Heterogeneity in the UK Labour Market: Using Machine Learning To Test Macroeconomic Models

We use machine learning techniques to investigate the main sources of heterogeneity in the UK labour market. A recent theoretical literature argues that accounting for differences in productivity between workers can help resolve some long-standing issues in the analysis of labour markets. This literature assumes that heterogeneity only reflects differences in productivity and not other types of heterogeneity, such as differences in age, sex, ethnicity, location, or the type of contract a worker has. We test this assumption. Applying a clustering algorithm to individual-level data, we find that cluster membership is mainly driven by productivity. This finding provides empirical support for the recent theoretical
literature. Our results also imply that differences in productivity are systematic rather than purely random; that there is a strong relationship between education and productivity and that UK labour markets are not fully segmented.