Gender, earnings, and the English skill acquisition of Hispanic workers in the United States.

Mora, Marie T. ; Davila, Alberto

I. INTRODUCTION

Recently, the number of English deficient residents has escalated in the United States. According to the 1980 and 1990 decennial censuses, individuals aged five years and older whose reported English verbal skills are "not well" or "not at all" rose from 4.22 million (2% of the population) to 6.67 million (3%). This percentage is more pronounced in the public schools. According to Han et al. [1997], over 2.12 million students (5.1%) of the 41.62 million public school students in grades K-12 were identified as being limited-English proficient (LEP) in the 1993-1994 school year. Other estimates, such as those made by Henderson, Abbott, and Strang [1993], put the public school LEP share at over 6%.

Increases in the LEP numbers mirror the growth in the immigrant population as well as the relatively high fertility rates of language minority Americans. The Hispanic and Asian/Pacific Islander ethnic groups account for much of this growth, as shown by the National Research Council [1997]. Hispanics, in fact, are projected by the U.S. Census Bureau [1996] to become this nation's largest racial/ethnic minority group in less than a decade. Policywise the growing LEP population raises economic and social questions concerning, for example, the best schooling programs for children in this group, as well as the impact of English proficiency in labor markets. Answers to some of these questions clearly require detailed and inclusive analyses.(1)

Indeed, many social scientists have linked the ability to communicate in English to labor market value and occupational selection.(2) The bulk of this research, however, primarily focuses on men. Less is known on whether gender differently affects the English proficiency outcomes of language minorities. Researchers in other social science fields, such as Stevens [1985, 1986, 1992], Cameron [1992], and Ellis [1994] have considered gender disparities with respect to the acquisition of English skills.(3) However, these investigations have not fully implemented some of the empirical advances made by economists to analyze English proficiency. Moreover, this extant work tends to focus solely on gender-related English-skill differentials at a single point in time.

In this study, we use recent empirical strategies in the economics field to analyze labor market facets associated with the English skill acquisition of Hispanic men and women. We also employ a two-period analysis that allows us to observe whether gender differences in the accumulation of English skills exist. With respect to our previous discussion, such knowledge facilitates the design of programs aimed at improving the English fluency of the LEP population.

In particular, we first track synthetic cohorts to investigate if the English proficiency acquisition made by Hispanic workers during the 1980s varies with respect to gender. We then test whether poor English skills differently affect the earnings and occupational sorting of Hispanic men and women. The 1% 1980 and 1990 Public Use Microdata Samples (PUMS) are utilized in these labor market analyses.

Our findings suggest that English skill procurement varies between Hispanic men and women, depending on age and schooling location. We also find that women face a lower English deficiency earnings penalty that rises more rapidly with education than the penalty obtained by their otherwise similar male peers. Our occupational selection results parallel this penalty difference, where English skills serve as a stronger sorting mechanism for Hispanic women than men. Policy prescriptions from such analyses suggest that male-oriented studies overestimate the earnings gains and understate the occupational sorting associated with acquiring English fluency.

II. ENGLISH SKILL ACQUISITION IN THE LABOR MARKET

Our first objective tests whether women acquire English proficiency at a different rate than their male peers. All of our empirical work relies on the 1% PUMS of the 1980 and 1990 U.S. decennial censuses. In addition to the usual information provided by most large data sets, PUMS data include categorical information on how well individuals from minority language households report speaking English; these categories include: "n/a; only English is spoken at home," "very well," "well," "not well," and "not at all."(4) Unfortunately, these data do not contain information on English reading or writing skills, such that our empirical investigations pertain to verbal ability.

PUMS Sample Selection

Our PUMS sample contains individuals aged 16-64 years, and includes immigrant and U.S.-native Hispanic workers who speak English or Spanish (or both), as well as U.S.-native non-Hispanic whites who only speak English at home. We drop from our sample individuals reporting self-employment, work outside of the U.S., military status, work without pay, enrollment in school, zero earnings, zero usual hours worked, or zero weeks worked. Finally, we exclude workers with allocated (imputed) earnings, weeks worked, and usual hours worked per week.

