We develop a quantitative theory of repeated political transitions driven by revolts and reforms. In the model, the beliefs of disenfranchised citizens play a key role in determining revolutionary pressure, which in interaction with preemptive reforms determines regime dynamics. We estimate the model structurally, targeting key moments of the data. The estimated model generates a process of political transitions that looks remarkably close to the data, replicating the empirical shape of transition hazards, the frequency of revolts relative to reforms, the distribution of newly established regime types after revolts and reforms, and the unconditional distribution over regime types. Using the estimated model, we also explore circumstances of successful democratization, finding that the sentiment of political outsiders is key for creating a window of opportunity, whereas the scope of the initial democratic reform is key for the survival of young democracies.
Does investment create more jobs in slack than in tight labor markets? We study this question using data from a photovoltaic investment scheme. Comparing counties with high and low unemployment over time and across space, we find that €100,000 of investment created 1.2 job-years in slack markets and fewer than 0.5 job-years in tight markets. This corresponds to labor earnings multipliers of 1.1 and below 0.5, respectively. These differences are not driven by changes in investment composition, capital-labor substitution, or regional migration. Consistent with crowding-out as a mechanism, investment leads to higher wage growth in tight than in slack markets.
Parties, Divided Government, and Infrastructure Expenditures: Evidence from U.S. States (with Stephan Fretz) — European Journal of Political Economy Vol. 61, 2020
Final, freely available version here
This paper examines the impact of divided government in U.S. states on infrastructure expenditures for transportation, education, and social services. As infrastructure investments, in particular for transportation, are a bi-partisan issue, we hypothesize that divided governments expand infrastructure spending compared to governments under true Democratic or Republican control. We test this hypothesis using U.S. state-level data over the period 1970 to 2008 and find that divided governments indeed increase expenditures for transportation. This finding holds both for split-legislature and split-branch governments, and the impact is most pronounced for capital outlays. In contrast, expenditures for education and social services are characterized by more partisan patterns.
Weather and the Psychology of Purchasing Outdoor-Movie Tickets (with Thomas Kolaska) — Management Science Vol. 63(11), 2017, p. 3718–3738
Media Coverage: Harvard Business Manager (scan, in German)
Penultimate, but freely available version here. This paper supersedes Projection Bias with Salient State-Dependent Utility
The consequences of many economic decisions only materialize in the future. To make informed choices in such decision problems, consumers need to anticipate the likelihood of future states of the world, the state-dependence of their preferences, and the choice alternatives that may become relevant. This complex task may expose consumers to psychological biases like extrapolative expectations, projection bias, or salience. We test whether customers are affected by such biases when they buy advance tickets for an outdoor movie theater, a real-world situation that – due to the availability of reliable weather forecasts – closely resembles a stylized decision problem under risk. We find that customers’ decisions are heavily influenced by the weather at the time of purchase, even though the latter is irrelevant for the experience of visiting the theater in the future. The empirical evidence cannot be fully explained by a range of candidate rational explanations, but is consistent with the presence of the aforementioned psychological mechanisms.
The Employment Effects of Countercyclical Public Investments (with Martin Watzinger) — Resubmitted, American Economic Journal: Economic Policy
Media Coverage: Oekonomenstimme (in German)
We estimate the causal impact of a sizable German infrastructure investment program on employment at the county level. The program focused on improving the energy efficiency of school buildings, making it possible to use the number of schools as an instrument for investments. We find that the program was effective, creating one job for one year for each €25’000 of investments. The employment gains reached their peak after nine months and dropped to zero quickly after the program’s completion. The reductions in unemployment amounted to two-thirds of the job creation, and employment grew predominately in the construction and non-tradable industries.
Sentiment and Firm Behavior During the COVID-19 Pandemic (with Jonas Dovern, Carla Krolage and Sebastian Link)
Policy brief: ifo Schnelldienst (in German)
How did sentiment about the policies for containing the COVID-19 pandemic affect firms’ business outlook and behavior early in the crisis? In a representative panel of German firms, we demonstrate that the variation in expectations regarding the duration of COVID-19 induced lockdowns plausibly reflects sentiment that is unrelated to fundamentals. This sentiment-driven variation in expectations was an important determinant of firms’ crisis response: Firms that perceived the shutdown to last longer were more likely to implement strong measures like layoffs or canceling investments. The implementation of soft measures, e.g., working from home, was unrelated to the expected crisis progression.
Sudden Stop: When Did Firms Anticipate the Potential Consequences of COVID-19? (with Carla Krolage and Sebastian Link), preprint in COVID Economics, 36
COVID-19 hit firms by surprise. In a high frequency, representative panel of German firms, the business outlook declined and business uncertainty increased only when the spread of the COVID-19 pandemic led to domestic policy changes: The announcement of nation-wide school closures on March 13 caused by far the largest change in business perceptions. In contrast, business perceptions hardly reacted to any other potential source of information: Firms did not learn from foreign policy measures, even if they relied on inputs from China or Italy. The local, county-level spread of COVID-19 cases affected expectations and uncertainty, albeit to a much lesser extent than the domestic policy changes.
This paper studies a new aspect of firms’ expectation formation by asking whether expectations primarily reflect aggregate, industry-wide information (e.g., industry trends) or disaggregate information (e.g., firm-specific information uncorrelated with industry trends). First, we show that disaggregate information is strongly associated with expectations even when controlling for aggregate information at high-dimensional industry levels. Moreover, aggregate and disaggregate information explain comparable shares of the variance in expectations. Second, we exploit a natural experiment to identify the causal effect of new information on expectations. The predictable demand effects for durable goods due to the German VAT increase of 2007 implied that, at the time, durable goods retailers had access to more reliable information about their future demand than non-durable goods retailers. Utilizing this observation in a difference-in-differences design, we find that “treated” firms were significantly more forward-looking ahead of the VAT-induced demand shifts. Overall, our results suggest that firms rationally incorporate disagreggate information into their expectations.
Do Lagged Expectations Determine Reference Points? A Test of Köszegi and Rabin’s Equilibrium Concepts
available on request
This paper tests the conjecture of Köszegi and Rabin (2006, 2007) that the reference point is given by recent probabilistic beliefs about outcomes. This entails the notion that expectation-based reference points adjust to new information with a lag. Because of this lag between the arrival of new information regarding feasible choices (and their implied distributions of outcomes) and the adjustment of the reference point to information, Köszegi and Rabin predict that the time structure of a decision problem affects preferences for risk. Specifically, individuals are predicted to be more risk-averse in situations in which the distribution of outcomes is known well in advance than in situations in which the choice set and the realization of risky options arrive unexpectedly. We test this prediction in an experiment in which we vary the time-span between the time at which subjects learn their choice set and the time at which they learn the realization of a risky choice option. The findings of the experiment do not support the hypothesis that a longer time-span between learning the choice set and the resolution of uncertainty increases aversion to taking risk.