N ~ O O N · . · . . . . · . · . 000~00000 - ~ N ~ OOCOcOOOOOOO  O O O O ~ O O O O O O OC)OOO`OOOOOO O O O 0 0~ 0 0 0 0 0 0 O O O O C~ O O O O O O 00000~000000 O O O O N O O O O O LS~ ~OOOOOOOOOON · . . . . . . . . . . ~O~J~J~N ~Nm=N~J--J =~mOJ ~J=~m J O~J ~N~ N CC ~LL L`-l Z c) ~ O - ~6 C~) h] C: Z ~-~ cn ~ Z ~ ~-6-O (: -~ L~J O I L" (~ C/) ~ Z 6 L~ a S . ~ Q c v, 4 - C a) Q a, - C C o C D . . - 00 cn a, C' ~C E o c' o 4 - c a, E Q a) ~n ·: C) C o 4 - C . _ C -  Q ~C U) 4 - o L) . _ . _ Q ~n C a) a, s o a m o .. LL ,c c :~:) - 0 0 (n c Q ~C ~7 ~7 4 - C a) Q

398 Z O N C C~ (>J ~ N ~ (>J ~0 O OC<<a"= OC`~O -oor~ - o - ~oo~ O oo ~ ~ ~ o: oo ~ o: o: ~O=~ - C'`~O - =O~ -O ~ ~ ~ ~ ~ ~ CO ~ L~ > 0 ~ ~ ON~ONL~C>~ ·n LJ C ~ ~ J °O O r~ ~ J O .......... O O ~ O ~ O O O O C: ~ LJ O Ox ~ ~ N ~ ~ J ~ <0 (: O ~ O ~ o: J ~ 1~ Z O ~ N ~C~ oD ~ 0N~J ~ O 1-~C) ~ J ~ O t~ 1- -O - N - ~ - O - O ~ ~ L~ ~ ON ~ ~ ~ L~ ~ Oa:)~JI-O~IJo:)~ > 0 ~ 1-N o3 O O ~J . . . . . . . . . . 00~0~0000 0N m - CO tJ ~5 Q . _ C) . _ a, C) o 11  L,J L~ cn c~ _ o cn z _ ~ > O ~ Q U) J C, a) ~5 C~ a, a) ~n . _ . _ . _ . . . _ . _ ~n a) . _ Q . _ ~n a) - C~ - 00 UJ J m ~: CDOOOOODOOOOO mooooaxooooo O O O O ~ O O O O O 0000~ 00000 OOOC~C000 O O O O ~ O O O O ~ OCOO=C000= 0000~0000~ . . . . . . . . . . ~LS\0>N=~0~°~ N O 1-0 `0 N l~ l~ ~ O  ~ Lr>J ~ ~ ~ ~ N ~ ~ CM ~ ~ J N O `0 J ~ ~ oO <) ~ N ~) oC = ~ N ~O L~ = O L~ O O ON ~ ~ ~ 1- 0\ L~\ \0 1-~ CJ~ L~ L~ N N 0 ) L~ ~ N ~ L~ ~ ~ 0N ~-=~:) ~) N N LS~ ~) ~ ~ O O N J N :t ~1-0 0 ~ - 1- . . . . . . . . . O O O ~ O O O O ~ 0000000000 0000000000 0000000000 0000000000 0000000000 0000~00000 ~ O O O O ~ G O 0 0 L~\ U) 0'0 0 0 1 - 0 0 0 0 1- . . . . . . . . . . ') ~ ~ N ~:) ~ O ~ oD O N O N O ~ N ~ L~ ~ ~ ~ LS\N N L ~ \0 1- N N~ - N ~LLJ L~ Z ~: O ~ ~ C: ~ LLJ ~ Z 6 -:E CO L3J Z C, ~ I <t-O -=' L~ ~ I L_ C) Ct) tn  . _ _C . ~ a) Q E 4 - c a C~ a) Q a) 4J E c . _ c o c CD Q . . 00 a) ~4J o O ~ O 4J C 5 a) ~ E ~ ~C~ Q ~ ~Q C) i~ ~05. . . C: C~ C) O c' 11 O ~ 4- O C Q ._ c~ c ~a) _ ~ a au a Q E t U, ~ 4 - , cn o .o ~._ ~, .o .. _ 9 ._ CL U) c~ ~  Q (t) ~ E C ~ 9 ._ a) ~Q c 4J ~ a. O c,, c, ~ Q a) m ° ~ .. a) 0D C ) E J ~ c m 0 0 ~ ' ~ _ ~ 0 ~ \0 ~ ~ ~ L~ ~ C~ L~\ ~ ~- ~ ~ 0 L~ 0 ~ ~ J u~ ~I L~ _ 0 ~ 0N ~ 1 \0 ~ ~! ~ 0 ~ C~ Cx (~1 N ~ 02) ~ 0 ~ CN O N L~ (\J \0 ~ O ~ ~ L~ (\J J c~ N J 0 O ~ J ~ O . . . . . . . . . O ~ 0 \0 C O O O L~ O O ~ 0 ~ 0:) 0N 0 0 0 ~ ~ o 0 ~ 0 ~ r~ Z O L~ O J oO ~ N O L~ 1- 6 O ~< N 0 ~ 0 ~ L~ -O 0 ~ ~ N O ~ L~ 0 J O ~ 0 ~ C~ ~0 J ~ ~ N 6 0 0 01-N O ~ O O (> > 0 ~ \0 ~ L~ O O N ~ O . . . . . . . . . . 0~0~0000 0 N ~ O O ~ O ~ O C/) 000000 OOOOJO =0000~0 hJ O O O O J O 000000 ~ O O O O O O _ LUOOOOOOO ~ ~ ....... C ) ~ L - \ O 0N ~) \0 ~) - O ~ L~ O ~ N ~ J ~) tl) Z N 1- 0 _ N ~ 6 N ct) > O ~0 ~ J ~ ~ J L~ J L~  - O O ~ N J ~ N ~ (:> ~ ~O ~ 0 sC) ~ 0 N N N ~) O ~ ~ 03 ~ ~ 0 0 L~ 0 (J) O ~S:) ~ ~ N ~ L~ 0 ~ 0 ~J Z O ~ O N ~) J ~ N Ca ~ O ~ N ~ O ~ 0 0 J ' ' - ~ ~ N co O O C~ CO . . . . . . . . . O O O ~ O O O O ~ N 000000000 OOCOOOCOO 000000000 000000000 030000000 5. ~ O O O L - \ O O O O ~ ~ =00~00~0 u) 0 0 0 0 0 0 0 G O . . . . . . . . . L~ 0 0N ~ :i ~ ~ 0\ L~ 0 N L~ ~ ~ N ~ ~ 0` ~ ~ ~N~ JOL~03t- N N ~ - N L ~~) ~L1J LU Z ~ ~O - ~6 c) a: L~2 ~Z 6 -~ cn Z COI 6-O ~ -~ LL] C I L~ ~ (t) ~ 6 . _ . a) Q E tn - a, c) Q E c o - ct) y C: 9 CO oo a) E o o o ~ o o o ~ o ~ o o o o ~4- O N C 3 O ~a . . . E 0N L~ 1- 4- 0~ - ~ L~ 0:) L ~Q 1 ~D  j · : C) C o 4J C c v, - a) Q E ~n o .o ? C) . Q a, Q ~ E C ~ C: g a' ~ c 4 ~ o a' m . . LL r: o ~n o a) 4 - E c c c

