The Wealth Effects of a Bankruptcy Filing Announcement on a Rivals Equity Claudia van Galen, 10210016
Economie en Bedrijfskunde Finance and Organization Philippe Versijp
17 February 2014
__________________________________________________________________________ Abstract
This paper examines the wealth effects on an industry rival portfolio in times of economic expansion and contraction. It investigates if there are effects present that correspond with the contagion or the competitive effect. The eleven-day cumulative abnormal return in times of economic expansion is significant at a five percent level. The eleven-day and the three-day cumulative abnormal returns in times of economic contraction are significant at a ten percent and a five percent level respectively. The signs of all cumulative abnormal returns are
conflicting with most results found by earlier research. Positive cumulative abnormal returns are found, while most papers show a negative cumulative abnormal return. The cumulative abnormal returns in times of economic expansion and contraction are compared to each other with a difference of means test. The hypothesis that follows from theory is that the cumulative abnormal returns should be stronger in times of economic contraction. There are probably larger consequences of bankruptcies in times of economic contraction but further research with a larger sample should first confirm this. The result of the difference of means test has a - 19.01% significance level. The cumulative abnormal returns in times of economic contraction are indeed stronger than in times of economic expansion but there can’t be said that there is a significant difference in the strength of the effects in both periods.
__________________________________________________________________________
1. Introduction
Many bankruptcy filings occur in the current unstable economic environment. The number of bankruptcy filings grew from 901.927 in the year preceding 31 march 2008 to 1.571.183 in the year preceding 31 march 2011 (Salzman and Hibschweiler, 2012). The effects of a bankruptcy filing reach beyond the bankrupt company to other stakeholders. A bankruptcy of a firm in a specific industry can have valuation implications for its competitors. Can they expect an increase or a decrease in their equity value and is this consistent with the
competitive or contagion effect? A change in the equity value can be caused by the fact that significant new information is released. This new information can be about the asset value of
the company that files for bankruptcy or it can be about the recognition of sizable unanticipated bankruptcy costs (Ferris et al., 1997). If the information that is released is about the asset value, other companies with the same industry assets can also experience a decrease in their equity value. This effect is called the contagion effect. The equity value of competitors can also be decreased because consumers, suppliers and regulators reassess the creditworthiness of the rivals as well because they have limited information available (Lang and Stulz, 1992). In the case of significant information about sizable unanticipated bankruptcy costs, a competitive effect can occur. For a Chapter 11 filing this means that the filing firm faces higher costs and will return to the market in a less competitive position, industry rivals have a cost advantage now (Ferris et al., 1997). When a firm files for Chapter 7 instead, it will totally disappear from the market and this leads to a competitive effect as well because the demand that the other firms face will rise as well. Expected is that which effect will outweigh the other depends on the characteristics of the filing firm and the industry. The current literature on this subject is not very extensive and just a couple of recent articles have been published.
It is important to recognize that a bankruptcy has far-reaching consequences, not only for the bankrupt firm but also for the industry competitors and even for customers and
suppliers. These consequences can be investigated by examining the abnormal returns of these subjects around the filing date. In this paper the wealth effects on the equity value of industry rivals in times of economic expansion and in times of economic contraction will be examined. It is not clear in advance if there will be an effect or not and if the observed effect will be in line with the competitive or contagion effect. Expected is that which wealth effect will occur is largely dependent on industry and firm-specific characteristics. For example, if all companies in an industry have similar assets or cash flows, than the information that is revealed by a bankruptcy filing of a specific company also reveals a lot of information about the other industry participants. In this situation the wealth effects that the competitors face will be stronger than when the bankrupt firm and its competitors have almost no similarities. Another example is the degree of concentration in an industry. In a highly concentrated industry there is much more to gain for the competitors when a rival faces problems than in an industry that exists of many small firms. These and other characteristics will be taken into account by carrying out cross-sectional regressions with independent variables that
represent the industry characteristics.
The expectation is that the wealth effects observed after a Chapter 11 filing will be predominantly consistent with the contagion effect and that the wealth effects after a Chapter 7 filing will be consistent with the competitive effect. This can be caused by the fact that when a Chapter 11 filing occurs, some bad information about the industry becomes public and/or that the bankrupt firm will come back as a stronger rival. A Chapter 7 filing on the
other hand will result in the disappearance of a company which is good news for the rivals. The first hypothesis that follows from theory is that after a bankruptcy filing, wealth effects occur that effect the equity value of the rivals in the industry. Next, the difference between the wealth effects in period of economic expansion and economic contraction will be examined. In times of economic contraction more bankruptcies occur which leads to more information transfers. Expected is that because of this the wealth effects that occur will be stronger. Hence, the second hypothesis is that the average wealth effects in times of economic contraction will be stronger than in times of economic expansion.
