the effects of bankruptcy filings on the competitors' earnings

The effects of bankruptcy filings on the

competitors’ earnings

Zahid Iqbal*

School of Business, Texas Southern University, Houston, TX 77004, USA

Received 25 July 2000; received in revised form 19 December 2000; accepted 19 March 2001

Abstract

In this paper, we investigate whether a bankruptcy filing explains the earnings of the bankrupt

firm’s competitors. We examine contagion and competitive effects on competitors’ earnings for a

sample of 165 bankruptcies filed between 1991 and 1996. Our findings show that the competitors

experience an increase in return on equity (ROE) in concentrated industries than in less concentrated

industries, suggesting a competitive effect. In addition, the competitive effect is observed in

industries with high leverage and multiple bankruptcy filings. Our data fail to provide support for

the contagion effect of bankruptcy on competitors’ earnings. D 2002 Elsevier Science Inc. All

rights reserved.

JEL classification: G33

Keywords: Bankruptcy; Earnings; Contagion effect; Competitive effect

1. Introduction

Prior empirical studies observe that a bankruptcy filing has a significant effect on the rival

firm’s stock prices. Lamy and Thompson (1986) conclude that Penn Square failure resulted in

negative stock price reactions among other banks. Gay, Timme, and Yung (1991) find

1059-0560/02/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved.

PII: S1059 -0560 (01 )00089 -2

* Tel.: +1-713-313-7737; fax: +1-713-313-7705.

E-mail address: [email protected] (Z. Iqbal).

International Review of Economics and Finance

11 (2002) 85–99

significant negative stock price reactions around bank failures in the banking industry in

Hong Kong. Later studies by Cheng and McDonald (1996), Ferris, Jayaraman, and Makhija

(1997), and Lang and Stulz (1992) show that the industry effects of bankruptcy depend on

certain industry characteristics. Lang and Stulz conclude that a negative stock price reaction

or contagion effect indicates that other firms in the industry experience financial difficulties as

well and that this effect is larger in industries with similar cash flow characteristics. There can

also be a positive reaction or competitive effect indicating that the competitors benefit at the

expense of the financial inefficiencies of the bankrupt firms. The competitive effect is larger

in highly concentrated industries.

Ferris et al. (1997) find evidence of a contagion effect when the competitors also file for

bankruptcy within 3 years of a bankruptcy filing in the industry. Their evidence on the

competitive is weak, however. Cheng and McDonald (1996) focus on the market structure

and contagion effect hypotheses to interpret stock price reactions to bankruptcy by the

surviving firms in the airline and railroad industries. They posit that the airline companies

possess market power with their ability to control price, quantity, and the nature of the

products sold. Thus, bankruptcy in the airline industry enhances the market power of the

surviving firms which results in higher market value. The railroads, on the other hand, are

interdependent in providing services to the customers. This interdependency weakens

operating performance of the surviving firms when a bankruptcy is filed in the railroad

industry, and, therefore, results in negative stock price reactions.

In this paper, we examine whether the earnings of the rival firms are related to

bankruptcy filings. We provide evidence on whether changes in earnings within the

industry are consistent with the intraindustry stock price effects of bankruptcy detected in

prior studies. Specifically, we report findings on changes in return on equity (DROE) for

a sample of 165 industries that experienced bankruptcy filings between 1991 and 1996

and for subsamples based on industry and bankruptcy-related characteristics. Although a

similar study by Kennedy (2000) examines the operating performance of rivals firms

following a bankruptcy filing, our study is different in the following ways: (i) We

incorporate the specifications used by Lang and Stulz (1992) for contagion and

competitive effects to examine the earnings performance of the competitors. (ii) Our

sample includes all the surviving firms in the industry, not just the five largest

competitors used in Kennedy’s study. (iii) We consider industry and bankruptcy-related

factors (e.g., industry leverage, number of bankruptcy filings in the industry, and size and

earnings of the bankrupt firm) that might affect competitors’ earnings as a result of a

bankruptcy filing.

Our findings indicate that the ROE increases in industries with high concentration of firms

than in industries with less concentration. This relationship is evident in highly leveraged

industries and in industries where more than one bankruptcy petition is filed. Our findings do

not provide support for the contagion effect of bankruptcy as we observe an increase in ROE

in industries with high stock return correlations.

