Labour market heterogeneity : wage determination and unemployment duration.

The last twenty years has witnessed an unprecedented growth in the development and application of non-competitive models of labour market behaviour. These models have emerged in response to the apparent failure of the competitive paradigm to adequately explain the observed stability of the wage structure and the persistence of unemployment in the past two decades. They also help to explain observed empirical regularities that lie outside of the orthodox domain. Identifying the role of competitive and non-competitive forces in the determination of labour market outcomes necessitates that differences between individuals are suitably accounted for. Our ability to control for such differences is limited in both cross-section and time-series data. Panel data, however, can provide new insights in that both observed and unobserved differences between individuals can be accounted for in empirical modelling.

This thesis addresses the importance of individual heterogeneity in the determination of wages and unemployment durations in the UK using longitudinal data drawn from the first eight waves of the British Household Panel Survey (BHPS). The BHPS provides comprehensive information at the individual and household level. It also provides extensive information on individual labour market activity both during the panel and retrospectively from labour market entry. We utilise both aspects of this data to fully explore the impact of individual heterogeneity. Our findings suggest that individual heterogeneity is of critical importance to observed labour market behaviour. We show that over 90 percent of the variation in individual earnings can be explained by observed and unobserved individual heterogeneity. We also demonstrate that the failure to incorporate individual heterogeneity results in a significantly upward bias in the magnitude and dispersion of inter-industry wage differentials. The neglect of individual heterogeneity is also observed to significantly bias econometric estimates that measure the conditional probability of exit from unemployment. We show that the failure to incorporate unobserved individual characteristics results in a significant downward bias in the hazard rate from unemployment, thus implying spurious negative state dependence.

The central conclusion of the thesis is that empirical work which is not founded on techniques which utilise panel data can give rise to seriously misleading conclusions regarding the operation of labour markets, and, in particular, the nature of the wage determination process and the likelihood of exit from unemployment.