Browse IFS
Publication types
Methods

Longitudinal data methods

Understanding individuals’ responses to changes in environment and to changes in risk and uncertainty requires structural analysis of sequences of individuals’ decisions and outcomes. Models currently used for analysis of longitudinal data employ such strong restrictions that our understanding of important aspects of behaviour is very limited. Research is needed to enable analysis using models with essential non-linearities, models with multiple sources of heterogeneity and models suitable for application in complex dynamic decision environments where people’s computational capabilities may be limited.

Research on longitudinal methods is needed to deliver tools that will maximise the returns from investment in the growing stock of high quality UK longitudinal data sources including the BHPS, ELSA, the British and Millennium Cohort Studies and Understanding Society. Valuable longitudinal data from commercial sources, such as panel data on auctions (from an anonymous online UK automobile auction firm) and panel data on consumer expenditures (Kantar World Panel) are becoming available.

Research projects
This research uses indirect inference to model earnings dynamics with participation to assess how non-random selection into work affects estimation of earnings dynamics.
Unobserved heterogeneity is important in econometrics. People who look alike in terms of education, income, or other variables, make different choices about consumption, and savings. etc. Economic theory provides no guidance as to how unobserved heterogeneity should be modelled nor how important it might be.
A key consideration when studying government policies towards higher education is how do the decisions of teenagers, the decisions of parents, the general economic environment and government policies interact to determine college enrolment decisions.