CFPB’s Complaint Data Disclosure:

(1)

MSc Marketing Intelligence

Thesis

CFPB’s Complaint Data Disclosure:

Consumer, Firm, and Stock Market behavior

University of Groningen

Faculty of Economics and Business

Department of Global Economics and Management

PO Box 800, 9700 AV Groningen (NL)

June 2018

by Thijmen van der Marel

s2207443

Course: Master’s Thesis MI

Course Code: EBM867B20

Supervisor: Bhattacharya, A.

(2)

1

Abstract

(3)

2

Introduction

The last financial crisis -commonly referred to as “the worst financial crisis since the Great Depression”- severely damaged economies and financial markets around the world, revealing major weaknesses in their financial regulatory systems. In January 2009, the Obama administration announced a plan to create new financial regulation. One year later, 21 July 2010, former U.S. president Obama signed the Dodd Frank Wall Street and Consumer Protection Act into law (DFA or Act thereafter). The Act, a legislative response to the 2007 -2009 American financial crisis, brings a new regulatory framework fit for the twenty-first century (Skeel, 2010). The passage of the Act marks a significant milestone in financial regulation. The Acts primary goal is “to promote the financial stability of the United States by improving accountability and transparency in the financial system, to end ‘too big to fail’, to protect the American taxpayer by ending bailouts, to protect consumers from abusive financial service practices, and for other purposes”.1

The DFA led to the formation of several new government agencies meant to oversee a number of components of the Act. Among these agencies is the Consumer Financial Protection Bureau (CFPB). Its goal, according to the official CFPB website is “to make sure banks, lenders, and other financial companies treat you fairly”2. The duties of this agency is to prevent predatory mortgage lending and to support consumers in their understanding of the terms associated with mortgages, before signing any paperwork. Other tasks involves preventing mortgage brokers from earning high commissions for closing loans with high interest rates or steering potential borrowers to a loan resulting in the highest payment for the originator. Besides, the law requires the CFPB to establish a single, toll-free telephone number, a website, and a database to facilitate centralized collection, monitoring, and response to consumer complaints regarding consumer financial products or services.3 An initial part of this database became publicly available when the CFPB published its database on their website on June 19, 2011. Consumers are confronting this type of data for the first time, leaving financial institutions unaware of what the consequences can be. Forbes (2008) describes customer complaints as a measure of dissatisfaction with the product. For financial firms,

1_{The Dodd Frank Wall Street Reform and Consumer Protection Act (Enrolled Final version – HR 4173).} 2_{Website - https://www.consumerfinance.gov}

(5)

4

whose products are a service, this would translate into dissatisfaction with their service quality. Service quality in turn is crucial to customer satisfaction, customer retention, and profitability (Andreassen, 1999). Besides, complaints and their resolution affect WOM which in turn affects customer loyalty. The corporate image of companies is vulnerable to all these factors (Andreassen, 1999). Hence, complaints and their resolution are a major topic in marketing and have been discussed/investigated in multiple literature. Disclosure of complaint data may also affects the stock market outcome of firms. Casado-Díaz, Mas-Ruiz, and Sellers-Rubio (2009), in a Spain-based study on stock market reactions to third-party complaints, found that the release of an annual report on Complaints Service negatively affects market’s assessment of future cash flow. An understanding of the consequences of a publicly available database on complaints is desirable for firms as it may affect their reputation, brand, and hence, stock market evaluation. Open access to certain financial and customer data enables customers to access previously undisclosed information, supporting them in making smarter decisions. Furthermore, it gives firms an opportunity for being more accountable with regard to their decision taking while they are also encouraged to adopt fairer practices. Literature focusing specifically on consumer complaints in the banking industry and the consequences for consumer behavior is only limited. The aim of this research is therefore to explore the effects of the public disclosure of CFPB complaint data. The research question is hence as follows:

“ How does open access to financial firms’ complaint data affect consumer complaint behavior, firm behavior, and long- and short-term stock market performance? ”

(6)

5

will provide an answer to the research question. This thesis concludes with some managerial implications, research limitation and recommendation for further research.

Literature review

This chapter contains an elaborate discussion of current literature on the topic of this study. First, an overview is presented of academic literature concerned with consumer- and firm-behavior. More specifically, I will analyze how both financial firms and its consumers might react to disclosure of the Act and the associated consequences for the future number of complaints and their resolution. Then, I present an analysis of current literature concerned with stock market’s reaction to events comparable to the DFA. Throughout the analysis several hypotheses are developed.

Consumer complaint behavior

(7)

6

subsequent complaints received. There is however reason to believe that the opposite will occur. It is not hard to imagine that, when banks’ complaint data comes publicly available, consumer sense a lack of ability in these banks’ future efforts to deliver proper service. Hence, consumers might lose trust in their financial institution and its image could be damaged. In both cases, the attitude in the consumers’ minds towards the banks seems to get more negative. Lee, Park, and Han (2008) confirm this in their study where they find that an increase in negative reviews increases consumers’ negative attitude towards the firm. A negative review is comparable to a complaint, as they are both written notions of discontent addressed to a certain firm. Lyons (1996), in her study on underlying factors of complaining behavior finds evidence for an increase of complaints in case of dissatisfaction among consumers. Considering that a negative attitude among consumers is similar or even caused by a certain degree of dissatisfaction, one can assume an increase of subsequent complaints.

(8)

7

benefits of complaining. With the introduction of the CFPB, the process of filing a complaint is easier now that there is a centralized place where consumers are able to file them. This means a reduction of the costs to complaint. The fact that a federal agency handles and monitors the complaints makes it likely that the probability of success increases. This implies that the marginal benefits for consumers to complain increases. After the first disclosure of complaints data, one can expect an increasing number of consumers that take note of the new complaint platform. Consumers, noticing the decrease in costs and increased benefits, are more motivated to file a complaint. Hence, an increase in the subsequent number of complaints post information disclosure may be expected.

