Lending: A Geographical Analysis
In this post, we will be exploring some geographical trends in the online lending environment. In particular, we will be looking into origination characteristics on a state-by-state basis, as well as relationships between some of these metrics across states. The information presented includes data from four consumer originators integrated into Orchard’s platform, with names of originators omitted to maintain confidentiality. The data covers one year’s loan originations, from April 2015 through March 2016.
In our first graph above, we display total loan originations by state. As we might expect, the most populous states – California, Texas, New York, and Florida – show the highest levels of loan originations. Note that Iowa is not covered in this or any of the graphs below, as lending is not offered there by any of the covered originators.
The graph above shows the weighted average interest rates (weighted by loan size) across states. The states show a large degree of dispersion, from an average of 12.7% in Maine to 19.1% in Idaho. This large range is not an issue of low sample size, as both of these values represent origination volumes in excess of $15 million. This dispersion indicates either significant differences in underwriting or in borrower populations across states.
Average loan size also varies by a large degree across states, from a low of $10,614 in Idaho to a high of $15,790 in North Dakota. Note that Maine has the second-highest average loan size at $15,597. The presence of Idaho and Maine at the extreme ends of both the average interest rate and average loan size spectrums indicates that there may be some relationship between these factors, which we will explore later on in this post.
Weighted average debt-to-income ratios also show very large differences across states. New York comes in with the lowest average DTI, at 18%, while Nevada shows the highest at 26.8%. We might expect borrowers in Nevada to have higher interest rates than those in New York, reflecting the higher risk associated with these DTI values, and if we check the interest rate graph above this is in fact what we see.
In the next few graphs, we will analyze some of the relationships between the above credit variables across states. In each case, we will present a scatterplot along with a best-fit regression line, regression coefficients, and the associated r-squared.
The relationship between interest rates and loan sizes across states is shown above. We see a strong negative relationship here – as loan sizes increase, interest rates tend to fall. This corroborates the initial relationship we saw earlier with Idaho and Maine on the extreme ends of both of these scales. The r-squared associated with our regression is .764, meaning that 76.4% of the variation in average interest rates can be explained by average loan sizes across states. This negative relationship between interest rate and loan size is somewhat counterintuitive, as higher loan sizes increase an originator’s concentration risk and should command higher interest rates, all things being equal. As we discussed in our recently published quarterly industry report, we believe this indicates that originators employ stricter underwriting criteria for borrowers at higher loan amounts. Originators may be willing to lend small amounts to borrowers who they would not otherwise consider for higher loan sizes. As loan size increases, approval criteria becomes stricter and we see average interest rates fall.
The above graph shows the relationship between interest rates and debt-to-income ratios across states. A strong positive relationship is seen here – as DTI increases, interest rates tend to increase as well. This trend makes sense, as higher DTI ratios tend to indicate riskier borrowers. The r-squared associated with our regression is .407, meaning that 40.7% of the variation in average interest rates can be explained by average debt-to-income ratios across states.
This post has demonstrated that credit characteristics vary considerably across states. In future posts, we hope to explore how various macroeconomic factors–unemployment, for example–might relate to state-level credit characteristics and performance.