Lender Companies in the Consumer Financial Protection Bureau’s Database of Auto Loan Complaints
Project Description and Rationale
Automobiles provide substantial access to opportunities in U.S. cities that have been developed around the automobile. However, there are numerous barriers to automobile ownership, particularly for low-income households, many of whom cannot afford to own and operate vehicles. Compounding these difficulties is the role of the lending sector, a sector that has routinely prioritized profit-making over fair business practices. Additionally, the government’s hands-off attitude has created an unregulated environment, where consumers are easily exploited. In this study, I use data from the Consumer Financial Protection Bureau’s (CFPB) Consumer Complaints Database to examine automobile loan complaints in the US.
In order to encourage more targeted policies for regulators, I want to find out if there are some lender companies that are more problematic than others. Hence, in this study, I am particularly interested in the types of lender companies that are most frequently mentioned in the complaints. I have several research questions
- What types of problems are consumers facing from interacting with these lender companies? Are they the same across all companies?
- Are some lender companies more likely to be complained about than others?
- Is there a strong link between the type of lender companies and the type of complaints?
The motivation for the study rests on research showing a strong correlation between automobile ownership and access to economic opportunities, particularly among low-income households (Blumenberg and Ong, 2001). For example, Blumenberg and Ong (2001) find that automobile ownership is strongly linked to benefits such as higher employment rates and shorter unemployment spells as a result of providing better geographic access to opportunities. However, low-income and non-white households are less likely to own automobiles than higher-income and white households (Brown, 2017) and, therefore, make up a high percentage of transit riders. At the same time, they are most likely to benefit from automobile ownership. Access to an automobile can provide them with the luxury to opt out of using public transit. Transit travel times tend to be long; transit can be inconvenient for individuals and families who must make multiple, sometimes linked, trips in a day; and transit can raise safety concerns, especially for women (Blumenberg and Manville, 2004).
Additionally, this project aims to add value to the complaints database that the CFPB has started building. I add value by adding two new categories that will enable myself and others to study the lender companies and themes of the complaints in greater detail. I also hope to be able to inform readers about the type of issues that consumers may encounter as auto loan owners. The reason is not to intimidate them but to raise consumer awareness in an industry that lacks transparency. This is important because almost every consumer will make a car purchase. It is also one of the most expensive purchases a consumer will make in their lifetime. For this reason, auto loan lenders have the upper hand in the market. Consumers who are not aware of their rights and the trend of practicing unfair conduct in this industry will be most affected.
There are four target audience groups for the project. Firstly, the project targets policymakers who are interested in gaining insights from the complaints database. There are around 22,000+ complaints from a random sample of consumers which can better inform policymaking. The second group is the Consumer Financial Protection Bureau who published the database. I have cleaned the data and added two new columns that might add value to their reports. These columns are the type of lender companies and the theme of complaints. These added columns are also central to this study. The third is consumers who are concerned about getting an auto loan. Hopefully, this project can give them an idea of what to look out for when dealing with different types of lenders. I also hope that it encourages consumers to shop for auto loans to get the best deals if they are able to, instead of choosing what seems to be the most convenient option. Lastly, the project targets lender companies who are looking to improve their services. Since this project is a representation of the consumers’ sentiments towards them, it is my hope that the consumers’ voices are amplified to the point where lenders are able to hear these concerns and address them.
I am using Excel to clean the data and add two new columns to the database:
- Type of lender companies:
In order to create the column “Lender Type”, I filtered the top 75% of complaints and identified the lender type for those companies. It turned out to be 29 companies that made up 75% of the data. I found that there were five categories which were apparent — banks, finance companies, captive finance (financial arm of automotive companies) companies, credit unions, and insurance companies.
- Theme of sub-issues/complaints:
For “Theme of Sub-issues”, I read through some random complaint narratives that consumers submitted in order to identify a theme for each sub-issue. The themes were “affordability”, “unfair” practices and “logistical” issues.
