Paper was presented at the Joint Statistical Meetings, Las Vegas, Nevada, August 1985.

The Survey of Income and Program Participation is designed to provide a dynamic, longitudinal view of educational, labor force, and demographic activity both during the two and one-half years of data collection and for individuals' lifetime experience prior to the initial data gathering point. This paper is the first in a series which will explore the applicability of hazards modeling techniques for the analysis of demographic, sociological and economic transitions using SIPP data.

Hazards models examine the time paths of events that occur to individuals. These models, also known as multivariate life tables or life tables with covariates, analyze the effects of independent variables upon the time dependent risk of experiencing an event, e.g.; death, divorce, marriage, or childbirth. Hazards models present a view of the factors related to both the number and the timing of such transitions of interest.

Information collected in the SIPP lends itself to studies of many types of transitions. Changes in family composition, marital status, program participation, labor force participation, and other such life events, are substantially documented over the sample period. Also, detailed persona1 hist1lries for individua1s covering 1abor force experiences, program participation and household composition changes are collected for the same individuals. For the purpose of this study the event of interest is the transition from the state of unemployment to employment. Differences in probabilities of moving from one state to another, and the timing of such movements by characteristics of individuals will be examined. Covariates in the model need not be limited to characteristics of an individual, which are unchanging over time, such as sex and race, but may also include characteristics measured at the beginning of the study and measures of change in these characteristics as ascertained over time. Hazards modeling techniques are also ideal for dealing statistically with problems of right-censored data, that is, the problems caused by the varying time at risk for individuals in the study due to sample attrition or the ending of the study.

At the present stage of data completion, information is limited to a four month period, with weekly data on labor force status in particular. Using this information it is possible to pinpoint the week in which an individual experienced a transition from one state to the other.

This study explores the adequacy of the SIPP for this type of analysis and suggests ways in which it could be made more suitable. This project also lays the groundwork for future application as more information becomes available. The work history fixed topical module, for example, provides a useful event history suitable to these purposes. Beyond the need for information about duration of the state in which an individual enters the sample, it is also desirable to know previous histories of event occurrences. Knowledge of the number of 1ike events experienced by each individual is important for the employment of hazard rate models.