The Effect of Countercyclical Capital
Buffers on the Riskiness of Swedish
Banks
Daria Serdiuk
11445173
University of Amsterdam
Faculty of Economics and Business
Bachelor Thesis
Magdalena Rola-Janicka
Statement of Originality
This document is written by Student Daria Serdiuk who declares to take full responsi-bility 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 com-pletion of the work, not for the contents.
UNIVERSITY OF AMSTERDAM
Abstract
Faculty of Economics and Business Bachelor Thesis
by Daria Serdiuk
Despite some evidence of potential unintended effects of countercyclical capital buffers, little research has been conducted to investigate them. Therefore, the aim of this paper is to fill out the existing gap in the literature and evaluate the impact of this regulation on banks’ riskiness in the context of Sweden. For this purpose, a difference-in-difference analysis is carried out. The findings present suggestive evidence that a 1 percentage point increase in countercyclical capital buffers on average reduces the growth of banks’ z-score by 5.56 percentage points. That is, the policy has the opposite effect than what was intended. This effect is uniform across the banks with different initial riskiness. Additionally, the analysis of cumulative abnormal returns around the date of policy announcement indicates a negative effect of this regulation on banks’ profitability.
Contents
1 Introduction 4
2 Conceptual Background 6
3 Literature Review 8
4 Research Questions and Hypotheses 11
5 Data and Methodology 14
6 Results 19
7 Discussion and Analysis 23
8 Conclusion 25
References 26
Chapter 1
Introduction
In the run-up to the global economic crisis of 2007 – 2008, banks all over the world were building-up their leverage and holding increasingly insufficient liquidity levels, accompa-nied by a reduction in the level and quality of their capital (Basel III,2011). After the crisis commenced, the question of financial sector stability and bank liquidity became more important than ever. In an attempt to “improve risk management and governance as well as strengthen banks’ transparency and disclosure” (Basel III, 2011, p. 1) and, thus, prevent the development of a similar scenario in the future, the Basel Committee issued Basel III rules in late 2010. The main goals of the regulation were ensuring ade-quate capital and liquidity levels in the banking sector as well as reducing the cyclicality of its performance.
One of the macro-prudential tools that were proposed as a measure of mitigating eco-nomic cycles in the banking sector were countercyclical capital buffers. The implemen-tation of this tool was meant to ensure that banks accumulate higher capital during good times which they would be able to use during bad times. Thus, the countercyclical aspect is reached by restricting banks from extending too much credit in periods of an economic upturn and creating resources to continue supply credit during an economic downturn.
Although countercyclical capital buffers have been used by several member states of the Basel Committee and one non-member state – Norway, for a substantial time, little research has been done to measure the effectiveness of this tool and its potential unin-tended impact on banks’ activities. However, there is some evidence that suggests that this regulation may have undesirable side-effects. For instance, Jim´enez et al. (2017), who investigated the effect of countercyclical capital buffers on the real economy in Spain, concluded that this regulation led to an unexpected shift in the banks’ portfolio towards more risky firms. Moreover, several other studies (Cerutti et al.,2017;Behn et
5
al.,2016;Vandenbussche et al.,2015) suggest that macroprudential policies can have a significant effect on the real economy such as slowing down the growth of housing prices or, in fact, inducing a credit crunch as was the case in Germany in 2008 (Behn et al.,
2016).
The goal of this paper is to look more closely at the implications of countercyclical capital buffers for the banking sector and to fill the existing gap in the literature. More specifically, this paper looks at the example of Sweden, which has been actively using countercyclical capital buffers since 2014. The main focus is on the riskiness of banks which is measured by the z-score and takes account of banks’ return on assets (ROA), their volatility as measured by the standard deviation of ROA and the Equity-to-Asset ratio. Therefore, the question that this paper aims to answer is what is the effect of an increase in countercyclical capital buffers in 2014 – 2016 on the riskiness of banks based in Sweden.
However, it is essential to remember that banks may have very different characteristics despite being headquartered in the same country. For example, they may have different preferences as to the amount and kind of risk that they decide to take on. In this case, banks can be classified into two subgroups: initially risky and initially less risky. There-fore, it is interesting to see whether the above regulation has somewhat different effects on these two groups of banks. Thus, another question that this paper investigates is whether this has been the case for Swedish banks in 2014 – 2016. Finally, countercyclical capital buffers can influence banks’ profitability and, thus, may affect banks’ abnormal returns. Therefore, this is another aspect that this paper tries to test.
In order to investigate the above questions, an event-study is carried out looking at the period from 2011 to 2017. More specifically, the difference-in-difference analysis on bank riskiness is applied to control for country-specific aspects. At the same time, the change in the abnormal returns is tested by comparing the cumulative abnormal returns of Swedish banks with the control group. For this purpose, Denmark has been chosen to be the control group since the economies of both countries are expected to follow a similar trend over time ceteris paribus. This expectation is based primarily on the facts that both Sweden and Denmark are situated close geographically and that neither of them is a member of the Euro Area. It is explored later in the paper whether Denmark is indeed a good choice.
