Research

Kyklos - International Review for Social Sciences (2026, Vol. 79(3), p. 608-628)

There have been many studies estimating the causal effect of an additional year of education on earnings. The majority employ administrative changes in the minimum school leaving age as the mechanism allowing identification. Here we survey 79 such estimates. However, remarkably, while the majority of these studies find substantial gains from education, a number of well-grounded studies find no effect. The average return from these studies still implies substantial average gains from an extra year of education: an average of 8.2%. But the pattern of reported returns shows clear evidence of publication biases: omission of studies where the return was not statistically significantly above 0, and where the estimated return was negative. Correcting for these omitted studies, the implied causal returns to an extra year of schooling will be only in the range 0-3%.

Published paper
Working paper
Data and appendix

Media coverage: Macro Roundup by Edward Conard, Marginal Revolution, Göteborg-Posten (in Swedish)

Kyklos - International Review for Social Sciences (Forthcoming)

In this response to Lee Crawfurd’s comment on The Returns to Education: A Meta-Study we address his three main criticisms.  We contend that estimates based on compulsory schooling laws constitute the most reliable evidence on causal returns to education.  We show that our parsimonious approach to correcting for publication bias is preferable to more complex approaches that have biases of their own.  Finally, we argue that the observed skewed distribution is most plausibly explained by selective omission of negative estimates.  We therefore conclude that the comment does not undermine our central finding that reported returns to education are substantially inflated because of publication bias.

Current Version

Social class descriptors are still widely used in social science. But does social class add any useful information in predicting social outcomes, compared to other status indicators such as occupational status? Utilizing two large databases on social outcomes for people in England 1700-2020, we show that granular measures of status perform much better as predictors of a wide range of social outcomes than does the classic allocation of people into classes. We learn almost nothing additional by categorizing people by class. This suggests that social class does not reflect the social stratification of society.  Social status lies along a continuum, so dividing people into a few lumpy classes reduces rather than adds explanatory power.

Conventional estimates of social mobility suggest that Denmark is one of if not the most socially mobile country in the world. This is often attributed to Denmark’s low economic inequality and generous social welfare policies. Using a surname approach that reduces errors of conventional methods, I conversely show strong persistence of social status in Denmark, 1750-2020. I estimate persistence rates utilizing changes in relative representation of surname groups in elite occupations and education, specifically doctors, lawyers, officers, company directors, members of parliament and high school graduates. This persistence of status is expressed both in slow upward mobility of lower-status individuals (carriers of patronymic surnames, which constitute a majority of the population) and slow downward mobility of a small group of upper-status individuals (carriers of noble surnames). The findings suggest an underlying persistence rate of 0.7-0.8. The rate is similar to other countries with different institutional setups, and only decreases slightly in the full period, despite significant economic and political changes in Denmark, including democratization and the introduction of social welfare state policies.

Do doctors live longer than other social elites because of medical knowledge? One strand of literature establishes that high social status individuals live longer, while other researchers point to the importance of health behaviours, not status in itself. Doctors likely adopted good health behaviours in their private life earlier than other social elites as medicine became a more scientific discipline throughout the 19th century, which potentially affected their longevity. Utilizing biographies of doctors (1479-1982) and other professionals like lawyers (1660-1936), this paper compares longevity differences over time for elites with and without medical knowledge. I argue that the 1884 medical conference in Copenhagen can be interpreted as the time, after which a broad set of Danish doctors adopted a scientific approach to medicine, the effects of which is investigated in a difference-in-differences setup. Preliminary results suggest that doctors have a larger longevity increase than other professionals after the conference.