A Detailed Investigation of State-Level Loan Pricing
Loan pricing–the setting of interest rates for individual loans–varies significantly based on geography in the consumer lending space. In particular, there are large differences among states, both in current interest rates and in the changes of interest rates over time. As of Q2 2017, state-level average interest rates range from a low of 13.30% in Vermont to a high of 17.34% in Iowa. Historically, the variance among states was even larger–in Q1 2011 average rates in Alaska were 12.59% while average rates in Nebraska were 22.61%.
In this post we’re going to explore the impact that geography has on interest rates, analyzing the distribution of current interest rates across states, state-level deviations from national averages, and changes in state-level interest rates over time. This post will primarily act as an exploratory analysis, presenting different ways to think about and visualize interest rate levels and changes in interest rates.
Above, we’ve created a table showing average Q1 interest rates by state for the consumer lending universe in every year since 2010. Lower interest rates are colored red, while higher interest rates are colored blue. This presentation helps us to easily identify aggregate changes in rates over the last seven years: the bright red left column gives way to the pale blue right column, showing a steady rise in interest rates over this period. We can also use this table to easily see changes in interest rates at the state level over this same period. For example, we can quickly see that interest rates in California have steadily increased from 13.2% to 15.2% while rates in Indiana fell from 19.0% to 15.7%. If we’re interested in seeing how rates vary among different states within a given quarter, however, this table is not the best presentation. While it’s true that we can easily see outliers–for example Vermont, New York, and Connecticut all clearly show low rates in Q1 2017–it’s harder to see additional differences. Instead, we find it helpful to use a table like the one below, where we’ve shown the deviation in interest rates from the average U.S. rate in Q1 for each year from 2010:
In this table we first present average United States interest rates in each quarter, then we show each state’s deviation from this average. Negative values correspond to lower interest rates than the national average and are colored red, while positive values show higher interest rates than the national average, and are colored green. We can now more easily identify those states with below-average interest rates or above-average interest rates. In Q1 2017, some of the below-average states include New York, Vermont, Connecticut, and Maine, while some above-average states include Iowa, Florida, Alabama, Arkansas, and Tennessee. With some reflection, we can identify some of the below-average rate states as those in the Northeast and some of the above-average rate states as those in the South, but this table does not lend itself to this type of regional analysis. For this, let’s instead turn to an animated heat map, which displays these deviations over time.
This heat map is useful because it helps us to see the geographical trends mentioned above much more clearly. For example, as the animation progresses we can identify a band through the southeast and the rust belt where interest rates are slightly higher than the national average. We see the opposite effect in the northeast, with states stretching from Delaware through Maine all showing below-average interest rates. This map also helps us show that over time interest rates have become more consistent across states, which is represented by the fact that the colors become less vibrant (indicating smaller deviations from zero) over time.
While this graph makes it easy to identify geographical trends like the higher rates in the southeast or the lower rates in the northeast, it doesn’t show clearly how rates within states have changed over time. For this, we can use a geographically faceted graph of historical interest rates:
This helps us to see a few things. First, we can identify states like Idaho, Nebraska, and Mississippi who have experienced the highest amount of variance in interest rates historically. We can also use this graph to visualize interest rate trends within particular states. For example, while most states have seen slight increases in interest rates over the last seven years, we can identify exceptions like Indiana, which showed a large drop in 2013, and Maine, which has fallen in recent months.
In the graph above we plot 2016 interest rates on the y-axis and the Q2 2016-Q2 2017 interest rate change on the y-axis. The slightly negative trend line shows that interest rates have experienced some reversion to the mean, with higher interest rate states tending to show larger reductions in interest rates while lower interest rate states have shown either small reductions or increases in interest rates over the prior year. This reversion to the mean effect corroborates our earlier finding above that interest rate differences across states have tended to reduce over time.
We know that there are differences in population characteristics across different regions–average incomes, credit scores, and education levels all vary geographically. Given these differences, we should expect that borrower state should play a role in aggregate loan-level pricing. There are many different ways to think about and visualize interest rate levels and changes, and they all help us to get a greater understanding of trends in pricing across the industry. While these discrepancies have fallen over the last several years, we believe that borrower state will continue to be a powerful predictor over the coming years, and we look forward to further explorations of the impact this factor has on loan pricing and borrower performance.