Lastly, another one of Srinivasan’s (1990) theoretical streams of consumer information search explains why we may expect an increase in subsequent number of complaints. This involves the internal motivation of a consumer to search for information. Olshavsky and Wymer (1995) define motivation in the context of information search as “The strength of the desire for information about a good is assumed to be directly related to the importance of the good to the consumer”. In the context of this study, ‘the good’ translates into the resolution of a complaint, while ‘the desire for information’ translates into the desire to complain. When consumers experience grievance regarding a service or product related good of their financial institution, the motivation to complain increases. The desire (motivation) to expend effort on a task (complain) comes from a mechanism that governs the movement from one state (i.e. problem with financial institution) to a desired end state (i.e. problem resolution) (Bettman, 1979; Simon, 1967). Grievance to seek redressal should motivate consumer to complain. Moorman (1996), in her research on consumers’ reaction to the introduction of the Nutrition Labeling and Education Act of 1990 requiring food manufacturers to provide nutrition information about their products, found evidence that consumer acquisition and comprehension of nutritional information increased post disclosure of the act. Assuming the same reaction occurs after disclosure of the DFA would mean that consumers would be more interested or motivated into filing a complaint. All of the combined theory above suggest that it is likely that more consumer will file a complaint in the nearby future. Hence, the first hypothesis:

(9)

8

Customer Resolution by Firms

(10)

9

work of Moorman, Ferraro and Huber (2012) on firm responses to the Nutrition Labeling and Education Act. As mentioned before, this act required food manufacturers to provide nutrition information about their products. Moorman, Ferraro and Huber found that the nutritional content of firms’ food items increased post disclosure of the act. Another study, concerned with the effect of information on product quality, found evidence for improved restaurant hygiene after hygiene-quality grade cards were displayed in restaurant windows (Jin & Leslie, 2003). What both studies have in common is that they find significant proof that public disclosure of information produces the desired alignment. In the context of this study, it would mean that the public disclosure of complaint data is giving banks a motivation to improve the quality of its complaint mechanism. If banks do not improve their service quality, it is likely that the public discovers this lack of improvement when subsequent disclosure follows, including the possible negative consequences. It may hence be assumed that banks improve its complaint handling mechanism post information disclosure.

Finally, bank could see their complaints as the route to greater contact between the firm and its customers as envisioned by Plymire (1991). Plymire, in his paper, argues that firms can create new opportunities and improved business processes when they view their complaints as a form of customer feedback. If banks take on this vision, taking on customer complaints as a form of feedback, it could well improve its service quality and other business processes. This may eventually result in an improved complaint handling mechanism and therefore increased customer complaint service.

The arguments above all come down to the assumption of banks developing an improved complaint handling mechanism. This makes it likely that banks are able to handle more customer complaints and come up with a resolution in timelier fashion. Hence the second hypothesis:

H2: Complaint resolution time decreases post information disclosure.

(11)

10

Employees that take on such an approach avoid contact with dissatisfied customers and try to ignore customer complaints. This behavior negatively affects customers’ “complaint satisfaction” which means to what extend the complainant perceives the complaint handling service to meet up to its expectations or not (McCollough, Berry, & Yadav, 2000). The distance between the banks and its customers will increase as a result. Overall customer satisfaction merely affects the customers complaint satisfaction level. As a result, the attitude of customers towards their banks gets worse. Banks could even lose its customers if they experience inadequate complaint handling, even though they are labeled as highly satisfied prior to the complaint (Andreassen, 1999). Customers who are faced with defensive replies from their banks become increasingly belligerent and are more likely to react in aggressive manner (Bach & Kim, 2012). Furthermore, employees might eventually act counterproductive with regard to the firm’ formal goals and objectives (Argyris, 1990). The combination of bad complaint handling and increasingly belligerent customers give reason to believe that the number of disputed cases will rise over time. Besides, the impact of the information disclosure can also result in a decrease of customer attitude towards the bank. Similar to the theory preceding the first hypothesis, disclosure of complaint data might give customers the idea that their bank is not able to deliver on its premises or customer expectations. If, as a result, customer satisfactions towards the banks deteriorates, the changes of a customer disputing a case increases. The third hypothesis is therefore:

H3: Consumers dispute more cases post information disclosure.

Data disclosure and stock market reactions

The disclosure of complaint data is likely to affect those firms whose complaints and related issues are now publicly available. The nature of publicly available data compels firms to react on it as it is a potential source of threat. Besides, investors will strictly follow what is going-on around banks and will take hold of the complaint data disclosure. The reactions of both banks and investor represent itself in the stock prices of these banks. A common used tool to measure the relationship between a certain event and the reaction of the stock market is the so called ‘event study’. The study by Agrawal and Kamakura (1995), about the economic worth of celebrity endorsers, is an example of a highly cited article in the field of marketing.

(12)

11

turned out to be in a negative position post the information disclosure needed to change. Firms cannot ignore negative reports that are publicly available because of the external pressure placed on these firms (Konar & Cohen, 1997). Similarly, the firms included in this study cannot ignore the publication of complaint data. Banks will have to take corrective measures when they are not satisfied with the numbers they are confronted with. Think of corrective measures like upgrading the complaint system, hiring more customer care personnel, or setting up complaint redressal centers. The costs that come along with these measures have a negative impact on the firm’s financial performance. Firm’s stock prices will likely incur this when investors take note of these measures. Besides the negative indirect effect the stock market might experience due to corrective measures, there is also an explanation for an negative direct effect of the data disclosure. This explanation follows from a study by Hamilton (1995) who measured the stock market reaction to the public disclosure of firm pollution figures. The firms in his sample experienced negative, significant abnormal returns upon the first public release of these pollution figures. Similar reactions might occur after the disclosure of complaint data of financial firms. Stockholders could lose faith in the performance of banks when finding out the number of complaints banks receive. This will then result in a decline of the stock prices. The assumed, negative, direct and indirect effect of complaint information disclosure leads to the following hypothesis:

H4: Stock prices of firms will drop post the public disclosure of bank complaint data.

(13)

12

by publicly disclosing information about complaints might therefor be beneficial for the firms stock returns.

