Working Papers

Regional economic disparities within countries have become increasingly large, often surpassing the disparities observed between countries. To address regional inequality, governments have been turning away from standard subsidies and are experimenting with public employment reallocation as a place-based policy. This paper estimates the causal effect of public employment reallocation on local labor markets. I study the ‘Heimatstrategie,’ which relocates around 3,000 public sector jobs from Munich to economically lagging regions in Bavaria, Germany. Using novel data on 60 agency relocations between 2015 and 2025, I exploit the government’s quantitative selection criteria for receiving municipalities and implement a long-differences design comparing treated Bavarian municipalities to Mahalanobis-matched control municipalities in other German states. My estimates show that relocations increased private sector employment shares by up to 2.3%, reduced unemployment rates by up to 11.9%, and increased local population by up to 1.6% without harming sending locations. These results correspond to a public-to-private jobs multiplier of 1.08. To assess general equilibrium effects the relocation program, I implement a quantitative spatial model with a two-sector (public and private) framework showing modest increases in amenities through the relocation counterfactual and negligible welfare effects.

Presented at: 12th ifo Dresden Workshop on Regional Economics, 10th CGDE Doctoral Workshop, IEB Barcelona IWIP Seminar,  AQR Research Seminar, University of Barcelona (invited), 13th European UEA Meeting, 2024 RSA Annual Conference, Aix-Marseille Research Seminar (invited), NBER Conference on Place-Based Policies, 14th European UEA Meeting, 6th World Labor Conference in Toronto, VfS Congress (in Sep 2025), 9th Potsdam PhD Workshop in Empirical Economics (in Sep 2025)


This paper examines whether place-based policies influence electoral outcomes by exploiting the largest public sector relocation program in recent decades in Germany. We analyze the effects of relocating over 60 public authorities and institutions and more than 3,000 civil service jobs to municipalities in structurally weak regions of Bavaria between 2015-2025. Using a difference-in-differences approach leveraging the staggered rollout of relocations, we find that the program significantly impacted voting patterns. The Christian Conservative party (CSU) that actively claimed credit for the policy experienced a strong incumbency advantage, gaining over 2 percentage points in state elections in treated municipalities. Meanwhile, their coalition partner, the Free Voters (FW), was punished at the polls, losing 1.6 percentage points in state elections. The relocation program also had some success in containing the rise of the far-right populist AfD party, reducing their vote share by 0.5 to one percentage point in treated areas. Additionally, we observe a small but significant decrease in voter turnout of 0.72 percentage points in the first election following a relocation, suggesting a modest demobilization effect. Our analysis reveals limited geographical spillover effects to neighboring municipalities, suggesting that the policy’s impact was largely confined to the host localities. We also find small spillovers to the federal elections, with the CSU gaining 0.7 percentage points and the FW losing 1.5 percentage points. Our results suggest place-based policies can shape electoral outcomes, potentially helping the incumbent party that claims credit for the policy win support in economically distressed regions while counteracting the rise of populist movements. The findings have implications for governing strategies and maintaining democratic stability in the face of growing regional economic disparities.

Presented at: 18th ifo Workshop in Political Economy, European Public Choice Society Meeting 2025


The vote share of far-right, populist parties has almost doubled in Western countries in the last decade. This sudden increase led to a growing body of literature, especially in Economics and Political Science. Yet, due to disagreement among the disciplines, measurement, and identification issues, the main drivers of far-right populism remain unclear. It is also unclear if some of the determinants analyzed so far might have a non-linear relationship with the far-right, populist vote share. Against this backdrop, we create an extensive new dataset with over 26 determinants of far-right populism in over 30 countries between 1980 and 2019. We then employ Bayesian Model Averaging (BMA) to identify the most robust drivers of populism across all possible models. We move on to apply Bayesian Additive Regression Trees (BART) to determine non-linearities in those drivers’ relationship with far-right populism. We find that while the trade-to-GDP ratio decreases far-right populism, imports from China increase it, however, only after a threshold value. Regarding the effect of immigration on far-right populism, we find that higher migrant stock decreases far-right populist vote share, confirming the contact hypothesis. In contrast, we find a threshold effect regarding refugee migration. Regarding culture, we find a strong positive relationship between religiosity and far-right populism, confirming the idea that religion forms a group identity and plays into the cards of populist rhetoric of the division of society into the moral people and the corrupt elite. 

Presented at: Annual Congress of the Swiss Society of Economics and Statistics 2024, Aix-Marseille Research Seminar (invited)


Xu (2022) estimates the causal impact of bank failures on the level of trades with a staggered difference-in-differences design and an IV strategy with Bartik instrument, using the 1866 banking crisis as a quasi-natural experiment. Findings, based on historical data on the trades and loans between London banks and banks around the world, show that countries exposed to bank failures in London immediately exported significantly less and did not recover their lost growth relative to unexposed places. Moreover, the effect lasted for decades. First, we reproduce the paper’s main findings by running the original code and uncover three issues, one of which that slightly affects the main estimates reported in the study. Second, we test the robustness of the results to (1) removing weights from the regressions, (2) using a spatial HAC correction for the standard errors, and (3) implementing a method for possibly heterogeneous treatment effects with a staggered difference-indifferences design. Overall, we conclude that the main findings are valid and robust.