Showing papers on "Panel data published in 1985"

Benefits and limitations of panel data

Abstract: Observations for a number of cross-sectional units over time have become increasingly available. The new data sources enable econometricians to construct and test more complicated behavioral models than a single cross sectional or time series data set would allow. The availability of new data sources, however, also raises new issues. In this paper we review some basic econo- metric methods that have been used to analyze such data sets. We also indicate areas of research where panel data may be useful.

The Permanent Income Hypothesis and Consumption Durability: Analysis Based on Japanese Panel Data

Abstract: The permanent income hypothesis with durability of commodities is tested on a panel of about 2,000 Japanese households for several commodity groups. Under static expectations about real interest rates and for some class of utility functions, consumption, which is a distributed lag function of current and past expenditure, follows a martingale. Main empirical results are (i) the durability of commodities usually classified as services is substantial, (ii) the hypothesis applies to about 85 percent of the population consisting of wage earners, and (iii) income changes explain only a small fraction of the movements in expenditure.

Models of well-being and ill-being

Abstract: The purpose of this paper is to present models of well-being and ill-being which integrate the effects of different types of independent variables. Using the more powerful techniques provided by the LISREL software, the paper replicates and extends analysis previously reported in SIR (Headey, Holmstrom and Wearing, 1984a).

Unemployment and Mental Health: Some Evidence from Panel Data

Abstract: Microdata are used in this paper to analyze the effects of unemployment on mental health. The analysis is done in two steps. First, cross-section data of labor force participants are analyzed. It appears that the unemployed have worse mental health than the employed. Next, panel data are used to control for "fixed" effects, that is, unobserved omitted variables that are constant over time. The model is also specified to allow both the occurrence of and duration of unemployment to affect mental health. Then we cannot reject the hypothesis that there are no effects of unemployment on mental health. However, some sensitivity tests indicate that the precision of our estimates is rather low.

Employment While in College, Academic Achievement and Post-College Outcomes: a Summary of Results

Abstract: This paper uses panel data that cover the 1972-1979 period obtained from the National Longitudinal Survey of the High School Class of 1972 to study how male college students' employment while in college influences their academic performance, persistence in school, decisions to enroll in graduate school, and post-college labor market success. The analytic framework employed treats in-school employment as endogenous and determines persistence by a comparison of expected utilities.

The use of an autoregressive model for the analysis of longitudinal data in epidemiologic studies

Abstract: Korn and Whittemore1,2 have presented methods for analyzing longitudinal data where the number of observations per individual is large relative to the number of variables considered for each subject. However, this is often not the case in epidemiologic studies, since one usually collects data at relatively few time points, and the quantity of data collected for each individual at each time point is typically extensive. We present here an autoregressive model for analyzing longitudinal data of this type for the case of a continuous outcome variable. Some of the important features of this model are that one can (1) in the same analysis, consider both independent variables that are time-dependent and those that are fixed over time, (2) partially use data for an individual where some examinations are missing, (3) assess relationships between changes in outcome and exposure over short periods of time, (4) use ordinary multiple regression methods. Anderson3 has considered this type of model, but, to our knowledge, the model has never been applied to biostatistical problems. We illustrate these methods with data from a longitudinal study that seeks to identify the role of personal cigarette smoking on changes in pulmonary function in children.

A Probabilistic Model For Testing Hypothesized Hierarchical Market Structures

Abstract: In this paper a methodology for hierarchical market structure analysis is derived and illustrated. A probabilistic model is developed which provides a general, flexible framework which can be used to test hypothesized hierarchical market structures. Because the general probabilistic model can be translated in terms of a restricted latent class model it can be easily implemented. Among its other benefits, the proposed methodology, utilizing panel data, tracks households over a number of switches, furnishes explicit statistical tests for alternative hypothesized tree structures, as well as individual parameters, and can potentially identify clusters of households who structure the market in similar ways.

