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.

Hispanic Youth in the Labor Market: An Analysis of High School and Beyond Roberto Fernan(lez INTRODUCTION The number of people of Spanish origin in the United States rose from 9.1 million in 1970 to 14.6 million in 1980 (Bureau of the Census, 1982:Table 3.21. In addition to this growth in absolute numbers, the relative share of the population accounted for by Hispanics grew from 4.5 percent in 1970 to 6.4 percent in 1980. Although part of these increases probably reflect changes in Census Bureau enumeration procedures (see Jaffe et al., 1980:311-313 and Appendix A) and an undercount of Hispanics in 1970 (Bureau of the Census, 1979a; U.S. Commission on Civil Rights, 1974), it is clear that Hispanics are a substantial and growing part of the population of the United States. Hispanics tend to be younger than non-Hispanic whites. According to the March 1977 Current Population Survey, the median age of the Spanish-origin population was 22.1 years versus 30.0 for non-Hispanic whites (Bureau of the Census, 1979b:Table C). Since Hispanics are disproportionately young, they are more likely than non-Hispanic whites to suffer the employment problems that youth in general face in the labor market, e.g., low employment and low labor force participation rates. In fact, the data show that regardless of age, rates of employment and labor force participation are lower for Hispanics than for non-Hispanic whites, but not as low as for native Americans or non-Hispanic blacks (U.S. Commission on Civil Rights, 1978:Table 3.1~. However, differences in population-age profiles cannot explain why Hispanic youths are less successful than white majority youths in the labor market. For example, among those aged 16-19 in 1981, Hispanics had an unemployment rate of 24.1 percent and a civilian labor force participation rate of 46.3 percent compared with 17.3 and 59.0 percent, respectively, for whites and 41.5 and 37.4 percent, respectively, for blacks (National Commission for Employment Policy, 1982:Table l). Data from the March 1980 Current Population Survey show that Hispanic youths encounter other barriers in the labor market, as well. Among those l Roberto M. Fernandez is assistant professor in the Department of Sociology, University of Arizona. 410

411 aged 14-19, Hispanics performed worse than non-Hispanic whites and blacks on three out of four indicators of "underemployment," i.e., Hispanic youths are more likely to experience involuntary part-time employment, live in households whose incomes fall below the poverty line, and receive inequitable pay than are non-Hispanic whites and blacks, although blacks are more likely than Hispanics to be inter- mittently employed (U.S. Commission on Civil Rights, 1982:Table 5.4; also see Clogg, 1979; Sullivan, 1979~. While it is clear that Hispanic youths are less successful than non-Hispanic whites in the labor market, the reasons underlying these disadvantages are less obvious. Determining the causes of Hispanic underachievement has important practical implications; the choice of relevant policies to ameliorate those conditions depends on under- standing the factors that lead Hispanics to fare less well than non-Hispanic whites in the labor market. In this paper, I undertake two tasks. First, I document the extent of the employment difficulties of Hispanics compared with non-Hispanic whites and blacks using data from High School and Beyond, a national longitudinal study of high school sophomores and seniors in 1980. Because respondents in this survey were enrolled in school in 1980, labor force statistics derived from the survey will not be directly comparable with statistics based on household surveys of the labor force, e.g., the Current Population Surveys. However, because respondents in High School and Beyond all started in high school, the survey is ideal for studying the transition of youths from school to work. Although past research has found that Hispanic youths fare less well than non-Hispanic white youths on many indicators of labor market success (e.g., wages, family income; see Mayers, 1980), I will focus on two important measures, i.e., labor force participation and unemployment rates. Also, because of the interdependency between youths' leaving school and their employment decisions during the school-to-work transi- tion (see National Commission for Manpower Policy, 1976; Stevenson, 1978b), I discuss the indicators of labor force status by school status, i.e., by high school dropout versus in-school youths for the sophomore cohort, and by out-of-school versus attending postsecondary institution for the senior cohort. My second task is to examine some of the presumed causes of the difficulties of Hispanic youths in the labor market. As with the descriptive analyses, labor force status will be studied in conjunction with school enrollment. Therefore, as a dependent variable, labor force participation has four categories: participating in the labor force and enrolled in school, participating in the labor force and out of school, out of the labor force and enrolled in school, and out of the labor force and out of school. Employment status is treated simi- larly and also has four categories: employed and enrolled in school, unemployed and enrolled, employed and out of school, and unemployed and out of school. Using logistic regression analysis, I predict these labor force and enrollment status indicators with measures of family background, school performance, language, immigration history, and other demographic variables.

412 The remainder of this paper addresses six topics: (1) the extant knowledge on the labor market status of Hispanic youth, (2) the char- acteristics of the High School and Beyond data set and the advantages of using this survey for studying Hispanic youths' achievement, (3) the findings of descriptive analyses of the various subpopulations under study, (4) the findings of causal analyses of labor force and enroll- ment status indicators, (5) the results of empirical analyses of labor force participation and employment, and (6) recommendations for policies to improve the status of Hispanic youths in the labor market. LABOR MARKET STATUS OF HISPANIC YOUTHS As the Hispanic share of the population has increased, the socio- economic achievement of Hispanics has increasingly become the object of policy discussion (see e.g., National Center for Education Statistics, 1980; National Commission for Employment Policy, 1982~. Unfortunately, research on Hispanics in general, and Hispanic youths in particular, has been hampered by a lack of suitable data (see Estrada, 19801. For this reason, information on the labor market status of Hispanic youths is poor relative to that available on non-Hispanic white and black youths (see, e.g., Freeman and Wise, 1982~. Because much research suggests that the decisions young people make on participating in the labor force and continuing in school are interdependent (see B. Duncan, 1965; Edwards, 1976; Ornstein, 1976), it is important to examine the causes of Hispanics' educational difficul- ties when considering the determinants of their underachievement in the labor market. These causes can be divided into two types: general and specific. General factors, such as sex and family socioeconomic status, are potentially important for explaining the school and labor market achievements of everyone in the United States, regardless of their race or ethnicity. Specific factors are characteristics that are particu- larly salient for some minority groups and are expected to affect those groups disproportionately. For Hispanics, specific factors are language skills and immigration history. Distinguishing between the effects of general and specific factors on the labor market achievements of Hispanics is important for policy purposes. For example, if Hispanics' labor market disadvantages are due primarily to their lower levels of family socioeconomic status, then general policies designed to help all poor people would help improve Hispanics' labor market status. However, if specific factors, such as language background, account for a large portion of Hispanics' school or labor market difficulties, then general policies are apt to do little to improve Hispanics' performance in school or in the labor market. In this case, policy instruments, such as bilingual education, may have to be targeted specifically on the Hispanic population to improve Hispanics' labor market achievements.

413 General Factors Recent studies identify Hispanics' low levels of education as one of the most important general factors that explain Hispanic youths' underachievement in the labor market (National Commission for Employment Policy, 1982~. Indeed, there is much evidence that Hispanics experience considerable educational difficulties. At each age level, school enrollment rates for Hispanics lag those for whites (National Center for Education Statistics, 1980:Table 1.08~. Hispanics also have significantly lower rates of high school completion than non-Hispanic whites (National Center for Education Statistics, 1980:Table 1.09~. Among those who remain in school, Hispanics are much more likely to have to repeat a grade as they progress through school than non- Hispanic whites (National Center for Education Statistics, 1980:Table 2.211. Hispanic educational difficulties extend to the postsecondary level, as well. Hispanics are underrepresented in undergraduate, graduate, and professional programs relative to their share of the population (National Center for Education Statistics, 1980:Table 3.01) and underrepresented among the nation's degree recipients (National Center for Education Statistics, 1980:Table 3.21~. There is much research, however, that suggests that these educa- tional difficulties are, in turn, caused by other general factors. In other words, Hispanics' low levels of education are an endogenous cause of their labor market difficulties. Other factors that influence Hispanics' educational attainments may also influence their labor market achievements directly, or indirectly through educational attain- ment. The most important of these factors is family socioeconomic background (Blau and Duncan, 1967; O. Duncan et al., 1972; Jencks et al., 1972~. This is generally interpreted to mean that higher income families, in which parents have high educational and occupational statuses, are more likely to support their children in educational endeavors. Less affluent families may not emphasize education for their children as much because the relative cost of college and higher education relative to the prospective returns on this investment do not justify the expenditure. In addition to the indirect effects of family background on labor market outcomes through education, most studies have also shown direct effects of family background on offsprings' labor market success (e.g., Blau and Duncan, 1967~. Unfortunately, the mechanisms by which these direct effects operate are not well understood in the case of occupa- tional status and earnings. A number of complicated and sometimes crosscutting processes appear to be operating to convert family background into occupational status and earnings (see Jencks et al., 1979:Ch. 3~. However, in the case of youths' labor force participation and employment, it has been shown that children of poorer families are likely to enter the labor force at earlier ages than offspring of wealthier families (Neugarten and Hagestad, 1976), even after the effects of educational attainment are controlled (Hogan, 1981:Ch. 5~. The direct effects of family background on labor force participation and employment have also been documented for high school students (Lewin-Epstein, 1981~.

414 A number of recent studies of various Hispanic subgroups have come to the same conclusion as the studies of the general population: family socioeconomic background is an important determinant of Hispanics' educational achievements (Aspire, 1976; Fligstein and Fernandez, 1982, 1985; Nielsen and Fernandez, 1982) and occupational achievements (see Tienda, 1981; McLaughlin, 1982; Stolzenberg, 1982~. Although there has been very little empirical research on the topic, family background factors have also been cited as important determinants of Hispanic youths' labor market difficulties (National Commission for Employment Policy, 19821. The most important of these background factors is thought to be family income (see, e.g., Aspira, 1976; Briggs et al., 1977~. Hispanics are much poorer than non-Hispanics. In 1977, the median family income of Hispanics was Sll ,421 compared with $16,284 for non-Hispanics (Bureau of the Census, 1979b). Hispanic families also tend to be larger than non-Hispanic families (3.88 persons versus 3.31; see Bureau of the Census, 1979b). Researchers argue that to help ease the family's financial burdens, Hispanic youths are more likely to enter the labor force than non-Hispanics. However, as Hispanic youths become increasingly involved in the world of work, they are correspon- dingly drawn out of school. Hence, they are presented with a self- reinforcing situation wherein they leave school to work, and then their lack of schooling becomes a major obstacle to their success in the labor market. Specific Factors Language problems often head the list of specific factors that may disproportionately affect Hispanics' educational and labor market achievement (U.S. Department of Health, Education and Welfare, 1974; Barrera, 1979; National Commission for Employment Policy, 1982~. For youths entering school from non-English language backgrounds, limited English proficiency can certainly constitute a barrier to effective learning in English-only school systems. Students who cannot understand what is being taught through the medium of the English langu- age are likely to have both psychological and substantive difficulties in their interactions with teachers and in their studies. AS a conse- quence, it is often argued, these students tend to have lower scholastic performance and are more likely to drop out of school (see, e.g., Hirano-Nakanishi and Diaz, 1982; Steinberg et al., 1982a). Survey research in this area tends to support these notions. For example, Lopez (1976) found that U.S.-born Mexican-Americans raised in Spanish- language environments had lower educational attainments than their U.S.-born Mexican-American counterparts raised in English-language environments. To the extent that Hispanics speak only or predominantly Spanish when they complete their schooling, studies suggest negative effects on work-related variables (Lopez, 1976; Chiswick, 1978; veltman, 1981; Garcia, 1983~. Because effective communication is an important component of any production activity, Spanish monolinguals' inability to communicate in English may make them less attractive to employers.

415 In addition, Spanish monolinguals are likely to receive lower wages (see Stolzenberg, 1982; McManus et al., 1983; Tienda, 1983) and to be underemployed and unemployed (Carliner, 1981~. For Spanish-dominant bilinguals, there is some evidence to suggest that accented or non- standard English may result in employers consciously or unconsciously showing bias against Spanish users (Garcia, 1983; Lopez, 1976~. The use of Spanish, or any non-English language, however, may not be intrinsically harmful to bilinguals' educational and work-related achievement. In fact, the effects of using Spanish, controlling for English proficiency, have been subject to debate. One argument empha- sizes the cost of bilingualism. In this view, the coexistence of two lexicons and two syntaxes in the mind of the bilingual represents a drain on a finite amount of mental energy, and less mental energy will be available, for example, for intellectual tasks in school. Another harmful consequence of bilingualism may be that the languages interfere with one another. This process is known as "code switching n (Albert and Obler, 1978~. In this view, Spanish proficiency and use should retard achievement in English-language schools. On the other hand, other studies have found that bilingual proficiency is an asset or does not hinder bilinguals either in school (Peal and Lambert, 1962; Lambert and Tucker, 1972; Cummins, 1976, 1977; Veltman, 1980; Fernandez and Nielsen, 1984) or in the labor market (Lopez, 1976; Tienda, 1981:Ch. 8; Garcia, 1983~. The fact that bilinguals have two codes for every concept may help them to realize that codes are arbitrary. Therefore, bilingualism may serve to stimulate intellectual development for abstract reasoning tasks, which should be expressed in higher scholastic achievement. Regarding the labor market, some studies have suggested that bilingualism is a form of human capital that may yield returns in the labor market (Carliner, 1976; Tienda, 1982~. Therefore, in areas where there is a demand for workers who can communicate in more than one language, bilinguals will be in an advantageous position in the labor market. Also, Lopez (1976) suggests that the knowledge of Spanish may aid bilinguals to find jobs in blue-collar job markets. Results from research on the effects of immigration patterns on achievement have been inconsistent. A substantial body of work docu- ments the fact that despite an initial lack of familiarity with language and customs, immigrants sometimes achieve higher educational and occupational levels than nonimmigrants (Blau and Duncan, 1967~. Chiswick's research (1977, 1978, 1979, 1980a, 1980b, 1982) tends to support these findings, although he shows that an initial adjustment period is needed before immigrants' attainments overtake those of nonimmigrants. Carliner's (1980) analyses support Chiswick's initial adjustment period: recent immigrants generally receive lower wages than second-generation workers, but second-generation workers receive higher wages than do third-generation workers. These findings have been taken to be indicative of a selection process whereby immigrants' high level of motivation manifests itself in higher socioeconomic attainment. Nielsen and Fernandez (1982) speculate that this high level of motivation may be passed on to the immigrants' children, thus explaining why progeny of more recent immigrants perform better in high school.

