Part 2: Do Borrowers Search for the Best Available Rates?

New Call-to-action

Last month we explored whether physical branch locations matter. We presented data suggesting that the average borrower lives closer to their lending institution than they do to work; evidence that the market for credit is more local than the labor market. We conjecture that branch locations still matter because, despite the increased frequency and ease of electronic search, the cost of searching for loans is non-trivial. In this blog, we present evidence that borrowers do not always seek out the lowest rates available to them. In next month’s blog we will explore why that is the case.

Using a sample of just under 4 million auto loans, we assess whether borrowers could have found a loan with more favorable terms by comparing any given borrowers interest rate to a sample of interest rates from borrowers in the same city, that borrowed at the same time, with similar credit profiles.  To simplify the interpretation of our analysis, we calculate a summary statistic we name “distance from lowest available rate (referred to henceforth using the acronym DLAR).”  We calculate DLAR in the following way.  First, we group borrowers within 5-point FICO score bins in a given metropolitan statistical area (MSA).  For example, all borrowers in the Fresno California metro area with FICO scores between 700 and 705 are placed in a bin. We bin all borrowers in an MSA in non-overlapping 5-point FICO bins, starting at 500-505, running up through 795-800. We further refine the bins to require that all loans in a FICO bin were originated within 2 quarters of each other.  Next, we require the bins to only include loan amounts within $1,000 of the other loans in the bin.  Finally, we require that borrowers have similar debt-to-income ratios; within 5 percentage points.  

To provide as much clarity as possible, our matching approach would create the following bins. We have one bin for all loans originated in Salt Lake City in the first two quarters of 2016 with FICO scores between 695-700 for loan amounts between $20,000-20,999 with DTI’s between 25-29. We would have a separate bin for all loans originated in Salt Lake City in the first two quarters of 2016 with FICO scores between 695-700 for loan amounts between $20,000-20,999 for borrowers with DTI’s between 20-24.  We loop through our data, holding all criteria constant while changing one attribute of the bin, for each of the 5 required bin attributes. 

Having implemented the location, time, FICO, loan amount, and debt-to-income screens we are left with 1,075,879 matched bins that have an average of about 4 loans in each bin. We then compare the interest rate on each loan in a bin against the loan in the bin that has the lowest interest rate, and calculate the average distance from the lowest available rate in the bin. This DLAR statistic thus captures the average difference between any given borrower’s interest rate and the lowest available rate in the MSA that the borrower conceivably could have found.

Our calculations yield insightful results. First, the average DLAR across the 359 MSA portfolios is 1.73 percentage points. Meaning, for the average matched portfolio of loans in the average MSA, borrowers originated loans with interest rates that were 1.73 percentage points higher than the lowest available rate in their matched portfolio. We also calculate that 33.2% of all matched loans have a loan in their matched portfolio with an interest rate at least one full percentage point lower.

The figure below plots a portion of the results as a function of FICO scores. The blue dots plot the average DLAR for each FICO bin and the red dots plot the average number of loans in each bin. The scale of the left y-axis, labeled “average borrower loss” is the average DLAR for each FICO bin.  The right y-axis measures the average number of loans in a bin The figure indicates that DLARs are largest for borrowers with FICO scores between 600 and 650. This result is not surprising given the dispersion in lending criteria for lower credit quality borrowers.


 What are the implications of this analysis? First, our results suggest that borrowers do not always shop for the lowest available rates. The fact that otherwise similar borrowers originate loans with meaningfully different interest rates suggests that either borrowers find it too costly to shop for the lowest available rate, or that features of the loan-originating process other than interest rates influence loan decisions. Second, our results indicate that the market for auto credit is not as efficient as one might expect. Were markets operating efficiently, borrowers of similar credit quality with similar demand would originate loans with similar interest rates. Finally, this analysis indicates, perhaps unsurprisingly, that lenders employ substantially different pricing rules.

In next month’s blog we dig deeper into the price dispersion phenomenon, exploring potential explanations for the surprising degree of price dispersion we find in the data. Hint: search costs are alive and well, even in a digital world!

Taylor Nadauld

Chief Economist