The rest of the paper is structured as follows: Chapter 2 sets out the conceptual back-ground of this research, Chapter 3 explains the previous findings in related literature and Chapter 4 discusses the research questions and hypotheses. In Chapter 5, the data and methodology are explained. Finally, Chapter 6 presents the results of the research, while Chapter 7 provides a discussion of the outcomes. Chapter 8 concludes.
Chapter 2
Conceptual Background
The main goal of Basel III is to improve the quantity and quality of the banks’ capital base, increase risk coverage, and prevent the growth and spread of systemic risk through the banking sector (Basel III,2011). More specifically, the purpose of the countercyclical capital buffers proposed by the Basel Committee is to reduce excess credit growth in the periods of an economic upturn and the subsequent increase in the system-wide risk. In other words, the regulation aims to ensure that the banking system can maintain a stable supply of credit without compromising banks’ solvency during bad times (Committee et al.,2010).
According to the Committee et al. (2010), this goal is achieved through an increase in the cost of credit supply during the periods of accumulation of excessive levels of credit as compared to a predetermined benchmark, and a decrease in the cost of credit when the supply of it is judged to be lower than a predetermined benchmark. This mechanism is to be put in place by a relevant authority in each jurisdiction, which would impose an appropriate buffer requirement based on their judgement of the current level of system-wide risk. Any increase in the countercyclical capital buffers should be announced 12 months before the regulation becomes effective in order to allow banks to accumulate the needed amount of capital. At the same time, any decrease in the countercyclical capital buffers becomes effective immediately. The level of countercyclical capital buffers is recommended to be reviewed quarterly.
Even though countercyclical capital buffers offer a side benefit of reducing the build-up of excess credit, managing economic cycles and asset prices is not the primary role of this regulation. Rather, countercyclical capital buffers should be seen as one of the tools in the authorities’ toolkit. Therefore, it is crucial that in the process of judging the appropriateness of a change in the level of countercyclical capital buffers, the authorities pursue a more holistic approach and consider a wide range of factors besides credit and
7
GDP growth such as multiple asset prices, the ability of entities to meet their debt obligations and credit conditions (Committee et al.,2010).
According to Basel III, countercyclical capital buffers are a weighted average of the existing buffers in the state where the bank operates. It is to be applied by the authorities with respect to the banks that are located in their jurisdiction, including those that are not headquartered in their country, in order to sustain the competitiveness of the market. Furthermore, another purpose of ensuring that all banks in the jurisdiction apply the same buffers regardless of where they are headquartered is to guarantee that the banks without countercyclical capital buffers do not compensate for the reduction in the credit supply from those adhering to the regulation and, thus, do not increase the systemic risk in the sector. Having said that, if the home authorities consider countercyclical capital buffers imposed by the host country as not sufficiently high, they can require that the supervised by them banks adopt higher buffers (Committee et al.,2010).
Chapter 3
Literature Review
Several studies have been conducted in recent years aimed at assessing the effects of capital requirements in general and countercyclical capital buffers in particular. Most of this research was concentrated on determining whether these policies have any significant impact on the real economy and, if so, whether they do what they were meant to do, i.e. reduce the cyclicality of the financial sector.
It has been demonstrated that most countries use policies aimed at reducing systemic risk and that some of them lead to a reduction in the growth rate in credit and house prices. This conclusion was drawn by Cerutti et al. (2017) who conducted a study on different macroprudential policies applied in 119 countries for 2000 – 2013. The authors concentrated their attention on 12 main policies and aimed to evaluate how effective those were in reducing the procyclicality in the banking sector. Moreover, they have found evidence that macroprudential policies are more effective when the credit growth rate is high and less effective during busts as well as that there is less effect of macroprudential policies on credit developments in financially open economies and those with more sophisticated financial systems.
A large number of related research has also concentrated on estimating the impact of macroprudential policies on credit supply. Recent research suggests that capital buffers reduce the credit supply cycle with positive effects on the real economy. More specifically, the effect of such regulation in good times was found to reduce the amount of credit supply but with little effect on firm performance. In bad times, the increase in capital lowered the amount of credit extended to firms and thus had a negative effect on their survival rate. This result was documented by Jim´enez et al.(2017) after looking at the effect of countercyclical capital buffers introduced in Spain in 2000 on the real economy.
