What the Data Can Really Tell Us
The difficult question in the U.S. bank-lending evidence is not whether the exchange-rate uncertainty story sounds plausible. It is whether the analysis measures exposure before the shock, compares banks in a meaningful way, and tests explanations that could produce the same rise in margins or fall in lending. This chapter explains how the paper uses lagged loan intensity to make that comparison, then matches DealScan, SLOOS, and Call Reports to the claims each can support. The goal is a disciplined judgment: robustness makes the channel more credible, but it does not turn an observational working paper into a fully exogenous experiment.
Why lagged loan intensity is an exposure proxy
The quantitative analysis uses U.S. commercial-bank Call Reports from 1995Q1 through 2019Q3. The panel is unbalanced and contains roughly 40–70 of the largest banks. The selected institutions have average assets above $10 billion, at least seven years of observations, and at least a 20 percent loan-to-asset ratio. Together, the sample represents about 60 percent of the U.S. commercial-banking asset base. Those criteria give the analysis a substantial view of large-bank lending, but they also define its population. The results speak most directly to large banks with meaningful loan activity, not automatically to small community banks or every form of U.S. credit.
The paper would ideally measure each bank’s direct participation in loans held for sale or in the syndicated loans market. In practice, the loans-available-for-sale field is not populated consistently across the observations. The author therefore uses lagged loans divided by interest-earning assets as the exposure measure. Loans and loans-held-for-sale have a reported positive correlation of 0.7 in the observations where both are available. A bank with a higher share of loans in its earning-asset portfolio is treated as more likely to face the distribution and funding pressures associated with a loan-intensive business.
This is a proxy, not a literal inventory of syndication commitments. A high loan-to-interest-earning-assets ratio may reflect other features of a bank’s business, and not every loan is sold through the syndicated market. The proxy is useful because it captures a pre-existing difference in the amount of loan exposure a bank must manage. It becomes informative when the bank’s outcomes respond more strongly to exchange-rate volatility than the outcomes of otherwise comparable banks with lower lagged exposure.
The timing of the exposure is central. The paper interacts the lagged share with the volatility of the dollar index and with the change in the index’s level. Because the exposure is measured before the current quarter’s exchange-rate movement, the current shock cannot have caused that particular lagged share by construction. The design therefore asks whether pre-existing exposure changes the bank’s response when volatility rises. It is a cross-bank comparison embedded in a panel, not a random assignment of banks to a currency episode.
The regressions include bank fixed effects and time fixed effects.