Browsing by Subject "Equality of opportunity"

Sibling correlations in long-run earnings are an omnibus measure of family background in equality of opportunities. This thesis provides updated estimates for sibling correlations in Finland and augments the standard sibling correlation model with the within family gender components. The thesis also explores a machine learning model predicting correlation in artificial outcome variables to raise the lower bound of the effect of family background on equality of opportunities. The thesis uses Finnish FOLK-dataset, tracking a random sample of people from 1970. The estimation models follow previous literature to keep the results comparable to the main results. A machine learning algorithm is tested as a proof of concept comparing its results with the regular sibling correlations. The sibling correlations in Finland in long-run earnings are estimated at 0.25 for brothers, 0.17 for sisters and 0.19 for pooled sample. Long-run disposable family income is also estimated at 0.25 for brothers, 0.28 for sisters and 0.22 for pooled sample. Within family gender components explain an additional 0.04 for both outcome variables. The artificial predicted outcome variable has correlations consistently higher than regular sibling correlations by up to 25 percent. The largest differences are in earnings for sisters and both outcomes for pooled samples. The correlations in long-run earnings are similar to the estimates for Nordic countries in previous studies. Large difference in results for sisters between the outcome variables suggests disposable family income as a more reliable outcome metric for women. The gendered components have a minor, but consistent effect. The artificial predicted outcome variable is able to explain a noticeably higher share of variance of outcome variables, especially in the samples where sibling correlations are relatively lower. This suggests that the lower bound of the estimate for equality of opportunity could be raised with this approach and it warrants further research.