PUMS Sample Summary Statistics for Hispanic Workers

Table I presents selected summary statistics for the Hispanic sample within each English skill category.(5) Note that with the exception of the "not at all" category, the percentage of women is higher in the more proficient categories. This finding reaffirms other social science research showing a gender differential in English skills (see footnote 3).

Other figures in Table I conform to previous research on English proficiency, as in footnote 2. Notably, hourly wages increase with English fluency (hourly wage = previous year's annual salary divided by usual weekly hours worked times annual weeks worked); English proficient workers have more education and less experience on average than their less English fluent peers;(6) Mexican Americans and Cubans tend to be overrepresented (and Puerto Ricans and Other Hispanics underrepresented) in the less proficient categories; and the share of immigrants inversely corresponds to English fluency.(7)

The Construction of the English Deficiency Index

We next collapse these English skill categories into an English deficiency index to be used in our analyses:

(1) EDI = (No English is Spoken)

+ [summation of] ([r.sub.j] * [Category.sub.j]) where j = NotWell to VeryWell;

for j = "Not Well," "Well," "Very Well," and 0 [less than or equal to] [r.sub.j] [less than or equal to] 1.

[r.sub.j] represents a normalized parameter ranging from zero to one, and [Category.sub.j] denotes the jth "intermediate" English proficiency - a binary variable that takes on the value of zero or one. The base group (the origin) of EDI contains individuals from monolingual English households (EDI equals zero). EDI equals one for completely English-deficient individuals (no English spoken). The coefficients on the intermediate English skill categories may be interpreted as each fluency level's percentage of"complete" English deficiency.(8)

To calculate the [r.sub.j]'s, we focus on Hispanics and use non-linear regression to estimate:

(2) ln(W) = x[[Alpha].sub.1] + (female * x)[[Alpha].sub.2]

+ (90PUMS * x) [[Alpha].sub.3]

+ (female * 90PUMS * x) [[Alpha].sub.4] + e,

where ln(W) represents the natural logarithm of hourly wages, and x depicts a vector containing EDI, education, experience, experience-squared, and several census-coded geographic indicator variables.(9) The terms (female * x), (90PUMS * x), and (female * 90PUMS * x) denote vectors of the variables in x multiplied to binary variables indicating if the individual is: female, from the 1990 PUMS sample, and female in the 1990 PUMS sample. The [[Alpha].sub.i]'s represent vectors of coefficients to be estimated, and finally the error term is e [similar to] N(0,[[Sigma].sup.2]I).

This empirical technique allows the constructed EDI to reflect earnings differences associated with poor English skills unexplained by education, job experience, and regional and temporal cost of living differences. The EDI also assesses English skills as a continuum, providing intuitive insights into the [TABULAR DATA FOR TABLE I OMITTED] labor market role of English fluency. Table II presents the estimated [r.sub.j]'s, and the remaining results can be obtained from the authors.

TABLE II

English Deficiency Index of Hispanic Workers in the 1980 and 1990
PUMS

English Skill Category Coefficient Standard Error

Only speaks English at home 0.000 -
Speaks English "very well" 0.107 0.024
Speaks English "well" 0.392 0.037
Speaks English "not well" 0.856 0.062
No English is spoken 1.000 -

Notes: Only the Hispanic sample is used to estimate this index.
Additional estimations of equations (1) and (2) have been used to
test whether the coefficients on the intermediate English skill
categories significantly changed between the two censuses, which
they did not. The test for the statistical difference was conducted
using the pooled 1980 and 1990 PUMS data in non-linear regression,
and the EDI coefficients were interacted with a binary variable for
90PUMS. The statistical significance was determined by the z-ratio
on the category-90PUMS interactions; the absolute value z-statistics
for "very well," "well," and "not well" are 0.129, 0.110, and 0.387.
Moreover, an F-test reveals that these EDI coefficients as a group
are not statistically different between the censuses; the calculated
F-value is 0.116.

