Research Statement: Robert Ulbricht June 2022* My research covers, broadly speaking, three main topics: (i) the consequences of uncertainty for the business cycle, (ii) the macroeconomics of job-to-worker allocation, and (iii) the long-run dynamics of political regimes. A recurring theme across these topics is the insight that people face information frictions in their decisions to save and to work, to search for jobs, to hire employees, to join a political revolution, or to contract agents. Methodologically, most of my research is theory-based. I strive to distill complex realities into simple yet insightful models. While abstracting from unnecessary detail, much of my work seeks rigorous quantification of the core mechanism that is being studied. I summarize the three areas of my research below (along with a fourth topic that is somewhat of an outlier). 1 Uncertainty and the Business Cycle The first strand of my research concerns the role of uncertainty for the business cycle. In “Robust Predictions for DSGE Models with Incomplete Information” (AEJ: Macroeconomics, forthcoming), my coauthor and I quantify the importance of expectational shocks for explaining aggregate fluctuations. One challenge in doing so is that there are myriad ways in which expectational shocks can propagate through an economy, and discriminating between them usually requires taking a precise stand on the structure of information available to agents. Unfortunately, there is no direct evidence to discriminate between the many ex-ante plausible information structures (information is inherently hard to observe). In our paper, we propose a simple workaround: Instead of taking a precise stand on the information available to each inhabitant of an economy (consumers, firms, etc), we allow the researcher to specify certain conservative bounds on their information (e.g., consumers observe the price of their own consumption basket, everyone observes GDP with some lag, etc). We then map these information-bounds into bounds on the dynamics of endogenous equilibrium outcomes such as GDP, inflation, or consumption. Econometrically speaking, our approach provides a “set identification” of the joint time series of macroeconomic variables within a given economic environment.1 Our paper offers a general characterization theorem that applies to the majority of business cycle models used in the literature. We then apply it to a specific business cycle model to quantify how much of aggregate fluctuations can be accounted for by waves of optimism and * An up-to-date version of my CV and PDFs of all completed papers can be found on my website: www.robertulbricht.com. 1 In our language, an “economic environment” fully specifies the non-information aspects of an economy (preferences, technologies, market structures), without taking a stand on the information that is available to agents when they form their expectations. 1 pessimism among consumers and firms. We find that if we allow for all types of uncertainty, expectational shocks can account for up to 51% of the U.S. business cycle. However, this hinges critically on firms being uncertain about idiosyncratic demand. If we shut down idiosyncratic demand uncertainty, all other sources of uncertainty (including uncertainty about idiosyncratic productivity and all aggregate variables) can together at most account for 3% of the U.S. business cycle. In conclusion, expectational shocks are a strong contender for being a primary driver of business cycles, but for them to fill that role it is essential that—while making their production plans—firms face uncertainty about the idiosyncratic demand that their products will receive. In “Endogenous Uncertainty and Credit Crunches” (Review of Economic Studies, R&R, 2nd round), my coauthor and I study the role of uncertainty about such idiosyncratic factors2 as a source of slow recovery from financial crisis. The starting point is the realization that when firms are unable to operate at potential (e.g., due to limited access to external funds), less information about the quality of forgone projects and their potential demand is being generated, increasing uncertainty. At the same time, the more uncertain lenders are about a firms’ potential, the less likely the firm obtains funding. Jointly, these two forces imply that an exogenous, but temporary, reduction in funding can morph into a persistent spiral of increased uncertainty about a firm’s potential, heightened credit risk, and the firm operating below potential. We study this mechanism embedded into a neoclassical general equilibrium model, in which firms are funded by a competitive banking sector. In this setting, the amplification and internal persistence of financial disruptions carries over to aggregate financial shocks that hit banks’ capacity to lend. Calibrated to U.S. data, we find that an exogenous financial shock with a half-life of 4 quarters translates into an endogenous decline of output with a half-life of 22 quarters. This discrepancy is caused entirely by the interaction between endogenous uncertainty and financial frictions: when shutting down the former, the half-life of output mimics the one of the exogenous financial shock. Besides offering an explanation for the persistent decline in output and employment during financial crisis, the theory further explains the contemporaneous rise in credit spreads and default rates, an increased cross-sectional dispersion of firm sales, the increase in measured uncertainty, and high levels of disagreement among forecasters. In “Endogenous Second Moments: A Unified Approach to Fluctuations in Risk, Dispersion, and Uncertainty” (Journal of Economic Theory, 2019), we study a related question: namely, why several second moments—such as cross-sectional dispersion, risk, volatility, or uncertainty—fluctuate (and co-move) over the business cycle. One possibility would be exogenous shocks to these second moments. In our paper, we offer an alternative interpretation, in which second moments fluctuate endogenously when the variables of interest are concave (or convex) transformations of some (random) fundamental. Figure 1 illustrates the mechanism for the simplest possible case where the fundamental X is exogenously “shifted left”. For concave transforms g, this leads to an increase in variance of the transformed 2 In this paper, uncertainty is about about firm-specific revenue-productivity. In line with the takeaway from the paper above, this includes the idiosyncratic demand uncertainties that we have found to be key in explaining aggregate fluctuations. 