Showing papers on "Panel data published in 1979"

Analyzing panel data

Abstract: An introduction to a variety of techniques that may be used in the analysis of data from a panel study -- information obtained from a large number of entities at two or more points in time The focus of this volume is on analysis rather than problems of sampling or design, and its emphasis is on application rather than theory

Using Panel Data to Improve Estimates of the Determinants of Educational Achievement.

Abstract: Single equation models of educational achievement that have been estimated by OLS have tended to ignore the effects of omitting relevant variables. These omitted variables may lead to biased and inconsistent estimates. Further, past research frequently obscures the correct interpretation of the estimated coefficients. This paper considers three frequently estimated models of educational achievement: the case when only cross-sectional data are available, the case when "pre" and "post" measures of achievement are available, and the case where one has more complete panel data. Starting from a general cumulative model of achievement, we outline sufficient assurmptions to produce readily interpretible OLS estimates which have good statistical properties. With only cross-sectional data, these assumptions are almost certainly not met in most research. With two measures of achievement, useful estimates with good statistical properties can be obtained if we make quite restrictive assumptions concerning the relationships among the explanatory variables. With panel data, simple OLS can produce coefficients with good statistical properties underfairly general conditions. The use of multiple measures, however, may lead to multicollinearity which one must balance off against the gain in efficiency.

The SRC Panel Data and Mass Political Attitudes

Abstract: One of the richest data sources for the study of public opinion is the Survey Research Center's panel study conducted in the late 1950s. Because the SRC interviewed its national panel of Americans three times over a four-year period, the SRC panel data allows the analysis of changes in survey responses over time. The most remarkable discovery from the SRC panel was that panelists changed their reported opinions on policy issues with considerable frequency when asked the same policy questions in different years. Moreover, the amount of observed change in the individual responses varied little with the time interval between responses. That is, the correlations between responses to the same issue item in 1956 and 1958 or in 1958 and 1960 (two years apart) were almost as low as the correlations to the same issue item in 1956 and 1960 (four years apart).

Employment trends of young mothers and the opportunity cost of babies in the united states

Abstract: The central concept of microeconomic theories of fertility is opportunity cost—the product of wife’s employment lost due to childbearing and the value of her employment. This paper presents a model for analyzing opportunity cost using panel data. The average loss of employment attributable to a second- or higher-order birth, calculated at around age 2, is over 400 hours per year. This time cost represents an income loss of about $1050 in 1969 dollars. Time cost is independent of such demographic factors as birth order and age of oldest sibling. Neither does time cost depend on husband’s wage rate or wife’s education or potential wage rate. This indicates that many microeconomic models of fertility have been seriously misspecified. The paper also compares results from static and dynamic models, explores possible problems due to simultaneity bias, investigates the relationship between changes in employment (including time cost) and initial employment level, and identifies the difficulties of theorizing about opportunity cost.

Population Heterogeneity and Inference from Panel Data on the Effects of Vocational Education

Abstract: This paper considers a model of earnings over time which incorporates individual effects and time effects without assuming that these effects are orthogonal to the variable of primary interest. The central coefficient is the effect of participation in a Manpower Development and Training Act training program on the earnings of trainees. Since the training status of (some) individuals in the sample changes during the period of the sample, both pre- and posttraining contrasts and trainee-nontrainee contrasts in earnings can be made. An estimate of the cross-section bias in a training coefficient can be made directly. The extent of the analogous bias in the education coefficient in regression studies is a point of current debate. The cross-section bias in the sample analyzed is large, and the estimated effect of training is small and positive.

The Distribution of Lifetime Labor Force Participation of Married Women: Comment

Abstract: Heckman and Willis (1977) apply a sophisticated methodology for the treatment of panel data on labor force participation of married women. Without entering the methodological question about criteria for an optimal statistical treatment of such data, we take issue with the substantive finding that "the distribution of participation probabilities is U shaped, indicating that most women have participation probabilities near zero or near one" (p. 27). The "finding" is remarkable for omitting any qualification concerning the length of the period over which the distribution is supposed to hold. In other words, women found working in a given period tend to work continuously, while women not found working in that period are highly unlikely to be found working at any time. Partly because of fixed costs of working, this hypothesis is plausible in the short run, and the several years over which the authors empirically confirm the hypothesis may be viewed as a short run. Partly it might reflect a cyclical effect. Business cycles do affect labor force turnover even in the short run, but they do it mainly by accelerating or retarding entries into the labor force (Mincer 1973), thereby affecting the probability that nonworking women in the base period (say, 1967) will work in subsequent years (up to 1972 in the Heckman and Willis sample). Because of the cyclical downturn which started in 1969 the probability of reentry of women not working in 1967 or 1968 was diminished. The proposition that participation probabilities are U shaped should be tested with data covering a long span, where fractions of time worked empirically approximate probabilities. Albeit retrospective, such data exist in the National Longitudinal Survey (NLS)