To explore the reliability of EDI, we present selected mean statistics of this index in Table III. Note that with the exception of college graduates, women appear slightly more proficient in English than their male counterparts, maintaining the sample statistics presented earlier. Other aspects in Table III support extant research in that individuals educated outside of the U.S. tend to be less fluent in English than their U.S.-schooled peers; English skills positively relate to the level of education; and English fluency varies with respect to age and schooling location. In short, the mean characteristics of EDI accord with conventional wisdom; hence, EDI serves as a reliable proxy for English deficiency.

Gender and the English Skill Acquisition of Hispanic Workers

Using EDI, we now test whether gender significantly relates to the English skill acquisition of Hispanic workers. Even though the PUMS data are not longitudinal, temporal changes may be approximated through the use of synthetic cohorts. Synthetic cohort analyses have been utilized by researchers such as Borjas [1994] in the economics of immigration domain.

To illustrate this technique, individuals between the ages of 16-54 years in 1980 can be synthetically tracked in the 1990 PUMS by those aged 26-64 years (excluding immigrants who arrived to the U.S. after 1980). We realize that because some immigrants in the 1980 sample will have emigrated out of the U.S. before 1990, the 1990 PUMS may not entirely reflect "true" immigrant cohorts. We acknowledge this potential problem by dividing the cohorts into samples of individuals solely or partly educated in the U.S., or completely educated abroad. Presumably, immigrants who attended U.S. schools have a lower emigration likelihood, such that the U.S.schooled synthetic cohorts should be particularly consistent with the "true" cohorts.

Moreover, because we exclude from our sample students enrolled in school, individuals attending college in 1980 are not in the 1980 PUMS sample, but could be in the 1990 PUMS sample. To reduce the shortcomings arising from this scenario, we further explore three smaller cohorts: I, J, and K. Cohort I includes workers aged 26-36 years in 1980 or 36-46 years in 1990. Similarly, individuals in Cohort J [Cohort K] were 36-46 [46-54] years old in 1980 or 46-56 [56-64] years old in 1990. We exclude immigrants migrating to the U.S. after 1980 in all cohorts.

To test whether the English fluency accumulated by Hispanic workers during the 1980s varies with respect to gender, we estimate:

(3) EDI = [[Delta].sub.1]female + [[Delta].sub.2] 90PUMS

+ [[Delta].sub.3] (female * 90PUMS) + (Other)D + [Epsilon],

where the [[Delta].sub.i]'s represent the coefficients of interest, and [Epsilon] is the normally distributed stochastic error term. D denotes a vector of additional coefficients associated with Other, which contains the additional explanatory variables of education; potential experience years workers had accrued in 1980 (= experience for workers in the 1980 PUMS, and experience - 10 for workers in the 1990 PUMS); ethnicity; immigrants' U.S.-tenure (in categories), and the school location and potential foreign experience of immigrants. Table IV presents the regression results from estimating equation (3) for the synthetic cohorts.

[TABULAR DATA FOR TABLE III OMITTED]

The two most important rows in Table IV for our discussion correspond to the female and female*90PUMS variables. While workers on average gain English fluency (i.e., become less deficient) even after entering the labor market, women appear to acquire English skills differently than men depending on age and schooling location. The younger female workers schooled in the U.S. enter the labor market with an English fluency advantage over their otherwise similar male peers. Such an English proficiency edge, however, appears to subside over time. As with other human capital, the English fluency advantage seemingly hits diminishing returns. The oldest cohort (Cohort K) suggests that U.S.-educated women in this group do not appear to have statistically different English skills than their male counterparts.

[TABULAR DATA FOR TABLE IV OMITTED]

Immigrant women educated abroad (particularly in the youngest cohort) initially have an English fluency disadvantage relative to men, although their English skills progress over time. One interpretation of this finding has been offered by Espenshade and Fu [1997], who note that lower English skills of immigrant women may reflect limited past expectations of female labor market participation. The English fluency of immigrant women surpassed that of men during the 1980s only in Cohort K. For our purposes, the synthetic cohorts in Table IV reveal the presence of gender differentials associated with the acquisition of English proficiency in the labor market.

The remaining variables reaffirm other work on the correlates of English skills. Namely, education and its location positively relate to English skills, the lower English proficiency of immigrants increases with U.S.tenure, and ethnicity associates with English fluency differences.

