Research

Published or forthcoming

High school dropout for marginal students: Evidence from randomized exam form, with Martin Eckhoff Andreasen. Statistics Norway discussion papers 894. [SSRN], [RepEc]. Conditional accept at  Journal of Labor Economics. 

We exploit the quasi-random assignment of exam form in a high-stakes Norwegian high school exam to estimate the impact of exam form on exam outcomes, later school performance, graduation and longer run outcomes. Results indicate that written exams significantly reduce exam grades and reduce the probability of passing relative to the alternative oral exam, particularly for initially low-performing students. Because passing the exam is mandatory to obtain a high school diploma, this translates into reduced high school graduation rates that remain significant over time, permanently shifting a group of marginal students to drop out of high school entirely. IV results on labor market earnings are close to zero, suggesting that the value of a marginal diploma is low, but these results are too imprecise to draw firm conclusions. 

Long Term Impacts of Class Size in Compulsory Schooling, with Edwin Leuven. Journal of Human Resources 2020, 55: 309-348. (IZA working paper) (Statistics Norway discussion paper)

How does class size in compulsory school affect peoples' long run education and earnings? We use maximum class size rules and Norwegian administrative registries allowing us to observe outcomes up to age 48. We do not find any indication of beneficial effects of class size reduction in compulsory school. For a 1 person reduction in class size we can rule out effects on income as small as 0.09 percent in primary school and 0.12 percent in middle school. Population differences in parental background, school size or competitive pressure do not appear to reconcile our findings with previous studies.

Costs of school closure in the face of COVID-19, with Martin E. Andresen and Simon S.Bensnes, Samfunnsøkonomen, 2/2020: 51-63.

The COVID-19 pandemic has led to numerous policy responses, some of which are directed at the education sector such as school closings. This analysis estimates the predicted costs of maintaining current policies in the Norwegian education system. Specifically, we estimate the cost of keeping all education institutions closed, with all teaching given through online platforms. Our estimates sum to a total of NOK 2.2 billion for higher delayed progression for a small share of students assuming higher education institutions remain closed. The value of lost lifetime income due to reduced human capital and reduced parental labor supply is estimated to NOK 1.7 billion each day schools remain closed. A total of NOK 77.7 billion will be lost if schools and kindergartens remain closed for the remainder of the school year.

Work in progress

School Spending and Financial Reform: Economic and Social Consequences of Shifting Resources to Disadvantaged Students with Simon S. Bensnes 

In this study we investigate the role of school spending by exploiting a school financing reform in Oslo in 2003. Prior to the reform schools were mainly financed on a per student basis. This lead to a situation where schools with a high share of resource demanding students became relatively poorly funded. These costs were associated with special language needs of students with immigrant backgrounds and social issues associated with children from low income households and social housing. The 2003 school nance reform shifted a substantial part of the spending to be determined by several indicators that were constructed to capture several socio-economic characteristics of the school districts. This new allocation had dramatic implications for per-child spending with some schools loosing (gaining) up to 8 000 NOK (5 000 NOK) per year.  Because the reform was budget neutral for the municipality, the the average funding per student remained constant on the aggregate level.

Family Planning and Economic Uncertainty with Edwin Leuven and Arnaud Maurel.

The fall of the Berlin Wall and the subsequent reunification of East and West Germany was arguably among the most significant geopolitical as well as economic shocks that took place in the last century. The fall of the Wall was associated with a drastic and sudden increase in economic uncertainty faced by the inhabitants of former East Germany, who underwent a fast transition from a centrally planned communist economy with full employment to a market economy. Fertility rates among East German women dropped sharply after the reunification. Although multiple mechanisms may play a role, this suggests that the sudden increase in uncertainty associated with the reunification process may have resulted in some of the East German women delaying childbearing. In this paper, we use the fall of the Berlin Wall as a “natural experiment” that provides a unique opportunity to identify and estimate the causal effect of economic uncertainty on fertility behavior. The empirical strategy of this paper combines several identifying variations. The first is the comparison of the fertility rates among East German women before and after the fall of the Berlin Wall. Second, we exploit that different cohorts of women are exposed to uncertainty at different segments of their fertile years. Finally, a we exploit variation in individual characteristics that to varying degrees shield women from the effects of uncertainty, such as having a “safe” civil servant job or risk attitudes and perceptions from surveys.

