Similar Loans, Different Interest Rates

Mary Amiti, Anil K. Kashyap, Anna Kovner, and David E. Weinstein. "Why Do Firms Pay Different Interest Rates on Their Bank Loans?" Federal Reserve Bank of Richmond Working Paper No. 26-03, February 2026.

Among similar commercial and industrial loans, there are significant differences in interest rates, and this dispersion does not appear to be driven by risk. A recent working paper by Anna Kovner of the Richmond Fed, Mary Amiti of the New York Fed, Anil Kashyap of the University of Chicago Booth School of Business, and David Weinstein of Columbia University documented these differences in interest rates and theorized that they could be affected by how costly it is for firms to search for loans. The authors found that firms appeared to differ widely in how they search for loans: Over a third did not comparison shop at all, about half obtained only one additional quote, and the remainder searched multiple times.

Using confidential supervisory data that provided detailed information on loan risk, the authors examined loan spreads — the difference between the interest rate on a newly originated loan and a reference rate — such as the London Interbank Offered Rate, or LIBOR. They grouped loans into 722 bins based on various factors associated with loan pricing, such as identical risk ratings, rate structures (floating or fixed), and nearly identical loan purposes, sizes, and maturities.

The main source of loan spread variation was not from differences between bins, but rather from differences within bins. However, even after including 7,000 controls, the authors determined that half of the variation in spreads within the typical bin was still unexplained. They also did not find much evidence that differences in spreads within bins were caused by differences in banks' estimates of loan risk. A bank's estimate of expected loss in case of default is calculated as the product of the probability of default, the loss in the event of default, and the exposure at default. When the authors controlled for expected losses, it barely reduced the unexplained variation in spreads. Thus, the dispersion in interest rates is unlikely to be due to differences in risk.

The authors then built a borrower search cost model to better explain the differences in loan interest rates observed in the data. They found that, on average, about 37 percent of borrowers received only one quote, about 48 percent searched twice for a quote, and the remaining borrowers searched multiple times for all available quotes. The borrowers who searched several times had far lower search costs, on average, than those who took the first quote they received.

Search costs also differed with borrower risk, loan size, and information transparency. Loan spread variation exhibited patterns related to loan quality and the ease of obtaining information about borrowers. Bins with only highly rated A loans showed about half the variation compared to bins with only lower rated B loans. Additionally, loans to publicly traded firms had less spread variation than loans to private borrowers, reflecting the fact that the financial statements of publicly traded firms are audited and more available to lenders.

Through the lens of the search cost model, higher variation suggests that these search costs were higher for borrowers who were harder to screen and monitor. This is consistent with the "screening and monitoring hypothesis," which posits that screening and monitoring costs for lenders decrease when they have more private information about borrowers. While this paper represents one of the first estimates of search costs, the authors compared their estimates to those reported by small business surveys and found similar magnitudes.

Interestingly, the search costs estimated through a similar model are much smaller for corporate bonds. By grouping corporate bonds into risk bins similar to those they used for loans, the authors showed that the spreads on these bonds have much less variation. This implies that when information about firms' risk is accessible (as is the case for corporate bonds, which are assessed by rating agencies) and borrowers can more easily search for lenders, the unexplained variation in interest rates is much lower.

In this paper, the authors documented significant dispersion in interest rates for similar loans. Firms differed widely in how they shop for loans, and search costs were higher for smaller, riskier, and private borrowers. Understanding the relationship between interest rate dispersion and these underlying search costs has broad policy implications. These results suggest that reducing screening and monitoring costs, such as through improved disclosure processes or standardization, could lower borrowing costs and reduce exposure to bank-specific shocks, such as a bank failure.