III. ENGLISH SKILLS, GENDER, AND EARNINGS

At this point, we have provided empirical evidence suggesting that the acquisition of English skills significantly varies between men and women. We now examine whether this linguistic difference projects into the labor market via earnings and occupational sorting.

English Skills and the Change in the English Deficiency Earnings Penalty

To test whether gender influences the earnings penalty associated with poor English skills, we use the English deficiency index from above and estimate:

(4) ln(W) = X[[Beta].sub.1] + (female * X)[[Beta].sub.2]

+ (90PUMS * X) [[Beta].sub.3]

+ (female * 90PUMS * X) [[Beta].sub.4] + u,

where the [[Beta].sub.i]'s represent vectors of coefficients to be estimated, and u is the stochastic error term where u[similar to]N(0, [[Sigma].sup.2]I). X depicts a vector that includes the variables in vector x from equation (2) as well as ethnicity, immigrants' U.S.-tenure (in categories), and the school location and potential foreign experience of immigrants. The remaining terms in equation (4) have a similar interpretation to their counterparts in equation (2). The functional form of equation (4) accounts for earnings differences between men and women in 1979 and 1989. Table V reports selected regression results from estimating this earnings equation, with U.S.-born monolingual-English non-Hispanic whites comprising the base group. For the sake of brevity, we only report the results associated with the variables of interest (such as female and EDI); the remaining results are available upon request from the authors.

The backdrop coefficients on EDI and its interactions support those found in related studies (see footnote 8). That is, the combined direct and indirect effects suggest that workers with poor English skills earn relatively less than their English fluent peers; this earnings penalty increases with higher levels of schooling (and also with experience in 1989).(10)

Table V also shows that the combined direct and indirect effects (through schooling) of poor English skills on earnings statistically differ between women and men. Specifically, Hispanic women face a significantly different English deficiency earnings penalty than their male peers, affirming that gender related linguistic differences ensue in labor market earnings. Most women receive a relatively smaller earnings penalty that rises more sharply with education. For example, the 1979 penalty received by an English deficient woman with 10 schooling years is about 13% lower [= 56.1 - (10*0.043)] than the penalty obtained by her otherwise similar male peers; a female high school graduate accrues an earnings penalty approximately 4.5% smaller [= 56.1 - (12*0.043)] than her male counterparts' penalty.

[TABULAR DATA FOR TABLE V OMITTED]

The relatively smaller penalty for women implies that overall earnings gains associated with enhancing English skills fall short of the gains estimated from exclusively focussing on men. Hence, studies solely utilizing men overvalue the increase in earnings associated with enhanced English fluency for women, particularly for the less educated. Moreover, our findings provide additional information on the well documented male-female earnings gap because the unrestricted sample means underscore this gap for men and women with the same reported proficiency.

The EDI*90PUMS interaction in Table V suggests that the earnings penalty of English deficient workers with little education fell significantly between 1979 and 1989. However, this penalty increased with respect to education and experience during this time, such that skilled workers faced a higher EDI earnings penalty than their less skilled counterparts by 1989. Overall, this finding supports Mora [1998], and further accords with the increase in returns to other human capital that occurred during the 1980s as discussed, for example, by Murphy and Welch [1992, 1989]. Note also that during the 1980s, the earnings penalty of English deficient women increased relative to men, but its rate of change with respect to schooling fell. In other words, male/female earnings differences associated with English proficiency somewhat narrowed during the 1980s, but did not vanish.

Other results from estimating equation (4) are standard, and we only summarize the highlights here. Education enhances earnings, and [TABULAR DATA FOR TABLE VI OMITTED] each additional year of experience increases earnings at a decreasing rate. As expected, these attributes were compensated with higher labor market rewards in 1989 than in 1979. Also, women obtain higher returns to schooling and lower returns to experience than their male counterparts; similar to the EDI results, however, these differences lessened during the 1980s. Moreover, the returns to experience significantly decline with foreign job training, although these lower returns dissipate for female immigrants in 1989.

