Krzysztof Pytka

Looking into the Consumption Black Box: Evidence from Scanner Data

(jointly with Daniel Runge)

• PDF
• 120min Slides
• BibTeX

This paper examines the heterogeneity and persistence of household non-durable consumption. We address three questions: (i) Do different consumer groups buy different products? (ii) How persistent are individual choices? (iii) What are the implications for structural models? We find minimal differences in basket composition between rich and poor households and high individual instability, with only 38% of products repurchased annually. To explain this, we propose a "shopping spree" model where products are perfect substitutes and baskets result from random sampling. Our findings serve as a cautionary note for structural models that emphasize product and consumer sorting.

Shopping Frictions with Household Heterogeneity: Theory & Empirics

Revise & Resubmit at the Review of Economic Studies

• PDF
• Slides
• BibTeX

This paper examines the impact of price dispersion on household consumption, highlighting the role of economic status in shaping purchasing behaviors. Leveraging detailed scanner data, I document high-earning employees pay 1.5 to 7% more than lower-earning ones for the same or similar goods. A causal link between income and the prices is established using the Economic Stimulus Act of 2008. The findings indicate that 8 to 22% of the increase in household spending following a transitory income shock is due to higher prices paid. Despite a broader variety in the consumption baskets of wealthier households, very few goods are tailored to specific income groups. Integrating consumer search with the savings problem, I propose a new model to reconcile the observed patterns and quantify the impact of retail-market frictions on consumption. Counterfactual analysis shows that over two-thirds of households face higher prices due to a price externality.

Understanding the Sources of Earnings Losses After Job Displacement: A Machine-Learning Approach

(jointly with Andreas Gulyas)

Accepted at the Journal of Labor Economics

• PDF
• BibTeX

Using generalized random forests in a difference-in-differences framework, we analyze heterogeneity in earnings losses following job displacement. We find substantial variation in short-term losses ranging 20-70%. While all workers face long-term losses, employment rather than wage changes primarily drives heterogeneity. The most important predictor for employment losses is worker's age, while the firm wage premium is the most important one for wage losses. We predict similar loss patterns for the population excluded by the typical sample restrictions in the literature. We further show how to create simple decision rules to target active labor market programs to high-loss individuals.

The Consequences of the Covid-19 Job Losses: Who Will Suffer Most and by How Much?

(jointly with Andreas Gulyas)

CEPR Covid Economics (47)

• PDF
• BibTeX

Using the universe of Austrian unemployment insurance records until May 2020, we document that the composition of UI claimants during the Covid-19 outbreak is substantially different compared to past times. Using a machine-learning algorithm from Gulyas and Pytka (2020), we identify individual earnings losses conditional on worker and job characteristics. Covid-19-related job terminations are associated with lower losses in earnings and wages compared to the Great Recession, but similar employment losses. We further derive an accurate but simple policy rule targeting individuals vulnerable to long-term wage losses.