In the most frequently used microdata sets, over a quarter of all respondents now refuse to answer some questions about their incomes. The Census Bureau has dealt with this problem, which has been increasing in severity over time, by imputing incomes of non-respondents. Their imputation procedure, called the "hot deck," essentially matches nonrespondents with demographically similar donors. In this paper we evaluate the census imputation methodology and raise some questions. First, the census procedure is tied to commonality of events in the population rather than the more appropriate informational content of regressors. Clearly, the census procedure severely understates income in certain occupations. Because it is based on the apparently invalid assumption that income does not affect reporting propensities, it most likely understates average incomes as well.

This paper (1) presents estimates of minimum wage effects on employment of teenagers 14-15, 16-17 and 18-19 years, and (2) decomposes these estimates into scale and substitution components for calculating effects of differential minima. In 1972 the House of Representatives approved an amendment to the Fair Labor Standards Act calling for a youth differential, but one was not included in the Senate version. These bills died when the House refused to submit to conference. In 1973 the Administration proposed another amendment containing a youth differential, but the amendment passed by Congress later that year eliminated it. Although attempts to enact a youth differential have failed,it is clear that support for these measures comes from a consensus concerning the relatively adverse effects of existing minima on teenage employment. It is also likely that the question of differential minima is not dead and it would be nice to have estimates of what effects might be if a differential were enacted. The point of departure for all minimum wage studies has been that those most adversely affected are those who in the absence of the minimum would have earned the lowest wage. In comparing teenagers to adults, there is fairly consistent evidence that minimum wages reduce teenage-adult employment ratios.' But there is less evidence among groups of teenagers themselves. The Labor Department Survey (1970) contrasted males and females, white and non-white, for employment of those 16-17 and for those 18-19. Results are mixed. For white males, estimates conform to expectations of more adverse employment effects for younger workers, but results for white females are inconclusive and there is no relationship for non-white teenagers. Jacob Mincer (1976) and Nori Hashimoto and Mincer (1970) combine all ages 16-19 into one group but distinguish whites from non-whites and find more adverse effects for non-whites. Similarly, Marvin Kosters and Finis Welch (1972) treat ages 16-19 as a single class and distinguish employment by sex and race. More adverse effects are reported for non-whites than for whites and for females than for males. Each of these studies uses time series estimates of teenage employment from Current Population Surveys. These data contain large sampling errors especially for age, sex, and race partitions.2 Students and part-time workers are not distinguished, wages are not available and because the data are for national aggregates, state minimum wage laws are ignored. Here, the data are from the 1 in 100 Public Use Sample of the 1970 Census and refer to teenage employment in the week before the Census was taken. Individual observations are aggregated to state totals. Nationally, students account for half of total teen employment but only one-third of hours worked and females work 95% as many hours as males. Aggregation weights reflect these differences in hours worked. The Census does not contain reliable wage information, and as is true of all previous studies, there is a problem in estimating legislative effects on costs of teenage employment. Cross-state observations enable us to include state wage laws in our estimates of these costs. Section II describes our procedure for inferring legislative effects on teenage employment costs and provides estimates of associated changes in employment. Section III provides empirical results for a decomposition of estimated effects into scale and substitution components and gives estimated effects of a 20% youth differential extended first to those 14-15 and then to those 14-17 year old. A summary follows. Received for publication July 2, 1976. Revision accepted for publication March 4, 1977. * University of California, Los Angeles, and The Rand Corporation. We are grateful to Ernst Stromsdorfer for suggesting this topic and Dennis DeTray and James P. Smith for their helpful comments. Support for this project was provided by a contract from ASPER/USDOL. 'See, for example, the paper by Mincer (1976) and the Labor Department Survey (1970). 2See Welch (1974), the comment by Siskind (1977) and reply by Welch (1977) for discussions of some of the peculiarities of these data.

The size and rapid growth of the Hispanic population in the United States raise many questions about the effects of language on the process of economic assimilation. This paper uses data on earnings for 1975 to explore the role of English language proficiency in the assimilation of Hispanic men into the U.S. labor market. The results are strong: differences associated with English language skills explain virtually all of the Hispanic wage differences usually attributed to ethnicity, national origin, and time in the United States. The analysis is completed by relating the measure of English language proficiency to assimilation-related variables.

In 1940 the average worker in the United States had 9.2 years of schooling; by 1990 the average had increased to 13.3 years. Although education levels show a dramatic upward trend, there is no corresponding long-term trend toward lower economic rewards from schooling. In fact, the return to schooling increased in three of the five postwar decades and is higher in 1990 than it was in 1940.1 In our opinion, the tremendous growth in the supply of more-educated labor with no corresponding reduction in relative wages is evidence that the demand for education increased greatly between 1940 and 1990 (see Murphy and Welch [1992] for a similar argument for the 1970-1989 period). In this paper we examine changes in the occupational wage and employment structure at ten-year intervals from 1940 to 1990 in an attempt to identify and quantify some of this increased demand for skill. Our analysis is based on the five Decennial Census files from 1940 to 1980 and the Current Population Surveys for 1978-1982 and 1989-1991. Not surprisingly, our findings support the view that the demand for skill increased significantly over the 1940-1990 period. Over the full period and for each of the five decades, we find that employment shifted from less-educated and lower-paid occupations toward more-educated and higher-paid occupations. We interpret this shift in employment patterns as evidence of a shift in demand. We do not find that the demand for skill grew particularly rapidly during the 1970's and 1980's, a period when wage inequality expanded in comparison to the three earlier decades when the trend was either toward less wage inequality or when wage inequality increased only modestly.

This article considers the representativeness of the Panel Study of Income Dynamics (PSID) over its 14-year history from 1968 to 1981 given the dynamics of entry and exit from the panel. By 1981, 40% of the original members had left the sample and were replaced by new entrants who joined either existing households or new households being formed by members of the original sample. We consider the distribution of demographic characteristics and earnings equations over time, and we compare the PSID with the Current Population Survey (CPS). By either approach we find little evidence that the PSID has become unrepresentative.

There are two sections to this paper. The first summarizes relative black-white earnings and wage ratios by schooling class and estimated time out of school (work experience) for both 1960 and 1970. The second presents a partial accounting giving order-of-magnitude estimates of import for some sources of black-white income differences as of 1970 and of changes between 1960 and 1970.