399 z 0 ~ as o~ ~ ~ ~ ~ ~ ~ ~ oooO~)Cu~o~ a:~o0JO -O O J ~ ~ ~ ~ ~) L~ 0` `0 o o ~ r~ o~ ~ as oo ~ oo3voJo:>J~ ~)L~I-O -OCUJ~0~ LS>~ O:)0OO0J" > 0 ~) ~) 0` ~ - 1¢ ~ =) N ~ hJ O ~ J J J ~ ~ J ~ J O ,~ . . . . . . . . . . 00- 0~00000 C) ~n L`J OCvJ OLt~ - JCxJ ~ O - 00 C) O0J`O~ 0~ ~J0`J Z O ~ O `0 t~ 1-r~ J ~ oO (U ~ 0 0~=N ~ J L~C~00~)J -O ~ (~ J 1-~) o:) oC) 0~ 0 N (Y O J o:) ~0 ~ 0:) (U 0~) O C ~ L`> 3 ~ ~ C~J J ~ ao 1- > O O O N J O O (~ ~ ~ ~ . . . . . . . . . . . 00~0000000 - C~ m 11 a, 4J a) . _ . _ . _ . . ~n . _ .m 4J  ~n . _ 4J Q . _ a) C) - 00 LLJ m ~S cnooooooooooo cooooomoooooo O O O O J O O O O O O C) O C O O N O O O O O O mOOOOC~OOOOOu~ cn ~ooooo~ooooo~ OOOOOC7~00000- - ~0000~000000\ ........... ~ ~ LrN ~ r~ ~ ~ ~ cr~ ox~o 0 o~ - 000~)L~0~L~0~1-J~ C: ~J ~ 1-~ ~ ~ ~J ~ ZoD oC)JI-J~0= ~ ~ ~ ON cn os ~o ~N _ ~ O>~N~O-~O O O~=OO`~N O~N-=J=~= ~ O=JN0~00 O OJ~O-~-N~ ZO--OJ= ~O~O=~N-~- LU O O J ~ 0\ 0 0 ........... 0 0 0 ~ 0 0 0 0 0 r~ 0000000000 00000~00000 00000~00000 OOOOO~OOOCO 0000~00000 ~OOOOJOOO=O ~OOOONOOOOO cooooo~oooooo . . . . . . . . . . ~ ox ~ ~ 0 ox cr, as~ 0 00~-J ~J=~J ~J~ =~< oo ~ Z ~ J O - cc <: ~ m ,~ c: z ~r -~ ~hJ Z ~ ~ = ~- -~ ·~ ~ _ ~ ~ ~ ~: Z ZO-=N~ OO-~JO ~-OJ~0<N~O~ _ ~O~N~J=0~-  Q ~O--OJ~- -O-~N=~-N >O=~N~JN~= ~O~JJ=O C) . . . . . . . . . . c: O O ~ O r~ n cn a' 4 - . _ O ~O~=N=~O~= OO=O<)O~-N~l-~L)O ZO~ ~O~I-~L~J~ ~et O U~ 0> 0 N J ~ J ~) 0 c ~-O N ~ ~ ~ ~ ~ L~ O ~C~ O ~ O ~ `0 ~ `0 J ~C) ~ ~ _ ~ O ~ N (O N N ~ ~ ~ J 0\ > 0 O ~-~ J O O N N N ~ . . ........... ~00~00~00000 00 . a' o C) (/) O O O 0 0\ 0 0 0 0 0 0 O O O O ~ O O O O O O 00000~000000 4 ~C) O O O O J O O O O O O a, l~ O O O O ~ O O 0 0 0 LS~ ~OOOO~OOC ~ON ~o0000000000~ 4 ~- LLI==O=L~0000003 ~ Q t/0 CY ~ ~ ~ \01-oC:) N ~1-0 J 1) - O ~ ~-J J J O O ~-O~ N C:a)~J~J~ J000\O <: ZJ ~ O ~ ~ N~ cna) ~ ~ox . - J h] C)~ > a' _ ~ _ O ~ ~O - ~N~ - O - - ) O O ~X) ~) ~ J ~ N ~ ~ 0 o;) LL U) O N oC) J O a:) J 1-~ ~ O . - l~l O<X)~o~JJ - - o:)= Ot~l~NMJOU~cO~O Z O k0 J ~ O N ~ ~ ~ ~ ~ to a) <: 0 ~) ko ~ ~ N ·~ ~ oD ~) oo O O ~ N ~ O O - ~5 ~.. . . . . ~000400000- (~) O) ~ N ~ ~n a) 0 ~ ~o o o o ~ o o o o o o .o ~oooomoooooo ._ 0 0 0 0 ~ 0 0 0 0 0 0  ._ 0 0 0 0 ox 0 0 0 0 0 0 .> ooooo~oooooo ~n C: ~OOOOC~OOOOOO ~. ~oooo~ooooou~ .~= .. ~OOOOOOOOOON Q ~.O ~=N~ D ~-OJOO=~=O = ~=~J~ =~ : Q J -~= ~-NJ =' C C a) C: ~._ ~Q Q C ._ 4- 0 ° a a, ~ C m - _ ~ C 00 ~hJ " ~ 6 =Q = ~ . ~ ~ -~ ~L~ O C ~Z ~ ~ ~ C:-O (/) ~_ ~ ~ ~ ~ ~ ~ ~ ~ Z . ~ a' Q C C a' a) Q - C C C o C 9 . . 00 a' C C o o C a' C Q a) j · : C o C C ~n Q C cn o .o a' L~ . _ ~  - ~,L ~ C : Cm ~C CQ 4- o a' ~S m o c c c ~n a, c

400 ZOO~=OJO-I OO~--~J~I -O-N-~-JI O~ON-~NJ ~O~J~O~- _ o~_ ~U~\O~= >ONJ~J~N~ OJMJO'-J=O ....... ~ . OO-OJ0000- _ In O~-J~O OO-O~ON-~= ZOO ~O-~N~N~ _o=~OO-~=N Led-J-~ O (\J ~O=O=~NJ~J >O-=N~OON-~ . . . . . . . . . . 