Does the EU bank stress test have a decreasing informational value for investors? An event study on the three EU bank stress tests

(1)

1

Does the EU bank stress test have a decreasing informational value

for investors? An event study on the three EU bank stress tests

By Martin Jan de Jong

Supervisor: Dr. M. Hernandez Tinoco

Co-assessor: Dr. J.O. Mierau

MSc International Financial Management MSc Finance

Faculty of Economics and Business University of Groningen

m.j.de.jong.12@student.rug.nl S1902873

MSc Economics and Business Faculty of Social Sciences University of Uppsala

Martin.Dejong.7319@student.uu.se 910416-P873

(2)

2 Abstract

This study aims to identify the information value for the stock market of publication of stress test results. By creating a unique database comprising the listed financial institutions subject to the EU and US stress tests, this study finds that, over time, the information is mostly incorporated in the market price prior to the publication of its results, indicating a decreasing value over time. However, for institutions failing a bank stress test, the market displays significant abnormal returns following the publication of the results. Moreover, this study pioneers in analysing trends in the EU stress test, finding that movements to a failing grade result in significant negative abnormal returns.

Keywords: Stress tests, information disclosure; banks; market efficiency; event study

(3)

3

Introduction

As a reaction to the decreased trust in the banking system after the 2008 financial crisis, the European Banking Authority (henceforth: EBA), implemented the EBA stress test. The first test in 2010 included 91 banks from 21 countries, encompassing close to 65% of the total assets of all European banks. These tests were used to assess the resilience of the European banks in highly adverse macroeconomic conditions and increase the transparency in the European Banking sector.

Alves et al. (2015) find that after the announcement of the results of the EU bank stress test of 2010 and 2011, cumulative abnormal returns (henceforth: CAR) occur in the five-day and ten-day period after the publication of the results. Banks with higher leverage display a better performance in comparison with the other institutions in the sample. Moreover, volatility of the stocks declined during this period in over 18 per cent of the stocks of banks that were tested and over 20 per cent in the stocks of financial institutions that were implemented as a control group in this study. Most notably, the CAR increased for banks included in the bank stress test between 2010 and 2011, compared to the control group of financial institutions. How this developed in the 2014 EU bank stress test is still unidentified. Therefore, this study aims to perform an in-depth analysis on the CAR of all three bank stress tests (2010,2011,2014) and their informational value to the market. This study builds on the research by Alves et al. (2015) and Morgan et al. (2014) by including the results of the 2014 EU bank stress test and CCAR 2011-2015, in order to investigate if any trends are occurring in the results of the stress test. Moreover, a comparison of the informational value of both stress tests is conducted. As previous literature concludes that higher leveraged banks perform better in the 2010 EU bank stress test, this result was more modest in 2011. This raises the question whether this is a trend that is reoccurring or that the 2014 EU banks stress test has a higher informational value, leading to higher CAR after the publication of the results, or does the stock market already incorporate the new information which will be published.

(4)

4

2014 show a higher CAR. From an investor’s perspective, not only can positive CAR be exploited, also negative CAR can be exploited by taking a short position. As previous literature does not emphasize the occurrence of trends in relation to the publication of new information, this research aims to identify whether trends are reoccurring after the publication of stress test results. Especially in the banking sector, Xiaolou & Woodlock (2011) indicate that most of the variance of total stock returns stems from unexpected returns driven by market conditions, it is interesting to research the CAR following the publication of the EU bank stress test results and compare this to the US counterpart.

This research finds that the inclusion of the 2014 stress test allows for significant trends to occur. Moreover, that banks nearly passing twice, results in higher positive CAR relative to its peer group as the positive information was not incorporated in the market prior to the publication. Subsequently, the market is reacting to previous results and becomes more efficient. As the post-event windows in 2010 and 2011 display significant positive CAR, the pre-event window in 2011 and 2014 display significant CAR as well. Literature indicates that the change from positive (negative) pre-event CAR to negative (positive) post-event CAR is a result from an overreaction in the market. Due to the increase of significant CAR in the pre-event windows over time, for both the EU stress tests and US stress tests, early incorporation of the new information is visible as the market becomes more efficient. Also, the impact of the leverage ratio becomes less significant in the 2014 stress test compared to the previous tests. This could indicate that the relationship between higher risk and higher returns, does not hold in the long term. The publication of the stress test results have a stronger impact amid riskier banks in 2010 and 2011, however, this impact is decreasing in the 2014 stress test. On the other hand, volatility in the market is decreasing after publication of the 2010 stress test results, while volatility increases after the publication of the 2011 and 2014 stress test results. Over time, this research finds that the EU bank stress test still has a higher informational value compared to its US counterpart. However, for both banks the informational value decreases over time.

(5)

5

the methodology, stating how the hypotheses will be tested. Section VI presents the results of the conducted analysis. Section VII will include the robustness checks of this research. Finally, Section VIII presents the conclusion of this paper.

I. EU Bank stress tests and US stress test

In May 2009, the Committee of European Banking Supervisors (henceforth: CEBS) was mandated to perform a forward looking stress test on the aggregate banking system. Most notably, the stress test was not aimed at identifying individual banks in need of recapitalization, because this remains a responsibility of the national authorities. The main purpose of stress testing is to increase the level of aggregate information among policy makers in assessing the resilience of the European financial system (CEBS, 2009). The stress test includes an assessment of credit risks based on two sets of commonly agreed economic scenarios, a baseline scenario and an adverse scenario. An adverse macro-economic scenario represents a severe but plausible shock, such as the one experienced in 2008. Secondly, a sensitivity analysis is conducted on the trading book and market risk positions, which is based on commonly agreed parameters (CEBS, 2009). The CEBS works closely with national supervisory authorities when creating the individual risk profiles. By proposing benchmark risk parameters, agreed scenarios and agreed guidelines in the process, the CEBS ensures consistency and comparability of the results.

Previous literature has studied the EU bank stress tests of 2010 (2011), in which 90 (91) banks were assessed on their credit risk and 7 (8) failed. However, the 2014 EU bank stress in which 123 banks were tested and an astonishing 24 banks failed the test remains unidentified. However, ten banks already took measures in the time between the stress test and the publication of the results.

(6)

6

From 2011 to 2015, the FED conducted the Comprehensive Capital Analysis and Review (henceforth: CCAR). In the CCAR, the banks are evaluated across five areas, namely: capital assessment and planning processes, capital distribution policy, plans to repay any government investment, ability to absorb losses under several scenarios and plans for addressing the expected impact of Basel III and the Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 (CCAR, 2011). The banks which are tested, present their capital plans according to a set of guidelines provided by the FED. Subsequently, the FED either objects, or does not object the provided capital plans. Prior to the publication of the results, the FED provides its judgement in private to the bank.

II. Literature Review

According to the efficient-market hypothesis (henceforth: EMH) by Fama (1970), asset prices should reflect all the available information in a market where the efficiency is strong. This would make it impossible for investors to benefit from under- and over-priced stocks, indicating that an investor is unable to outperform the market. Subsequently, price changes occur as new information causes prices to shift. Tetlock (2010) measures the asymmetric information of a firm’s stock price following financial news events. The results indicate that public news plays a key role in informing some investors, but not all. Certain investors are able to predict and trade on the information that will be released. As news will decrease the information asymmetry, positive news will increase the stock holdings of investors in a particular firm. Although most research is focusing on positive news, Ball and Kothari (1991) and Kothari (2001) find that the market reacts strongly to the publication of negative earnings surprises.

(7)

7

direction in which the bank moves and that this information is incorporated in the stock returns. In contrast to Pettway (1980), Petrella & Resti (2013) find that market prices react significantly to supervisory announcements and inspections analysing the 2011 EU stress test, concluding that the market was unable to incorporate all relevant information. The findings of Petrella & Resti (2013) are in line with the recent paper on the 2010 and 2011 EU stress test by Alves et al. (2015).

Baumann & Nier (2004) add that the increased information disclosure of banks is both valuable to banks and investors, concluding that stock volatility has decreased in the long-term with increased information disclosure. This could be compared to the findings by Alves et al. (2015) where similar results are obtained for the short-term.

(8)

8

Also, Neuhierl et al. (2013) find that the time stamp on press releases accurately finds captures the time at which the market first learned the news. By studying press releases on a wide variety of corporate aspects between April 2006 and August 2009, the research finds that volatility in the market significantly increases following most type of announcements. Moreover, the authors find that this volatility is related to news-induced valuation uncertainty. This would imply that after the results of the stress test are made public there is no case of under-reacting, as suggested by Daniel et al. (1998) and Morgan et al. (2014), as the CAR are higher in the 0-5 day period then the 0-10 period.

Comparably, Morgan et al. (2014) find that investors partly anticipate the outcome of the US SCAP in 2009. Most notably, the CAR do not occur most significantly with the publication of the results, but occurred with the clarification of the SCAP methodology, including a testimony by its chairman Bernanke stating that the banks failing the SCAP will not be nationalized. The investors anticipated which banks would have a significant capital gap, which lead to negative CAR in the period around the chairman’s statement. The authors indicate that publication of the results is informative, and results which display larger positive or negative results, could lead to CAR’s. Complementary, Quijano (2014) finds similar results by analysing the bank bond returns after the publication of the 2009 SCAP results, indicating a decrease in information asymmetry.

(9)

9

the informational value of the CCAR. As the 2009 SCAP finds significant CAR around several announcement dates, it is interesting to compare the informational value the SCAP and CCAR have to the European Bank Stress test conducted by the EBA. To investigate this theory, the following hypotheses will be tested.

III. Hypotheses

Based on the previous literature review focussing on the information disclosure and behavioural finance, the following hypotheses are proposed.

H₁: Stock returns do not incorporate the outcome of the stress test in their prices.

As literature concludes, the CAR’s are mostly apparent in the tested periods after the publication of the results, as new information is presented to the market. As previous literature finds that for the 2010 and 2011 European Bank stress test and the 2009 SCAP, the market did not anticipate the outcome, this research tests this hypothesis on the previously researched stress test, but adds to the literature by extending the research to the more recent European bank stress test and CCAR.

H₂: Banks that fail show negative abnormal returns.

As this phenomenon is not extensively discussed in previous literature due to data availability, it remains an opportunity for investors to benefit from negative abnormal returns. By taking a short position on the security which is probable of failing, investors are able to benefit from a negative CAR. Also, the informational value of the group failing a stress test could be of significant value to the market, depicted in a negative CAR. As in line with Pettway (1980), this research expects banks that fail the stress test to have negative abnormal returns.

H₃: Banks that have a higher leverage have higher abnormal returns.

(10)

10

2011 EU stress test, which raises the question whether a decreasing trend is occurring for the market to incorporate the stress test results. Therefore, including the third stress test could be of great value to analyse whether the decrease will persist. As these results were significant in both cases, the expectation is that high leveraged banks will continue to have a higher CAR.

H₄: Banks that pass or fail multiple times in a row, have higher abnormal returns compared

to banks that pass or fail for the first time.

According to the behavioural finance field, especially Barberis et al. (1998), a series of good (bad) news produce an increased positive (negative) CAR. As this is still under-researched in relation to the stress test, this is where this study can provide a significant insight for investors to benefit from a series of positive or negative results. Also, the analysis of trends could identify whether the efficiency in the market is increasing. This will be elaborated on in eighth hypothesis.

H₅: Banks that move from passing (failing) to failing (passing) show higher negative

(positive) abnormal returns.

As banks move from a positive scenario to a negative scenario, or the other way around, the expectation is that this will cause higher abnormal returns than banks that pass. As literature proposes that unexpected positive news, moving from failing to passing, leads to higher CAR’s, this hypothesis will test whether this holds for the publication of the bank stress test results. Subsequently, it is expected that banks that move from passing to failing, will have higher negative abnormal returns.

H₆: Volatility of banks included in the stress test decreases due to decreased

information asymmetry.

(11)

11

both prior to the event and shortly after the event. Although the decline in volatility decreased between the first and second stress test, the expectation is that this will still be apparent when including the third bank stress test. This will be tested by performing a GARCH (1,1) test, on which will be elaborated in the methodology section.

H₇: The CAR following the US stress test are lower compared to the EU stress test.

As Morgan et al. (2014) indicate that the CAR around the publication date of the 2009 SCAP results are lower, compared to the CAR around the statement made by chairman Bernanke. Subsequently, as the SCAP and CCAR results are provided privately to the banks, the informational value of the publication of the results could be lower, as this information might be available earlier in the market.

H₈: Over time, the CAR’s are decreasing with the amount of stress tests conducted.

The more stress tests are conducted, the lower the informational asymmetry should be between the supervisors and the market. The market will become more efficient (Fama & French, 1970), and consequently, the informational value of the results decreases. Therefore, this study argues that the CAR will decrease over time, as information asymmetry decreases.

IV. Methodology

As recent literature proposes, the estimation period will start 130 trading days prior to the publication of the relevant stress test and will end 10 days prior to the event date. As stock prices must react to the event and subsequently, have time to recover from the event of over-reaction / under-reaction to the event (Daniel et al., 1998). For this purpose, there will be four event windows to investigate whether investors had prior knowledge of the outcome by including two event windows shortly before the publication of the results at t₀, and shortly after the publication of the results as in line with Alves et al. (2015).

(12)

12

Warner, 2007; MacKinley, 1987). The event date will be the date when the EBA announces the results of the bank stress tests e.g. October 26th 2014. For calculating the returns, this research follows the paper by Brown & Warner (1985), the simple returns are defined by:

𝑅_𝑖,𝑡 = 𝑃𝑖,𝑡 − 𝑃𝑖,𝑡−1

𝑃_{𝑖,𝑡−1} (1)

Subsequently, the abnormal returns will be identified by the market model:

𝐴𝑅𝑖𝑡 = 𝑅𝑖,𝑡− (𝑎𝑖+ 𝛽𝑖 × 𝑅𝑚𝑡) (2)

where 𝐴𝑅𝑖𝑡 is the abnormal return on stock i at time t and 𝑅𝑚𝑡 is the return on the market at time t. The daily return on the Eurostoxx 600 banks is taken as the market return. As previous research is indecisive of using the home market index (Petrella & Resti, 2013) or an industry specific market (Morgan et al. 2014), this research applies the industry specific market. Moreover, this research aims to identify whether the information value to the tested banks is higher than the control group. Therefore, one single industry specific market can be applied and differences can be measured accordingly. The 𝑎 and 𝛽 of the bank will be calculated by running an Ordinary Least Squares regression. The Eurostoxx 600 banks has the closest proximity as a market indicator based on the test sample. The CAR are computed as follows:

𝐶𝐴𝑅_𝑖[𝑡₀; 𝑡₁] = ∑ 𝐴𝑅_𝑖𝑡 𝑡₁

𝑡=𝑡₀ (3)

(13)

13

t-₁₃₀ t-₁₀ t-₉ t-₅ t₀ t₅ t₁₀

Estimation Period (1) (2) (3) (4)

Figure 1: Timeline

The pre-event windows function as a mean to research the anticipation by investors on the results. If the results are anticipated, CAR will occur during the pre-event windows (1) and (2) as experienced by Morgan et al. (2014) in their research on the SCAP. If the results are not anticipated, as indicated by H₁, the CAR will occur during post-event window (3) and (4). We assume that 𝐶𝐴𝑅[𝑡₀; 𝑡₁] ≠ 0 for the indicated windows as previous literature has shown (Morgan et al. 2014; Alves et al. 2015). However, the statistical analysis will indicate whether this is significant or not.

Subsequently, the stress tested banks will be separated in groups. Group “Pass” consists of the tested banks that have passed the test. Group “Near Pass” will contain banks that have received a warning for low, but still passable, capital ratios. Group “Fail” will contain banks that failed the stress test.

To assess whether the average CAR is significant statistically, the research applies the T-test.

𝑡_{𝑠𝑡𝑎𝑡} = 𝐶𝐴𝑅[𝑡₀; 𝑡₁]

𝜎𝐶𝐴𝑅[𝑡₀; 𝑡₁] ~𝑡(120 − 2) (5)

For other event studies, it can be assumed that the 𝐶𝐴𝑅𝑖𝑡 is independent and identically distributed. When this is applied to an analysis on organizations within the same industry with the same influential event, this might cause spill-over effects and co-movements of the stock returns. Therefore, this research applies two correction methods suggested by literature; the cross section crude dependence adjustment (Brown & Warner, 1980) and the cross sectional average correlation method (Kolari & Pynnönen, 2010). The variance, as used in the cross sectional average correlation method, is calculated as follows:

𝜎2 ₌ 𝜎̃²

𝑁 × (1 + (𝑁 − 1) × 𝑝̅) ₍₆₎

(14)

14 𝜎̃²[𝐶𝐴𝑅[𝑡₀; 𝑡₁]] = 1 𝑁²× (∑ 𝜎²(𝐶𝐴𝑅𝑖[𝑡₀; 𝑡₁]) 𝑁 𝑖=1 ) (7)

where 𝑝̅ is the average correlation of the CAR within a group. Based on these two equations, the variance can be computed for the cross sectional average correlation method. The variance in the cross-section crude dependence adjustment is expressed as

𝜎2_{[𝐶𝐴𝑅[𝑡}0_{; 𝑡}1_{]] =} 1 𝑇 − 1× (∑ 𝑇 𝑇=1 (1 𝑁∑ 𝐴𝑅𝑖𝑡− 𝐴𝑅̅̅̅̅̅𝑖 𝑁 𝑖=1 ) ²) (8)

On the other hand, literature discusses the process of handling a non-normal distribution of the abnormal returns. Therefore, a Chi-squared test will be performed to investigate the distribution of the CAR. When a normal distribution is found in the sample, a non-parametric test needs to be applied to the abnormal returns. This research will follow the Cowan (1992) Generalized Sign test to deal with the non-normal distribution. The Cowan (1992) test is an extension of the Corrado (1989) test, creating a test for dealing with a non-normal distribution for a longer event window. The test statistic uses the non-normal approximation to the binomial distribution with parameter.

𝑝̂ = 1 𝑛∑ 1 120 𝑛 𝑗=1 ∑ 𝑆𝑗𝑡 𝐸₁₂₀ 𝑡=𝐸₁ (9) where 𝑆_𝑗𝑡= {1 if 𝐴𝑅𝑗𝑡 > 0 0 Otherwise

The Z-value will be calculated as follows by the Generalized Sign test statistic: 𝑍_𝐺 = 𝑤 − 𝑛𝑝̂

[𝑛𝑝̂(1 − 𝑝̂)]0.5

(10)

(15)

15 𝑍𝑟 = 𝑑 1 2 𝐾𝐷− 70.5 [∑140𝑡=₁(𝐾𝑇− 70.5)² / 140 ] 1/2 (11)

Where 𝐾𝐷 is the average rank across the n stocks and d days of the event window and is the average rank across n stocks on day t of the 140 day combined estimation and event period. This extension treats the estimation window and event windows as one single time series. Rank 1 represents the smallest abnormal return, rank 140 the largest. The mean rank of this time series is 70.5. Finally, to asses volatility changes, a GARCH (1,1) will be performed to analyse volatility changes in the four event windows previously indicated. The volatility is assessed by

𝑅_𝑖𝑡 = 𝑎_𝑖 + 𝛽𝑖 × 𝑅_𝑚𝑡+ 𝑣_𝑡 (12)

and

𝑣_𝑡² = 𝑦₀+ 𝑦₁× 𝑣_𝑡−1+ 𝑙_𝑖 × 𝜎_𝑡−1² + ∅ × 𝑑𝑡

(13) where dt is the binary variable which takes a value equal to 1 in the event window and 0 in the other periods, 𝑣𝑡 represents the conditional variance and 𝜎² the unconditional variance. The ∅ indicates the volatility in the event window. If positive and statistically significant, it indicates an increase in volatility. When it is negative and statistically significant, it indicates a decrease in volatility.

V. Database description

The sample is based on the official documents from the European Banking Authority and FED available on their websites. For the European sample, the ISIN numbers are ran through DataStream to receive the daily return data and relevant debt and equity values. Banks that did not have a daily return for a complete year were deleted from the sample. This results in 47 listed banks which were tested in 2010, 48 listed banks in 2011 and 59 listed banks in 2014. Only in 2014 there are multiple banks (13) failing the stress test, while in 2011 only one bank failed in this sample (EFG Eurobank Ergasias). As conclusions cannot be generalized by reviewing one bank failing a stress test, a “Fail Group” is omitted from the 2010 and 2011 stress test.

(16)

16

“Pass” represents the banks that have passed the stress test the respective year. Banks that have failed the stress test are presented in the “Fail” group. Tangential results are identified when a tested bank has a CT1R ratio that is between 0 and 1 per cent higher than the requirement for passing the stress test over a 2-year time horizon. In the stress test of 2010 and 2011 this results in a CT1R ratio between 5% and 6%. For the 2014 stress test a tangential result is between 5.5% and 6.5%. This group is identified by the “Near pass” group.

For the EU sample, a control group of 115 banks is created, consisting of banks that have not been subject to any of the stress tests in this sample. The European banks are selected based on their “Total assets” value in Bankscope. Subsequently, the return data is gathered from the DataStream database. Appendix A displays an overview of all the groups in this study. Table 1 in the following section displays the descriptive statistics.

For the US sample, the returns are exported from Compustat. As proposed by previous literature, the S&P 500 Financials, which is available through Yahoo Finance, is used as the market return. For the 2009 SCAP, the banks will be divided in a “GAP” group, identifying banks which are assessed with a capital short, and a “No GAP” group, representing banks which have a sufficient capital buffer. For the CCAR, the 2011 CCAR will only have a “Tested” group. As in the 2011 CCAR, firm-specific results were not disclosed publicly (CCAR 2011). For the following CCARs, the banks are divided in “Non-objection” groups, encompassing banks to which the FED did not object the capital plans, and an “Objection” group, containing banks to which the FED objected the capital plans of that specific year. For the US database, no control group is conducted, as this database functions as a comparison to the EU stress test, which is the main subject of analysis. An overview of the groups is presented in Appendix A.

VI. Results

(17)

17

Based on the methodology in the previous sections, the CAR’s are calculated for the European and US stress tests and displayed in Table 1.

A detailed overview of which bank belongs to which group, is presented in Appendix A. Surprisingly, the average CAR of the “Near Pass” group in 2011 and 2014, yields the highest CAR of all groups in the respective years. This indicates that this information was unexpected by the market. Also, the pre-event windows in 2011 display negative CAR’s throughout the sample, while, except for the control group, the groups have positive CAR’s in the post-event windows. This is reversed in 2014, where the pre-event window CAR’s are all positive, while most surprisingly, the CAR for the treatment group is negative. In both cases, as indicated by previous literature (Barberis et al., 1999; Antweiler & Frank, 2004; Daniel et al., 1999; Petrella & Resti, 2013; Alves et al., 2015) , this displays a clear “overreaction” in the market.

Informational Value

As H1 states, this research argues that the market did not incorporate the outcome of the stress test. When reviewing Table 2A, we can accept H1, indicating that stock returns do not incorporate the outcome of the EBA stress tests. This can be concluded due to the large fraction of significant CAR in the two post-event windows. Although the CAR is significant in most of the pre-event windows of the 2011 and 2014 events, the sign changes of CAR. This indicates that the stock returns were overly optimistic in 2014, resulting in significant positive CAR in the pre-event windows, while portraying negative or insignificant CAR in the post-event windows. This indicates that the informational value is significant, as it was not incorporated in the stock prices prior to the announcement. Subsequently, the market shows signs of overreaction to the news, which can be identified by the change from positive CAR in the pre-event windows, to negative CAR in the post-event windows. On the other hand, stock returns in the pre-event windows of 2011 were overly pessimistic, resulting in significant negative CAR. However, when the results of the EBA stress test were published, the stock returns display significant positive CAR, most notable in the “Near Pass” group, indicating that the information in this group has the highest value to the market.

(18)

18 Table 1A: Descriptive

statistics European Banks Number of

Stocks:

Average CAR: Max CAR: Min CAR:

CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) 2010 Stress Test: Full Sample 162 0,62% -0,86% 0,40% 1,00% 28,98% 9,77% 16,74% 17,24% -7,31% -6,44% -17,00% -37,33% Control Group 115 0,27% -0,76% -0,14% 0,49% 26,23% 6,82% 16,74% 17,24% -7,31% -6,44% -17,00% -37,33% Treatment Group 47 1,47% -1,10% 1,70% 2,23% 28,98% 9,77% 12,03% 12,32% -5,89% -5,95% -8,33% -9,60%

Near Pass Group 7 1,42% -2,00% 1,87% 1,86% 14,87% 2,55% 10,26% 10,91% -2,85% -4,78% -3,30% -7,31% Pass Group 40 1,48% -0,94% 1,67% 2,29% 28,98% 9,77% 12,03% 12,32% -5,89% -5,95% -8,33% -9,60%

2011 Stress Test:

Full Sample 163 -1,52% -1,40% 0,28% -0,26% 8,60% 7,85% 18,45% 17,61% -23,51% -18,46% -15,37% -14,35% Control Group 115 -0,59% -1,20% -1,08% -1,25% 8,60% 6,50% 16,09% 17,61% -20,14% -13,73% -15,37% -14,35% Treatment Group 48 -3,80% -1,87% 3,54% 2,10% 7,01% 7,85% 18,45% 13,65% -23,51% -18,46% -6,29% -5,06%

Near Pass Group 7 -6,53% -1,92% 5,48% 3.53% -0,52% 7,85% 10,56% 13,65% -11,56% -6,19% 1,40% -5,06% Pass Group 40 -3,04% -1,73% 2.99% 1,68% 7,01% 5,04% 18,45% 12,37% -23,51% -18,46% -6,29% -4,42%

2014 Stress test:

Full Sample 174 1,24% 0,95% -0,11% 0,06% 25,30% 16,43% 10,70% 21,99% -8,39% -11,35% -45,96% -30,47% Control Group 115 0,69% 0,68% 0,97% 1,46% 9,56% 6,83% 7,34% 11,52% -8,39% -11,35% -16,23% -13,43% Treatment Group 59 2,35% 1,51% -2,20% -2,66% 25,30% 16,43% 10,70% 21,99% -7,75% -5,90% -45,96% -30,47% Near Pass Group 4 2,97% 3,16% 1,69% 2,73% 10,94% 11,69% 5,35% 4,93% -2,39% -1,01% -1,08% 1,18%

(19)

19 Table 1B: Descriptive statistics

American Banks

Number of Stocks:

Average CAR: Max CAR: Min CAR:

CAR ` (-9; -1) CAR (-5; -1) CAR (0;5) CAR

(0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10)

(20)

20 Table 2A: Cumulative Abnormal Returns (CAR) EU

stress test

CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) 2010 Stress Test:

Full Sample 0,616% -0,855% 0,397% 0,996%**/**

Control Group 0,268% -0,757%*/ -0,137% 0,493%

Treatment Group 1,469%*/* -1,095% 1,703%**/** 2,225%***/***

Near Pass Group 1,418% -1,995%*/** 1,866%**/** 1,862%**/**

Pass Group 1,477% /* -0,938% 1,674%**/** 2,288%***/***

2011 Stress Test:

Full Sample -1,517%**/* -1,397%**/* 0,280% -0,260%

Control Group -0,589% -1,202%*/ -1,082%*/ -1,246%**/*

Treatment Group -3,799%***/*** -1,865%*/* 3,544%***/** 2,103%*/**

Near Pass Group -6,531%***/*** -1,921%*/** 5,481%***/*** 3.530%**/***

Pass Group -3,043%**/** -1,734%*/** 2.985%**/*** 1,682%*/**

2014 Stress test:

Full Sample 1,239%*/* 0,947%*/* -0,106% 0,061%

Control Group 0,692% 0,675% 0,965% 1,459%**/*

Treatment Group 2,346%**/** 1,509%*/** -2,198%*/* -2,661%**/**

Near Pass Group 2,965%*/** 3,158%*/** 1,685% 2,734%**

Pass Group 2,446%**/** 0,845%*/* 0,282% -0,120%

Fail Group 1,832%*/* 3,146%/** -11,406%***/*** -12,531%***/***

Note: ***/** and * denote statistical significance at 1%, 5% and 10% respectively. The significance level is based on the Cross Sectional Average Correlation method / Cross section Crude dependence adjustment. The table displays the behaviour of the CAR during the two pre-event and two post-event windows.

Most of the CAR occur during the pre-event windows, indicating that the information which is published on the event date, is already incorporated into the stock price prior to its announcement. Moreover, the 2014 CCAR does not result in any significant CAR in all of the pre-event and post-event windows, implying that the publication of the results did not have any informational value for the market. On the other hand, there is a significant difference between the Non-Objection and Objection groups in the 2012 and 2013 CCAR.

Results of failing the stress test

(21)

21

post-event windows as can be derived from table 2A. Moreover, when reviewing table 3A, the percentage of statistically significant negative CAR in the post-event window is over sixty percent. This indicates that the market did not expect these banks to fail the stress test as this group has significant negative CAR in the post-event windows. The reaction of the market, moving from positive CAR in the pre-event windows and negative CAR in the post-event windows, display signs of overreaction in the market.

1_{Although Morgan et al. (2014) use a different event window of [𝑡₋₁; 𝑡₁] the results presented in this paper}

remain comparable to prior research. As this study has found comparable CAR in the 2009 SCAP event windows proposed by Morgan et al. (2014).

Table 2B: Cumulative Abnormal returns (CAR) US SCAP / CCAR1

Event window: CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10)

2009 SCAP: Tested Group: 1,820% 4,491%**/*** 0,253% -1,636% GAP group: 4,809%*/** 8,291%***/*** 1,645% -1,367% No GAP group: -1,168% 0,691% -1,139% -1,905%*/** 2011 CCAR: Tested Group: 1,396%*/** 1,329%*/** -0,581% -0,678% 2012 CCAR: Tested Group: -1,134%*/** -0,048% 1,063%*/* 0,894% Non-Objection -1,010%*/** 0,150% 1,787%**/** 2,259%**/*** Objection -1,601% /*** -0,789%**/ -1,652%***/** -4,228%***/*** 2013 CCAR: Tested Group: 1,419% */*** 0,773% /** 0,074% 0,199% Non-Objection 2,375%**/*** 1,261%*/*** 0,449% 0,645% Objection -2,166%*/*** -1,056% /* -1,331% /* -1,473% /** 2014 CCAR: Tested Group: 0,496% 0,485% 0,234% -0,451% Non-Objection 0,559% 0,539% 0,228% -0,427% Objection 0,165% 0,207% 0,269% -0,578% 2015 CCAR: Tested Group: 1,041%**/** 0,996%**/* 0,153% 0,234% Non-Objection 1,159%**/** 1,009%**/* 0,226% 0,072% Objection -0,105% 0,876%/* -0,555% 1,801% /*

(22)

22

Although the technicalities are different for the US SCAP and CCAR compared to the EU stress test, objection to the capital plans from the FED can be compared to failing thestress test. In both cases, the capital buffer in the near future is insufficient and there is a need for revised capital plans. The “Objection” group displays inconsistent CAR’s from 2011-2015. However, most of the CAR’s in the pre-event and post-event windows of the CCAR are negative and significant, indicating that the market expects the results to be negative. These results are in line with previous research (Pettway, 1980; Ball & Kothari, 1991; Kothari 2001), finding that the market reacts strongly to the publication of negative financial news of a bank. Subsequently, this research accepts H2.

Leverage

To test whether leverage has an impact on the CAR, the treatment groups have been divided into quartiles based on their debt-to-equity ratio. The “First Quartile” includes the banks with the highest DE ratio, while the “Fourth Quartile” includes the banks with the lowest DE

Table 3A: Percentage of statistically significant CAR at 5%

CAR > 0 CAR < 0 CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) 2010 Stress Test: Full Sample 12,30% 4,30% 20,20% 27,60% 16,60% 17,80% 14,70% 17,80% Control Group 9,60% 2,60% 16,50% 24,30% 16,50% 17,40% 18,30% 21,70% Treatment Group 19,10% 8,50% 29,80% 36,20% 17,00% 19,10% 6,40% 8,50% Near Pass Group 14,30% 0,00% 28,60% 28,60% 14,30% 14,30% 0,00% 0,00% Pass Group 20,00% 10,00% 30,00% 37,50% 17,50% 20,00% 7,50% 10,00%

2011 Stress Test:

Full Sample 12,90% 6,10% 9,80% 10,40% 29,40% 25,80% 16,00% 19,60% Control Group 16,50% 7,00% 3,50% 6,10% 24,30% 26,10% 20,90% 26,10% Treatment Group 4,20% 4,20% 25,00% 20,80% 41,70% 25,00% 4,20% 4,20% Near Pass Group 0,00% 12,50% 25,00% 37,50% 62,50% 25,00% 0,00% 0,00% Pass Group 4,20% 2,10% 20,80% 14,60% 29,20% 18,80% 4,20% 4,20%

2014 Stress test:

(23)

23 Table 3B: Percentage of statistically significant

CAR at 5% CAR > 0 CAR < 0 CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) 2009 SCAP: Tested Group: 11,10% 16,70% 11,10% 16,70% 22,20% 11,10% 33,30% 44,40% GAP group: 11,10% 22,20% 22,20% 11,10% 0,00% 0,00% 44,40% 33,30% No GAP group: 11,10% 11,10% 0,00% 22,20% 44,40% 22,20% 22,20% 55,60% 2011 SCAR: Tested Group: 22,20% 22,20% 16,70% 16,70% 5,60% 5,60% 22,20% 22,20% 2012 SCAR: Tested Group: 10,50% 5,30% 15,80% 21,10% 31,60% 10,50% 5,30% 21,10% Non-objection 7,10% 7,10% 21,40% 28,60% 28,60% 7,10% 0,00% 14,30% Objection 25,00% 0,00% 0,00% 0,00% 50,00% 25,00% 25,00% 50,00% 2013 SCAR: Tested Group: 36,80% 31,60% 21,10% 21,10% 15,80% 10,50% 21,10% 26,30% Non-Objection 46,70% 33,30% 26,70% 26,70% 0,00% 6,70% 26,70% 20,00% Objection 0,00% 25,00% 0,00% 0,00% 75,00% 25,00% 0,00% 50,00% 2014 SCAR: Tested Group: 38,70% 32,30% 16,10% 19,40% 16,10% 12,90% 6,50% 25,80% Non-Objection 42,30% 30,80% 19,20% 19,20% 19,20% 15,40% 3,80% 26,90% Objection 20,00% 40,00% 0,00% 20,00% 0,00% 0,00% 20,00% 20,00% 2015 SCAR: Tested Group: 40,60% 12,50% 6,30% 21,90% 15,60% 9,40% 6,30% 25,00% Non-Objection 41,40% 13,80% 6,90% 20,70% 17,20% 6,90% 6,90% 24,10% Objection 33,30% 0,00% 0,00% 33,30% 0,00% 33,30% 0,00% 33,30%

ratio. These results are comparable to previous research, which finds significant CAR in the highest leverage quartiles in the post-event windows for both the 2010 and 2011 stress test.

(24)

24

However, the 2014 stress test resulted in more than double the amount of failed banks compared to 2010 and 2011. Therefore, negative CAR are the most likely consequence and this is visible throughout all the post-event windows of the 2014 stress test. Therefore, H3 is accepted for the 2010 and 2011 stress test. However, due to the unexpected results of the 2014 and earlier incorporation of the stress test results in the market, H4 is rejected for the 2014 stress test as the impact of leverage on the CAR’s is higher in lower quartiles of the DE ratio. Subsequently, this challenges the assumption that higher risk is associated with higher returns.

Multiple Fails/Pass

As H4 states, this research expects banks that fail (pass) multiple times, to have lower (higher) CAR in the post-event windows. Unfortunately, the databases consulted do not allow for this research on multiple fails to be performed with accuracy. In the sample, only one bank that is listed has failed the 2011 stress test (Eurobank Ergasias) that has also failed the 2014 stress test. No listed banks have failed the 2010 stress test from this database. Therefore, this research will not draw any conclusions on this part. However, the database

Table 4: CAR by leverage quartiles

CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10)

2010 Stress Test: First Quartile -0,035% -0,198% 0,656% /** 0,626% **/* Second Quartile 0,694% -0,243% 0,643% /* 1,050% */** Third Quartile 0,060% -0,560% 0,373% 0,299% Fourth Quartile 0,750% -0,091% 0,031% 0,250% 2011 Stress Test: First Quartile -4,822%**/*** -2,289% /* 4,986% */** 3,855% **/** Second Quartile -2,525%**/** -1,040% 2,370%**/** 1,580%*/* Third Quartile -2,616%*/* 0,080% 2,305%**/** 2,395%*/** Fourth Quartile -5,210%**/*** -4,205%**/*** 4,705%**/** 0,780% 2014 Stress test: First Quartile 2,538%*/** 1,620% /* -3,910%**/** -2,270%*/* Second Quartile 2,620%*/** 2,470% /** -2,200% /** -3,516% **/*** Third Quartile 2,750%**/** 2,730%**/** -4,665%***/*** -5,750%***/*** Fourth Quartile 0.972% -0.322% 0.714% -1.617%*/**

Note: ***/** and * denote statistical significance at 1%, 5% and 10% respectively. The significance level is based on the Cross Sectional Average Correlation method / Cross section Crude dependence

(25)

25

allows to investigate whether banks that pass for the third time outperform the “Pass” group of 2014. The CAR for this test are presented in Table 5. The difference between both groups is incremental, as the largest difference occurs in the 0-10 days post-event windows. This allows for discussion of the efficient market hypothesis, as the market does not depict any statistical significant CAR at the 5% level for this group. The market does incorporate the news of this group prior to the publication of the result, indicating an increased market efficiency. Therefore, based on this finding, H4 is partly rejected as market efficiency has increased and the market did incorporate the results of the stress test prior to its publication.

Trends in the stress tests

As H5 hypothesizes that banks that move from passing (failing) to failing (passing) show higher negative (positive) abnormal returns, these trends are analysed in Table 5. One of the main contributions of this research is the addition of the 2014 stress test. This allows for analysing trends occurring over time. As the literature indicates possibilities for higher CAR after a series of good news or following an unexpected event, this research will identify whether this holds for the publication of the EU stress test results. For the analysis of trends, this research focuses on banks that have shifted from a “Near Pass” to “Pass” or reversed. Also, a series of “Near Pass” or banks that have passed all three stress tests can be analysed to review the influence of the series of news.

(26)

26

Table 5: Trend CAR

Group:

Number of Stocks:

Average CAR:

CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10)

Triple Pass: 24 1,648%* 0,656%* 0,805%* 0,659% Pass 2014: 42 2,446%** 0,845%* 0,282% -0,120% Double Near-Pass: 3 -7,106% -0,291% 8,801%** 8,078%** Near pass 2011: 8 -6,531%*** -1,921%** 5,481%*** 3,503%*** Pass 2010 - Near Pass 2011: 4 -6,099%** -0,309% 2,991%** 1,002% Near Pass 2011: 8 -6,531%*** -1,921%** 5,481%*** 3,503%***

Near Pass 2010 - Pass

2011: 3 -5,933%** 1,405% 5,309%** 0,690%

Pass 2011: 40 -3,043%** -1,734%** 2,985%*** 1,682%

Pass 2011 - Fail 2014: 3 1,215% 4,771% -20,662%*** -14,235%***

Fail 2014: 13 1,832%* 3,146% -11,406%*** -12,531%***

Pass 2011 - Near Pass

2014: 2 -0,982% -0,146% 2,777%* 2,320%*

Near Pass 2014: 4 2,965%* 3,158%* 1,685% 2,734%**

Near Pass 2011 - Pass

2014: 3 2,000% 0,057% -5,073%* -8,441%*

Pass 2014: 42 2,446%** 0,845%* 0,282% -0,120%

Note: ***/** and * denote statistical significance at 1%, 5% and 10% respectively. The significance level is based on the Cross Sectional Average Correlation / Crude dependence adjustment. The table displays the behaviour of the CAR during the two pre-event and two post-event windows. Although the Cowan (1992) test is for positive CAR, the calculation can be applied to negative CAR as well. This is how the significance levels are attained for the negative CAR

CAR occur. As these values are lower compared to their peer group, this could indicate that the market was expected certain banks that move to a “Near Pass” in 2014 after passing the 2011 stress test.Moreover, in the extreme case when banks have passed the 2011 stress and failed the 2014 stress test, they exhibit even higher statistically significant negative CAR compared to the regular “Fail” group of 2014. Insignificant positive CAR occur in the pre-event windows, indicating that this group was most likely unexpected to “Fail” significantly.

(27)

27

displays a negative CAR in the pre-event windows, suggesting that these banks were most likely perceived to have failed the stress test. When eventually nearly passing the stress test, this group has a higher CAR compared to their peer group, and consequently, accepting H5 in this case.

Volatility

As H6 indicates, volatility of banks included in the stress test decreases due to decreased information asymmetry. For this analysis, a GARCH (1,1) with a binary variable for the event windows is computed. The resulting percentages of statistically significant positive and negative volatility changes are presented in Table 6. During the event windows of the 2010 stress test, a small percentage of the complete sample (4,32%) experienced an increase in volatility during the post-event windows. However, the significant volatility changes were mostly decreasing volatility in the pre-event windows for the full sample and the “Control Group”. For the treatment group, the decrease in volatility was the highest in the post-event window of 0-10 days, implying that the informational asymmetry decreased in this period as stock returns became less volatile. However, the volatility changes in the 2011 EU stress test are more in line with existing literature (Neuhierl et al. 2013), finding that there is an increase in volatility during the post-event windows of the full sample. Finally, the volatility surrounding the publication of the 2014 EU stress test results, indicates significant lower volatility between 10% and 20% for the control group, while the “Treatment group” only has a comparable amount of lower volatilities in the 0-5 pre-event window. Moreover, in around 10% of the stocks of the “Control Group”, the volatility increases in the post-event windows, while this 32,2% and 37,29% for the “Treatment group”, indicating a significant rise in the volatility after the publication of the stress test results in 2014.

(28)

28

Table 6: Structural break test in the GARCH (1,1) volatility. Percentage of stocks without stability in the variance equation

∅i > 0 (higher volatility) ∅i < 0 (lower volatility)

Event window: (-9; -1) (-5; -1) (0;5) (0;10) (-9; -1) (-5; -1) (0;5) (0;10)

2010 Stress Test:

2011 Stress Test:

2014 Stress test:

Full Sample 12,07% 12,64% 17,82% 21,26% 15,52% 21,84% 13,22% 9,77% Control Group 8,70% 6,09% 10,43% 13,04% 19,13% 22,61% 15,65% 10,43% Treatment Group 18,64% 25,42% 32,20% 37,29% 8,47% 20,34% 8,47% 8,47% Near Pass Group 0,00% 0,00% 0,00% 0,00% 0,00% 0,00% 0,00% 50,00% Pass Group 19,05% 4,76% 26,19% 30,95% 9,52% 28,57% 11,90% 6,78% Fail Group 7,69% 0,00% 61,54% 69,23% 7,69% 0,00% 0,00% 0,00%

SCAP/CCAR compared to EU stress test

(29)

29

with a larger interval between 2011 and 2014, the informational value of 2014 might be larger compared to the 2014 CCAR as, the CCAR is also performed in 2012 and 2013. Nevertheless, the overall CAR in the post-event windows of the SCAP and CCAR is less significant compared to the EU stress test post-event CAR, therefore accepting H7. Consequently, the FED could cooperate more intensively with the EBA to discuss the technicalities of the EU stress test, as it still provides informational value to the market.

Decreasing value over time

Finally, in H8, this research argues that the informational value of stress tests decreases over time, as the market becomes more efficient with the publication of earlier results. In the case of the SCAP and CCAR, the post-event windows with significant CAR occur primarily in the 2009 SCAP and 2012 CCAR, and are diminishing in the post-event windows moving forward to the 2015 CCAR. However, the pre-event windows, with exception of the 2014 CCAR, indicate significant CAR, implying an early incorporation of the results into the stock prices.

Comparably, in the EU stress test, a similar trend is occurring. Most of the significant post-event CAR occur in the 2010 and 2011 EU stress test. The most significant group in the 2014 EU stress test is the “Fail Group”. However, also this is the first time a “Fail Group” could be tested. While, on the other hand, significant CAR are apparent throughout all stress tests in the pre-event windows. This could hint that the market becomes more efficient by incorporating the new information prior to its publication, decreasing the informational value of the publication itself. Also, when reviewing the trends in Table 5, the group of banks that have passed all three EU stress test, do not have a significantly higher CAR compared to its peer group. Therefore, this research accepts H8. Therefore, policy makers could review their processes, as the publication of their results has a decreasing informational value to the market.

VII. Robustness:

(30)

30

event, this might cause spill-over effects and co-movements of the stock returns. Therefore, this research applies two correction methods suggested by literature; the cross section crude dependence adjustment (Brown & Warner, 1980) and the cross sectional average correlation method (Kolari & Pynnönen, 2010). Subsequently, this research calculates the significance level of the CAR’s based on these two parametric tests. However, as literature states that the abnormal returns may have a non-normal distribution (Corrado, 1989; Cowan 1992; Kolari & Pynnönen, 2011), a Chi-Squared test is designed to assess the distributions of the abnormal returns. As can be reviewed in Appendix B, a large proportion of the abnormal returns are non-normally distributed. Therefore, this research applies the non-parametric Cowan (1992) Generalized Sign Test and Rank test as a robustness check presented in Table 7.

The significance of the CAR in the EU stress test, are robust to the parametric test proposed by Brown & Warner (1985) and Kolari and Pynnönen (2010). Comparable

Table 7: Cumulative Abnormal Returns (CAR) EU stress test

CAR (-9; -1) CAR (-5; -1) CAR (0;5) CAR (0;10) 2010 Stress Test:

Full Sample 0,616% -0,855% 0,397% 0,996%

Control Group 0,268% -0,757%*** -0,137%* 0,493%

Treatment Group 1,469% -1,095%*** 1,703%* 2,225%***

Near Pass Group 1,418% -1,995%** 1,866% 1,862%

Pass Group 1,477% -0,938%*** 1,674%* 2,288%***

2011 Stress Test:

Full Sample -1,517%** -1,397%** 0,280% -0,260%

Control Group -0,589%* -1,202%** -1,082%** -1,246%** Treatment Group -3,799%*** -1,865%*** 3,544%*** 2,103%* Near Pass Group -6,531%*** -1,921%** 5,481%*** 3.530%***

Pass Group -3,043%** -1,734%** 2.985%*** 1,682%

2014 Stress test:

Full Sample 1,239% 0,947% -0,106% 0,061%

Control Group 0,692% 0,675% 0,965% 1,459%*

Treatment Group 2,346%** 1,509%** -2,198% -2,661%*

Near Pass Group 2,965%* 3,158%* 1,685% 2,734%**

Pass Group 2,446%** 0,845%* 0,282% -0,120%

Fail Group 1,832%* 3,146% -11,406%*** -12,531%***

(31)

31

significance levels are found with non-parametric testing compared to the parametric tests performed. Consequently, there is only a slight difference in the significance of the 2010 EU stress test. The non-parametric test accepts a higher amount of pre-event window CAR, which increases the validity of a decrease in the information value of the EU stress test. the non-parametric Cowan (1992) Generalized Sign Test and Rank test as a robustness check presented in Table 7. As this study focuses on the EU stress test, the robustness checks are applied to that sample.

Comparison to previous literature

When comparing the results of the 2010 and 2011 stress test, this research finds many similarities to the previous research on the EU stress test and the SCAP. When reviewing the statistical results of this study, comparable significance levels and CAR for the 2010 and 2011 EU stress test are found (Alves et al., 2015: Petrella & Resti, 2013). This builds the robustness of the results of the incorporated 2014 EU stress test. Subsequently, when the SCAP event windows used by Morgan et al. (2014), are applied to this database comparable significant CAR are found for the “Tested sample” and “Gap” and “No GAP” groups. However, the event window applied by Morgan et al. (2014) do not allow for in-depth testing of the informational value. Nevertheless, the CAR and significance when this window is applied, is comparable to previous research. Therefore, the results of the following CCAR are of a comparable quality compared to previous research.

VIII. Conclusive Discussion

Overall, this research confirms the informational value of the stress test to the stock market. As significant CAR are apparent in most post-event windows of the EU stress test. In 2011, the market displayed negative CAR in the pre-event windows while the post-event windows show significant positive CAR. In the 2014 pre- and post-event windows, this sequence is reversed, showing significant positive CAR in the pre-event windows and significant negative CAR in the post-event window both of these events are implications of market overreaction to the published news as previous literature indicates.

(32)

32

subsequent the CAR of this group. As expected by previous market reaction literature, this group shows the most significant and highest, although negative, CAR in this research. The maximum CAR identified in this stress test even resulted in -45% in the post-event window. The influence of this group is significant throughout the treatment group, as it displays insignificant CAR, while the control group actually has a positive significant CAR.

Another contribution to the existing literature, is that this research pioneers by comparing the EU bank stress test, to its US counterparts, the SCAP and CCAR. Moreover, little research has yet been conducted on the CCAR. The results of this paper, contribute by finding that the informational value to the market is higher in the EU bank stress test compared to the SCAP and CCAR. One possible explanation for this difference is the intermediate release of information to the objected bank in the CCAR. As a bank’s capital plans might be objected by the FED, the bank is informed of this prior to the publication of the results. Therefore, this information could flow into the market prior to the publication of the results, leading to more significant CAR in the pre-event windows. Another possible explanation for this difference is the time between the publications of the stress test results. As the EU stress test is performed in 2010, 2010 and 2014, the US SCAP is performed in 2009, and the CCAR from 2011-2015. The market could become more efficient based on prior publications of the stress test results, moreover, when these are published yearly.

This statement indicates that the stress tests would have decreasing value to the market over time. As more information flows into the market through supervisory stress testing, the market becomes better informed, decreasing the need for new information. This hypothesis is confirmed in this study by reviewing the occurrence of significant CAR throughout the stress tests in Europe and the United States. Subsequently, banks that pass all three EU stress tests, do not show higher CAR compared to the “Pass” group of the 2014 EU stress test.

(33)

33

of the 2014 stress test results that banks in the third quartile (with lower DE ratios) show a negative CAR in the post-event windows. Consequently, rejecting the assumption that higher risk leads to higher returns.

When analysing the possible trends over the three stress tests, some interesting results are investigated. As literature expects banks that move from “Near Pass” to “Pass”, to show a higher CAR, the opposite is occurring. As banks that move from “Near Pass” to “Pass”, actually show a lower (and even negatively significant) CAR in the event windows compared to their peer group. On the other hand, when banks receive a “Near Pass” twice, they show significantly higher positive CAR for the post-event windows, indicating an expectation of the market that this group was more probable to failing the stress test.

Finally, this study confirms prior research on the volatility in post-event windows following financial news. As only the 2010 EU stress test results in more decreasing volatility in the event windows, the opposite occurs in the event windows for the 2011 and 2014 tests, when a higher percentage of stocks experience an increase in volatility.

All in all, this study makes a significant contribution to the existing literature by investigating both the EU stress test and the US counterparts, the SCAP and CCAR. This study pioneers by comparing the results of the stress test and its informational value to the market. While previously, literature could not analyse trends occurring over time in stress test, this study includes the 2014 EU stress test and is able to pioneer in the analysis of trends occurring in the stress tests. Therefore, this study provides a building block for research to follow on the informational value of stress tests and stock market reactions to the publication of the stress test results.

An implication for further research is to assess whether the process of a stress test has an influence on the informational value for the market. As over time, the market becomes more efficient and the results are incorporated into the market prior to its publication, policy makers could review their processes to reduce the information asymmetry in the market.

(34)

34

Final implication for further research could be the creation of a prizing model around the publication of a stress test. As multiple stress tests in the EU and US have been conducted, it is interesting to review how the Carhartt 4-Factor model, especially the momentum factor, relate to the publication of the results and its subsequent returns.

As indicated, this research provides implication for policy makers in the EBA and FED. As the informational value of their stress test provides a decreasing value to the market, these supervisory could review their processes. This is especially applicable to the FED, where the publication of their results do not provide significant informational value to the market participants.

References:

Alves, C., Mendes, V., Pereira da Silva, P., (2015), Do stress tests matter? A study on the impact of the disclosure of stress test results on European financial stocks and CDS markets,

Applied Economics, 47(12), pp. 1213-1229.

Antweiler, W., Frank, M.Z., (2006), Do US stock markets typically overreact to corporate news stories? Available at SSRN 878091

Ball, R., Kothari, S.P., (1991), Security returns around earnings announcements, Accounting

Review, 66, pp. 718–738.

Barberis, N., Shleifer, A. Vishny, R., (1998), A model of investor sentiment, Journal of

Financial Economics, 49, pp. 307–343.

Baumann, U., Nier, E., (2004), Disclosure, volatility, and transparency: an empirical investigation into the value of bank disclosure. Economic Policy Review, 10(2), pp. 31-45.

(35)

35

Brown, S.J., Warner, J.B., (1980), Measuring security price performance, Journal of Financial

Economics, 8, pp. 205-258.

Brown, S.J., Warner, J.B., (1985), Using daily stock returns; The case of event studies, Journal

of Financial Economics, 14, pp. 3-31.

Corrado, C. J., (1989), A nonparametric test for abnormal security-price performance in event studies, Journal of financial economics, 23(2), pp. 385-395.

Cowan, A., (1992), Nonparametric event study tests, Review of Quantitative Finance and

Accounting, 2, pp. 343–358.

DeBondt,W., Thaler, R., (1985), Does the stock market overreact?, The Journal of Finance, 40, pp. 793–805.

Dyckman, T., Philbrick, D., Stephan, J., (1984), A comparison of event study methodologies using daily stock returns: A simulation approach, Journal of Accounting Research, 22, pp. 1-30.

Fama, E.F., (1970) Efficient capital markets: a review of theory and empirical work, The

Journal of Finance, 25, pp. 383–417.

Fama, E.F., MacBeth, J., (1973), Risk, Return and Equilibrium; empirical tests , The Journal of

Political Economy, 81(3), pp. 607-636.

Hirschhorn, E., (1987), The informational content of bank examination ratings, Federal

Deposit Insurance Corporation Banking and Economic Review, pp. 6–11.

Kolari, J., Pynnönen, S., (2011), Nonparametric rank tests for event studies, Journal of

(36)

36

Kothari, S. P., (2001), Capital markets research in accounting, Journal of Accounting and

Economics, 31, pp. 105–231

Kothari, S.P., Warner, J.B., (2007), Econometrics of event studies in: Ekbo, B.E., Handbook of

Corporate Finance: Empirical corporate finance, Amsterdam: Elsevier, pp. 5-36.

MacKinley, A.C., (1997), Event studies in Economics and Finance, Journal of Economic

Literature, 18(3), pp. 13-39.

Morgan, D.P., Peristiani, S., Savino, V., (2014), The Information Value of the Stress Test,

Journal of Money, Credit and Banking, 46(7), pp. 1479-1500.

Neuhierl, A., Scherbina, A., Schlusche, B., (2013), Market reaction to corporate press releases, Journal of Financial and Quantitative Analysis, 48(4), pp. 1207-1240.

Petrella, G., Resti, A., (2013), Supervisors as information producers: Do stress tests reduce bank opaqueness?, Journal of Banking & Finance, 37, pp. 5406-5420.

Pettway, R., (1980), Potential insolvency, market efficiency, and bank regulation of large commercial banks, The Journal of Financial and Quantitative Analysis, 15, pp. 219–236.

Quijano, M., (2014), Information asymmetry in US banks and the 2009 bank stress test,

Economics Letters, 123(2), pp. 203-205.

Tetlock, P. C., (2010), Does public financial news resolve asymmetric information?, Review of

Financial Studies, 23(9), pp. 3520-3557.

Xiaolou, Y., Woodlock, P., (2011), What drive banks’ abnormal returns?, Banking & Finance

(37)

37 Official stress test documents:

EU stress tests2:

CEBS, (2009), CEBS’S Press Release on the Results of the EU-wide stress testing exercise CEBS, (2010), Aggregate outcome of the 2010 EU-wide stress test exercise coordinated

by CEBS in cooperation with the ECB

EBA, (2011), European Banking Authority 2011 EU-wide stress test aggregate report EBA, (2014), Results of 2014 EU-wide stress test; Aggregate results

US stress tests3:

FED, (2009), The Supervisory Capital Assessment Program: Overview of Results FED, (2011), Comprehensive Capital Analysis and Review: Objectives and Overview

FED, (2012), Comprehensive Capital Analysis and Review 2012: Methodology and Results for

Stress Scenario Projections

FED, (2013), Comprehensive Capital Analysis and Review 2013: Assessment Framework and

Results

2

All available at: http://www.eba.europa.eu/risk-analysis-and-data/eu-wide-stress-testing/ [accessed 22.01.2015]

3

(38)

38

Appendix A:

Groups EU 2014:

Treatment Group: Pass: Near Pass: Fail:

Erste Group Uncredit Erste Group Allied Irish RBS Dexia

Dexia Intesa Sanpaolo KBC Group Uncredit Lloyds Bank of Greece

KBC Group Banco Popolare Commerzbank Intesa Sanpaolo IKB Bank Piraeus

Hellenic Bank UBI Deutsche Bank UBI Mediobanca Eurobank Ergasias

Commerzbank Mediobanca Aareal Bank Credito Emiliano Banco Popolare

Deutsche Bank Banca ppo Emilia Sydbank Bank of Valletta Banca ppo Emilia

Aareal Bank Banco ppo Milano Danske Bank ING Banco ppo Milano

IKB Bank Banco Carige Jyske Bank DNB Nor Banco Carige

Sydbank Banco ppo di Sondrio BBVA Getin Nobile Banco ppo di Sondrio

Danske Bank Credito Emiliano Caixabank Handlowy Banco Piccolo

Jyske Bank Banco Piccolo Banco Sabadell Bank BPH Monte dei Paschi

BBVA Bank of Valletta Banco Popular Alior BCP

Caixabank ING Bankinter BOS Hellenic Bank

Banco Sabadell DNB Nor Banco Santander PKO

Banco Popular Getin Nobile BNP Paribas Espirito

Bankinter Handlowy Credit Agricole Banco BPI

Banco Santander Bank BPH Societe Generale Nordea

BNP Paribas Alior Alpha Bank SHB

Credit Agricole BOS HSBC SEB

Societe Generale PKO Barclays Swedbank

HSBC Espirito OTP

Barclays BCP Bank of Ireland

RBS Banco BPI

Lloyds Nordea

Bank of Greece SHB

Piraeus SEB

Eurbank Ergasias Swedbank

(39)

39

Appendix A2:

Groups EU 2011:

Treatment Group: Pass: Near Pass:

Erste Group Bank of Ireland Erste Group Bank of Valletta Banco Sabadell

Dexia Allied Irish Dexia ING Banco Popular

KBC Group Uncredit KBC Group SNS Bankinter

Bank of Cyprus Intesa Sanpaolo Bank of Cyprus DNB Nor Piraeus

Commerzbank Banco Popolare Commerzbank PKO Banco Popolare Deutsche Bank UBI Deutsche Bank Banco BPI Espirito Landesbank Berlin Bank of Valletta Landesbank Berlin Nordea BCP

Sydbank ING Sydbank SHB

Danske Bank SNS Danske Bank SEB

Jyske Bank DNB Nor Jyske Bank Swedbank

BBVA PKO BBVA Nova Kreditna

Banco Sabadell Espirito Banco Santander Monte dei Paschi

Banco Popular BCP OP Pohjola

Bankinter Banco BPI BNP Paribas

Banco Santander Nordea Credit Agricole

OP Pohjola SHB Societe Generale

BNP Paribas SEB HSBC

Credit Agricole Swedbank Barclays

Societe Generale Nova Kreditna RBS

HSBC Monte dei Paschi Lloyds

Barclays Bank of Greece

RBS Alpha Bank

Lloyds OTP

Bank of Greece Bank of Ireland

Piraeus Allied Irish

Eurbank Ergasias Uncredit

Alpha Bank Intesa Sanpaolo