Working papers

Moving beyond expectations: From cohort-component to microsimulation projections, with Zhiyang Jia and Stefan Leknes. Statistics Norway discussion papers 999.

Population projections are predominantly made using the cohort-component method (CCM). The opportunities for further development within that framework are limited. Lately, with advances in technical and computational capacity, the microsimulation framework has become a serious contender. In contrast to CCM, it allows for rich complexity of behavior and provides insights on projection uncertainty. Still, demographers have been reluctant to apply this framework, which may be due to lack of guidance. We contribute by clarifying underlying CCM assumptions, translating a multi-regional version of the model into a dynamic spatial microsimulation model, and discuss the usefulness of prediction intervals for planning. Using data for Norway, we demonstrate that the results for the two models are equivalent, even for very small subgroups, and converge with relatively few simulations. The model can easily be amended with additional individual heterogeneity, facilitating more accurate representations of population dynamics. 

Empirical Bayes estimation of small area fertility schedules with a flexible three-level model, with Stefan Leknes. 

(Statistics Norway  notes 2020/03, early version)

There is high demand for reliable local demographic schedules, which are typically used for research, planning and business purposes. Small area problems pose a challenge to estimating latent demographic behavior. Using hierarchical linear models, we retrieve empirical Bayes estimates of small area fertility schedules and study the trade-off between sampling variability and bias from misspecification of hierarchy. We find that introducing a third “global” geographic level reduces concerns of misspecification and yields superior performance in estimating true rates. Our specification is flexible with nested levels. It resembles a variance decomposition exercise, and applying the estimated rates reflect (produces the number of births and) the total fertility observed in the sample. The positive properties of the model is demonstrated through simulations and estimations on full-count Norwegian population data.

Long-run Effects of School Spending: Evidence from Exiting Cohort Size Variation , with Audun Langørgen

This paper investigates the long-term effects of local government education spending on child outcomes, including income, educational attainment, and family formation in adulthood. We propose a novel identification strategy which exploits quasirandom variation in demographic trends when there is strong inertia in local government spending on compulsory schooling. Specifically, size of the exiting cohort that finishes compulsory schooling just before entry of the treated cohort is used as a source of exogenous variation. First, we show that exiting cohort size displays a significantly positive effect on per-pupil spending during school years of the treated cohort. Second, we argue that causal effects of school spending can be identified by utilizing exiting cohort size to instrument for school spending. In implementing this strategy, school spending is found to exhibit sizable and significant effects on income in adulthood for boys, with estimates that are relatively large for children from low- and middle-income families. By comparison, the effects of education spending are small and insignificant for girls. 

The Effects of School Spending on Long Term Earnings: Evidence from a Regression Discontinuity in Local Elections 

This paper investigates the effect of school spending on the long-term earnings of the pupils. Variations in spending caused by a discontinuity in party control in Norwegian local elections are allocated among pupils at different stages in the compulsory school system. These exogenous shifts in spending can then be exploited in order to estimate the causal effect of school spending on the long-run earnings of the children. The elasticity of school spending on long-term earnings is found to be around 0.3. This means that a 10 percent increase in total school spending will increase the adult earnings by 3 percent. Most of this increase in earnings is found to be attributable to human capital induced productivity gains.

The Dynamic Adjustment of Local Government Spending

This paper investigates how local governments reallocate resources when desired allocations change. We use a linear expenditure model of local government spending behavior in which municipalities are allowed to have preferences for short term savings and place it in a partial adjustment framework that remains consistent with the budget constraint. By estimating this model on Norwegian municipality data for the years 1988-2009 we find a much higher inertia to adjustment than previous studies have found. We find that the adjustment process is relatively faster in service sectors aimed at the elderly rather than the young, which severely limits reallocation towards elderly user groups in a period in which the elderly share of the population is growing. In addition, we find that unemployment reduces the adjustment speed of most service sectors since the lower income tax revenue resulting from unemployment inhibits reallocation.