I am a PhD student and Carl J. Friedrich fellow at the Department of Government, Harvard University. I am interested in the political economy of long-term policymaking, particularly in how democratic institutions and political actors shape policy choices over time and how policy feedback influences these dynamics. My substantive focus is on climate change and the politics of the green transition.
Methodologically, I employ causal inference and computational methods. More broadly, I am interested in the identification of causal effects in dynamic settings, particularly in the presence of feedback loops and temporal spillovers.
My current work examines wind power expansion in Germany and how perceived costs and benefits moderate local communities’ support for the energy transition over time. Alongside this, I am also working on methods for parameter estimation in constrained spatial models.
I hold an MSc in Statistics and Data Science from Ludwig-Maximilians-Universität (LMU), an MSc in Public Policy and Administration from the London School of Economics and Political Science (LSE) and a BSc in Psychology from LMU.
Research
Working Papers
“Wind Power, Compensation and Electoral Backlash: Evidence from 35 Years of Wind Power Expansion in Germany” (draft available on request). [Abstract]The expansion of renewable energy infrastructure is essential to the green transition, yet its local implementation can generate electoral opposition. This paper examines whether wind power expansion produces dynamic electoral responses and whether economic benefits can mitigate local backlash. First, using 35 years of municipal wind turbine construction and election data, I investigate whether electoral responses follow an initial implementation dip. This presents a distinct econometric challenge, as treatment is staggered, non-absorbing and non-binary. I therefore compare two modern Difference-in-Differences estimators developed by Callaway and Sant’Anna and de Chaisemartin and D’Haultfœuille. The results provide inconclusive evidence of an implementation dip, but suggest a small, statistically significant electoral backlash against pro-renewables, reflected in declining Green Party vote shares. Second, I investigate whether economic benefits can mitigate local opposition, leveraging the introduction of mandatory financial compensation for communities hosting wind turbines in Brandenburg ahead of the 2025 federal elections. Exploiting subnational policy variation, I find that newly constructed wind turbines increase the vote share of the AfD, the main political opponent of renewable energy expansion, in states without compensation, but not in Brandenburg. This suggests that financial compensation can mitigate electoral backlash, although it does not translate into active support for renewable energy. Taken together, the findings illustrate that electoral responses to the green transition are shaped both by the temporal dynamics of infrastructure deployment and by how its local economic costs and benefits are distributed. They also highlight the challenges of identifying such dynamics empirically, particularly when treatment is continuous, staggered and non-absorbing.
“How to Hack Democracy: Parameter Estimation for Constrained Spatial Models” (with Caroline Belka, Vanessa Kleisch, and Carl Müller-Crepon).
[Abstract]Simulation-based spatial models are the dominant tool for evaluating redistricting plans, yet they treat constraint weights, governing criteria such as district compactness, as fixed hyperparameters. This means they can assess whether a proposed map is unusual, but cannot estimate the parameters that most plausibly generated it, leaving the underlying priorities and intent of mapmakers unidentified.
We reframe this as a parameter estimation problem and implement three strategies for inferring constraint weights directly from an observed redistricting plan. The estimators are evaluated through a Monte Carlo experiment on real-world maps, in which known parameters are used to generate plans that the estimators then attempt to recover. Results show that all three approaches successfully recover the true underlying parameters under controlled conditions, establishing proof of concept for this inferential framework.
At the same time, we document substantial challenges for practical application. The proposed methods are computationally intensive, and the estimators exhibit considerable variance, producing wide confidence intervals even under favorable conditions. We discuss what these limitations imply for the reliability of inference in real-world settings and outline directions for future methodological development.
Published
Straub, N., Goretzko, D., & Sterner, P. (2025). “Assessing Measurement Invariance With Exploratory Factor Analysis Trees: A Practical Guide.” European Journal of Psychological Assessment. [Abstract] | Open Access Paper | Open CodeMeasurement invariance (MI) describes the equivalence of measurement models of a latent variable across different groups or time. It is an important prerequisite for subsequent analyses such as group comparisons and diagnostic decisions. Therefore, establishing MI should be a key goal of psychometric research. Exploratory Factor Analysis trees (EFA trees) were developed to address MI already in early stages of measurement scale development. This tool, combining EFA and model based recursive partitioning, performs well in situations with many groups and continuous covariates. The goal of EFA trees is that they are an easy-to-use tool for assessing MI which can be applied both in questionnaire development and prior to group comparisons, especially when no specific hypotheses on the non-invariance between groups exists. To facilitate the application of EFA trees in practice, this paper provides a step-by-step tutorial of the new R package EFA tree that explains how to run the analysis and evaluate the results. A large dataset from a multi-country study is used to illustrate each step, and the results are compared with conventional methods of assessing MI. Advantages and limitations are discussed and further research ideas are presented.