9
Similarly, capital measures such as minimum capital adequacy ratio and the maximum ratio of household loans to share capital as well as nonstandard liquidity measures like marginal reserve requirements on foreign funding and those linked to credit growth have been found to affect house prices. Vandenbussche et al. (2015), concluded this after looking at the effect of macroprudential policies on inflation and housing prices in the context of Central, Eastern and Southeastern Europe. Another research that found evidence that risk-sensitive capital regulation influences credit supply was carried out by Behn et al. (2016). The study looked at the case in Germany. It concluded that procyclical policies, such as an increase in capital charges, have a significant effect on credit extension and that the procyclical policy was a major contributor to the credit crunch of 2008.
Finally, there is evidence that suggests that mortgage origination shifts from more to less affected by countercyclical capital buffers banks. After conducting research aimed at identifying the effect of countercyclical capital buffers introduced in Switzerland in 2013 on the lending behaviour of banks and more specifically on the composition of mortgage suppliers,Basten(2020) has concluded that banks with lower capital cushions and higher mortgage specialization increased their mortgage prices by 8-9 basis points. Some researchers have also aimed to investigate the effect of macroprudential policies on the stability of the financial sector. It has been found that policies targeted at borrow-ers and countercyclical buffborrow-ers with regard to leverage and assets are rather effective at reducing system vulnerabilities in the period of an economic upswing (Claessens et al.,
2013). However, after analyzing the impact of macroprudential policies on the vulner-ability of the banking system especially in emerging economies Claessens et al. (2013) have concluded that few policies can reduce declines in banking activities during bad times.
Despite plentiful evidence pointing to the effectiveness of macroprudential policies in influencing the real economy, the question about the unintended effects of such policies remains open. For instance, banks could start extending more credit to riskier firms after the introduction of countercyclical capital buffers which was documented byJim´enez et al. (2017). Furthermore, it has been demonstrated that capital exercise banks raise their capital ratios by lowering their risk-weighted assets (i.e. the denominator) by asset shrinking and not risk reduction. After investigating the effect of a change in the required capital ratio on banks’ balance sheets, their lending behaviour, and the subsequent impact on the real economy, Gropp et al.(2019) suggested that it could be due to asymmetric information and debt overhang.
Additionally,Koehn & Santomero(1980) researched the effect of bank capital regulations on banks’ portfolio risk. They concluded that the variance of the total risk in the industry
10
increases with an increase in the bank capital requirements. The authors constructed a mathematical model and used it to show that as a result of an increase in required capital ratio, banks will reshuffle their portfolios in order to maximize their returns given a certain level of risk, which causes the overall riskiness of the bank’s portfolio to increase. The degree of reshuffling is subject to the bank’s risk-aversion. Therefore, the initially more risky banks increase the riskiness of their portfolio to a greater degree than the less risky banks. Similarly,Kim & Santomero(1988) show that imposing higher capital ratio requirements does not reduce bank riskiness; it has a contrary effect due to banks’ risk preferences. However, it is important to note that both of these papers look at the leverage ratio of banks that takes into account non-risk-weighted assets, whereas countercyclical capital buffers use risk-weighted assets.
Finally, an increase in countercyclical capital buffers may reduce the profitability of banks by forcefully changing their capital structure. After investigating the effect of involuntary equity issuance due to increased capital requirements on banks’ abnormal returns as compared to the voluntary increase in equity, Cornett & Tehranian (1994) found evidence that imposing higher capital requirements indeed gets translated into negative abnormal returns around the time of the announcement albeit these returns are less negative than in the case of voluntary equity issuance.
Chapter 4
Research Questions and
Hypotheses
The goal of this paper is to answer four research questions. The first research question is whether the increases in the level of countercyclical capital buffers in Sweden in De-cember 2014, September 2015 and June 2016 had any effect on the riskiness of banks based in Sweden. That is,
Hypothesis 1:
H0: The average increase in the z-score of banks based in Sweden was the same before and after the introduction of countercyclical capital buffers;
H1: The average increase in the z-score of banks based in Sweden was different after the introduction of countercyclical capital buffers then before.
As it has been mentioned earlier, z-score is a measure of banks’ riskiness and indicates the probability of banks failure. In this case, a higher z-score would mean a lower probability of failure and, thus, a lower risk. A lower z-score would, on the other hand, indicate a higher probability of failure and higher riskiness. Thus, if there is not enough evidence to reject the null hypothesis, it would be concluded that the introduction of countercyclical capital buffers has no effect on bank riskiness. If there is enough evidence to reject the null hypothesis and the result indicates that an increase in countercyclical capital buffers increases the banks’ z-score, it would be concluded that the policy works as intended and reduces banks’ riskiness. If the result is the opposite similarly toJim´enez et al.(2017), there would be enough evidence to conclude that the given policy has an unintended effect on banks’ riskiness and causes it to increase.
12
The second research question is whether the Swedish banks that were initially riskier were affected more than those that were initially less risky.