The wealth effects will be examined by a short-term based event study based on the market model. To measure the industry wealth effects an industry portfolio is constructed. This is a portfolio consistent of all firms that have the same SIC code as the bankrupt firm in Compustat and have enough equity returns around the event date. The parameters of the variables of the market model will be analysed using Ordinary Least Squares regression. To take into account the industry characteristics three cross-sectional regressions will be conducted. The dependent variable is the three-day abnormal return measured over a [-1,1] day event window. The independent variables that are used are variants of leverage, variants of the Herfindahl-index and the industry rating. In the different models leverage and the Herfindahl-index are used as continuous variables and dummies. The data can be found in Compustat and the data twelve months preceding the event date is used. Expected from literature is a significant and negative one percent cumulative abnormal return. The wealth effects in times of economic contraction are hypothesized to be stronger than in times of economic expansion. It was not possible to examine the effects of Chapter 7 and Chapter 11 separately because there were too little Chapter 7 filings in the subsample. The contribution of this paper to the current literature is the comparison of the wealth effects in times of economic expansion and economic contraction.
The structure of the remainder of the paper is as follows. Section II gives an overview of the existing literature in the field and states the hypotheses. Section III describes the data used and explains the method that is used. Section IV presents the empirical findings. The conclusion is drawn in section V and the results are discussed here.
2. Literature
One of the first important papers about the effects of bankruptcy announcements on the stock price of rivals is written by Lang and Stulz (1992). They investigate if there are valuation effects observed that are consistent with the competitive and/or the contagion effect. The papers that followed afterwards are almost always linked to this paper. The
contagion effect is the change in the value of competitors’ equity that cannot be attributed to wealth redistribution. A common view is that the contagion effect occurs because people become more cautious about investing in an industry when there is a bankruptcy filing in that industry. Another view is that a bankruptcy filing provides information about other firms that have similar cash flows as the bankrupt firm. A bankruptcy filing can also bring good news for the industry. Value can be redistributed from the bankrupt firm to the firms in the industry which can lead to a rise in equity value; this is called the competitive effect (Lang and Stulz, 1992).
In the current literature, researchers take into account the fact that abnormal returns should be dependent on industry characteristics. Variables that are commonly used to capture the effects of industry characteristics are leverage, the Herfindahl index, the
correlation in cash flows between the competitors and the bankrupt firm and the rating of the industry. Leverage is an important industry characteristic because there is a high correlation with leverage and the contagion effect because of elasticity (Tang, 2010). An announcement that has a negative effect on the value of a company will have a larger effect on a company with high leverage because the wealth effect is born by the equity holders and not by the debt holders. Higher leverage means more risk for the equity-holders which leads to a larger effect on the stock price. So firms who compete in a highly levered industry are more
vulnerable to changes in the industry environment that can be caused by bankruptcies and they benefit less from the failure of their competitors. This implies that higher leverage results in a stronger contagion effect, so the coefficient on leverage is hypothesized to be negative. More concentrated industries benefit more from the competition effect (Hertzel et al., 2007) The disappearance of a rival shifts a larger percentage of demand to the competitors than it would shift in a highly competitive industry. The degree of competition can be measured by the Herfindahl-index and the coefficient on this variable is expected to be positive because there is an inverse relation between the Herfindahl-index and the degree of competition. A high degree of similarity in cash flows leads to stronger contagion effects. The bankruptcy filing of an industry participant gives a lot of information about the cash flows of all the other industry participants. This similarity in cash flows can be measured by the correlation of equity returns of the bankrupt firm and the competitors which has a hypothesized negative coefficient. However, there is a high negative correlation between this correlation and the Herfindahl-ratio (Lang and Stulz, 1992). Greater vulnerability to the competition effect is expected from industries that have a lower credit rating (Zhang, 2010). The expected coefficient on the industry rating is negative.
Assets and size are commonly used as company-specific factors. Jorion and Zhang use the natural log of the total liabilities of the bankrupt firm for the variable size. The effect of size on the abnormal return is different for Chapter 7 and Chapter 11 filings. A Chapter 11
filing of a large firm will convey more information about industry specific cash flows which leads to a coefficient with an expected negative sign. The larger the market share of the bankrupt firm, the better for the industry rivals when a firm files for Chapter 7. So the
coefficient of size is positive in this case. The natural log of total assets can also be used as an explanatory variable of size (Tang, 2010).
