Impact of counter-cyclical capital buffer on returns of banks

(1)

Bachelor Thesis:

IMPACT OF COUNTER-CYCLICAL

CAPITAL BUFFER ON RETURNS OF

BANKS

JULY 15, 2020

Written by: Ngoc Lan Nguyen

Student number: 11798831

(2)

Statement of Originality

This document is written by Student [Ngoc Lan, Nguyen] who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Following the Global Financial Crisis in 2008, the Counter-Cyclical Capital Buffer (CCyB) policy was introduced via Basel III, which aims to increase resilience of the financial sectors and reduce the excessive risk-taking. However, by increasing the capital requirement, banks’ performance would likely be affected, and by extension their market value. In this study, we looked into how the increase in CCyB rate affects banks’ market returns in France. The results show an inconsistent result with our hypothesises, in which there is no significant change in market returns of banks after the CCyB rates were raised. The study also considers two features of banks, their sizes and capital ratio prior to the raise in CCyB rate, which can affect how banks’ market returns would react to the raise in CCyB rate. Evidence shows there is no significant difference in market returns between small banks and large banks, or between banks with higher capital ratios and banks with low capital ratios.

(4)

I. Introduction

Following the damaging effects of various economic crises such as the 2008 housing crisis or the most recent COVID-19 Pandemic, policy makers are generally being pressured to improve the resilience of economic systems and prepare for future market failure. As a result, macroprudential policies have become more common among emerging markets and advanced countries. The main aim of macroprudential policies is to safeguard the financial system against systematic risks and procyclicality (Cerutti, Claessens & Laeven, 2017).

After the 2008 financial crisis the European Union implemented Basel III rules which introduced the policy of “countercyclical capital requirements - CCyB.” (European Systemic Risk Board, 2020). This policy is meant to increase the resilience of the financial sector and reduce excessive risk-taking by smoothening the credit supply cycle. In the first phase, banks build up a buffer of required capital during prosperous times. Throughout the second phase the buffer is then released to lower the realised losses during crises, hence reducing the need for governmental bailouts and other costs (Jiménez, et.al., 2017). This policy is meant to be an improvement to the previous version of capital requirement, which was introduced in the Basel II framework. The 2-phase approach of CCyB policy reduces the procyclicality feature caused by the previous capital requirement policy (Shim, 2013).

Currently, there are already 31 countries within Europe (including all European Union members and other European countries) that implemented the CCyB policy (European Systemic Risk Board, 2020). This study aims to deepen the understanding of how the market value of banks (which is measured by stock returns) is affected by the introduction of CCyB nationwide. The result of this research will contribute to the long-standing literature on how bank capitalization affects bank value. According to the theory of the frictionless market by Modigliani and Miller (1958), a change in capital structure, by increasing capital and reducing leverage, would not affect the value of the firm. However, given this frictionless market assumption does not hold in reality, our findings try to show that the market value of banks would decrease with the increase in capital requirement. Furthermore, by comparing banks with different conditions (high and low capital ratio, large and small banks) prior to the announcement of CCyB policy, we will have a better understanding of to what extend different banks’ value are affected by the change in capital structure.

In this study, we will look at France, one of the biggest economies and banking systems in EU, and how its banks were affected by CCyB policy. France first raised its capital buffer to 0.25% on July 1st_{, 2018}

and this rate was implemented a year later on July 1st_{, 2019 (European Central Bank, 2019). It then went}

on to increase the buffer to 0.5% on April 3rd_{, 2019 (Haut Conseil de Stabilite Financiere (HCSF),}

2020). Then in response to the recent COVID-19 crisis, the authority made an announcement to reduce the CCyB rate to 0% on April 1st_{, 2020. The research question of this study is:}

(6)

Does increasing the CCyB rates have an impact on the stock market performance of French banks between 2018 and 2020?

Using event study methodology, we will analyse the abnormal returns of French banks’ stocks around the announcement date of July 1st_{, 2018 and April 3}rd_{, 2019. Another event study will then be carried}

out around the same announcement dates focusing on German stock returns. Germany has a similar economy to France, but only introduced the rate at 0.25% for the first time on July 1st_{, 2020 (Federal}

Financial Supervisory Authority, 2020). By comparing the results of the abnormal returns between the banks in two countries around the same announcement dates, we account for other factors that apply more broadly to market value of European banks, for instance in Germany, and are not directly linked to CCyB announcement dates in France. Regarding the most recent announcement date (April 1st_{, 2020),}

the market was extremely volatile due to the COVID-19 crisis and would be more difficult to measure the impact of the change in CCyBs on banks’ stock price alone using an event study. Therefore, this announcement date is out of this study’s scope. The results found is inconsistent with our hypothesis, as there is no significant abnormal returns of French banks surrounding the announcement dates of CCyB.

The study also went on to examine if certain features of banks prior to the announcement, more specifically banks’ size and banks’ capital ratio, can affect their abnormal returns around the announcement. Our results show an insignificant effect from the two factors (size and capital ratio) on banks’ abnormal returns. A possible explanation for the inconsistenccy between our hypotheses and the results found might be due to the small sample size of this study.

The paper is structured as follow: Chapter 2 provides literature backgrounds and the context of the CCyB policy, and theories on its potential effects on banks’ market returns. Chapter 3 starts off with an introduction to the hypotheses of this paper, then the data collection procedures and the event study methodology used in this study are explained. Followed by Chapter 4 which presents and explains the results of the study. The paper is concluded in Chapter 5 with recommendations for further research.

(7)

II. Theoretical Framework

The purpose of prudential policies is to ensure the resilience and stability of the financial system (Beyer, et.al., 2017). Micro-prudential policies aim to tackle the problem from inside out, by monitoring individual credit institutions (firm-specific risks) and therefore contributing to the stability of the system as a whole. Macroprudential policies, on the other hand, monitor the systematic risks in the financial system and absorb the damages of the cyclical downturns as much as possible (Beyer, et.al., 2017). Therefore, macroprudential policies are meant to complement monetary policies to ensure the stability of both economic and financial systems. Macroprudential policies specifically tackle three externalities sourced from the financial sector. According to a study of De Nicoló, Favara & Ratnovski (2012), the three externalities are: (1) “externalities related to strategic complementarities” where banks take excessive risks and hence cause procyclicality, (2) “externalities related to fire sales” causing a shock in asset price, and lastly (3) “externalities related to interconnectedness” which are externalities spilled over from other institutions.

According to a cross-country study across 16 Eastern and Southeastern European (CESEE) countries, there are 29 categories of macroprudential measures (Vandenbussche & Detragiache, 2015). These measures can be divided into 5 groups of “capital measures, provisioning measures, liquidity measures, loan eligibility requirements, and other quantitative restrictions” (Vandenbussche & Detragiache, 2015). Their study suggested that capital measures and liquidity measures are effective in smoothening the credit supply cycle. Among these 5 groups, CCyB would be considered a capital measure. Capital requirement measures aim to set out a minimum capital for lending institutions (Vandenbussche & Detragiache, 2015). This policy tries to address externalities by monitoring the supply side (financial institutions), especially in reducing the excessive risky lending behavior of banks (Cerutti, Claessens & Laeven, 2017). However, there is a potential problem with macroprudential policies, and more specifically capital requirements, which is the possibility of an increase in cross-border borrowing, thus a reduction in the effectiveness of the policies on the targeted banking system (Cerutti, Claessens & Laeven, 2017).

