When Volatility Withdraws the Lenders
This chapter examines the syndicated-loan evidence behind the exchange-rate uncertainty channel. The specific question is whether higher volatility in the dollar’s foreign value is associated with the combination the mechanism predicts: less participation and loan volume, but higher prices for the loans that still trade. The reader’s task is to treat those movements as clues about funding supply, while keeping the DealScan sample, the exposure comparison, and the limits of causal inference in view.
What the DealScan sample can see
The data come from Thomson Reuters Loan Pricing Corporation DealScan, a facility-level database covering 1990 through 2018. For each observed loan, the paper can examine features such as the amount, type, currency, purpose, borrower, lender, and members of the syndicate. That makes DealScan useful for the first part of the proposed chain. It can show who supplied funds, how much credit was arranged, and what price was attached to the facility. It cannot, by itself, reveal every balance-sheet choice made by the individual banks involved.
The paper’s selected sample is designed around loans with a plausible connection to real corporate financing. It focuses on senior loans with completed deal status, keeps syndicated observations and sole-lender observations, and removes outliers in loan volume and spreads. The analysis also concentrates on corporate purposes, debt repayment, working capital, takeovers, leveraged buyouts, and recapitalizations. Those choices make the sample more interpretable for the question of commercial credit, but they also define what the results do not cover. This is not a complete census of all U.S. loans, all bank assets, or every type of credit contract.
The sample’s composition explains why it is a useful setting for the currency question. Approximately 99 percent of the loans in the U.S.-syndication sample are denominated in dollars. About 40 percent of loans are held by foreign banks or institutional investors. Roughly 60 percent are revolving loans or credit lines, with the remainder term loans. The borrowers are predominantly connected to the United States: about 98 percent are U.S.-based, and 88 percent are corporations. These figures are descriptive facts about the selected historical sample, not current market shares. Their significance is structural. A U.S. corporate borrower can be relying on a funding chain that includes institutions whose capital and obligations are expressed in other currencies.
The distribution of responsibility inside a syndicate matters too. Lead arrangers have an average bank allocation of about 30 percent in the paper’s sample, while non-lead participants or other agents have an average stake of about 7 percent. The lead bank is therefore not merely an administrator. It usually carries a larger share, helps determine the terms, and has a stronger interest in finding participants for the facility. If a foreign participant reduces a small allocation, the effect may be absorbed. If the market becomes less willing to take the exposure the lead bank expected to distribute, the originating bank’s own pipeline becomes more consequential.
The paper tracks the price of a syndicated loan with AllInDrawn. This is the amount the borrower pays in basis points over LIBOR for each dollar drawn, combining the loan spread with any annual or facility fee charged by the bank group. It is not the same as a U.S. bank’s net interest margin. AllInDrawn is a facility-level price paid by a borrower when funds are drawn; net interest margin is a bank-level income measure that compares interest income and interest expense with interest-earning assets. Keeping those objects separate is essential when the two parts of the paper are read together.
DealScan can thus answer a focused set of questions. Does the amount of lending move with exchange-rate volatility after the dollar-level change and loan type are considered? Does the average facility price move in the opposite direction? Do foreign and U.S. bank stakes change together? And does a lender with more exposure to a facility show a different spread response? Those are not four versions of the same statistic. Together, they show whether the market’s composition and terms move in a pattern that makes the proposed supply bottleneck plausible.
Why lower volume and higher spreads point to supply
A data set of this kind is best understood as a lens on the loan market, not as a direct reading of a bank’s intentions. A facility record tells us that a deal was arranged and provides a price, amount, purpose, and set of lenders. The lender records tell us how the exposure was divided. Linking the two lets the paper ask whether a change in uncertainty is associated with different participation decisions, but it does not reveal every conversation in which a bank considered a deal and declined it. Unobserved potential loans are especially important: a fall in completed volume can reflect loans that were never launched, not only facilities that reached the market at a smaller size.
The paper’s sample restrictions sharpen one question while narrowing another. By focusing on completed senior loans and corporate purposes, it reduces noise from contracts that are less relevant to business financing. It can then compare facilities with broadly interpretable uses, such as working capital, debt repayment, takeovers, recapitalizations, and other corporate purposes. But excluding incomplete deals and removing outliers means the data do not describe all attempted financing or all extreme episodes. The sample can show how observed transactions changed; it cannot count every borrower that was unable to obtain a loan or every bank that quietly withdrew from a prospective syndicate.
The combined foreign-bank and institutional-investor share also needs care. Institutional investors are not foreign banks, and they may have different mandates, funding structures, and currency exposures. The paper groups them in some descriptive comparisons because both are non-U.S.-bank sources of funds in the selected market. That grouping supports the claim that an important part of the syndicate is outside the balance sheet of a domestic commercial bank. It does not show that all those institutions respond to exchange-rate volatility through the same mechanism. Foreign-bank retrenchment is the paper’s central interpretation; the broader non-U.S. participation share is context for why that interpretation could matter.
The price variable has a similar scope. AllInDrawn is calculated for dollars drawn and includes the loan spread plus an annual or facility fee. It is closer to the borrower’s cost of using the facility than a bare coupon spread, but it is still not a complete measure of the cost of credit. A borrower may value an undrawn commitment, face collateral or covenant requirements, or receive a different price because the facility’s risk and maturity changed. The paper controls for loan type and, in additional checks, for primary purpose, borrower industry, and borrower type. Those controls make the spread comparison more meaningful, but they do not eliminate every form of contract redesign.