Most papers find a negative abnormal return with a value around one percent, but the results are almost always barely significant. Jorion and Zhang (2007) try to solve this
problem by investigating contagion and competitive effects in the light of credit risk using credit default swaps. The standard is using stock markets but they use CDS data because the observed effects are much stronger with this data. A disadvantage of this method is that the data is harder to obtain, because CDS are less liquid due to less frequent trading. Another approach is looking at both the short-term and the long-term industry stock
responses. This research is performed by Zhang (2010). Tang (2010) looks at the existence of intra-industry information transfers by examining three different solution outcomes. By distinguishing between Chapter 7 and Chapter 11 filings, the contagion and competitive effects can be partly separated. In Chapter 7 filings the competitive effect probably
dominates because of the disappearance of a competitor. In Chapter 11 filings the contagion effect is expected to be dominating because the firm emerges mostly as a stronger
competitor and the filing reveals negative information about the industry. In the current literature a rival portfolio is used to examine the effects of a bankruptcy filing on the equity value of the industry. Such a rival portfolio consists out of firms with the same SIC code in Compustat as the bankrupt firm. Furthermore, the firms in the portfolio should have at least data in CRSP one year before and after the filing date. The variables which are used to indicate the industry characteristics and the company-specific factors are the factors already described in this section.
First an analysis on the abnormal returns for each period will be conducted
separately. Wealth effects are expected to occur because a bankruptcy filing reveals new information which results in revised expectations of the performance of competitors and this information will be incorporated in the stock price. The alternative hypothesis is expected to be true.
H0: AARit = 0 H1: AARit ≠ 0
The analysis of abnormal returns in periods of economic expansion and contraction that will be conducted in this paper will make a comparison between two different states of the
economy. It will provide insights into the extent to which a contraction increases the effects of the information transfers. This comparison is not yet made in the literature and this paper will
make a new contribution to the recent literature by comparing the wealth effects in a period of economic expansion and in a period of economic contraction. The difference in effects will probably become clearer if Chapter 7 and Chapter 11 are examined separately, so that the effects cannot cancel each other out. Lang and Stulz (1992) already mentioned that the wealth effects are stronger in industries which experienced more than one bankruptcy. In times of economic contraction more bankruptcies occur than in times of economic expansion. This will lead to more information transfers. The hypothesis that follows from the current literature is that because of the increased number of information transfers the wealth effects are stronger in times of economic contraction than in times of economic expansion. The alternative hypothesis is expected to be true.
H0: AARi economic expansion= AARi economic contraction H1: AARi economic expansion ≠ AARi economic contraction
If Chapter 7 and 11 filings are examined separately, the differences between the effects in times of economic expansion and contraction are expected to become even larger. This hypothesis is in accordance with the findings of Jorion and Zhang.
3. Data and method
A list of 272 bankruptcy filings in the U.S. from 2003 to 2006 and a list of 485 filings from 2008 to 2011 are collected from www.bankruptcydata.com. This is a website with information about bankruptcy data and news. The list contains the firms that had an explicit number for the value of assets and liabilities. Companies for which there was only an estimated range on assets and liabilities available are not part of the sample because the size of the liabilities has to be known for research. A selection of large firms is made by only retaining the filing firms that had liabilities worth more than 120 million dollar. This cutoff is also used by Lang and Stulz (1992) and Hertz et al. (2007). By using this subsample we distinguish large firms from small firms. Only large firms are relevant because the wealth effects of small firms who file for bankruptcy are expected to be negligible.
A rival portfolio is constructed for each firm that filed for bankruptcy by finding firms with the same SIC code in Compustat. A requirement is that the firms have enough equity return data available. Instead of equity return, operating income could also be used. In this paper equity returns are used because the data on equity returns is easily accessible and straightforward. A three year period for both the period of economic expansion and the period of economic contraction is used. The reason for this is that the effects of the crisis probably became clear at the second half year of 2008 and not yet at the beginning. The
sample that is used in this paper contains 71 bankruptcy filings in times of economic expansion and 107 bankruptcy filings in times of economic contraction. A firm that is in financial distress can continue existing after reorganization (Chapter 11) or can be liquidated (Chapter 7). What will happen depends largely on the strategic position of the firm. The option that results in the highest value for the creditors will be chosen. The legal aspect is also important. In the United States most companies file for a Chapter 11 bankruptcy while in some European countries Chapter 7 is more common (Kraft and Steffensen, 2009). The number of Chapter 7 filings in sample is excessively low; there were only four Chapter 7 filings. There were, however, some companies who first filed for Chapter 11 bankruptcy but who were later transformed to Chapter 7. Indro et al. (1999) state that companies use a Chapter 11 filing sometimes as a strategic choice. They choose to file for Chapter 11 bankruptcy even if they are not yet in a financial insolvent position. In this way they try to benefit from the protective environment of Chapter 11. After they are back on the market, they have a better financial position and the chance that they have to file for Chapter 7 diminishes. There is a possibility that this can partly explain the unequal proportion of Chapter 7 and Chapter 11 filings in the sample. Unfortunately because of the lack on Chapter 7 filings the difference in wealth effects between Chapter 7 and Chapter 11 filings cannot be investigated. Panel A of table I gives a description of the distribution of the final sample.