There are two main groups of cross-country studies regarding the effect of macroprudential policies on financial stability. The first group focuses on the link between macroprudential policies and credit growth, and the other focuses on the link between macroprudential policies and systematic risks in banks (Cerutti, Claessens & Laeven, 2017). The effectiveness of macroprudential policies can also be studied through event study methodology (Vandenbussche & Detragiache, 2015). In this study, we will focus on the later method, event study.

Capital requirements are first introduced into the European Union in the Basel II framework, introducing the model-based capital charges for risky assets. This means the riskier the assets that banks

(8)

hold, the higher capital the bank is required to have. The policy is meant to monitor bank’s excessive risk; at the same time aims to build up a capital cushion for banks to absorb the losses during downturns, allowing banks to continue lending even during recession. However, it was soon revealed by many researchers from 2001 to 2008 that the capital requirement policy has procyclical feature. (Furfine, and Lowe, 2001; Goodhart, Hofman, and Segoviano 2004; Gordy Rochet 2008 as cited in Behn, Haselmann & Wachtel, 2016) Instead of reducing the effect of the business cycle, it amplifies the fluctuations as banks reduce lending due to the capital charges. According to the policy, if banks’ assessment to asset risk is responsive to the economic environment, the capital changes will increase during recession. In general, there are two ways for banks to increase their capital ratios to match with the regulation: (1) increasing their capital, or (2) reducing their risk-weighted assets (Gropp, Mosk, Ongena & Wix, 2019). According to the study on German banks during the implementation of the Basel II regulations, banks gravitated towards the second option to cope with the increase in capital requirement during cyclical downturns (Behn, Haselmann & Wachtel, 2016). The study also shows that the procyclical effect is even stronger at the firm level, as firms received less loans during these periods. As the availability of credit decreases significantly during periods of crisis, the procyclical effect of the capital requirement policy leads to a credit crunch and a deepening of the recession (Behn, Haselmann & Wachtel, 2016). Instead of reducing systematic risks, the capital requirement in Basel II backfired and caused an even more pronounced procyclical effect. Possible causes for this may be due to “debt overhang” and “asymmetric information” within banks (Gropp, Mosk, Ongena & Wix, 2019). Due to the high level of debt, bank shareholders would rather be forgoing any positive NPV projects since they believe the cash flows would eventually go to debtholders instead. While information asymmetry refers to how banks are afraid that issuing equity would signal to the market that their stock price is overvalued (Gropp, Mosk, Ongena & Wix, 2019). Therefore, to meet the required capital, banks would reduce lending action and increase the instability of financial sector.

Learning from the limitation of the policy in Basel II the new policy, countercyclical capital buffer (CCyB), was introduced in the Basel III framework. There are two main objectives of the policy: (1) a buffer for downturns of the financial cycle and preventing “credit crunch” and (2) mitigate credit-led booms by increasing the cost of lending during good times (Jiménez, et.al., 2017). The same is true with the original policy of capital regulation, the goal being to mitigate and prevent excessive credit growth and leverage. The difference between these policies is the 2-tier structure of the policy. This means a buffer will be built-up during good times, and during the downturn regulators can flexibly reduce the required capital, releasing the buffers for banks to make-up for the losses (Jiménez, et.al, 2017). The 2-tier structure of the policy allows banks to build-up their buffers, by raising lending costs gradually and limiting the sudden raise in lending costs when downturns or credit shocks hit (Benes and Kumhof, 2015). With sufficient capital and a steady flow of credit, the banking sector helps reduce the damages

(9)

of recession in the economy, and hence increases the wellbeing of other sectors. According to a study by Jiménez, et.al (2017), the welfare created by the policy particularly increases firm-level credit supply, employment, and survival rate in times of crisis.

However, there are some concerns regarding the limitations of the policy. There has been no proven significant impact on mitigating credit-led booms during good times, as firms can still seek out banks that are outside the jurisdiction of the regulations (Jiménez, et.al, 2017). This limitation is also one of the biggest issues faced by capital regulation measures as mentioned before. Furthermore, according to Behn, Haselmann & Wachtel (2016), when an unexpected shock happens, such as the case of the rapid spread of COVID-19 pandemic, regulators might not be able to anticipate and adjust the buffer in time. Once the adjustment is eventually implemented banks have already suffered severe consequences. Another point worth mentioning, is that the 2-tier structure of the CCyB seems promising in reducing systematic risk by increasing the cost of lending, but is the measure actually effective in reducing bank risks? According to a study by Jiménez, et.al. (2017), after the policy shock, the higher capital requirements can cause banks to focus more on firms with higher interest paid and leverage before the event. These firms have higher default rates and an increased risk to the banks (Jiménez, et.al., 2017). Therefore, the risk-level of banks may not decrease as expected by the measure. The risk level and risk changes of banks before and after the implementation of CCyB policy need to be examined. However, this may be an avenue for future research since the main focus of this paper is to examine the market returns of banks but not their risks. Therefore, in this paper, we would assume that the CCyB policy effectively reduces banks’ risk as it should.

As mentioned previously, when CCyB was implemented or the rate is raised, banks are forced to increase their capital ratio by either increasing capital or reducing lending. For the first option, increased capital can be accomplished through issuing out new equity or using retain earnings. Issuing out new equity would be more costly due to higher transaction costs or share price reductions, which explains why banks prefer to finance their capital through retain earnings (Myers and Majluf, 1984). As a result, banks might try to pass on the new burden of building-up the buffer to borrowing firms by raising the lending costs, and most likely would cause a lower growth rate in their lending portfolio. Therefore, both decisions would help to slow down the credit-led boom cycle, the desired effect of the CCyB policy. The decrease in lending activities and growth leads to a decrease in future profits, and hence banks’ market value. Furthermore, according to Baker and Wurgler (2015), as capital requirement increases and banks’ risk decreases, the cost of capital of banks would increase, leading to lower market value. This theory, however, seems to go against the long-standing belief of CAPM-model, as when the bank risks decrease, so does its cost of equity, and in turn its cost of capital. The reason why the theory stated that the cost of capital ends up increasing, instead of decreasing, is due to a market inefficiency:

(10)

the “low-risk anomaly” (Baker and Wurgler, 2015). With both of the above theories, market value of banks would then decrease with the increase in capital requirement.

