Should I Stay or Should I Go? Bandwagons in the Lab with T.-R. Heggedal & Leif Helland (JEBO June 2018)

We experimentally test the seminal platform selection model of Farrell and Saloner (1985). At the core of this model is the presence of irreversible actions and private valuations. In general, our data support the model. While complementarities in actions strongly determine follower behavior, there is a reluctance to lead not accounted for by theory. We explain observed deviations from the neoclassical equilibrium by injecting some noise in the equilibrium concept. We find that allowing cheap talk messages improves efficiency while policies aimed at insuring failed leadership or subsidizing joint choice of the challenger platform reduce efficiency.

Working Papers

Dispersion Over the Business Cycle (with M. Carlsson and A. Clymo)

We use rich Swedish micro-data to show that increased dispersion during recessions is primarily a demand-side phenomenon. The key novelty of our analysis is that we use goods level data on prices to estimate firm-level demand shocks, and production-line-level data on reported capacity utilization to accurately measure firm-level supply (TFPQ) shocks. We document that the dispersion of both TFPQ and demand growth across firms rose during the Great Recession, but that the increased dispersion in TFPQ growth is reduced by up to 1/4 after controlling for capacity utilization. We then perform a semi-structural variance decomposition exercise for firm-level sales growth. We show that 2/3 of the increased dispersion in sales growth in 2009 is explained by the increased dispersion of demand, while TFPQ dispersion plays essentially no role. Key to this finding is that we estimate a low level of passthrough from TFPQ shocks to prices, limiting the ability of increased TFPQ shock dispersion to affect sales dispersion. Consistent with this, we find evidence that demand curves are “kinked”

Convergence Analysis for Experimental Economics (with S. Groenneberg)

In the analysis of experimental time series, economists are often interested in whether or not behavior converges to a particular level. Answering questions about convergence enables researchers to evaluate theory and to establish new stylized facts. If an experimental institution reliably reproduces the same outcome or distribution of outcomes, this is a basis for economic knowledge. In this paper, we address issues related to convergence that are relevant for experimental economists. We begin by considering methods rely on a fixed rate of convergence to assess the long-run behavior of experimental time series. We show that this class of methods have a number of deficiencies. Some of these deficiencies can be corrected. In particular, we work out the standard errors needed for correct inference. However, we also show that the such methods are sensitive to misspecification. Researchers should therefore only use such methods unless they believe that convergence has this form. Next, we propose an alternative characterization of convergence. In this alternative characterization, the goal is to identify a point beyond which the time-series no longer exhibits any systematic trends. Based on this characterization of convergence, we suggest an approach for assessing convergence that may useful to researchers. Our approach is simple to implement, intuitive, and facilitates disciplined conclusions about when behavior has become regular.

Markups from Inventory Data and Export Intensity (with M. Carlsson and T. B. Milicevic)

This paper studies the relationship between export intensity and the price-cost markup using detailed Swedish .rm-level inventory data on the quarterly frequency. We proceed in the spirit of the production function approach to measure markup variation, but impose a minimal set of assumptions on the firm’s environment. With access to a unique firm-level measure of capacity utilization, the data then allows us to cross check the identifying assumptions. Besides providing supporting evidence in favor of previous findings in the literature from a new data source, the quarterly data allows us to provide novel results on the rapid dynamic adjustment of the markup related to changes in export intensity.

Market Entry with Frictional Matching and Bargaining

This paper studies an experimental labor market that incorporates elements from the standard theory of equilibrium unemployment. Specifically, we test in a controlled lab setting a novel market entry game that includes matching frictions and wage bargaining. The model predicts that firms will enter up to the point at which stochastic rationing of workers equalizes the value (of a vacancy with its costs. Between treatments, we vary productivity and how wages are determined. Consistent with theory, we find that increases in productivity increase job creation and thereby reduce unemployment. We also reproduce the expected outcomes associated with different wage institutions. When wages are determined by bargaining after match, firms face a hold-up problem. As a consequence, job creation collapses when workers have excessive bargaining power. In contrast, when wages are determined prior to entry, workers moderate their wage claims to induce vacancy creation. Although our findings tend to align with theory, we observe some deviations. In particular, there is a systematic bias in the aggregate level of entry. There is too much vacancy creation when productivity is low and too little vacancy creation when productivity is high. To explain this bias and to account for heterogeneity at the individual level, we estimate a quantal response equilibrium in which we allow for idiosyncratic preferences for entry.

Discrimination in Small Markets with Directed Search

To understand the consequences of discrimination for labor market outcomes, it is critical to account for how discrimination interacts with labor market frictions. In an important contribution in this regard, Lang  et  al. (2005) show how discriminatory hiring leads to labor market segregation in a large (continuum) market environment with posted wages. The essential reason for this outcome is that discriminated workers apply to low wage jobs to avoid competition from workers of the preferred type. This paper makes two contributions: First, I investigate the theoretical properties of a small market analog of the model of Lang et al. (2005). Second, I test a simplified version of the model in the laboratory. In the lab, I find evidence of the postulated segregation effect. Firms learn to take advantage of discrimination and the experimental markets become more segregated over time.  Preferred worker types apply almost exclusively to high wage firms and the income of discriminated workers is reduced by about 30% relative to the income of prioritized workers. Relative to treatments without discrimination, firms also increase their payoffs by about 25%. Although we do not reproduce the complete segregation predicted by the theory, we show that the presence of discrimination affects both worker search and firm wage-setting.


  • The Causal Impact of Trade on Workforce Composition (with Lena Hensvik)
  • Monopoly, Resale, and Arbitrage: An Experimental Study  (with Christian Riis)