Browsing by Person "Ruberg, Tim"

Now showing 1 - 1 of 1

Three essays on causal inference in education and health
(2025) Ruberg, Tim; Osikominu, Aderonke
This dissertation presents three essays on the causal impact of education and health programs in different countries. In the first essay, co-authored by Aderonke Osikominu and Gregor Pfeifer, I use individual-level data for Switzerland to estimate the causal effect of a STEM university degree on wages, allowing for heterogeneous returns to observed and unobserved characteristics. In the second essay, co-authored by Hideo Akabayashi, Chizuru Shikishima, and Jun Yamashita, I exploit a unique feature of the Japanese preschool system to estimate the causal effect of an education-oriented against a care-oriented preschool on children’s cognitive and socioemotional development in adolescence. In the third essay, co-authored by Lester Lusher, I investigate the impacts of a public warning system for extreme heat events on behavioral and health outcomes in Japan. 1st Essay: Characterizing Returns to STEM: Evidence on Marginal and Policy-Relevant Treatment Effects (with Aderonke Osikominu and Gregor Pfeifer) Most developed countries have been fostering education in the STEM field to increase innovation. In this study, we estimate heterogeneous returns to a STEM education in Switzerland based on individual-level data, exploiting the regional distribution of relative distances to technical and cantonal universities as a cost factor driving college major choice. A clear setting in the 1990s where prospective students could freely choose universities and majors allows us to exploit the regional distribution of technical and cantonal universities for identification in an instrumental variable framework. On average, individuals strongly gain in terms of wages from a STEM education. Descending Marginal Treatment Effect (MTE) curves suggest positive selection on gains, implying that individuals with a low resistance for a STEM education gain the most. This positive selection is driven by both heterogeneity in the returns to a non-STEM education as well as a STEM education. By simulating affirmative action policies that increase STEM enrollment and estimating corresponding policy-relevant treatment effects (PRTEs), we show that the effectiveness of such policies strongly depends on who is affected and, thus, on their observable and unobservable characteristics. We also show how such policies must be designed to increase STEM enrollment and benefit targeted individuals, most importantly, women. 2nd Essay: Education-Oriented and Care-Oriented Preschools: Implications on Child Develop-ment (with Hideo Akabayashi, Chizuru Shikishima, and Jun Yamashita) It is well known that experiences in early childhood are critical in shaping children's cognitive and noncognitive development. This paper estimates the causal effect of education-oriented vs. care-oriented preschools on child development. In Japan, parents can freely choose to send their children into either an education-oriented preschool (i.e. kindergarten) or a care-oriented preschool (i.e. nursery school). Education-oriented preschools are half-day facilities that provide about four hours educational instruction per day, while care-oriented preschools are full-day facilities that provide care for parents who are unable to provide care for their child for work or other reasons. Using a unique quasi-experimental setting in Japan that exploits plausibly exogenous regional and temporal variation in the relative availability of those preschools, we find that attendance at an education-oriented preschool is associated with significant improvements in mathematical and linguistic achievement that manifest later in adolescence. Positive effects can also be found for socioemotional measures. Ascending marginal treatment effect (MTE) curves suggests an inverse selection pattern: children that are least likely to enroll in the education-oriented preschool gain the most from it. These children are more likely to come from disadvantaged backgrounds. Heterogeneity in treatment effects is mainly due to specific features of education-oriented preschools (i.e. educational orientation, the interaction with parents due to shorter operating hours, and peer effects), while gains from enrollment in care-oriented preschools appear more homogeneous. 3rd Essay: Unveiling the Unseen Illness: Public Health Warnings and Heat Stroke (with Lester Lusher) The world has been experiencing the hottest summers ever recorded in human history. Research suggests that the health consequences from extreme heat are dire, yet still potentially severely underestimated. We investigate the behavioral and health responses to the first comprehensive heat-health warning system in Japan, where alerts were issued in a region when the forecasted wet bulb globe temperature (WBGT) exceeded 33°C. These alerts inform the public about the risk of heat stroke and encourage people to stay at home, drink a lot, take salt, avoid exercise, use air conditioning, and report to others if they are in poor physical condition. We estimate the impact of these heat alerts using plausibly exogenous region-day variation in the difference between actual and forecasted WBGT (i.e. forecasting errors). By conditioning on local daily WBGT, we separately disentangle the effects of a high heat warning from high heat itself using exogenous variation in forecasting errors. We find that the alerts led to a large and precisely estimated increase in heat stroke counts of around 17%. The effect exists across the spectrum of severity, age groups, and places of incidence. Using data from Google Trends, we find that heat alerts increased general awareness as reflected by increases in searches for the terms “heat stroke,” “heat stroke alert,” and “temperature.” Using additional data from Google Trends, Google Mobility Reports, energy consumption behavior, the population of ambulance records, and our own survey, we rule out any “adverse” response and substitution in health diagnoses away from other sudden illnesses. Therefore, our evidence suggests that the effect is driven by increased reporting of otherwise unidentified cases. Back-of-the-envelope calculations suggest at least 2,550 additional heat strokes in 2021 and 2022, or 10.2% of all heat strokes recorded on these high-heat days are detected. The effect in low-income compared to high-income neighborhoods is four times the effect in magnitude, highlighting severe environmental inequalities in reporting behavior.