Michael Peters

Work in Progress

The TFP Losses from Financial Frictions in Indonesia
That financial frictions can reduce economic efficiency and aggregate TFP is well known. Recently it has been argued that general equilibrium models with heterogeneous firms, uninsurable idiosyncratic productivity shocks and collateral constraints can quantitatively account for a substantial part of the cross-country variation in TFP. Using plant-level data from Indonesia, I show that the micro-implications of such models are hard to reconcile with the firm-level evidence. When the model is disciplined to be consistent with the micro data, the implied TFP losses from financial frictions are modest.

Endogenous Skill-Biased Technological Change: Evidence from a Natural Experiment
This paper exploits a large-scale population transfer to test a theory of endogenous skill-biased technological change. Between 1946 and 1947, in the aftermath of World War II, roughly 8 million Germans were expelled from the German Eastern Territories and transferred to Western Germany. The settlement in Western Germany was not uniform, but showed substantial heterogeneity in the allocation of refugees across Germany's 260 counties. Given that in the pre-war era the Eastern Territories were much more agricultural than their western counterparts, I exploit this cross-county variation in refugee shares as quasi-experimental variation in local skill supplies and study the long-term effects on wages, sectoral employment patterns and the type of technologies used on the factory floor.

Costly Information Processing: Empirical Evidence (with Stefan Hoderlein)
How do individuals form expectations? In this paper we focus on two aspects. We present a general framework that allows us to (1) characterize the content of individual information sets and (2) test if information processing is costless. To do so we only require common regularity conditions on agents' preferences and agents' information sets can be partly unobserved by the econometrician. In particular, our approach does not require individuals to have rational expectations. As an empirical application we analyze income expectations from a large cross-section of households. Preliminary results suggest that individuals' information sets are coarse. We further argue that our findings are hard to reconcile with a theory where information processing is costless.