WORKING PAPERS

This paper investigates the role of mortgage refinancing in shaping the estimates of marginal propensity to consume (MPC) and its implications for fiscal policy. Using U.S. household data, we find that MPCs decrease during the year of mortgage refinancing and stabilize afterward, particularly among households with lower liquid assets, higher debt-to-income ratios, and valuable illiquid assets. The empirical evidence suggests that refinancing provides extra liquidity, reducing MPCs. We leverage on a partial equilibrium model to quantitatively assess these effects and explore the role of home equity extractions for fiscal policy. Our findings highlight a new dimension for the efficacy of cash transfers: targeted programs that consider higher MPCs of no-refinancers generate savings between 4 and 12% compared to non-targeted programs. These estimates imply $30 billion in potential savings under the CARES Act of March 2020.

Presented at: Boston College Macro Lunch Seminar (2023), Mini-Workshop - Research Task-Force on Heterogeneity in Macroeconomics and Finance at the European Central Bank (2024), 3rd Naples School of Economics PhD Workshop (2024), 1st Tor Vergata Ph.D. Conference in Economics (2024),  4th SASCA Ph.D. Conference in Economics (2024), Boston College Macro Reading Group (2024)*, 19th Economics Graduate Student Conference at Washington University in St. Louis (2024)*, European Central Bank DG-Research Internal Seminar (2024), European Central Bank DG-Economics Seminar (2024), Central Bank of Ireland Economics Seminar (2025).

The heterogeneous sensitivity of firms to aggregate fluctuations influences the dynamics of the business cycle. We study how firms’ outcomes (sales, debt, investment, and market value) respond to aggregate fluctuations (business cycle, monetary policy, uncertainty, and oil shocks) based on eight observed characteristics using the Generalized Random Forest algorithm. Using micro-level data from Compustat, we document three micro-facts about the cross-sectional heterogeneity in firm sensitivity: (1) while the linear OLS benchmark provides good estimates of the average effect, there is significant cross-sectional heterogeneity in firms’ sensitivities; (2) the importance of firm characteristics varies across dependent variables and shocks, but on average, non-financial characteristics play a larger role in explaining the heterogeneity in sensitivity; (3) sensitivity to aggregate fluctuations exhibits non-linear patterns with respect to firms’ characteristics. We study the implications of cross-sectional heterogeneity in firm-level sensitivity, developing an aggregation theory and leveraging the estimated random forest model to generate counterfactual firm-level sensitivities. We show that: (1) the heterogeneity in sensitivity amplifies the aggregate response to macroeconomic variables; (2) the impact of estimated non-linearity at the micro-level on the aggregate response is negligible; (3) non-financial characteristics shape aggregate responses more than financial characteristics; and (4) state-dependent heterogeneity in firm characteristics plays a marginal role in driving the aggregate response to macroeconomic variables.

Presented at: SNDE-IIF Workshop for Young Researchers (2024)*, Reading Group at Bank of Italy (2024)*, CFE-CMStatistics (2024)*, AI in Economics Seminar ECB (2025).

I investigate the role of high-frequency identified monetary policy shocks in the context of classic exchange rate puzzles. I use the identification strategy by Jarociński and Karadi (2020), who decompose “pure” and “information” monetary policy shocks from the Fed announcements. I show that a pure monetary policy shock - cleaned of any information effects - is able to replicate classic puzzles such as delayed overshooting, failure of perfect consumption risk-sharing (Backus-Smith), and predictable excess returns for the G6 countries vis-à-vis the U.S. dollar. The information shock generates no differences in the impulse responses of the exchange rate for the G6 countries, whereas the exchange rate depreciates instead of appreciating for a sample of emerging market economies. My results suggest that isolating the information component from central bank announcements is crucial for the dynamics of the exchange rate conditional on the sample of countries. However, the information shock does not generate robust statistics in line with the classic puzzles.

Presented at: Boston College Macro Lunch Seminar (2023).