Hypothesis 2:
H0: The average increase in the z-score of the initially risky banks based in Sweden after the introduction of countercyclical capital buffers was the same as the average increase in the z-score of the initially less risky banks;
H1: The average increase in the z-score of the initially risky banks based in Sweden after the introduction of countercyclical capital buffers was lower than the average increase in the z-score of the initially less risky banks.
This hypothesis is motivated by Koehn & Santomero(1980) findings suggesting that as a result of the introduction of countercyclical capital buffers the initially risky banks could become even riskier compared to the initially less risky banks. In this case, if there is not enough evidence to reject the null hypothesis, it can be concluded that the relative riskiness of banks remained the same after the increase in countercyclical capital buffers as it was prior to the policy introduction. If, on the contrary, there is enough evidence to reject the null hypothesis, it can be concluded that the relative riskiness of the initially risky banks has increased after the increase in countercyclical capital buffers and the result is similar to Koehn & Santomero (1980).
Next, the effect of the introduction of countercyclical capital buffers on the banks’ ab-normal returns is investigated. Therefore, the third research question is whether the implementation of countercyclical capital buffers in Sweden had any effect on the banks’ abnormal returns.
Hypothesis 3:
H0: Banks based in Sweden experienced no cumulative abnormal returns around the time of countercyclical capital buffer increase announcement;
H1: Banks based in Sweden experienced significant cumulative abnormal returns around the time of countercyclical capital buffer increase announcement.
This hypothesis aims to test whether the introduction of countercyclical capital buffers had an effect on banks’ abnormal returns similarly to Cornett & Tehranian (1994). In this case, the choice of the model that is used to estimate abnormal returns is critical. This is because a model that is not good at explaining stock returns may produce abnormal returns when there are none. Thus, there is essentially a dual hypothesis of whether the model is good at explaining stock returns and whether abnormal returns exist. However, assuming that the model used to estimate abnormal returns is correct,
13
if there is not enough evidence to reject the null hypothesis, it can be concluded that the policy has no effect on banks’ profitability. If, on the other hand, there is enough evidence to reject the null hypothesis, this could indicate that the introduction of countercyclical capital buffers affects banks’ profitability. In this case, positive abnormal returns would signal a favourable effect on profitability, while negative abnormal returns would indicate a negative effect.
Finally, research question four is whether the change in the abnormal returns was differ-ent for the initially more risky banks as compared to the initially less risky banks based in Sweden.
Hypothesis 4:
H0: The cumulative abnormal returns of the initially risky banks based in Sweden around the time of countercyclical capital buffers announcement were the same as the cumulative abnormal returns of the initially less risky banks;
H1: The cumulative abnormal returns of the initially risky banks based in Sweden around the time of countercyclical capital buffers announcement were not the same as the cu-mulative abnormal returns of the initially less risky banks.
The goal of this hypothesis is to analyze more in-depth the effect of countercyclical capital buffers on banks’ abnormal returns by looking at the two subsamples of banks according to their initial riskiness. If there is not enough evidence to reject the null hypothesis, it can be concluded that the effect of countercyclical capital buffers of banks’ profitability was the same for both the initially risky and initially less risky banks. If there is enough evidence to reject the null hypothesis, it can be concluded that the relative effect of the new policy of the profitability of the two subsamples was different.
Chapter 5
Data and Methodology
For this research, a combination of resources has been used. In order to answer the first two of the research questions, namely, what is the effect of countercyclical capital buffers on bank riskiness (Model 1 and Model 2 ), an initial smaller dataset was used (further Dataset 1 ). Later, more information is added to investigate the two latter questions, i.e. what is the effect of countercyclical capital buffers on bank’s abnormal returns (Model 3 and Model 4 ). It will further be referred to as Dataset 2. To construct Dataset 1, Osiris database from theVan Dijk (2015) has been used to get information about the banks’ net income, total assets and total equity on an annual basis. In total, a sample of 30 banks has been created with 24 banks being headquartered in Denmark and 6 banks in Sweden. The information was obtained for a period of 7 years, from 2011 to 2017. The total number of observations amounted to 210 with 168 observations attributed to Denmark and 42 to Sweden. Total equity and net income have been winsorized at 5 and 95 levels to account for possible outliers. Additional summary statistics can be found in
Table 5.1.
Obs Mean Std. Dev. Min Max
Total Assets 210 405,059,574 907,097,636 912,530 3,539,528,000 Total Equity 210 19,672,890 41,543,827 79,574 133,572,000 Net Income 210 1,979,875 4,499,875 -195,581 15,184,000
Table 5.1: Descriptive Statistics 1
In order to investigate research questions three and four, Dataset 2 was created by adding more information to Dataset 1. First of all, the realized returns of listed banks have been added for the period from 2011 to 2017 (FactSet, 2020). As a result, the sample of banks was reduced to 28, of which 6 are Swedish, and 22 are Danish with 48,211 observations in total. Then, the 10-year government annual bond yields have
15
been used as an approximation of the risk-free rates for both countries (Denmark,2011). Additionally, OMX Stockholm 30 and OMX Copenhagen 20 indices have been used to approximate realized market returns for Sweden and Denmark, respectively.