Besides looking at an overall relationship between banks’ market value (stock returns) and CCyB rate, this paper also analyses to what extend different banks are affected by the policy. When banks are forced to build up a larger capital buffer, increasing their cost of lending, it does not mean that all banks faced the same costs and reduction in Net Income (Basten, 2020). As mentioned by Jiménez, et.al (2017), the increase in lending price might cause borrowing companies to shift from banks more affected by the policy to those less affected. Less affected banks are banks with lower “capital cushions” – the difference between their pre-existing capital ratios and the new capital requirement (Basten, 2020). Thus, these banks would raise the lending cost less than banks with higher cushions. Some possible motivations which may lead banks to hold higher capital are (1) higher retain earning – due to “pecking order” theory, as banks’ retained earnings increase so do their capital; (2) “economic capital” – banks try to match their capital ratio with their level of systematic exposures; and (3) “acquisition plans” – banks save up their capital for future acquisition opportunities (Berger, DeYoung, Flannery, Lee & Öztekin, 2008). An empirical study by Basten (2020) regarding the effect of CCyB on mortgage pricing in Switzerland, further strengthens this theory. Banks with lower pre-exiting capital ratios are under higher pressure to increase their capital or reduce their risk-weighted assets more to meet the capital requirement compared to banks with higher pre-exiting capital ratio. These more affected banks increase their lending price much more than less affected banks. Even though the study by Basten (2020) did not show a significant decrease in lending activities of more affected banks, the higher cost of lending may result in more affected banks to have a lower Net Income in the future. By comparing the market value of banks with above-median existing capital ratios to banks with below-median pre-existing capital ratios, this paper examines if the profitability of banks that have higher capital cushion would be more affected by raising CCyB rate compared to banks with lower capital cushion.

Lastly, our study also contributes to the smaller body of literature surrounding how CCyB policy affects different sized banks. According to a study on French banks by Coffinet, Coudert, Pop & Pouvelle (2012), during good times, large banks tend to hold less capital compared to others. This evidence can be explained by the “too-good-to-fail hypothesis,” in which the larger the bank, the more likely that the government would intervene in case of default. This would lead to larger banks having less need to keep more capital buffer. With the lower pre-existing capital ratio, larger banks may be more affected by the CCyB policy, effecting their market value. Berger, DeYoung, Flannery, Lee & Öztekin (2008) also had the same result in their research regarding the relationship between banks’ size and targeted capital ratio prior to CCyB policy. The research also adds another motivation to why large banks hold less capital. Large banks are more diversified and have an advantage of economic of scale in risk management, leading to them having a lower cost of raising more equity, even over a short time period. However,

(11)

this can also imply that with the increase in CCyB rate, large banks are less affected in term of their stock value compared to small banks even though they might hold less capital. Therefore, the size of banks can also directly affect the stock returns of banks around the announcement of CCyB.

On December 30th_{, 2015, the Haut Conseil de Stabilité Financière (HCSF) of France decided to}

implement the CCyB policy for the first time with the rate of 0% (Federal Financial Supervisory Authority, 2020). HCSF continued to monitor broad credit to different sectors, “as percentage of GDP, growth rates or gap against its long-term trend.” At the beginning of 2018, France’s financial cycle started to show signs of an increasing trend and even overpassed it long-term averages, driven by non-financial private sector and residential housing. On July 1st_{, 2018, HCSF decided to increase the CCyB}

rate to 0.25%. And on April 3rd_{, 2019, HCSF announced it would increase the rate again to 0.5%}

(Federal Financial Supervisory Authority, 2020). The reason we chose the French market as the studied population is due to the fact that the majority of French banks are mutual banks, in which their lending growth is highly affected by the size of the capital buffers (Coffinet, Coudert, Pop & Pouvelle, 2012). Therefore, the lending behavior of French banks, and their market value, are more sensitive to the change in CCyB rates.

(12)

III. Methodology

To test the two hypotheses stated below, this study uses Event Study methodology to test if there are any abnormal returns of banks pre- and post-announcement date.

3.1 Hypothesis

According to efficient market hypothesis by Farma French (1991), stock prices reflect all available information in the market. Therefore, with the announcement of raising the countercyclical buffer (to 0.25% on July 1st_{, 2018, and increase to 0.5% on April 3}rd_{, 2019), banks would incur costs of raising}

equity in such a short period of time, or reduce their lending activities to meet the requirement rates. This would lead to a decrease in Net Income and thus be reflected in their market value. However, it is uncertain if these announcements are unexpected by the market or not, a question we will return to in the Conclusion.

As a result, stock returns of French banks would decrease more than expected after the announcement dates. To exclude the effect of the policy from other external factors that may have occurred at the same time and affected European banks, we will then compare the abnormal returns around the announcement dates to the abnormal returns of German banks.

Hypothesis 1:

H0: French banks experienced no abnormal returns around the announcement dates of raising the countercyclical capital buffer rates.

H1: French banks experienced negative abnormal returns around the announcement date of raising the countercyclical capital buffer rates.

Hypothesis 2:

H0: There is no statistical difference in the abnormal returns of the French and German banks around the announcement of raising the countercyclical capital buffer rates.

H1: French banks experienced more negative abnormal returns than German banks around the announcement of raising the countercyclical capital buffer rates.

Next, the paper examines if banks with different features react differently to the raise in CCyB rate. To do this, we focus on two characteristics: banks’ capital ratio and banks’ size. According to previous empirical studies of banks in different countries, such as France and the United States, there is a significant relationship between banks’ size and their capital ratio. We would want to test this theory on our data as well before understanding how these two factors can affect the abnormal returns of banks during the announcement dates of CCyB.

(13)

Hypothesis 3:

H0: There is no relationship between banks’ size and their capital ratio, and hence these two variables are independent.

H1: There is a relationship between banks’ size and their capital ratios, and hence these two variables are dependent.

As banks with higher capital ratio prior to the announcement of CCyB rate face less pressure to build up their capital buffer, their stock returns would be less affected by the announcements compared to banks with lower pre-existing capital ratio. With that, we have the third hypothesis,

Hypothesis 4:

H0: Banks with different pre-existing capital ratios have the same abnormal returns around the announcement of raising the countercyclical capital buffer rates in France.

H1: Banks with higher pre-existing capital ratio would have smaller abnormal returns compared to banks with lower pre-existing capital ratio around the announcement of raising the countercyclical capital buffer rates in France.

Lastly, we would test the relationship between firm size and the abnormal market returns of banks around the announcement the CCyB rates. It is more difficult to predict which direction the correlation would go since previous empirical studies show that they could move in either direction.

Hypothesis 5:

H0: Banks with different sizes have the same abnormal returns around the announcement of raising the countercyclical capital buffer rates in France.

H1: Banks with different sizes have different abnormal returns around the announcement of raising the countercyclical capital buffer rates in France.

3.2 Data collection

Since the subject of study is French banks pre- and post-announcement date, this study collected the stock prices of all public banks available from February 1st_{, 2018 to June 15}th_{, 2020. The data of banks’}

stock price was collected through the dataset “Compustat – Capital IQ”. In total, the data from 34 French banks were collected. Among those, the stock price data of four banks (Ashler & Manson, BNP Paribas branch 1, Credit Agricole du Languedoc branch 2 and Credit Lyonnais) seems to be unchanged throughout the whole period. The constant stock prices throughout the whole seven-month period seems to be unlikely in reality. This abnormality can be explained by acquisition event (Credit Lyonnaise),

(14)

shutting down of bank’s branches or no real evidence was found (Ashler & Manson). Regardless, after running the event study analysis for both cases, including the 4 banks’ stock prices and excluding the 4 banks’ stock prices, the data of these four banks were taken out of the study as they did not significantly affect the final results of the study. Furthermore, we also took out the data of Locindus bank, as it was announced to enter acquisition on 26th_{June 2018, right before the announcement date of CCyB. This}

would make it difficult to determine if the abnormal returns (or the lack thereof) from Locindus bank was due to the announcement of the CCyB rate or due to the acquisition. Therefore, the study sample of French banks consists of 29 banks in total.