The rest of the paper is organized as follows. Section 2 provides background and the main

hypotheses for this paper. Section 3 describes sample and data collection. The empirical

findings of this study are given in Section 4. Section 5 concludes the paper.

Z. Iqbal / International Review of Economics and Finance 11 (2002) 85–9986

2. Background and expected intraindustry earnings effect

The hypotheses for our study are based on the arguments and empirical evidence presented

by Lang and Stulz (1992). They propose a contagion effect of bankruptcy on the premise that

financial difficulties of the bankrupt firm indicate financial problems in the industry as well.

In addition, they point out that other firms in the industry can be adversely affected in their

dealings with customers, regulators, and suppliers due to a bankruptcy filing in the industry.

This negative contagion effect is stronger in industries with a high degree of cash flow

correlations between the bankrupt firm and nonbankrupt firms.

The competitive effect is based on the argument that the rival firms derive benefits from

the financial difficulties of the bankrupt firm. For example, a decreased demand for the

bankrupt firm’s products may translate into higher demand for its competitors. Lang and Stulz

(1992) argue that this competitive effect is positively related to the degree of concentration

(i.e., inversely related to the degree of competition). Firms in highly concentrated (i.e., less

competitive) industries can increase the prices of their outputs even when they experience an

increase in demand for their products.

Lang and Stulz (1992) provide empirical evidence consistent with their assertions. They

observe that, in general, the competitors’ stockholders react negatively to bankruptcy

announcements. The negative reactions are stronger in industries with similar cash flow

characteristics and in industries with less concentration (measured by Herfindahl index).

Their regression results indicate a positive relationship between abnormal stock price

reactions and degree of concentration and a negative relationship between abnormal stock

price reactions and stock return correlation.

Given the above findings on the valuation effects within the industry, we expect similar

effects of bankruptcy on the earnings of the rival firms. Specifically, we propose the

following hypotheses:

Hypothesis 1: For bankruptcies in general, we can expect either a contagion (negative)

or a competitive (positive) effect on rival firms’ earnings measured by ROE.

Hypothesis 1a: Contagion effect is stronger in industries with highly correlated stock

returns. There is a negative relationship between DROE of the rival firms and their stock

return correlations with the bankrupt firms.

Hypothesis 1b: Competitive effect is stronger in highly concentrated industries. There is

a positive relationship between DROE of the rival firms and Herfindahl index.

3. Sample and data

We identify 170 Chapter 11 bankruptcy filing announcements in the Wall Street Journal

Index (WSJI) from 1991 to 1996. Of these 170 bankrupt filings, we deleted five cases because

of missing or inconsistent industry median values of ROE in either the filing year or any of

the two subsequent years. Table 1 provides some descriptive information on the final sample.

Z. Iqbal / International Review of Economics and Finance 11 (2002) 85–99 87

Table 1

Descriptive statistics of the 165 firms filing bankruptcies and their industries

Bankruptcy filing firms listed on Research Insight database as

Active companies as of 12/31/98 38

Research companies (deleted from the active file) 127

Deleted due to:

Bankruptcy 81

Liquidation 10

Acquisition 21

Leveraged buyout 1

Other reasons 14

Yearly breakdown of the sample

Year of bankruptcy filing 1991 1992 1993 1994 1995 1996 Total

Number of firms filing bankruptcy 54 39 28 17 22 5 165

Industry statistics

Number of industries in the sample 165

Number of nonbankrupt firms in the

sample industries

5688

Mean of nonbankrupt firms

per industry

34

Median of nonbankrupt firms

per industry

22

Frequencies of bankruptcy filings per industry

Frequency Single

filing

Multiple

filings

Number of industries 72 93

Relative asset size of the filing firms

Number of firms with total assets �respective industry median

67

Number of firms with total assets <

respective industry median

90

Missing data 8

Total liabilities of the filing firms

Number of firms with total liabilities �US$120 million

78

Number of firms with total liabilities <

US$120 million

78

Missing data 9

Earnings of the filing firms

Number of firms with negative earnings 1 year

prior or in the year of filing

150

Number of firms with positive earnings 1 year

prior and in the year of filing

15

(continued on next page)


Of the 165 firms that filed bankruptcy petitions, 38 are listed as active companies at the

end of 1998 in the Standard & Poor’s Research Insight database. The remaining 127 research

firms are no longer in business, with 81 firms deleted due to bankruptcy, 10 due to

liquidation, 21 due to acquisition, 1 due to leveraged buyout, and 14 for other reasons.