00~ 0~0000- - Cal m a) E 11 4. Q ._ . _ o 11 a Q a) a) en ._ . _ . _ · . . _ 4- ._ Cal en a) . _ Q  ._ In a) - 00 UJ m In ooooo~ooooo noooomooooo OO=O====OO OOOOONOOOOO m0000~ O000L~ OOOO~OOOON ~ OOOOON0000 - 0000000000 ~ ~ · In ~ LOONY-Cat-PLATO ~ - O~=N-O~J-00 OOJ Len ZN =~- -~J J N O L`J - O-NN~J--~ O=~NN-~- (/) O ~ ~ O _ ~--\0 C\l ~J 0 0 _ 1-~ 0~ - 0~00~0~=- O ~ O 1 - 0 J_O-O O--~=NN~- ON~-OO~ , · . . · . . . . · ~ -O00~00000_ 0000000000 O O O O O O O O O O 0000000000 OCOOOOOOOO OCOOOOOOOO O O O O ~ O O O O O ~ O O O O ~) 2 0 0 0 L~ (t)OOOO~OOOO0J · . . . . . . . . . ~0~0-- 0 ~ ~ 0N L~ OJ~-= O~ N N-~ --~ N C ~L~  Z ~ J ~ O - cY ~ O a, I~J c) Z ~=) -~ U) Z C~-~-O -~ L`J C: I L" 1~ ~) :E ZO--J-~-~ OO=~JO- ~-O~J~_=N~ _ ~ 0mo0~00xoo0~- ~O~-~J -ON~-O~NN~ >O=~-J-_~- ~ON-JJ__JNO · · . . . · · . . a, 0 0 _ 0 r~ 0 0 0 0- a' Q 4 - . _ c O LLJ O L~ 1-~ ~ ~ 1 - 0 L~ ~ c C'OOO`Joo`00~0 - ZO-=O ~ON-- 0000~0 ~-O=~-N J N~ m ~O~J J~O _ <: O ~ ~ 0 CO _ ~ O ~ oo .. >OOJ-- OONOO _· · · · . . . . . . ~OO-OJ0000 00O cn a)~n . - _C 0a) coooooo~ooooo oE cooooo~ooooo a) 0 0 0 0 ~ 0 0 0 0 0 IL c~oooor~ooooo ~OOOON0000 cn ~oooo~oooor~ ~00000 - 00000 C - m0000~0000- Q Q ~ ~ ._ - O-~-O-JN-~- OO-J~= ~0- ZN ~ ~ -~ _ ~n ~ox vo _ CL _ : a' ~ > O _ 11 t~ C _ O ~ ~ O~J~-- O ~ 0~0-L~NOO - <X)~= ~Q ~ O-~N-O~ ._ ~ ~ O ~ ~ ~ O ~ O ~ ~ 0\ C J ~ OJJ-~=J~ ZO-NOO~J=N~ ~O=~J--~-~ ~ ~OO=N~OON~- `_ ~ : C ~000~0000 -  N a, ~ U, O ~ ~C0000000000 .o .mO O O O O O O O O O ._0000000000 ~0000000000 a) ._oooooooooo C~:~0000~00000 · ..=0000~0000 ~ ~ncoooooo~oooo~ ._ ,, C~.......... Q ~._-~-JJO~= · o `~,--OO- JN--N ._ O ~ O ~ \0 L~ 0 oo : Q ~J N N - _ L1 - 3 ~N L~ cn ~. ~n C C ~ .> CQ Q ~ ~. C) O O ~ G) E _ ~ c c m - _ c: J O ~ 0 ~ ~m hJ c: z ~s: ~ . ~ ~ _ ~ cr) L~ O -' ~z ~ ~ ~ <r-0 c: ~ ~ _~< a) Q E c a a) Q 4 - E . _ o c - Q - 00 - a' C~ o o a) Q G) U) j · : - o c c  - a, Q 4~ o .C: . _ ~ _ ~ Q ~ : Q 4- V) C cQ 4~ o :D a m . . LL o ~n o a) 4J . _ c o c - ~n c

401 out of the labor force, as one might expect for students at the middle of the spring term in 1980--56.2 percent of all white male students and 74.2 percent of the black male students did not participate in the labor force. The black-white ratio is only 1.32, just half the ratio of participation rates for black and white nonstudents reported in the last section. Employment-to-population ratios, however, are worse for black students relative to white students than they were for black nonstudents relative to white nonstudents. The fraction of all white male students who had jobs was 38.6 percent; this was 1.97 times the fraction of all black male students who had jobs. The ratio for male nonstudents was only 1.67. The employment-population ratio for white female students was 38.2 percent, 2.06 times the ratio for black female students. However, 2.06 is almost as high as 1.94, the analogous ratio for female nonstudents. Labor Force Participants Of the 4,856 white male students, 2,125 were labor force partici- pants; 2,123 of the 8,239 black students were participants. Even though the original sample contained 2.2 million people, only 505 unemployed black male teen students made it into this final group. Of the 2,125 white student participants, 205 were unemployed. This means that the black unemployment rate among teenage students was 23.8 percent, while the white unemployment rate was 11.8 percent. Note that the ratio of the black rate to the white rate is worse for male students than it was for male nonstudents: 2.02, compared with 1.65 as computed from Table 4, or compared with 1.68, as computed from the LPM regressions on the 0.1 percent sample reported in Table 5(a). Table 8(h) shows that 2,013 of the 4,811 white female students were labor force participants; part (g) of the same table shows that 2,065 of the 8,505 black female students participated in the labor force. Some 176 of the whites and 489 of the blacks were unemployed, which yields female student rates of 8.7 percent and 23.7 percent, respec- tively. The black-white ratio of female student unemployment rates computed from this table is 2.71. Since parts (g) and (h) of Table 4 imply a nonstudent ratio of 2.38, the racial ratio of unemployment rates is worse for female students as well as for male students. Gross Effect of Race on Unemployment In Tables 9(a)-9(d) are linear probability models exactly like those in Table 5, but this time they are estimated for students. "MODEL01" of Table eta) shows that the gross racial differential in male unemployment-population ratios was 1.55 percent and of marginal statistical significance. 3 ° Part (c) of the table shows a slightly 3 ° Like the LPM coefficient, the logistic coefficient is small and barely significant at the 7 percent level. In the sample restricted to labor force participants, the logistic coefficient on race is large and has an asymptotic t-statistic of 5.8. J

402 TABLE 9(a) Linear Probability Models: Civilian, Student Male Teenagers - MO[)Fl: MOF)~-l()1 SSF:,?87.511459 F ilATIO3.28 DFE5660 PROR>F0.(17()~4 DEP VAR: U MSE0.050797 R-SQUARE().()()()G VARIAL31EDEF8TIMA1EERROR -r RAr loPRot~>l T I 1 N l ~ i\Ct- i' 11() . 051483(-).0()3234298 15.9177().()()()1 COLOR1().()15515(-).()()85723()1 1.8()'39().()-7()14 MODEI: MODEL() ~SSE1,?72.lt8 ~ RATIO124.' [)FE5660 PROB>F().()()()1 DEP VAR: M MSF0.224820 R-SQUAREt-).()215 VARIABLEE)rESTIMATEERROR T RAI-IOPROB>ITI INIE-~(,FPl1().38612()C).0(-)68()lt213 5t, . 74 72() . ()(:)() 1 COl()I\1-().2()1257().018(134 -11.1598().(:)(-)()1 MO[~LI: M()~tI()3 SSE1346.9G6 ~ RAl-Io1()().,21 DFE5G6() I'F?~.E().()()(11 DEP VAl<: N t4SE0.23798() R-SQUARE().(-)17lJ VARIAl3IF[)tES1IMA1EE[lf<Ol' r RATIOPR()f3>IIl lN[Ll\(~Ll,!1().562391().~(:)7()(-)()527 8(-).33Gl~(:).(~()~)1 COIOFl1().185-lit2(-).(-)1855Il 1().()1()6().()()()1 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 9(b) Linear Probability Models: Civilian, Student Labor Force Participants, Male Teenagers MODEL: MODEL2 1 SSE 260. 2237()3 F RAT I O 36. 46 DF E 2326 PROB>F t).00()1 DEl' VAR: U MSE 0.111876 R-SQUARE 0. ()154 PAKAM E r E R S TAN DAR D VAR I ABl ~DE EST I MATE ERROR T RAT I O PROB> I T I I Nr~Rcr ~r 1 (1. 1176li7 0.007255864 16.2141 ().00()1 OOLOR 1 (:).148363 0.(:)24572 6.()38() (-).(:)0(-)1 __ __ __ _ ___ _____ ___ _ _____ _ __________________ _ __ _________ ________ _ ___ _ ___ h10[)E 1: MOr)t L,?? SSE 260.223703 F RAT IO 3G.4G DFE 2326 PROB>F ().()(:)01 r)EP VAll: M MSE ().111876 R-SQUARE 0. ()154 PARAMETER STANDARD VAR I AE3I ~[)F FST I MAIE ERROR T RAT I O PROB> I T I INIERCEPI 1 ().882353 0.007255864 121.G055 0.00(-)1 COI ()~< 1 -().148363 ().()24572 -6.038() 0.0()()1 MO[)E l: MOL)[ L 3 1 SSE 259.708068 ~ RAT I 0 1 3. 7() D r E 2324 P RO B> F () . C) 001 DEP VAR: U MSE 0.11175() R-SQUARE (). (:)174 PARAME T E ~STANDARD VAfl I Al3LE- ~DE ~sr I MAI E ERROR T RAT I C PROB> I l- I 1 N r ~ R(,~ i' ~1 () .12713(1 O .00855751 14.8559 O . ()0() 1 SC) 1 -(). (!33641 0.016118 -2.0871 (). ()37() C()l 0~< 1 0.153358 (~.()3789'' 4.()469 (-).()()()1 IN1-RACT 1 0.()()9351145 0.050459 (:).1853 ().853() MOL)El: MO[)LI 32 SSE 259. 708()68 F RAT I O 1 3. -1() DFE 2324 PROB>F 0.0001 DEP VAR: M MSE 0.111750 R-SQUARE 0.0174 PARAME r ER STANDARD VAR IABl E DF ESt I MATE ERROR T RAT I O PROB> I T I  1 N r F KCE ~-r 1 0 . 872870 0.00855751 102.0005 O .00(-) 1 SO 1 0.033G41 0.016118 2.0871 0.0370 COLOR 1 -().153358 ().037895 -4. (:)469 0.00() 1 1 N T RAC T 1 - (-) .00935114 0.050459 - 0.1853 0.8530 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

403 TABLE 9(c) Linear Probability Models: Civilian, Student Female T. t1ODEL: MODEL01 SSE211.346319 F RATIO3.77 DFE5683 PROB>F0.0524 F)EP VAR: U MSE0.037189 R-SQUARE0.0007 VARIABLEDFESTIMATEERROR T RATIOPROB>ITI INTERCEPT10.0365830.002780295 13.15790.0001 COLOR10.0137600.007090884 1.94060.0524 tA()DEL: MODEL02 SSE1256.516 F RATIO155.97 DFE5683 PROB>F0.0001 ()tP VAR: M MSE0.221101 R-SQUARE0.0267 VARIABLEDFESTIMATEERROR T RATIOPROB>ITI INTERCEPT10.3818330.006779186 56.32440.0001 O()LOR1-0.2159290.017290 -12.48890.0001 MODEL: MODEL03 SSE1318.858 F RATIO130.26 DFE5683 PROB>F0.0001 DEP VAR: ~MSE0.232071 R-SQUARE0.0224 VARIABLEDFESTIMATEERROR T RATIOPROB>ITI INTERCEPT10.5815840.006945323 83.73750.0001 (,OLOR10.2021690.017713 11.41330.0001 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 9(d) Linear Probability Models: Civilian, Student Labor Force Participants, Female Teenagers FlODEL: MODEL21 SSE194.368636 F RAT1041.33 DFE2200 PROB>F0.0001 DfP VAR: U MSE0.088349 R-SQUARE0.0184 PARAMETERSTANDARD VARIABLEDFESTIMATEERROR T RATIOPROB>ITI INTERCEPT10.0874320.006624908 13.19740.0001 COLOR10.1453730.022613 6.42870.0001 MODEL: MODEL22 SSE194.368636 F RATIO41.33 DFE2200 PROB>F0.0001 [)~-P VAR: M MSE0.088349 R-SQUARE0.0184 PARAMETER V,~RIABLEDFESTIMATE T RATIO PROB>ITI INTERCEPT 1 0.912568 0.006624908 137.7481 0.0001 C()LOR 1 -0.145373 0.022613 -6.4287 0.0001 _______________________________________________________________________ F1()DEL: MODEL31 SSE194.338377 F RATIO13.88 DFE2198 PROB>F0.0001 OFP VAR: U MSE0.088416 R-SQUARE0.0186 PARAMETERSTANDARD VARIABLE DFESTIMATEERROR T RATIOPROB>ITI INTERCEPT 10.0898330.007846721 11.44850.0001 SO 1-0.008377060.014656 -0.57160.5677 C()LOR 10.1400520.032831 4.26590.0001 1~TRACT 10.0137860.045803 0.30100.7635 MODEL: MODEL32 SSE 194.338377 F RAT10 13.88 DFE 2198 PROB>F 0.0001 DEP VAR: M MSE 0.088416 R-SQUARE 0.0186 PARAMETER STANDARD VARIABLE DF ESTIMATE ERROR T RATIO PROB>ITI INTERCEPT 1 0.910167 0.007846721 115.9933 0.0001  SO 1 0.008377063 0.014656 0.5716 0.5677 (-OLOR 1 -0.140052 0.032831 -4.2659 0.0001 1NTRACT 1 -0.013786 0.045803 -0.3010 0.7635 SOURCE: Bureau of the Census, ''Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

404 more significant racial differential of 1.38 percent for females. However, as was the case for nonstudents, a regression for a sample restricted to labor force participants tells a very different story. Table 9(b) gives as the gross differential between black and white male student unemployment rates a highly significant 14.8 percent. This difference is greater than the white unemployment rate, 11.8 percent. When computed from this regression, the ratio of black to white student unemployment rates is 2.26. When computed from Table 8 the ratio is 2.02. Once again, the difference between the two computa- tions stems from the reduced sampling error in the