Ethnic parity in labour market outcomes for benefit claimants

A significant gap exists in the UK between the employment rate for Ethnic Minorities and that for Whites. From a policy perspective, it is important to know whether this gap is due to differences in the characteristics of White and Ethnic Minority groups (which reduce the employability of Ethnic Minority groups relative to Whites) or whether it results from some form of discriminatory behaviour in the labour market. In this paper, we use administrative data to estimate ethnic differences in employment and benefit receipt amongst individuals who began claiming a Jobcentre Plus benefit in 2003. In contrast to much of the previous UK literature, we use a number of different quantitative techniques to estimate this gap, and show that in a lot of cases the estimates obtained are very sensitive to the techniques used. We argue that for the questions we are interested in and the data we have, propensity score matching methods are the most robust approach to estimating ethnic parity. We compare this preferred approach with estimates derived using alternative approaches commonly used in the literature (generally regression-based techniques) to determine the extent to which more straightforward methods are able to replicate those produced by matching. In many cases, it turns out not to be possible to calculate satisfactory quantitative estimates even with matching techniques: the characteristics of Whites and Ethnic Minorities are simply too different before the Jobcentre Plus intervention to reliably estimate the parameters of interest. Moreover, for a number of the groups, results seem to be very sensitive to the methodology used. This calls into question previous results based on simple regression techniques, which are likely to hide the fact that observationally different ethnic groups are de facto being compared on the basis of parametric extrapolations. Two groups for which it was possible to calculate reasonably reliable results are incapacity benefit (IB) and income support (IS). For these groups we find that large and significant raw penalties almost always disappear once we appropriately control for pre-inflow background and labour market characteristics. There is also a good degree of consistency across methodologies.