About one-third or 54 of the sample filings are in 1991 with the number decreasing over time.

Since our empirical analysis of earnings is based on industry medians, we identify all

nonbankrupt firms in the bankrupt firm’s industry to compute the industry median. We ensure

that the bankrupt firms are not included in the sample of nonbankrupt firms during the sample

period. This essentially excludes a firm that filed for bankruptcy and subsequently emerged

from it, in measuring industry median in a later year. The total number of nonbankrupt firms

in all 165 industries are 5688 with the mean number of firms per industry being 34 and

median number being 22.

There are 93 multiple-filing industries, i.e., industries with two or more filings and 72

industries with single filings. One hundred fifty bankrupt firms have negative earnings

(before extraordinary items) in the year of filing or 1 year prior to the filing and 15 bankrupt

firms have positive earnings in both the years. Most bankruptcies in our sample are, therefore,

associated with financial difficulties. Sixty-seven bankrupt firms have total assets greater than

or equal to the respective industry medians 1 year prior to the filing, while 90 bankrupt firms

have total assets less than the respective industry medians. Seventy-eight bankrupt firms meet

Lang and Stulz’s (1992) sample selection criterion of total liabilities in excess of US$120

million. Almost half of our sample bankrupt firms are, therefore, comparable in size to Lang

and Stulz’s bankrupt firms.

Table 1 also presents the industry means and medians of Herfindahl index, stock return

correlations, and debt-to-assets ratio, which we use in subsequent empirical analysis.1

Herfindahl index for each industry is computed as the sum of squared proportions of total

sales 1 year prior to the filing year for all nonbankrupt firms in that industry. The mean

Herfindahl index for the industries is 0.45 and the median is 0.27. Stock return correlation in

an industry is defined as the Pearson’s correlation coefficient of monthly percentage change

in stock prices between the bankrupt firm and median value of the bankrupt firm’s industry.

1 We include debt-to-assets ratio in our analysis because Ferris et al. (1997) and Lang and Stulz (1992) argue

that industry leverage may impact competitors’ stock returns around bankruptcy filings. Higher leverage is

expected to strengthen contagion effect, while higher leverage may either increase or decrease competitive effect.

Table 1 (continued )

Number of

industries

Missing

data

Mean Median

Stock return correlations and Herfindahl index

Correlations 153 12 0.24 0.22

Herfindahl index 165 0 0.45 0.27

Debt-to-assets ratio 165 0 0.57 0.56


The mean value of the Pearson correlation coefficients for all industries is .24 and the median

value is .22. The mean debt-to-assets ratio for the industries is 0.57 while the median is 0.56.

4. Empirical findings

Table 2 presents drift-adjusted DROE in the year of the bankruptcy filing (Year 0) and two

subsequent years (Years 1 and 2). Following Benartzi, Michaely, and Thaler (1997) who used

Table 2

Five-year drift-adjusted changes in DROE for nonbankrupt firms in industries where bankruptcy petitions have

been filed

Year 0 Year 1 Year 2

Sample N Mean %Pos Mean %Pos Mean %Pos

Total sample 165 � 0.456 41.2## � 0.349 54.6 � 1.047 55.8

Correlation

Above median 76 � 0.604 34.2### 0.948 56.6 1.346* 64.5###

Below median 77 � 0.257 48.1 � 1.613** 51.9 � 3.622 46.8

Difference � 0.347 2.561b 4.968

Herfindahl index

Above median 82 0.409 46.3 0.518 63.4# � 2.520 54.9

Below median 83 � 1.302** 36.1## � 1.207* 45.8 0.408 56.6

Difference 1.711a 1.725a � 2.928

Leverage

Above median 82 � 0.141 41.4 � 0.142 59.8# � 1.806 56.1

Below median 83 � 0.766 40.9# � 0.554 49.4 � 0.297 55.4

Difference 0.625 0.412 � 0.509

The results are based on the median DROE in the bankrupt firm’s industry. We report the mean values of the

industry medians and the percentage of the positive industry medians in the year of the bankruptcy filing (Year 0)

and in two subsequent years (Years 1 and 2). The results are reported for (a) the total sample of 165 industries, (b)

subsamples of industries partitioned by the median value of stock return correlations between the bankrupt and

nonbankrupt firms, (c) subsamples of industries partitioned by the median value of Herfindahl index, and (d)

subsamples of industries partitioned by the median value of debt-to-assets ratio. N is the number of industries.a Mean values of two subsamples significantly different at the 10% level using a two-tailed t test.b Mean values of two subsamples significantly different at the 5% level using a two-tailed t test.c Mean values of two subsamples significantly different at the 1% level using a two-tailed t test.

