Business dynamism – the process of firm entry, growth and exit – lies at the heart of modern endogenous growth models. While productivity differences have traditionally been seen as the main driving forces of business dynamism, a growing body of evidence suggests that customer acquisition frictions are at least as important. In light of this evidence, we propose a novel endogenous growth model in which innovating firms must first acquire customers to sell their products. Estimating our model with aggregate and firm-level data, we find that expansions of firms’ customer bases (market sizes) boost their incentives to innovate and shift resources towards high-growth businesses ("gazelles"). Combined, these effects explain over 1/3 of aggregate growth and substantially change predictions about the efficacy of growth policies. Finally, we document support for key model predictions using firm-level micro-data.

How do different firms respond to R&D incentives and, in turn, shape aggregate growth? We develop a novel empirical framework, grounded in endogenous growth theory, allowing us to measure firms’ responsiveness to R&D incentives and to aggregate such responses. After validating the predictions of our framework using several micro-datasets, we apply it to Compustat data. We find that (i) ignoring firm heterogeneity severely under-states the aggregate effectiveness of R&D incentives, (ii) per dollar spent on R&D incentives, young (rather than small) firms raise aggregate growth the most, and (iii) our results are robust to knowledge spillovers, dynamics, and borrowing constraints.

Consider a union of atomistic member states, each faced with idiosyncratic business-cycle shocks. Private cross-border risk-sharing is limited, giving a role to a federal unemployment-based transfer scheme. Member states control local labour-market policies, giving rise to a trade-off between moral hazard and insurance. Calibrating the economy to a stylized European Monetary Union, we find notable welfare gains if the federal scheme's payouts take the member states' past unemployment level as a reference point. Member states' control over policies other than unemployment benefits can limit generosity during the transition phase.

Using a large and representative panel survey of German firms, we document sizable forecast errors in employment growth which decline with firm age and which are related to investment and R&D activity. Motivated by this evidence, we build an endogenous growth model with heterogeneous firms which learn their productivity from noisy signals, decide about innovation activity, employment, and exit. Aggregate productivity growth responds to a selection channel via firm entry and exit and to an innovation channel via R&D investments of heterogeneous firms. We calibrate the model to replicate the realized and expected firm growth rates over the firms' lifecycle in our data. We use the calibrated model to quantify the role of information frictions in the selection and innovation channels behind aggregate productivity growth.

The increase in the relative supply of college-educated workers has transformed the labor force in every developed economy. How does this secular trend affect the characteristics of firms in the economy? To answer this question, I develop a general equilibrium model in which both workers and firms are heterogeneous. In the model, firms of different sizes rely on different types of workers due to capital-skill complementarity in production. I estimate the model using administrative linked employer-employee data from Germany. The model predicts that the changes in the labor force composition entail the reallocation of production towards firms with a larger capital stock, which tend to be older and less dynamic. The quantitative results indicate that the skill composition of the labor force can account for most of the recently documented shift in the size distribution of firms, the falling number of new firms, and the increasing market concentration. The patterns of business dynamism across German industries provide reduced-form empirical support for the model's predictions.

This paper develops a computationally efficient method for minimum distance estimation in a class of high-dimensional, non-convex, black-box problems for which likelihood function does not exist in a closed form. As a side effect, the method enables to apply global uncertainty quantification methods at close to zero additional computational cost.