Finally, the aggregate bin-scatter relationship is not an experiment in which the same borrower and bank are observed under two randomly assigned volatility levels. High-volatility quarters may contain other changes that affect the market. The paper’s next layer of analysis gains leverage by comparing lender stakes and asking whether exposure changes the spread response. That is a stronger test of the proposed path than an unqualified time-series correlation, because it uses variation in how closely a lender is tied to a facility. It remains a test within observational data, however, and must be read together with the later bank-level design rather than treated as the final answer.
The first pattern is a joint one. In the paper’s aggregate evidence, higher exchange-rate uncertainty is associated with lower total loan volume in the syndicated market and higher average AllInDrawn spreads. The analysis controls for changes in the dollar’s level and for loan type; additional checks control for available borrower, lender, primary-purpose, and industry characteristics. The result is not that every loan becomes more expensive or that every period of lower volume has the same cause. It is that the price and quantity of syndicated credit move in the direction expected when the supply of loanable funds becomes tighter.
Why is the combination more informative than either movement alone? Lower volume could reflect weaker demand. If companies postpone investment or reduce working-capital needs, fewer loans may be arranged even if banks are willing to lend. Higher spreads could reflect greater borrower risk, higher expected defaults, or a change in the mix of loans. But when volume falls while the price of the remaining credit rises, a supply restriction becomes one plausible explanation: funds are scarcer, and borrowers must pay more to secure them. It is a clue, not a verdict, because demand, risk, and composition can still move at the same time.
The lender-stake evidence supplies a more direct connection to the proposed mechanism. When exchange-rate uncertainty is higher, the paper finds a greater stake for U.S. banks and a lower stake for foreign banks. The author interprets the pattern as foreign-bank retrenchment and a larger pipeline for U.S. institutions. This is different from saying that U.S. banks won a market-share competition because they became more efficient. The relevant question is what happened to the missing foreign capacity. If U.S. banks take a larger stake while total volume is falling, they may be absorbing more of a shrinking market rather than expanding into abundant demand.
The paper then uses differences in lender exposure to examine spreads more closely. For each loan-bank pair, exposure is measured as the logarithm of the loan amount multiplied by the bank’s allocation. The analysis interacts that exposure with exchange-rate volatility and with the change in the dollar’s level, while absorbing time and loan-type effects. The reported quarterly effect of a one-standard-deviation increase in volatility at an average lender stake is about 52 basis points for lead U.S. banks and 79 basis points for lead foreign banks. The spread response grows with the lead arranger’s stake, and the estimate is larger for a foreign-led syndicate.
Those numbers require several qualifiers. They come from a facility-level DealScan exercise, not from the balance sheets of all U.S. banks. They are quarterly AllInDrawn effects, not annualized changes in net interest margin. The paper calls the exercise suggestive because DealScan lacks enough bank-level information to tie the observed spread response cleanly to an individual bank’s balance-sheet behavior. A reader should therefore use the estimates to understand exposure and market pricing, not combine them with the later approximately 10-basis-point NIM estimate as though both measured the same outcome.
The logic of the evidence can be represented as a sequence of states:
The sequence should not be read as proof that the data literally observe every arrow in real time. DealScan directly supplies transaction, composition, and price information. The arrows connecting those observations to foreign-lender risk and U.S. pipeline burden are the paper’s interpretation, supported by the market structure described earlier. The value of the sequence is that it makes the interpretation specific enough to challenge. A general claim that “uncertainty hurts credit” would not require a change in foreign participation or a stronger response at more exposed lenders.
There is also an important distinction between a lender’s stake and a bank’s overall exposure. A large allocation to one facility tells us that the bank is more exposed to that transaction. It does not tell us how important syndicated loans are relative to the bank’s complete asset portfolio, how much the bank hedges the currency risk, or whether it can replace the participant with another source of funding. That is why the paper moves to Call Reports for its quantitative bank-level analysis. DealScan identifies the market relationship; it does not close the entire causal chain.
Nor does the observed shift prove that foreign banks were responding only to exchange-rate volatility. The dollar’s volatility can coincide with changes in financial sentiment, credit risk, interest rates, or the economic outlook. The paper controls for the dollar’s level and several observable loan characteristics, but control variables cannot transform a descriptive market pattern into a randomized experiment. The later SLOOS and Call Reports evidence matters because it asks whether the same uncertainty measure appears in bank officers’ reported standards and in the balance sheets of banks with different pre-existing exposure.
For now, the strongest conclusion is deliberately narrower. In the historical DealScan sample, periods of higher exchange-rate uncertainty line up with a less foreign-heavy syndicated market, lower aggregate loan volume, and higher facility spreads. Exposure-based comparisons make the supply interpretation more credible, because the lenders most connected to the facilities show stronger price responses. Yet the result remains evidence consistent with a funding bottleneck, not a final causal demonstration. The next question is whether U.S. banks’ own reported terms, margins, and quantities show the corresponding response once the loan pipeline reaches their balance sheets.