We begin by analysing the estimated values of intergenerational income persistence at different stages in the lifecycle. Table 1 reports both the income elasticities and rank correlations for our three samples using both a single year measure of parental income (panel A) and a two-year average measure (panel B). Full regression output tables are available in the appendix.
A number of features stand out. Firstly, there is a clear ranking of elasticities across the three samples regardless of the lifecycle stage and the measure of parental income used. The North of Britain exhibits far less income persistence than the South, hence inferring that the North has greater levels of mobility. The baseline estimates for Britain as a whole lie between its two regional subsamples, which is to be expected given it is essentially an average measure of the two. The alternative rank correlation measure, which removes the differences in income dispersion, uncovers the same pattern across the three samples, suggesting that the lower mobility seen in the South is not due to increased growth in intergenerational income inequality.
The rank correlation measure has mixed effects relative to the size of the respective elasticity estimate. In some cases, the rank is higher than the elasticity, implying that the variance in income distributions increases from parents to sons, while in others the opposite is true. In panel (B) however, the rank correlation is much more consistent, being smaller than the respective elasticity in all bar one estimate.
The elasticities based upon two-year averages of parental income (panel B) are consistently higher than their point-in-time counterparts (panel A). This indicates the presence of temporary income fluctuations or "transitory shocks" that downwardly bias the estimated value of β (Solon 1992). The two-year average therefore eliminates some of these transitionary components of income, thus getting closer to a true measure of lifetime parental income and alleviating the measurement error to an extent.
Comparing both elasticities and rank correlations across the lifecycle for all three samples proves that earnings measures at age 42 are much better proxies of lifetime earnings than earnings measures at age 30, and that the age 30 measure often greatly underestimates the true value of income persistence.
Figure 1 provides a represents visually the elasticity and rank correlation estimates for all three samples using the averaged measure of parental income.
All estimates were significant at the 1% level, but all also had low values of R2. This however, is not concerning as low R2 values are generally a feature of empirical intergenerational mobility studies (e.g., Blanden et al. 2007) and imply potential missing variables. Indeed, the Ramsey Regression Equation Specification Error tests that were ran on the regressions reported concluded that variables were indeed missing from the model. This was to be expected as it is evident from the literature review that a number of mechanisms influence intergenerational income persistence and it is not simply a direct relationship between parental income and sons' earnings. Nonetheless, we can conclude from these results that our first hypothesis, that the North of Britain has greater levels of mobility relative to the South, is true.
Educational Inequality and Returns to Education
Building on the estimations of the strength of intergenerational persistence above, the focus now shifts to unpacking the mobility process by considering the role that education inequalities and earnings returns to education play in intergenerational income persistence across the three samples.
As described in the data section, a fourfold categorisation of educational attainment was constructed and applied to each sample. Table 2 describes the earnings returns to the highest educational level of sons at age 30 (panel A) and at age 42 (panel B). The results can be interpreted as the percentage lower earnings associated with achieving a given level of educational attainment compared to the baseline category (university degree).
There is a substantial advantage to obtaining a university degree in all samples, at both ages, with said advantage decreasing with each additional higher level of education attained. Like the estimates for intergenerational persistence, there is a clear a ranking for the returns to education at the "Less than O-Level" and "O-Level" categories, with the South exhibiting the highest returns and the North the lowest. However, at the "A Levels" category, this pattern reverses, although the small frequency of cohort members who fall with this category likely cloud the true returns. Indeed, the "A Levels" category is the only one which returns non-significant results, in both ages.
The returns to education are consistently greater at age 42 than at age 30. This is logical as the possession of a university qualification increases lifetime income (Tamborini et al. 2015) and so the wage disparities between a graduate and a person with no qualifications will be greater at age 42 than at age 30.
The results for the earnings returns to education appear to support the view that education is behind the lower mobility seen in the South, but this is also dependent on the relationship
between parental income and educational attainment, reported in table 3. The interpretation is the effect of a doubling of average parental income on the probability of being in each one of the education groupings compared to the other three categories. At age 30, the probability of obtaining a university degree is 21% higher for a doubling of average parental income for The North, much higher than the 14% in The South, meaning that degree attainment in the North is more dependent on parental income. Conversely, the chance of obtaining qualifications below O-Level is around 12% lower in The South, compared to 5% lower in the North. The difference in the income effect on the risk of having the highest vs. the lowest education level in all three samples.
Decomposing for the Role of Education
After having analysed both the association between parental income and education, and the returns to education, we now turn to decomposing the income association. Table 4 reports the formal decomposition outlined in the methodology section.
Figure 2 illustrates the decompositions for elasticities calculated using averaged parental income reported in Table 1 above. The decomposition separates out the part of the income associations that is mediated by education.
When viewed as a percentage, there is a clear ranking in that education plays a greater role in intergenerational persistence in the South than it does in the North, accounting for around an additional 5% of total persistence for both ages in the South. Given the patterns revealed in tables 2 and 3, it can be inferred that the lower persistence through education in the North originates in the lower returns to education, rather than the smaller inequalities in education by parental income.