416 However, when considering Hispanic immigrants specifically, others (e.g., Featherman and Hauser, 1978; Borjas, 1982; Tienda, 1983) find that Hispanic immigrants are at a socioeconomic disadvantage (relative to long-time residents), which these researchers attribute to dif- ficulties of language, cultural adjustment, and transferability of skills. In addition, using census data, Jaffe et al. (1980) have shown that Hispanic immigrants have lower levels of education than other immigrants, which can result, through the general mechanisms described above, in lower educational and occupational achievements for themselves and their children. In addition to the above research, which focuses on the characteris- tics of immigrants that lead them to achieve well or poorly in the United States, a number of researchers have emphasized that the political and economic climate of the United States at the time of immigration may be an important determinant of how well and how quickly immigrants are assimilated. The Cubans are an example here. It has been argued that the particular historical circumstances under which the initial wave of Cuban immigration took place--the climate of general acceptance by the host population, the legal status of Cubans as political rather than economic migrants (Pedraza-Bailey, 1980; Wilson and Portes, 1980), and supportive governmental policies at the time of Cuban settlement (see, Rogg, 1974; Pedraza-Bailey and Sullivan, 1979; Sullivan and Pedraza-Bailey, 1979; Jorge and Moncarz, 1980~-- explain Cubans' relative advantage over other Hispanic subgroups (see, e.g., Borjas, 1982; Nielsen and Fernandez, 1982; Portes, 1982~. A number of researchers have also argued that the fact that Cuban immigrants have largely settled in an ethnic enclave (Miami) made up of previous immigrants (see Wilson and Portes, 1980; Wilson and Martin, 1982) who own about 10 percent of the businesses and employ 50 percent of Cuban males in the area (see Clark, 1977; Portes et al., 1977, 1981) has had beneficial effects on Cubans' socioeconomic achievements (see Portes and Bach, 1980; Portes, 1982~. Finally, there is a substantial literature that suggests that ethnicity, viewed as analytically separable from language and immigra- tion factors, is related to lower achievement among Hispanics. Akin to arguments regarding the disadvantages that blacks face, it is often argued that racial-ethnic prejudice or cultural and socialization differences between majority-minority groups help to explain achievement differentials (see, e.g., Carter and Segura, 1979; Noboa, 1980; for a review, see Duran, 1983~. Although measuring the effects of racial or cultural discrimination in school or in the workplace is extremely difficult, discrimination is often cited as a major reason for Hispanic youths' school and labor market difficulties (see Carter and Segura, 1979; National Commission for Employment Policy, 1982~. In the case of labor market discrimination, inferences have been made on the basis of the different earnings returns to education for whites and Hispanics (National Commission for Employment Policy, 1982~. Such Hispanic-white differentials in returns to education have also been offered as a reason for Hispanic youths' lower levels of schooling: Hispanic youths are less likely to judge each additional year of schooling to be worth the investment, and hence, they are more likely to drop out.

417 DATA AND VARIABLES The High School and Beyond Data Base The data analyzed in this paper are from the first two waves (1980 and 1982) of the National Center for Education Statistics (NCES) study, High School and Beyond, a longitudinal study of U.S. high school sophomores and seniors in 1980. The data were collected for NCES by the National Opinion Research Center at the University of Chicago. The base-year (1980) sample consists of 30,030 sophomores and 28,240 seniors in 1,015 high schools; the overall response rate of 84 percent. Of the respondents, 25,875 sophomores and 10,815 seniors were surveyed again in 1982. Hispanic schools were oversampled in the base year, and respondents in those schools had very high probabilities of being included in the follow-up sample (see Frankel et al., 1981~. Three features of High School and Beyond make it ideal for studying Hispanic youths' labor market achievements. First, because it is a longitudinal study of the sophomore and senior high school classes in 1980, respondents can be tracked through their transition from school to work. In addition to providing information on respondents' labor force status, the study provides detailed data on respondents' educational backgrounds and on how respondents combine their school and labor force activities. Second, because Hispanics were oversampled, the study contains sufficient numbers of Cubans, Puerto Ricans, and Mexican-Americans for separate analyses. This is important because past research has shown that Hispanic subgroups differ in their school and labor market achievement profiles (Newman, 1978; Jaffe et al., 1980; National Center for Education Statistics, 1980; National Commission for Employment Policy, 1982; Nielsen and Fernandez, 1982~. Third, High School and Beyond is rare in that it includes many detailed questions about the linguistic practices of the respondent and his or her family (see Nielsen, 1980:App. B and C, for descriptions and discussions of the language data available from the survey). The study also provides information especially relevant to Hispanics, such as nativity and length of U.S. residence. Definition of Comparison Groups One of the main goals of this paper is to provide statistics showing how Hispanic youths compare with non-Hispanic youths on different measures of employment status. To this end, I have divided both the sophomore and senior samples into groups of Hispanics, non-Hispanic whites, and non-Hispanic blacks. Self-identification was used in the survey to classify respondents' ethnic identity." This was done for both theoretical and practical - iDetailed coding information on the definition of the comparison groups and both the dependent and independent variables can be found in the appendix.

418 reasons. First, the use of self-identification to define ethnic identification is in agreement with the emerging theoretical consensus on what constitutes "ethnic" identity (Barth, 1969~. Second, self- identification of ethnicity is particularly well suited for use in surveys. Smith (1980) has shown that of the various methods of classification (i.e., natal definitions, such as those based on the respondent's country of birth; behavioral definitions based on some objective cultural criterion, such as the use of a language other than English; and subjective criteria involving self-identification by the respondent), self-identification is the most efficient technique for eliciting a positive national-origin identification from respondents in the general population. (Also see Smith, 1983; for research regarding the identification of Mexican-Americans, see Hernandez et al., 1973.) Mauve ~_Am~'~ ~c~ Dependent Variables Two dependent variables are analyzed: labor force participation and unemployment. For both variables, the statistics reported are for those in the civilian labor force; those enlisted in the military are counted as out of the labor force. Because school-leaving and employ- ment decisions are interdependent, I treat labor force and school status as simultaneous events. Therefore, for both sophomores and seniors, the two dependent variables each have four categories. For labor force participation, the four categories for sophomores are participating in the labor force and enrolled in high school; partici- pating in the labor force and not enrolled in high school; out of the labor force and enrolled in school; out of the labor force and not enrolled in school. The variable is defined similarly for seniors with the exception that the relevant school-continuation decision is used, i.e., enrollment in postsecondary education rather than enrolled versus not enrolled in high school. The unemployment variable is defined in analogous fashion for both cohorts, i.e., among those participating in the labor force, respondents were distinguished as employed versus unemployed and enrolled in school versus not enrolled. Independent Variables Corresponding to the discussion in the literature review section, the independent variables are divided into two groups: general and specific. Among the general predictors of labor force and school enrollment status are family socioeconomic background, scholastic performance, demographic variables, and a measure of past labor force involvement.

419 For both sophomores and seniors, I measured family socioeconomic background with a composite variable derived from a number of measures of parental background and family resources. 2 To assess the effects of scholastic performance on school retention and employment propensity, I also included among the general predictors of labor force and school enrollment status two measures of scholastic achievement: self-reported grades and a standardized-test composite. As measures of scholastic achievement, grades and test scores differ in that grades do not vary across schools, while test scores vary both within and between schools. Three demographic variables are also included as general predictors sex, age, and marital status. Respondents' sex is measured by a dummy variable coded 1 = male and 0 = female. Because younger respondents are expected to be less likely to participate in the labor force and more likely to be enrolled in school, I also included a measure of the respondent's age, coded in years, in the models discussed below. Marital status was included as a demographic variable to test the hypothesis that the increased financial responsibilities that accompany marriage are likely to force respondents into the labor force. Finally, to assess the effects of past labor force experience on youths' labor force and enrollment status (see Stevenson, 1978a), I included a dummy variable measured in the base-year survey of past work experience. Consistent with the discussion above, I also included six variables that are likely to affect Hispanics disproportionately as predictors: respondent's, father's, and mother's length of U.S. residence (measured in years); a dummy variable for whether the respondent is bilingual; proficiency in the non-English language; and proficiency in English. (See appendix for coding details.) Regarding the language measures, I considered respondents bilingual if a language other than English was given in response to at least one of three questions: mother tongue of respondent (first language spoken), second mother tongue (other language spoken before schooling), respondent's usual language. These criteria clearly distinguish those students who have never used a language other than English from those who have had at least some natural exposure to another language. Note that this is unlike the criteria used in the Bilingual Education Act (as amended in 1974) to define children of limited English proficiency in that it does not hinge on students' level of English proficiency or nativity (see O'Malley, 1981:Ch. 2~. My definition also excludes respondents with only indirect contact with languages other than English, such as those who studied a language in school as an academic subject. 2Replacing the socioeconomc status composite with measures of father's and mother's education and family income does not change the substantive results reported here. The summary measure was used because of the large numbers of missing values on parental education (15 to 20 percent) and family income (12 to 18 percent).

420 The non-English language proficiency scale used in the survey is based on the student's self-assessed ability to understand, speak, read, and write in the non-English language. 3 These questions are contained in a separate language questionnaire and are only asked of students who indicated some exposure to a non-English language. Finally, English proficiency is measured by performance on a standardized vocabulary test. Note that using vocabulary-test per- formance as an indicator of English proficiency builds in a correlation with the standardized-test composite that is used as a measure of the student's scholastic achievement. Although it would have been preferable to have independent measures of a student's English proficiency and scholastic ability, I chose this specification because the alternative self-reported measure of English proficiency (based on a set of items parallel to the proficiency in other language items) showed very little variance. The fact that the measure of English proficiency is correlated with the composite test measuring scholastic achievement is not of itself disturbing. Indeed, it is difficult to imagine any measure of English proficiency that is uncorrelated with these tests of scholastic achievement since the tests are written in the English language and purport to measure knowledge and skills that are largely taught in the schools through the English language. In addition, my experience in past research (Nielsen and Fernandez, 1982; Fernandez and Nielsen, 1984) and in the preliminary stages of these analyses has shown that the pattern of results is the same if one uses the vocabulary test as a measure of English proficiency and the mathematics test as a measure of scholastic achievement, or the vocabulary test with the composite test (i.e., reading, vocabulary, and mathematics) as a measure of scholastic achievement, as I have done here. DESCRIPTIVE ANALYSES Sophomores For High School and Beyond, sophomores were interviewed in 1980 and two years later, regardless of whether they were still in high school. Table 1 presents high school dropout rates, by sex and population subgroup, for the sophomores. 4 3Self-reported measures of language practices have been found to be highly reliable and valid (see Fishman, 1969; Fishman and Cooper, 1969; Fishman and Terry, 1969~. Fishman and Terry (1969) attribute these qualities to the fact that respondents are forced to perform a global assessment of their linguistic behavior. Many objective measures capture more fragmentary aspects of language usage and have correspondingly lower validity. 4The standard errors reported in the descriptive analyses have been corrected for the effects of sample design.

421 TABLE 1 Dropout Rates, by Sex and Population Subgroup, for Sophomore Cohort . = Male Female Population Standard Sample Standard Sample Subgroup Percent Error Size Percent Error Size All Hispan ics 18 ~ 5 1. 32, 280 18.1 1.32, 210 Mexican American 21.4 1.81,288 20.8 1.81,270 Cuban 1~.6 4.2184 26.5 5.2189 Puerto Rican 24 .0 4 .3258 2 ~ ~ 5 4 .3240 Other Latin American 12 .0 2 .2550 10 .8 2 .2511 Non-Hispanic blacks 20 .3 1.61, 685 14.2 1.31, 961 Non-Hispanic whites ~ 3 .4 0.69, 226 11.6 0 .69, 340 SOURCE: Data from High School and Beyond. The high school dropout rate for Hispanic males overall (18.5 percent) is lower than the rate for blacks (20.3 percent) and higher than the rate for whites (13.4 percent). Consistent with past research on high school noncompletion (National Center for Education Statistics, 1980:Table 2.31) among males, Puerto Ricans have the highest dropout rate (24.0 percent), followed by Mexican-Americans (21.4 percent). "Other Latin Americans" have the lowest dropout rate among males, lower than whites (12.0 versus 13.4 percent), and the rate for Cuban males (14.6 percent) is slightly higher than the rate for whites. Among females, Hispanics overall have the highest dropout rate (18.1 percent, compared with 14.2 percent for blacks and 11.6 percent for whites). Cuban females have the highest dropout rate of any subgroup of either sex, 26.5 percent. The pattern for the remaining Hispanic subgroups is the same as that for males: the rate for Puerto Ricans is highest (21.5 percent), followed by Mexican-Americans (20.8 percent) and other Latin Americans (10.8 percent). The mechanisms underlying these differences in dropout rates are unclear. In part because of problems of data availability, very little empirical research exists on the causes of these different dropout rates. However, the limited research available suggests that Hispanics are likely to drop out in order to work and help support the family (National Council of La Raza, 1980~. At least for males, the dropout statistics in Table 1 are consistent with this hypothesis: the dropout rates for the various subgroups increase as the median family income of the subgroup decreases (National Center for Education Statistics, 1980~. The same pattern holds for females, with the exception of blacks, who drop out less than one would expect, and Cubans, who drop out more than one would expect.

422 Table 2 also lends support to the idea that Hispanic males tend to drop out for financial reasons. Table 2 shows labor force status by school enrollment status for the sophomores. Among out-of-school males, Hispanics overall show a higher degree of labor force attachment than do whites or blacks: 85 percent of Hispanic males were in the labor force compared with 82.5 and 73.1 percent, respectively, for whites and blacks. The relatively poor Mexican-Americans show the highest, and the relatively rich Cubans the lowest, degree of labor force involvement among the out-of-school males. In agreement with past research (Ryscavage and Mellor, 1973; Newman, 1978), the poorest subgroup of all, the Puerto Ricans, show a very low rate of labor force participation. However, this is probably due to their very high rate of military enlistment (see Table 2~. A number of researchers have noted that because Puerto Ricans are heavily concentrated in New York City, which has had a declining economy in recent years, job oppor- tunities for Puerto Ricans have worsened (Newman, 1978; National Council of La Raza, 19801. Enlistment in the military is common among those faced with bleak job prospects. Considering females' labor force participation rates among those who are out of school, Hispanics overall again have a lower rate of participation than either whites or blacks. However, unlike past research on the adult population that has shown that the labor force participation rate of Puerto Rican females is especially low (Ryscavage and Mellor, 1973; Newman, 1978) and declining (Santana-Cooney, 1979; Santana-Cooney and Warren, 1979; National Commission for Employment Policy, 1982), Table 2 shows that among youths, Puerto Rican females have the highest rate of labor force participation, even higher than white females (67.8 versus 66.0 percent). Also contrary to the past research on the adult population that shows that Cuban females have a high rate of labor force participation relative to other Hispanic subgroups (see Ryscavage and Mellor, 1973; Newman, 1978:Table 1; National Commission for Employment Policy, 1982), the data in Table 2 show Cubans have the lowest labor force participation rate among female youths. 5 While out-of-school Hispanics are more likely than out-of-school whites to participate in the labor force, Hispanics are less successful that whites in finding employment. For both sexes, unemployment rates among out-of-school Hispanics are considerably higher than those of out-of-school whites (males: 30 versus 21.8!percent; females: 34.9 versus 26.6 percent), albeit not as high as among out-of-school blacks (36.8 percent for black males and 47.4 percent for black females). This is consistent with past research on the general population (see McKay, 1974; Newman, 1978~. Also, consistent with past research on the adult population (Newman, 1978; National Commission for Employment 5Aside from differences in the age groups studied, the discrepancies between the results in past research and the analyses here are probably due to differences in the target population. Note that none of these sources reports data on out-of-school youths.