2 g fX1 fg(X1 ) fX2 fg(X2 ) Figure 1: Illustrating the mechanism in “Endogenous Second Moments: A Unified Approach to Fluctuations in Risk, Dispersion, and Uncertainty”. The plot shows the pdf of a random variable X and its transform g(X). For concave transforms g, a downward shift from X1 to X2 increases the variance of g(X). variable g(X). The methodological contribution of our paper is to provide precise conditions on the nature of the fundamental shocks for which second moments are pro- or countercyclical (going beyond the simple case of a translation). The applied contribution is to highlight one important non-linearity that can explain the business cyclicality of several second moments, casting doubt on the common practice of using them as evidence for second-moment shocks. Specifically, we show that when the input-elasticity between capital and labor is strictly below unity, log-employment and log-output become concave in a firm’s log-productivity, while (under certain additional conditions) investment becomes convex. Empirically, this is indeed the relevant case, with estimated input-elasticities ranging from 0.25 to 0.53.3 Hence, first-moment shocks to the productivity distribution are very likely to endogenously induce counter-cyclical fluctuations in the cross-sectional dispersion of employment and output, and procyclical fluctuations in investment dispersion (all in line with the data). Moreover, if an econometrician were to infer firms’ productivities using a Cobb-Douglas production function (falsely assuming a unit elasticity), then the inferred productivity dispersion also becomes counter-cyclical (even though the true productivity dispersion is constant). Calibrated to U.S. data, these endogenous fluctuations account for a large share of the empirical cyclicalities, suggesting caution when using the latter as evidence for second-moment shocks. 3 These estimates are smaller than corresponding estimates for the input-elasticity at the macro level, reflecting that unlike its counterpart at the macro level the micro-elasticity does not capture substitutions across firms and sectors. 3 Relatedly, in “Mismatch Cycles” (Journal of Political Economy, forthcoming), my coauthors and I study the impact of another type of uncertainty on the business cyclicality of mismatch between workers and jobs, and the consequences for labor productivity. I describe the details of this paper below. 2 The Macroeconomics of Job-to-Worker Allocation In a second strand of my research, I study frictions that impede the allocation of workers to jobs, and explore their implication for the individual worker and for the economy as a whole. In “Mismatch Cycles” (Journal of Political Economy, forthcoming), my coauthors and I study how well workers in the U.S. are matched to jobs in the sense of how well their specific skills line up with the specific skill requirements of their jobs. In the data, we see a lot of “mismatch” between these. Some of it might emerge ex-post, due to skills or requirements changing after a match is formed. But that doesn’t explain the prevalence of mismatch right upon match formation. One explanation is that matching is random, in which case mismatch is just an intrinsic property of labor markets. Our work explores another possibility where mismatch emerges endogenously due to skill uncertainty (the paper provides direct survey-based evidence in support of this). To do so, we develop a tractable framework, in which workers are characterized by multidimensional skill bundles and learn how skilled they are for specific tasks as they use these specific skills in production. As workers revise their beliefs, they sort into jobs that differ in both the type of skills they require (a “career” choice) and how intensely they make use of them (pinning down “job rungs”). We estimate the framework using a combination of worker-level data from the NLSY79 and occupation-level descriptors of job requirements (O*NET). At the worker-level, skill learning (prolonged by search frictions) creates mismatch that is highly inert and carries over to highly inert job rungs and earning dynamics. Quantitatively, the inertia is strong enough to explain an unemployment scar of 19 percent five years after displacement, fully accounting for its empirical counterpart. At the macro-level, the dynamics of mismatch are shaped by two opposing forces. On the one hand, there is a cleansing of matches that are mismatched during recessions. On the other hand, workers that are cleansed from the bottom job rungs have a high propensity to look for new careers, resulting in sullying during recessions. The relative magnitude of these two opposing effects depends on the size of a recession (cleansing is relatively larger the larger a recession) and whether it is broad or concentrated to only a few sectors (sullying is relatively larger the more concentrated). Calibrated to the “average” recession, we find that cleansing dominates, which we independently confirm in the data. Through the lens of the model, this corresponds to an aggregate increase in labor productivity of 0.4 percent during recessions (but with a very heterogeneous incidence in the cross-section, ranging from an increase in labor productivity by 1.5 percent for workers that have been continuously employed for two years to a decline by 1 percent for new hires). In “The Colocation Problem: Dual-Earner Job Search and Labor Market Efficiency” (in progress), my coauthor and I study the role of another friction for the 4 job-to-worker allocation, namely the fact that dual-earner couples need two jobs in one location. Previously, this has been conjectured to discourage workers to search for jobs that entail migration, forgoing potential earning gains (Mincer 1978). Closely related, when search isn’t discouraged, the same friction has been argued to generate trailing spouses and contribute to gender inequality (Venator 2021). The goal of our work is to quantify the impact of this “colocation friction” on job mobility, migration rates, gender inequality, and social welfare. To do so, we develop a dual-earner framework with directed search and multiple job applications across commuting zones. At the core of our study is a (fictitious) benchmark, in