IV. GENDER, ENGLISH DEFICIENCY, AND OCCUPATIONAL SORTING IN 1990

One possible explanation for the relatively smaller English deficiency earnings penalty received by women might hinge on the role of English skills in occupational sorting. To test this proposition, we estimate a multinomial logit for the Hispanic sample with occupation (composed of six major categories) as the dependent variable; the base group of comparison consists of professional, executive, and managerial jobs. The right-hand side variables include EDI, female, EDI*female, and the variables in vector X from equation (4). The EDI*female interactions should indicate whether vocational selection related to English skills varies with gender.

Table VI reports selected results from estimating this multinomial logit using the 1990 PUMS. We do not utilize our 1980 sample because of potential structural changes that may have occurred in the labor market during the 1980s. However, we replicate our results for 1980 (not shown); the same basic results hold.

Table VI indicates that poor English fluency occupationally sorts away from professional positions, reaffirming earlier studies like Jasso and Rosenzweig [1989] and McManus [1990]. Moreover, this sorting significantly favors women in all vocations except technical, sales, and administrative support trades. English proficiency seems to filter women beyond its effect on the careers of men, such that English fluent women are occupationally "crowded." This crowding might diminish the relative English skill returns of women, which is consistent with our earnings analysis in Table V. As noted by Bergmann [1989], if occupational segregation exists between men and women, market forces decree two sets of wages based on gender. Then, if women are vocationally crowded with respect to skill (say, English proficiency), their smaller job array should yield lower skill returns owing to excess female labor supply in these occupations. In short, our findings for men underscore the influence of English skills on female job selection.

Other occupational findings reaffirm extant research. For example, the employment odds of women fall short of men's odds in the blue collar sector relative to professional positions, ceteris paribus. Moreover, education and job experience affect career choice, where skilled workers tend to be employed in white collar vocations. Finally, immigration status, foreign human capital, and ethnicity relate to occupational sorting.

V. CONCLUDING REMARKS

Although research in the economics literature on the labor market value of English proficiency traditionally focuses on men, other social scientists have recognized that linguistic skills vary with respect to gender. We empirically reaffirm the importance of this claim. For example, we find that Hispanic female workers maintain their relative verbal command of English over time (albeit at a diminishing rate) if they were educated in the U.S. Female immigrants educated abroad are not as proficient as their male peers, but appear to have acquired English fluency at a faster clip during the 1980s. These results not only show that the acquisition of English proficiency occurs in the labor market, but that such skill procurement varies between language minority men and women.

Gender differentials along the English-skill dimension also surface in earnings and occupational sorting. We find that English deficient Hispanic women receive a lower earnings penalty that rises more rapidly with education than the penalty obtained by their otherwise similar male peers. This penalty difference relates to our occupational selection results, where English skills act as a stronger sorting mechanism for Hispanic women. The relative vocational "crowding" of English proficient women may diminish their English proficiency returns (and hence reduce their English deficiency earnings penalty).

One possible explanation for the gender differential in English fluency may be that language minority women have a greater necessity to acquire basic majority language skills to effectively compete in the labor market. Indeed, sociolinguists such as Cameron [1992] have suggested that many language minority women perceive a more narrow occupational array than men because of fewer construction and other physically demanding job opportunities; our occupational sorting results fit with this claim. Another explanation of the apparent gender difference in English proficiency accumulation may involve females' expectations of household activities; for example, Stevens [1985, 1986] observes that women transfer language skills to children more than men. Future research should explore this issue by focussing on the intergenerational properties of majority language skill acquisition.

In short, our findings suggest that research ignoring gender may yield an incomplete picture of the English skill outcomes in the U.S. Moreover, male-oriented labor market studies may overestimate the potential earnings gains and understate the occupational sorting associated with enhancing the English fluency of women. Given the growing number of limited-English proficient individuals, labor market, educational, and immigration policies targeting language issues will affect much larger segments of the population. As such, underlying aspects of these policies should be fully addressed to understand the potentially far-reaching economic and social consequences.

