The Policy Evaluation Lab applies and advances causal methods to evaluate the impacts and inform the design of public policies addressing societal challenges, such as climate change, biodiversity conservation, air quality, and health.
We combine quasi-experimental methods and structural modeling rooted in econometrics with the potentials of machine learning and novel data to contribute to rigorous policy assessments enabling informed decision-making. Our research covers a broad selection of policies, including carbon pricing, command-and-control regulations, and subsidies, which are key instruments in the building, energy, industry, transport, and land use sector. We analyze their environmental and distributional effects as well as their broader co-benefits for health and well-being.
We publish our research findings in general-interest and field journals in economics (e.g. AEJ:Economic Policy, JAERE) as well as interdisciplinary outlets (e.g. Science, PNAS).
Highlights
1. Using credible quasi-experiments based on rich administrative microdata to uncover novel causal relationships for policy-making
In a quasi-experimental study published in the American Economic Journal: Economic Policy, we examine the staggered introduction of Low Emission Zones—local driving bans designed to improve air quality. We evaluate how the policy affected the health of children born before and after its implementation, compared to newborns in cities that have not yet introduced their own Low Emission Zone. Public health insurance data allow us to analyze pharmaceutical prescriptions for asthma and other respiratory diseases for approximately one-third of all children born in Germany between 2006 and 2012. By tracking these children from birth to school enrollment, we show that children exposed to even modest air quality improvements during pregnancy and their first year of life require persistently less medication. Importantly, the rich administrative microdata help us reveal policy effects that would otherwise be overlooked: health benefits that develop gradually over time and that are too subtle for standard measures of infant health to capture. Identifying these previously neglected, large, and persistent effects on childrens’ respiratory health enables us to more accurately weigh the costs and benefits of Low Emission Zones.
2. Developing ML-methods to evaluate the causal impact and effectiveness of policy
Much of the debate about climate policy centers on which policy instruments reduce emissions, and which do not. Yet, prior evaluations have focused on a limited range of headline policies, neglecting hundreds of other measures. In our recent study in Science, we use machine learning to provide the first global, systematic ex-post evaluation identifying effective climate policies out of the universe of about 1,500 policies implemented between 1998 and 2022 across 41 countries from six continents. We combine a novel, meticulously collated climate policy database from the OECD with a machine learning-based extension of the common difference-in-differences approach to evaluate which policies caused emission reductions. The methodological innovation is that we do not resort to a subjective a-priori selection of particular policies from the large candidate pool of 1,500 potentially effective interventions, but instead focus on identifying those policies that demonstrably led to large emission reductions in a transparent and reproducible statistical framework.
3. Leveraging structural econometrics to provide theoretically grounded evidence on the potential of policy alternatives
We not only show the actual impacts of policies as they are implemented but also inform policymakers of potentially more efficient policy alternatives and their social welfare effects. For instance, in a study published in the Journal of the Association of Environmental and Resource Economists, we use structural methods for welfare analysis of alternative options for just transition policy in an archetypal fossil industry—coal mining. We present a structural job search model to calculate lifetime welfare costs of job loss from a coal phase-out. Based on the universe of German employment biographies, we show that unemployment is a small factor in the welfare costs of job loss, with wage differentials and job security much more important. We then contrast the current labor market policy of subsidizing early retirement for older workers with a counterfactual just transition policy of subsidizing wages for workers who move to other industries. Because the latter policy encourages labor market mobility while automatically benefiting those most affected by job loss, we find that a wage insurance scheme can greatly reduce the welfare costs of decarbonization at little fiscal cost.
4. Bringing the hitherto unavailable information on behavioural patterns of people and firms into environmental economics
In several studies, we use novel, spatially granular, and near real-time mobility data that is automatically generated from cell phones and GPS tracking applications to more comprehensively capture complex and diverse travel behavior in response to policy interventions. In particular, we evaluate urban policy designs, such as different road pricing schemes, to determine which ones recover most welfare lost due to road congestion. Similarly, we analyze how large-scale public transport subsidies, such as Germany’s Deutschlandticket, affect overall mobility and travel mode choices.