which spouses can correlate their individual matching outcomes through the use of “correlated submarkets”. E.g., if spouses search in perfectly correlated submarkets, then this ensures that if one of them gets a job offer in a certain commuting zone, their spouse does so as well, effectively shutting down the colocation friction. Using this benchmark as a “measurement device”, we identify job searches for which the colocation friction is binding as job searches where spouses would prefer to correlate their matching outcomes if given the choice. In a general theorem, we provide exact conditions when this is the case. Intuitively, the friction is binding whenever the costs of one spouse becoming a single-earner outweighs the “hedging gains” from uncorrelated search. We identify four forces that contribute to the friction becoming binding: (i) both spouses being well-matched in their current location, (ii) steep job ladders/high returns on job experience, which raise the cost of becoming a trailing spouse, (iii) large migration costs, (iv) large declines in matching rates from “narrowing” the job search to a single location. The quantitative exploration (in progress) leverages a novel indirect inference method, exploiting a “super block-recursive” structure of our model. Specifically, we show that the solution to the system of Hamilton–Jacobi–Bellman equations in our model neither depends on the cross-sectional distribution, nor does it depend on certain location-specific preference shocks. This allows us to sidestep one of the computationally most expensive steps of our estimation procedure by directly inverting the ergodic distribution of the model over commuting zones and matching it to its empirical counterpart (backing out the necessary distribution of location-specific preferences shocks along the way). Methodologically, this strand of my research also includes “Optimal Delegated Search with Adverse Selection and Moral Hazard” (Theoretical Economics, 2016). In this work, I explore how to optimally delegate a stopping-problem à la McCall (1970) to an agent (e.g., a recruiting agency) subject to two informational frictions: (i) “moral hazard”, meaning that the agent’s search behavior is unobserved (or cannot be contracted upon); and (ii) “adverse selection”, meaning in this case that the distribution of job candidates and their skills itself is only known to the agent. In light of these frictions, the agent is naturally inclined to search little and blame lackluster results on there only being bad job candidates out there. I show that the second-best optimal way to address such gaming motives is to offer the agent a menu of simple bonus contracts. Going beyond the literal application laid out here, I argue that this can rationalize the ubiquitous use of bonus contracts in a variety of situations that feature the dual-constraints of both moral hazard and adverse selection. 5 3 Long-run Dynamics of Political Regimes In “A Quantitative Theory of Political Transitions” (Review of Economic Studies, 2020), my coauthor and I document five empirical regularities about long-run regime dynamics, and provide a quantitative framework of political transitions that accounts for these facts. Specifically, the five regularities are: 1. The evolution of political systems is shaped by both revolts and democratic reforms, with revolts being about three times as likely as reforms. Other modes of transition are secondary. 2. Transition hazards are declining in regime maturity. Newly established regimes are about three times as likely to be overthrown by a revolt and about six times as likely to implement a democratic reform compared to regimes older than 10 years. 3. Transition hazards are inverse “J-shaped” in the inclusiveness of political systems: political systems at the extremes of the autocracy–democracy spectrum have smaller transition hazards than regimes near the center of the spectrum; full-scale democracies are overall most stable. 4. Revolts establish autocratic regimes; reforms establish democracies. Political systems near the center of the autocracy–democracy spectrum are unlikely to arise from either mode of transition. 5. The distribution of regime types is bi-modal, with mass concentrated towards the extremes of the autocracy–democracy spectrum. We then develop a quantitative model of political transitions that accounts for all five regularities. We demonstrate the quantitative potential by fitting the model to data on the universe of political regimes existing between 1946 and 2010. The model matches the data remarkably well. It is not only able to account for the above-listed regularities, but also quantitatively replicates the shape of transition hazards, conditional outcome distributions, and the stationary distribution of regime types (all untargeted). 4 Dynamic Oligopoly Pricing Finally, unrelated to the previous themes, “Dynamic Oligopoly Pricing: Evidence from the Airline Industry” (International Journal of Industrial Organization, 2020) documents regularities on the impact of competition on pricing dynamics. For this paper, my coauthor and I implemented a webcrawler to collect data on the price path of airline ticket prices. We use the data to shed light on how competition affects the scope of intertemporal price differentiation. We document three main findings. First, the rate at which prices increase towards the scheduled departure date is significantly reduced in more competitive markets. Second, the sensitivity of the intertemporal slope to competition increases in the heterogeneity of the customer base. Third, ex-ante predictable advance purchase discounts account for 83 percent 6 of within-flight dispersion in prices and for 17 percent of cross-market variation in pricing dynamics. While this is primarily an empirical paper, our findings hint towards intertemporal price discrimination as a likely source of advance purchase discounts. Intuitively, monopoly airlines discriminate against late booking customers with inelastic demand, but are restrained in their ability to do so in more competitive environments. In line with this intuition, one would expect that the intertemporal price gradient is particularly sensitive to competition when there is a high potential to discriminate against late booking customers in the first place, which is precisely what our findings regarding customer heterogeneity show. 7