ABBREVIATIONS

EDI: English Deficiency Index LEP: Limited-English Proficient PUMS: Public Use Microdata Samples

1. While our study concentrates on the labor market implications of English fluency, educational policies designed for LEP children also have increasing importance. In fact, evidence suggests that LEP individuals tend to be relatively young; a perusal of census data reveals that the share of 5-year-olds [15-year-olds] in the English skill distribution monotonically decreases [increases] with English proficiency. Lopez and Mora [1997] and Glenn [1997] represent recent studies that discuss some potential long-term consequences of programs like bilingual education.

2. The study conducted by McManus, Gould, and Welch [1983] represents a classic analysis of English proficiency in the U.S. labor market. More recent work includes Jasso and Rosenzweig [1989]; McManus [1990, 1985]; Davila, Bohara, and Saenz [1993]; Chiswick and Miller [1995]: and Mora [1998]. Examples of empirical investigations into socioeconomic factors related to English fluency can be found in McManus, Gould, and Welch [1983]; Chiswick and Miller [1995]; Espenshade and Fu [1997]; Espinosa and Massey [1997]; and Mora [1998].

3. For example, Ellis [1994] notes that women seem to have a greater potential for acquiring standard proficiency in a second language because they rapidly adopt newer linguistic forms and tend to rely on their first language skills as a cognitive aid more than men.

4. The PUMS questionnaire first asks whether an individual speaks a non-English language at home. If not, the individual disregards the question on English fluency. We realize some individuals may speak a non-English language outside of the home while only speaking English at home; however, we feel that the impact of this potential bias, if any, on our results is quite small.

5. The 1990 PUMS data provide statistical weights to be used in analyzing the data; we utilize these weights in all of our analyses. The number of observations presented, however, is not weighted. The 1980 PUMS data are self-weighted.

6. The 1990 PUMS data do not provide the actual number of schooling years; instead, they present the highest education level attained in categories. These categories closely parallel the schooling coding presently utilized by the Current Population Survey (CPS); however, in February 1990, the CPS provided education both continuously and categorically. Because the same individuals gave information for both schooling codes, Park [1994] estimates schooling years using the educational categorical coding; we use Park's translation to measure completed schooling years. Experience equals age - education - 5. We realize that because of differences in the labor force participation between men and women, the experience proxy may not completely capture "true" labor market training. Unfortunately, the PUMS do not provide an alternative.

7. Because the labor-market consequences of job training and education may differ when acquired abroad, we construct foreign experience to approximate the maximum potential years of foreign training. Specifically, we compare the minimum number of years since migration (using migration-interval categories consistent between both PUMS of 0-5, 6-10, 11-20, 21-30, and 31 or more years) with total years of potential training. If the immigrant's working years exceed the time since migration, the difference between experience and U.S. tenure comprises foreign work years; otherwise, we assume the individual did not work abroad. Similarly, we compare the individual's estimated schooling completion age with these migration intervals to approximate the location of educational attainment. Although imprecisely measured, such variables provide additional insight into the labor-market experiences of Hispanic immigrants in the U.S.

8. This empirical technique was cultivated by McManus, Gould, and Welch [1983]. See McManus [1990, 1985] and Mora [1998] for examples of other work that utilize this method.

9. These areas include the West (base group), Mountain, West North Central, West South Central, East North Central, East South Central, South Atlantic, Middle Atlantic, and New England regions.

10. To illustrate, Hispanic men with ten schooling years who do not speak English earned approximately 29% less { = [0.176 - (0.047 * 10)] * 100} than their otherwise similar peers from monolingual English households in 1979. The EDI of individuals with "moderate" English skills can be found by multiplying together the coefficient from Table II and the complete EDI earnings penalty. For example, men with ten schooling years who speak English "very well" earned about 3% less (= 0.107 * 29) in 1979 than their male counterparts from monolingual English households.

REFERENCES

Bergmann, Barbara R. "Does the Market for Women's Labor Need Fixing?" Journal of Economic Perspectives, Winter 1989, 43-60.

Borjas, George J. "The Economics of Immigration." Journal of Economic Literature, December 1994, 1,667-717.

Cameron, Deborah. Feminism and Linguistic Theory, 2nd ed. New York: St. Martin's Press, 1992.