We then categorized these banks under two factors: banks’ size and their capital ratio a year prior to the announcement dates (2017 and 2018). Banks’ capital ratio is measured by the ratio of Core Equity Tier 1 (CET 1) to the bank’s risk-weighted assets (RWA). Among the 29 observed French banks, we ould only obtain information on capital ratio of 10 banks for 2017 and 15 banks for 2018. Banks were then sorted according to their capital ratios, as either high capital ratio (above median) or low capital ratio (below median) as shown in Appendix 1. To measure the size of firms, we looked at their total assets a year prior to the announcement dates (2017 and 2018). We collected the total assets data for 19 banks out of 29 banks in 2017 and 2018. The banks are then sorted as a large firm or small firm according to the size of their assets (Appendix 1).

Since the paper focus on analyzing the data of French banks more, only information of stock returns of German banks was collected. In total, stock price data of 13 German banks were collected from the same period (from February 1st_{, 2018 to June 15}th_{, 2020).}

The announcement date is the most important input needed for this study since the returns of banks would be observed and studies around this date. Even though the first announcement date of the raising CCyB rate to 0.25% for French banks is in July 1st_{2018, the event date of this study was set to be in}

July 2nd_{, 2018 since it is the nearest trading day from the announcement date (Federal Financial}

Supervisory Authority, 2020). The second announcement date is April 3rd_{, 2019.}

3.3 Event study methodology

According to Armitage (1995) and Park (2004), the results of an event study are not significantly affected by varying the estimation window length as long as the length exceeds at least 100 days. The decision of setting estimation window length is important as it needs to balance the tradeoff between accuracy and bias. As the longer the estimation window is the more accurate the estimation becomes, but it also suffers from bias more due to too much noise in the data. With that in mind, the estimation window length of this study is set to be around 100 trading days or more precisely 5 months before the announcement date. Fifteen days prior to the announcement date were excluded from the estimation window to avoid any contamination by the event. Therefore, the event window consists of 21 days, 10

(15)

days prior to the announcement date, the announcement date and 10 days after the announcement date [-10,10].

This event study analyses the abnormal returns of French banks and German banks within the event window. The abnormal returns are calculated by deducting the returns that would be expected if the event did not taken place (normal returns) from the actual returns of the banks’ stocks. For the actual returns, the collected stock returns of both Germany and France banks, as discussed in the section above, are used. Meanwhile, the normal returns need to be estimated using expected return models. This study uses “market model”, which is based on CAPM risk-factor, to estimate normal returns. Even though “market model” received some criticism regarding its underlying assumptions, the model is chosen as it is a widely accepted and used model. Furthermore, “market model” so far produces “smaller variances of abnormal returns and smaller correlations across securities” (Coutts, Mills & Roberts, 1994). Equation (1) shows that formula of the market model:

𝑁𝑅𝑖,𝑡= 𝛼̂𝑖+ 𝛽̂𝑖𝑅𝑚 (1) The normal return (NR) in equation (1) is estimated using data in the estimation window. Then the abnormal return (AR) is calculated by subtracting actual returns of banks (𝑅𝑖,𝑡) to expected normal return as shown in equation (2).

𝐴𝑅𝑖,𝑡= 𝑅𝑖,𝑡− 𝑁𝑅𝑖,𝑡 (2) To further analyze the total impact of the event over a particular period of time within the event window, daily ARs are added up to get the cumulative abnormal returns (CAR) for each bank as shown in equation (3). In this equation, 𝑡1= −10, which is when the event window starts and ending at 𝑡2 = 10.

𝐶𝐴𝑅𝑖= ∑𝑡𝑡=𝑡2 1𝐴𝑅𝑖,𝑡 (3) Then we calculate the average of CARs across all the banks to get cumulative average abnormal returns (CAARs), where N is the total number of banks:

𝐶𝐴𝐴𝑅𝑖 = 1

𝑁∑ 𝐶𝐴𝑅𝑖 N

i=1 (4) There are two sets of CAARs for each event (respectively for French banks and German banks), so four sets in totals.

3.4 Testing abnormal performance

Lastly, tests of significance were carried out to test if the abnormal returns calculated with the procedure above are statistically significant. In another word, it would conclude if any detection in the abnormal stock returns caused by the announcement is significant. Thus, the two hypotheses would be tested are:

(16)

Hypothesis 1: French banks experienced negative abnormal returns around the announcement date of raising the countercyclical capital buffer rates.

𝐻0: 𝐶𝐴𝐴𝑅𝐹𝑟𝑒𝑛𝑐ℎ 𝑏𝑎𝑛𝑘𝑠= 0 𝐻1: 𝐶𝐴𝐴𝑅𝐹𝑟𝑒𝑛𝑐ℎ 𝑏𝑎𝑛𝑘𝑠 < 0

To test for the statistical hypothesis above, we use a one sample t-test to check if the abnormal returns found in French banks around the announcement dates are significantly larger than 0. Different significant levels are used from 1%, 5% to 10%. The formula of t-statistic is as follow:

t = √N𝐶𝐴𝐴𝑅

𝑠 (5) Where N is the number of French banks and s is the standard deviation of CARs across different banks. To give a more insightful result, the t-test is also then carried out to test the significant of CAR of each individual banks.

Hypothesis 2: French banks experience more negative abnormal returns than German banks around the announcement dates

To test if the abnormal returns around announcement dates are due to the impact of CCyB policy, we need to set a control group to compare with the abnormal returns of French banks around the announcement dates of CCyB. However, it is not possible to find a control group in French banks population, as the policy was implemented nationwide. Therefore, we choose German banks as our control group which did not experience any change in CCyB rate between 2018 to April 2019. Germany remained their CCyB rate at 0% and only announced to increase the rate to 0.25% on June 28th_{, 2019.}

The same event studies around the two announcement dates (July 2nd_{, 2018 and April 3}rd_{, 2019) were}

carried on the German banks to find the abnormal returns around the event time. Then a linear regression analysis was run on CARs of each individual banks across both countries (26 French banks and 13 German banks) against the dummy variable: France. France variable indicates which country does the CAR of the banks belong to. France = 1 indicates French banks while France = 0 if it is German banks. With that, we have the equation:

𝐶𝐴𝑅𝑠𝑖 = 𝛽0 + 𝛽1× 𝐹𝑟𝑎𝑛𝑐𝑒𝑖 + 𝜀 (6) Where CARs are the cumulative abnormal returns of banks in both countries, Germany and France, 𝛽0 is the constant variable and 𝛽1 is the coefficient of France variable. The robust standard error is added to solve any heteroskedasticity in the data. The regression analysis is carried for both event dates. Therefore, the second statistical hypothesis is formulated as:

(17)

𝐻1: 𝛽1< 0

3.5 Testing for factors affecting abnormal performance of banks

As mentioned in previously in the Hypothesies section, we suspected there are two factors that can affect the abnormal returns of French banks around the announcement dates: banks’ size and capital ratio. First, to test for hypothesis 3, we wpould test whether there is a significant relationship between the two categorical variables – banks’ size and capital ratio – using the Fisher’s exact test instead of the Chi-square test. The reason we chose the Fisher’s exact test is due to our small sample size, with most of the expected values in the contingency table below 5 (Appendix 2). Looking at Table 1, the number of large banks have low capital ratio prior to announcement dates are much higher in both years (2017 and 2018). P-value is then calculated to find the probability of getting the observed data, assuming the null hypothesis is true. We would reject the null hypotheis if p-value of the test is smaller than 0.05, showing that there is a correlation between the two variables.

Table 1: Contengency table of French banks’ size and capital ratio prior to announcement dates

2017 Size Total

Small Large

Capital ratio Low 1 4 5

High 2 3 5

Total 3 7 10

2018 Size Total

Small Large

Capital ratio Low 3 5 8

High 4 3 7

Total 7 8 15

Next, we would examine how the two mentioned variables, size and capital ratio, can affect the abnormal returns of French banks. As the number of banks with asset information and capital ratio information are different (10 and 19 respectively), we first tested the effect of these two variables on bank CARs in two separate linear regressions, hence optimising the data usage. For the robustness test, we performed a joint regression, which has a smaller sample size, of both varaibles against the CARs of banks around the announcement dates. However, if the result of the Fisher’s exact test of independence above show that there is a correlation between the two independent variables (size and capital ratio), it will be impossible to preform this joint regression.

(18)

The first regression is:

𝐶𝐴𝑅𝑠𝑖 = 𝛽0 + 𝛽1× 𝐻𝑖𝑔ℎ 𝑐𝑎𝑝𝑖 + 𝜀 (7) In equation (7), we have a dummy variable: High cap, with High cap = 1 if the bank has higher pre-existing capital ratio and High cap = 0 if the bank has lower capital ratio.

The second regression is:

𝐶𝐴𝑅𝑠𝑖 = 𝛽0 + 𝛽1× 𝐿𝑎𝑟𝑔𝑒𝑖 + 𝜀 (8) Where the Large = 1 means large banks and Large = 0 represents small banks.

And the joint regression:

𝐶𝐴𝑅𝑠𝑖 = 𝛽0 + 𝛽1× 𝐻𝑖𝑔ℎ 𝑐𝑎𝑝𝑖 + 𝛽2× 𝐿𝑎𝑟𝑔𝑒𝑖+ 𝜀 (9) Where 𝛽1 is the coefficient of the High cap variable and 𝛽2 is the coefficient of the Large variable.

(19)

IV. Data Analysis

4.1 Testing the first hypothesis

The average cumulative abnormal returns (CAARs) across French banks are calculated in both announcement events (Table 2). In both cases, CAARs have a positive value of 0.01 and 0.044. This result is against our hypothesis of which CAARs should be negative. In the first event, CAAR across all banks is not statistically significant due to a relatively large p-value (0.23) of the t-test. This means that null hypothesis is not rejected and on average, there is no abnormal returns around the first announcement. Cumulative abnormal returns (CARs) are also calculated for every individual firm. Only two firms are found to have positive significant abnormal returns at 5% level.

In the second announcement date, CAAR across all banks is significant at 1% level and hence the null hypothesis is rejected. However, the direction of the result is opposite with our alternative hypothesis. Looking at the CARs at firm level, there are only four firms have significant CARs out of 29 firms (one firm is taken out from the analysis of the second event as it is registered to be delisted in 2019).

Table 2: CAR’s broken down by banks of France

This table represents the cumulative abnormal returns by each French bank in each event and the p-value of the t-test for each coefficient. Event 1 represents the first announcement date on July 2nd_{, 2018.}

Event 2 represents the second announcement date on April 3rd_{, 2019. There are in total 29 banks being}

observed. The last row represents the average CARs across all banks. * stands for a significance level of 10%, ** for 5% level and *** for 1% level.

Event 1 Event 2

Banks Estimate p.value Estimate p.value

BNP PARIBAS 1 0.02513 0.56482 0.08678 0.10587 BNP PARIBAS 2 0.02666 0.54048 0.09344 0.07672* BNP PARIBAS 3 0.02407 0.59610 0.08864 0.10840 BNP PARIBAS 4 0.02485 0.57038 0.09218 0.08956* CA TOULOUSE 31 CCI -0.00795 0.78144 -0.04078 0.22136 CRCAM ATLANTIQUE VENDEE -0.02939 0.36280 0.04736 0.17425 CRCAM ALPES-PROVENCE -0.01313 0.74527 -0.01274 0.71620 CRCAM BRIE PICARDIE -0.07667 0.02856 -0.01212 0.73011 CRCAM ILLE VILAINE 0.03506 0.21756** -0.02191 0.63896 CRCAM LOIRE HAUTE LOIRE -0.04462 0.21399 0.03486 0.36695 CRCAM MORBIHAN SA -0.02969 0.64775 0.12602 0.03605**

CRCAM NORD -0.00917 0.83851 0.07266 0.13769

(20)

CRCAM PARIS D'ILE FRANCE -0.03980 0.30288 0.04697 0.16201 CRCAM SUD RHONE ALPES -0.01775 0.51535 -0.10312 0.00822*** CRCAM TOURAINE POITOU -0.00722 0.86310 -0.01229 0.82751 CREDIT AGRICOLE DU LANGUEDOC -0.06559 0.12550 -0.02520 0.61643 CREDIT AGRICOLE SA 1 0.03375 0.39958 0.08918 0.10440 CREDIT AGRICOLE SA 2 0.04364 0.41648 0.09397 0.16332 CREDIT AGRICOLE SA 3 0.03461 0.44302 0.09121 0.10302 CREDIT AGRICOLE SA 4 0.03656 0.38349 0.09023 0.11201

LOCINDUS 0.13936 0.03274** N/A N/A

SOCIETE GENERALE GROUP 1 0.04370 0.39019 0.07145 0.28188 SOCIETE GENERALE GROUP 2 0.04724 0.38558 0.07561 0.24209 SOCIETE GENERALE GROUP 3 0.04786 0.37973 0.07503 0.25839 SOCIETE GENERALE GROUP 4 0.04482 0.50529 0.06066 0.46687

NATIXIS SA 0.00999 0.85466 0.08437 0.17573

AXA BANQUE 0.01208 0.85637 0.03043 0.34560

UNION FINANCIERE DE FRANCE

BANQUE SA 0.00338 0.94913 -0.03975 0.60791

ROTHSCHILD&CO 0.06026 0.30102 0.07397 0.32321

CAAR of French banks 0.01035 0.23018 0.04408 0.00001***

Looking at the plot of the average of abnormal returns (AARs) across all banks within the event window of the second announcement date, there seems to have a small increase in AARs after the announcement was made (from day 0 to day1). However, the significant CAAR that is discussed above more likely to come from a big jump in abnormal returns between day 6 and & after the announcement date (Figure 1). The AAR went from -0.001 in day 6 to 0.03 in day 7.