Complaints are not necessarily bad news for a firm. Robert Johnston (2001), in his study linking complaint management to profit, argues that complaints can be either good or bad for a company. It all depends on how the companies deal with their complaints. One way of dealing with your complaints is to see them as an opportunity. He mentioned that complaints can be received as an opportunity for firms to build a stronger relationship with their customer. When the complaint handling process is smooth, and a resolution is found, the customer likely feels increasingly satisfied with the company and becomes more loyal. In this scenario, firms are able to generate more sales from what started as a customer complaint. Customer complaints could also benefit the firm when they view complaints as useful feedback. Pyon, Woo, and Park (2011) found that firms in the financial service industry can improve service by business process management using customer complaints. These are all valuable ways of dealing with the incoming complaints which result in the following hypothesis:

H5: stock prices of firm will increase post the public disclosure of bank complaint data.

It may be assumed that not all banks in the sample pursue the same strategy when it comes to complaint management. According to Bach and Kim (2012), high-performing businesses tend to have a proactive approach to resolving consumer complaints, while a low performer tends to have a more defensive approach. A defensive approach seems to backfire because of dissatisfied customers who lose trust in the banks, damaging their attitude towards the bank. When banks take on such an approach, the disclosure of complaints will likely result in negative stock returns. A proactive approach on the other hand is more likely to result in satisfied customers. Banks could generate more sales, improve its relation with the customers and improve consumer trust. Taking on such an approach will likely lead to an increasing stock price. The performance of a bank thus seems to moderate the relationship between the data disclosure and firm stock price. This leads to the sixth and final hypothesis:

(14)

13

Methodology

In this section I will discuss the research methodologies that are used to test the above-derived hypotheses. First, I will briefly elaborate on the data sources, variables, and samples that together form the input for the empirical tests.

Data source and sample.

To measure the first three hypotheses, I gathered secondary data on US bank complaints from the CFPB website4. The nominal data from this database are transformed into interval data which results in multiple variables. These variables tell how many complaints banks received pre and post the data disclosure, how often firms replied to complaints in a timely manner, and how many times customers decided to dispute a case. All this data is computed on a monthly basis. The initial sample contained data for 378 different US based banks. There were no missing data points in this sample. I choose to trim the data so that the sample only includes firms that received at least one complaint before and after the event date. As a results, the final sample include a total of 167 firms.

The data for the event study and the cross-sectional regression analysis comes from Wharton Research Data Services (WRDS). WRDS’ stock market database is one the most complete databases (Christou, 2008). The WRDS database collects its data from independent sources specializing in distinctive historical data. WRDS retrieves most of its stock market data from the Center for Research in Security Prices. The sample size to conduct the event study comprises of stock return data for a total of 48 banks located in the US. The data for the regression analysis to test for a moderator effect are retrieved from Compustat. Compustat is a database providing a broad range of financial, statistical and marketing related information from public global companies around the world. Data gathered from this database is matched to the sample used for the event study.

T-test analyses

Several paired-tests measure whether there are differences between the pre- and post-information disclosure number of complaints that banks receive, how often banks replied in timely manner, and the number consumer disputed a case. A paired t-test compares the means of a population that are divided in two samples. The number of complaints that a particular firm receives

(15)

14

information disclosure is placed in one sample, and the post-disclosure number of complaints received by the same is placed in the second. The same procedure applies to the number of timely responses and disputed cases. With five steps, performed by statistical software program SPSS, you get the results of this test.

Event study

For the second part of this study I use the event study methodology to test the reaction of the stock market on the disclosure of the CFPB bank complaint data. The event study methodology as suggested by MacKinlay (1997), with extensions by Carhart (1997), are the basis for this study. Hence, the estimation window for the daily stock return data is as long as 120 trading days ([t-139, –

t-20]) preceding to the event date ([t0]). The stock market’s response on the disclosure of complaint

data is reflected through the stock market prices around the event. Thus, In order to cover a considerable wide range of possible empirical impacts, I tested several so-called event windows: (i) [t-1,- t+1]; (ii) [t-2,- t+2]; (iii) [t-5,- t+5]; and a single post-event window (iv) [t+1,- t+6]. The estimation

window for monthly stock return data is as long as 18 months, ending 7 months prior to the event date ([t-22, – t-7]). To check if the information disclosure has a more long-term effect on firms stock

prices, post event windows of [t0,- t+1]; and [t1,- t+6] are considered. A more detailed discussion on

the event study methodology, including its historical development, is available through MacKinlay (1997).

To test whether the public disclosure of complaint data has any effect on the US bank’s stock prices, the abnormal returns (AR) of these banks are calculated. In order to retrieve useful AR’s, one needs to compute both actual returns and expected returns. The Carhart four-factor model, which is an extensionof the Fama-French Three Factor Model (Fama & French, 1995) forms the basis from which the abnormal returns are computed. The Carhart four-factor model is a model that computes firm’s expected return on assets. It may be considered a rather up to date model as it incorporates several factors that, according to literature, are essential for computing more realistic expected returns (Jegadeesh & Titman, 2001). The estimation window, running from t0 to

t1,is set to measure the expected returns (see Figure 1 on the next page). This resulted in the

following formula:

(16)

15

where Ri,t ‒ RF,t stands for the excess return of the market portfolio i, on time t, minus the risk-free rate (𝑅𝑅f). βi, M is the Beta of the assets, RM, t is the return of the stock market, βi, SMB is the coefficient for SMBt, which is short for Small market capitalization Minus Big and measures the excess returns of small caps over big caps. βi,HML is the coefficient for HMLi, which is short for High book-to-market ratio Minus Low, measuring the excess returns of value stocks over growth stocks. Finally, βi,MOM is the coefficient for MOMt, which is short for Momentum, and measures the excess returns of the ‘winners’ that went up minus the ‘losers’ that lost value. The inclusion of these factors allows a better detection of effects of the event date. The intercept, 𝛼𝛼I, is a measure of abnormal performance following Carhart methodology. For more details concerning the development of the Carhart Four-Factor Model see Carhart (1997).