For this research, z-score has been used as a measure of a bank’s risk, which essentially shows the bank’s probability of failure by estimating the number of standard deviations below the mean by which the bank’s profits have to fall before the value of its capital goes to zero. Thus, the lower the z-score of a given bank, the higher the probability of it failing. Z-score is calculated as
z-score = ROA +
Total Equity Total Assets
σROA
where ROA has been calculated by dividing a bank’s net income for a specific year by total assets in that year, and σROAis the standard deviation of ROA. Similarly toYeyati
& Micco(2007), σROAhas been calculated on a two-year rolling basis. In order to reduce
the number of missing values for the test for parallel trends only, the standard deviation of each bank in the initial year was substituted for the rolling standard deviation in the following year. Finally, the annual change in the z-score per bank was calculated by subtracting the z-score of the previous period from the current z-score and dividing the result by the z-score of the previous period i.e.:
Change in z-score = z-scoren− z-scoren−1 z-scoren−1
Table 5.2presents descriptive statistics of the generated variables.
Obs Mean Std. Dev. Min Max
ROA 210 0.0060899 0.0081288 -0.042492 0.0228186
Std. Dev. ROA 180 0.0029159 0.0046964 0.0000124 0.0424902
z-score 180 215.6072 547.1498 2.25096 5422.193
Change in z-score 150 2.908182 13.87499 -0.9817486 163.5678
Table 5.2: Descriptive Statistics 2
Next, all Swedish banks were sorted into five bins according to the average level of their z-score from 2011 to 2013 inclusive. Thus, banks that end up in the bins with the lowest average z-score can be classified as initially risky – risk category 1 and 2 – and banks that are assigned to the bins with the highest average z-score can be classified as initially less risky – risk category 4 and 5 (seeTable 5.3). Hence, a dummy variable is generated that takes on a value of 1 if a given bank was initially risky and 0 otherwise.
16
Mean Std. Dev. No. Risk Category 1 96.441705 78.028731 2 Risk Category 2 138.92322 101.54841 1 Risk Category 3 145.43845 106.90958 1 Risk Category 4 267.34668 285.585 1 Risk Category 5 272.14421 351.76493 1
Table 5.3: Summary Statistics of the Z-score of Swedish Banks by Risk Category
Finally, the number of control variables in the regressions has been limited to one in order not to reduce the significance of the resulting coefficients in a small sample. Similarly toBoyd et al.(2006), the Herfindahl-Hirschman Index was used to approximate the size of firms within a given industry and, as a consequence, the level of market competition. The index has been calculated by first finding the size of the industry approximated by the sum of total assets across all banks for a given year and country. Then, the individual weight of a given bank in a given year was calculated by dividing the total assets of this bank for a given year by the industry size. Finally, the Herfindahl-Hirschman Index is computed as a sum of the squared weights of each bank for a given year and country. Two models have been used to investigate the effect of an increase of countercyclical capital buffers on the banks’ riskiness. Model 1 is aimed at determining the overall impact of this new regulation by utilizing a difference-in-difference analysis on panel data. More specifically, the change in the z-score is the dependent variable, and the independent variables are the country the bank is headquartered in, the level of the countercyclical capital buffers in a given year and the interaction variable between the country and the level of the countercyclical capital buffers in that year. The Herfindahl-Hirschman Index is used as a control variable.
Model 1:
Change in z-scorei,t = β0+ β1yeart+ β2CCyBt+ β3SEi+ β4CCyBt∗ SEi+ β5HHIi,t+ εi,t
where i denotes a bank, t shows the year, change in z-score is the annual change in bank’s riskiness, year is the year in consideration, CCyB is the level of countercyclical capital buffers in a given year, SE is a dummy variable that takes on a value of 1 if the bank in question is headquartered in Sweden and 0 otherwise, CCyB*SE is the interaction variable between the level of the countercyclical capital buffer in the year under consideration and the country where the bank is headquartered, and HHI is the estimation of the Herfindahl-Hirschman Index.
17
Hypothesis 0: β4 = 0
Hypothesis 1: β4 6= 0
The purpose of Model 2 is to broaden the analysis provided by Model 1. This is done by means of difference-in-difference analysis on panel data for banks that are head-quartered in Sweden. Model 2 allows for determining not only whether an increase in countercyclical capital buffers had an effect on bank riskiness but also whether this effect was different for the initially more risky banks as compared to the initially less risky banks based in Sweden.
Model 2:
Change in z-scorei,t= β0+ β1yeart+ β2CCyBt+ β3Riskyi+
β4CCyBt∗ Riskyi+ β5HHIi,t+ εi,t
where Risky is a dummy variable that is equal to 1 if the bank in question is an initially risky bank and 0 otherwise and CCyB*Risky is the interaction variable between the level of countercyclical capital buffer and the dummy variable representing the initial riskiness of the bank.
The coefficient of the main interest is β4 and the hypotheses are:
Hypothesis 0: β4 = 0
Hypothesis 1: β4 < 0
In order to answer the two latter research questions, the Capital Asset Pricing Model (CAPM) was used to estimate stock betas over the period from January 1st 2011 to August 1st 2014 and used to calculate normal returns for later periods. Next, these returns were used to calculate the abnormal returns of banks’ stocks by subtracting the normal returns predicted by CAPM from the realized returns in the event window from -5 days to +5 days around the date of each announcement of countercyclical capital buffer increase, i.e. around December 10th 2014, September 8th 2015 and June 20th 2016. Then, the cumulative abnormal returns were computed by calculating the sum of abnormal returns for each event window per bank.
Again, two models have been utilized to estimate the effect of the change in the level of countercyclical capital buffers on banks’ cumulative abnormal returns. Similarly to Model 1 and Model 2, Model 3 measures the overall effect of the new policy on the abnormal returns of Swedish banks. In contrast, Model 4 does so for initially more risky banks in comparison to initially less risky banks based in Sweden.
18
Model 3:
CARi,e= β0+ β1SEi+ εi,e
where CAR stands for cumulative abnormal returns and e indicates each event window. The coefficient of the main interest is β1, and the hypotheses are:
Hypothesis 0: β1 = 0
Hypothesis 1: β1 6= 0
Model 4:
CARi,e= β0+ β1Riskyi+ εi,e
This model is applied to Swedish banks only.
The coefficient of the main interest is β1 and the hypotheses are:
Hypothesis 0: β1 = 0
Hypothesis 1: β1 6= 0
Finally, Model c was constructed as a robustness check. It is identical to Model 1 with the exception that a lead CCyB was used instead. That is, the change in the bank riskiness is estimated with the level of countercyclical capital buffers in the following year:
Model c:
Change in z-scorei,t = β0+ β1yeart+ β2lead CCyBt+ β3SEi+
β4lead CCyBt∗ SEi+ β5HHIi,t+ εi,t
where lead CCyB is the level of CCyB at t + 1 and lead CCyB*SE is the interaction variable between the level of countercyclical capital buffers in the following year and the dummy variable indicating the country. In this case, a statistically significant coefficient β4 would indicate the existence of the endogeneity problem, i.e. that an increase in
countercyclical capital buffers could have been caused by an increase in the overall riskiness of banks. The hypotheses are:
Hypothesis 0: β4 = 0
Chapter 6
Results
As it has been discussed earlier, to test Model 1, this paper uses a difference-in-difference analysis where Sweden is the treatment group, and Denmark was chosen to be the control group. The choice of the control group was decided by the fact that both countries are located close geographically, and neither of them is part of the Euro Area. Nonetheless, it is crucial to evaluate how fruitful is the comparison of Sweden and Denmark before discussing the results of the four models presented earlier.
In order to investigate whether Denmark is indeed a good choice for the control group, a test for parallel trends was used. For this purpose, a difference-in-difference analysis was carried out for the period prior to the introduction of countercyclical capital buffers. More precisely, Model a looks at whether the change in the banks’ riskiness was the same in Sweden and Denmark before the new regulation:
Model a:
Change in z-scorei,t = β0+ β1timet+ β2SEi+ β3timet∗ SEi+ εi,t
where time takes on the value of 1 if it is the year 2013 and the value of 0 if it is the year 2012. If the two countries followed a similar trend in the years before the regulation, then the coefficient β3 should not be significantly different from zero. From Table 6.1,
β3 is not significantly different from zero.
Additionally, it is necessary to test whether both the initially risky and initially less risky banks in Sweden also followed the same trend before the introduction of countercyclical capital buffers in order to fulfil the assumptions of Model 2. Model b represents such a test for parallel trends, i.e.:
20
Model b:
Change in z-scorei,t = β0+ β1timet+ β2Riskyi+ β3timet∗ Riskyi+ εi,t
This model is only applied to banks based in Sweden and should present a result in which β3 is not significantly different from zero in order to point to the existence of
parallel trends. The coefficient of interest is not significantly different from zero (see
Table 6.1).
Table 6.1: Test for Parallel Trends
Model (a) Model (b)
time 7.8341 10.2187 (6.8759) (10.0928) SE -0.1123 (0.1044) time*SE 0.1246 (8.0145) Risky 0.0280 (0.1469) time*Risky -1.7845 (12.4052) cons 0.1323 -0.0233 (0.0834) (0.0448) F-statistic 2.059 0.812 Observations 60 10
Note. The dependent variable in Model a and b is the change in the z-score, which is a measure of the bank’s riskiness. The regressions include the following independent variables: time, a dummy for whether a country is Sweden or not, the interaction variable between time and country dummy and whether the bank is initially risky or not. Robust standard errors in parentheses.