Table I. Distribution of firms that filed for bankruptcy
This table describes the distribution of the final sample of firms that filed for bankruptcy. Panel A shows the number of firms that filed for bankruptcy and the number of industries in terms of their 4-digit SIC code over which these firms are spread. Panel B shows the summary statistics of the rivals within an industry portfolio.
Panel A. Summary Statistics of Events
Year Number of firms
that filed for bankruptcy Number of industries Chapter 7 filings Chapter 11 filings 2003 33 30 0 33 2004 17 17 0 17 2005 21 18 0 21 2009 65 46 1 64 2010 21 17 1 20 2011 21 18 2 19 Total 178 119 4 174 7
Panel B. Number of Rivals within an Industry Portfolio
Year Number of
Events
Mean Std. Dev. Min. Median Max.
2003 33 19.06 20.08 1 11 85 2004 17 14.70 16.20 1 9 53 2005 21 43.37 32.03 1 40 111 2009 65 49.23 69.41 1 21 250 2010 21 74.24 104.85 1 10 250 2011 21 55.10 73.99 2 31 250 Total 178 43.30 64.69 1 17 250
To examine the wealth effects, measured as abnormal returns, on industry rivals we need to construct a rival portfolio. For each firm that filed for bankruptcy an equally-weighted portfolio is constructed. An equally-weighted portfolio has been chosen because this was more convenient to make the calculations. For the results it doesn’t matter if a
value-weighted or an equally-value-weighted portfolio is used (Lang and Stulz, 1992). The requirements for a firm to be part of the rival portfolio are the following: 1) the SIC code should be the same as the filling firm, and 2) stock returns in the CRSP daily database for [-12, 12] months should be available. Panel B of table I summarizes the statistics of the rivals within an industry portfolio.
To analyze the abnormal returns around the filing date, which we call the ‘event date’ from now on, a short-term event study based on the market model is used. This method is proposed by F. de Jong (2007). The event date is obtained from www.bankruptcydata.com. When the event does not takes place at a trading-day, the first trading-day after the event is taken as the event date. The abnormal return, AR, is the difference between the industry portfolio return and the expected return estimated by the market model. The market model uses the value-weighted equity return in CRSP as a proxy for the market. The parameters of the market model are estimated with Ordinary Least Squares over a 220 trading-day period, so the estimation window is [-250,-31] days. The abnormal returns are evaluated over an eleven day period around the event date, this results in an event window of [-5,5] days. Also cumulative abnormal returns, CAR, are estimated. This is done over a [-1,1] day and [-5,5] day event window. Significance is tested using a t-test which follows a student t-distribution.
Next, a cross-sectional Ordinary Least Squares regression is made with the [-1,1] day average cumulative abnormal return as the dependent variable and variants of the variables INDRTG, LEV and HERF as independent variables that indicate the industry characteristics. The regression is conducted separately for the period of economic expansion and the period 8
of economic contraction. Panel A of table II summarizes the descriptive statistics of the independent variables used. Panel B of table II shows the correlation between the variables. Three different models are estimated to see what the differences are if the variables are threated in a different way. The coefficients on the independent variables are evaluated using a t-test which follows a student t-distribution. For all calculations in this section STATA and/or excel are used.
Table II. Descriptive Statistics of Variables
Panel A shows the descriptive statistics of the independent variables. INDRTG is the average S&P industry rating. HH is a dummy which is 1 if the industry has an above average level of leverage and Herfindahl-index and 0 otherwise. LH is a dummy which is 1 if the industry has a below average level of leverage and an above average level of the Herfindahl-index and 0 otherwise. LL is a dummy which is 1 if the industry has a below average leverage level and Herfindahl-index. LEV is the average level of the industry leverage. HERFpos is the
Herfindahl-index multiplied by the dummy POS which is 1 if the Herfindahl-index is smaller than or equal to 1 and 0 if it is larger than 1. LEVdummy is 1 if the industry has an above average level of leverage and 0 if the level of leverage is below the average. HERFdummy is 1 if the industry has an above average Herfindahl-index and 0 if the Herfindahl-index is below the average. Panel B shows the correlations between the independent variables. INDRTG has a very weak correlation with all other variables. LEV and LEVdummy have a strong positive correlation which is logical because they should measure the same effect. This is also the case for HERFpos and HERFdummy. HH has a weak positive correlation with LEV, HERFpos, LEVdummy and HERFdummy. This is as expected because HH measures high leverage and a high Herfindahl-index . The other way around holds for LL, there is a weak negative correlation observed with all this variables. LH has a negative middling correlation with LEV and LEVdummy and a positive middling correlation with HERFpos and HERFdummy, all these signs are as expected. All other correlations not yet mentioned are weak.
Panel A. Summary Statistics of Main Variables
Economic expansion (N=71) Economic contraction (N=107)
Variable Mean Std.