tenfold larger sample summarized in the earlier table. "MODEL31" shows that, once again, the racial differential is larger outside the South than in the South. Part (d) of Table 9 shows a gross female student racial unemployment differential that is very close to the male differential; it is 14.5 percent and highly significant. "MODEL31" shows something different for women, however. The implied ratios of black to white student unemployment rates are 2.89 in the South and 2.56 outside the South. For men, the ratio was larger outside the South. Effect of Additional Explanatory Variables Tables 10(a)-10 (d) show the same equations for students that were estimated for nonstudents in Table 6. There are two dramatic changes. First, for male students but not for female students, race loses its significant effect entirely, regardless of whether age is included. Second, for male students but not for female students, household income, net of the teenager's earnings, emerges from insignificance as a strong explainer of variation in unemployment. When age is included in the equation, household income is the greatest reducer of chi-square for young men, stronger even than education. SUMMARY This paper presented a very brief review of the economic literature on unemployment, in particular the implications for empirical work on youth unemployment and labor force participation. New structural models for use with microdata were developed. These models may reduce two important sources of bias in estimates of the impact of race on unemployment: simultaneity and ecological correlation. Original empirical work based on 1980 Census microdata shows, using simple, single-equation methods, that the labor force participation decision cannot be ignored in estimating the impact of race on unemployment. For students and nonstudents, male and female, there is no gross racial differential to be explained if unemployment is measured as the ratio of the number of unemployed to the size of the population. However, when samples are restricted to labor force participants, large and significant racial differentials emerge. These racial differentials vary by sex, region, and school enrollment status. A racial difference

405 TABLE 10(a) Conditional Employment Probabilities Using Age as a Regressor, Civilian, Student Labor Force Participants, Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2328 OBSERVATIONS 304 POSITIVES 2024 NEGATIVES 7 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1804.1$ CONVERGENCE OBTAINED IN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.4640D-05. MODEL CHI-SQUARE= 73.14 WITH 10 D.F. D=0.031. -2 LOG L= 1731.04. P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P D INTERCEPT 1.76G78657 1.173324442.27 0.1321 EDUCATN -0.12846921 0.062821744.18 0.0409 0.002 AGE -0.10924862 0.084958741.65 0.1985 0.001 HH1NCOME -0.01426909 0.0042830611.10 0.0009 0.005 FAM -0.11351487 0.777431070.02 0.8839 0.000 D1SAB1L 0.94116955 0.366565856.59 0.0102 0.003 SO -0.38920228 0.161541405.80 0.0160 0.002 CC -0.20908409 0.172255551.47 0.2248 0.001 SOXCOL 0.46637668 0.388383681.44 0.2298 0.001 CCXCOL 0.92305443 0.395068085.46 0.0195 0.002 COLOR 0.24863929 0.392565500.40 0.5265 0.000 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 10(b) Conditional Employment Probabilities, Civilian, Student Labor Force Participants, Male Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2328 OBSERVATIONS 304 POSITIVES 2024 NEGATIVES , O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKEllHOOD FOR MODEL CONVERGENCE OBTAINED IN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.2649D-05. MODEL CHl-SQUARE= 71.46 ~lTH 9 D. F. CONTAINING INTERCEPT ONLY= 1804.18 D=0.030. -2 LOG L= 1-732.72. P=0.0 VARIABLEBETA STD. ERROR CHI-SQUARE P D INTERCEPT 0.40037328 0.50537597 0.63 0.4282 EDUCATE -0.1829C)315 O.C)45268()(3 16.33 O.C)()()1 0.()()7 HHINCOME -O.C)1358426 0.00422821 10.32 0.0013 0.004 FAM -0.19656901 0.77373926 0.()6 0.7995 ().()()()  D1SAB1L 0.92588989 0.36581134 6.41 0.0114 0.()()3 SO -0.388~)3809 ().16150578 5.77 ().0163 0.0(-)2 CC -0.21812()70 C).172C)960() 1.61 ().2(95~) 0.()()1 SOXCOL 0.44365482 0.38765566 1.31 ().252l' 0.()()1 CCXCOL 0.91135256 0.39467425 5.33 C).()2()9 ().C)~)2 COLOR 0.25223778 0.39166524 0.41 0.5196 O.O()C) SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

406 TABLE 10(c) Conditional Employment Probabilities Using Age as a Regressor, Civilian, Student Labor Force Participants, Female Teenagers LOCI s r IC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2202 OBSERVATIONS 1982 U = 0 220 U = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1430.79 MODEL CHI-SQUARE= 61.55 WITH 10 D.F. (SCORE STAT.) P=0.0 . CONVERGENCE OBlAINED IN 6 ITERATIONS.R= 0.147. MAX ABSOLUTE DERIVATIVE=0.1323D-11.-2 LOG L= 1379.95. MODEL CHI-SQUARE= 50.84 WITH 10 D.F.(-2 LOG L.R.) P=0.0000. VARIABLEBETA STD. ERROR CHI-SQUARE P R INTERCEPT 1.75277707 1.39271281 1.58 0.2082 EDUCATN -0.07694844 0.09504671 0.66 0.4182 0.000 AGE -0.18154598 0.11397084 2.54 0.1112 -0.019 HH1NCOME -0.00251025 0.00444995 0.32 0.5727 0.000 FAM 0.20474210 0.45725416 0.20 0.6543 0.000 D1SAB1L 1.04000882 0.47274892 4.84 0.0278 0.045 SO -0.10730316 0.17961398 0.36 0.5502 0.000 CC 0.05299589 0.18402893 0.08 0.7734 0.000 SOXCOL 0.15040473 0.40052623 0.14 0.7073 0.000 CCXCOL -0.44346755 0.40144515 1.22 0.2693 0.000 COLOR 1.40902455 0.36442253 14.95 0.0001 0.095 _ SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample. TABLE 10(d) Conditional Employment Probabilities, Civilian, Student Labor Force Partici pa nts, Female Teenagers LOGISTIC REGRESSION PROCEDURE DEPENDENT VARIABLE: U 2202 OBSERVATIONS 1982 U = 0 220 U = 1 O OBSERVATIONS DELETED DUE TO MISSING VALUES -2 LOG LIKELIHOOD FOR MODEL CONTAINING INTERCEPT ONLY= 1430.79 MODEL CHI-SQUARE= 59.45 WITH 9 D.F. CONVERGENCE OBTAINED lN 6 ITERATIONS. MAX ABSOLUTE DERIVATIVE=0.0 . MODEL CHI-SQUARE= 48.14 WITH 9 D.F. (SCORE STAT.) P=0.0 . R= 0.145. -2 LOG L= 1382.65. (-2 LOG L.R.) P=0.0000. VARIABLEBETA STD. ERROR CHI-SQUARE P INTERCEPT -0.25593762 0.641359740.16 0.6899 EDUCATN -0.18882782 0.0573203710.85 0.0010 -0.079 RH1NCOME -0.00182823 0.004408510.17 0.6784 0.000 FAM 0.17669396 0.455456500.15 0.6981 0.000 D1SAB1L 0.99716482 0.470698124.49 0.0341 0.042 SO -0.12178188 0.179837480.46 0.4983 0.000 CC 0.03780462 0.183842860.04 0.8371 0.000 SOXCOL 0.11184032 0.399101050.08 0.7793 0.000  CCXCOL -0.45539480 0.400617711.29 0.2557 0.000 COLOR 1.41227175 0.3629883715.14 0.0001 0.096 SOURCE: Bureau of the Census, "Public-Use Microdata Sample C" (Washington, D.C.: U.S. Department of Commerce, 1983): blacks, noninmate 1 percent sample; whites, noninmate 0.1 percent subsample.

