There are many ways to measure intergenerational mobility. One prominent approach is to measure the strength of the association between the lifetime earnings of fathers and that of their sons. This makes sense because lifetime earnings are such a fundamental aspect of economic wellbeing. Another strength of such measures is that they facilitate simple comparisons between countries, and of change over time for individual countries. The focus on males is a clear limitation of this approach. It does however avoid the complications associated with changing female labour force participation over time, and differences in female participation rates between countries.
That said, the data required to directly estimate this association are simply not available for many countries, including Australia. Such data will be available eventually for Australia if the Household, Income and Labour Dynamics Australia (HILDA) panel survey continues to run. But for now, there is no available data source where the earnings of a representative sample of men can be matched to their fathers’ earnings, even for one point in time, let alone for the whole life course.
In this context, one can impute fathers’ earnings on the basis of their occupation, which is observed. But this comes with major measurement error, which almost certainly biases the estimate towards zero due to attenuation bias. It also makes international comparisons problematic. To navigate this issue, one can use the same approach with U.S. data, for which the extent of bias will perhaps be similar, and then to adjust the Australian estimates by the extent to which the U.S. estimates differ from an external benchmark that is estimated using the best U.S. data available. This is the approach that underlies the international comparisons shown in Corak (2013). It is also the approach used by Leigh (2007) for Australia.
In this paper, we generate up-to-date and internationally comparable estimates of the association between fathers’ and sons’ earnings. We closely follow Leigh’s approach, but we use considerably more data for Australia (twelve waves of HILDA) and for the USA (four waves of PSID). Our preferred estimate of intergenerational elasticity (0.35) is considerably higher than implied by Leigh’s study, and is less subject to sampling variation. This estimate implies that 10% higher earnings for a father are associated with 3.5% higher earnings for his son. In an international context, intergenerational mobility in Australia is not particularly high, and is consistent with its relatively high level of cross-sectional inequality.
We also consider other summary indicators of mobility. For this we need to make a number of additional non-trivial assumptions. We estimate the intergenerational correlation to be around 0.23. This is considerably smaller than the elasticity estimate, due to a major increase in earnings inequality over the last generation. We also show indicative estimated probabilities that a son achieves high (low) earnings, as a function of his father’s earnings. For example, consider a father whose earnings are at the 5th percentile. His son is four times more likely to have earnings in the bottom decile than in the top decile (17.8% compared to 4.3%).