Dev.
Min. Median Max. Mean Std.
Dev.
Min. Median Max.
INDRTG 5.382 0.785 2.5 5.433 7 5.372 0.799 2 5.338 9 HH 0.211 0.411 0 0 1 0.150 0.358 0 0 1 LH 0.155 0.364 0 0 1 0.243 0.431 0 0 1 LL 0.296 0.460 0 0 1 0.308 0.464 0 0 1 LEV 0.622 0.160 0.291 0.637 0.957 0.638 0.172 0.122 0.626 0.952 HERFpos 0.296 0.237 0 0.254 1 0.266 0.237 0 0.188 1 LEVdummy 0.549 0.501 0 1 1 0.449 0.500 0 0 1 HERFdummy 0.408 0.495 0 0 1 0.393 0.491 0 0 1 9
Panel B. Correlation between Main Variables
INDRTG HH LH LL LEV HERF
pos LEV dummy HERF dummy INDRTG 1 HH 0.0504 1 LH -0.0173 -0.2287 1 LL 0.0150 -0.2664 -0.3339 1 LEV -0.0410 0.3250 -0.4558 -0.4530 1 HERF pos 0.1249 0.3963 0.5500 -0.2776 -0.2296 1 LEV dummy 0.0010 0.4296 -0.5325 -0.6201 0.7862 -0.2085 1 HERF dummy 0.0616 0.5109 0.6405 -0.4697 -0.1556 0.6723 -0.1113 1 4. Empirical results
First the industry stock reaction to a bankruptcy filing announcement is examined. The analysis is conducted as an event study based on the market model. The results of this analysis are presented in table II. In the table the mean and the median of the AR and the CAR are reported in percentages. The numbers in parenthesis under the mean are the t-statistics.
Table III. Abnormal returns of the rival portfolio
This table shows the valuation effects on an equally-weighted rival portfolio after a bankruptcy filing
announcement. To calculate the abnormal returns, the market model is used. The table presents the mean and median of the AR and CAR and the percentage of negative returns. The market model is estimated over a [-250,200] days estimation window. The CAR is estimated over a [-1,1] day and a [-5,5] day event window. The value-weighted market return of CRSP is used as a proxy for the market.
Economic Expansion (N=71) Economic Contraction (N=107)
Day Mean
(%)
Median (%) Day Mean
(%) Median (%) - 5 0.291 (1.88)** 0.139 - 5 (0.306) (-1.93)** (0.141) - 4 0.134 (0.96) 0.031 - 4 0.203 (1.32)* 0.125 - 3 (0.189) (-1.18) (0.099) - 3 (0.130) (-0.66) (0.167) 10
- 2 0.120 (0.92) 0.134 - 2 (0.002) (-0.01) (0.017) -1 0.230 (1.45) 0.201 -1 0.242 (0.98) 0.037 0 (0.110) (-0.85) (0.121) 0 0.265 (1.17) (0.037) 1 0.072 (0.45) 0.048 1 0.223 (1.55)* 0.072 2 0.308 (2.07)** 0.138 2 0.156 (0.83) 0.040 3 0.206 (0.94) 0.09525 3 (0.009) (-0.06) (0.219) 4 (0.079) (-0.53) (0.000) 4 (0.060) (-0.34) (0.251) 5 (0.058) (-0.40) 0.055 5 0.450 (2.11) ** 0.153 [-1,1] 0.194 (0.70) 0.220 [-1,1] 0.730 (1.84)** 0.069 [-5,5] 0.928 (1.75)** 0.384 [-5,5] 1.030 (1.58) * (0.185)
*** Significance level of one percent ** Significance level of five percent * Significance level of ten percent
The abnormal returns of an industry portfolio should depend on the characteristics of the industry because there is probably a large difference in the wealth effects across different industries. This will be examined by cross-sectional regressions in which we take the three-day CAR as the dependent variable. I will estimate the following models:
(1) CARj = α0 + β1*INDRTGj * β2*HHj + β3*LHj + β4*LL j + ɛj (2) CARj = α0 + β1*INDRTGj * β2*LEV j + β3*HERF j *POS j + ɛ j
(3) CARj = α0 + β1* INDRTG j + β2*LEVdummy j + β3*LEVdummy j + ɛ j
The independent variables used are a selection of the variables mentioned in section 2. INDRTG is the S&P rating of the industry. A+ is indicated by 1, A by 2, A- by 3, B+ by 4, B by 5, B- by 6, C by 7 and D by 8. Because the average industry rating is needed, INDRTG is used as a continuous variable. If only rounded numbers could be used that indicate each rating the results will be less accurate. If a variable is 4,45 rounding would lead to a rating of B+ while it is almost in between B+ and B. So by using INDRTG as a continuous variable this
problem is solved. But using INDRTG as a continuous variable gives a problem as well. The increase in chance of bankruptcy is not linear while in this paper the variable is used as if the chance of bankruptcy is linear. For this reason it is important to look at value of the
coefficient of INDRTG with caution but for the sign it doesn’t matter. HH, LH and LL are dummies which indicate different combinations of leverage and the Herfindahl-index. The first character shows if leverage is higher (H) or lower (L) than the average leverage in the period of interest and the second character shows if the Herfindahl-index is higher (H) or lower (L) than the average Herfindahl-index in the period of interest . For example, if an industry portfolio has a leverage ratio that is below the mean and a Herfindahl-index that is above the mean, the coefficient on the variable LH is 1 and the coefficients on the variables HH and LL are 0.