∗
p < 0.1,∗∗p < 0.05,∗∗∗p < 0.01
Moving on to Model 1, the coefficient of the interaction variable CCyB*SE is negative and statistically significant at the 5% level. Thus, there is enough evidence to reject the null hypothesis. As to Model 2, the coefficient of interest (CCyB*Risky) is not statis-tically significant; therefore, there is not enough evidence to reject the null hypothesis (Table 6.2).
From Table 6.3, when estimating Model 3, the coefficient of SE is negative and sta-tistically significant at the 1% level. Thus, there is enough evidence to reject the null hypothesis. Similarly, the estimation of Model 4 produced a positive and statistically significant coefficient of Risky at the 1% level, so there is enough evidence to reject the null hypothesis.
21
Table 6.2: Effect of Countercyclical Capital Buffers on the Riskiness of Swedish Banks
Model (1) Model (2) year 0.7700 1.2284 (1.0078) (2.1867) CCyB 49.9614 -690.7238 (306.4774) (760.6180) SE 48.0793 (44.1529) CCyB*SE -555.8269∗∗ (217.4827) HHI 116.6912 226.1449 (123.0222) (392.9257) Risky -1.6678 (8.9751) CCyB*Risky 141.3180 (517.9971) cons -1623.608 -2527.675 (2012.7693) (4305.5962) F-statistic 2.398∗∗ . Observations 150 25
Note. The dependent variable in Model 1 and 2 is the change in the z-score, which is a measure of the bank’s riskiness. The regressions include the following independent variables: year, the level of CCyB, a dummy for whether a country is Sweden or not, the interaction variable between the level of CCyB and country dummy and whether the bank is initially risky or not. Robust standard errors in parentheses.
∗
p < 0.1,∗∗p < 0.05,∗∗∗p < 0.01
Finally, Table 6.4shows the outcomes of Model c, which is a check for reverse causality. The coefficient of interest is negative and statistically significant at the 1% level.
22
Table 6.3: Effect of Countercyclical Capital Buffers on the Cumulative Abnormal Returns of Swedish Banks
Model(3) Model(4) SE -0.0452∗∗∗ (0.0045) Risky 0.0244∗∗∗ (0.0012) cons -0.0261∗∗∗ -0.0908∗∗∗ (0.0021) (0.0005) F-statistic 100.842∗∗∗ 385.741∗∗∗ Observations 588 35
Note. The dependent variable in Model 3 and 4 is the bank’s cumulative abnormal returns. The regressions include the following independent variables: a dummy for whether a country is Sweden or not, a variable that indicates the bank’s initial riskiness. Robust standard errors in parentheses.
∗
p < 0.1,∗∗p < 0.05,∗∗∗p < 0.01
Table 6.4: Test for Endogeneity
Model (c) year -0.5281 (0.9142) lead CCyB 965.0833∗ (519.0840) SE 90.3588∗ (51.8782) lead CCyB*SE -1492.594∗∗∗ (489.4240) HHI 183.6231 (133.8755) cons 933.3407 (1826.8643) F-statistic 2.303∗∗ Observations 120
Note. The dependent variable in Model c is the change in the z-score, which is a measure of the bank’s riskiness. The regressions include the following independent variables: year, a dummy for whether a country is Sweden or not, the level of CCyB in the following year, the interaction variable between the level of CCyB in the following year and country dummy and Herfindahl-Hirschman Index. Robust standard errors in parentheses.
∗
Chapter 7
Discussion and Analysis
As it is shown in Table 6.2, the estimation of Model 1 has produced a statistically significant coefficient of interest of -555.8. This means that all other things equal, a 1 percentage point increase in the level of countercyclical capital buffers on average reduces the growth (improvement) in the z-score of a bank by 5.56 percentage points. In other words, the introduction of countercyclical capital buffers increases the riskiness of an average bank which is the opposite outcome than what is intended. This result is similar to Jim´enez et al. (2017).
Interestingly, this result is uniform across the banks with a different level of initial risk-iness, which is demonstrated in Model 2. That is, contrary to the prediction inspired by
Koehn & Santomero(1980) model, all banks in Sweden are affected by the introduction of countercyclical capital buffers in a similar way regardless of their initial riskiness. Additionally, the announcement of an increase in the countercyclical capital buffers has a statistically significant negative effect on banks’ cumulative abnormal returns around the time of the policy announcement. More specifically, all other things equal, the announcement of an increase in countercyclical capital buffers on average reduces banks’ cumulative abnormal returns by 4.52 percentage points; a similar conclusion was reached by Cornett & Tehranian (1994).
