The Effect of a Downgrade on a Bank’s
Share Price
Abstract:
Credit rating agencies have endured a lot of criticism during the recent financial crisis. The criticism pointed to the matter that credit rating agencies were too slow in downgrading banks. This paper studied when the correction of the share price of a bank occurs in relation to a downgrade. An event study was conducted and the parameters of periods prior to the
downgrade were tested against parameters of periods after the downgrade. This paper finds that the correction of the share price mainly occurs on the day of downgrade but also in the period before the downgrade and more specific shortly before the downgrade.
Name: Max Engbers
Student number: 10002139
Supervisor: Timotej Homar
Table of content
1. Introduction ……… 3
2. Related literature …... 3
3. Data collection and description ……… 6
4. Results and discussion ……….. 11
5. Conclusion ………. 21
1. Introduction
In the recent financial crisis there has been a lot of economic trouble: governments had to intervene in their country’s economies, firms went bankrupt, governments had to
restructure financially to prevent bankruptcy. Criticisms from all kinds of origins were argued as to how, and more specific how the banking sector, got the economy into this recession.
One kind of criticism targeted the credit rating agencies. Among other criticism economists argued that they were too slow in downgrading banks in their credit rating (Rom, 2009). The explanation for this criticism is that the market already knows a bank’s
creditworthiness has deteriorated but the credit rating agencies haven’t corrected for this yet. It can be expected that this should be reflected upon in the share price of the bank, because if the market already knows a bank’s creditworthiness has deteriorated it may be expected that its share price also deteriorates. Although Manso (2013) refutes this and suggests that credit rating agencies aren’t slow in terms of downgrading there is reason to believe that the market knows a bank’s characteristics have already deteriorated. Boot et al. (2006) and Bannier & Hirsch (2010) argue that a firm, or for this paper a bank, is put ‘on watch’ or on a watchlist before a downgrade. This could explain why a share price is corrected before a downgrade occurs. It could also be that the market knows the bad news about a bank’s creditworthiness even before the bank is being put ‘on watch’, so that the share price is corrected even before there is any news of the bank being put ‘on watch’.
This study will examine when the effect of downgrade of a banks on its share price occurs and test whether the effect in the period before the downgrade differs from the period afterwards. This will be tested by conducting event studies with various time windows and performing several Wald-tests. The abnormal returns of a variety of banks will be regressed on dummy variables that indicate a period before the downgrade and a dummy that indicates the day of the downgrade itself and the period afterwards. Different time windows will be used to discover a possible pattern. Also the influence of the financial crisis will be tested.
This paper continues with the discussion of related literature in section 2. Section 3 explains the method of collection of the data and also describes the collected data. Section 4 presents and discusses the results and section 5 concludes this paper.
2. Related literature
In the research of Boot, Milbourne and Schmeits (2006) the authors suggest that the financial economics literature is disagreeing on whether or not the credit ratings are of
importance. Boot et al. (2006) contribute to this discussion by refuting that credit ratings are not important. They propose a different way to look at credit ratings by presenting “an
understanding of the way credit ratings come about and the role credit ratings actually play in financial markets”. In practice there is often contact between a credit rating agency and the management of the firm when a credit rating agency observes changes in the firms
characteristics. After contact with the credit rating agency the firm implicitly engages in a series of actions to avoid being downgraded by the credit rating agency (the recovery effort). Meanwhile the firm is put ‘on watch’ by the credit rating agency in a way that the market can observe this to and reacts to it. When the recovery effort is made by the firm, the credit rating agency then communicates to the market whether the recovery effort was effective or not. So new information is given to the market and the market reacts to it.
Boot et al. (2006) refer to past literature on the effect of bond downgrades on security prices. The research of Hand, Holthausen and Leftwich (1992) and of Goh and Ederington (1993) provide support for the hypothesis that a downgrade has a significant negative effect on the stock price of a company.
Bannier & Hirsch (2010) conducted a study about the effect of the introduction of the watchlist. They compared the effects of the introduction of the watchlist on credit rating changes. Which relates to the research in this paper in a way that if the watchlist is of
significant influence on credit rating effects it is possible that the effect of a rating change on the share price is occurring in the time prior to a downgrade. Bannier and Hirsch use credit rating changes from Moody’s from 1982 until 2004. For the first part of their research the authors use both downgrades and upgrades. They investigate the size (single or multiple notches), distribution and number of rating changes and find significant evidence that the effect of credit rating changes on the market value of firm equity is stronger in the post-watchlist era than before the post-watchlist era. The results indicate that credit ratings are more informative since the introduction of the watchlist. In the second part of their research they provide a multivariate framework, in which they only use downgrades, and where they test the effect of the introduction of the watchlist by adding a dummy variable that indicates if a downgrade is done before or after the introduction of the watchlist. With this model they find that the effect of a downgrade is only statistically significant after the introduction of the watchlist. Furthermore, the results show that the information content of downgrades has a much stronger effect after the introduction of the watchlist.
Becker and Milbourne (2011) argue that the increased competition among credit rating agencies (Fitch started to compete with S&P and Moody’s) did not specifically have positive
effects on the quality of credit ratings: they claim that rating levels went up, the correlation between credit ratings and market-yields fell and the ability of credit ratings to predict defaults worsened. If the quality of credit ratings fell one could argue that this should reflect in the correction of the share price in a manner that the effect of the downgrade on the share price is weak during the days after the downgrade.
Another research that indicates that the effect of a downgrade is larger in the period prior to the downgrade is the study of Güttler and Wahrenburg (2007). Their research concerns the issue of biases in credit ratings and in lead-lag relationships for near-to-default issuers. They examined changes in ratings of firms that subsequently defaulted, using a database that contains defaulted firms in the period of 1997 until 2004. Güttler and
Wahrenburg used data on the long-term, senior unsecured and historical data of the watchlist. The authors had data on 407 companies that announced default on the previously mentioned types of credit ratings. For the study only credit ratings of Moody’s and S&P were used. Further, they used different types of indicators on creditworthiness and macro economical variables. With this data it was found, among other findings, that if a company gets
downgraded, a subsequent downgrade is of a greater magnitude in the short term. The authors also found that if a company gets downgraded harshly by one rating agency it gets even harsher downgraded in the same direction by the second agency. Additionally, they find serial correlation in rating changes up to 90 days subsequent the rating change of interest after controlling for rating changes by the other rating agency. These findings could all apply to the data of this, since the effect of prior downgrades in the time window of another downgrade is also covered in this paper.
The findings of Manso (2013) refute the by the author called, recent criticism on credit rating agencies. According to Manso the recent criticism on credit rating agencies regarded, among others, the issue of credit rating agencies being biased in advantage of the borrower and they were too slow in downgrading companies with deteriorated quality of credit. That last matter might indicate that the majority of the effect of a downgrade on share prices takes place in the period after a firm is being downgraded. In his research Manso uses a
performance-sensitive-debt (PSD) model. One of his findings is that in this PSD model soft-rating-agency equilibrium is better in terms of welfare than a tough-soft-rating-agency
equilibrium. Also Standard & Poor’s argues that what would happen to a company if a company gets downgraded is not determinant for the downgrading of that company.
All the researches introduced until now are not bank specific. However, the research of Morgan (2002) addresses the difference between banks and most firms. The author points
out that banks are more opaque than any other firm except for insurance companies. Morgan argues that the most basic difference between a bank and another firm is a bank’s financial nature. Non-banking firms hold a considerably bigger relative amount of physically fixed assets than a bank does. Because of the large amount of financial assets a bank holds, investors that invest in banks are generally more uncertain about the underlying collateral than an investor of any other firm (again, insurance companies aside). The financial assets a bank holds, brings agency problems to the table as well. Among their assets a bank has loans to smaller borrowers. These borrowers are not as transparent as bigger borrowers so they need to be monitored. But this brings a disadvantage; monitors might not be very transparent themselves. What follows is that a bank loans to an opaque borrower and so the bank
becomes more opaque itself. Words from analysts of Moody’s about this matter are that with holding bank securities come unrecognized asset quality problems. Because there are no guidelines for handling this an analyst has to simply judge upon the impression he or she gets. One could reason that the analysis of a rating is not that clear and because the market
recognizes this, the market may not wait for a credit rating change by an agency and correct the share price itself before the actual rating chance occurs.
These studies all have their own theories on how the market reacts to information given by credit ratings. This paper examines the point in time that the market reacts to information given by credit ratings.
3. Data collection and description
For the event study conducted in this paper different datasets were used. One dataset is the dataset that was collected using the Thomson One database. In Thomson One the data on credit ratings of the 25 largest banks of the US and the 25 largest banks of Europe was collected. The credit ratings of S&P, Moody’s and Fitch were used, as their ratings add up to 95 percent of all credit ratings (Becker & Milbourn, 2011). Only the subordinated, long-term issuer, bank financial strength (Moody’s only) and bank individual (Fitch only) ratings were used. In reality it turns out that the downgrades of these kinds of credit ratings coincide with other kinds of credit ratings. To determine the 25 largest banks in the US, FED data
containing the total amount of deposits outstanding was used1. Some banks were from
European origin so I chose to ignore these banks in the US bank list. Other banks that weren’t used in the dataset were banks that were a subsidiary of another bank that was already in the
1 See reference list: Federal Reserve Statistical Release
dataset or banks that had no data on credit ratings and/or stock prices. Similar documents as for the US banks of two financial web pages were used for the European banks2. Two sets of European banks were combined to compose a list of the 25 largest banks of Europe. Since the list will not be very inaccurate, this won’t have a big impact on the results. After all data on credit ratings for each bank was collected all downgrades were paired to the accurate date. All data before 2005 was dropped, because this paper focuses on a more recent period so that calculating variables like beta-values is more reliable. This summed up to 429 downgrades in the dataset. In comparison if the data before 2005 weren’t dropped there would have been 531 downgrades, so the period after 2005 contains far more downgrades.
Subsequently, the stock prices of the banks were collected. Via Wharton Research
Data Services (WRDS) access was gained to CRSP Daily Stock and the stock prices were
requested stock for each US bank from the 1st of January 2005, until the 31st of December 2012, since CRSP didn’t have data after December 2012. For the banks that were founded after 2005 stock price data from the first available date until 2012 was collected. Since there wasn’t an accessible database that had data on European stock markets, those stock prices were collected from a different source. Hence, for the European stock prices ‘Yahoo Finance’ and ‘Google Finance’ were consulted. The collected data ranged from January 2005 until December 2013 for as far as possible. For the US benchmark the S&P 500 was used and for the European benchmark data on the S&P European 350 was used. Both benchmarks were accessed through ‘Yahoo Finance’. Market indices from January 2005 until December 2013 were collected. For the risk-free rate return access was gained to the Fama-French database and the daily returns of the one-month US Treasury note was collected.
Following, (after removing all data before 2005) the datasets were merged on date and other dataset-specific key variables. Subsequently, the variables relevant for this paper were produced. First, the stock returns for each bank were produced and outliers were replaced by ‘winsorizing’. These outliers were mostly caused by emissions of shares or some odd and unexplainable values in the dataset. Winsorizing the share returns was done by replacing the values smaller than the 1st-percentile by the value of the 1st-percentile and replacing the values larger than the 99th-percentile by the value of the 99th-percentile. Second, the dummies
indicating the time windows were generated. The dummy variable MinXdays indicates the period of X days prior to the downgrade, the dummy variable PlusXdays indicates the day of the downgrade and the X days after the downgrade and the dummy variable DaysX contains
2 See reference list: SNL Financials and Standard & Poor’s Research
the time periods of MinXdays and PlusXdays combined. Thereafter, the beta-values were calculated by regressing a bank’s return on the market return, saving the regression coefficient and using them as betas. Next, the betas were winsorized in the same manner as the stock returns. With the risk-free rate return, for which the one-month US Treasury note returns were used, the expected returns were subsequently calculated via the CAPM formula for each bank from 2005 until 2012 or 2013 depending on the specific bank. The expected returns were also winsorized as they showed some unexplainable high and low values. The risk-free rate return wasn’t winsorized as winsorizing it didn’t imply any changes to the observations of the variable. Finally, the abnormal returns were calculated and winsorized for the same reason that the expected returns were winsorized. Odd values for abnormal and/or expected returns could be caused by the betas since these were calculated for the whole period between 2005 and 2012 or for some banks even 2013 and were used to estimate daily returns. To allow measuring the difference of the effect of a downgrade on the bank’s share price in crisis time and in better economic times, the dummy variable Crisis was generated. The dummy variable is equal to 1 for all observations from the 15th of September 2008 and on and 0 for all other observations. This date was indicated as the fall of Lehman Brothers, because it filed for bankruptcy3 and from that point on governments acknowledged the need for action. Finally, the variables PriorDowngradeMinX and OverlapMinX were generated. These variables are presented in table 1 and will be discussed in section 4.
Presented below are some tables to describe the collected data. Table 1 gives
information about the variables that were used in the regressions, table 2 displays the number of downgrades per country per year and table 3 displays some numerous information about variables that were either used in the regressions or that were used to calculate important variables. Table 2 shows that America has the most downgrades overall and that in 2009 the most downgrades occurred. Although, America has the most downgrades in table 2 can also be seen that as a continent Europe has the most downgrades as opposed to North America. Table 3 indicates, among other, the 1st-percentille and the 99th-percentille for every variable, since all variables are winsorized these values correspond with the minimum and maximum of the variable.
3 See reference list: United States Bankruptcy Court
Table 1: Variables and their descriptions
Table 2: Number of Downgrades per Country per Year
Country 2005 2006 2007 2008 2009 2010 2011 2012 2013 Total America 3 3 6 40 56 13 21 12 5 159 Belgium 0 0 0 2 3 1 2 1 0 9 Canada 0 1 1 0 2 2 1 0 2 9 Denmark 0 0 0 0 4 0 4 2 0 10 France 0 0 3 4 8 4 12 10 1 42 Germany 0 1 3 8 9 1 9 9 3 58 Italy 2 0 2 5 5 6 10 15 9 43 Netherlands 0 0 0 0 9 0 0 0 1 10 Russia 0 0 0 0 3 0 0 0 2 5 Spain 0 0 1 0 2 1 7 12 0 23 Swenden 0 0 2 3 4 0 1 5 2 17 Switzerland 0 0 0 1 1 0 1 0 2 5 UK 0 0 1 6 15 4 7 6 1 40 Total 5 5 19 72 121 32 75 72 28 429
Variable Description, (formula)
DayOfDowngrade Dummy variable that indicates the date of
downgrade of a bank.
MinXdays Dummy variable that indicates the number
of X days prior to a downgrade.
PlusXdays Dummy variable that indicates the number
of X days after.
DaysX
Dummy variable that indicates the time period of MinXdays and PlusXdays combined.
AbnRet Daily abnormal return,
(Abnret=BankRet-ExpRet)
Crisis Dummy variable that indicates if the
observation is in time of crisis or not.
PriorDowngradeMinX
Dummy variable that indicates the number of downgrades prior to the measured downgrade in the of X days before the measured downgrade
OverlapMinX
Dummy variable that measures the effect of overlap between time windows of two separate downgrades if days prior to the second downgrade is X.
Table 3: Descriptive statistics
Variable Observations Mean Std dev. 1
st -percentille 99th -percentille Bank return 103,906 -.0001826 .0254449 -.0801001 .0736515 Risk-free rate return 99,938 .0000652 .000077 0 .00022 Beta 103,934 1.057681 .3363067 .3950048 1.599049 Market return 100,831 .0001611 .0141367 -.0494037 .0440021 Abnormal Return 98,410 -.0003317 .0195809 -.0617321 .06164
4. Results and discussion
Table 4 presents the results of a variety of separate regressions. All regressions include a constant.
Table 4: Different time windows regressed on AbnRet
Table 4. *=significant at 1%-level **=significant at 5%-level ***=significant at 10%-level Within the parentheses are the standard deviations
In regression (3) the dependent variable AbnRet is regressed on Min10days and Plus10days. The regression indicates that the 10 days prior to the downgrade have a negative effect of .0003916 on abnormal returns. With a standard deviation of .0003346 this coefficient is not significant at any level. As for the 10 days subsequent to the downgrade the regression shows that this window of time has a positive effect of .0000698 with a standard deviation of
.0003354. This coefficient also shows no sign of significance at any level. The effect is
Dependent variable: AbnRet Independent variable (1) (2) (3) (4) (5) (6) (7) (8) Constant -.0002857* (.0000636) -.0002867* (.0000637) -.0003086* (.0000646) -.0003079* (.0000648) -.00029* (.0000664) -.0002877* (.0000668) -.0002367* (.000068) -.0002581* (.0000687) DayOfDowngra de -.0028612* (.0010145) -.0029853* (.0010146) -.0029511* (.0010145) -.0028531* (.0010144) Min5days -.0011882* (.0004595) Plus5days -.0006616 (.0004573) Days11 -.0011359* (.00032) Min10days (.0003346) -.0003916 Plus10days .0000698 (.0003354) Days21 -.0003333 (.0002427) Min20days -.0002899 (.0002541) Plus20days (.0002489) -.0001672 Days41 -.0003463*** (.0001874) Min30days -.0010287* (.0002132) Plus30days .0001675 (.0002131) Days61 -.0004242* (.0001649) R2 0.0068 0.0002 0.0001 0.0000 0.0001 0.0000 0.0003 0.0001 #Obsvervations 98410 98410 98410 98410 98410 98410 98410 98410
smaller than the effect of the ten days prior to the downgrade, -.0003916% versus .0000698%. If AbnRet is regressed on Days21, which combines the time windows of Min10days, DayOfDowngrade and Plus10days, then the regression shows that this 21-day time period has a negative effect of -.0003333 on AbnRet with a standard deviation of 0.0002475. Days21 is insignificant even at any significance level.
So what can be concluded from these two regressions is that the effect of a downgrade on the abnormal return for this time window is insignificant. This could be caused by a variety of reasons. It could be that the effect of the two time windows combined, ten days prior until ten days after, is not significant because not all of the correction in the stock price is captured in this period of time. It might be that this time period only captures a part of the correction that occurs. Or on the contrary, the time window might be too long. In this case the time window would capture more than just the correction. It would also capture observations of some time in which none of the correction takes place so that the abnormal returns of the corrections get evened out.
To test these possible explanations, a variety of time windows were used. Table 4 shows these time windows. The dummy variables DayOfDowngrade Min5days and Plus5days, Days11, DayOfDowngrade, Min20days and Plus20days, Days41,
DayOfDowngrade, Min30days and Plus30days and Days61 were regressed on AbnRet in the same manner as the first and second regression.
In regression (1), where AbnRet is regressed on Min5days, DayOfDowngrade and Plus5days, it is shown that the five days prior to the downgrade have a clear significant and negative effect on the abnormal return. The effect is significant at a 1% level. Plus5days, although negative, the effect isn’t significant. The negative sign of both coefficients can mean two things. One meaning is that it has a reducing effect on abnormal returns. Another
explanation is that it intensifies abnormal returns negatively. Looking at the related literature, discussed in section 2, the latter explanation is more likely. Regression (1) suggests that the correcting effect of a downgrade on the share price mostly occurs in the period before the downgrade, hence the effect of the variable indicating the period before the downgrade is greater and more significant compared to the effect of the variable indicating the effect after the downgrade.
The three variables combined generate a negative effect of -.0011359. With a standard deviation of .00032 the coefficient of the variable Days11 is significant at a 1% level.
Regression (5) ascertains no significance of the variables Min20days or Plus20days at any level. Again, just as with the earlier regression, the coefficients are small.
On the contrary, in regression (6) the combined time window, expressed as Days41, is significant at a 10% significance level. To find out if there is a pattern to be discovered there are two additional regressions.
Regression (7) shows that the 30 days prior to a downgrade have a significant negative effect of -.0010287 on abnormal returns, at a 1% level. The 30 days after a downgrade show no significant effect on abnormal returns, hence there is no sign of significance of a correction in the share price. However, taking a longer period of time also has consequences for the results of the regression. An advantage of a short time window is that other effects than the effect of the downgrade have a limited probability of affecting the share price, thus the abnormal returns. The probability of capturing other effects than the effect of the downgrade rises when extending the time window. So although the result of Min30days is significant it could capture effects that aren’t related with the downgrade of the bank. Opposing, one could reason that the effect of the downgrade isn’t captured in a time window of twenty days. The market could already know for a while that a bank is less creditworthy without any
downgrade of a credit rating agency. The question is for how long does the market know that a bank is less creditworthy? And even if this is known and the right time window is captured, the regression isn’t protected against effects other than the effect of a downgrade.
In regression (8) the two periods are combined again and regressed. The regression shows a significant, at a 1% level, negative effect of .0004242.
Finally in all regressions discussed above the variable DayOfDowngrade is significant at a 1%-level and has a stronger effect than any other variable in that same regression.
If we take a look at the data produced so far we can discover a pattern. In every separate regression produced the variables for days prior to the downgrade and the day of downgrade are systematically stronger and more significant compared to the variables indicating the days subsequent to the downgrade. This then indicates that the correction of a downgrade in the share price occurs stronger and mostly before the downgrade itself. This finding is in line with the articles of Boot et al. (2006) and Bannier & Hirsch (2010). Boot et al. (2006) suppose that a firm is put ‘on watch’ before it gets downgraded. The market anticipates this before the actual downgrade and the share price of the bank is corrected. Subsequently, in some cases, the downgrade occurs but the share price is already corrected and the effect after the downgrade is weak. This is also a finding in this paper. Secondly this paper finds that the effect on the day of downgrade is also negative and significant.
The paper of Bannier & Hirsch (2010) studies the effect of the introduction of the watchlist. The findings of this paper indicate that the watchlist plays a role in the process of the correction of the share price for a downgrade. The correction occurs mostly and stronger in the days prior to a downgrade instead of after the day of the downgrade. If a bank is put on the watchlist by a credit rating agency, the market will react to this and share prices will drop.
Considering the results produced in different time windows, there is no obvious pattern to discover. Starting from the smallest time window and working our way up we cannot state that results are getting more significant or larger. Although the time window of 11 days (5 days prior until 5 days after the downgrade) captures stronger and more significant effects than the time window of 61 days, the time windows of 21 days and 41 days show weaker and less significant effects. Even if the subsections of both time windows show less significant and weaker results. The other way around, starting from the largest time window and working our way down, there isn’t a pattern to discover either. Even though the time window of 11 days, and the subsections the day of the downgrade and the period of 5 days prior to the downgrade, are stronger and more significant than any other time window, there is no clear correlation of pattern in the results that indicates a stronger effect as the time window reduces. A clear correlation appears when Min30days is excluded form the results. The shorter the days prior to the downgrade gets the stronger and more significant the effect becomes, with as shortest, strongest and most significant the day of downgrade itself.
But there is a possibility that the time windows taken into consideration thus far are not the correct ones. It might be that the effect of a downgrade on the share price takes place longer before the estimated time windows. S&P has monthly reviews of companies/banks that are put ‘on watch’ and these companies/banks have worsened in creditworthy before they are put on watch.
Table 5: Longer time windows regressed AbnRet
Dependent variable: AbnRet
Independent variable (1) (2) Intercept -.0001868 (.0000671)* -.0001474 (.000068)** DayOfDowngrade -.0027325 (.0010146)* -.0026688 (.0010147)* Plus5days -.0005127 (.0004584) -.0004643 (.0004585) Min40days -.0009989 (.0001898)* Min50days -.0011069 (.000176)* #Observations 98410 98410
Table 5. *=significant at 1%-level **=significant at 5%-level ***=significant at 10%-level Within the parentheses are the standard deviations
So it is established now that a long time window generates a strong and significant effect. This might be, at least partly, caused by a bank being put on watch or it is caused by effects that are not or less related to the downgrade of a bank. Fact of the matter is that if the time window is divided in days prior and days after, excluding day of downgrade, it is obvious that the longer days prior the stronger and more significant the coefficient becomes. This could indicate that the correction in the share price takes place longer before the
downgrade. This could be due to the effect of a bank being put ‘on watch’.
To analyze the robustness of the regressions, Table 6 below, presents the number of downgrades that were measured within different time windows prior to the downgrade being analyzed. It is shown that if days prior to the downgrade is increased the number of
downgrades prior to the analyzed one increases considerably. It means that the effect of days prior to the downgrade is influenced by a prior downgrade of which the effect is not taken into account in the regression. This is especially the case with 20 days or more prior to the downgrade as the measured prior downgrades add up to 25% or more of the total amount of downgrades. These observations are a footnote to the found results in the regressions.
Table 6: Summary of Second Downgrades
Variable #Observations Percentage of total
sample (429 obs.) PriorDowngradeMin5 62 14.45% PriorDowngradeMin10 82 19.11% PriorDowngradeMin20 108 25,17% PriorDowngradeMin30 134 31,24% PriorDowngradeMin40 164 38,23% PriorDowngradeMin50 187 43,59% Table 6
Additionally, presented in table 7 are regressions that test the effect of an overlap of time windows of two separate downgrades. The variable OverlapMin5 measures the effect of
the overlap of the 5 days after the first downgrade and the 5 days prior to the second downgrade, OverlapMin10 measures the effect of the overlap of the 5 days after the first downgrade and the 10 days prior to the second downgrade and so on. The results in table 7 show that OverlapMin5 and OverlapMin10 do not have a significant effect. When the number
of days prior to the second downgrade is increased to 20 and 30 the effect becomes stronger and significant. This is simply explained by the fact that the increased days measured prior to
the second downgrade now also measure more days after or maybe even prior to the first downgrade.
Table 7: The Effect of Overlap of time windows Independent variable (1) (2) (3) (4) Constant -.0002974* (.0000631) -.0003063* (.0000637) -.0003002* (.0000646) -.0002232* (.0000658) DayOfDowngrade -.0029138* (.0010139) -.0029782* (.0010139) -.0029543* (.0010139) -.0027866* (.0010139) Min5days -.0012131** (.0004862) Min10days -.0003961 (.0003477) Min20days -.0001675 (.0002604) Min30days -.0008959* (.0002158) OverlapMin5 -.0001051 (.0014436) OverlapMin10 .0001203 (.0011864) OverlapMin20 -.002214** (.0009959) OverlapMin30 -.0018917** (.0008696) #Obsvervations 98410 98410 98410 98410
Table 7. *=significant at 1%-level **=significant at 5%-level ***=significant at 10%-level Within the parentheses are the standard deviations
Since this paper focuses on the effect of a downgrade on the share price captured in a time window the regressions do not take into account variables that potentially have an effect on abnormal returns. If the amount of time is kept small the chance that anything else but the downgrade will affect the share price of the bank is limited. By enlarging the time window this chance increases. So although some caution is needed when drawing conclusions from results of a larger time window, this paper indicates that there that the correction in the share price caused by a bank being put on a watchlist occurs already about thirty days or more prior to the downgrade. However, this needs to be studied in more detail. The most important finding of this study is that the correction in the share price partially occurs in the period of 5 days prior the downgrade followed by a further correction on the day of downgrade itself. However, in the period following the day of downgrade further correction was not measured.
The R-squares of the regressions are very small. A possible reason for this is the lack of control variables. Since this is an event study, these control variables go beyond the scope of this paper. For this type of study it is not unusual not to take control variables into account. See for example, the study of Galil & Soffler (2011), the study of Hull, Predescu and White (2004) and the study of Norden & Weber (2004).
4.2 Wald-test
To test whether the variable that indicates the days prior to a downgrade statistically differ from the day of the downgrade and the days subsequent this paper proposes a Wald-test. The Wald-test tests the parameter of interest (days prior to downgrade) against a
proposed parameter (day of and after downgrade). The test uses a chi-squared distribution and it is calculated according to the next formula:
Wald-test = (βparameter of interest-βproposed parameter) / variance(βparameter of interest)
Presented below (see table 8) are the values for the Wald-tests between the parameters of the same time window and the corresponding regressions. It is shown that the variables Min5days, Min30days, Min40days and Min50days differ significantly from the variables that indicate the 5 days subsequent. The reason why time windows with different lengths are tested against each other is that it is shown in previous regressions that the time windows longer than 5 days after the downgrade do not have a significant and not even a negative effect. The shorter period after the downgrades is also more convenient to work with, because the chance of a second downgrade in the period after the first downgrade reduces.
Table 8: Wald-tests and their regressions
Independent variable (1) (2) (3) (4) (5) (6) Constant -.0002857* (.0000636) -.0002945* (.0000641) -.0002889* (.0000651) -.0002143* (.0000661) -.0001868* (.0000671) -.0001474** (.000068) DayOfDowngra de -.0028612* (.0010145) -.0029227* (.0010145) -.0029209* (.0010145) -.0027818* (.0010144) -.0027325* (.0010146) -.0026688* (.0010147) Plus5days (.0004573) -.0006616 (.0004577) -.0006938 (.0004578) -.0006906 (.0004579) -.0005669 (.0004584) -.0005127 -.0004643* (.0004585) Min5days -.0011882* (.0004595) Min10days -.0003515 (.0003343) Min20days -.0002822 (.0002531) Min30days -.0009816* (.0002109) Min40days -.0009989* (.0001898) Min50days -.0011069* (.000176) Wald-test 0.0100 0.1622 0.1515 0.0000 0.0000 0.2040 #Obsvervations 98410 98410 98410 98410 98410 98410
Table 8. *=significant at 1%-level **=significant at 5%-level ***=significant at 10%-level Within the parentheses are the standard deviations
4.3 Effect of the crisis
To study the effect of the crisis on a downgrade on a bank’s share price, two
regressions were made for every model. In the first regression of each model the variables of the specific time window and a constant are regressed on abnormal returns if the dummy variable Crisis is equal to one. In other words effective for all downgrades after the 15th of September 2008. In the second regression of each model the variables of the specific time window are regressed on abnormal returns if the Crisis is equal to zero. Thus, containing all observations before the 15th of September 2008.
There are a few differences between the two periods. When in crisis, the largest time window has a more significant effect on the days prior to the downgrade and in the combined time window. For the smallest time window the days prior to the downgrade are more
significant in times of financial prosperity. Also the time window of 21 days and the 10 days after a downgrade are more significant when there isn’t a crisis. This could be caused that in times of economic prosperity the market doesn’t see the downgrade coming. The time
window of 61 days and 30 days prior to a downgrade is stronger and more significant in times of a financial crisis. One could imagine that in times of financial crisis the of effect of a downgrade on the share price occurs longer before the downgrade itself compared to times of no financial crisis. People are a lot more skeptical and cautious in times of crisis so the
confidence in the economy and especially in banks is less people are more suspicious before a downgrade. Another explanation of a strong effect in the largest time window lies in a
subsequent downgrade. It is likely that after some period of time a bank is downgraded for a second time by the same or by another credit rating agency, in times of crisis. Or it could be that after a bank is downgraded, another bank in the same country, i.e. on the same market is downgraded and this reflects in the share price of the bank. This is the so-called ‘domino effect’. This could be reflected in the values of the coefficients of the largest time window.
For the data on the crisis the pattern seems to be that in times of crisis the correction in the share price starts earlier, because Min30days is significant in times of crisis and not in times of economic prosperity. This is a logical consequence of financially uncertain times. People are more skeptical and tend to sell themselves out once there is a hint of negative news about a bank, which then reflects in the share price of the bank. In times of no financial crisis the emphasis seems to lie on the days after a downgrade.
Table 9: Effect of the Crisis Dependent variable: AbnRet Independent variable (1) Crisis (1) No Crisis (2) Crisis (2) No Crisis (3) Crisis (3) No Crisis (4) Crisis (4) No Crisis Constant -.0003015* (.0000965) -.0002651* (.000075) -.0003057* (.0000967) -.0002624* (.000075) -.0003417* (.0000987) -.0002665* (.0000755) -.0003435* (.0000991) -.0002641* (.0000756) DayOfDowngr ade -.0026394* (.0012458) -.0041582** (.0021024) -.0027696** (.001246) -.004262** (.0021014) Min5days -.0009304 (.0005677) -.0025859* (.0009353) Plus5days -.0008185 (.000574) .0000559 (.000854) Days11 -.0010317* (.0004014) -.0016005* (.0006126) Min10days (.0004173) -.0002097 -.0012618*** (.0006553) Plus10days .0000646 (.0004244) .0001243 (.0006083) Days21 -.0002139 (.0003087) -.0007792*** (.0004458) #Observations 98410 42127 98410 42127 98410 42127 98410 42127 #Downgrades 363 66 363 66 363 66 363 66
Table 9. *=significant at 1%-level **=significant at 5%-level ***=significant at 10%-level Within parentheses are the standard deviations
Table 9: Effect of the Crisis (continued)
Table 9 continued. *=significant at 1%-level **=significant at 5%-level ***=significant at
10%-level
Within parentheses are the standard deviations
Dependent variable: AbnRet Independent variable (5) Crisis (5) No Crisis (6) Crisis (6) No Crisis (7) Crisis (7) No Crisis (8) Crisis (8) No Crisis Constant -.000321* (.0001027) -.0002525* (.0000764) -.0003133* (.0001038) -.0002581* (.0000765) -.000229** (.0001065) -.0002443* (.0000772) -.000261** (.0001081) -.000255* (.0000774) DayOfDowngr ade -.0027429** (.0012457) -.0042141** (.0021019) -.0026391** (.0012453) -.0041736** (.0021025) Min20days (.0003212) -.0002045 (.0004738) -.0006359 Plus20days -.000095 (.0003174) -.0004219 (.0004448) Days41 -.0002865 (.000243) (.0003315) -.0005323 Min30days -.0011869* (.0002721) -.0003724 (.000387) Plus30days (.0002735) .0003316 (.0003754) -.0005257 Days61 -.0004227*** (.0002175) -.0004214 (.0002833) #Observations 98410 42127 98410 42127 98410 42127 98410 42127 #Downgrades 363 66 363 66 363 66 363 66
5. Conclusion
This study examined the occurrence of the correction of a bank’s share price
associated with a downgrade of the credit rating of a bank in the period of 2005 until 2013. This time span resulted in 429 downgrades available for analysis. This is examined by multiple event studies and the parameters of these event studies were also tested by a Wald-test. A variety of time windows were used in the event studies. The abnormal returns were regressed on the variety of time windows and secondly the financial crisis was taken into consideration. From the variety of time windows that were studied it was deduced that the correction takes place in period of 5 days prior to the day of downgrade, but also on the day of downgrade itself. On the contrary, in the period after the day of downgrade no significant correction was determined. Also a smaller but significant effect was measured in the time window of at least 30 days prior to the downgrade. The Wald-test pointed out that the
variables indicating the days prior to the downgrade were significantly different for 5, 30, 40 and 50 days before the downgrade. These periods of time were tested against the period of 5 days after the downgrade.
These results are in line with the study of Boot et al. (2006). They argue that the credit rating agencies serve as ‘focal points’. When a bank is put ‘on watch’, the bank has an
opportunity to engage in a series of actions that may prevent the bank from being
downgraded. This could result in a ‘correction’ of the share price in the period prior to the downgrade. The findings of this study, that the correction in the share price predominantly occurs in the period before the downgrade, support the authors’ argument.
Bannier & Hirsch (2010) study the effect of the introduction of the watchlist on credit ratings. They find that a credit rating change is much more informative after the introduction of the watchlist. That is also what this study confirms, at least for downgrades. This study finds that the correction in the share occurs at least 30 days before the downgrade, which is at least partly caused by the bank being put on a watchlist.
Additionally, the sample was divided in two samples. One sample contained all the downgrades before September 15th, 2008 (fall of Lehman Brothers and begin of the financial crisis) and the other sample contained all downgrades after that date. In this way the influence of the crisis on the effect of a downgrade was measured. With these subsamples it was found that during the recent crisis the effect of the downgrade on share prices occurred earlier before the date of downgrade. This can be explained be the economical uncertainty in the recent crisis.
6. Reference list
Bannier, C. E., & Hirsch, C. W. (2010). The economic function of credit rating agencies – What does the watchlist tell us? Journal of Banking & Finance, 34(12), 3037–3049. Becker, B., & Milbourn, T. (2011). How did increased competition affect credit ratings?
Journal of Financial Economics, 101(3).
Boot, A. W. A., Milbourn, T. T., & Schmeits, A. (2006). Credit Ratings as Coordination Mechanisms. Review of Financial Studies, 19(1), 81–118.
Federal Reserve Statistical Release, Large Commercial Banks As of June 30, 2013, from: http://www.federalreserve.gov/Releases/Lbr/current/default.htm
Galil, K., & Soffer, G. (2011). Good news, bad news and rating announcements: An empirical investigation. Journal of Banking & Finance, 35(11), 3101–3119.
Goh, J. C., & Ederington, L. H. (1993). Is a Bond Rating Downgrade Bad News, Good News, or No News for Stockholders? The Journal of Finance, 48(5), 2001.
Güttler, A., & Wahrenburg, M. (2007). The adjustment of credit ratings in advance of defaults. Journal of Banking & Finance, 31(3), 751–767.
Hand, J. R. M., Holthausen, R. W., & Leftwich, R. W. (1992). The Effect of Bond Rating Agency Announcements on Bond and Stock Prices. The Journal of Finance, 47(2), Hull, J., Predescu, M., & White, A. (2004). The relationship between credit default swap
spreads, bond yields, and credit rating announcements. Journal of Banking & Finance,
28(11), 2789–2811.
Manso, G. (2013). Feedback effects of credit ratings. Journal of Financial Economics, 109(2), 535–548
Morgan, D. (2002). Rating Banks : Risk and Uncertainty in an Opaque Industry, American
Economic Review, 2002, Vol.92(4), pp.874-888.
Norden, L., & Weber, M. (2004). Informational efficiency of credit default swap and stock markets: The impact of credit rating announcements. Journal of Banking & Finance,
28(11), 2813–2843.
Rom, M. C. (2009). The Credit Rating Agencies and the Subprime Mess: Greedy, Ignorant, and Stressed? Public Administration Review, 69(August), 640–650.
SNL Financials, Top 20 largest banks in Europe, from:
http://ww1.prweb.com/prfiles/2013/06/07/10814184/euro%20bank%20top%2050%20ra nking.gif
Standard & Poor’s Research, Ratings for Europe’s Largest 100 Banks Show The Widest Range In Creditworthiness In 30 Years, April 18, 2011, from:
http://www.bng.nl/DocsComb/Publicaties/S&P%20ratings%20Europes%20largest%201 00%20banks.pdf
United States Bankruptcy Court, Southern District of New York, Lehman Brothers Filing for Bankruptcy September 14, 2008 from:
