University of Groningen
Faculty of Economics and Business MSc Finance Thesis
Reputational risks for banks due to litigation post 2008 financial crisis
Tomi Goldenberg
Supervised by Dr. S.S.H. Eriksen
Abstract
This Master’s thesis investigates whether banks originating in the United States of America incur
reputational damage due to litigation announcements, which involve settlements or fines from
governing bodies. To do so, a sample has been taken of 124 of these events, for which both
regression analyses and an event study have been conducted using both the Capital Asset Pricing
Model and the Market Model. The banking index and a general market index have been applied
as market returns. The event study yielded significant negative abnormal returns at the day after
an event, as well as significant positive abnormal spillover effects to the banking index. This may
indicate that the investors view litigation events cause reputational damage to banking companies,
but improve the prospects of the industry as a whole.
Acknowledgements
I would like to express gratitude to my supervisor Steffen Eriksen for his guidance and patience
during the process of writing this thesis. I could not have delivered this thesis in time without his
support, ideas and useful tips. I would also like to thank my family and friends for their support.
Contents
1 Introduction 5
2 Literature review 7
3 Analysis 10
3.1 Data . . . 10
3.2 Methodology . . . 12
3.2.1 Regression analysis . . . 13
3.2.2 Event study . . . 15
4 Results 17 4.1 Regression analysis . . . 17
4.2 Event study . . . 20
5 Discussion 21
6 Conclusion 23
A Non-parametric tests 27
1 Introduction
The banking system is highly dependent on public trust (Algan and Cahuc, 2010). Nevertheless, banks have a controversial history by breaking this fragile trust on several occasions by engaging in fraud and other illegal schemes. Since the financial and economic crisis of 2008, litigators have employed more strict regulations and control which have led to an increase in fines handed out to banks (Grasshoff et al., 2017).
A variety of papers have been written regarding the influence of the announcements of litigation and settlements in the period before the financial crisis, which all indicate that significant reputational losses are incurred by banking institutions while having negative spillover effects on the industry as a whole as well (Perry and De Fontnouvelle, 2005; Cummins et al., 2007; Fiordelisi et al., 2013;
Sturm, 2013). Since the financial crisis of 2008 has lowered the public’s trust in the banking system, it is interesting to research whether investors react differently to announcements of fines or settlements by banks due to non-compliance to regulations than before the crisis (Roth, 2009;
Guiso, 2010).
Due to the timing of these papers, which all originate in the period before the financial crisis of 2008, no research has been conducted on the reputational damage caused by litigation and settlement announcements after the financial crisis. Therefore, the following research question has been derived:
Do announcements of litigation or settlements cause reputational damage in the period after the financial crisis of 2008?
These fines and settlements by banks do not only have a tangible impact on the banks due to the sum they have to pay up to the regulatory body in question, causing operational losses, but might hurt their reputation and trust by the public, customers and investors which is rather intangible.
Following economic theory, this increased perception of risk of a bank increases it’s weighted
average cost of capital due to higher costs of equity and thus lowers its valuation and the market
capacity of the bank (Fama and MacBeth, 1973). As a result of this drop, the returns of banking
securities are lower than that they are expected to be in the absence of this so-called event. To
measure whether these announcements have significant effects on the stock returns, we assessed the
abnormal returns with an event study and ran regressions with the events set as dummy variables.
These abnormal returns are the actual returns subtracted by the estimated returns which are generated by the Capital Asset Pricing Model (CAPM) and market model. The aforementioned effects and the resulting stock price drops are expected to result in negative abnormal returns, since these events could not have been anticipated by investors. This resulted in the following hypotheses which have been tested:
H0: There are no negative abnormal returns after an announcement of a litigation related fine or settlement against banks after the financial crisis of 2008
H1: There are negative abnormal returns after an announcement of a litigation related fine or settlement against banks after the financial crisis of 2008
The null hypothesis is supported by the aforementioned papers, which is why the alternative hypothesis states that for the period after the financial crisis of 2008 litigation events will result in significant negative returns. The outcomes of past studies before the crisis will be compared with our results in an attempt to find explanations for possible differences.
As independent variables, for both the regression analysis and the event study, a banking index (BIX) and a regular stock index (SNP) were used. Since the banking index seems like the natural fit due to the fact that it closely relates to the financial institutions which are examined in this thesis. However, reputational damage by litigation and/or settlements with government agencies could spill over to other banks represented in the banking index. Therefore, the results to tests with the banking index are possibly not accurate and a general market index was used as well to create the estimation models for the regression analysis and the event study. Lastly, the same tests were applied to the banking index with the general market index as independent variable to assess whether there were significant spillover effects to the banking industry of reputational damage caused by a litigation process or settlement against a single bank to the banking index.