Chiswick, Barry R. and Paul W. Miller. "The Endogeneity between Language and Earnings: International Analyses." Journal of Labor Economics, April 1995, 246-88.

Davila, Alberto, Alok K. Bohara and Rogelio Saenz. "Accent Penalties and the Earnings of Mexican Americans." Social Science Quarterly, 74(4), 1993, 902-16.

Ellis, Rod. The Study of Second Language Acquisition. Oxford: Oxford University Press, 1994.

Espenshade, Thomas J. and Haishan Fu. "An Analysis of English-Language Proficiency Among U.S. Immigrants." American Sociological Review, April 1997, 288-305.

Espinosa, Kristin E. and Douglas S. Massey. "Determinants of English Proficiency among Mexican Migrants to the United States." International Migration Review, Spring 1997, 28-50.

Glenn, Charles L. "What Does the National Research Council Study Tell Us About Educating Language Minority Children?" READ Abstracts, May 1997. Amherst, Mass.: Institute for Research in English Acquisition and Development.

Han, Mei, David Baker, Carlos Rodriguez and Peggy Quinn. A Profile of Policies and Practices for Limited English Proficient Students: Screening, Methods, Program Support, and Teacher Training. Washington, D.C.: U.S. Government Printing Office, 1997.

Henderson, Allison, Catherine Abbott and William Strang. Summary of the Bilingual Education State Educational Agency Program Survey of States' Limited English Proficient Persons and Available Educational Services: 1991-1992. Washington, D.C.: Office of Bilingual Education and Minority Language Affairs, August 1993.

Jasso, Guillermina and Mark R. Rosenzweig. "Language Skill Acquisition, Labor Markets and Locational Choice: The Foreign-Born in the United States, 1900 and 1980," in Migration and Labor Market Adjustment, edited by J. Van Dijk, F. Hendrik, H. W. Herzog Jr, and A.M. Schlottman. Dordrecht, The Netherlands: Kluwer Academic Press, 1989, 217-39.

Lopez, Mark Hugo and Marie T. Mora. "Bilingual Education and Labor Market Earnings Among Hispanics: Evidence Using High School and Beyond." Paper presented at the 19th Annual Association for Public Policy Analysis and Management (APPAM) research conference, Washington, D.C., November 1997.

McManus, Walter, William Gould and Finis Welch. "Earnings of Hispanic Men: The Role of English Language Proficiency." Journal of Labor Economics, 1(2), 1983, 101-30.

McManus, Walter S. "Labor Market Costs of Language Disparity: An Interpretation of Hispanic Earnings Differences." American Economic Review, 75(4), 1985, 818-27.

-----. "Labor Market Effects of Language Enclaves: Hispanic Men in the United States." Journal of Human Resources, 25(2), 1990, 228-52.

More, Marie T. "Did the English Deficiency Earnings Penalty Change for Hispanic Men between 1979 and 1989?" Social Science Quarterly, forthcoming, 1998.

Murphy, Kevin M. and Finis Welch. "Wage Premiums for College Graduates: Recent Growth and Possible Explanations." Educational Researcher, 18, 1989, 17-26.

-----. "The Structure of Wages." Quarterly Journal of Economics, February 1992, 285-326.

National Research Council. The New Americans: Economic, Demographic, and Fiscal Effects of Immigration, edited by J.P. Smith and B. Edmonston. Washington, D.C.: National Academy Press, 1997.

Park, Jin Heum. "Estimation of Sheepskin Effects and Returns to Schooling Using the Old and New CPS Measures of Educational Attainment." Working Paper No. 338, Department of Economics, Industrial Relations Section, Princeton University, August 1994.

Stevens, Gillian. "Nativity, Intermarriage, and Mother-Tongue Shift." American Sociological Review, February 1985, 74-83.

-----. "Sex Differences in Language Shift in the United States." Sociology and Social Research, October 1986, 31-36.

-----. "The Social and Demographic Context of Language Use in the United States." American Sociological Review, April 1992, 171-85.

U.S. Bureau of the Census. Resident Population of the United States: Middle Series Projections, 2001-2005, by Sex, Race, and Hispanic Origin, with Median Age. Washington, D.C., March 1996.