Effect heterogeneity is central for social mobility research, focusing on whether subgroups have converging or differential mobility rates. Still, standardized approaches are lacking, and conventional methods suffer from drawbacks – e.g., covariate imbalance and overfitting. I propose a three-stage evaluation process that uses prediction methods, complementing conventional examinations of group mean and effect differences, that allows for gauging subgroup-specific changes in explanatory power. The analytical framework additionally focuses on identifying what heterogeneity components (parental SES, gender, or migration background) drive any differences produced. While drawing on traditional subgroup analysis and intersectionality theory, the empirical focus is on universality vs. interactive models of social mobility using detailed Swedish register and census data. Although the conventional interpretation of interactive models shows strong support for subgroup heterogeneity and varying social mobility rates, predictive results indicate minimal improvements in fit and little support for an effect heterogeneity by parental SES. In sum, the interpretation of coefficients and graphical analysis of interactions can be contextualized by expanding the criteria for assessing subgroup heterogeneity. Ultimately, this leads to a better understanding of subgroup particularities, guiding future research toward richer answers with a focus on both effect estimates and how well competing models fit the data.

Co-authors: Kai Liu, Kjelll G. Salvanes

How does a large structural change to the labor market affect education investments made at young ages? Exploiting differential exposure to the national decline in routine-task intensity across local labor markets and using population wide panel data, we show that the secular decline in routine tasks causes upskilling among high school students, with a decrease in vocational-trades and in dropout, along with increases in college. Low school performers and low-SES students are the most responsive and, as a result, intergenerational mobility in college education increases. Routine-biased technical change impacts the supply of skills of the next generation.

While social skills seem to gain importance in the workplace, other skills may become less relevant. The evolution in skill demand and supply directly affects wage returns to skills over time. However, estimating returns to skills is challenging: a potential bias comes from unmeasured ability differences, and there is an indirect return through college. This paper estimates direct and indirect returns to skills, controlling for unmeasured ability differences, using a novel dynamic model with endogenous cognitive, social, and diligence skills. In Germany, across recent cohorts, returns to social skills grew by 6 percentage points across cohorts. Due to routine task displacement and sorting into routine-intensive occupations, returns to diligence skills for low-cognitive individuals dropped by 10 percentage points.