* Significantly different from 0 at the 10% level using a two-tailed t test.

** Significantly different from 0 at the 5% level using a two-tailed t test.

*** Significantly different from 0 at the 1% level using a two-tailed t test.# Proportion of positive value significantly different from equal proportions at the 10% level using a chi-

square test.## Proportion of positive value significantly different from equal proportions at the 5% level using a chi-

square test.### Proportion of positive value significantly different from equal proportions at the 1% level using a chi-

square test.


drift-adjusted changes in earnings in their study, we compute 5-year drift-adjusted change in

ROE as:

DROEy ¼ ðROEy � ROEy�1Þ �ROE�1 � ROE�5

4ð1Þ

For each nonbankrupt firm, we first compute DROE for Years 0, 1, and 2. We then

determine the median DROE in each of the 165 industries. The mean value of the industry

medians and the percentage of positive industry medians are reported in Table 2. The

findings for the total sample indicate that the mean DROEs are not significantly different

from 0 in any year. Only the goodness-of-fit test shows that the percentage of positive

DROE in Year 0 (41.2%) is significantly less than the expected value of equal proportions.

The evidence of any bankruptcy effect on the competitors’ earnings for the whole sample is,

therefore, weak.

The findings for subsamples partitioned by the median values of stock return correlation,

Herfindahl index, and debt-to-assets ratio are also presented in Table 2. For industries with

correlation values higher than the median, the percentage of positive DROE in Year 0 (34.2%)

is significantly less than the expected value, which provides some support for a contagion

effect in the year of bankruptcy. However, the positive DROE of 1.346% in Year 2 for highly

correlated industries and the negative DROE of � 1.613% in Year 1 for less correlated

industries do not support the contagion effect. Overall, we fail to observe a strong evidence of

negative earnings (i.e., contagion effect) in industries that have high stock return correlations

between the bankrupt firm and its competitors.

The findings partitioned by the median Herfindahl index show that the below median

industries have negative and statistically significant DROEs of � 1.302% and � 1.207% in

Years 0 and 1, respectively. In these 2 years, the DROEs for the below median industries are

also significantly less than those of the above median industries. The proportion of positive

DROE (36.1%) is significantly less than expected in Year 0 for industries with low Herfindahl

index, while the proportion of positive DROE (63.4%) is significantly greater than expected

in Year 1 for industries with above median Herfindahl index. These findings based on

Herfindahl index support our hypothesis that a bankruptcy in a highly concentrated industry

has a greater competitive or positive effect on other firms’ earnings than a bankruptcy in a

less concentrated industry.

The findings partitioned by the median debt-to-assets ratio show that the mean DROE of

neither subsample is significant in any year. In Year 0, however, the proportion of positive

DROE (40.9) for the low-leverage industries is significantly less than the expected value. In

addition, in Year 1, the proportion of positive DROE (59.8) for the high-leverage industries is

significantly higher than the expected value. Thus, there is a weak evidence that leverage has

a positive influence on the DROE of bankrupt firms’ competitors.

To examine whether certain bankruptcy-related variables, e.g., number of bankruptcies in

the industry, relative size of the bankrupt firm, liabilities of bankrupt firm in excess of

US$120 million, and earnings of the bankrupt firm explain competitors’ earnings, we present


DROEs in Table 3 for subsamples partitioned by these variables.2 In Year 0, the proportions

of positive DROE are significantly less than the expected value for the following categories:

single bankruptcy filings, bankrupt firms with asset size less than industry median, bankrupt

firms with liabilities less than US$120 million, and bankrupt firms with earnings losses. In

Year 1, the proportion of positive DROE is significantly higher for firms with asset size less

than industry median. In Year 2, multiple bankruptcy industries and bankruptcies by small

firms seem to have higher proportions of positive DROE. In addition, the mean DROE of

1.067% in Year 2 for the bankrupt firms whose assets are less than median is significantly

different from zero.