Results yielded by this research could have implications for financial literature while also influencing decision making of investors. Along with filling a gap in the literature, we can assess whether the attitude of investors towards reputational damage by litigation has changed after the financial crisis of 2008. Investors could use this information to alter their strategies for when these events occur.
The thesis starts with a literature review on the topic of reputational risks due to litigation,
while touching on literature about banking regulation and spillover effects. This is followed by a
section on the data, describing the collection methodology while laying out the sample statistics and characteristics of the collected data. The methodology section touches on the methodologies employed to test the hypotheses by making a distinction between the regression analysis and the event study. Following the methodology, the results of both the regression analysis and the event study are presented in the results section and analysed and explained in the discussion. Lastly, a summary will be provided in the conclusion section which will lay out the main findings of the thesis.
2 Literature review
Ever since the crisis of 2008, banking regulation has evolved rapidly. In response to the crisis, the Dodd-Frank Wall Street Reform and Consumer Protection Act (the Dodd-Frank act) was signed into law in 2010 as a reaction to the crisis, which combined with rigid law enforcement resulted in around 321 billion US dollars (USD) in fines over the period of 2009 until 2016 (Grasshoff et al., 2017). According to research by Grasshoff et al. (2018), global banks currently have to cope with an average of 200 regulatory revisions every day, which triples the same figure when compared to 2011. Currently, the United States of America experiences a wave of banking deregulation, with acts such as the Financial CHOICE Act which removes many of the regulations imposed by the Dodd-Frank act (Grasshoff et al., 2018).
Non-compliance to the regulations imposed by governments results in litigation. The research by Grasshoff et al. (2018) indicates that in the period 0f 2009 until 2017, around USD 220 billion in fines were handed out to North American banks. This phenomenon results in a multitude of papers written on the topic of reputational risks due to litigation of banks. When considering litigation of banks, we look at the events of regulatory authorities fining banks for illegal activities.
Walter investigated this topic in 2005, defining reputational risk as ”the risk of loss in the value of a firm’s business franchise that extends beyond event-related accounting losses and is reflected in a decline in its share performance metrics ” (Walter, 2008). When defining this as such, we can investigate the reputational risk by looking whether the returns of a bank decrease by more than only the sum of the fine or settlement.
Research by Perry and de Fontnouvelle (2005) measure reputational losses using the same definition
as Walter (2008). To estimate the returns in the event period for the event study, a single-factor market model was employed, while the corporate governance index as described by Gompers (2003) was employed to determine whether the different firms had weak or strong shareholder rights (Gompers et al., 2003).They found that the market values of firms fall over twice as hard when the litigation case involves internal fraud, except for firms with weak shareholder rights where they did not find differences between internal or external fraud. For firms with strong shareholder rights, robust evidence was found that the market capitalisation falls more than one-to-one after internal fraud which means that reputational losses have been incurred (Perry and De Fontnouvelle, 2005).
Another analysis of the reputational losses due to operational events, in this case both internal fraud cases and events concerning ”clients, products and business practices”, was conducted by Gillet et. al (2010). Using an event study methodology, significant reputational damages were found in cases of internal fraud (Gillet et al., 2010). In contrast to other papers on this topic, three distinct event dates are identified: the first press cutting date, the recognition date by the company and the settlement date.
P. Sturm (2013) analysed a sample of German Public Sector Banks from January 2000 until December 2009, where he found significant negative stock price reaction, even after accounting for the nominal loss due to the fines or settlements which means that reputational losses have been incurred (Sturm, 2013). Additionally, multivariate analysis suggested that reputational damages where influenced by firm characteristics rather than the characteristics of the event. This conclusion does not necessarily contradict Perry and de Fontnouvelle (2008) however.
According to Fiordilisi, quoting the Basel Committee on Banking Supervision (2009), reputational
risk can be described as the ”risk arising from negative perception on the part of customers, counter
parties, shareholders, investors, debt-holders, market analysts, other relevant regulators that can
affect a bank’s ability to maintain existing, or establish new, business relationships and continued
access to sources of funding” (Basel, 2009; Fiordelisi et al., 2013). This definition is slightly
different from the definition used by Walter (2008), but it does not result in a different research
construct (Walter, 2008). However, Fiordelisi et al. (2013) adds to the literature on reputational
risks by detecting six determinants driving reputational loss being bank riskiness, profitability,
level of intangibles, capitalisation, size, the entity of the operational loss and the business units
that suffered this loss. Additionally, an analysis is included of the event windows and the influence
of it on the outcome of the research. The event window is the window for which the estimation is compared to the actual returns to detect abnormal returns. An event window is constructed around the event date, which is situated at day zero. These normally are longer than only the event date, since either a slight delay of the market reaction or early rumours come up which influence the returns ahead of the event date.
Besides literature on the reputational damage on firms caused by their own litigation events, articles have been written on the spillover effects of such events on other firms within their industry.
Cummins et al. (2007) conduct a study on the spillover effects in the financial industry caused by operational loss announcements, of which litigation events are a subset. Significant negative spillover effects due to reputational damage in the financial industry have been detected. This research split the financial industry in three distinct sectors; commercial banking, investment banking and insurance and found, using an event study methodology, strong negative intra and inter-sector spillover effects. According to the paper, this could be explained by several reasons such as that operational risk events could indicate that signal that these events also take place at firms which do not announce these events (Cummins et al., 2007). This reasoning does not hold for the litigation effects, since these are always announced by the litigators and thus publicly available. However, the events could result in reputational damage as described by Perry and de Fontnouvelle (2005), which leads to customers shying away from financial institutions (Cummins et al., 2007). Alternatively, the study of Lang and Stulz (1992) on the spillover effects of bankruptcy announcements introduced the competitive effect of events. The competitive effect can be explained as a shift in demand, which cannot be met by supply from the bankrupt firms and thus will be met by the competitors, resulting in a positive spillover effect (Lang and Stulz, 1992).
Due to the high market capacities of banks relative to the sums that are to be paid to institutions in settlements and fines, we do not take these fines into account when calculating the returns.
Single factor market model Perry and de Fontnouvelle (2005). In our research, we look at the first
press cutting date which also often falls on the same day as the recognition date and the settlement
date. This could apply as well to litigation announcements, since the reputational damage incurred
by an individual bank could lead to a demand shift to the competitors which therefore experience
positive spillover effects. In line with previous literature, an event window of (0;1) was selected
since (-5;5) and (-20;20) windows applied by Fiordelisi (2013) did not yield significant results for
any of the determinants (Fiordelisi et al., 2013). The estimation window, used for gathering data to construct an estimation for the event window, has been set at (-251;-1), leaving 250 trading days (approximately a trading year) which is in line with the estimation windows applied by the literature by Fiordelisi (2013), Perry and de Fontnouvelle (2005) and Gillet et al. (2010) for the aforementioned purposes.
When looking past 2008, we find a remarkable lack of literature. This is a gap that will be filled by this paper allowing for comparisons with the pre-crisis researches. Additionally, we look at the spillover effects of reputational damage on the banking index as a whole, which is comprised of both commercial and investment banks. Since this analysis was not conducted by the previously mentioned papers for after the financial crisis of 2008, we can expand the literature on that front as well.
3 Analysis
3.1 Data
For the study to be conducted, an essential step was the gathering of event data. Events are characterised by being unanticipated and the fact that one can put them in time. By applying this definition, we can categorise the announcements of settlements and fines as events. This means that the market has not yet included this new information in the pricing of the assets. Therefore, the stock returns of firms for which an event happens will deviate from the normal expected pat- tern. This deviation is measured by looking at the abnormal returns, which is the actual returns less the estimated returns at the event window.
Given the characteristics of events, data for this research was gathered not only from dependable
sources, but also with an eye on whether it was the first publication of this event. Since the set-
tlements and fines are published by government agencies such as the U.S. Securities and Exchange
Commission (SEC), the U.S. Department of Justice (DoJ) and the Financial Industry Regulatory
Authority (FINRA), their websites are the primary and most reliable sources used for gathering
data. These sources have been coupled to major news outlets such as Bloomberg, Reuters and the
Wall Street Journal which often report financial news as soon as it breaks to the market. This
second check serves to ensure the validity of the data.
These sources have been leveraged to find the dates of events influencing the top 25 publicly traded banks operating in the United States of America starting from January 2009. Using this method- ology and scope, a dataset of 210 events has been created. This was narrowed down to 124 events by leaving out all events by firms traded outside of the United States of America, all cases were two events occurred at the same day and omitting all events of firms which where not traded for the entire period of 2009 to 2018. Also, some events were announced simultaneously (e.g. multiple announced fines in one day due to foreign exchange manipulations, where the bank in question received a fine from both the American and the British institutions), and thus left out of the sample available for the analysis. Along with these events, data has been gathered regarding the stock market returns of the relevant banks and two control groups (Standard and Poor’s Banks Select Industry Index and the Standard and Poor’s 500 Index) for the same period from Yahoo Finance.
Estimation window Event window Events
Code Bank Mean return Window Mean return Window Number First Last Mean (million USD)
BAC Bank of America 0.16% (-251;-1) 0.47% (0;1) 9 2009 2018 61
BK Bank of New York Mellon 0.03% (-251;-1) -0.35% (0;1) 3 2015 2017 17
C Citigroup 0.05% (-251;-1) 0.07% (0;1) 12 2010 2018 97
GS Goldman Sachs 0.01% (-251;-1) 0.25% (0;1) 14 2010 2016 634
JPM JP Morgan 0.10% (-251;-1) 0.16% (0;1) 17 2009 2016 1249
HSBC HSBC 0.03% (-251;-1) -0.28% (0;1) 15 2010 2018 264
ML Meryll Lynch 0.13% (-251;-1) -1.38% (0;1) 8 2009 2018 11
MS Morgan Stanley 0.06% (-251;-1) -0.13% (0;1) 24 2009 2018 210
WFC Wells Fargo Chase 0.01% (-251;-1) 0.23% (0;1) 22 2009 2018 630
Table 1: Summary statistics of the sample
Table 1 gives a summary of the data used for the event study analysis. The highest mean returns
during the estimation and event windows are generated by Bank of America. Most events were
recorded for Morgan Stanley, which span the whole period of data collection (2009-2018). Meryll
Lynch seems to have the largest mean return drops at the times of events, while it has the lowest
average fine or settlement size. JP Morgan, on average, has the largest mean event size with about
USD 1,249 million in fines or settlements on average for every event. This, however, is slightly
skewed by an exceptionally high settlement in 2013, where it paid USD 13 billion in settlements
to a variety of regulatory institutions.
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Number 6 8 8 15 6 21 7 14 9 9
Mean size USD, millions 19.9 99.5 50.1 93.6 3054.8 416.7 105.4 676.7 15.2 94.8 Median size USD, millions 5.0 9.5 12.0 2.4 247.0 20.0 50.0 33.2 8.0 10.5 Min. size USD, millions 0.5 0.65 1.4 0.6 108.0 4.0 1.0 1.0 0.0 0.0 Max. size USD, millions 75.0 550.0 285.0 1256.0 13000.0 3015.0 267.0 5060.0 53.0 765.0
Table 2: Fine statistics over researched years
Figure 1: Event size distribution
The event size distribution, as we can see in Figure 1, is heavily skewed. We can also perceive that the previously named settlement of USD 13 billion, is the highest by a large margin. The events in our sample are laid out over the years with descriptive statistics per year in Table 2. The median sizes show us that over the years, except for 2013 where we perceive a spike in all event sizes, the settlement or fine sizes are rather evenly distributed.
3.2 Methodology
This paper employs two distinct research methods to test the previously stated hypotheses. Firstly,
an ordinary least squares regression was employed with dummies representing the dates of the event
window. Alternatively, an event study assessing the abnormal returns on the day of the event and the first day after the event. This was done both using the Standard and Poor’s Banks Select Industry Index (BIX) and the Standard and Poor’s 500 Index (SNP) as markets. These were chosen since the BIX offers a good indication on the performance of banks listed in the United States of America, while the SNP indicates the performance of a large selection of companies listed in this country. The estimation window and event window have been set as (-251;-1) and (0;1) respectively, which is in line with the aforementioned papers since the event windows of (-5;5) and (-20;20) applied by Fiordelisi did not yield significant results for any of the determinants Perry and De Fontnouvelle (2005) Gillet et al. (2010) Fiordelisi et al. (2013). Since the sums being paid by banks are very small relative to their market capacities, we do not subtract the relative size of the fine from the banking security returns. The estimation models, both for the regression analysis and the event study, have both been based on a single factor market model employed by Perry and de Fontnouvelle (2005) and the CAPM which allows us to create a better estimation by adding the risk free rate to the equation.
3.2.1 Regression analysis
For the regression analysis, we employed the following equation to regress an estimation for indi- vidual events using the market model:
Rr
t= α + βM
t+ γD
t+ δG
t+
t(1)
For the CAPM, we used the following equation to estimate the returns:
Rr
t= Rf
t+ α + β(M
t− Rf
t) + γD
t+ δG
t+
t, (2)
where Rr
t, Rf
tand M
tstand for the returns of a security in the regression analysis, the daily risk
free rate and the market returns, with the market being represented by the Standard and Poor’s
Banks Select Industry Index and the Standard and Poor’s 500 Index, at time t. The dummy
variables can can be explained as:
D
i,t=
1, if firm i experiences an event on time t 0, else
G
i,t=
1, if firm i experiences an event on time t + 1 0, else.
By running these regression, and looking at whether the dummy variables are significant, we were able to assess whether the events significantly affect the returns for every bank (Tables 3 and 4 ADD CAPM REGRESSIONS). Additionally, regressions for all events individually were run to investigate whether the dummies were significant for individual events and a regression was run for all banks and all events, by applying multivariate multiple regressions Binder (1998). Since we analysed the returns of nine major American banks for multiple events, Rr
i,t, Rf
t, M
t, D
i,tand G
i,twere vectors:
Rr
1,1Rr
1,2...
Rr
1,T...
Rr
F,1Rr
F,2...
Rr
F,T
= α
1
1 1 ...
1 ...
0 0 ...
0
+ · · · + α
F
0 0 ...
0 ...
1 1 ...
1
+ β
1
M
1M
2...
M
T...
0 0 ...
0
+ · · · + β
F
0 0 ...
0 ...
M
1M
2...
M
T
+ γ
D
1,1D
1,2...
D
1,T...
D
F,1D
F,2...
D
F,T
+ δ
G
1,1G
1,2...
G
1,T...
G
F,1G
F,2...
G
F,T
+
1,1 1,2...
1,T...
F,1 F,2...
F,T
,
where F is the number of firms, T is the end time, R
i,tis the return of firm i on time t and M
tis
the market return on time t.
3.2.2 Event study
The methodology for an event study is laid out by McKinlay (1997). After linking the events to their respective banks and events, estimation models have been constructed for the individual events. These estimation models are based on the market model, setting the market (being the Standard and Poor’s Banks Select Industry Index) as the only explanatory variable along with the alpha and error term. The market model is to be applied, since it removes returns related to market variance (McKinlay, 1997), and is shown in equation 3:
ER
i,t= α
i,t+ β
iM
i,t+
i, (3)
where ER
i,tstands for the estimated returns with the market being represented by Standard and Poor’s Banks Select Industry Index for security i Eckbo (2008). This estimation is using market and return data from estimation window (-250;-1) to account for a trading year before the event.
Equation 4 depicts the formula for the returns of an event
R
i,t= ER
i,t+ AR
i,t, (4)
where R
i,trepresents the returns, while the abnormal returns (AR) equal the actual returns less the estimated returns Eckbo (2008). The average abnormal returns and cumulative average abnormal returns are calculated using equation 5 and equation 6:
AAR
t= 1 N
N
X
i=1
AR
t, (5)
CAAR
t1,t2=
t2
X
t=t1
AAR
t, (6)
where AAR is the average abnormal returns and CAAR represents the cumulative average abnor-
mal returns. The event window is (0;1), as employed by Merchant and Schendel (2000), Gleason
et al. (2003) and Chiou and White (2005). To test for the significance of the abnormal returns,
a simple t-test can be applied. For testing the CAAR we used a cross-sectional t-test MacKinlay
(1997) Eckbo (2008):
t − value = CAAR
t1,t2[σ
2(t
1, t
2)]
12, (7)
with
σ
2(t
1, t
2) = Lσ
2(AR
t), (8)
and
L = t
2− t
1+ 1. (9)
Using this methodology, we tested whether the AAR and CAAR are significantly different form zero.
All tests have been conducted both using R and Excel
1. In Excel, to determine the alpha for individual events in excel, an intersect function has been applied for the estimation periods, with the bank returns as the y-variable and market returns as x-variable. The beta was to be determined by simply dividing the covariance with the market variance over the same estimation window.
This way, estimations were derived for all the events in the event windows. Subtracting the returns yielded by the estimation from the actual returns in the event yields the abnormal returns and dividing this by the standard error t-statistics were easily retrieved. The same applied for the average abnormal returns in the respective days of the event windows, and the cumulative abnormal returns in day one. For R, the methodology applied was similar. Only here, a short code of instructions was written in RStudio to run the event study automatically in a loop. Tests for significance where run on both individual levels, the average abnormal returns and the cumulative average abnormal returns in day 1 (since the cumulative average abnormal returns in day 0 of the event window are equal to the average abnormal returns of the same day).
1
The data was managed and sorted in Microsoft Excel, where the coupling of each of the sourced events and the
company returns has taken place by a simple combination of index and match formulas. An additional sheet was
used for output to R, to assure a smooth import procedure of data.
4 Results
4.1 Regression analysis
For every bank, individual regressions have been run using Equations 1 and 2 by applying the Banking Index and the General Market Index as market variable in the model. This had to be done, to account for the spillover effects of the litigation and/or settlement announcements which could be influencing the results of the banking index variable. The individual regression results yielded using the General Market index are depicted in Table 3. We can see that individually, both the dummies of day 0 and day 1 are not significant except for day 1 of the bank JP Morgan Chase.
Individual regressions using the General Market Index, result in the figures which are shown in Table 4. In this case, we can see that the dummies are significant at the day of the announcement for Goldman Sachs and the day after the announcement for JP Morgan Chase.
BAC BK WFC MS ML JPM HSBC C GS
coef.D 0.556 0.556 0.384 −0.435 −0.961 −0.944 −0.973 0.337 0.442 p - value D 0.578 0.578 0.701 0.663 0.336 0.345 0.331 0.736 0.658 coef.G −0.180 −0.180 0.272 0.065 −0.433 −1.981 0.858 −1.222 −0.432 p - value G 0.857 0.857 0.786 0.948 0.665 0.048 0.391 0.222 0.666
Table 3: Individual regression outputs (Banking Index as market) for all banks in sample
BAC BK WFC MS ML JPM HSBC C GS
coef.D 0.991 −0.879 0.241 −0.693 −0.229 −0.587 0.863 0.604 2.562 p - value D 0.380 0.578 0.810 0.488 0.819 0.557 0.388 0.546 0.010 coef.G 0.507 −0.564 1.822 −0.851 −1.444 1.273 −0.924 −0.383 −0.363 p - value G 0.612 0.573 0.069 0.395 0.149 0.203 0.355 0.701 0.716
Table 4: Individual regression outputs (General Market Index as market) for all banks in sample
For the multivariate regression, the dummy variables did not turn out to be significant for the
announcement day and the day after the announcement. The regression output of the multivariate
multiple regression with the Banking Index set as market variable in the market model is shown in
Table 5. Both dummy variables were not significant on any conventional level. The multivariate
regression ran using the regular Standard and Poor’s 500 Market Index, depicted in Table 6 did not yield any significant results for the dummy variables either.
Estimate Std. Error t value Pr(>|t|)
βBAC 0.0005 0.0003 1.33 0.1834
βBK 0.0002 0.0006 0.36 0.7191
βC -0.0003 0.0003 -1.14 0.2536 βGS -0.0003 0.0003 -1.33 0.1852
βHSBC 0.0000 0.0002 0.06 0.9523
βJ P M 0.0003 0.0002 1.47 0.1409
βM L 0.0001 0.0003 0.30 0.7619
βM S 0.0002 0.0002 0.80 0.4261
βW F C 0.0001 0.0002 0.56 0.5774
BIX BAC 1.2292 0.0113 108.95 0.0000 BIX BK 0.8895 0.0398 22.36 0.0000 BIX C 1.1429 0.0180 63.49 0.0000 BIX GS 0.8418 0.0148 56.97 0.0000 BIX HSBC 0.4904 0.0114 43.07 0.0000 BIX JPM 0.9379 0.0128 73.23 0.0000
BIX ML 1.1636 0.0145 80.03 0.0000
BIX MS 0.9888 0.0084 118.17 0.0000 BIX WFC 1.0642 0.0091 117.36 0.0000 D -0.0003 0.0014 -0.24 0.8114 G -0.0018 0.0014 -1.28 0.1999 Table 5: Multivariate multiple regression output Banking Index
To assess whether banking index is influenced by the events, we regressed it as well by regressing
it with the regular market index as market variable in the model. These results are shown in Table
7. We were not able to observe significant spillover effects on the market.
Estimate Std. Error t value Pr(>|t|)
βBAC 0.0003 0.0004 0.65 0.5180
βBK 0.0003 0.0007 0.46 0.6468
βC -0.0003 0.0003 -0.91 0.3640 βGS -0.0005 0.0003 -1.59 0.1115
βHSBC -0.0002 0.0003 -0.81 0.4177
βJ P M 0.0002 0.0003 0.57 0.5716 βM L -0.0002 0.0004 -0.60 0.5511 βM S -0.0001 0.0002 -0.29 0.7733 βW F C -0.0001 0.0002 -0.57 0.5695 SNP BAC 2.1564 0.0304 70.89 0.0000
SNP BK 0.0407 0.0449 0.91 0.3649
SNP C 1.6785 0.0354 47.39 0.0000 SNP GS 1.3217 0.0285 46.44 0.0000 SNP HSBC 1.0963 0.0269 40.72 0.0000 SNP JPM 1.5774 0.0301 52.44 0.0000
SNP ML 1.7811 0.0333 53.53 0.0000
SNP MS 2.0961 0.0196 107.08 0.0000
SNP WFC 1.5556 0.0208 74.91 0.0000
D 0.0014 0.0017 0.83 0.4055
G 0.0002 0.0017 0.14 0.8849
Table 6: Multivariate multiple regression output Market Index
Estimate Std. Error t value Pr(> |t|) α -0.0001 0.0001 -1.90 0.0576 βSN P 1.4931 0.0068 219.09 0.0000
D 0.0017 0.0012 1.37 0.1696
G 0.0019 0.0012 1.53 0.1259
Table 7: Banking index regression on Standard and Poor’s 500 Index
4.2 Event study
The t-tests conducted for the average abnormal returns and the cumulative average abnormal re- turns yielding the results depicted in Table 8.
Banking Index General Market Spillover test Day 0 Day 1 Day 0 Day 1 Day 0 Day 1 AAR -0.0003 -0.0021 0.0013 -0.0007 0.0016 0.0125 CAAR -0.0003 -0.0025 0.0013 0.0006 0.0016 0.0028 Median -0.0002 -0.0011 0.0009 0.0005 0.0008 0.0011 Min. -0.0631 -0.0910 -0.0955 -0.0858 -0.0432 -0.0434
Max. 0.0445 0.0404 0.0440 0.1320 0.0381 0.0974
df 123 123 123 123 123 123
t-statistic AAR -0.355 -1.895 0.901 -0.369 1.783 1.0515 p-value AAR 0.723 0.060 0.369 0.713 0.077 0.295
Table 8: Results of the event studies
These results indicate that at for day 1 of the event period, which is the day after the announcement of the litigation process, we find abnormal negative stock returns when applying the Banking Index as market to calculate abnormal returns. For day zero, we find that the Banking Index itself has significant abnormal returns when applying the General Market Index to calculate the abnormal returns. These first results allow us to reject the null hypothesis that the stock returns of banks are not affected by the litigation and settlement announcements.
When looking at a graphical representations in Figures 3 and 2 of the abnormal returns, we may
observe that that the returns do not seem normally distributed. As mentioned in the methodology
section, in the case of a non-normal distribution of the abnormal returns non-parametric tests
had to be applied. However, the central limit theorem applies here due to the high number of
independent samples. Therefore the mean approaches a normal distribution and non-parametric
tests are not needed for this study.
Figure 2: AR distribution on event day 1 (Banking Index as market)
Figure 3: AR distribution on event day 0 (Banking Index as market)
5 Discussion
The results detected by the regressions were surprising when regarding the literature. Individually, none of the returns banks, except for JP Morgan Chase in the day after the announcement, where significantly affected by the litigation and/or settlement when regressed with the Standard and Poor’s Banking Index. The same applied for the multivariate multiple regressions with dummy variables for the days in the event window, where we also concluded that there was no significance.
Replacing the Banking Index with the Standard and Poor’s 500 Index did not change the results.
Also, we found no spillover effects on the Banking Index by the litigation announcements by re- gressing the Banking Index with the general Market Index, which would (if in place) explain the lack of significant results for the Banking Index regressions.
The event study results, in contrast to the regression analysis, are more in line with what we have
seen in literature. We find evidence supporting the findings of P. Sturm (2013) and Perry and de
Fontnouvelle (2005), were significant negative abnormal returns were detected in the event win-
dows. The security returns, when regressed utilising the Banking Index as the market parameter, showed negative significant abnormal returns in the day after the announcement. This indicates that, with the delay of one day, the investors react to the litigation news. The analysis on possi- ble spillover effects yielded unexpected results. The average abnormal returns on the day of the announcements were significant positive. This is surprising since it indicates that investors react positively on news which should deteriorate the trust in the sector. The results could be caused by the fact that the competing banks gain an advantage due to the bad performance of one of their peers. However, we have no proof justifying this possible explanation and we were not able to find literature on this specific issue.
To address the discrepancies between our research outcomes and the literature, several reasons could be brought up. An explanation could be that the announcements by the SEC and alike are not picked up by investors until the day after this has concurred. This would be strange for professional investors, since the major news agencies mostly report it at the same day as the official announcements and this was accounted for when retrieving the data. Less involved parties could react with a slight delay due to later arrival of this information causing the significant abnormal negative returns in the trading day after the announcement. Alternatively, it is a possibility that due to the data being collected as primary research, without a validated database, errors could be situated in the data. This is highly unlikely due to the careful selection of data and the coupling of multiple data source to determine the exact event date. Also, when mistakes are being made with event dates, it is most likely due to the fact that there has been a prior announcement of an event. In this case it seems to be the other way around since we perceive significant abnormal returns in the day after the announcement.
This thesis, despite being carefully constructed, does have several limitations. The data which was analysed was collected by primary research and does not originate from a validated database.
Therefore, errors could come up regarding the accuracy of the data. Additionally, we consider operational losses (being the paid fines and settlements) as too small to affect the security returns of the banks to simplify the research setup. However, in reality these operational losses slightly affect the returns which might lead lower abnormal returns.
Rejecting the null hypothesis but not yielding significant abnormal returns for the day of the an-
nouncements brings up questions on why this one-day delay is in place after 2008 while it was
not the found in earlier studies. This will require additional research to investigate whether this has been the case in other regions as well as in the United States. Also, more qualitative research on this question may add value by assessing investors’ views on reputational damage to banks by these events. The unlikely results yielded by the analysis of spillover effects will need further research as well. This would mainly be by looking at whether this was the case with data from the earlier studies, and if so we would need to dig deeper into the reasoning of the investors by more qualitative research.
6 Conclusion
Previous literature on the topic of reputational damage to banking firms caused by litigation events all focuses on the period before the financial crisis of 2008. This formed a gap in literature, which resulted in this thesis which has assessed these litigation events after the financial crisis of 2008.
To look into whether reputational damage was caused by these events, both a regression analysis and an event study has been conducted with 124 events sampled of the 25 largest banks in the United States of America. This was done utilising both the Standard and Poor’s Banking Index and the Standard and Poor’s 500 Index, to assess whether spillover reputational damage affected the banking industry.
The regression analyses did not yield any significant results. This was surprising when looking at previous studies regarding this topics, which all find significant negative relations between the events and the returns of the banks. We also did not find proof of spillover effects of reputational damage on the banking industry as a whole.
The AAR for day zero and day one where -0.002 and -0.003 respectively, with the CAAR of day 1
being -0.004 when applying the Banking Index for the estimation procedure. The results proved to
be significant on all levels for day one as well as on a cumulative level, which means that the null hy-
pothesis stating that there are no abnormal returns after litigation and settlement announcements
can be rejected. However, these results differ from the literature originating before the financial
crisis of 2008, where significant abnormal returns were detected on the announcement day itself as
well. When applying the regular Standard and Poor’s 500 Index, we found no significant results in
the event window. When looking at the spillover effects of litigation announcement on the Banking
Index, we found a significant positive relationship. This could be the case due to the fact that investors, in stead of losing trust in the industry as a whole, find other banks more attractive as an investment opportunity as a result of the event.
The results of this thesis can be used for a variety of applications. Asset managers and investors
can apply the knowledge to determine their investment strategies related to securities held in bank-
ing firms. Additionally, banks themselves can consider these results when assessing their internal
litigation risks and the results of non-compliance.
References
Algan, Y. and Cahuc, P. (2010). Inherited trust and growth. American Economic Review, 100(5):2060–92.
Basel (2009). Revisions to the Basel II market risk framework. Bank for International Settlements.
Binder, J. (1998). The event study methodology since 1969. Review of quantitative Finance and Accounting, 11(2):111–137.
Campbell, C. J., Cowan, A. R., and Salotti, V. (2010). Multi-country event-study methods. Journal of Banking & Finance, 34(12):3078–3090.
Cummins, J. D., Wei, R., and Xie, X. (2007). Financial sector integration and information spillovers: Effects of operational risk events on us banks and insurers. The Journal of Risk and Insurance.
Eckbo, B. E. (2008). Handbook of Empirical Corporate Finance, 2 volume Set. Elsevier Ltd.
Fama, E. F. and MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. Journal of political economy, 81(3):607–636.
Fiordelisi, F., Soana, M.-G., and Schwizer, P. (2013). The determinants of reputational risk in the banking sector. Journal of Banking & Finance, 37(5):1359–1371.
Gillet, R., H¨ ubner, G., and Plunus, S. (2010). Operational risk and reputation in the financial industry. Journal of Banking & Finance, 34(1):224–235.
Gompers, P., Ishii, J., and Metrick, A. (2003). Corporate governance and equity prices. The quarterly journal of economics, 118(1):107–156.
Grasshoff, G., Mogul, Z., Pfuhler, T., Coppola, C., Wiegand, C., Villafranca, V., and Vonhoff, V.
(2018). Future-proofing the bank regulatory agenda. Boston Consulting Group.
Grasshoff, G., Mogul, Z., Pfuhler, T., Gittfried, N., Wiegand, C., Bohn, A., and Vonhoff, V.
(2017). Global risk 2017: Staying the course in banking. Boston Consulting Group.
Guiso, L. (2010). A Trust-driven Financial Crisis.Implications for the Future of Financial Markets.
EIEF Working Papers Series 1006, Einaudi Institute for Economics and Finance (EIEF).
Lang, L. H. and Stulz, R. (1992). Contagion and competitive intra-industry effects of bankruptcy announcements: An empirical analysis. Journal of financial economics, 32(1):45–60.
MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of economic literature, 35(1):13–39.
Perry, J. and De Fontnouvelle, P. (2005). Measuring reputational risk: The market reaction to operational loss announcements. Federal Reserve Bank of Boston.
Roth, F. (2009). The effect of the financial crisis on systemic trust. Intereconomics, 44(4):203–208.
Sturm, P. (2013). Operational and reputational risk in the european banking industry: The market reaction to operational risk events. Journal of Economic Behavior & Organization, 85:191–206.
Walter, I. (2008). Reputational risk and conflicts of interest in banking and finance: the evidence
so far. In Variations in Economic Analysis, pages 75–97. Springer.
A Non-parametric tests
In the event study, parametric tests were employed since normality was assumed, while non- parametric tests have to be applied when abnormal returns are not normally distributed. However, we can test whether we have to employ non-parametric tests, we look at the Jarque-Bera statistic of the abnormal returns. Campbell et al. (2010) has proposed several of these, where the Cowan sign-test and Corrado’s rank test are best suitable for event studies. Cowan (1992) reported the test to be very powerful in random NYSE-AMEX samples. Corrado’s rank test developed by Corrado (1989) does not allow for missing values, which is why Campbell et al. apply the same procedure as Corrado and Zivney (1992). The Corrado rank test first requires a ranking of the abnormal returns represented by Campbell et al. (2010):
K = rank(AR
i,j) (10)
with
AR
i,j> AR
i,k→ K
i,j> K
i,k. (11) With these specifications, the test statistic is:
CT =
1
N