423 Policy, 1982), Puerto Rican males have the highest unemployment rate among Hispanic subgroups. However, the employment situation of other Latin Americans who are out of school is significantly better: their unemployment rates for both sexes are relatively low, for males even lower than the unemployment rate for whites. Somewhat of a surprise, out-of-school Cuban females show the highest jobless rate in Table 2, 52.5 percent. The employment situation of out-of-school Puerto Rican females is also relatively poor, albeit not as bad as for Cuban and black females who are out of school. Finally, out-of-school Mexican- American males and females show very similar unemployment rates (32.6 versus 32.3~. Turning now to students, labor force participation rates among males enrolled in school do not vary much among ethnic subgroups (75.5 to 79.5 percent). For female students, the variation in labor force participation rates across ethnic subgroups is considerably more than for males (67.3 to 77.7 percent) but is much less than the ranges for high school dropouts of either sex (males: 70.2 to 90.1 percent; females: 47.4 to 67.8 percent). But while rates of labor force participation do not vary much, chances of employment do. Among male students, Puerto Ricans have the highest unemployment rate of any subgroup (27.5 percent). Only black female students have a higher unemployment rate, 32.6 percent. Among male Hispanics, Mexican-Americans have the lowest unemployment rate (14.8 percent), and among female Hispanics, other Latin Americans have the lowest unemployment rate (16.3 percent). Comparing students to dropouts, no simple pattern emerges for labor force participation rates among males. In some cases, e.g., Mexican- Americans, dropouts have a higher degree of labor force attachment than students (90.1 versus 77.2 percent), but in other cases, such as Cuban males, students have a higher level of labor force involvement than dropouts (75.5 versus 70.2 percent). However, the unemployment statis- tics for males show a clear-cut pattern: once in the labor force, high school dropouts have a more difficult time finding work than youths who remain in school. This pattern could reflect employers' responses to dropouts' relative lack of education. An alternative explanation for this pattern is that high school students and dropouts seek different kinds of jobs. For example, high school students largely seek part- time employment (Lewin-Epstein, 1981), while dropouts are more likely to look for full-time work (Bows, 1983~. Differences in unemployment rates may simply reflect differences in the job markets in which students and dropouts search for work. Youths who are looking for full-time work may be more disadvantaged than youths searching for part-time jobs because those who seek full-time employment are likely to be competing with adult workers who have considerable labor force experience. In contrast, the job market for part-time work is likely to be less competitive. Considering the statistics for females, the pattern is clear across all subgroups: dropouts are less involved in the work force than students. Part of this pattern may be due to a discouraged worker effect. Because female dropouts have relatively poor employment prospects, as evidenced by their very high unemployment rates, females

424 sit o o Q U) o .,' o Pi X U) ~n U] o ~: o o U] :^ Q U] :' cn S~ o E4 ~ O O ,~ O a O O  .= ~ O E~ cn S" 0 ~ · - · - [: C~ o o C) U) `:: ·,' a 0 4~ o C) o ~: ~ c0 C) O · - S~ CJ O · - eD ~ td L ~m  O a, . - . O S~ bO ~ D O C~ a N ~ Ct? U] O C: 4 CQ 4 a) C) a P~ a) CL N · - ~ U] U] O S~ ~ W V? C) S~ a a) N CO S" O c: U]  S:: C) LrN ~L ~O ~ ~(\] ~- 1~)(~)Lf) O ~O ~a0 ~O e ~1 ~OO r~kDO~tt-JJ ····... ~OO ~00 a) OL ~O ~C~ kDO ~-1(NJa ~:= ~r ~J 0JOr~1J ~·e U~ O ~0 00 ~e ~e O ~0o~D 0-0 ~J <X) 0~1 ~- 1 ~~) O ~In CNJJO0~J(\J ~e ~ee J ~a~~) ~(NJ O ~J ~L ~1 e ~OO ~ o00> ~C~00~- 00 a' a) ~J~)( - ~JO~D ~- ,) ~J ~1¢ OOOO03t~ ~ee OOOOO ~O OOOOO0 ~e ~ OOOOO ~O w) ~L~ ~- a ~1 U ~3 ~a: L~  C1C)0~03t ~O~ ·.····~ O0o ~00 ~D ····~ ~0 Jr~00 U~ 3~ ·~ J(\J ~Joc)~ C ~U) ~U~ O ~03 ···~ U ~30 ···.··. a ~0~D L~Lr ~D ~: C) C) .,] u ~a ~U) s ~a) ~: a) C) · ~a) C.) ~ C) ~· - ~· - .,- ~4 , ~· - ~ S ~C: D 3 S ~S: D ;t ~d c: o ~ C.) ~O ·r~1 C) C) C) [3 C) · - C) C) ·r ~¢ ·r~1 J ~· - · - · - ~. - 1 =; ~c: ~I [= ~O ~O r ~C) = ~r ~cO v] C) S: ~L a] ~ ~·r ~C4 ~·r ~I, ~.~] ·~1 :c x ~x ~=: x ~a) ~x a) :5 :5 -) 1 1 c) : ~4 ~1 1 C' ~O ~X C ~O ~ O O ~O O '¢ ~Z; Z C' Z Z

425 o o s C) .,4 a) o a: A_ · - o C) _' He m em  e o it: c: 0 C) · - 0 ~ .~' C) 0 · - A: 0 · - ::5 0 : 0 N E ~1 ~ C/? CO So o c: S V] a a P~ Q N td C/7 U] L O c: C~ CO a a 0a ~1 a: ~c ~0 ~o ~oc _ _ oo ~oo ~oc ·eeeeee  ~0 ~O O00 ~00~ eeeeeee 0 c ~cr~ ~0 ~O u~ao ~=t0 _ __ OU)~1 ~00 eee kDln C~U~O ~L~ e ~ U)~D ~0> a a ~3C ~1 - ··~ ~0 00 cr\O O~ e~_ 0o ···~ ~O ~c ~ ······~ cr ~0c~ CSi~D ~00 ~cx: cn ~c~ __ 00 ~O~ ······~ O 0O ······~ C) C) .,' ~· - DO (1) [: L ~: G) tI5 a) C) 4 ~10 (1) C) c' ~·- c' e ~ o-4 e ~q .S ·~- e ~q S S ~3 ~{ c' e c' ·- c' c' c) Ei c' ~c' c ·- e ~or ~-1 ~ct ~- 4. ~·~- ·~ S: 1 =; t ~1 C ~tM ~  Q. ~O Q. C Q 10 0 ~Q ~c) ~a ~c ~; 4 ~a ·,l ~{ ~s ~a) -1 · - · - ·- ~a) ~ x .O ~) ~=: ~x ~C) a) 5:: := =: (1) ~1 1 ~3 ~t 1 0 C: {:: ~x ~a. 0 s:: c: ~O O ~O O e ~z z e ~z z

426 choose to stay out of the labor force. A second explanation for this pattern is related to the reasons they left school in the first place. Since many females dropped out because they were pregnant or getting married (see Borus, 1983), it is reasonable to expect that many of them chose the rode of homemaker; therefore, they are not counted in traditional definitions of labor force participation. Seniors High School and Beyond also followed up, two years later, on respondents who were seniors in 1980. Table 3 describes the seniors' postsecondary school activities by sex and population subgroup. Hispanics are underrepresented in postsecondary education relative to their share of the population (National Center for Education Statistics, 1980:Table 3.01~. However, Hispanics who have graduated from high school have been found to go on to college at a rate equal to (Peng, 1977; Duran, 1983) or higher (Fligstein and Fernandez, 1982, 1984) than non-Hispanic whites. Peng (1977) speculates that this pattern is due to the success of affirmative action programs. Nielsen (1980), however, offers the intriguing interpretation that this pattern is actually a consequence of the significant barriers to Hispanic achievement in high school. Because high school is a difficult process for Hispanics (evidenced by their very high dropout rates; see Table 2), the "survivors" of the process, he argues, are a more select and highly motivated group than whites who do not encounter the same obstacles in high school. Regardless of which of these interpretations is correct, past research shows that Hispanics compare favorably with other groups in their ability to gain access to higher education once they make it through high school. Olivas (1979), however, thinks that the equivalence of college-going rates is due to the tendency for Hispanics disproportionately to attend junior and two-year community colleges. Olivas (1981) and others (e.g., Duran, 1983) argue that this is because Hispanic high school graduates are relatively poorer than their non-Hispanic counterparts and, thus, are less able to afford four-year colleges. The data reported in Table 3 do not support these past results. Both male and female Hispanic high school graduates are less likely to go on to college than whites. These Hispanic-white differences in rates of postsecondary attendance are mainly due to Hispanic underrepresentation in four-year institutions. Because of the small sample sizes, the standard errors for the Hispanic subgroups are very large, which makes inferences for the Hispanic subgroups difficult. However, the following patterns emerge among the subgroups. The percentage not attending postsecondary school is particularly high for Puerto Rican males (57.8 percent), but is also large for Mexican-Americans and other Latin Americans. Only Cuban males have a higher rate of postsecondary attendance than whites (82.7 versus 62.2 percent).

427 o o C) o ,, V] o Q4 o s~ Q V] o ·_' Q4 o U] Q ~5 o a U]  U] o P~ ~; E~ ~ ~o O o ~ ~ E · C) ~ CC C~ :^ {. O ~ C C) c: ~' ~ ~ u ~ r; U) a' O O Z 1a) ~ e., C': O C~ o C) CO ~ L ~ C [: o ~ ~ ·,1 ~ ~: U) C) ~*: a) bO a' CO' :^ 1 O  a Q ~ ~, ~ tI: U] s" C c: ~ ~ ~: V) a) . ., ~: bO tQ Q) O ~ C a) ~ ~ ~ ~: I t/-J O S~ O o. . - ~ J~) c) ~ S~ bl) ~ :, O U) P~ a) O C~ l ~D LrN O C\J J ~ C\l ~C~U~J ~r ao ~o~ LCNao ·~ r~ ~U) kD ~0~DLr) ~o ~J~ trico~o ~J  ......~ oo ~C ~o~o J~D cr ~D (U LrN U ~o r ~Lr · · ~ u~ ~J a~ O ~tD ~ J (~1 t ~C ~D ~cr~ ~D ~00 J Lr ~J L ~c ~rr) C\J C~ c) ·,1 ~C .,' ., :r: ~ a ~X; cI 0oCO ·~ ~: 1 (d C) . - L E~ ¢:5 C) .,1~ ~ ~ C) O· - ~L L L~ a) Q)~: E :3 O C) ·.- s: Q V] ~ · - C) :~ 1 o ~o ~o  C ~C ~C~ ~oo · ·~ J ~o ~J~rc\J ···~ CM r ~u~ · ···~ coe ~r ~r ~r . oo ~O u~0a~ L ~O ~CM o~ ·~ ~l ~o~ ~o0 c~ a~ ··- JJo C~ ·~ 0J C~ o 0a ~c~ ·..~ ~a ~c~ L~O~D ··~ ~a~ o ~u ~tJ ~ LrN ·~ o C~C\l ~D oo o <~ ·~ o~J ~a~ ·~ r~ r) J C) C) · - C) C) ·0 c: · - ~ s: · - 1 ~C: Q ~ ·-i o ~ u' Q c~ o~ . ~o~ s ~0 a) ~ C) ~ 4 X ~S ~ ~=-- a) ¢ ~:5 ~ ¢ 1 ~1 ,C: >: C' ~O ~ ~ 3 O O Z :~: .

428 Although I cannot resolve the issue here, it is my speculation that the results reported differ from those of past research because of differences in sample design. Unlike the data reported in most other studies, High School and Beyond is a longitudinal study of a grade- based cohort of students, i.e., seniors in 1980.6 Other studies report percentages of high school graduates in household surveys (e.g., the Current Population Surveys; see Duran, 1983:Table 1) who go on to college. Sometimes an age restriction is used to define the survey population, but it is typically a broad range; for example, Duran (1983) uses the population aged 18-340 If, by being poorer than whites, Hispanics are more likely to have discontinuities in their educational careers and therefore to take longer to make the transition to college, 7 college-going rates based on studies of grade cohorts, such as in High School and Beyond, will show Hispanics lagging in their rates of college-going. If it is the case that Hispanics go on to college at rates equal to or higher than whites but that it takes longer for them to do so, studies of broad age cohorts, such as Duran'S (1983), that do not examine the question of whether Hispanics are overage compared with whites will show that Hispanics have reached parity with whites. If the discrepancy between these results based on High School and Beyond and those based on other studies is due to Hispanics' taking longer to get to college, then the discrepancy should diminish as the High School and Beyond cohort ages. Table 4 shows seniors' labor force status by postsecondary school attendance, sex, and population subgroup. Here, too, the small sample sizes make inferences concerning the Hispanic subgroups difficult, and caution should be exercised in interpreting differences among the Hispanic subgroups. Similar to the results for the sophomores, Puerto Rican males have a high rate of military enlistment two years after high school graduation (23.6 percent). Black males also have a high military enlistment rate (19.1 percent). Although much lower in absolute size than the rate for Puerto Rican males, the corresponding rate for Puerto Rican females is also high relative to white females (5.2 versus 2.3 percent). The military enlistment rate among black females is similar to that of white females (2.2 versus 2.3 percent). Considering the labor force participation of males who are not enrolled in school, rates of participation in the labor force are very high (greater than 90 percent) and do not vary much across ethnic subgroup. Overall, Hispanic males participate in the labor force more than either whites or blacks (95.7 compared with 92.9 and 92 percent, respectively). Almost all civilian out-of-school Puerto Rican males are 6Peng (1977) is an exception here. His results are based on the Class of 1972 National Longitudinal Study. 7For evidence that family socioeconomic status is inversely related to school discontinuities in the general population, see Featherman and Carter (1976~. For evidence on socioeconomic status and the timing of educational transitions, see Hogan (1981~.

429 either working or seeking employment (98.8 percent). Mexican-American males also have a very high labor force participation rate, 96.6 percent. Although not as high as the rates for males, the labor force par- t~cipation rates of out-of-school females are fairly high. However, unlike the males, there is considerable variation across ethnic subgroup in the rates for females. Female Hispanics participate in the labor force at a rate that is almost equal to that of white females (83.3 versus 84.4 percent), but is somewhat higher than the rate for black females (78.2 percent). Cuban females are substantially more likely to participate in the labor force than white females: almost 95 percent of out-of-school civilian Cuban females were employed or looking for work, compared with 84.4 percent for white females. Puerto Rican females, who showed the highest labor force participation rate among out-of-school sophomores, had a relatively low rate of participation, i.e., 79.7 percent when followed up two years later. Turning to the unemployment rates for out-of-school males, Hispanics overall have an unemployment rate that is slightly higher than that of whites (18 compared with 14.8 percent), but substantially lower than that of blacks (29.3 percent). Among male Hispanic subgroups, Puerto Ricans have the highest and Cubans the lowest unemployment rates (19.5 and 14.1 percent, respectively). Among out-of-school females, Hispanics' employment prospects are much poorer than those of whites, but not as poor as those of blacks. More than 40 percent of black females are unemployed, compared with 17.8 and 27.7 percent of white and Hispanic females, respectively. Among Hispanic subgroups, Cuban females have the highest rate of unemployment (40.5 percent)--the highest unemployment rate in Table 4. Because of their small sample size, statistics for the Cubans should be interpreted with caution. Looking at those who are enrolled in~postsecondary education, labor force participation rates are very low. Among males, only 44.5 percent of whites, 38.8 percent of blacks, and 52.2 percent of Hispanics are employed or seeking work while attending postsecondary education. For both sexes, Mexican-Americans have the highest rates of labor force participation (males: 59.9 percent; females: 58.6 percent). Puerto R,cans of both sexes show the lowest labor force participation rates among Hispanic subgroups: 39.9 percent for males and 54.2 percent for females. Unemployment rates for youth enrolled in postsecondary education follow the same pattern found for the other populations: the unemploy- ment rate for Hispanics is higher than that for whites and lower than that for blacks. Consistent with the results for other populations, Puerto Ricans show the highest unemployment rate among male Hispanics (27.6 percent). Among female Hispanics, Mexican Americans show the highest rate of unemployment (15.7) percent. Finally, comparison of the labor force status of seniors (Table 4) with that of sophomores (Table 2) reveals a number of interesting patterns. For one, a comparison of high school dropouts with seniors who have not gone on to college shows that the seniors have uniformly higher labor force participation rates and uniformly lower unemployment