Table 3

Five-year drift-adjusted DROE for (a) multiple and single bankruptcy industries, (b) industries grouped by

bankrupt firm’s assets above and below industry median, (c) industries grouped by bankrupt firm’s total liabilities

above and below US$120 million, and (d) industries grouped by negative and positive earnings of bankrupt firm

Year 0 Year 1 Year 2

Sample N Mean %Pos Mean %Pos Mean %Pos

Multiple bankruptcies 93 0.053 44.1 � 0.086 55.9 0.930 59.1#

Single bankruptcy 72 � 1.113 37.5## � 0.690 52.8 � 3.601 51.4

Difference 1.166 � 0.604 4.531

Asset size >median 67 � 1.095 44.8 � 1.199 47.8 � 4.011 46.3

Asset size <median 90 � 0.013 38.9## 0.151 60.0# 1.067 * 62.2##

Difference � 1.082 1.350 � 5.078a

Liabilities > US$120million 78 � 0.139 44.9 � 0.889 55.1 0.381 55.1

Liabilities < US$120million 78 � 0.795 38.5## 0.039 55.1 � 2.555 56.4

Difference � 0.656 � 0.928 2.936

Negative earnings 150 � 0.546 40.7# � 0.424 54.0 � 1.256 55.3

Positive earnings 15 0.448 46.6 0.399 60.0 1.039 60.0

Difference 0.098 � 0.823 � 2.295

N is the number of industries.a Mean values of two subsamples significantly different at the 10% level using a two-tailed t test.b Mean values of two subsamples significantly different at the 5% level using a two-tailed t test.c Mean values of two subsamples significantly different at the 1% level using a two-tailed t test.



*** Significantly different from 0 at the 1% level using a two-tailed t test.# Proportion of positive value significantly different from equal proportions at the 10% level using a chi-

square test.## Proportion of positive value significantly different from equal proportions at the 5% level using a chi-

square test.### Proportion of positive value significantly different from equal proportions at the 1% level using a chi-

square test.

2 Ferris et al. (1997) provide evidence that the relative size of the bankrupt firms explains intraindustry effects.

They also indicate that their results differ from those of Lang and Stulz (1992) because Lang and Stulz include

bankrupt firms with liabilities in excess of US$120 million only.


Although most of the findings in Table 3 are statistically insignificant to draw any valid

conclusions, they can be summarized as follows: Industries with multiple bankruptcy filings

seem to have an increase in ROE when compared to industries with single filings. For small

firm bankruptcies, the industry ROE decreases in Year 0, but increases in two subsequent years.

In addition, the industry ROE decreases when a firm with earnings losses file for bankruptcy.

Table 4

Results of regressions where the dependent variable, DROE, is regressed on Herfindahl index, stock return

correlation, and debt-to-assets ratio

Dependent variables: drift-adjusted DROE

Independent variable Year 0 coefficient (t value) Year 1 coefficient (t value) Year 2 coefficient (t value)

Constant � 3.135 (� 1.29) � 4.778 (� 2.11** ) � 2.785 (� 0.39)

Herfindahl index 2.383 (3.19*** ) 2.888 (4.19*** ) � 0.227 (� 0.10)

Return correlation � 1.068 (� 0.38) 5.681 (2.17** ) 6.428 (0.78)

Debt-to-assets ratio 3.295 (0.81) 3.044 (0.81) 0.278 (0.02)

F value 3.732*** 7.978*** 0.205

R2 .070 .138 .004

The sample size is 153 industries (12 industries are deleted due to missing stock return correlation data).