The actual returns are derived from the Center for Research in Security Prices (CRSP) stock return data base. They may also be computed by means of the continuously compounded returns, similar to the MacKinley methodology. The formula is as follows:

𝑅𝑅it = ln(Pit/Pit-1), (2) where 𝑅𝑅it is the continuously compounded return of the security for firm i at time t, Pit is the price of the security for firm i at time t and Pit-1 is the same stock price for firm i but at time t – 1. Now that the expected and actual return have been computed they can be used to obtain the abnormal return. The abnormal return is basically the difference between the expected and actual return and it formulated as follows:

AR = Rit ‒ ( Ri,t ‒ RF,t ), (3) where the first part of the formula, Rit, represents the actual return and the last part, Ri,t ‒ RF,t, represent the expected return based on the Carhart Four-Factor Model. When the abnormal returns are aggregated over multiple event windows you can spot how they develop over time. This will result in cumulative abnormal returns (CAR) and the equation looks as follows:

(17)

16

where CARi (𝜏𝜏1,𝜏𝜏2) denotes the cumulative average return (CAR) of stock i in the according event windows (figure 1). Since there are multiple event windows included in this research that need to be compared the cumulative average abnormal returns (CAAR or mean CAR) are computed as well to increase comparability:

(5)

Significance tests

(18)

17

Standardized cross-sectional test

The standardized cross-sectional test (StdCSect Z), introduced by Boehmer et al. (1991) is an extension of the Patell or Standardized Residual Test. The Patell test (Patell, 1976) is a widely used test statistic in event studies. Here, the abnormal returns are standardized by the forecast-error corrected standard deviation before any test statistics are calculated. The Patell test assumes cross-sectional independence which is not valid when the firms of the sample originate from the same industry. The StdCSect Z test compensates for a potential variance increase by including a cross-sectional variance adjustment. This significance test is hence robust to variance induced by the event.

Time-series standard deviation test

The times-series standard deviation- or ‘crude dependence adjustment’ test (Brown & Warner, 1980) differs from the StdCSect Z test in that it uses the entire sample for variance estimation. It thus applies a single variance estimate for the entire portfolio. One downside of this technique is that potential unequal return variances across firm’s stocks is not taken into account. It does however solve the problem of cross-sectional independence, by avoiding the possible problem of cross-sectional correlation of stock returns.

Rank Test

The Corrado’s (1989) Rank Test (Rank Z) is a nonparametric significance test. The first step here is to transform abnormal stock returns into ranks. The abnormal returns for the event period and also the estimation period are incorporated in the ranking process. The midrank is used in case of tied ranks. The ranks are standardized by the number of non-missing values plus 1 to take into account any missing values (Corrado & Zinvey, 1992). The Rank test is originally described for a one-day event window but the effect of ignoring dependence in short event windows is negligible.

Jackknife Z

(19)

18

Cross-Sectional Regression

In an attempt to further explain issues with respect to the studied disclosure of complaint data, multiple regression analysis are employed to examine the final hypothesis. The search for a moderator effect is conducted similarly as the work by Cornwel, Pruitt, and Clark (2005) who studied the relationship between major-league sports’ official sponsorship announcements and the stock prices of sponsoring firms. In order to do this regression, I use the CARs of the firms for the event day [t=-1, t=+1] as dependent variable. Besides, the post event day [t+1, t+6], and post event months [t0, t+1], and [t+1,t+6] serve as additional dependent variables. Following the sixth hypothesis, the independent variable applied in the model is ‘earnings per share’ (EPS), which serves as a proxy for firm performance.

Results

In this section I will present the output of the different statistical analyses performed to test the hypotheses. The first subsection gives an general overview of the data in the form descriptive statistics. Next, to provide an answer to the first three hypotheses, I will present the outcome of multiple T-tests. Clarification on the fourth and fifth hypothesis comes from the results of the event study. The regression test output to test for an moderator effect finalizes the results, after which an overview of the hypotheses concludes this section.

Descriptive statistics

(20)

19

(21)

20

potential for finding significant results.

Table 1. Descriptive statistics

# Complaints Observations Mean St. Deviation Min Max

(22)

21

T-tests results

In order to answer the first three hypotheses I performed multiple t-tests. In these tests I check whether there are significant difference in the mean of the of the pre- and post-information disclosure volume of complaints, timely responses and disputed cases. First, I test whether there are significant differences over a total of 6 different time windows. In table 2 you can find the results. The first time window, pre_Di_M1/post_Di_M1, shows whether the first month post information disclosure displays significantly different means compared to the average of complaints (Co), timely responses (TR), and disputed cases (DC) of 1 month pre information disclosure. The second variable (pre_Di_M2/post_Di_M2) displays the same but than for the mean of a total of two months post information disclosure and the equivalent of two month pre-information disclosure. This continuous upon the total of 6 months of complaints, TR’s, and CD’s pre- and post-disclosure where.

(23)

22

In sum, when you consider a total of six months post ID, banks did indeed receive significantly more consumer complaints then before the ID. Especially in the first three months there are clearly more consumers that send their bank a complaint. Although the effect of information disclosure seems too fade over time, the first hypothesis; ‘the number of complaints increases post information disclosure’ can be confirmed.

(24)

23

Table 2. Paired-samples T Test - Months combined

Month 1 Complaints Timely Responses

Consumer Disputed

Pre Disclosure (mean) 24,93 23,71 5,81

Post Disclosure (mean) 28,73 27,46 7,49

Paired T test -3,136 -2,972 -3,363

P-Value ,002 ,003 ,001

Month 1-2

Paired T test -3,058 -3,378 -3,208

P-Value ,003 ,001 ,002

Month 1-3

Paired T test -2,876 -3,212 -3129.00

P-Value ,005 ,002 ,002

Month 1-4

Paired T test -2,902 -3,219 -3,268

P-Value ,004 ,002 ,001

Month 1-5

Paired T test -2,646 -2,933 -3,085

P-Value ,009 ,004 ,002

Month 1-6

Paired T test -2,394 -2,716 -2,455

P-Value ,018 ,007 ,015

(25)

