Age-Period-Cohort Models for Electoral Survey Data: An Application to British General Elections, 1964–2010

Abstract

This case study underscores the importance of using age-period-cohort models to disentangle life cycle, generational and election-specific effects when examining the determinants of long-term political change. I illustrate how these models can be used to study the impact on turnout of the decline in the competitiveness of British elections over the last 50 years, while controlling for other factors that may mask the relationship of interest. A key issue with age-period-cohort models is that standard statistical techniques and software packages are unable to handle relatively large data sets such as the one used in this study, with roughly 40,000 observation from thirteen UK general elections. During the course of the project I also had to make hard methodological choices to tackle some of the common challenges that arise when working with electoral surveys, such as turnout over-reporting, high proportion of missing values, and the difficulties in operationalizing relevant – dependent and independent - variables. A fundamental lesson from this case study is that careful consideration of alternative measurement strategies and estimation methods is crucial for the application of age-period-cohort models to repeated cross-sectional electoral surveys, as decisions in these areas can radically affect the substantive conclusions drawn from the data. At a more practical level, the study highlights that researchers may not always be able to rely on canned estimation routines to fit their models. Being able to tailor the estimation approach to the data at hand may thus be critical for the project’s success.