LEV indicates leverage and is measured by total liabilities divided by total assets. This ratio is calculated using the average of assets and liabilities in the industry portfolio in the previous year. HERF is used as a proxy for the degree of imperfect competition and is measured by the Herfindahl-index. The Herfindahl-index is calculated using data on
operating income before depreciation over the previous year because data on sales were not available in Compustat. This method caused a problem for some industry portfolios. The largest competitor had a considerable share of total operating income before depreciation of the industry and other industry rivals had negative operating income. This caused the Herfindahl-index to be greater than 1. This problem is solved by introducing the dummy variable pos. The coefficient on pos is 1 if there is no negative operating income that causes problems for the Herfindahl-index. If due to negative operating income the Herfindahl-index becomes larger than 1, the coefficient on the variable pos becomes 0 and the variable
HERF*pos becomes 0 as well. In this way the problem is solved by isolating the observations that caused a problem. Now the cases that caused problems don’t have to be excluded from the research. In the second model a continuous variable for leverage and the Herfindahl-index is used which is equal to the ratio or Herfindahl-index itself. The variable HERFpos that is used in model 2 is an interaction variable between HERF and pos as mentioned above. In model 3 the Herfindahl-index and leverage are measured as dummy variables LEVdummy and HERFdummy on which the coefficient is 1 if the variable is larger than the average and 0 otherwise. For more information about the variables see appendix A.
Table IV. Signs on the coefficients of the independent variables
This table provides the signs that on the coefficients of the independent variables that are expected from theory. The expected signs on INDRTG, LEV and LEVdummy are negative. The sign on HERF and HERFdummy are positive. The sign on LL is positive, for LH it could be either positive or negative and for HH it is expected to be negative.
Panel A: Expected signs by theory
Variable INDRTG LEV LEVdummy HERFpos HERFdummy LL LH HH Expected
Sign
- - - + + + + or - -
Panel B: Signs found by Research in this Paper in Times of Economic Expansion
Variable INDRTG LEV LEVdummy HERFpos HERFdummy LL LH HH Signs model 1 - + + + Signs Model 2 + - + Signs Model 3 + - +
Panel C: Signs found by Research in this Paper in Times of Economic Contraction
Variable INDRTG LEV LEVdummy HERFpos HERFdummy LL LH HH Signs model 1 - + + + Signs Model 2 - - + Signs Model 3 - - +
Industries with lower credit ratings are more fragile and consequently face stronger competitive effects, which will lead to an expected negative coefficient on the variable INDRTG. The coefficients on LEV and LEVdummy are expected to be negative because the percentage fall in equity of the rivals increases with leverage. This is caused by the fact that greater leverage leads to a greater elasticity of equity with respect to total firm value and to an increase bankruptcy costs. Leverage magnifies the contagion effect but it doesn’t affect the competitive effect. The competitive effect increases when the degree of competition, which is measured by HERF and HERFdummy, is decreasing. This is because companies in more concentrated industries have a larger gain from the exit of a large competitor. The degree of competition and the contagion effect are uncorrelated. So the average abnormal return is expected to be higher for industries with a high Herfindahl index and the coefficient
on HERF and HERFdummy are expected to be positive. The competitive effect is expected to be strongest in the subsample of industries with both low leverage and a low Herfindahl-index, which results in a positive coefficient on LL. The contagion effect should be strongest in the subsample of industries with both high leverage and a high Herfindahl-index, which results in an expected negative sign on HH. The sign on LH is depend on the fact if leverage or the Herfindahl-index will dominate. CORR is the correlation between returns of the event firm and the industry portfolio. A high correlation means similar cash flows which means that the coefficient on CORR is hypothesized to be positive. Unfortunately, there were not enough equity returns available for the firms that filed for bankruptcy so this coefficient had to be left out of the models. Table III presents the descriptive statistics of the variables used in the three models that are estimated. Table IV panel A shows the expected signs for the coefficients on the independent variables.
Table V. Ordinary Least Squares Regressions on Industry Characteristics
This table shows the coefficients on the independent variables in the cross-sectional Ordinary Least Squares regressions of three different models.
