Chapter 8: Longitudinal data analysis in the sociology of education: key concepts and challenges
Restricted access

In the sociology of education, surveys collecting longitudinal data are on the rise. This chapter gives an overview of survey designs and instruments that produce either episode data or panel data. Episode data is characterized by origin and destination states with a starting and ending point. Panel data is a collection of information from the same individuals repeatedly at different points in time. The two data types are distinct and require different techniques of data analysis. For episode data, survivor functions and transition rate models are generally the first choice. As non-parametric survivor functions may create problems when there are competing destination states, I also discuss cumulative incidence functions, an ideal but rarely used alternative. With respect to parametric transition rate models, the chapter focuses on time-varying covariates. Panel data analysis is discussed in the framework of multilevel analysis, which entails repeated measurements of the same individuals on the same variables at the lowest level, followed by individual-specific information at the next level and in some cases also context information on higher levels. A common model for analysing hierarchical data is the growth curve model, which focuses on gains or losses over time, using indicators such as socioeconomic background. Most studies using this type of model fall short, however, in capturing changes at the lowest level as a source of potential effects. I therefore address an increasingly popular model type in the social sciences but one that is still used rarely in the sociology of education: fixed effects panel regression. It focuses solely on changes at the individual level and eliminates all time-constant, observed or unobserved higher-level characteristics. The final section discusses comparative advantages and disadvantages of methods for analysing episode and panel data, and comments on whether current longitudinal surveys provide suitable data for these methods.

You are not authenticated to view the full text of this chapter or article.

Access options

Get access to the full article by using one of the access options below.

Other access options

Redeem Token

Institutional Login

Log in with Open Athens, Shibboleth, or your institutional credentials

Login via Institutional Access

Personal login

Log in with your Elgar Online account

Login with you Elgar account
Edited by Rolf Becker
Handbook