Job Market Paper
When interest rates rise, standard models predict lending should decline and vice versa. However, a number of papers have empirically documented aggregate commercial and industrial (C&I) loan volume puzzlingly increases (decreases) in response to contractionary (expansionary) monetary policy. This paper revisits this idea of the "C&I loan puzzle," allowing for asymmetric responses to monetary policy shocks. I show that aggregate C&I loans fall in response to a contraction as standard models expect, but C&I loans still puzzlingly decline about 15% after an expansionary shock. I find this decline is concentrated in the top 1% of C&I lenders, while most banks are unresponsive to the shock. Because large firms tend to borrow from large banks, I use firm-level Compustat data to provide evidence consistent with two possible mechanisms. Large firms substitute away from bank borrowing and toward corporate bonds, consistent with bond rates adjusting more quickly than bank rates. These firms also refinance higher-rate bonds at the lower prevailing market rates. This paper rejects the existence of a contractionary C&I loan puzzle and provides a plausible explanation for the expansionary C&I loan puzzle.
with Jim Rauch
Large firms are dominated by conglomerates: firms that own other, legally independent firms. Conglomerates form by acquiring existing firms and generating new firms. To be able to discern generation from acquisition we study the formation of new conglomerates, distinguishing between "C-conglomerates'' that have peak firms with employees and "H-conglomerates" that have holding companies at their peaks. We use data from Brazil for the period 1987-2014. We find that existing firms dominate the economic activity of new H-conglomerates and that new firms, mainly the peak firms, dominate the economic activity of new C-conglomerates. Defining granules as the largest four entities in their two-digit industries, we identify 35 or 42 new C-conglomerates and 9 or 12 new H-conglomerates as granules. Less than 15 percent of new Brazilian conglomerate granules, nearly all of them H-conglomerates, would not qualify as granules if reduced to their largest firms. Our results indicate that in Brazil most new granules are safely modeled as unified, superstar firms.
Undergraduate Thesis
In financial markets, banks play a key role in transforming illiquid assets into more liquid assets. However, their ability to spread the risk of liquidity shocks over a body of agents generates a positive probability for non-efficient bank runs. Building off of the classic Diamond-Dybvig framework, this paper uses an agent based model to observe the two equilibria, efficient risk sharing and the bank run. While previous literature has looked at under what conditions could a bank run equilibrium occur, this proximity based learning model (PBLM) focuses on the development of a panic driven bank run in light of limited information, proximity based learning, and localized interactions among heterogeneous agents. This simulation approach is novel in that it allows for the inclusion of more realistic conditions (e.g. heterogeneity and learning) that would make such a model difficult to solve, if not mathematically intractable. This paper finds proximity based learning to be an effective method of communication and a panic transmission mechanism when consumers only have limited information.
with Miguel Ramirez
This paper presents indirect evidence that absolute purchasing power parity (PPP) may hold in the long-run between Mexico and the U.S., but due to data limitations, the relationship could not be tested directly. Thus it is not clear if absolute PPP holds in the long run between the U.S. and Mexico. Given that relative PPP is a necessary, but not sufficient, condition for absolute PPP to hold, this study tests the relationship between the change in the log of the exchange rate, and the changes in the log of the U.S. producer price index (PPI) and the Mexican PPI. Here, the absence of relative PPP would indicate that absolute PPP could not hold. Given that all the relevant variables in first difference log are stationary, PPP in its relative form holds and OLS can be applied directly in a VAR model setting, viz., treating all variables initially as potentially endogenous. The estimates indicate one-way Granger causality from the percentage change in the exchange rate to the percentage change in the Mexican price level, which is not an implausible result for an emerging nation such as Mexico which imports a significant fraction of (dollar denominated) intermediate products and capital inputs.