24

Table 3. Paired-samples T Test - Months individually

Month 1 Complaints Timely Responses

Paired T test -3,136 -2,972 -3,363

P-Value ,002 ,003 ,001

Month 2

Paired T test -2,616 -2,950 -2,807

P-Value ,010 ,004 ,006

Month 3

Paired T test -1,980 -2,242 -2,195

P-Value ,049 ,026 ,030

Month 4

Paired T test -1,470 -1,901 -1,144

P-Value ,144 ,059 ,254

Month 5

Paired T test ,693 ,181 1,460

P-Value ,490 ,856 ,146

Month 6

Paired T test ,560 -,045 ,964

P-Value ,576 ,964 ,067

Note: P-value set at 5%

Event study results

(26)

25

total of 30 firms (62,5%) have positive CARs against 18 firms (37,5%) having a negative CAR. More banks have higher stock prices around the data disclosure compared to the expected stock price under normal circumstances (p>,01). 13,33 % of all the firms turned out to have significantly positive CAR while only 5,55 % display significantly negative CARs. Overall, the firms in the sample had an average CAR of 0,65 % over the 1-day event window as shown by the MCAR. These positively skewed CARs are only amplified over a 2-day event window. A total of 36 firms (75%) have positive CARs against 12 negative (25%). Again, more firms seem to benefit from the data disclosure compared to firms that do not benefit (p< ,001). Although most firms managed to have a positive CAR, only 8,33 % of the firms in the sample have a significant positive CARs against 4,12 % of the firms that got a negative CAR. Overall, over a 2-day event window, the firms in the sample had a mean CAR of 1,02%. In the 5-day event window there are again more firms with positive CARs (33), compared to firms with a negative CAR (15). The difference is significant (P< ,05) From the 2-day event window to the 5-day event window there is thus a slight decrease of firms with a positive CAR. 4,12 % of the firms got a significantly positive CAR while 2,1% of the firms display a significant negative CAR. Taking all firms together over a 5-day event window results in a positive mean CAR of 1,07%. Besides the event windows surrounding the event I also checked the ARs of the first week after the event date ([+1, +6]). More firms (26) ended up having a positive CAR compared to firms with a negative CAR (22). The difference is not significant in this case. There are 4,12% of these firms with a significant positive CAR while not a single firm got a significant negative CAR. The mean CAR over this period is positive (0,34%). The conclusion from this data is that the disclosure of complaint data causes an initial overall significant positive effect on firms stock returns. Hypothesis 5 which states that the stock prices will go up after the disclosure of complaint data is thereby confirmed.

Table 4. Event study results - Day level

Days N Mean CAR Positive:Negative

P-value Std. T Test

Firms sig.

CAR Sig. Negative CAR

-1, +1 48 ,65% 30:18 p < ,01 8,33% 2,1%

-2, +2 48 1,02% 36:12 p < ,001 8,33% 4,12%

-5, +5 48 1,07% 33:15 p < 0,05 4,12% 2,1%

(27)

26

Besides testing whether there would be an initial reaction to the disclosure of complaints, it is relevant to test the effect of information disclosure over the long term. In table 5 you can find the results of this second event study. First I check the abnormal returns over a period of 1 month post information disclosure. From the event day until 1 month later, a total of 30 firms note positive CARs against 17 negative. This positive difference is significant (p<,10). Only one firm has a significant positive CAR and two firms a significant negative CAR. On average, the firms in the sample seem to benefit from the first month post information disclosure as the mean CAR is positive (1,25%). A drastic change, however, occurs when a more long-term view is tested. You find more firms with a negative CAR (38 )over a period of 6 months post information disclosure. For the firms time there are significantly more firms with a negative CAR instead of a positive (p<,10). A total of 36,17% firms turn out to have a significant negative CAR over the 6 months post information disclosure against only 14,89% with a significant positive CAR. All firms combined have a lower stock return than expected over the 6 months post information disclosure period with a mean CAR of -5,05%. On the shorter term the stock prices go up post information disclosure but in the long term, the stock prices go down. The fourth hypothesis stating that the stock prices go down post information disclosure is therefore partially confirmed since there is a long term effect.

Cross-sectional regressions

To test the for a moderator effect between firms’ CARs and the data disclosure, I performed a regression analysis. Firm performance is in this case the moderator. Firms’ earnings per share is used as independent variable while a multitude of CARs over different event windows are the dependent variables. The results of this analysis are in table 6. Unfortunately, the regression results indicate that firm performance does not act as a moderator for the relationship between the disclosure of complaint data and firms’ stock return. The regression turned out to be highly insignificant. The fifth hypothesis, stating that higher performing banks benefit more from the data disclosure compared to lower performing banks, does not hold.

Table 5. Event study results - Month level

Months N Mean CAR Positive:Negative

P-value Std. T Test

Firms sig.

CAR Sig. Negative CAR

0, +1 47 1,25% 30:17 p < ,100 2,13% 4,26%

(28)

27

Table 6. Regression output - Moderator

Dependent variable EPS EPS EPS EPS

AR_-1,+1_D -0,269 [0,593] AR_+1, +6_D 0,737 [0,685] AR_0,+1_M 1,225 [2,898] AR_+1,+6_M 0,293 [7,293] Obs. (N) 47 47 47 47 R-squared 0,005 0,025 0,004 0 F-value 0,206 1,157 0,179 0,000

Note: Control variables are inlcuded in the regression analysis

* p < ,1 ** p < ,05 *** p < ,01 *** p <0,001

Robustness checks

(29)

28

second hypothesis is found in the fifth and sixth month post information disclosure. Both months appear to have significant higher means for the variable ‘timely response’ compared to an average month pre-information disclosure. Overall, the nonparametric t-test increases robustness of the results since all significant variables are confirmed by these tests .