Migration and wage growth: a human capital approach

Abstract: The purpose of this paper is to analyze migration as an investment in human capital using panel data. [The authors] focus on two issues. The first is the economic motivation for migration and the reasons some movers receive high returns relative to other migrants. The second issue and the major focus of the paper is a test of the human capital model explaining wage profiles of nonmovers first-time and repeat migrants. The data are from the National Longitudinal Survey of Young Males 1966-1971 cross-section and concern the United States. (EXCERPT)

Housing Purchases and Transitory Income: A Study with Panel Data

Abstract: In this paper we explore the role of transitory income in the housing purchase decision. Because of moral hazard considerations banks typically require downpayments to be financed internally, hence transitory income is potentially important in overcoming the downpayment constraint. Using a novel approach to the measurement of permanent and transitory income, we estimate a two-stage model in which households decide whether to purchase housing in the first stage and the quantity to be purchased in the second. The results indicate a significant role for transitory income in both decision stages.

Stochastic Panel-Data Models of Urban Shopping Behaviour: 4. Incorporating Independent Variables into the NBD and Dirichlet Models:

Abstract: In this paper the authors describe the way in which the NBD and Dirichlet models of consumer behaviour may be extended to include independent or exogenous variables. Such methods allow the predictions obtained from these models to be disaggregated, that is, to be made conditional on the values of the exogenous variables. Empirical examples are presented of repeat buying of a branded good and at an individual store (for the NBD model), and of multistore purchasing (for the Dirichlet model).

Collecting Panel Data in Developing Countries: Does It Make Sense

Abstract: This Working Paper reviews a number of aspects of the collection and use of panel data from households in developing countries. Sampling issues are discussed in Section 1. The authors conclude that there are likely to be real, if modest, benefits from incorporating some panel element into household survey data collection in developing countries. The recognition that panel data are likely to be subject to substantial errors of measurement does not invalidate this conclusion. Section 2 discusses the measurement of income dynamics, an issue that cannot be addressed without panel data. Recent research using U.S. data is reviewed to show that comparable work for developing countries would add an important dimension to discussions of poverty, inequality, and development. It is in the third area of review, that of econometric analysis, that the real benefits of panel data appear most fragile. While it is true that panel data offer the unique ability to deal with the contamination of econometric relationships by unobservable fixed effects, the presence of measurement error can compromise the quality of the estimates to the point where it is unclear whether cross-section or panel estimators are superior.

Employment While in College, Academic Achievement and Post-College Outcomes: A Summary of Results

Abstract: This paper uses panel data that cover the 1972-1979 period obtained from the National Longitudinal Survey of the High School Class of 1972 to study how male college students' employment while in college influences their academic performance, persistence in school, decisions to enroll in graduate school, and post-college labor market success. The analytic framework employed treats in-school employment as endogenous and determines persistence by a comparison of expected utilities.

The Covariance Structure of Earnings and Income, Compensatory Behavior and On-the-Job Investments

Abstract: Observationally alike individuals who make different choices about on-the-job investments should have earnings profiles that differ in systematic ways. In particular, investments in non-specific human capital should result in lower initial earnings but higher earnings growth rates. Human capital models of this sort admit testing, then, by examining the covariance between the level of earnings and the growth rate of earnings. This paper reports estimates of this covariance using the sample covariance among income observations across time for the same individuals. The sample covariances are drawn from the Utah Panel Data, a panel of some 16,000 households with income and wealth observations at various intervals over the period from 1850 to 1900. The parameter of interest is negative. This estimate is robust to various specifications of the model. I also reexamine earlier work by Lillard and Weiss and Hause, who use data on earnings, and conclude that there is strong support for the on-the-job investment hypothesis using data from thre equite different sources covering different economies and different time periods.

Evaluating the effects of changes in AFDC: Methodological issues and challenges

Abstract: Analysts are commonly called upon to perform the difficult task of evaluating the effects of specific changes in public policy upon the behavior of individuals, such as a change in the provisions of the Aid to Families with Dependent Children (AFDC) program relating to the benefits of those in the program who find work. When charged with such a task, analysts commonly try to answer the question by tracing the behavior of a fixed panel of individuals, comparing the experience of the group before and after the change in policy. That approach, however, risks major errors; in the case of the AFDC program, for instance, changes in the work benefit provisions affected the decisions of some who might have come into the program, a consequence that would not be picked up by a fixed panel of initial recipients. Cross-sectional data drawn independently from a general population at points in time before and after a policy change can often provide a more valid measure of the effects of the policy change than can panel data; moreover, cross-sectional data are usually less expensive and more readily available.

Estimation and testing of dynamic econometric models from panel data.

Abstract: This research develops methods of estimation and test statistics of dynamic single equation models from panel data when the errors are serially correlated. It is assumed that the number of time periods is fixed while the number of cross-section observations is large. This makes it possible to consider prediction equations of the initial observations based on the exogenous variables corresponding to all periods available in the sample, as well as to leave unrestricted the covariances of the prediction errors with the remaining errors in the model. The concentrated likelihood function is derived both for cases where the prediction error is left unrestricted and where it is assumed to have the marginal distribution of the stationary process. The performance of maximum likelihood methods is investigated, either for correct models or under several misspecifications, by resorting to Monte Carlo methods using antithetic variates. Dynamic models from panel data can be seen as a specialisation of a triangular system with covariance restrictions. In this context, the asymptotic distribution of the estimators that maximise the gaussian likelihood function is derived when normality holds and also when the errors are non-normal. In particular, it is shown that in the latter case the estimator that takes into account the covariance restrictions is not generally more efficient than the estimator that leaves the co-variance matrix unrestricted. The possibility of obtaining consistent estimates of the unrestricted intertemporal covariance matrix is used to develop test statistics of covariance restrictions arising from various random effects specifications. A Wald test and a minimum chi-square test, which are robust to the non-normality of the errors, and appropriate asymptotic probability limits for the quasi-likelihood ratio test are proposed. Monte Carlo experiments are conducted to study the performance of these test criteria. In order to illustrate these procedures, QML estimates of dynamic earnings functions from the Michigan Panel are obtained.-4-Joint minimum distance estimators of slope and covariance parameters are defined that are generally efficient relative to QML estimators when normality is not imposed and the covariance matrix is restricted. Finally, it is shown that there exist separate minimum distance estimators of the covariance parameters and generalised least squares estimators of the slope parameters that are efficient. A simulation is also carried out to examine the performance of these methods.

The Covariance Structure of Earnings and Income, Compensatory Behavior and On-the-Job Investments

Abstract: Observationally alike individuals who make different choices about on-the-job investments should have earnings profiles that differ in systematic ways In particular, investments in non-specific human capital should result in lower initial earnings but higher earnings growth rates Human capital models of this sort admit testing, then, by examining the covariance between the level of earnings and the growth rate of earnings This paper reports estimates of this covariance using the sample covariance among income observations across time for the same individuals The sample covariances are drawn from the Utah Panel Data, a panel of some 16,000 households with income and wealth observations at various intervals over the period from 1850 to 1900 The parameter of interest is negative This estimate is robust to various specifications of the model I also reexamine earlier work by Lillard and Weiss and Hause, who use data on earnings, and conclude that there is strong support for the on-the-job investment hypothesis using data from thre equite different sources covering different economies and different time periods

Two-wave panel models and causal analysis: A criminological example

Abstract: This paper illustrates some general analytic procedures which may be helpful in assessing various assumptions underlying alternative models in the causal analysis of panel data The role of two basic mechanisms of change, temporal inertia and entropic decay, is specifically examined The models developed are evaluated using data on changes in the politicalization of 172 male felons over a six-month period following their incarceration in a maximum security institution in the southeast Several issues and problems in drawing causal inferences from panel data are addressed

Labor and Investment Demand at the Firm Level: a Comparison of French, German and U.S. Manufacturing, 1970-79

Abstract: We investigate how labor and investment demand at the firm level (gross as well as net and replacement investment separately) differs in French, German and US manufacturing, and has changed since the 1974-75 crisis We use three consistent panel data samples of large firms for1970-79, and rely on simple models of the accelerator-profits type We find that the accelerator effects and the profits effects did not vary much between 1970-73 and 1976-79, and were quite comparable in the three countries, the former being of a more permanent nature and the latter more transitoryTo a large extent these effects account for the important changes and differences in labor and investment demand between the two subperiods and across the three countries

The Performance of Coal-Burning Electric Generating Units in the United States: 1960-1980

Abstract: This paper studies determinants of the thermal efficiency and reliability of coal-burning electric generating units using a new, comprehensive, unbalanced panel data set. For two major technological groups consistent and efficient estimates are obtained of the effects of unit age, vintage, scale, operating practices, and coal quality. Large utilities integrated into design and engineering appear to obtain superior unit performance. Implications of our results for life-cycle generation costs, regulators' performance norms, and the complex pattern of technological change in this industry are discussed.