For data exploration and data visualization, I chose to use Tableau because I have a lot of categorical data. Hence, I wanted to be able to use the “Hierarchies” function to view the data in different levels. For example, I am able to view the data in terms of “Issues” and in the more granular level of “Sub-issues”. With Tableau, I can present the type of complaints that consumers are facing. Next, I chose R to test the statistical significance of my results. Namely, I use R to test the relationship between the type of lender companies, the theme and number of complaints. With R, I conduct a t-test, Chi-square test and ANOVA test.
I made several adjustments to the data for this project. The most important distinction is that I added two new columns “Lender Type” and “Theme of Sub-issues”. For the theme of sub-issues, I filtered out complaints talking about “logistical” issues out of the dataset because I wanted to focus on loans being unaffordable or unfair. Additionally, I also chose to focus on data from 2017–2020, filtering out data from 2021. This is because I wanted to have a complete dataset to work with.
Description of work plan
I start with data exploration on Tableau to provide the material to answer the first question, “What types of problems are consumers facing from interacting with these lender companies?” I filter out the sub-issues that I am interested in, and generate a table of the counts of complaints to identify the most common problems. Additionally, before diving into more specific questions about lender companies, I generate a table showing the lender companies who are involved and also show the distribution of complaints across lender companies. This is important to provide some context to the study.
In my project, I then use the t-test to find out if the mean difference of the number of complaints between two types of lender companies is 0. This is where I start to answer my second question, “Are some lender companies more likely to be complained about than others?” If I find the mean difference is 0, I can say that the complaints are equally distributed between the two companies. The questions I will ask are, “Is there a difference between the results of banks vs captive finance companies? Is there a difference between the results of captive finance companies vs finance companies? How about banks vs finance companies?” Although I do not know what sets them apart in detail, I want to learn if there is variation across the different types of companies. The reason only these three companies were chosen for the t-test is that they have significantly more complaints than other companies.
The ANOVA test is conducted with two variables : the type of lender companies and the theme of sub-issues. It determines whether there are any statistically significant differences between the means of number of complaints in banks, finance, captive finance, credit union and insurance companies as a whole. It takes into account variance within and between groups. In contrast to the t-test, it tests the impact of type of lender company on the the number of sub-issues for the entire dataset, instead of just comparing between two companies. It also helps to answer the question, “Are some lender companies more likely to be complained about than others?”
I then use the Chi-Square test to test the hypothesis that there is a relationship between the types of lender companies and the type of complaints. This is to answer the last question, “Is there a strong link between the type of lender companies and the type of complaints?” In other words, it tests whether the two variables are independent or not. This finding will be important for me to determine whether it is worth suggesting to policymakers to create more targeted policies that have varying impacts based on the type of lender.
After establishing whether there is a link or not, I go back to Tableau and create data visualizations to present a more granular study on the theme of complaints across different lender companies. In this section, I will show the breakdown of which lender types are more apparent for sub-issues with the theme of “affordability” or “unfair” practices. This is where I answer the question, “What types of problems are consumers facing from interacting with these lender companies? Are they the same across all companies?
The data visualizations below present an overview of the sub-issues that are being complained about, as well as their counts in the dataset. Table 1 shows the 15 out of 27 sub-issues that I chose to focus on, since they have the themes of affordability and unfair practices. It can be seen that the top 5 sub-issues are “Denied request to lower payments”, “Unable to receive car title or other problem after the loan is paid off”, “Loan balance remaining after the vehicle is repossessed and sold”, “Lender trying to repossess or disable the vehicle” and “Problem with paying off the loan”. It is important for consumers to be aware of these sub-issues so they can avoid falling prey to common practices of lender companies.
I divided the sub-issues into themes “Affordability” and “Unfair” Practices. Graph 2 shows a bar graph of sub-issues concerning the theme “Affordability”. 4/5 of the top 5 sub-issues come from this theme, indicating that most consumers find auto loans unaffordable. Consumers complain about being denied requests to lower payments the most, followed by issues with their loan balance remaining at the end of the loan, repossessions as well as problems managing the loan.