Multiple parametric and nonparametric significance test are performed to increase the robustness of the event study. In most cases the results from these tests confirm the findings of the initial measurement. The results of the day level event windows [t-1,- t+1]; [t-2,- t+2] and [t-5,- t+5] are

confirmed by the parametric StdCsect test and the nonparametric Jackknife test . The results for event window [t+1,- t+6], similar to the initial test, are non-significant among all extra test included

(table C). The results for the daily event study are hence robust. The results of the extra test for the monthly event study are divided (table D). The results for the 1 month post-event window ([t0,-

t+1]) are only significant for the initial conducted t-test. The two extra parametric and

nonparametric test display insignificant results. The results of the extra test, except for the Rank test, for the long-term post-event window ([t1,- t+6]) confirm the significant negative effect of the

information disclosure. What stands out is that not one of the performed Rank tests display a significant result. A plausible explanation for this is that the Rank test is originally created to the test 1-day event windows while the event windows considered in this study are greater than 1 day.

(30)

29

Discussion and conclusion

The following section provides a general discussion of the results from this study. Besides, I will present the managerial implications that follow from this discussion. Then limitations, further research suggestions, and an overall conclusion will finalize this thesis.

CFPB complaint data

(31)

30

more customers than before become aware (i.e. by means of media attention) of this increase of benefits and decrease in costs. This leads to more customers that are motivated to complain. The second stream of information search theory believes that more consumers complain because of increasing internal motivation to complain (Srinivasan, 1990). This boost of internal motivation to complain derives from increased product involvement after the disclosure of an act (Moorman, 1996). The increase of complaints, motived by both theories of information search, is initiated by the disclosure of complaint data. This initial effect that the disclosure of data has on customers will fade over time. It explains why the first months post information disclosure shows an significant increase in complaints and then returns to numbers similar to pre-information disclosure. These findings contribute to the marketing literature as new back-up is found for two theoretical streams of the consumer information search theory.

The theories explaining the increase in timely responses and disputed cases find back up as well. Similar to a recent study by Nishimura and Okamuro (2018), a business process (in this case complaint handling), shows improvements while under governmental monitoring. Banks could take measures to improve their complaint handling or to make sure that a complaint receives a timely response. Older research where public information disclosure results in improved quality finds backup too (Jin & Leslie, 2003; Moorman, Ferraro, & Huber, 2012). Another argument for the improved timely response comes from Plymire (1991) who argues that organizations can improve business processes if customer complaints are viewed as feedback. Improved service quality and complaint resolution reduces the number of customers who defect due to dissatisfaction (Fornell & Wernerfelt, 1987; Hirschman, 1970).

(32)

31

The banks’ stock market reaction

The results from the event study confirm the contradictory hypotheses regarding the directions of firms’ stock market upon the disclosure of complaint data. The initial effect of this disclosure on firms’ stock price is positive. The positive stock market reaction in the days post the information disclosure is possibly a results from an increased image of transparency (Fishman & Hagerty, 2003). Increased transparency in the banking industry is an opportunity for investors because it contributes to more precise measures of stock market indications. Furthermore, now that there is an external monitor in place, in the form of government and customers, there is a reduction of information asymmetry between firm and public. In general, a reduction of information asymmetry improves the business reputation (Hölmstrom, 1979). Firms’ improved business reputation and increased image of transparency are in this case a direct result from the information disclosure. Investors pick up these signals and it will translate in positive reaction on the stock prices. This might thus explain why there is an initial positive stock market reaction upon the complaint data disclosure.

(33)

32

determinant factor customer purchasing behavior. Hence, especially banks will face a drop in sales when consumer trust is damaged. Consumers lose trust in the first months post information disclosure because of the increasing number of complaints and the sales will drop. The stock price of banks will eventually represent this loss in sales which can explain why the overall stock price is lower than expected over the 6 months post information disclosure. That the number of complaints decrease in the 5th and 6th month post information disclosure is hence a good sign for banks. Customers might pick this up and get the feeling that banks react on the past months by improvements of their services. Consumer trust will restore and the sales and stock prices improve.

Managerial implications

No doubt, the public disclosure of bank complaint data affects both banks and customers. Managers must deal with this government provided data since its disclosure is mandatory. In an ideal situation, firm’s customer complaints decrease while its customer base increases. This would be a sign of adequate complaint handling and overall service. For that reason it is of paramount importance for managers to avoid DOB within the firm. Employees could easily fall for DOB if consumer complaints increase post disclosure and the workload increases. Instead, managers should spread the idea of complaints being a source for potential future profits. It is an opportunity to reduce satisfaction and encourage repeated business (Ngai, Heung, Wong, & Chan, 2007). This requires a firm to view complaints as a form of feedback instead of a problem that needs a rapid resolution. According to Johnston (2001), process improvements and employees are the strongest link between complaint processes and financial performance. Thus, from an operations perspective, it entails that managers develop complaint management procedures and provide training to its staff. Eventually, a firm should strive to create a complaint culture which focuses on process improvement, its employees, and not solely on customer satisfaction. The business improvements created with help of customer complaints will leave other customers with a lower probability of bumping into the same problem. The same goes for complaints that eventually result in a customer disputing the case. When firms get this done it will likely result in higher customer trust and all the benefit that come with it. This way, there is change that the negative long term stock prices will recover and financial performance improves, even though the CFPB publishes complaint data.

Limitation and further research

(34)

33

exact date is known of investors’ stock market changes (Brown & Warner, 1980). This implies that the stock market reactions of a certain event should all be captured in a single moment. There is indeed a single first day when the complaint data was published but this dataset was only limited in size and has grown over time. The reactions on the stock market following this disclosure could hence be limited. Besides, preliminary announcements of the Dodd-frank act already confirmed that such a database will be published which reduces the eventual shock effect on the disclosure date. It is hence likely that not all stock market changes caused by the disclosure of complaint data are wrapped in the event date of this study. Nonetheless, it is still very likely that investors created certain expectations on the event day which are represented in the results of the daily event study. Despite this critique, the event study methodology is extensively used in academic research for both financial and marketing related subject.

Another limitation concerns the limited availability of stock return data and firm performance indicators. If case of a bigger sample size, preferably similar to the sample size of the complaint data, more insightful analysis could be performed. The changes of finding a moderator effect would also increase because of improvements in statistical power. To counter this limitation, the event study yields significant results from which useful conclusion are drawn.

