Auto Title Lending Data

Todd Zywicki has written several articles (here and  in a fuller version here) on auto title pledge lending that cite default rates on auto title loans are 14-17%, while repossessions occur only in 4%-8% of cases and in 20% of those cases the borrower redeems the car. These numbers are cobbled together from several disparate sources, so they might not all fit together, but from this (and borrower characteristics) Todd concludes there's no basis to claims that auto title lending is predatory.

These numbers long seemed too good to be true to me—they would imply either huge profit margins (which Todd disputes) or huge overhead costs (Todd's explanation). They also seemed highly skewed by the fact they were counting loans rather than borrowers. Title loans are 30-day loans that can be rolled over, but a roll-over counts as a new roll, which effectively inflates the denominator for default rates. 

I came across a rather obscure Tennessee Department of Financial Institutions study (actually cited by Todd) that has some numbers from examinations of title lenders, and it seems to tell a somewhat less rosy story about title lending.

Following a change in Tennessee's auto title pledge law, the Tennessee Department of Financial Services put out two reports (here and here) on auto title lending in the state based on its examinations of title lenders. The 2008 report has data from 2006. There were 139,319 new title pledge loans made in Tennessee in 2006 (roll-overs excluded), all for $2,500 or less. The average loan was $557.70.

By my calculation, the duration of the average loan was 4.95 months (it actually might be longer, but the report lumps together all longs with 11 or more rollovers—I treated them as all having exactly 11 rollovers for my calculations), meaning there were on average approximately 4 roll-overs or that the real term of a title loan was 5 months on average. 88% of loans outstanding had rolled over at least once. 

Tennessee title pledge lending resulted in 18,199 vehicles being repossessed in 2006. While there is not perfectly matched data on loans and repossessions (repossessions in 2006 include loans made in both 2005 and 2006), if we assume a steady level of auto title lending (which probably isn't right--the number of outstanding loans was 20% higher in Dec. 2006 than Dec. 2005), this translates into 13% of loans (18k/139k) or approximately one in seven resulting in a repossession (not default) rate. That alone is significantly higher than the rate reported by Zywicki.

I had previously been taking this 13% and multiplying it by the roll-over rate, which resulted in a eye-popping 65% repossession rate.  Jim Hawkins, in a very helpful comment rightly noted that there's no reason to multiply this figure by the roll-over rate as the denominator is only "new" loans, not roll-overs.   The number that I've been interested in is not the percentage of loans that result in repos, but the percentage of borrowers who lose their cars annually.  The 139k figure is still likely an inflated denominator for my purposes, as it double counts individuals who took out multiple title loans in a year, but whatever the resulting increase, it surely doesn't match the roll-over rate. Let's say it means a 15%-18% repossession rate. If so, that's quite different from 4%-8% (but hardly 65%...). Todd also cites a range of LTVs on title loans from 33% to 100%; if these loans are at the lower-end of that range, that implies an awful lot of equity stripping going on, which makes me think that title lending is often predatory.   

A second point of interest--Tennessee has an (extremely high) pricing cap for title loans:  22% per month.  But only about 60% of lenders are pricing at the cap. A substantial minority are pricing beneath it. And the industry has healthy profit margins--13% on average before taxes. And this doesn't include money taken out by owners in the form of salary or other benefits). That doesn't fit with Todd's explanation of low margins due to high overhead costs and stiff competition. To me these sort of margins suggest that the usury cap isn't restricting credit in a major way and that it might be too high.  

Now maybe title lending in Tennessee is just different from everywhere else. For example, if Tennessee permitted much higher rates than other states, then maybe the industry is able to accommodate a riskier borrower base, which would translate into more defaults and repossessions. Tennessee does have a very high bankruptcy filing rate.  It's not clear whether there's any correlation between title borrowers and bankruptcy filers (Skiba and Tobacman find a correlation for payday and bankruptcy, albeit of relatively small magnitude if I recall), so maybe Tennessee title borrowers are an unusually risk group, which would account for the high repo rate. I have no way of knowing if Tennessee is unique or typical, but the Tennessee data certainly makes me hesitate to accept Todd's conclusions about title lending.

A final thought: this Tennessee Department of Financial Services report is really good. It's a shame we don't see more of this sort of study from state bank regulators, and that they don't coordinate studies. I think this sort of data is a real service for policy issues. Hopefully this sort of examination-based study is what we will see the CFPB producing.

So here's a question for Credit Slips readers: am I reading this Tennessee data correctly? Any thoughts on its interpretation? 

[Note:  I've updated this post.  1.4.11 11:50pm.]

Comments

[Don Lloyd]

Any time you come up with an interest rate of ~900% you should realize that your analysis isn't going to be worth anything.

What's needed is take a typical case and work strictly in dollars and cents, and carefully understand what alternatives, if any, the borrower has.

If I start a business which loans out cash at 100% a month, I'm still not going to make a profit unless the loan balances are larger than a couple of bucks.

Regards, Don

[Jim Hawkins]

I read the report differently, although it is possible I am misreading it.

I assume you calculated the repossession rate by dividing the number of repossessions by the total number of loan divided by the rollover rate (or 18,199 vehicles being repossessed / (139,319 new title pledge loans / 4 rollovers on average)). I think the real repo rate would be just the number of vehicles repossessed divided by the total new loans (or 18,199/139,319), which is around 13%. The reason I think rollovers should not reduce the denominator is that the report makes clear that the 139, 319 number is just new loans and not rollovers. Or, said another way, 139,319 is the number of title loan customers in Tenn. in 2006. It says "For reporting period January 1, 2006 to December 31, 2006, a total of 139,319 new title pledge agreements were made. This figure reflects new agreements made and does not include renewals of these initial agreements."

I asked a major title lender today what its repo rate was per customer (not per loan), and it indiciated it tracks that information by customer and its repo rate is between 5 and 6%.

But, I wonder if I am missing something in reading the report since I got such a different number.

[Adam Levitin]

Jim--thanks! I've incorporated your comment into a revised posting.

Don--yes, APRs are a terrible measure of cost for short-term, small-dollar loans. So let's put some numbers on it. To borrow $557.70 for 30 days, you'd have to pay $122.69 at most shops in Tennessee. If you rolled that loan over 5 times (the average), you'd have to pay $1281.21 to have borrowed $557.70 for 180 days. Does that look so much better than a 900% APR?

[Knute Rife]

In my jurisdiction, the folks who take out title and payday loans are not the ones who file bankruptcy. They can't afford it, or at least they believe they can't. Instead, they improvise what transportation they can after they lose the car and hope they can hide their wages from garnishment.

@Jim: You might want to ask the lender if its repo figure includes assignments to collection or only counts repos by its in-house crew.

[Jim Hawkins]

Good question. Thanks!