For “Unfair” practices, the top sub-issue as seen in Graph 3 is “Unable to receive car title or other problem after the loan is paid off”. The other sub-issues burden the consumers by charging fees incorrectly or unfairly, offering fraudulent loans, causing problems with additional products purchased with the loan, confusing marketing, changing terms of the loan unfairly, pressuring consumers and misusing their information. It is easy to imagine how burdensome these issues might be, especially for consumers of lower-income who do not have the resources to simply wait for these issues to be resolved.
Now that there is a clearer picture of what consumers are complaining about, I move on to identifying the type of companies in the dataset. Table 4 shows the top 75% of companies in the dataset and the lender types that the 29 companies fall under, based on my research. The key on the right hand side shows that the darker the colour of the bar, the greater the count of complaints about that company. Some of the companies with the darkest colors include finance company “Santander Consumer USA Holdings Inc”, banks “Ally Financial Inc”, “Wells Fargo & Company” and “Capital One Financial Corporation”.
Graph 5 below shows the distribution of complaints in regards to the type of lender companies. Most of the complaints are about finance companies, followed by banks, captive finance companies, and a significant drop in the number of complaints about insurance companies and credit unions.
In order to provide context to the distribution of complaints, Graph 5 shows a bar graph of auto loan lenders according to market share. For the purposes of making the comparison easier, I will include Buy Here Pay Here (BHPH) companies in the figure as finance companies, making the percentage of finance companies an average of 19%. Graph 6 shows that banks typically dominate the auto loan industry with around 32% market share, followed by captive finance companies at 29%. Credit unions are third at around 21% while finance companies make up 19% of market share as mentioned above. If the number of complaints was proportional to the frequency of interacting with these companies as referenced by their market share, we should expect similar percentages of complaints from the named institutions.
I then move to R to run some tests on the data. Although the data visualizations show that there are clearly some lender types more represented than others, it is useful to conduct a statistical test that takes into account variance across different sub-issues. I start with the t-test to test whether the mean difference of the number of complaints between two types of lender companies contested against each other is 0. Since it is clear that banks, finance and captive companies dominate the dataset, I only perform t-tests with pairings consisting of these three groups. I answer three questions with reference to Table 7 :
- Is there a difference between the number of complaints about banks vs captive companies?
The table shows that p-value is smaller than alpha. I reject the null hypothesis and accept that it is probable there is a positive difference between number of complaints in banks and captive finance companies, based on the positive t-statistic. I can draw the conclusion that banks are being complained about more than captive finance companies, with a difference of a value which lies within the confidence interval, ranging from 26 to 192. This is surprising as based on Graph 6, banks and captive finance companies have similar market share. Overall, the t-test shows that banks are more commonly complained about, suggesting that banks are more problematic than captive finance companies.
- Is there a difference between the number of complaints about banks vs finance companies?
The table shows that the p-value is greater than alpha. As a result, I fail to reject the null hypothesis and accept that it is probable there is no difference in number of complaints in banks and finance companies. From this, it can be derived that banks and finance companies are equally commonly complained about. This is surprising as based on Graph 6, finance companies have a significantly smaller market share than banks. This potentially suggests that finance companies in the dataset are overrepresented in the data, despite having fewer interactions with customers. Hence, it is worth investigating finance companies in greater detail.
- Is there a difference between results in the number of complaints about finance vs captive finance companies?
The table shows that the p-value is smaller than alpha, leading me to reject the null hypothesis that there is no difference between the mean of complaints in finance and captive finance companies. The positive values of confidence interval endpoints and t-statistic suggest that there are more complaints from finance companies than captive finance companies. This is surprising as based on Graph 6, captive finance companies have significantly larger market share than finance companies, yet they are receiving less complaints than finance companies. From the analysis in number 2. as well, it can be deduced that finance companies have fewer interactions with customers, but are far more likely to be complained about.
With the ANOVA test, I test if there is a correlation between the types of lender companies and the number of complaints, having taking into account variance within and between groups. This is different from the t-test as I am comparing across all five lender groups, instead of just between two groups. Based on R output as shown in Table 9, I find that p-value is smaller than alpha, prompting me to reject the null hypothesis that there is no correlation. The R output also shows that there is a significance code of 0.001, indicating a strong correlation. This is not surprising given how uneven the graph of the distribution of complaints is in Graph 5.