(1) CARj = α0 + β1*INDRTGj * β2*HHj + β3*LHj + β4*L j + ɛj (2) CARj = α0 + β1*INDRTGj * β2*LEV j + β3*HERF j *POS j + ɛ j (3) CARj = α0 + β1* INDRTG j + β2*LEVdummy j + β3*LEVdummy j + ɛ j
CAR is the dependent variable. This is the cumulative abnormal return of the rival portfolio over a [-1,1] days event window. INDRTG is the average S&P rating of the industry. HH is the dummy variable which is 1 if both leverage and the Herfindahl-index are high. LH is the dummy variable that is 1 if leverage is lower than the means and the Herfindahl-index is higher than the mean value. LL is the dummy variable that is 1 if both leverage and the Herfindahl-index are low. LEV is the average industry leverage. HERFpos is the Herfindahl-index multiplied by the dummy variable pos which is 1 if the Herfindahl-index is smaller or equal to 1. LEVdummy is a dummy variable that is 1 if the average industry leverage is higher than the average leverage in all industries.
HERFdummy is a dummy variable that is 1 if the average index is higher than the average Herfindahl-index in all industries. The numbers in parenthesis are the heteroskedastic-robust standard t-statistics. The significance levels are indicated by the superscripts ***, ** and * which means 1%,5% and 10% significance level respectively.
Economic Expansion (N=71) Economic Contraction (N=107)
(1) (2) (3) (1) (2) (3) Constant (0.003) (-0.12) 0.004 (0.20) 0.020 (0.10) 0.055 (2.19)** 0.089 (2.87)*** 0.062 (2.42)** INDRTG (0.000) (-0.02) 0.000 (0.10) 0.000 (0.05) (0.011) (-2.31)** (0.012) (-2.57)** (-0.011) (-2.29)** HH
(1 if high leverage and high Herfindahl index; 0 otherwise) 0.004 (0.56) 0.012 (1.03) 14
LH
(1 if low leverage and high Herfindahl index; 0 otherwise) 0.012 (1.44) 0.014 (1.40) LL
(1 if low leverage and low Herfindahl index; 0 otherwise) 0.005 (0.64) 0.009 (0.96) LEV (0.011) (-0.59) (-0.036) (-1.56) HERFpos 0.006 (0.48) 0.011 (0.67) LEVdummy (0.006) (-1.06) (0.006) (-0.86) HERFdummy 0.005 (0.80) 0.008 (1.06) *** Significance level of one percent
** Significance level of five percent * Significance level of ten percent
5. Discussion and conclusion
In the period of economic expansion there are three significant returns. The average return at day – 5 and day 2 are positive and significant at a five percent level and the cumulative abnormal return over an eleven-day interval is also significant at a five percent level. In the period of economic contraction there are more average abnormal returns that are significant. Also both cumulative abnormal returns are significant. The results are not corresponding with the main findings in the current literature. Most researches find significant negative returns around one percent. In this paper in both periods the abnormal returns are predominantly positive which resulted in positive cumulative abnormal returns as well. There is a big chance that this is caused by the fact that there are not much observations. The problem can maybe be solved by examining a larger sample that contains more observations for all years. By looking at more observations, the standard errors will decrease which will lead to an increase in the significance level. Another problem can be that the bankruptcy is already anticipated by the market so that the stock price reactions are not so large at the actual moment of the bankruptcy filing. This possible cause can be investigated by looking at stock price returns in a longer period before the filing actually happened.
The period of economic expansion and the period of economic contraction are
compared by a one-sided paired difference of means t-test carried out by Stata. The means of the cumulative abnormal returns in period of economic expansion are compared with the means of the cumulative abnormal returns in the period of economic contraction. Expected is that the wealth effects in times of economic contraction will be stronger than in times of economic expansion. The t-score that follows from the test is -1.47 with a 19.01%
significance level. Based on this information we can’t conclude that there is a difference in the strength of the wealth effects in period of economic expansion and economic contraction. The effects in times of economic contraction are indeed stronger but the results are not significant at a five percent level.
Next we will look on the industry characteristics. The coefficient on industry rating has a negative sign in four out of six estimations. In the other two estimations the coefficient of the variable is positive but very insignificant. Generally the coefficient of INDRTG is in accordance with theory, in the period of economic contraction the coefficient is also significant at a five percent level in all models. The coefficient on HH in both periods is positive which is not in accordance with the expectations. A high level of leverage and a high Herfindahl-index should lead to a negative wealth effect. However, both coefficients are insignificant. The coefficients on LL are also positive and not significant. The coefficients on LEV and LEVdummy were expected to be negative. All of them are indeed negative in both models but again the coefficients on the variables are insignificant. The last variables are HERFpos and HERFdummy. Both are expected to be positive and this research shows indeed positive coefficients.