(21)

Figure 1: AARs plot across the event window of Event 2

To test if the significance of CAAR in the second event is due to the spike of abnormal return in day 7 as mentioned above, a robustness check is conducted with event window reduces from 21 days to 11 days, meaning 5 days before and after the announcement date. CAARs reduces from 0.045 to 0.026 but it is still statistically significant at 1% level (see results in Appendix 3).

Therefore, for the first hypothesis of this paper, there is no significant abnormal returns of French banks around the first announcement date. While on the second announcement date, there is a significant increase in market returns of French banks (by 0.045) due to the raise in CCyB rate to 0.5%.

4.2 Testing the second hypothesis

Event study analysis is carried out in similar manner to the stock returns of German banks during the two announcement dates. Looking at the CAARs of German banks across two events, the CAAR around the second announcement date is statistically significant at 5% level with the value of 0.06 (See Appendix 2 for results). Therefore, there seems to be other external factors that also affect German banks’ returns during the same period. These external factors might also affect French banks as well. Therefore, to explicitly test the impact of CCyB on French banks’ returns, we carry out linear regression of CARs between two countries.

(22)

Table 3: Regression on CARs across German and French banks

This table describes the regression of CARs of both French and German banks on France variable with robust standard errors. The three level of statistical significance are tested on, * represents 10%, ** represents 5% and * represents 1%.

Estimate Standard error p-value

Event 1 France -0.015268 0.025754 0.5571

Constant 0.022000 0.023773 0.3611

Event 2 France -0.015266 0.021057 0.473140 Constant 0.060437 0.017642 0.001547**

The regression results show that, surprisingly, in both events, the coefficient of Country variable is negative (Table 2). This means that, in both cases, CARs of French banks are smaller than CARs of German banks. However, none of the coefficient is significant. Therefore, null hypothesis is failed to be rejected. This means there is no statistical difference in the abnormal returns of the French and German banks around both announcement dates. The significant abnormal returns of French banks around the second announcement date we found in the previous section seems to happen to German banks as well. Therefore, the abnormal returns found in French banks might not be due to the impact of CCyB policy but just from the fluctuation of the economy in general.

4.3 Testing for the relationship between banks’ size, capital ratio prior to the

announcements and CARs of banks

First, as explained in previous section of Methodology, we would want to test the theory if our two independent variables: size and capital ratio are correlated. The result of the Fisher’s Exact test shows the p-value for 2017 data to be 1 and 0.6193 for 2018. In both cases, the p-values are too high to reject null hypothesis. This evidence is contrasted with previous literatures by Berger et.al. (2008) and Coffinet et.al. (2012) which proven that larger banks tend to hold less capital, hence having lower capital ratio. It is important to take note here that the tests have a small sample size (10 for 2017 data and 15 for 2018 data), hence the statistical power of this result is relatively low. Therefore, if the study is performed in a larger sample group, we might get a different answer. However, for now, we would assume that there is no correlation between two independent variables in this specific study.

The results of the two separate regressions of CARs of French banks on Large variable (represents the size factor) and CARs on High cap variable (measuring capital ratio factor) are presented in Table 4 and 5 respectively. Looking at Table 4, the coefficient of Large variable in both years are positive (0.000 and 0.032), showing that large banks are more affected by the announcement of CCyB compared to smaller banks. However, since the p-values are relatively high (0.996 and 0.196), these coefficients are

(23)

not statistically significant and hence there is no difference between large banks’ and small banks’ abnormal returns around both events.

Table 4: Regression of the size of banks on CARs

This table describes the regression of CARs of French banks on Large variable with robust standard errors. The three level of statistical significance are tested on, * represents 10%, ** represents 5% and * represents 1%.

Event 1 Large 0.00009 0.01905 0.99610

Constant -0.00749 0.01149 0.51050

Event 2 Large 0.03180 0.02360 0.19550

Constant 0.01759 0.01799 0.34170

The result in Table 5 shows that for both events, the coefficients of High cap are negative (-0.012 and -0.025). This means that banks with higher capital ratio prior to the announcements in general

experience smaller abnormal returns around the announcement dates. However, since both

coefficients are not significant (p-values are 0.625 and 0.408), we cannot reject the null hypothesis, hence there is no significant difference in the abnormal returns that banks with different capital ratio experience.

Table 5: Regression of the capital ratio of banks on CARs

This table describes the regression of CARs of French banks on High cap variable with robust standard errors. The three level of statistical significance are tested on, * represents 10%, ** represents 5% and * represents 1%.

Event 1 High cap -0.01175 0.02311 0.62502

Constant 0.01687 0.00615 0.02538*

Event 2 High cap -0.02534 0.02966 0.40835

(24)

V. Discussions and Conclusion

The results of event studies on the two announcement dates show that there is no significant impact from the introduction of the CCyB on stock returns of French banks. As the second announcement of increasing CCyB rates from 0.25% to 0.5%, even though a significant positive impact on the stock returns is found (with CAAR equals to 0.045) when running regression of these abnormal returns and the abnormal returns of German banks, we found no significant different between the two. Therefore, despite the abnormal returns, it might the result of other economic events during the same period, since Germany who has a similar economic structure with France, also experienced similar abnormal returns as the French banks. This study also further examined if the size of banks and the capital ratio level of banks prior to the announcements affect the French banks’ abnormal returns around the announcement dates. However, the results came back negative, as we did not find any significant difference in abnormal returns between large banks and small banks, or banks with high capital ratios and banks with low capital ratios. In conclusion, this study has not found a significant impact of the CCyB policy on market value of banks in France. However, it is important to consider some factors may have contaminated the study’s data and affected the results of the test statistics.

Firstly, an important assumption is made under event study methodology is that the announcement must be a surprise event in which no one in the market, except insiders, know about beforehand. However, this assumption may not be valid under the July 1st_{, 2018 case. The CCyB policy has been introduced}

in the Basel III framework and implemented by many other EU countries, starting with Norway in December 2013, before France announced to implement the policy into their banks for the first time in December 2015 at 0% (European Systemic Risk Board, 2020). Therefore, it might not be that surprising for French banks to receive the news that the rate would be increase to 0.25% in 2018 and to 0.5% in 2019. It is difficult to tell how much of the stock returns reactions were affected by the information prior to the announcement date. One way to cross-check the impact of announcement dates on the stock returns is to look at the Net Income and Pro-forma Net Income of French banks post-announcement date. If the Net Income decreases in the period following the raise in CCyB rate, then banks expected to have a decrease in profit and hence future value. Therefore, even though the abnormal returns detected within the event window are not statistically significant, the CCyB may still have a negative impact on the future value of French banks. To explain the lack of significant abnormal returns detected within event window, the stock returns might adjust to the news gradually prior to the announcement dates. Being able to analyze French firms’ profits would provide much more insightful conclusion to our hypotheses. However, due to the lack of availability in accounting reports from French banks from 2018 to 2020, it is not possible to test for this, which could be a potential avenue for future research. Reflecting on the study’s methodology, other than the assumption regarding announcement date can be problematic, the method to estimate expected returns of the event study can be potentially improved.

