Research

Abstract. Differences in plants’ average products of capital are widely interpreted as evidence of market distortions. Yet, because plants adjust infrequently, observed capital stocks often deviate from desired levels, such that differences in average products reflect both market distortions and optimal waiting-to-adjust behavior. This paper develops a plant investment model with non-convex adjustment costs and heterogenous, time-varying distortions, alongside a Bayesian state-space framework to recover plant-specific desired capital levels from panel data. Applying this method to Portuguese manufacturing plants, I find that distortions shape both desired capital levels and adjustment incentives: plants with distortions that generate high target average products are particularly reluctant to downsize when holding excess capital. In a variance decomposition of average products of capital, market distortions account for 104 percent of total variance, while deviations from target average products account for 2 percent. A negative covariance between targets and deviations contributes -6 percent, reflecting heterogeneous reluctance to downsize and partially offsetting the other components. Dynamically, this interaction makes aggregate shock transmission depend on the distribution of distortions. These findings show that recovering individual-level distortions and accounting for their impact on optimal inaction is essential for understanding static misallocation and aggregate investment dynamics.

Abstract.  Are recessions cleansing or sullying? The answer depends on whether recessions steer resources away, in particular labor, from low-productivity firms and towards high-productivity firms. This steering capacity depends on the severity and on the nature of labor market frictions. In this project, we ask how labour market duality, that is the co-existence of permanent and temporary employment contracts with distinct adjustment costs, affects the reallocation of workers across firms and therefore the cleansing or sullying nature of recessions. Using Portuguese matched employer-employee data, we document the following facts: (i) temporary contracts represent around 30 percent of total contracts, (ii) permanent employment shows signs of irreversibility while temporary employment is flexible, and (iii) within narrowly defined sectors there is substantial heterogeneity in the use of temporary employment. We propose a firm dynamics model that incorporates these two employment types. The model features heterogeneous firms and contract-specific adjustment costs. We use this framework to characterize aggregate productivity and reallocation dynamics in downturns.  We find that following a 1 percent negative aggregate demand shock there is an initial cleansing, as low-productivity firms shed their flexible temporary workforce rapidly, followed by a prolonged sullying  as rigid permanent contracts prevent low-productivity  firms from contracting, while high-productivity firms are reluctant to rebuild their permanent workforce.

Abstract. We test a sufficient-statistics prediction for the propagation of aggregate productivity shocks in economies with lumpy investment, summarized by two steady-state moments of firms' capital-productivity gaps: cross-sectional variance and covariance with time since the last adjustment. On the macro side, we identify permanent productivity shocks from labor productivity and estimate sector-level capital responses using local projections. On the micro side, we recover the productivity process and the sufficient statistics from the size and timing of plant-level investment adjustments. Using manufacturing data from Chile and Colombia, we find that sectors with larger predicted sufficient statistics exhibit more persistent capital responses to aggregate productivity shocks. Our results show that steady-state microeconomic investment behavior contains substantial information about aggregate transitional dynamics.

Abstract.  We analyze administrative data covering the near-universe of French non-financial firms over 1997-2019. Aggregate productivity dispersion (polarization) rises only modestly, masking substantial industry heterogeneity: most manufacturing industries experience an increase in polarization, whereas in services rising and falling industries are roughly balanced, in contrast with the services-led pattern documented for other European economies. Our main contribution is to decompose the contribution to polarization from different firm dynamic factors. Contrary to traditional narratives that an increase in polarization is driven by technology developments that make it more difficult to challenge the dominance of top firms, we find that industries with larger increases in polarization did not experience greater persistence of top firms or larger declines in entry, and that polarization is driven by compositional turnover at the top of the productivity distribution rather than by further advances by industry leaders. This is particularly evident in the services sector, where the productivity premium of firms that remain at the top narrows relative to the median, while turnover at the top boundary pushes polarization in the opposite direction. In particular, top dropouts tend to leave from the lower part of the upper tail, leaving behind a more selected group of top firms. A firm dynamics model with entry, exit, and incremental and radical innovation best matches the observed decomposition when polarisation rises because of shocks to entry conditions, namely higher entry costs and higher volatility of entrants' productivity. Overall, the evidence is difficult to square with a "superstar firms" view and suggests that the link between technological change, firm dynamics and polarization is more complex than previously understood.