SYLLS: synthetic data estimation for the UK longitudinal studies

Belinda Wu, Adam Dennett and Nicola Shelton, University College London

(Project no. 30158)

The eventual aim is to create a dataset which closely resembles the three UK Longitudinal Studies but contains no real longitudinal data - this will allow researchers to try out code and hypotheses on a freely-available data source before applying to use the real LS'.

We aim to generate transition probabilities from 1991 to 2001 across a range of commonly-used LS variables, starting with individual rather than household characteristics. We will take a baseline microdata set from the SARs, and apply the transition probabilities deduced from the LS to it, with a view to obtaining a distribution at 2001 which reflects the distribution of the real LS variables. We later aim to extend this process to the Scottish and Northern Irish LS'. This will be done at UCL using intermediate outputs from the LS and employing specially-written microsimulation software (these outputs will be probabilities rather than raw data).