(25)

As explained previously, we used the “market model” to calculate the expected returns and then calculating the abnormal returns around the announcement dates. However, to perform “market model”, we make the underlying assumption that that the beta of the banks (𝛽̂𝑖) within the CAPM formula of (𝛼̂𝑖+ 𝛽̂𝑖𝑅𝑚) remains constant throughout the whole event window. However this might not be the case, since as mentioned in the theoretical framework, bank risk should decrease with the implementation of CCyB policy. This means the expected returns calculated might be overestimated, leading to underestimation of abnormal returns. Therefore, in future research an additional study of the bank risks pre- and post-announcement dates should be conducted, adjusting the abnormal returns accordingly. Furthermore, it would also be interesting to examine the relationship between the change in stock price and risk of banks. According to Baker and Wurgler (2015), an increase in capital requirement, which is aimed to reduce the risk of banks, can cause “low-risk anomaly”– a case when stocks with lower risks turn out be have higher returns. This however, seems to go against the long-standing belief of CAPM-model, as when the bank risks decrease so does its stock return.

Lastly, the data’s small sample size of is a significant limitation to the study. There were only 29 French banks and 13 German banks are studied. The abnormal returns of French banks might not be insignificant as found, but the abnormality is just too small to be detected in such a small sample. Therefore, there needs to be a follow-up study that carries out the same analysis, but at a larger scale to have more precise results.

Furthermore, in this study, we assumed that only publicly traded banks are affected by the CCyB policy. However in reality, there are also private banks which were affected by the change in capital requirements as well. Therefore, the study suffers from selection bias, as only publicly traded banks were examined. Since these privately held banks do not have stock price to reflect their future profitability, and how it changes after the announcement of CCyB, future research needs to look into other methods to study the impact of CCyB on the profitability of these banks.

Another suggestion for future research is comparing the impact of CCyB policy on the market value between different types of banks in the French market. In the French banking system, we can divide banks into three types as Coffinet, Coudert, Pop, and Pouvelle (2012) suggested in their study: “(1) commercial banks; (2) mutual, saving banks and credit cooperatives; and (3) financial and investment firms”. By comparing the abnormal returns of French banks in these three categories, we can have more insight into to what extent different types of banks react to the raise in CCyB rates. This would further contribute to the literature of how banks react to the new capital requirement policy.

(26)

References

Armitage, S. (1995). Event study methods and evidence on their performance. Journal of economic surveys, 9(1), 25-52.

Baker, M. and Wurgler, J. (2015): Do strict capital requirements raise the cost of capital? Bank

regulation, capital structure and the low risk anomaly, American Economic Review: Papers and

Proceedings 105, 315–320

Basten, C. (2020). Higher Bank Capital Requirements and Mortgage Pricing: Evidence from the

Counter-Cyclical Capital Buffer. Review of Finance, 24(2), 453-495.

Behn, M., Haselmann, R., & Wachtel, P. (2016). Procyclical capital regulation and lending. The Journal of Finance, 71(2), 919-956.

Benes, J., & Kumhof, M. (2015). Risky bank lending and countercyclical capital buffers. Journal of Economic Dynamics and Control, 58, 58-80.

Berger, A. N., DeYoung, R., Flannery, M. J., Lee, D., & Öztekin, Ö. (2008). How do large banking

organizations manage their capital ratios?. Journal of Financial Services Research, 34(2-3), 123-149.

Beyer, A., Nicoletti, G., Papadopoulou, N., Papsdorf, P., Rünstler, G., Schwarz, C., ... & Vergote, O. (2017). The transmission channels of monetary, macro-and microprudential policies and their

interrelations (No. 191). ECB Occasional Paper.

Bundesbank. (2020,03,18). Statement on the countercyclical capital buffer by the German Financial

Stability Committee. https://www.bundesbank.de/en/tasks/topics/statement-on-the-countercyclical-capital-buffer-by-the-german-financial-stability-committee-828822

Cerutti, E., Claessens, S., & Laeven, L. (2017). The use and effectiveness ofmacroprudential policies:

New evidence. Journal of Financial Stability, 28, 203-224.

Coffinet, J., Coudert, V., Pop, A., & Pouvelle, C. (2012). Two-way interplays between capital buffers

and credit growth: Evidence from French banks. Journal of international financial markets, institutions

and money, 22(5), 1110-1125.

Coutts, J. A., Mills, T. C., & Roberts, J. (1994). The market model and the event study method: a

synthesis of the econometric criticisms. International Review of Financial Analysis, 3(2), 149-171.

De Nicoló, M. G., Favara, G., & Ratnovski, L. (2012). Externalities and macroprudential policy. International Monetary Fund.

(27)

European Central Bank. (2019,3,27). Shelter from the storm: recent countercyclical capital buffer

(CCyB) decisions. https://www.ecb.europa.eu/pub/financial-stability/macroprudential-bulletin/html/ecb.mpbu201903_04~b8ad0c67e9.en.html#toc1

European Systemic Risk Board. (2020,04,17). Countercyclical capital buffer.

https://www.esrb.europa.eu/national_policy/ccb/html/index.en.html

Fama, E. F. (1991). Efficient market hypothesis. The Journal of Finance, 46, 383-417.

Federal Financial Supervisory Authority. (2020,04,22). Countercyclical capital buffer.

https://www.bafin.de/EN/Aufsicht/BankenFinanzdienstleister/Eigenmittelanforderungen/Kapitalpuffe r/antizyklischer_kapitalpuffer_node_en.html

Gropp, R., Mosk, T., Ongena, S., & Wix, C. (2019). Banks response to higher capital requirements:

Evidence from a quasi-natural experiment. The Review of Financial Studies, 32(1), 266-299.

Haut Conseil de Stabilite Financiere (HCSF). (2020,3,18). Communique de presse.

https://www.economie.gouv.fr/files/files/directions_services/hcsf/HCSF_20200318_Communique_de _presse_de_seance.pdf

Jiménez, G., Ongena, S., Peydró, J. L., & Saurina, J. (2017). Macroprudentialpolicy, countercyclical

bank capital buffers, and credit supply: evidence from the Spanish dynamic provisioning experiments.

Journal of Political Economy, 125(6), 2126-2177.

Modigliani, F. and Miller, M. (1958): The cost of capital, corporation finance and the theory of

investment, American Economic Review 48, 261–297.

Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have

informationthat investors do not have (No. w1396). National Bureau of Economic Research.

Park, N. K. (2004). A guide to using event study methods in multi‐country settings. Strategic Management Journal, 25(7), 655-668.

Shim, J. (2013). Bank capital buffer and portfolio risk: The influence of business cycle and revenue

diversification. Journal of Banking & Finance, 37(3), 761-772.