430 o ·,, o ~o CO a) U) v o o v cn S~ o v U2 U2 o Q [Q S~ O ~ O U] C) a V O k-~ ·- O ~q O Q =0 d4 o ~1 ~ ~ Q E~ U] :^ ~: o  ~' a) V) o ~: .,1 a o L 4~ o :^ (: s~ a 4~ ~ · - U) ~ · - · - ~ ~: ~ a) 0 C) · -  ·,! C) O · - ~ 4~ J tl, O Pe · - O 00 :~ CL O C~ a) Pe N . - ~ U] tn S" S~ C) c: S~ U) C) S~ a a) ~ N ~ ~ ~ O S:: u, c: C) a' ~ ~l .- rd U] cO s~ ~ s~ ~o o C:  ~ s~ co c) s~ a O)3~1C~3L~ ~C~ Lf ~O - ~L~LfNa ~ ....... ~Jo ~m1 ~N ....... 00~Gi ~l~C~(Nl ~D ~0a0 ~0 ~c0 3 ~cO ~r ~00 ~ ~e ~e (~1:=~03 ~(~)~1 O ~Lr ~0o ·e O ~-.=o~o~\ ~3 ~,~j ~1 ~(NJ~1 ~0a) ~0 U-.O ~O ~L~ 0- ~CO ~r ` C) ( ~OC) 3 0 . . . . . . . UN ~0 0 0 3 ~J -1 = ~(~> G ~O ~0 (1' O LO(\J ~0 (\J 3 ~J 3 ~1 ~N - ·- OO rr) ·- OO 30 L~ U~ OO ~C~ ····- OU~OOO O(\J~J(NJ ····~ O ~O ~C~ C~0- ~0 C~3CS~ ~DJ ~r ~C ·....·~ Jl ~0 r~0 ······~ GU~b~O (\J~J3 ~JJ  U~0-0 ~0~ ~13 ~J ~O ~DJ ~(~J - ~OL ~ ··..- (U ~ C~3coe~CNJc\13 ······~ C<1t~3C ~=)3 ooc~oa ~co c: ~: C) C) ·.- a ~c, ~·- ~co [ ~00 ~L ~: a ~a) v ~a) c' ~ c) ~· - . - cr ~· - ¢ S ~S, 3 L ~Q 3 ~a) ~u ~a, V ~. - c) ~ ·,1 ct ·,' 4 ~· - · - · - cI: ·,1 ~· - · - c: 1 o:; nS ~C 1 ~tt c: Q, ( ~o c a a ~0 Q U] ~ ~L ~L ~ · - · - (d L a) ·r ~·~1 · - · - (:d L (V ~· - X X ~ aD S ~5: ~X D a) ~: ~: X a) ~ :5 ~1 1 ~:: ~::5 ~1 1 c~ ~O ~:: ~>: c: P ~O ~ ~0 0 ~0 0 e ~:Z Z ct Z Z

431 o o ~_ C) C!) :^ L o C) a U] 4~ U] o `:: ·, ,4 o C _% · -  o C) m ~: o a) co ~ . - S~ ~ O C)~ .- ~ C) O · - fd S~ L~ c O 0- · - ~ O ba ~ D O C~ Q N e ·- ~n S~ s°- ~ S~ CO c) S~ a) a' N · - ~ C~ U) S~ ~ O cn c: a  C) S~ a ~JO ~1¢L ~C~ N~J3 ~L~kD ~C) O~0c~t~c~3 0 ·e ~·e ~00 ~O1 ~N a~3a`\ ~(~)~0 kD L~tDt-kD-1 3 ~1· - ~(~.1(\J(\J ~1 O ~(~)~1(~)LC\ 0- ~O 3(~\J ~13 ~J 3~1 ~(\J ~C~ ~a 3 ~OkD(~( - ~1 0 ~00 U~ e ~e J( ~0 ~ L~U~)~ ~t~J(~) 3 00 ~tDa~ 300=) ~C e O ~03 t-O ~( -0~1 ··..··~ ~U~O0 O<)30(~J ~(~ 3C ~3(NJtDt- ~r~ 0~1~3 ··~ ~U~0~ ~CC ··~ ~00 ~ ~D ~r~L[N ~ ·.·... ~00~S)Jl~3 L(>l~Li~U ~3 S: ~ C) C) ·,1 ~CO · - CO C ~dV ti:: L a) C) ~a) c' 4 C) ~· - . - cC , - ,= · - =: 5- ~, D :` ~C D ~cn a' ~`:: C) ~a ~· - -4 ct ·,4 · ~- ~- ~-d ¢ ·, ~J 1 ~: ~C C: ~1 ~; ~ rd O Q ~CL ~O Q C ~c> ~4 ~C) c: 4> ~u ~a .,' .,- ~. - . - . - . - ~. - X X D ~S :I: X ~: X .O Q) =~  a) : ~1 1 a) :5 ~1 1 X ~O ~O ~ ~O O ~O Ct Z: Z =: Z Z

432 rates. This is not surprising given that youths in the senior cohort are older than youths in the sophomore cohort (on average, 19 versus 17) and are high school graduates rather than high school dropouts. But if we consider youths who are in school from both cohorts, the pattern for labor force participation reverses. Students in high school (sophomores) have much higher rates of labor force participation than students in postsecondary schools. However, among those in school and participating in the labor force, the chances of employment are not systematically different for members of the sophomore and senior cohorts. These patterns imply that high school students are much more attached to the labor force than students enrolled in postsecondary schools. There are two possible explanations for these patterns: (1) postsecondary study allows fewer opportunities for labor force involvement, and (2) self-selection is operating so that students who attend postsecondary education are the ones who wish to concentrate on their schooling. These results may be due to the need of Hispanics to participate in the labor force more than whites because Hispanics are poorer and are less able to afford the costs of postsecondary education. But, perhaps because of self-selection, Hispanic high school students participate in the labor force at higher rates than postsecondary students. Comparisons of youth who are out of school from both cohorts suggest that caution should be exercised when interpreting the causal analyses of seniors' labor force status (presented in the next section). Unlike the sophomore cohort, the senior cohort does not include high school dropouts. Thus, the causal analyses that follow are subject to selection bias (Heckman, 1979~. This problem is compounded by the fact that the various population subgroups have markedly different selection rates due to dropping out (see Table 2~. The data for the sophomore cohort are not subject to this problem, because pre-sophomore year attrition rates are small and do not vary for Hispanics and non-Hispanics (Rumberger, 1983:Table 1~. MULTIVARIATE ANALYSES Analysis Strategy In this section, I develop models of labor force participation and employment for whites and Hispanics. The purpose is to test a Note that I do not estimate models for blacks. This is for two reasons. First, the logistic regression analyses presented in this section are estimated by maximum likelihood techniques and are therefore very expensive. Eliminating blacks from consideration has the advantage of simplifying the number of comparisons that must be made in the analysis and cuts computation time by a third. Second, a major focus of this analysis is the assessment of the effects of

433 number of hypotheses derived from the literature concerning the causes of Hispanics' underachievement in the labor market. Specifically, the purpose is to test whether Hispanic-white differences in general background factors, such as family income or scholastic achievement, account for Hispanics' difficulties, or whether specific factors that differentiate Hispanics from the white majority, such as language or recency of migration, explain these difficulties. My strategy is to first specify separate models of labor force participations for white and Hispanic sophomores and seniors. Because of the small numbers of Cubans and Puerto Ricans in the sample, the various Hispanic subgroups have been aggregated and dummy variables have been included to distinguish subgroup membership. Although it would have been preferable to explore subgroup interactions with respect to the models developed here, my preliminary analysis has shown that the numbers of Cubans and Puerto Ricans are very small and therefore likely to yield unreliable estimates. Because there is evidence that decisions about school continuation and labor force participation are interrelated (Duncan, 1965; Edwards, 1976; Ornstein, 1976), I treat labor force participation and school enrollment status as joint dependent variables.~° Therefore, the dependent variable has four categories: in the labor force and in school, in the labor force and out of school, out of the labor force and in school, and out of the labor force and out of school. Three dummy varibles are created for membership in these four categories. They are labeled LFP1, LFP2, and LFP3 and correspond to the first three categories above. The excluded (base) category is out of the labor force and out of school. Each of these three dummy variables is linguistic patterns on labor force and school enrollment status. Although there is evidence that linguistic factors are important in determining black students' school achievement (see Dillard, 1973:Ch. 7; Harber and Bryen, 1976; Labov, 1976), the literature focuses on the use of nonstandard English dialects, i.e., "Black English." Since the language data in High School and Beyond does not contain any information about dialects, but is geared toward the identification of foreign-language users, the language issue for blacks cannot be properly addressed. 9As with the descriptive analyses, those who are enlisted in the military are defined as being out of the labor force. Therefore, the equations presented predict participation in the civilian labor force. Iran alternative here would be to use school enrollment status as a predictor of labor force status. However, if it is true that decisions about school continuation and labor force participation are made jointly, the results of such a specification would suffer from simultaneity bias (Their, 1971:429-432~.

434 predicted by means of logistic regression analysis.ll The coefficients estimated from these models represent the effects of independent variables on the probability (log-odds) of being in a particular labor force-school enrollment status (i.e., LFP1, LFP2, LFP3) as opposed to being in the base category (i.e., out of the labor force and out of school). The next step is to specify models for employment versus unemploy- ment for whites and Hispanics in both cohorts. Parallel to labor force participation, employment and school enrollment are treated as jointly determined variables. The dependent variable has four categories: employed and in school (labeled EMP1), employed and out of school (EMP2), unemployed and in school (EMP3), and unemployed and out of school (the base category). A set of logistic regressions is then run to predict membership in the first three employment-school enrollment statuses (i.e., EMP1, EMP2, and EMP3~. Because employment is defined only for those who participate in the labor force, the estimates derived from the logistic regressions for employment are conditional on participation in the labor force. Results Tables 5 and 6 show the number of cases used in the analysis and the means and standard deviations of the independent variables for labor force participation and employment models for white and Hispanic sophomores and seniors. The data in Table 5 confirm a number of findings of past research (see above). Hispanic youths tend to come from poorer families than white youths. Hispanics show a shorter length of U.S. residence on all three length-of-residence variables. Hispanics are also much more likely to be bilingual and, among bilinguals, to report a greater facility with the non-English language (i.e., Spanish) than whites. Hispanics also do poorly in school relative to whites: they have lower grades and score less well on standardized tests. These patterns are the same for both the sophomore and senior cohorts. These results are also similar for those respondents who are in the labor force (Table 61. Labor Force Participation of Sophomores Table 7 presents the results of the logistic regression analyses for white and Hispanic sophomores' labor force participation-school enrollment. For both Hispanics and whites, only one sex effect because the dependent variables are dichotomous, ordinary least squares regressions would produce estimates that are not minimum variance unbiased estimates because of heteroskedasticity. A logit specification solves this problem (see Theil, 1971:631-633~.

435 TABLE 5 Means and Standard Deviations for Variables in Labor Force Participation-School Enrollment Analysis Sophomore s Sen ions White Hispanic White Hispanic Mean S. D. Mean S. D. Mean S. D. Mean S. D. LFP1·73·44.65.48.25.43.28.45 LFP2.04.20.07.26.36.48.42.49 LFP3.21.41.26.44.30.46.21.41 Sex (1 =male).47.50.48.50.46.50.44.50 Age15.47.5915.62.7417.43.5717.57.69 Mexican American--.53.50 Cuban-~.10.30-_.12.33 Puerto Rican--.10.30--.08.27 Other Latin Amer lean-_.27.44__.23.42 Bilingual (1 =yes).04.20.51.50__.63.48 Proficiency in non-Engli sh Language.06.351.041.12.07.361.341.15 Voca bu lary Test Score50.629.3645.529.1555.189.4749.239.76 Composi be Test Score53. C28.4246.497.9452.908.2346.607.94 G race Po int Average2.84.782.53.763.02.702.80.69 Marital Status.03.17.04.20.11.31.13.34 Worked During Base Year.45.50.34.47.64.48.58.49 Length of Residence15.311.2014.862.4317.271.2616.413.13 Father's Length of Residence41.685.2035.6911.7743.864.9235.9512.93 Mothe r ' s Leng th of Residence38.884.6933.1810.9440.854.9333.5012.10 Soc ioeconomic Stat us.12.69- .36.72.04.71- .50.74 ( N=3, 389 ~( N=2,211 ~( N=4, 340) ( N= 1,623) surfaces in Table 7: males are significantly more likely than females to be in the labor force and enrolled in school. Considering the other demographic variables, the results for age are as expected: older youths are more involved in the labor force and less involved in school. Among Hispanics, older youths are more likely to be in the labor force and out of school (see equation for LFP2) and are less likely to be in school and out of the labor force (equation LFP3) than out of the labor force and out of school. The results for whites follow a similar pattern, although one effect is statistically significant, i.e., the effect of age in the

436 TABLE 6 Means and Standard Deviations for Variables in Employment Status-School Enrollment Analysis Sophomores Sen iors Wh i te Hi span ic White Hispanic Mean S. D. Mean S. D. Mean S . D. Mean S . D. EMPl EMP2 EMP3 Sex ( 1 =male ) Age Mexican Arner loan Cuban Puerto Rican Other Lat. in Amer lean Bilingual ( 1 =yes ) Proficiency in non -English Language Vocabulary Te st Score Composite Test Score Grade Point Average Mar ital Status Worked During Base Year Leng th of Rest dence Father's Length o f Residence Mother's Length o f He si dence Soc ioeconor~c Status .80 .04 . 1 4 .48 15 .48 .04 .40.72 .20.07 .35.19 .50.50 .5915.65 .52 .09 -. 1 0 .20 .06 .33 50.55 9.23 52 .90 8 .20 2.82 .02 .77 .14  .50 .50 15.34 1.13 41.75 5.02 38.93 4.56 .67 .45 .36 .26 .49 .39 .05 .50 .45 .75 17 42 .50 .29 .30 .45 .50 1.02 .04 .07 8.98 53.89 46.317.84 2.61 .03 .40.49 14 .88 36 .061 1 .64 33 . 371 0 .90 _.35.71 .752.92 .17.13 .70 2.4217.29 .48 .50 .22 .50 .56 .20 .35 .35 .48 .05 .43 17.57 .60 . 1 2 .07 .22 .62 1.31 9 .44 48 .22 51.508.13 .70  .34 .46 1. 1 2 43.864.91 40.845.00 -.07.66 .48 .50 .22 .50 .70 .50 .33 .26 .41 .49 1.14 9.34 45 .857 .58 .68 .33 .62.49 16.443.07 36 . 1 4 33 .6512 12.93 . 1 5 _ .55 .71 (N=2,613) (N=1,580) (N-2,664) (N=1, 127) equation for LFP1: younger white youths are more likely to be in the labor force and in school than in any of the other categories. The independent effects of marital status on labor force participa- tion and school enrollment are similar for whites and Hispanics, even for this very young group. For both whites and Hispanics, having been married decreases the chances of being in the labor force and in school and increases the odds of being in the labor force and out of school as opposed to being in either of the out-of-the-labor-force categories (i.e., the base category and LFP3~. For both whites and Hispanics, married people find it particularly difficult to participate in both school and the labor force.

437 ^= c~oor<~ _ ~ ~ ~ co~ ~o ~, ~ ~ 433 oo~ o o o ~o o o o om O ~4 C/) · · · e· ·e~ U] _] a~ :' ~ * N * P ~ CS~ ~) ~0 ~N ~O ~ ~l ~N ~U~ ~I O N ~ OO OOOO N ~J O OO O IS)- U] i i i i i 1 1 1i~ =: _ ~O ~ N N1 ~00NN m) ~l ~N ~ L~ _ C) ~ ~ N ~N N ~OO ~ ~O O OO ~ O: . - N CO . . . .. .... . . . . .. . O ~ ~1 ~ t ~:= * * *N aD ~ ~ ~ ~N ~N ~1( ~t ~N OO N O I _ ~N ~0 L~ O ~N ~OO ~ ~O OO O O~ 8 1 1 i i 1 i~ U] _ 1 ~u~ ~ ~ ~.o ~ ~ ~ o ~o o o ~oo o ~o o oo o ~ .= ~a~ ~U~ ~C * * * ~# ~ .~. S) ~D O L~ N ~Oa ~ N N O ~O ~ N O ~OOO ~ ~O OO 0 - 1 ~_` %_ LrN ~a ~ co ~L ~ O V ~C/) O O I I I N N O O O ~O O O O O J ^ OV J 0 * * ~J * 1 S~1 D `0 N =t ~ 0 l ~O ~ ~ ~- o I I I I o ~ ~ O c ~ O ~O O O ~ I _ V _ ~O J O J N N J ~D ~' O N ~D J ~ N I I I J ~o o ~ ~o N O O ~ l~ (,, 1 ~o- 1 ,, .. ,, . . 1 Q ~J O. .,1 ~N J I ~0 J O O O t ~N N O O ~ 0 C l l ~o o ~ ~N ~o o o ~o o o o o a) ov ~cn . .' ~ ~. . . . . . . . ~ £v ~o S a) ~ ~* * _ ~N ~1 ~N 0 0 J N ~N l~ ~0 H 0 I l o ~ I , o ~o o o o ~o o o o ~- ~cl 1 1 ~ m = ~S S ~ ~C: ~ a, ~ a) = . C D ·- S :~ 00 t0 C) 00 C) C) o O t ~ ~ ~ ~ ~ ~ ~ ~O O  c: =.- ~ c :~.- ~ ~ ~ ~ ~ ~ ~ ~ e s t- ~0) ~c-) ·,1 C C) ~ ~ ~ O O ~: ~ ~ a) c' ~ ~ ~ ~ 0 ·- Vl ·- ~ ~ ~ ~ C ~0 bo t~ c-) ~ C~ ~ a) c/~ :3 ~ ~ c: · - ·- ~Q C :- ~ ~ ~ ~ ~ ~ C ~ ~ ~ ~ ~ o ~ ~ o ~ ~ ~ ~ ~ o ~ ~ ~ {J) H~ ·-~ ~ ttS ,-~ 0) ~:~ - ~ - ~ ~ ~ C ~: _ ~O S~ bO ~ C) ~ bO ~ 4~ U] ~ ~ ~ ~ ~ S ~ ~ ~ S~ ~ ~ ~ ra ·- ~l ~- c J~ a~ c: 11 ·- ~ ~ ,~ V) O ~ a) a) - ) (L) ~ ~ Vt O O O ~ ~ _1 m _ ~ ~ a) ~ ·- ~ ~ 0 ~ ~ ~ Q a, ~ ~ ·- ~ m ~ ~ s ~ s ~ .,_ X J) D 0) S ¢ ~ ~ O ~ .~ c) E~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O - 4 ~a) bO ~ ~ ~ ·- ~ O O S~ ,~ O ~ ~ o 0 o 0 o ~ c: ~ o m ~:> ~c ~ 3 ~cn c: ~

438 The scholastic achievement variables show significant effects for both whites and Hispanics. For Hispanics, the higher the base-year grade-point average, the greater the probability of being in school and out of the labor force and the lower the chances of being in the labor force and out of school (see equations for LFP2 and LFP3~. This same pattern surfaces for whites as well, but the t-test for the coefficient in the equation for LFP1 fails significance. Performance on the battery of standardized tests is not related to labor force participation or school enrollment once the other predictors in the model are controlled. The equation for Hispanics predicting LFP2 is an exception: better performance on the test battery lowers the chances of being out of school and in the labor force. The lack of significant effects for the composite test score suggests that between-school variation in scholastic achievement is largely irrelevant to dropout and labor force decisions. 2 The significant effects of grade-point average, which only vary within schools, strongly suggest that the effects of scholastic achievement on dropout and labor force participation decisions are highly contextual. It is only students' scholastic achievement relative to others in their school context that affects their decisions to leave school and/or participate in the labor force. The last of the general variables, i.e., previous work experience, has similar effects on labor force participation and school enrollment for Hispanics and whites. Those respondents who worked at the time of the base-year survey are less likely to be exclusively in school (see the equations for LFP2), although the effect for Hispanics fails to be significant. Previous work experience also increases the chances that both whites and Hispanics combine school and labor force activities (LFP1) and decreases their chances of being out of school and in the labor force (LFP31. Therefore, unlike previous studies that find that high school students who work suffer significant costs in terms of their schooling (Steinberg et al., 1982b), these data show no tendency for either Hispanics or whites to be pulled out of school and into the labor force by virtue of having worked during their sophomore year. Considering the effects of specific factors on youths' labor force participation and school enrollment, none of the length-of-U.S.- residence variables (i.e., mother's, father's, respondent's) signifi- cantly distinguishes among the four categories of the dependent variable. ~ ,~ residence in the equation for LFP1 for Hispanics: respondents whose father have been in the United States longer are more likely to be in the labor force and in school. The only exception is the coefficient for father's length of Recall that test performance varies both between and within schools, while grade-point average only varies within school. Because grade-point average is controlled in these models, test score performance largely taps the effects of between-school variation in scholastic achievement.

439 In terms of the effects of the language variables, exposure of Hispanics to Spanish during their upbringing does not significantly predict school continuation or labor force participation. Although compared with Hispanics relatively few whites had been exposed to another language (see Table 7), exposure of whites to a non-English language raises the probability of their being in the labor force and out of school. Contrary to expectations, none of the language variables signifi- cantly distinguishes among the four categories of the dependent variable for Hispanics. Why the effect of non-English-language background appears for whites but not for Hispanics is unclear. Last among the language variables, the effects of the measure of English-language proficiency (vocabulary test score) on labor force participation and school continuation are nil for both Hispanics and whites. This is most likely because the main effects of English- language proficiency for these youth are probably through scholastic achievement (see Nielsen and Fernandez, 1982), which has been controlled in these models. Finally, the dummy variables for Hispanic subgroup show only one effect. After the other variables in the model are controlled, Cubans are more likely to combine school and labor force activities and are less likely to be in the labor force and out of school than any of th other Hispanic subgroups. The lack of significant effects for the dummy variables for Hispanic subgroup implies that the other variables in the model have explained the subgroup variation in school continuation and labor force participation. Most important among the variables that have been found to account for differences in achievement among Hispanic subgroups is family socioeconomic background. For example, the relative affluence of the Cubans (see Jorge and Moncanz, 1980) is often cited as a major reason for Cubans' greater success in school and the labor market (see Nielsen and Fernandez, 1982. 3 However, other variables also explain the dependent variables, and consequently, differences among Hispanic subgroups in labor force participation and school enrollment are the same ones that are important for whites, i.e., scholastic achievement, previous work experience, and marital status. According to these results, the processes by which Hispanics and whites decide to stay in or leave school and participate or not participate in the labor force are very similar. The "specific" variables that I hypothesized would be necessary to explain Hispanics' underachievement have proven to be insignificant. denote that the latest wave of Cuban immigrants, the Mariel refugees, are not as affluent as early waves (see Bach, 1980~. However, these data do not contain any of these refugees because the High School and Beyond sample was drawn prior to the Mariel boat lift. e

440 Employment of Sophomores Table 8 shows the coefficients of models predicting employment and school enrollment as joint dependent variables for sophomores. Similar to the results for labor force participation, sex does not significantly distinguish among the categories of the dependent variable. Considering the other demographic variables, age is a significant predictor in two equations, i.e., EMP1 and EMP2 for Hispanics. Older Hispanics are less likely to be employed and in school and more likely to be employed and out of school than younger Hispanics. Similar to the pattern for labor force participation, marital status is a strong predictor of employment and school continuation for both whites and Hispanics, independent of the other variables in the model. Being married increases the chances that the respondent is employed and out of school and lowers the probability of being employed and in school for both whites and Hispanics. These results imply that both whites and Hispanics are more likely to be unemployed and out of school or unemployed and in school than being in school and employed. Apparently, employment and schooling~are an either-or proposition for those whites and Hispanics who are married. Looking at family socioeconomic background, socioeconomic status is not a significant predictor for either whites or Hispanics. The fact that the effects of family socioeconomic background are weaker for employment than for labor force participation for whites is not sur- prising. Family socioeconomic background may make it more-or-less desirable to seek employment, but actually securing a job involves convincing an employer that one is worth hiring. Especially in the youth labor market, family background is unlikely to be an important market signal to employers (see Spence, 1974~.~ 4 Although the low-wage, low-skill, high-turnover structure of the youth job market (see Osterman, 1980; Borus, 1983) is likely to make employers' hiring decisions less dependent on productivity-related criteria, employers are probably more likely to pay attention to the effects of past work experience and the characteristics measured by the second set of general predictors, i.e., scholastic achievement. Similar to the pattern of results for labor force participation, past work experience increases the chances of being in the two employed categories (i.e., EMP1 and EMP2) and lowers the probability of being unemployed and in school (EMPTY. This pattern is similar for both lobsterman (1980) shows data to support the argument that parents are crucial in helping many youths get started in the job market by providing youths with networks of personal contacts that help them find jobs. The effects of such job contacts on youths' probability of employment is certain to be positive, but this process is probably only marginally related to these family background factors. Such network variables may account for the significance of mother's and father's presence in the home in increasing youths' labor force participation and employment.

441 C l D1~ o ~ ~o ~o :r ~O ~cD ~O vo ~ ~a ~O ~O O ~ -1 0 0 0 0 ~ ~mo .~1 C~ ~* ~ D ~ t~ ~ (M ~ ~O ~O ~1 O O O ~o 0 o o o ~O O O O O i 1 1 1 1 1 i 1 1 1 i 1~. aD O ~1 ~N N l~ ~1¢ ~N t~ U) ~ ~ ~ ~ N ~J (NJ O O ~ ~O O O J ~| C ~1 l l N ~.,, ~,,, * 0- ~ lr, a) D L~ t:~ J U ~t ~O ~ax t ~~) O O - 1 0 ~ ~ m= ~O 0 - 0 ~O O 0 0= 1 _' l l l ~ t ~t ~D o ~_ D cO ~J ~J ~ ~D ~D ~ ~O O ~O O O ~O 'O O O O ~ O S ~ C) * ~* * * ~o ~ rot 0 Cl) ~ ~ C) ~J 0 ~co o ~0 ~0 0 ~) 1 l O ~ C~' C<l O O O O O J ~D ~) o O O O oo- U] . i ~1 i 1 i i i ::' ~_% t') ~D ~0 0 ~a~ ~D 0 N N CJ~ ~ 4.J I ~O r- ~1 ~0 0 0 (\1 0 0 0 0 0 0\ 1 (IJ ~ ~* ,= o O ~ ( - 0 L ~O Lt~ k0 ~O ~ t~ ~ 1 ~| - - ~O O O N ~O O O O V ~CxJ N t ~<\J ~J ~ ~0 - 1 ~ 0 `0 rn ~N ~ ~) I I I ~tD O O ~ ~J ~J O O ~ 0 {J)1 3 ~ 1 ~; Q * * * * * a-:) .~1 ~J U ~1 ~~) ~) ~J ~J C ~J ~1 0 cO I ~1 ~ J I I I ~ll ~O O CJ~ r<) (M C~J O O ~--' i i 1 i ~i i~ ~1 _ ~U ~oc~ ~ 0 0 tz] O ~ 1 1 ~~) ~O O O ~O O O O O c~ 1 ~ 1 ~''' -^1 a) ~u~ ~J ~o ~Oo ~O *~ * * H o ~ I I I o o o o ~ 0 ~o o o o U] i i i i i i O aJ ~. ~Q o  V O -~1 ~1 s s 1 ~ cn ~ ,- C ~:5 bO 00 ~ bO a) C) O U~ a) ~_ ~[: u ~Q) ~ C ~C C ~ C- ~O 0 v CL ~a) ~ C: ~ .- s" s~ ~ ~ ·- ~a) a) ~ ~ a) ~s _ a) s ~- c~ ·- ~c~ ,- a) ~ 0 0 c: ~ s~ ~ ~ ~ ~ ~ O ., ~Vl ~ · - ~ ~· - ~ ~ ~ ~ C b~ ~ S~ c' a) C) ·- c' ~ ~ ~ ~ ~ ~ · - C ~ s:: c ~ ~ c) ~ u, ~ ~ ~ m m-m o bO c~ o Q, a' ~ v' ~ o m ~ ~ r- ~ ~ (~ .-~ (1) ~ - a~ - a~ c) ~ c ~_ ·w ~o ~ bl) :^ C) I bl) :3 4~ ~ ~ S~ (d ~ a) S ·-t L ~: ~ =; IV ~ ~ ·- I J p. - ~ ~ ~ ~ C I' · - ~ ~ D ~ O ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O ~ ~ ~ rn ~ L a' ~ . - ~ ~ O ~ ~ ~ ~ ~ ~ ~ ~ ~ m ~ ~ s ~ s ~ ~ ~ ~ I 2 - . x a' D a' S ~¢ ~1 ~ O C ,~ CD ~ e E~ ~ =: L S~ s: ~ ~ O ~ O C) U) C: bO .~1 a) bO ~ ~ ~ ~ ~ O O L t~l O ~ td O O O O E~ m cn ~ ~ ~ O m ~ ~ c: C~ ~ ~ _: ~. ~ co c~ _~

442 whites and Hispanics, although the coefficient in the EMP2 equation is not significant for Hispanics. Here, too, there is no evidence of work experience drawing students out of school. General scholastic achievement, as measured by performance on the test battery, is unrelated to the dependent variables for whites. For Hispanics, better performance on the tests raises the probability of being employed and in school. The pattern for the test-score coeffici- ent in the other two equations implies that better students are more likely to be in school, but neither of these effects is significant. However, two of the three coefficients for grades are significant for both whites and Hispanics. Higher grades increase the probability of being employed and in school and decrease the chances of being employed and out of school. The fact that grade-point average is a significant predictor of employment suggests that employment choices are also made within the context of school. But unlike the case with labor force participation, wherein students choose whether to look for work, employment choices also reflect employers' choices among competitors looking for work. Because of the highly local nature of the youth job market (see Borus, 1983), especially for younger youths (see Osterman, 1980), it is possible that employers' hiring decisions are also made with reference to the same school context that students refer to when making their labor force participation decisions. Therefore, while better school performance increases students' school attachment and lowers their probability of labor force participation (see Table 7), employers try to choose the best students from among those who do choose to par- ticipate in the labor force--if not for their skills, then simply for their better discipline (for a similar argument regarding education and discipline, see Bowles and Gintis, 19771. In terms of specific variables, none of those measuring language patterns or immigration history significantly distinguishes the four cells of the dependent variable. The only exception to this pattern is the effect of non-English-language background in the equation for EMP2 for whites. Contrary to my predictions, Hispanics' special circum- stances play no role in explaining their school continuation or employment. If these results are to be trusted, this would imply that employers do not find these specific characteristics relevant criteria on which to base their hiring decisions. Finally, unlike the results for labor force participation, none of the subgroups is significantly different in its employment behavior. Apparently, the advantages that Cubans have in the transition into the labor force do not appear in employment, once the other variables in the model have been controlled. Labor Force Participation of Seniors Table 9 shows the results of the logistic regressions predicting labor force participation for members of the senior cohort. As mentioned above, the main differences between the senior and sophomore cohorts are that the seniors are, on average, two years older than the