There are contradictions with the predictions made by theory. Some shortcomings can probably be resolved by taking a larger sample. To get more significant results it will probably help to distinguish between Chapter 7 and Chapter 11 filings. To do this analysis, more Chapter 7 filings in the sample are needed. One way of getting more Chapter 7 filings is by relaxing the assumption that the liabilities have to be higher than 120 million. The biggest part of Chapter 7 filings in the sample did not have high liabilities. When this is done there is a chance that the effects will not be very strong because the bankrupt company was not big and important enough. Another suggestion is looking at the date of conversion from Chapter 11 to Chapter 7. There were quite some firms that filed for Chapter 11 but who were conversed to Chapter 7 later. At the date of conversion there will probably be a wealth effect too. Finally, a subsample of European countries instead of companies in the U.S. can be investigated. Kraft and Steffensen (2009) state that which bankruptcy filing is used depends partly on the legal setting of a country and that in the U.S. most companies file for Chapter 11 while in Europe some companies file for Chapter 7. The sample excludes the bankrupt firms that had only an estimated range of liabilities and assets. There could probably be some bias because of this selection. A suggestion for more reliable results is can also be
including these companies by finding the liabilities and assets in year rapports for example. However, in this research it was not possible to do this because it will be really time
consuming and it will probably be a problem too to find all old year rapports. The last
suggestion for further research is not looking at all industries at the same time but specify the analysis for homogeneous industry subgroups. This can also provide valuable insights.
References
Ferris, S. P., Jayaraman, N., Makhija, A. K., 1997. The response of competitors to announcements of bankruptcy: An empirical examination of contagion and competitive effects. Journal of Corporate Finance. Vol 3: 747-767.
Indro, D. C., Leach, R. T., Wayne, Y., 1999. Sources of Gains to Shareholders from Bankruptcy Resolution. Journal of Banking and Finance. Vol 23: 21-47. Jong, de, F., 2007. Event Studies methodology. University of Tilburg.
Hertzel, M. G., Li, Z., Officer, M. S., Rodgers, K. J., 2007. Inter-firm Linkages and the Wealth Effects of Financial Distress along the Supply Chain. Journal of Financial Economic. Vol. 87: 374-387.
Jorion, P., Zhang, G., 2007. Good and Bad Credit Contagion: Evidence from Credit Default Swaps. Journal of Financial Economics. Vol. 84: 860-883.
Kraft, H., Steffensen, M., 2009. Asset Allocation with Contagion and Explicit Bankruptcy Procedures. Journal of Mathematical Economics. Vol. 45: 147-167.
Lang, L. H. P., Stulz, R. M., 1992. Contagion and Competitive Intra-industry Effects of Bankruptcy Announcements. Journal of Financial Economics. Vol. 32: 45-60.
Salzman, M. L., Hibschweiler, A. M., 2012. Timing Considerations of Discharging Taxes in a Chapter 7 Bankruptcy. The Tax Adviser. Vol 43: 104-113.
Tang, T., 2010. Effects of Announcements of Reorganization Outcome. Applied Economics. Vol. 42: 1113-1124.
Zhang, G. Emerging from Chapter 11 Bankruptcy: Is It Good News or Bad News for Industry Competitors? Financial Management. Vol 4: 1719-1742.
Appendix
Appendix A: Variable list Variable
name
Variable description Sort variable Way of measurement
Scale
INDRTG The average S&P rating of the industry portfolio.
Categorical and discrete
1=A+, 2=A, 3=A-, 4=B+, 5=B, 6=B-, 7=C, 8=D
Between 1 and 8
LEV The average industry leverage measured by the average industry liabilities divided by the average industry assets.
Continuous Both assets and liabilities are
measured in dollars
Between 0 and 1
HERF The market
concentration measured by the average industry Herfindahl-index. Operating income before depreciation is used for the
Herfindahl-index
Continuous Operating income before depreciation is measured in dollars.
Between 0 and 1
Pos Variable that indicates if the Herfindahl-index is smaller or equal to 1 for which the
coefficient is 1 or larger than 1 for which the coefficient is 0.
Dummy 0 or 1
LEVdummy Variable that has a coefficient of 1 if the average portfolio leverage is larger than the total average ratio and 0 otherwise.
Dummy 0 or 1
HERFdummy Variable that has a Dummy 0 or 1
coefficient of 1 if the average portfolio Herfindahl-index is larger than the total average Herfindahl-index and 0 otherwise. HH Dummy variable that is
1 if both leverage and the Herfindahl-index are above average and 0 otherwise.
Dummy 0 or 1
LH Dummy variable which
is 1 if leverage is below average and the Herfindahl-index is above average and 0 otherwise.
Dummy 0 or 1
LL Dummy variable that is 1 if both leverage and the Herfindahl-index are below average and 0 otherwise.
Dummy 0 or 1
