Longitudinal and cross-sectional data collection and analysis in the study of crime and delinquency

Abstract

In response to recent debates about the utility of longitudinal as opposed to cross-sectional designs in the study of crime and delinquency, we examine empirical evidence on the extent to which longitudinal and cross-sectional data may be used interchangeably without altering substantive conclusions. We distinguish between longitudinal or cross-sectional methods of data collection and longitudinal or cross-sectional data, and acknowledge that longitudinal data may be collected by using a cross-sectional method. We examine the extent to which longitudinal data collected cross-sectionally are compatible with longitudinal data collected in a prospective longitudinal design. We then compare correlations and predictive models that use cross-sectional and longitudinal data to study the same substantive problem. We conclude with an examination of the issue of temporal order as a test of competing hypotheses. We find that longitudinal data collected in a cross-sectional design do not consistently produce the same results as longitudinal data collected in a prospective longitudinal design, and that longitudinal and cross-sectional data do not consistently yield the same substantive results in correlation and prediction models. Longitudinal data allow us to generate strong tests of competing hypotheses which would be impossible without longitudinal data. We conclude that cross-sectional data, or longitudinal data collected in cross-sectional designs, are inadequate substitutes for prospectively collected longitudinal data in the study of crime and delinquency.