443 ~_ ~ ~ ~ ~ ~ O ~ ~o ~oo ~oooo ~ o ~Lr~ U] J 3 D J (~1 3 0 ~J ~(~J ~- 1 ~J 1 ~J ~J 0D 0> O ~ 0 3 ~O ~O O ~ ~O O O ( - ~0 CJ) 1 ~1 1 1 i ~) ~_' ~ a~ ~ ~- ~ oo ~ cr c~ _ C) W O O ~ ~. O O O O O O O O O l~ .- -1 tQ · · · · · · e O Ud ~J S ~a~ J u~ * * * JO * * ] ~ ~ 0 u ~0 ~0 ~0 ~ cr~ 0 O O ~ ~O ~O O U) ~O O O mm 8 1 1 i 1 i 1 1 i im O ~D ~ l _ ~D O ~ t` ~J ~O ~) ~L) ~) ~ a~ (- O ~W O ~O O ~ ~O O O O O CO .= J cr~ ~0 Ql * .H * ~* ~* 1 .~1 D J CO ~ ~U ~ ~- ,: ~ .= U~ ~ t) ~ C\l O C~J ~O ~O O ~ ~J O O O ~ 0N S~ ~_% ~_ J 0O 0 (NJ ~ ~J J ~ 1~ t') a) ~) w O O I ~ I N N O O O ~O O O O O ~) So J ~ [4 N S ~S, ~ ~Lr ~0` N a) ~ 0 ~I O ~ ~I I I O O O O l~ a ~N O O O t~ O -~ 1 ~1 O ~ ~ oo O U ~ ~ tD ~ ~1 O O I I I N ~J O O O O O O O O 0 - 1 U] a) P" c/] · · 1 1 1 · · · · . · . . . . . . ~ ,Q ~I . ~ I ~1 ·~ - ~m1 a ~a ~1 ~L ~a~ ~c\J N ~ J ~) O O I I I O LrN O O ~ ~O O O ~ U~ i i i 1 i i i i im a) ~J a ~ ~J J ~ U~ ~ 15 W O 0 1 1 1 ~O O o 0 0 0 o 0 0 ~ ~· · 1 1 1 · · ·. . . . . . . . . Ql ~ J O N I I I O J OO ~ ~ N ~O O ~ · =~ CJ) · . 1 1 1 · . . . . . . . . . .~ O C) U] 'S O P4 3 Q~,cn ~CV] ~s s Co ~ -~.1 "a ~O? ~ a) ~ a>= ·- ~ 3 ~0 ~0 O 00 C) C) O a) (d ~a ~0 a) ~) ss ~ C ~ 1 0 O o. ~a ~ c: - L L ~ ·- S ~c~ aD C~ ~ ~ E S ' ~c: ·- c c' ~ ~ ~ o o c ~ ~ ~c' J ~ J ~ o ·- Vl - . ·u ~ ~· - ~ ~ ,- ~ ~ ~ ~ ~ ~ ~ ~ · - ~ c ~ ~ ~ ~ · - · - C ~E ~: ~ c~ ~ u~ O C ~ ~ cO ~ ct) O ~ ~ >4 0 a~ v~ oD v~ cn o ~ 0  , J ·- ~ a) ·- ~ :~ ~ ·- ~ ~ ~ ~ ~ a) c~ c r. e>e >. c) I bc :~ ~ a) ~ ~ ~ ~ a~ s ·- ~ ~ s~ a: g? ~ ·- ~ ~c: ~ s~ 0) ~ 11 · - c: ~ D o O oD ~ a) ~ a' U, ~ a' a) ~ ~ ~ _3 m - - ~ ~ a) E ·- ~ q~ o ~ ~ a) ~ a) ~ :> ·- Y ra bO Q) S ~ S ~ · - a, - 4 X ~ D a) S m: ~ '~ o ~ ~ c) E~ ~ ~ ~ ~ ~ ~ m bl) ~ :5 ~ ·- S~ O O ~ ~ O ~ ~ O O O O E~ ~cn ~Z ~ ~ 0 m ~C: C: ~: 3 J ~CO C~) J

444 sophomores (compare Tables 7 and 8) and the seniors are all high school graduates; this means that for seniors, school enrollment refers to participation in postsecondary education at any time in the two years after the base-year survey. Sex differences in labor force-school enrollment status are stronger for the seniors than the sophomores. Among Hispanics, one effect of sex appears: males are less likely to be in the labor force and in school (LFP1) than females. Two sex differences surface as significant predictors for whites. White males are more likely than white females to be out of the labor force and in school (LFP3) and less likely to be in school and in the labor force. Considering the effects of the other demographic variables, marital status is a strong predictor of labor force participation and post- secondary school enrollment for both whites and Hispanics. Whites who are married are (in order) most likely to be: (1) in the labor force and out of school (LFP2~; (2) out of the labor force and out of school (the base category); (3) in school and in the labor force (LFP1~; or (4) out of the labor force and in school (LFP3~. For Hispanics, being married clearly affects postsecondary school attendance: Hispanics are most likely to be in the two out-of-school categories (LEP2 and the base category) and least likely to be in the two in-school categories (LFP1 and LFP3). Similar to the patterns for sophomores' labor force participation, family socioeconomic background is a significant predictor of both white and Hispanic seniors' labor force participation. For both whites and Hispanics, respondents from more affluent families are most likely to be attending postsecondary school and not be in the labor force (LFP3) and are least likely to be in the labor force and out of school (LFP2~. Finally, whites from more affluent family backgrounds have higher chances of combining school and labor force participation (LFP1), although this is not as likely an outcome as LFP3. The results of the scholastic-achievement variables for seniors, in contrast to the results for sophomores, reveal significant effects of both test scores and grades. Test scores are significant here probably because colleges routinely use performance on standardized tests (such as the Scholastic Aptitude Test), which are likely to be correlated with the test battery used in High School and Beyond,~5 as screening devices. It is not surprising, then, that better performance on the standardized tests increases the probability of being exclusively enrolled in postsecondary education for whites, although the cor- responding effect for Hispanics fails to be significant (see equations for LFP3~. Better test performance also serves to lower the probability of respondents' being out of school and in the labor force for both whites and Hispanics (see the equations for LFP2~. But, whereas Hispanics who score well on the standardized tests are more likely to combine school and labor force activity, whites are not (see equations for LFP1). lathe test battery for High School and Beyond was developed by Educational Testing Service, Princeton, N.J.

445 Independent of performance on the tests, grades are a strong predictor of school and labor force activities for both whites and Hispanics. Here, too, the effect is probably due to colleges' using grades as an admittance criterion. For both whites and Hispanics, higher grades increase the chances of being in either of the in-school categories (LFP1 and LFP3) and decrease the chances of being in either of the in-labor-force categories. Last among the general predictors of achievement, previous work experience has strong effects in the expected directions for both Hispanics and whites. For both groups, previous work experience increases the likelihood of being in the labor force regardless of whether respondents are in school. Considering the effects of the specific variables, among the language variables, proficiency in English (as measured by the vocabulary test) is unrelated to either postsecondary attendance or labor force participation for whites. However, English proficiency does distinguish among some of the categories of the dependent variable for Hispanics. Greater English proficiency lowers the chances of being in the labor force and in school (LFP1) but increases the probability of being in school and out of the labor force (LFP3) for Hispanics. On the other hand, proficiency in a non-English language shows some effects for whites, but not for Hispanics. Among whites, better non- English language proficiency increases the chances of combining postsecondary education and labor force participation (LFP1) and decreases the probability of being out of school and in the labor force (LFP2~. The length-of-residence variables indicate only two significant effects. Hispanics whose mothers are long-time residents of the United States are less likely to be in school and in the labor force (LFP1~. Among whites, respondents who are long-time residents of the United States are more likely to combine labor force participation and postsecondary education (LFP1~. Finally, unlike the pattern in the analyses for the sophomores, the dummy variable for the Puerto Rican subgroup indicates that they are significantly more likely than other Hispanics to be out of the labor force and in school. Seniors' Employment Table 10 presents the results of the models of employment and postsecondary enrollment for seniors. As noted above, these estimates are for respondents who are in the labor force and who are high school graduates. Examining the effects of the demographic variables indicates that there is only one effect of sex (on EMP3 for Hispanics). Among Hispanics, males are significantly more likely to be unemployed and in school than females. There are two significant effects of age, i.e., predicting EMP1 for white and predicting EMP3 for Hispanics. For whites, older respondents are less likely to be employed while in school. Among Hispanics, older

446 C ^ ~o,= ~ ~ C~ C ~ N ~) ~) U ~(~) O O~I L ~oO O ~' .,, ~1 =1 * * D ~C) a) ~ c() CJ ~O 1¢00 ~O L~ ~D O ~1 ~)3 ~J' O OL~DOO O ~1 ~ ~ ~ ~ I ~ C) _ ~O ~ ~ N N ~O O ~ ~O O O O ~ a) .= C\J C/) ~a, ~ .- * ~ * ** ~m~ ~ O ~OJ ~= ~ - 0 1 r~l | I l l O N · ~ - ~J ~oO N ~O O O O ~ ~ -' O 1 ~53 `_ oo ~ ~o ~0 ~u ~ Lr) a) ~ ~ 0 ~1 ~O ~ ~ N N ~OO ~ ~O O O O ~ N O ~ U] * * * * * ~ oO 1 ~ r ~0 3 ~ 3 ~N ~CnJ r~ ~ 1 U] ~O ( ~(M O 0 3 ~-1 0 0 0 ~ 3 ~ U) '_ _ ~c~ ~o ~c ~ 0 a ~C7i ~c~ ~ ~ c 1 1 ~1 ~1 o. ~ 1 1 ~° ° ~ ~° ° ° ° ~ ~ 1 ~I~ O * * -*L~ L ~O ~N~ ~1 ~N ~ t~ ~I O ~I I I ~N O OO ~N ~O O N i i 1 1 i 0 ~1 1 U] ~O ~ ~ I I ~ O O O O O O O O O ~ | .~ | ~I ~I. . I I . . . . . . . . . . . . I O ~ I I I O ~ O O U~ 0 0 ~O O U~ ~I ~; · I I I · ; 1; 1 · · ; · ; 1 m~ ~ ~0 ~ O ~kD ~ ~ ~ c 1 ~°. ~1 1 ~1 ~ o 0 0 ~,o 0 0 0 0 ~ 1 ~O H ~N ~J ~N 3 0 ~L) t ~N ~ `£) 0 O N I I I ~a ~O O 0~ 0 ~O O t~ 0 O ii i i i L~ O ~ ~ c 1 ~ ~1 ~s ~ ~ ~ ~ =1 ~ ~·-] C: :5 ~b~ c) bC C) C) O tl) ~^ C tr1 a) Q) -) ~ ~C r: C C:~- O O O ~=.- ~ ~`-- L ~ ~-- ~ ~ ~ ~ ~ e s ·- c c' ~ ~ ~ 0 0 ~ ~ ~c' ~ ~ ~ ~ 0 ·- Vl ·- ~ ~ ~ ^ C 00 bD S~ ~ Q) C) ·- a) co ~ ~ C~ c: ·- ·- ~c~ => E~ ~ ~ ~ ~ ~ ~ ~ ~ ~ O ~ a 0 ~ ~ ~ ~ ~ o~ t ~H I'_] · - 3 Q) ·-Cz] ~ ~ ·- a' ~ ~ Q) ~ - 0) - 0) CJ C  ~ V ~O S~ bO ~ C) I ~0 :~ ~ V] ~ L. (~ ~ V) .C ·- L. ~ ~. ~ ~ td ·~1 1_ ~ ~ ~ 1l ·- S:: ~ D e~ O a~ a) a) ~ Q) ~ 4 ~ a~ Q~ a) O ~> rrl~ ~ au ~ ·~1 ~ ~ O rd ~ () Q a) ~ ~ · - .~: ~ ~0 a) ~ ~ C ~ · - ~ I ¢ ·-X Q) D ~ ~C C ~ ~ S~ ~ ~ Co) E~ ~ ~ ~ ~ ~ o ~ ~ o O O O E~ T cn =: ~ ~ 0 m ~c: c ~r: 3 ~L ~: ~C/) C~) ~

447 respondents are more likely to be unemployed while attending postsecondary education. As has been the case in all the analyses, marital status emerges as an important predictor of seniors' employment and school enrollment. For both whites and Hispanics, married respondents are less likely to be employed and in school (EMP1) or unemployed and in school (EMP3~. However, this latter effect is insignificant for Hispanics. For both Hispanics and whites, married respondents are more likely to be employed and out of school (EMP2~. Turning to the effects of family socioeconomic background, the effects are similar to those found for labor force participation. Among whites, respondents from more affluent family backgrounds are more likely to be in the two in-school categories of the dependent variable (EMP1 and EMP3) and less likely to be in the two out-of-school categories (EMP2 and the base category). The pattern is the same for Hispanics, but the effect of family background in the equation for EMP3 is not statistically significant. Considering the effects of the scholastic achievement measures, the pattern for both whites and Hispanics is familiar. Higher grade-point averages increase the probability of being employed and in school (EMP1) and lower the chances of being employed and out of school (EMPTY. Better performance on the test battery has similar effects. As discussed above, these patterns are probably due to college selection criteria. The effects of the final general variable considered--previous work experience--are also the same as those found in the other analyses. For both whites and Hispanics, respondents who worked during the base year are more likely to be employed and in school (EMP1) and less likely to be unemployed and in school (EMPTY. Examining the specific variables, there is some evidence of langu- age effects among whites, but not among Hispanics. Among whites, greater facility in a non-English language significantly increases the chances of being employed and in school (EMP1) and lowers the chances of being employed and out of school (EMPTY. Finally, English pro- ficiency, as measured by performance on the vocabulary test, increases the probability of postsecondary school enrollment and employment (EMP1) . Among the variables measuring the length of U.S. residence, only one effect appears for Hispanics: respondents whose mothers are long-time residents in the United States are less likely to be employed and enrolled in school (EMP11. Among whites, two effects surface, i.e., respondents who are long-time U.S. residents are more likely than recent immigrants to be employed and in school (EMP1) and less likely to be employed and out of school (EMEND. Finally, none of the dummy variables for the Hispanic subgroups is significantly related to the dependent variable. This implies that the subgroup differentials in unemployment rates found in Table 4 have been explained by the model.

448 SUMMARY, CONCLUSION, AND POLICY RECOMMENDATIONS The descriptive analyses in this paper have shown that Hispanics fare worse, overall, than whites, but not as poorly as blacks, in the schools and in the labor market. Hispanic youths drop out of high school at a higher rate than white youths and a lower rate than black youths. Similarly, the unemployment rate for Hispanic youths is higher than the rate for white youths and lower than the rate for black youths. These statistics for the overall Hispanic population mask considerable heterogeneity among the various Hispanic subgroups. Specifically, Cubans and other Latin Americans fare relatively well when compared with whites, but Puerto Ricans and Mexican-Americans fare relatively poorly. Puerto Rican youths have particularly severe employment problems and often have unemployment rates that are as high as or higher than the rates for black youths. The descriptive analyses also show that Hispanic-white disparities in labor force participation and unemployment are more severe among high school dropouts than among students in school. These differ- entials are even smaller for the population of high school graduates. The multivariate analyses that attempt to explain labor force participation, unemployment, and school enrollment for whites and Hispanics show a number of patterns. For both whites and Hispanics in the sophomore and senior cohorts, family socioeconomic background is consistently related to labor force participation and school enrollment; it is related to employment for seniors, but not for sophomores. With a few exceptions, the specific factors of language and family- immigration history are not consistently related to school and labor market achievements for either Hispanics or whites. The two most important determinants of labor force participation, employment, and school continuation for both white and Hispanic youths are scholastic achievement and previous employment experience. For both white and Hispanic sophomores, grade-point average is a consistent predictor of these school and labor market variables. For seniors, both grades and performance on standardized tests are related to the outcome variables for both whites and Hispanics. Previous work experi- ence is also strongly related to the dependent variables for both white and Hispanic youths. In conclusion, it appears that the root of Hispanic youths' labor market problems lies in their education. These results would suggest that policy efforts should be directed toward solving the problem of Hispanic underachievement in the schools. However, the positive independent effects of previous work experience also suggest that youth employment programs are likely to have beneficial results for Hispanic youths. Therefore, a two-pronged approach--through the schools and in the labor market--is likely to be most fruitful in tackling Hispanic youth employment problems.

449 APPENDI X CODING INFO~TION Respondents are classified as Hispanic in this paper on the basis of their answer to the following question from the High School and Beyond follow-up questionnaire: "What is your origin or descent? (If more than one, please mark below the one you consider the most impor- tant part of your background)." Under the general heading of "Hispanic or Spanish" were grouped four possible answers: (1) Mexican, Mexican- American, Chicano; (2) Cuban, Cubano; (3) Puerto Rican, Puertoriqueno, or Boricua; and (4) Other Latin American, Latino, Hispanic, or Spanish descent. For simplicity, these have been labeled Mexican-American, Cuban, Puerto Rican, and other Latin American. Respondents are considered white if their response is something other than Hispanic to the national-origin question and "white" to the question "What is your race?" Respondents are defined as black in a similar fashion. The terms "white" and "black" as used in this paper, then, refer to whites and blacks not of Hispanic origin. Hispanics were not differentiated further on the basis of race, because the distinction between concepts of race and ethnicity is blurred in the case of Hispanics. Many of the respondents answered "Other" to the race question, implying that they view their group as a distinct "race" (Nielsen and Fernandez, 1982:Table 1.3~. Regarding the measurement of the dependent variables (labor force participation, employment, and school enrollment status), respondents' labor force status is classified on the basis of their responses to the following questions. Sophomores were asked two items in the follow-up survey: (1) "Did you do any work for pay last week, not counting work around the house?" and (2) "Whether or not you already have a job, were you looking for a job last week?" Response categories of "Yes" and "No" were offered for both questions. Respondents' military enlistment (see Table 3) was determined from this question on the dropout survey: "What were you doing the first week of February 1982?" Among the answers offered was "On active duty in the Armed Forces (or service academy)." Youths who chose this option, regardless of their responses on the labor force status questions, are counted as being enlisted in the military. For the civilian population (i.e., those who did not choose the "On active duty in the military" option), respondents are defined as employed if they answered "Yes" to question (1) above. Civilian respondents are classified as unemployed if they answered "No" to question (1) and "Yes" to question (2~. Civilians who answered "No" to both questions are defined as being out of the labor force. Finally, school enrollment status for sophomores is based on whether the respondent was part of the dropout or the in-school follow-up sample. Parallel to the sophomores, senior cohort respondents who chose the "On active duty in the Armed Forces (or service academy)" option of the question "What were you doing the first week of 1982?" are treated as being enlisted in the military (see Table 5), regardless of their choosing other employment- or school-related options. The employment

450 related options are (1) "Working for pay at a full-time or part-time job," (2) "With a job but on temporary layoff from work or waiting to report to work," and (3) "Looking for work." Civilians are classified as employed if they chose the first option, unemployed if they chose the second or third option, and out of the labor force if they did not choose any of these options. The school-related options were (1) "Taking academic courses at a two- or four-year college" and (2) "Taking vocational or technical courses at any kind of school or college (for example, vocational, trade, business, or other career training school)." Civilian respondents are classified as enrolled in postsecondary education if they chose either of the school-related options, regardless of whether they chose any of the employment-related options. The type of postsecondary school that respondents were enrolled in (see Table 4) was not determined by the above school-related item. Rather, respondents were asked to provide the names and addresses of the postsecondary schools that they had attended since leaving high school. Those names and addresses were then matched with data on the characteristics of postsecondary educational institutions (the 1982-1983 Institutional Characteristics Survey of HEGIS, Higher Education General Information Survey, collected by the National Center for Education Statistics). These data were used to group respondents into the four types of postsecondary school enrollment. Note that the data on type of postsecondary enrollment refer to the school that respondents were enrolled in at the time of the follow-up survey (February 1982) or, if not enrolled at that time, the last postsecondary school they were enrolled in. Regarding the measurement of family socioeconomic status, the variable is a linear composite derived from measures of father's occupation, father's and mother's education, family income, and a set of questions that ask whether the respondent's family receives a daily newspaper; whether the family possesses an encyclopedia or other reference books, typewriter, automatic dishwasher, two or more cars or trucks, more than 50 books, or a pocket calculator; and whether the respondent has his or her own room. Coding on this variable is based on a linearly weighted combination of the above family background measures, where the weights are derived from the non-missing data. If a case has missing data on any of these background variables, the composite is computed from the non-missing data for that case (see Jones et al., 1983:62~. Grades are measured by the question, "Which of the following best describes your grades so far in high school?" Eight response categories were offered from "Mostly A" (a numerical average of 90 to 100) to "Mostly below D" (below 60~. The variable was recoded on a four-point scale so that "Mostly A" is assigned "4," "About half A and half B" is coded "3.5," and so on, down to "Mostly below D," which is coded ".5." The standardized test scores used in these analyses are a composite of reading, vocabulary, and mathematics tests administered during the base-year survey [see Heyns and Hilton (1982) for a detailed discussion of the High School and Beyond cognitive tests]. For both the sophomore and senior cohorts, each individual test was standardized within cohort

451 to have a mean of 50 and a standard deviation of 10. For sophomores, the composite was computed by taking the mean of the non-missing test scores. This procedure was slightly modified for seniors because they were administered two vocabulary tests. Items from the two vocabulary tests were combined before the vocabulary test was standardidized (see Jones et al., 1983:Section 6.9~. The test composite was then computed by taking the mean of the standardized non-missing reading, vocabulary, and mathematics test scores. Regarding the demographic variables (i.e., age, sex, and marital status), age and sex were measured by base-year items. However, because marital status was not measured directly in the base-year survey for seniors, a question from the follow-up survey was used: "What was your marital status the first week of February 1982?" Responses were recoded so that 1 = ever married (i.e., married, divorced, separated, widowed) and 0 = never married. Because sophomores were not asked their marital status directly in either the base-year or follow-up surveys, the following question from the follow-up survey was used to distinguish respondents who had been married from those who had not been married. Respondents were presented a question worded "At what age do you expect to ... ," which was completed with a number of items, including "Get Married?" Among the response categories for this question is "Have already done this." Respondents who chose this response to the "Get Married" item were coded in parallel fashion to the seniors, i.e., 1 = ever married, versus 0 = never married for those who did not choose this response. Both sophomores and seniors were asked, "Did you do any work for pay last week, not counting work around the house?" Responses of "yes" and "no" were offered and are coded here as one and zero, respectively. Regarding parent's length of U.S. residence, students were asked in the base-year survey how much of their mother's and father's lives have been spent in the United States. Each variable had five response cate- gories: (1) about 1-5 years; (2) about 6-10 years; (3) about 11-20 years; (4) more than 20 years, but not all; and (5) all or almost all. Categories (1) through (3) were recoded to the midpoint (3, 8, and 15.5 years, respectively). Categories (4) and (5) presented more of a prob- lem because they implicitly refer to the parent's age, for which High School and Beyond does not have a measure. The values for these two categories were imputed by using the modal age of mother's childbearing (25) and adding the student's modal age (15 for sophomores and 17 for seniors) and assigning that to the fifth ("All or almost all") cate- gory. Therefore, the value imputed for sophomores is 40 and for seniors, 42. The midpoint of the fourth category then became defined as 29 years for sophomores and 31 years for seniors. This procedure was repeated for father's length of residence, but three years were added to account for a typical three-year difference in age between husbands and wives. Thus, the fourth and fifth categories for father's length of residence were recoded to 43 and 30.5, respectively, for sophomores, and 45 and 32.5 for seniors. Students were also asked to report how much of their lives had been spent in the United States. The response categories were (1) about 1-5 years; (2) about 6-10 years; (3) more than 10 years, but not all; and

452 (4) all or almost all. Since available data included the student's age, all the categories were well defined and recoded as follows: (1) 3 years; (2) 8 years; (3) (10 + student's age)/2; and (4) student's age. If the student's age was not available, it was imputed for use in the student length-of-residence variable as the modal age--for sophomores 15 and for seniors 17. This was done for only a few cases. Language questions were administered through a separate question- naire to all respondents (i.e., not just Hispanics) who passed a filter of five questions that asked about the respondent's mother tongue and languages presently spoken at home. Those students who reported a language other than English in response to one of the five questions regarding language background were asked to choose on a four-point scale how well they understood, spoke, read, and wrote the non-English language. The response categories are "Not at All," "Not Very Well," "Pretty Well," and "Very Well" and were coded from zero to four. Exploratory factor analysis of the survey's pretest data showed that the four items clearly load on one factor, with each of the indicators contributing equally (see Fernandez, 1980~. The composite index was formed by taking the mean of the four items. Note that the coding is positive, ranging from a low of zero (indicating no proficiency in the other language) to a high of three (indicating high proficiency). Those students who did not pass the language background filter (i.e., were monolinguals) were assigned a zero on the scale for proficiency in non-English language. When combined with the dummy variable for language background, this coding has the effect of creating a spline for the proficiency-in-other-language scale. English proficiency is measured by the student's performance on the base-year standardized vocabulary test. To simplify across-cohort comparisons, the scores used are based on the subset of test items that were identical in the sophomore and senior test batteries. The test is standardized to a mean of 50 and a standard deviation of 10.

453 REFERENCES AND BIBLIOGRAPHY Albert, Martin L., and Loraine K. Obler 1978 The Bilingual Brain: Neuropsychological and Neurolinguistic Aspects of Bilingualism. New York: Academic Press. Ashford, J., and R. Snowden 1970 Multivariate probit analysis. Biometrics 26:535-546. Aspira 1976 Social Factors in Educational Attainment among Puerto Ricans in U.S. Metropolitan Areas, 1970. Puerto Ricans and Education Report No. 1. New York: Aspira of America. Bach, Robert L. 1980 The new Cuban immigrants: their background and prospects. Monthly Labor Review 103(October):39-46. Barrera, Mario 1979 Race and Class in the Southwest. South Bend, Tnd.: - University of Notre Dame Press. Barth, Fredrick, ed. 1969 Ethnic Groups and Boundaries. Boston, Mass.: Little, Brown. Blau, Peter M., and Otis Dudley Duncan 1967 The American Occupational Structure. New York: John Wiley & Sons. Borjas, George J. 1982 The earnings of male Hispanic immigrants in the United States. Industrial and Labor Relations Review 35, no. 3 - (April):343-353. Borus, Michael E., ed. 1983 Tomorrow's Workers. Lexington, Mass.: D.C. Heath. Borus, Michael E., J.E. Crowley, R.W. Rumberger, R. Santos, and David Shapiro 1980 Pathways to the Future: A Longitudinal Study of Young Americans. Youth Knowledge Development Report 2.7. Washington, D.C.: U.S. Department of Labor. Bowles, Samuel, and Herbert Gintis 1977 Schooling in Capitalist America. New York: Basic Books. Briggs, Vernon M., Walter Fogel, and Fred Schmidt 1977 The Chicano Worker. Austin: University of Texas Press. Bureau of the Census 1979a Coverage of the Hispanic population of the United States in the 1970 census: a methodological analysis. Current Population Reports. Series P. 23, no. 82. Washington, D.C Bureau of the Census. 1979b Persons of Spanish origin in the United States: March 1978. Current Population Reports. Series P. 20, no. 339. Washington, D.C.: Bureau of the Census. 1982 Population profile of the United States: 1981. Current Population Reports. Series P. 20, no. 374. Washington, D.C.: Bureau of the Census. Carliner, Geoffrey 1976 Returns to education for blacks, anglos and five Spanish groups. Journal of Human Resources 11:172-184. _ · -

454 1980 Wages, earnings, and hours of first, second, and third generation American males. Economic Inquiry 18:87-102. 1981 Wage differences by language group and the market for language skills in Canada. Journal of Human Resources 16:384-399. Carter, Thomas P., and Robert D. Segura 1979 Mexican-Americans in School: A Decade of Change. New York: College Entrance Examination Board. Chiswick, Barry 1977 Sons of immigrants: are they at an earnings disadvantage? American Economic Review 67(February):376-80. 1978 The effects of Americanization on the earnings of foreign-born men. Journal of Political Economy 86:897-922. 1979 The economic progress in immigrants: some apparently universal patterns. Pp. 357-399 in William Fellner, ea., Contemporary Economic Problems. Washington, D.C.: American Enterprise Institute for Public Policy Research. 1980a An analysis of the economic progress and impact of immigrants. Washington, D.C.: U.S. Department of Labor. 1980b Immigrants' earnings patterns by sex, race, and ethnic groupings. Monthly Labor Review 103~0ctober):22-25. 1982 The Employment of Immigrants in the United States. Washington, D.C.: American Enterprise Institute for Public Policy Research. Clark, Juan M. 1977 The Cuban Exodus. Why? Report to the Cuban Exile Union, Miami, Fla. Photocopy. Clogg, Clifford C. 1979 Measuring Underemployment: Demographic Indicators for the United States. New York: Academic Press. Cummins, James 1976 The Influence of Bilingualism on Cognitive Growth: A Synthesis of Research Findings and Explanatory Hypotheses. Working Papers on Bilingualism No. 9. April:1-43. 1977 Psycholinguistic evidence. Pp. 78-89 in Center for Applied Linguistics, ea., Bilingual Education: Current Perspectives. volume 4, Education. Arlington, Va.: Center for Applied Linguistics. Dillard, J.L. 1973 Black English: Its History and Usage in the United States. New York: Vintage Books. Duncan, Beverly 1965 Dropouts and the unemployed. Journal of Political Economy 73:121-134. Duncan, O.D., D.L. Featherman, and B. Duncan 1972 Socioeconomic Background and Achievement. New York: Seminar. Duran, Richard P. 1983 Hispanics' Education and Background: Predictors of College Achievement. New York: College Entrance Examination Board. Edwards, Linda 1976 School retention of teenagers over the business cycle. Journal of Human Resources 11, no. 2 (Spring):1-13.