(35)

34

in complaint handling processes. The goal, eventually, is to understand more precisely what factors play a role in the effects caused by the disclosure of CFPB complaint data

Conclusion

(36)

35

Appendix

Table A. Wilcoxon Signed Ranks test - Months combined

Wil.Sig.Rank Test -3,268 -3,481 -4,318

P-Value ,001 ,000 ,000

Month 1-2

Wil.Sig.Rank Test - Z -5,255 -5,767 -5,559

P-Value ,000 ,000 ,000

Month 1-3

Wil.Sig.Rank Test - Z -5,105 -5,649 -5968.00

P-Value ,000 ,000 ,000

Month 1-4

P-Value ,000 ,000 ,000

Month 1-5

P-Value ,000 ,000 ,000

Month 1-6

P-Value ,000 ,000 ,000

(37)

36

Table B. Wilcoxon Signed Ranks test - Months individually

P-Value ,001 ,000 ,000

Month 2

Wil.Sig.Rank Test -4,054 -4,740 -3,478

P-Value ,000 ,000 ,001

Month 3

P-Value ,012 ,001 ,000

Month 4

P-Value ,002 ,000 ,005

Month 5

Wil.Sig.Rank Test - Z -1,317 -1,979 -,170

P-Value ,188 ,048 ,865

Month 6

Wil.Sig.Rank Test - Z -1,359 -2,027 -,336

P-Value ,174 ,043 ,737

(38)

37

Table C. Event study results - Day level

Days N Mean CAR Positive:Negative

StdCsect Z (CDA) t Rank Test Z Jackknife Z -1, +1 48 ,65% 30:18 2,498*** 0,849 0,793 2,425*** -2, +2 48 1,02% 36:12 3,437**** 1,030 1,143 3,233**** -5, +5 48 1,07% 33:15 2,272** 0,729 0,928 2,093** +1, +6 48 0,34% 26:22 0,947 0,317 0,718 1,600* * p < ,1 ** p < ,05 *** p < ,01 **** p < ,001

Table D. Event study results - Month level

Months N Mean CAR Positive:Negative

StdCsect Z (CDA) t Rank Test Z Jackknife Z 0, +1 47 1,25% 30:17 4,215 0,532 0,168 1,016 +1, +6 47 -5,05% 19:28 -1,685** -1,283* -1,167 -1,792** * p < ,1 ** p < ,05 *** p < ,01 **** p < ,001

Bibliography

Agrawal, J., & Kamakura, W. (1995). The economic worth of celebrity endorsers: An event study analysis.

The Journal of Marketing, 56-62.

Anderson, E., & Weitz, B. (1989). The use of pledges to buld and sustain commitment in distribution channels. Journal of Marketing Research, 18-34.

Andreassen, T. (1999). What Drives Customer Loyalty With Complaint Resolution. Journal of Service

Research , 324-332.

Argyris, C. (1990). Overcomming organizational defensive-facilitating organization learning. Boston: Allyn & Bacon.

Ba, S. (2001). Establishing online trust though a community responsibility system. Decision Support

(39)

38

Bach, S., & Kim, S. (2012). Online consumer complaint behaviors: The dynamics of service failures, consumers' word of mouth, and organization-consumer relationships. International Journal of

Strategic Communication, 59-76.

Beatty, S., & Smith, S. (1987). External Search Effort: An Investigation Across several Product Categories.

Journal of Consumer Research, 83-95.

Bettman, J. (1979). An information processing theory of consumer choice. Reading, MA: Addison-Wesley, k.

Bloemer, J., Ruyter, K. d., & Peeters, P. (1998). Investigating drivers of bank loyalty: the complex relationship between image, service quality and satisfaction. International Journal of Bank

Marketing, 276-286.

Boehmer, E., Musumeci, J., & Poulsen, A. (1991). Study Methodology under Condition of Event-Induced Variance. Journal of Financial economics, 253-272.

Brown, S., & Warner, J. (1980). Measuring Security Price Performance. Journal of Financial Economics, 205-258.

Brown, S., & Warner, J. (1985). using Daily Stock Returns: The Case Of Event Studies. Journal of Financial

Economics, 3-31.

Carhart, M. (1997). On Persistence of Mutual Fund Performance. The Journal of Finance, 57-82. Carter, L., & Bélanger, F. (2005). The utilization of e-government services: citizen trust, innovation and

acceptance factors. Information Sysytems Journal, 5-25.

Casado-Díaz, A., Mas-Ruiz, F., & Sellers-Rubio, R. (2009). Stock market reactions to third-party complaints. International Journal of Bank marketing, 167-183.

Christou, N. (2008). Enhancing the teaching of statistics: Portfolio theory, an application of statistics in finance. Journal of Statistic Education.

Collins, D., & Dent, W. (1984). A comparison of Alternative Testing Methodologies Used in Capital Market Research. Journal of Accounting Research, 48-84.

Copeland, M. (1917). Relation of Consumer's Buying Habits of Marketing Methods. Harvard Business

Review, 282-289.

Cornwell, T., Pruitt, S., & Clark, J. (2005). The relationship between major-league sports' official

sponsorship announcements and the stock prices of sponsoring firms. Journal of the Academy of

Marketing Science, 401-412.

(40)

39

Corrado, C., & Zinvey, T. (1992). The specification and power of the sign test in event study hypothesis test using daily stock returns. Journal of Financial and Quantitative analysis, 465-478.

Dean, D. (2004). Consumer reaction to negative publicity: Effects of corporate reputation, response, and responsibility for a crisis event. . The Journal of Business Communication, 192-211.

Fama, E., & French, K. (1995). Size and Book to Market Factors in Earnings and Returns. Journal of

finance, 131-155.

Fishman, M., & Hagerty, K. (2003). Mandatory versus voluntary disclosure in markets with informed and uninformed customers. Journal of Law, Economics, and organization, 45-63.

Flavián, C., Guinalíu, M., & Torres, E. (2005). The influence of corporate image on consumer trust.

Internet Research, 447-470.