When it comes to the effect on the abnormal returns of the initially risky banks as com-pared to the initially less risky banks in Sweden, the results indicate that the cumulative abnormal returns of a risky bank were on average 2.44 percentage points higher than that of an initially less risky bank, all things equal. This is consistent with the result of Model 2 and most likely represents a risk premium.
It is also important to consider the limitations of this research. Even though it has been demonstrated earlier that the tests for parallel trends for both Sweden and Denmark as
24
well as the initially risky and initially less risky banks indicate that each of the groups was following a similar development path prior the introduction of countercyclical capital buffers, there still could be an issue of self-selection. This is because the introduction of countercyclical capital buffers is not mandatory, and countries can choose whether to implement it or not. Therefore, there could be some unobservable differences between Sweden and Denmark that would make the results biased.
Additionally, it has been indicated by the test for endogeneity that there is a problem of reverse causality. In other words, it is likely that not only the increase in countercyclical capital buffers affects the riskiness of banks but also the introduction of countercyclical capital buffers was provoked by rising riskiness of banks. This problem can bias the coef-ficients and can be solved by introducing an instrumental variable. Thus, the presented above results should be interpreted with caution.
Chapter 8
Conclusion
The main goal of this paper was to investigate the effect of an increase of countercyclical capital buffers on banks’ riskiness in the context of Sweden. Additionally, the effect of this regulation on the riskiness of banks with different initial riskiness was investigated as well as the impact on banks’ abnormal returns around the announcement day. The results indicate that the introduction of countercyclical capital buffers had the opposite effect from what was intended with respect to bank riskiness with a uniform effect on banks with different initial riskiness. Furthermore, the announcement of the new regulation is associated with negative cumulative abnormal returns suggesting its negative effect on banks’ profitability.
Having said that, it has been indicated that the results should be treated with caution due to the problem of reverse causality, which may bias coefficients. Therefore, one of the suggestions for further analysis is to investigate the same question with the help of an instrumental variable to rule out the current issue of endogeneity. Furthermore, it could be beneficial to carry out a similar analysis with a larger sample by, for instance, includ-ing information about the banks that are operatinclud-ing in Sweden but are headquartered in other countries. This would then also allow for the inclusion of more control variables. Additionally, other asset pricing models such as Fama and French Three-Factor Model can be used to reduce the issue of model misspecification.
To conclude, this paper has produced some suggestive evidence that countercyclical capital buffers may create unintended consequences on banks’ riskiness, but further research is needed to establish more robust results.
References
Basel III, B. C. B. S. (2011). A global regulatory framework for more resilient banks and banking systems revised version june 2011. Basel Committee on Banking Supervision. 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.
Boyd, J. H., De Nicol`o, G., & Jalal, A. M. (2006). Bank risk-taking and competition revisited: New theory and new evidence.
Cerutti, E., Claessens, S., & Laeven, L. (2017). The use and effectiveness of macropru-dential policies: New evidence. Journal of Financial Stability, 28 , 203–224.
Claessens, S., Ghosh, S. R., & Mihet, R. (2013). Macro-prudential policies to mitigate financial system vulnerabilities. Journal of International Money and Finance, 39 , 153–185.
Committee, B., et al. (2010). Guidance for national authorities operating the counter-cyclical capital buffer. Basel, Switzerland .
Cornett, M. M., & Tehranian, H. (1994). An examination of voluntary versus involuntary security issuances by commercial banks: The impact of capital regulations on common stock returns. Journal of Financial Economics, 35 (1), 99–122.
Denmark, S. (2011). Statbank denmark. Statistics Denmark’s website (www. dst. dk). FactSet. (2020). Price history. Retrieved fromhttps://my.apps.factset.com/navigator/
company-security/price-history
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.
References 27
Jim´enez, G., Ongena, S., Peydr´o, J.-L., & Saurina, J. (2017). Macroprudential policy, countercyclical bank capital buffers, and credit supply: evidence from the spanish dynamic provisioning experiments. Journal of Political Economy, 125 (6), 2126–2177. Kim, D., & Santomero, A. M. (1988). Risk in banking and capital regulation. The
journal of finance, 43 (5), 1219–1233.
Koehn, M., & Santomero, A. M. (1980). Regulation of bank capital and portfolio risk. The journal of finance, 35 (5), 1235–1244.
Vandenbussche, J., Vogel, U., & Detragiache, E. (2015). Macroprudential policies and housing prices: A new database and empirical evidence for central, eastern, and south-eastern europe. Journal of Money, Credit and Banking , 47 (S1), 343–377.
Van Dijk, B. (2015). Osiris database. Bureau Van Dijk, Amsterdam.
Yeyati, E. L., & Micco, A. (2007). Concentration and foreign penetration in latin american banking sectors: Impact on competition and risk. Journal of Banking & Finance, 31 (6), 1633–1647.