Vandenbussche, J., Vogel, U., & Detragiache, E. (2015). Macroprudential policies and housing prices:

A new database and empirical evidence for Central, Eastern, and Southeastern Europe. Journal of

(28)

Appendix

Appendix 1: Categorizing French banks in 2017 and 2018 by their size and capital

ratios

2017 2018

Capital ratio High SOCIETE GENERALE SA NATIXIS SA

CRCAM ATLANTIQUE VENDEE

BNP PARIBAS

CRCAM NORD SOCIETE GENERALE SA

ROTHSCHILD & CO CRCAM ALPES-PROVENCE

CRCAM PARIS D'ILE FRANCE CRCAM TOULOUSE 31 CCI

CRCAM TOURAINE POITOU

CRCAM ILLE VILAINE

Low AXA BANQUE CRCAM MORBIHAN SA

UNION FINANCIERE DE FRANCE BANQUE SA

CRCAM BRIE PICARDIE

NATIXIS SA CRCAM PARIS D'ILE FRANCE

BNP PARIBAS CREDIT AGRICOLE DU

LANGUEDOC

CREDIT AGRICOLE S.A. CRCAM ATLANTIQUE VENDEE

CRCAM NORD

CREDIT AGRICOLE S.A.

CRCAM LOIRE HAUTE

LOIRE

Bank’ size Large BNP PARIBAS BNP PARIBAS

CREDIT AGRICOLE S.A. CREDIT AGRICOLE S.A.

SOCIETE GENERALE SA SOCIETE GENERALE SA

AXA BANQUE AXA BANQUE

NATIXIS SA NATIXIS SA

CRCAM PARIS D'ILE FRANCE CRCAM PARIS D'ILE FRANCE

CRCAM NORD CRCAM NORD

CRCAM BRIE PICARDIE CRCAM BRIE PICARDIE

CREDIT AGRICOLE DU LANGUEDOC

Small CRCAM ATLANTIQUE VENDEE

CRCAM ATLANTIQUE VENDEE

CRCAM ALPES-PROVENCE CRCAM NORMANDIE SEINE

CRCAM NORMANDIE SEINE ROTHSCHILD & CO

CRCAM TOURAINE POITOU CRCAM ILLE VILAINE

CRCAM ILLE VILAINE CRCAM TOURAINE POITOU

ROTHSCHILD & CO CRCAM LOIRE HAUTE LOIRE

CRCAM LOIRE HAUTE LOIRE CRCAM TOULOUSE 31 CCI

CRCAM TOULOUSE 31 CCI CRCAM MORBIHAN SA

CRCAM MORBIHAN SA CRCAM ALPES-PROVENCE

UNION FINANCIERE DE FRANCE BANQUE SA

(29)

Appendix 2: Contingency table of the expected frequencies for banks’ size and banks’

capital ratio

Size Capital ratio Small Large 2017 Low 1.5 3.5 High 1.5 3.5 2018 Low 3.7 4.3 High 3.3 3.7

Appendix 3

: CAR’s broken down by banks of French around the second announcement date

This table represents the robustness check of the CAAR around the second announcement date with the event window reduces from 21 days to 11 days. There are in total 26 banks being observed. The last row represents the average CARs across all banks. * stands for a significance level of 10%, ** for 5% level and *** for 1% level.

Event 2

Banks Estimate p.value

BNP PARIBAS 1 0.05152 0.18854

BNP PARIBAS 2 0.05201 0.17943

BNP PARIBAS 3 0.05402 0.17965

BNP PARIBAS 4 0.05207 0.19128

CA TOULOUSE 31 CCI -0.01975 0.43822 CAISSE REG DE CRE AGRI MUT 0.02420 0.34287 CRCAM ALPES-PROVENCE -0.00668 0.80642 CRCAM BRIE PICARDIE -0.02242 0.37251 CRCAM ILLE VILAINE -0.00494 0.88632 CRCAM LOIRE HAUTE LOIRE -0.01153 0.69988 CRCAM MORBIHAN SA 0.07242 0.09872*

CRCAM NORD 0.08766 0.02094**

CRCAM NORMANDIE SEINE -0.01360 0.51689 CRCAM PARIS D'ILE FRANCE 0.00666 0.81491 CRCAM SUD RHONE ALPES -0.08359 0.00450*** CRCAM TOURAINE POITOU 0.01929 0.65045 CREDIT AGRICOLE DU

LANGUEDOC -0.02700 0.46762

CREDIT AGRICOLE SA 1 0.05914 0.12601 CREDIT AGRICOLE SA 2 0.06255 0.18832

(30)

CREDIT AGRICOLE SA 3 0.05985 0.12499 CREDIT AGRICOLE SA 4 0.05694 0.15031

LOCINDUS N/A N/A

SOCIETE GENERALE GROUP 1 0.02918 0.54918 SOCIETE GENERALE GROUP 2 0.03518 0.45860 SOCIETE GENERALE GROUP 3 0.03444 0.47942 SOCIETE GENERALE GROUP 4 0.04025 0.51775

NATIXIS SA 0.04906 0.27111

AXA BANQUE 0.01691 0.45803

UNION FINANCIERE DE FRANCE

BANQUE SA -0.01609 0.77582

ROTHSCHILD&CO 0.10989 0.03912** CAAR of French banks 0.02647 0.00034***

Appendix 4:

CAR’s broken down by banks of Germany

This table represents the cumulative abnormal returns by each German bank in each event. Event 1 represents the first announcement date on July 2nd_{, 2018. Event 2 represents the second announcement}

date on April 3rd_{, 2019. There are in total 13 banks being observed. The last row represents the average}

CARs across all banks. The * stands for a significance level of 10%, ** for significance level of 5% and *** for the significance level of 1%.

Event 1 Event 2

Banks Estimate p.value Estimate p.value

COMMERZBANK_1 0.07747 0.11402 0.10104 0.08668* COMMERZBANK_2 0.07149 0.15317 0.10195 0.07574* COMMERZBANK_3 0.01031 0.84818 -0.02132 0.79535 COMMERZBANK_4 0.05774 0.46096 0.10539 0.28773 HSBC TRINKAUS & BURKHARDT -0.00213 0.98199 0.13559 0.27159 DEUTSCHE PFANDBRIEFBANK AG_1 0.04390 0.58296 0.10984 0.27075 DEUTSCHE PFANDBRIEFBANK AG_2 0.05135 0.50736 0.04019 0.60126 COMDIRECT BANK AG 0.04238 0.50559 0.09690 0.17989 AAREAL BANK AG_1 0.04184 0.50466 0.09397 0.20339 AAREAL BANK AG_2 -0.00570 0.91070 0.05460 0.61413 UMWELTBANK AG -0.11901 0.07231* 0.02610 0.76798 MERKUR BANK KGAA -0.07863 0.20418 -0.07658 0.25221 PROCREDIT HOLDING AG & CO -0.01114 0.91758 0.01801 0.80412

(31)

Impact of counter-cyclical capital buffer on returns of banks

Bachelor Thesis: