RewardExpert investigates political and social patterns of consumer finance complaints across the US.
Financial services are subject to government regulation, and are affected by policy dictated by elected politicians. Naturally, financial institutions have political preferences, but do these preferences affect you, the consumer? Are the banks playing politics with you and your money? Here at RewardExpert, we wanted to find out, and we soon discovered a pattern that suggests the answer is yes.
RewardExpert looked at the consumer complaints data set from the CFPB, which includes complaints lodged against financial institutions about credit cards, bank accounts, and mortgages, and used GIS software to analyze the geographical distribution of complaints throughout the country. We wanted to see where people have more problems with financial institutions on average, and we observed that high complaint rates were concentrated in diverse, urban regions that typically vote for the democratic candidate.
We then visualized per capita rates of complaints about credit cards, bank accounts, and mortgages filed with the Consumer Finance Protection Bureau on a map of the United States. We noticed that this map looked strikingly similar to maps of county-by-county results of the last few presidential elections.
Republican-voting regions typically were regions in which few complaints per capita were received, and democratic-voting regions were ones with high per capita rates of consumer complaints. This pattern held true for both the 2012 and 2016 presidential elections.
Key Findings
- Per capita rates of consumer complaints accurately predicted the outcome of both the 2016 and 2012 elections in approximately 75% of the 3142 counties in the US. This is an indicator that, to our knowledge, has not entered into the calculations and prognostications of mainstream election forecasting.
- The so-called “Blue Wall” states of Wisconsin and Michigan were significantly more likely to vote republican than the data and election results for 2012 indicated.
- Minnesota and Maine showed up as closer to flipping than polling and conventional wisdom suggested, and this was truly the case in the actual election.
Predictions based on per capita consumer complaints are accurate up to a point (about 75%), with certain key shortcomings.
Estimated 2016 Election Results Based on Complaints
With a few key exceptions, the incidence of consumer complaints mirrors the partisan preference of US voters. In fact, it predicts precisely what the polls missed: that the Upper Midwest was more of a battleground than had been expected. On the other hand, complaint data suggested that Georgia and Arizona were going to be more of a battleground than they ultimately turned out to be, and inaccurately suggested that New Mexico would flip to vote republican for the first time since 2000. It is possible that high population growth rates in Georgia and Arizona, along with a correspondingly increased rate of urbanization and ethnic diversification plays a role here.
Actual 2016 Election Results by County
Consumer complaint data did not accurately forecast the outcome of the 2016 election in Florida, and was confounded, perhaps, by the continued high rates of problematic mortgage loans in the state. Florida was the epicenter of the subprime mortgage crisis, and the significantly higher complaint rates in this state may reflect the ongoing aftermath. On the other hand, compared to the complaint-based prediction for the 2012 election, the data did predict that Florida would vote less democratic – as it did.
Estimated 2012 Election Results Based on Complaints
Consumer complaints accurately predicted statewide outcomes in heavily democratic states such as New Jersey and Maryland, as well as in swing states like Florida and Virginia, that voted democratic in 2012. They were unable, however, to distinguish county-level partisan preference in such densely populated states, and in densely populated regions of these states. Likewise, our predictions were more bullish on Nevada and Colorado voting democratic, suggesting wider margins than we ultimately saw. This is possibly due to the fact that lenders such as Synchrony Financial have been sanctioned for discriminatory practices targeting the Latino community. Since Colorado and Nevada have Hispanic populations significantly higher than the national average, actual discriminatory practices may account for these excess complaints.
Actual 2012 Election Results by County
Overall, consumer complaint data fell short in rural areas where democratic candidates perform well, and suburban areas where republicans perform well. This may reflect cultural similarities that transcend partisan affiliation, or it may reflect geographical bias in financial institutions’ business practices.
Predicted Outcome of Congressional Elections 2016
The incidence of consumer financial complaints proved to be predictive not only of the county-level results of the US presidential elections, but also at the level of elections for the House of Representatives, and at a comparable rate of accuracy.
Actual Outcome of Congressional Elections 2016
Once again, a key shortcoming of the complaint-based model on the level of congressional elections is that it tends to overestimate the performance of democratic candidates in densely populated, urbanized regions, and underestimates it in traditionally democratic-voting regions, such as Minnesota,the Upper Mississippi River basin, and South Texas.
Conclusions
People living in the cities and suburbs, and residents of traditionally democratic states complain of unfair business practices on the part of financial institutions at a much higher rate than those who live in rural, red America.
Banks have been known to engage in unfair and discriminatory practices in cities, and with minority populations.
There are three possible explanations: 1) Financial institutions preferentially prey upon people likely to vote for democratic candidates, and indirectly support the election of republicans by treating their constituents better, 2) Liberals whine more than conservatives, who don’t complain nearly as much, or 3) Both.
Methodology
RewardExpert analyzed consumer complaint data from the Consumer Finance Protection Bureau for all complaints categorized as concerning bank accounts, credit cards, and mortgage loans. We aggregated complaints to 3-digit zip code areas and calculated the average complaints per capita rate, then calculated a percentile rank score ranging from 0 to 1 for all areas. We then aggregated data at the county and congressional district level and calculated an average percentile ranking for each geographical unit, and used this score to estimate the fraction of the vote allocated to the republican candidate. We then compared our estimate to the results of the 2016 general election. We repeated this process, filtering out all complaints post-2012 and compared our estimates to the results of the 2012 general election.