407 in labor force participation tends in each case to offset almost com- pletely the racial difference in the unemployment rate, so that there are no significant differences in unemployment-population ratios by race. Two very different structural interpretations of these findings are higher reservation wages for blacks and discouraged-worker effects. REFERENCES Abowd, John M., and Mark R. Killingsworth 1984 Do minority/white unemployment differences really exist? Journal of Business and Economic Statistics 2~1~:64-72. . . . Betsey, Charles L. 1978 Differences in unemployment experience between blacks and whites. American Economic Review 68~2~:192-197. Brown, Charles 1981 Estimating the effects of a youth differential on teenagers and adults. Pp. 389-427 in Report of the Minimum Wage Study Commission. Vol. 5. Washington, D.C.: Minimum Wage Study Commission. Bureau of the Census 1983 Census of Population and Housing, 1980: Public-Use Microdata Samples Technical Documentation. Data User Services Division. Washington, D.C.: Bureau of the Census. Cave, George 1981 The Incidence and Duration of Unemployment. Final Report to the U.S. Department of Labor under Grant No. 91-17-78-17. Springfield, Va.: National Technical Information Service. 1983 Job rationing, unemployment, and discouraged workers. Journal of Labor Economics 1~3~:286-307. Clark, Kim B., and Lawrence H. Summers 1982 The dynamics of youth unemployment. In Richard Freeman and David Wise, eds., The Youth Labor Market Problem: Its Nature, Causes, and Consequences. Chicago, Ill.: University of Chicago Press. Conant, James B. 1961 Social dynamite in our large cities. Pp. 26-42 in Social Dynamite: The Report of the Conference on Unemployed, Out-of-School Youth in Urban Areas. Washington, D.C.: National Committee for Children and Youth. Demsetz, Harold 1961 Structural unemployment: a reconsideration of the evidence and the theory. Journal of Law and Economics 4(October):80-92. Fearn, Robert M. 1968 Labor Force and School Participation of Teenagers. Ph.D. dissertation. Department of Economics, University of Chicago. Feldstein, Martin S. 1976 Temporary layoffs in the theory of unemployment. Journal of Political Economy 84~5~:937-957.

408 Fleisher, Belton, and George Rhodes 1976 Unemployment and the labor force participation of married men and women: a simultaneous model. Review of Economics and Statistics 58~4~:398-406. Richard B. Economic determinants of geographic and individual variation in the labor market position of young persons. Pp. 115-154 in Richard Freeman and David Wise, eds., The Youth Labor Market Problem: Its Nature, Causes, and Consequences. Chicago, Ill.: University of Chicago Press. Freeman, Richard B., and David A. Wise 1982 The youth labor market problem: its nature, causes, and consequences. Pp. 1-16 in Richard Freeman and David Wise, eds., The Youth Labor Market problem: Its Nature, Causes, and Consequences. Chicago, Ill.: University of Chicago Press. Gilman, Harry J. 1965 Economic discrimination and unemployment. American Economic Review 55~5, Part 1~:1077-1096. Hanoch, Giora 1976 Hours and weeks in the theory of labor supply. R-1787-HEW. Santa Monica, Calif.: Rand Corporation. Heckman, James J. 1974 Shadow prices, market wages and labor supply. Econometrica 42(4):679-694. 1979 Sample selection bias as a specification error. Econometrica 47(1):153-161. Hey, John D. 1981 Economics in Disequilibrium. Oxford: Martin Robertson. Kalachek, Edward 1969 Determinants of teenage employment. Journal of Human Resources 4~1~:3-21. - Kiefer, Nicholas, and George Neumann 1979 An empirical job-search model, with a test of the constant reservation-wage hypothesis. Journal of Political Economy 87(1):99-107. 1981 Estimation of wage offer distributions and reservation wages. In S. Lippmann and J. McCall, eds., Studies in the Economics of Search. Amsterdam: North-Holland. Killingsworth, Charles Ce 1978 The fall and rise of the idea of structural unemployment. Industrial Relations Research Association Proceedings 31(August):1-13. Levy, Frank 1982 Comment. P. 340 in Richard Freeman and David Wise, eds., The Youth Labor Market Problem: Its Nature, Causes, and Consequences. Chicago, Ill Lucas, Robert E., Jr. University of Chicago Press. 1978 Unemployment policy. American Economic Review 68~2~:353-357.

409 Nerlove, Marc, and S. James Press 1973 Univariate and Multivariate Log-Linear and istic Models. Rand Report R-1306-EDA/NIH. Corporation. Santa Monica, Calif.: Pindyck, Robert S., and Daniel L. Rubinfeld 1981 Econometric Models and Economic Forecasts. York: McGraw-Hill. 2d edition. New Rosen, Richard J. 1984 Regional variations in employment and unemployment during 1970-82. Monthly Labor Review 107(2):38-45. Rosenfeld, Carl 1977 Job search of the unemployed, May 1976. Bureau of Labor Statistics, Special Labor Force Report No. Review 100~11~:39-43. Stephenson, Stanley P., Jr. 1976 The economics of youth job search behavior. Review of Economics and Statistics 58(1):104-111. Weiss, Andrew 1980 Job queues and layoffs in markets with flexible wages. Journal of Political Economy 88~3~:526-538. 210. Monthly Labor