Forbes, S. (2008). The effect of service quality and expectations on customer complaints. The Journal of

Industrial Economics, 190-213.

Fornell, C., & Wernerfelt, B. (1987). Defensive Marketing Strategy by Constomer Complaint Management: A theoretical Analysi. Journal of Marketing Research, 337-346.

Fornell, C., & Wernerfelt, B. (1988). A Model for Customer Complaint Management. Marketing Science, 271-286.

Giaccotto, C., & Sfiridis, J. (1996). Hypothesis Testing in Event Studies: The Case of Variance Changes.

Journal of Economics and Business, 349-370.

Halstead, D., & Page, T. (1992). The effects of satisfaction and complaining behavior on consumer repurchase intentions. Journal of Consumer Satisfaction, Dissatisfaction and Complaining

Behavior, 1-11.

Hamilton, J. (1995). Pollution as news: media and stock market reactions to the toxics release inventory data. journal of environmental economics and mangement, 98-113.

Harrington, S., & Shrider, D. (2007). All events induce varianceL Analyzing abnormal returns when effects vary across firms. journal of Fiancial And Quantitative Analysis, 229-256.

Hirschman, A. (1970). Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Cambridge: Harvard University Press.

Hölmstrom, B. (1979). Moral hazard and observability. The Bell journal of economics, 74-91.

Homburg, C., & Fürst, A. (2005). How organizational complaint handling drives customer loyalty: An analysis of the mechanistic and the organic approach. Journal of Marketing, 95-114.

(41)

40

Jegadeesh, N., & Titman, S. (2001). Profitability of Momentum Strategies: An Evaluation of Alternative Explanations. The Journal of Finance, 699-720.

Jin, G., & Leslie, P. (2003). The effect of information on product quality: Evidence from restaurant hygiene grade cards. The Quarterly Journal of Economics, 409-451.

Johnston, R. (2001). Linking complaint management to profit. International Journal of Service industry

Management, 60-69.

Katz, J., Higgins, E., Dickson, M., & Eckman, M. (2009). The Impact of External Monitoring and Public Reporting on Business Performance in a Global Manufacturing Industry. Business and Society , 489-510.

Kolari, J., & Pynnönen, S. (2010). Event study testing with cross-sectional correlation of abnormal returns. The Review of financial studies, 3996-4025.

Konar, S., & Cohen, M. (1997). Information as regulation: The effect of community rigth to know laws on toxic emissions. Journal of environmental economics and management, 109-124.

Lee, J., Park, D., & Han, I. (2008). The effect of negative online consumer reviews on product attitude: An information processing view. Electronig Commerce Research and Applications, 341-352.

Liu, X., & Wei, K. (2003). An empirical study of product differences in consumers' e-commerse adoption behavior. Electronic Commerce Research and Applications, 229-239.

Locke, R., Qin, F., & Brause, A. (2007). Does Monitoring Improve Labor Standards? Lesson from Nike.

Industrial and Labor Relations Review, 3-31.

Lyons, J. (1996). Getting customers to complain: a study of restaurant patrons. Australian Journal for

Hospitality Management, 37-50.

Mackinlay, A. (1997). Event studies ineconomics and finance. Journal of Economic Literature, 13-39. McCollough, M., Berry, L., & Yadav, M. (2000). An empirical investigation of customer satisfaction after

service failure and recovery. Journal of Service Research, 121-137.

Mizerski, R. (1982). An attribution explanation of the disproportionate influence of unfavorable information. Journal of Consumer Research, 301-310.

Moorman, C. (1996). A quasi experiment to assess the consumer and informational determinants of nutrition information processing activities: The case of the nutrition labeling and education act.

Journal of Public Policy & Marketing, 28-44.

(42)

41

Ngai, E., Heung, V., Wong, Y., & Chan, F. (2007). Consumer complaint behaviour of Asians and non-Asians about hotel services: An emperical analysis. European Journal of Marketing, 1375-1391.

Okamuro, H., & Nishimura, J. (2018). Governance and Performance of Publicly Funded R&D Consortia.

Innovation in the Asia Pacific, 147-159.

Olshavsky, R., & Wymer, W. (1995). The desire for new information from external sources. Proceedings

of the Society for Consumer Psychology (pp. 17-27). Bloomington: Printmaster.

Patell, J. (1976). Corporate forecasts of earnings per share and stock price behavior: Empirical test.

Journal of accounting research, 246-276.

Plymire, J. (1991). Complaints as opportunities. Journal of Consumer Marketing, 39-43.

Porter, K. (2012). The Complaint Conundrum: Thoughts on the CFPB's Complaint Mechanism. Brooklyn

Journal of Corporate, Financial & Commercial Law.

Pyon, C., Woo, J., & Park, S. (2011). Service improvements by business process management using customer complaints in financial service industry. Expert systems with applications, 3267-3279. Ratnasingham, P. (1998). The importance of trust in electronic commerce. Internet Research: Electronig

Netrworking Applications and policy, 313-321.

Schmidt, J., & Spreng, R. (1996). A proposed model of external consumer information search. Journal of

the Academy of Marketing Science, 246-256.

Simon, H. (1967). Motivational and emotional controls of cognition. Psychological Review, 29-39. Sirdeshmukh, D., Singh, J., & Sabol, B. (2002). Consumer Trust, Value, and Loyalty in Relational

Exchanges. Journal of Marketing, 15-37.

Skeel, D. A. (2010). The New Financial Deal: Understanding the Dodd-Frank Act and its (Unintended)

Consequences. John Wiley & Sons.

Srinivasan, N. (1990). Pre-purchase external search for information. Review of Marketing, 153-189. Welch, E., Hinnant, C., & Moon, M. (2004). Linking citizen satisfaction with e-government and trust in

government. Journal of public administration research and theory, 371-391.

Zeithaml, V., Berry, L., & Parasuraman, A. (1996). The behavioral consequences of service quality. The

(43)

42

(44)

(45)

(46)

(47)

(48)

(49)

CFPB’s Complaint Data Disclosure:

MSc Marketing Intelligence

Thesis