Conference Paper International Association for Research on Income and Wealth Conference
Are Short Panels Good Enough to Study Intergenerational Mobility? Evidence from the BHPS
23 Aug 2004
Data requirements to estimate the extent of intergenerational mobility are stringent. For all studies that use standard household panel surveys with repeated observations on matched parent-child pairs, one of such requirements is that the data are sufficiently long (at least twenty years, and preferably thirty or forty years). This sharply restricts the set of data and countries that can be analyzed and, even when long panels of data are available, naturally leads to small samples (because of the sample attrition present in such data). The aim of this paper is to assess how severe the problem of short panels can be for the estimation of intergenerational mobility. For this purpose we introduce a new methodology which we apply to a short panel using the first eleven waves of the British Household Panel Survey (1991-2001).
Our method exploits the fact that all respondents in the BHPS are asked to report the occupation of their parents when they were aged 14. This information is based on retrospective questioning but does not rely on the matching of parents to children, that is, on the possibly highly selected group of children who are observed to live with their parents during the panel years. Using a continuous index of occupational prestige that relates strongly to wages for the full sample, we then estimate intergenerational elasticities in occupational prestige that are free of matching selection bias. Clearly, by matching parents and children over the panel years we obtain a smaller selected sample, with which we can analyze intergenerational mobility in incomes and earnings (as long as both parents and children report valid information on such variables) and not only in occupational prestige. This selected sample however may suffer from matching selection bias. To gauge the size of this bias we compare the intergenerational elasticities in occupational prestige obtained from the selected sample to those obtained earlier from the full sample.
We then evaluate two approaches that correct for this selection bias generated by the possibility of matching children to their parents. The first belongs to the general class of Heckman-type sample selection correction, which we perform using both parametric and semi-parametric estimation procedures. The second is within the class of models based on propensity score estimation, which we estimate using both matching and weighting procedures. Finally we conduct some sensitivity analyses to assess whether the elasticities estimated on the selected sample are robust to varying the length of the panel or the age range of children.