- DLW 2012 Markups and Firm-Level Export Status
- DLEU 2020 The Rise of Market Power and the Macroeconomic Implications
- Imbens, G. and Xu, Y. (2024). Lalonde (1986) after nearly four decades: Lessons learned.
- Arkhangelsky, D. and Imbens, G. (2024). Causal models for longitudinal and panel data: a survey.
I aim to address selection bias in the average comparisons between private-sector-only firms and government contractors. First, I assume that selection depends only on observable lagged outcomes and pre-treatment covariates, within a trimmed sample with sufficient overlap in the propensity score. Second, I control unobservable time-invariant confounders using recent causal panel data methods. My analysis demonstrates a declining trend in the treatment effect for government contractors, offering empirical evidence in favor of reforms aimed at curbing single-bidding practices and reducing political favoritism in public procurement in the Czech Republic.
Abstract:
I document the evolution of market power using firm-level data from the Czech construction sector since 2006. Contrary to the global trend of rising markups, I find that aggregate markups have decreased, declining from 40% above marginal cost in 2006 to 30% in 2021, driven primarily by firms in the upper tail of the markup distribution. By linking this data with government tenders, I examine the relationship between markups and public procurement. I find that markups are significantly higher when controlling for unobserved productivity; government contractors have price-to-marginal-cost ratios that are 0.3 higher than those of private-sector firms; and firm-level markups increase by 12% upon a firm’s entry into public procurement.
Bachelor's thesis, Charles University in Prague. DOI: dspace.cuni.cz
For my undergraduate thesis, I estimated production functions for firms in the Czech construction sector, addressing endogeneity in productivity shocks and variable input usage using the control function approach and GMM. A key contribution of this work was the creation of a novel dataset as well as the structural inference of the markup distribution.