The Consumer Finance Protection Bureau (CFPB) recently announced that it was releasing consumer complaints filed with the bureau against financial institutions — more than 400,000 complaints sent to nearly 3,000 companies, to be exact.
Here at Sunlight, we were excited not only by the fact that these data were being released, but also by the multiple easy-to-use formats — you can search, get bulk data and pull from an API — and the variety of data being made available. This newfound information is a great avenue for the CFPB to amplify the voice of consumers and open up useful data in a useful way. So, we did what we do best: Dig in!
One aspect of user-generated data we often find interesting is how the results might be impacted by the limitations of an interface or platform. So we investigated one of the first choices you have to make when you file a complaint with the CFPB: What financial product are you complaining about?
There are several financial services and products to choose from in the dataset, including a variety of loans, credit cards and more. We started our investigation by looking at mortgages, however, given the CFPB’s origins in the subprime-mortgage-fueled financial crisis.
Looking at the spike and the subsequent fall in mortgage complaints in 2013 made the wheels in our head start turning. Mortgage rates hit a record low in late 2012. Were adjustable-rate mortgages causing people trouble? Could it be that many homeowners had just refinanced and were just starting to have negative experiences with new mortgage companies?
No, it turns out. The most likely explanation for the Great 2013 Mortgage Complaint Spike is much more prosaic: It has to do with the choices CFPB made in their complaint-filing interface.
When we plot all products, we observe that complaints about debt collection skyrocketed in 2013 — right as the mortgage complaints were falling. In fact, there were zero debt collection complaints before that date, leading us to conclude with relative confidence that the CFPB added it as a possible term at the beginning of 2013. We believe that the most likely explanation for the upswing in mortgage complaints is that people were learning about the new tool, and starting to use it. The most likely explanation for the fall is that many of those complaints were really about mortgage-related debt collection. As soon as debt collection was an option, users started classifying their complaints in that more appropriate category.
Indeed, the availability of categories may have influenced the kinds of complaints that consumers filed. Prior to 2013, there were no complaints at all about companies like [Encore Capital Group](http://www.encorecapital.com/), [Enhanced Recovery Company, LLC](http://www.ercbpo.com/), [Expert Global Solutions, Inc.](http://egscorp.com/), and [Portfolio Recovery Associates, Inc.](http://www.portfoliorecovery.com/), all of which specialize in collections. After 2013, these companies broke into the top 50 targets of complaints.
The full text of complaints is available from the CFPB’s data source starting in 2015. If it were available earlier, we could use a text classification algorithm to attempt to reclassify complaints into debt collection and other new categories (credit reporting and payday loans were also added since the portal was launched). Text is available for complaints starting in 2015, so that should be possible for future additions to CFPB’s schema.
In conclusion, we’re really excited to see this data become public and can’t wait to see how the CFPB and others use it. It’s a great way for the CFPB to show how it’s impacting consumers, and for consumers to see how companies interact with regulatory infrastructure. In addition, the way complaints were filed in this dataset is a great reminder that the architecture of a system can have a serious impact on top-line results that we hope is taken into account in all opinion-gathering contexts.
Find something interesting in the data yourself? Let us know in the comments below!