455 Estrada, Leobardo 1980 Review and analysis of Hispanic youth employment data collection systems and reporting methods. Pp. 21-50 in National Council of La Raza, ea., Hispanic Youth Employment: Establishing a Knowledge Base. Knowledge Development Report 10.1. Washington, D.C.: U.S. Department of Labor. Featherman, David L., and T. Michael Carter 1976 Discontinuities in schooling and the socioeconomic life cycle. Pp. 133-161 in W.H. Sewell, R.M. Hauser and D.L. Featherman, eds., Schooling and Achievement in American Society. New York: Academic Press. Featherman, D.L., and R.M. Hauser 1978 Opportunity and Change. New York: Academic Press. Fernandez, Roberto M. 1980 Selected analyses of the pretest data. Appendix D in Francois Nielsen, Hispanic Youth in U.S. Schools: A Design for Analysis. Washington, D.C.: National Center for Education Statistics. Fernandez, Roberto M., and Francois Nielsen 1984 Bilingualism and Hispanic Scholastic Achievement: Some Baseline Results. Unpublished manuscript. Department of Sociology' University of Arizona. Fishman, Joshua A. 1969 A sociolinguistic census of a bilingual neighborhood. American Journal of Sociology 75, no. 3:323-339. Fishman, Joshua A., and Robert C. Cooper 1969 Alternative measures of bilingualism. Journal of Verbal Learning and Verbal Behavior 8:276-282. Fishman, Joshua A., and Charles Terry 1969 The validity of census data on bilingualism in a Puerto Rican neighborhood. American Sociological Review 34:636-650. Fligstein, Neil, and Roberto M. Fernandez 1982 The Causes of Hispanic Educational Attainment: History, Patterns, and Analyses. A report to the National Commission for Employment Policy, Washington, D.C. 1985 Educational transitions of whites and Mexican-Americans. Pp. 161-192 in Marta Tienda and George Borjas, eds., Hispanics in the U.S. Economy. New York: Academic Press. Frankel, Martin, Luane Kohnke, David Buonanno, and Roger Tourangeau 1981 High School and Beyond Sample Design Report. Chicago, Ill.: National Opinion Research Center. Freeman, Richard, and David Wise, eds. 1982 The Youth Labor Market. Chicago, Ill.: University of Chicago Press. Garcia, Phillip 1983 Dual Language Characteristics and Earnings: Male Mexican Workers in the United States. Los Angeles Population Research Laboratory' University of Southern California. Photocopy. Harber, Jean R., and D.N. Bryen 1976 Black English and the task of reading. Review of Educational Research 46~3~:387-405.

456 Hauser, Robert M. 1971 Socioeconomic Background and Educational Performance. - Washington, D.C.: American Sociological Association. Heckman, James J. 1979 Sample selection bias as a specification error. Econometrica 47:1S3-162. Hernandez, J., L. Estrada, and D. Alvirez 1973 Census data and the problem of conceptually defining the Mexican-American population. 53(4):671-687. Heyns, Barbara, and Thomas L. Hilton 1982 The cognitive tests for High School and Beyond: an assessment. Sociology of Education 55(April/July): 89-102. Hirano-Nakanishi, Marsha, and R.L. Diaz 1982 Differential Educational Attainment among "At-Risk" Youth: A Case Study of Language Minority Youth of Mexican Descent and Low Socioeconomic Status. Los Alamitos, Calif.: National Center for Bilingual Research. Hogan, Dennis P. 1981 Transitions and Social Change: The Early Lives of American Men. New York: Academic Press. Jaffe, A.J., Ruth M. Cullen, and Thomas D. Boswell 1980 The Changing Demography of Spanish Americans. New York: Academic Press. Jencks, C.S., M. Smith, H. Acland, M.J. Bane, D. Cohen, H. Gintis, B. Heyns, and S. Michelson 1972 Inequality: A Reassessment of the Effect of Family and Schooling in America. New York: Harper and Row. Jencks, C.S., S. Bartlett, J. Crouse, D. Eaglesfield, G. Jackson, K. McClelland, P. Mueser, M. Olneck, J. Schwartz, S. Ward, and J. Williams 1979 Who Gets Ahead? The Determinants of Economic Success in America. New York: Basic Books. Jones, Calvin, M. Clarke, G. Mooney, H. McWilliams, I. Crawford, B. Stephenson, and R. Tourangeau 1983 High School and Beyond: 1980 Senior Cohort First Follow-up (1982) Data File User's Manual. Washington, D.C.: National l Center for Education Statistics. Jorge, Antonio, and Raul Moncanz 1980 Cubans in South Florida: no. 3 (Fall):37-87. Labov, William 1976 Lanquaqe in the Inner CitY: Social Science Quarterly a social science approach. Metas 1, Studies in the Black English Vernacular. Philadelphia: University of Pennsylvania Press. Lambert, Wallace, and Richard Tucker 1972 Bilingual Education of Children: Rowley, Mass.: Newbury House. The St. Lambert Experiment. Lewin-Epstein, Noah 1981 Youth Employment During High School. An Analysis of High School and Beyond: A National Longitudinal Study for the 1980's. Washington, D.C.: Statistics. National Center for Education

457 Lopez, David E. 1976 The social consequences of Chicano home/school bilingualism. Social Problems 24, no. 2 (Dec.~:234-246. Mayers, Raymond Sanchez 1980 Characteristics of Hispanic youth. Pp. 51-88 in National Council of La Raza, ea., Hispanic Youth Employment: Establishing a Knowledge Base. Knowledge Development Report 10.1. Washington, D.C.: U.S. Department of Labor. McKay, Roberta V. 1974 Employment and unemployment among Americans of Spanish Origin. Monthly Labor Review 97tApril):12-16. McLaughlin, Steven D. 1982 English Language Proficiency, Occupational Characteristics and the Employment Outcomes of Mexican-American Men. Seattle, Wash.: Battelle Human Affairs Research Centers. McManus, Walter, William Gould, and Finis Welch 1983 Earnings of Hispanic men: the role of English language proficiency. Journal of Labor Economics 1~2~:101-30. National Center for Education Statistics 1980 The Condition of Education for Hispanic Americans. Washington, D.C.: National Center for Education Statistics. National Commission for Employment Policy 1982 Hispanics and Jobs: Barriers to Progress. Washington, D.C. National Commission for Employment Policy. National Commission for Manpower Policy, ed. 1976 From School to Work: Improving the Transition. Washington, D.C.: National Commission for Manpower Policy. National Council of La Raza 1980 Hispanic Youth Employment: Establishing a Knowledge Base. Knowledge Development Report 10.1. Washington, D.C.: U.S. Department of Labor. Neugarten, Bernice, and G.O. Hagestad 1976 Age and the life course. Pp. 35-55 in R. H. Binstock and E. Shanas, eds., Handbook of Aging and the Social Sciences. New York: Van Nostrand Reinhold. Newman, Morris J. 1978 A profile of Hispanics in the U.S. work force. Monthly Labor Review (December):3-14. Nielsen, Francois 1980 Hispanic Youth in the U.S. Schools: A Design for Analysis. Washington, D.C.: National Center for Education Statistics. Nielsen, Francois, and Roberto M. Fernandez 1982 Achievement of Hispanic Students in American High Schools: Background Characteristics and Achievement. Washington, D.C.: National Center for Education Statistics. Noboa, Abdin 1980 Hispanics and desegregation: summary of Aspira's study of Hispanic segregation trends in U.S. school districts. Metas 1, no. 3 (Fall):1-24.

458 Olivas, Michael 1979 The Dilemma of Access: Minorities in Two Year Colleges. Washington, D.C.: Howard University Press. 1981 Financial Aid: Access and Packaging Policies. Institute for Research on Educational Finance and Governance. Palo Alto, Calif.: Stanford University. O'Malley, J.M. 1981 Children's English and Services Study: Language Minority Children with Limited English Proficiency in the United States. Rosslyn, Va.: InterAmerica Research Associates. Ornstein, Michael D. 1976 Entry into the American Labor Force. New York: Academic Press. Osterman, Paul 1980 Getting Started: The Youth Labor Market. Cambridge, Mass.. MIT Press. Peal, Elizabeth, and Wallace Lambert 1962 The relation of bilingualism to intelligence. Psychological Monographs: General and Applied 76~546~:1-23. Pedraza-Bailey, Silvia 1980 Political and Economic Migrants in America: Cubans and Mexicans. Ph.D. dissertation. Department of Sociology, University of Chicago. Pedraza-Bailey, Silvia, and Terry Sullivan 1979 Bilingual education in the reception of political immigrants: the case of Cubans in Miami, Florida. In Raymond V. Padilla, ea., Ethnoperspectives in Bilingual Education Research: Bilingual Education and Public Policy in the United States. Department of Foreign Languages and Bilingual Studies. Ypsilanti, Mich.: Eastern Michigan University. Peng, Samuel S. 1977 Trends in the entry into higher education: 1961-1972. Educational Researcher 6~11:15-9. Portes, Alejandro 1982 Immigrants' attainment: an analysis of occupation and earnings among Cuban exiles in the United States. Pp. 91-111 in R.M. Hauser, F. Mechanic, A.O. Haller, and T.S. Hauser, eds., Social Structure and Behavior: Essays in Honor of William Hamilton Sewell. New York: Academic Press. Portes, Alejandro, and R.L. Bach 1980 Immigrants earnings: Cuban and Mexican immigrants in the United States. International Migration Review 14~3~:315-342. Portes, Alejandro, J.M. Clark, and R.L. Bach 1977 The new wave: a statistical profile of recent Cuban exiles to the United States. Cuban Studies 7tJanuary): 1-32. Portes, Alejandro, J.M. Clark, and M. Lopez 1981 Six years later: the process of incorporation of Cuban exiles in the United States: 1973-1979. Cuban Studies 11, no. 2 (July):1-24.

459 Reimers, Cordelia W. 1984 Sources of the family income differentials among Hispanics, blacks, and white non-Hispanics. American Journal of Sociology 89(4):889-903. Rogg, Eleanor M. 1974 The Assimilation of Cuban Exiles. New York: Aberdeen Press. Rumberger, Russell W. 1980 High school dropouts. Pp. 267-277 in M. Borus, J. Crowley, R Rumberger, R. Santos, and D. Shapiro, Pathways to the Future: A Longitudinal Study of Young Americans. Center for Human Resource Research. Columbus: Ohio State University. 1983 Dropping out of high school: the influence of race, sex, and family background. American Educational Research Journal 20, no. 2 (Summer):199-220. Ryscavage, Paul M., and Earl F. Mellor 1973 The economic situation of Spanish Americans. Monthly Labor Review 96(April):3-16. Santana-Cooney, Rosemary 1979 Intercity variations in Puerto Rican female participation. Journal of Human Resources 14:222- 35. Santana-Cooney, Rosemary, and A.E. Warren 1979 Declining female participation among Puerto Rican New Yorkers: a comparison with native white nonSpanish New Yorkers. Ethnicity 6:281-297. Sewell, William H. 1971 Inequality of opportunity for higher education. American Sociolocical Review 36:793-809. Sewell, W.H., and R.M. Hauser 1975 Education, Occupation, and Earnings: Achievement in the Early Career. New York: Academic Press. Shea, Brent 1976 Schooling and its antecedents: substantive and methodological issues in the status attainment process. Review of Educational Research 46:463-625. Smith, Tom W. 1980 Ethnic measurement and identification. Ethnicity 7:78-95. 1983 Problems in ethnic measurement: over-, under- and misidentification. American Statistical Association - Proceedings:107-116. Spence, A. Michael 1974 Market Signaling: Information Transfer in Hiring and Related Screening Processes. Cambridge, Mass.: Harvard University Press. Steinberg, L.D., P.L. Belinde, and K.S. Chan 1982a Dropping Out Among Language Minority Youth: A Review of the Literature. Los Alamitos, Calif.: National Center for Bilingual Research. Steinberg, L.D., D. Greenberger, L. Garduque, and S. McAuliffe 1982b High school students in the labor force: some costs and benefits to schooling and learning. Educational Ev~luation and Policy Analysis 4(3):363-372.

460 Stevenson, Wayne 1978a The relationship between early work experience and future employability. Pp. 93-124 in Avril V. Adams and Garth L. Mangum, ea., The Lingering Crisis of Youth Unemployment. Kalamazoo, Mich.: W.E. Upjohn Institute for Employment Research. 1978b The transition from school to work. Pp. 65-92 in Avril V. Adams and Garth L. Mangum, eds., The Lingering Crisis of Youth Unemployment. Kalamazoo, Mich.: W.E. Upjohn Institute for Employment Research. Stolzenberg, Ross M. 1982 Occupational Differences between Hispanics and Non-Hispanics. Report to the National Commission for Employment Policy. Santa Monica, Calif.: Rand Corporation. Sullivan, Teresa A. 1979 Marginal Workers, Marginal Jobs: The Underutilization of American Workers. Austin: University of Texas Press. Sullivan, Terry, and Silvia Pedraza.Bailey 1979 Differential Success Among Cuban-American and Mexican- American Immigrants: The Role of Policy and Community. Washington, D.C.: U.S. Department of Labor, Employment and Training Administration. Theil, Henri 1971 Principles of Econometrics. New York: John Wiley & Sons. - Tienda, Marta, ed. 1981 Hispanic Origin Workers in the U.S. Labor Market: Comparative , Analyses of Employment and Earnings. Washington, D.C.: U.S. Department of Labor. 1982 Sex, ethnicity and Chicano status attainment. International Migration Review 16~2~:435-473. 1983 Market characteristics and Hispanic earnings: a comparison of natives and immigrants. Social Problems 31, no. 1 (October):59-72. Trieman, D.J., and K. Terrell 1975 Sex and the process of status attainment: a comparison of working women and men. American Sociological Review 40tApril):174-200. U.S. Commission on Civil Rights 1971 The Unfinished Education: Outcomes for Minorities in the Five - Southwestern States. Washington, D.C.: U.S. Commission on Civil Rights. 1974 Counting the Forgotten: The 1970 Census Count of Persons of Spanish Speaking Background. Washington, D.C.: U.S. Commission on Civil Rights. 1978 Social Indicators of Equality for Minorities and Women. _ Washington, D.C.: U.S. Commission on Civil Rights. 1982 Unemployment and Underemployment Among Blacks, Hispanics, and Women. Washington, D.C.: U.S. Commission on Civil Rights.

461 O.S. Department of Health, Education and Welfare 1974 A Study of Selected Socioeconomic Characteristics of Ethnic Minorities Based on the 1970 Census. Volume I: Americans of Spanish Origin. Publication No. (OS) 75-120. Washington, D.C.: U.S. Department of Health, Education and Welfare. Veltman, Calvin 1980 Relative Educational Attainments of Minority Language Children. Washington, D.C.: National Center for Education Statistics. 1981 The Role of Language Characteristics in the Socioeconomic Attainment Process of Hispanic Origin Men and Women. Washington, D.C.: National Center for Education Statistics. Wilson, K.L., and W.A. Martin 1982 Ethnic enclaves: a comparison of the Cuban and Black economies in Miami. American Journal of Sociology 88~1~:135-60. Wilson, Kenneth L., and Alejandro Portes 1980 Immigrant enclaves: an analysis of the labor market experiences of Cubans in Miami. American Journal of Sociology 86:295-319.