Next, I conduct a Chi-Square test in order to test the hypothesis that the theme of complaints is independent of the type of lender companies that are being complained about. In this section, I answer the question, “Is there a strong link between the type of lender companies and the type of complaints?” In order to do this, I construct a 2 by 5 Chi square table in R as there are 5 lender types and 2 themes of sub-issues in the data. What I find is that I get a large Chi-Squared value of 117.61 and a p-value smaller than alpha. This shows that there is evidence to reject the Null Hypothesis. Hence, there seems to be a strong correlation between the types of lender companies and the theme of complaints. In other words, the theme of complaints is not independent of the types of lender companies.
Since I have established that there is a clear correlation between the type of lender companies and the number of sub-issues as well as the theme of complaints, I go back to Tableau to zoom in on this occurrence. Here, I hope to answer the “how”, that is, how do the theme of complaints differ across different companies? As seen in Graph 10, it is surprising to find that there is quite a stark difference between the lender companies. For the theme of affordability, almost half of the complaints in the US are about finance companies, followed by banks.
In terms of unfairness, the subject of most complaints are banks in the US, followed by finance companies and captive finance companies. It is clear that banks and finance companies are the most problematic. What is interesting is how they are problematic in slightly different ways, with one causing more issues with affordability and the other with unfair practices.
Here are the research questions I aimed to answer in my study :
- Are some companies more likely to be complained about than others?
- What types of problems are consumers facing from interacting with these companies? Are they the same across all companies?
- Is there a strong link between the type of companies and the type of complaints?
In conclusion, it is clear that the both the number and type of complaints do differ across types of lender companies in the US. Based on the t-test, banks and finance companies are most commonly complained about, followed by captive finance companies. The t-test shows an interesting result, that is finance companies are more likely to be complained about despite having a smaller market share. The implication for this result is that the companies in the dataset should be investigated further to identify the reasons why consumers feel compelled to complain about them. The finance company which has the highest number of complaints is “Santander Consumer USA”. The result could suggest that it is not all finance companies, but some large market share holders that are causing more problems than others. The ANOVA test also shows that there is a strong correlation between the type of lender companies and the number of complaints. In terms of the type of complaints, the Chi-Square shows that I can accept the hypothesis that the type of complaints is dependent upon the type of lender company. In particular, the results seem to suggest that banks are more likely to conduct more “unfair” practices while finance companies are more likely to raise issues of “affordability” towards consumers.
Moving forward, there is a lot that can be studied from the dataset in order to better inform regulation of the auto loan industry. Since social mobility is dependent on mobility in a literal sense, more has to be conducted to ensure that access to auto loans is as equitable as possible. I would suggest for there to be a deeper investigation into the companies that are most complained about in terms of their regular conduct, especially banks and finance companies. Instead of simply implementing blanket policies such as caps on interest rates and disclosure laws which have been found to have loopholes, more targeted policies must be implemented in order to protect all consumers.
Blumenberg, E., and Manville, M. (2004). Beyond the Spatial Mismatch: Welfare Recipients and Transportation Policy. SAGE Journal of Planning Literature. https://journals.sagepub.com/doi/10.1177/0885412204269103
Blumenberg, E., and P. Ong. (2001). Cars, buses, and jobs: Welfare participants and employment access in Los Angeles. Journal of the Transportation Research Board 1756:22–31.
Brown, A.E. (2017). Car-less or car-free? Socioeconomic and mobility differences among zerocar households. Transport Policy, 60(Supplement C), 152–159. https://doi.org/10.1016/j.tranpol.2017.09.016
Consumer Financial Protection Bureau. (2021). Vehicle Loan or Lease. Consumer Financial Protection Bureau. https://www.consumerfinance.gov/data-research/consumer-complaints/
Zabritski, M. (2020) AUTOMOTIVE INDUSTRY INSIGHTS Q2 2020. Experian. Retrieved March 6, 2021, from