*** Significantly different from 0 at the 1% level using a two-tailed t test.

Table 5

Results of regressions where the dependent variable, DROE, is regressed on Herfindahl index, stock return

correlation, debt-to-assets ratio, number of bankruptcy filings in an industry (coded 1 if multiple filings and coded

0 if single filing), relative asset size of bankrupt firm (coded 1 if higher than industry median and coded 0 if less

than industry median), bankrupt firm’s liabilities (coded 1 if higher than US$120 million and coded 0 if less than

US$120 million), and bankrupt firm’s earnings (coded 1 if negative earnings in Year 1 or 0 and coded 0 if positive

earnings in both Years 1 and 2)


Independent variable

Year 0 coefficient

(t value)

Year 1 coefficient

(t value)

Year 2 coefficient

(t value)

Constant � 2.591 (� 0.84) � 4.763 (� 1.67* ) 1.043 (0.12)


Correlation � 2.245 (� 0.68) 7.079 (2.31** ) 4.414 (0.46)

Debt-to-assets ratio 0.763 (0.61) 4.104 (0.35) � 8.548 (� 0.63)

Number of bankruptcies (coded) 0.949 (0.78) � 2.242 (� 0.22) 4.798 (1.36)

Bankrupt firm asset size (coded) � 1.286 (� 0.99) � 0.969 (� 0.80) � 7.043 (� 1.86* )

Bankrupt firm liabilities (coded) 1.578 (1.14) � 1.136 (� 0.89) 5.650 (1.410)

Bankrupt firm earnings (coded) 0.380 (0.18) 0.197 (0.10) � 0.747 (� 0.12)

F value 1.851* 3.549*** 0.995

R2 .086 .153 .048

The sample size is 145 industries (20 industries are deleted due to missing data).





Table 6

Regression results for subsamples partitioned by (a) debt-to-assets ratio, (b) number of bankruptcies in the

industry, (c) relative asset size of the bankrupt firm in the industry, (d) total liabilities of the bankrupt firm, and (e)

earnings of the bankrupt firm at the time of filing


Independent

variables

Year 0 coefficient

(t value)

Year 1 coefficient

(t value)

Year 2 coefficient

(t value)

Debt-to-assets

>Median (N= 72) Constant � 1.571 (� 1.34) � 3.300 (� 2.43** ) � 5.033 (� 0.89)

Herfindahl index 2.506 (3.56*** ) 3.094 (3.81*** ) 0.296 (0.88)

Correlation 0.853 (0.25) 5.854 (1.48) 10.800 (0.66)

F value 6.404*** 8.467*** 0.221

R2 .157 .197 .006

<Median (N= 81) Constant � 0.658 (� 0.41) � 2.533 (� 2.02** ) 0.361 (0.36)

Herfindahl index 1.280 (0.49) 1.529 (0.76) � 3.138 (� 1.96* )

Correlation � 2.801 (� 0.62) 5.964 (1.71* ) 2.217 (0.80)

F value 0.296 1.864 2.129

R2 .008 .046 .052

Number of bankruptcies in the industry

Multiple (N = 85) Constant � 4.388 (� 1.65) � 6.680 (� 2.44** ) � 1.909 (� 0.71)


Correlation 0.114 (0.04) 5.377 (1.70* ) 1.713 (0.55)

F value 5.286*** 9.808*** 0.720

R2 .162 .264 .026

Single (N= 66) Constant � 0.181 (� 0.04) � 0.759 (� 0.19) � 1.811 (� 0.11)

Herfindahl index 1.268 (0.44) � 0.725 (� 0.30) � 0.976 (� 0.09)

Correlation � 3.866 (� 0.72) 6.309 (1.37) 9.062 (0.46)

F value 0.226 0.678 0.092

R2 .011 .031 .004

Asset size of bankrupt firm

>Median (N= 62) Constant � 3.649 (� 0.77) � 9.660 (� 2.69*** ) � 4.805 (� 0.27)

Herfindahl index 5.193 (2.41** ) 4.771 (2.92*** ) 0.244 (0.03)

Correlation � 0.947 (� 0.17) 9.987 (2.36** ) 15.668 (0.76)

F value 2.013 6.072*** 0.210

R2 .093 .236 .011

<Median (N= 83) Constant � 2.363 (� 0.88) � 1.741 (� 0.58* ) � 1.151 (� 0.40)

Herfindahl index 1.759 (2.49** ) 2.350 (2.98*** ) � 0.655 (� 0.86)

Correlation � 1.835 (� 0.51) 3.456 (0.86) 2.348 (0.60)

F value 2.316* 3.296** 0.570

R2 .080 .110 .021

Liabilities of bankrupt firm

>US$120 million (N= 68) Constant � 4.433 (� 1.16) � 9.055 (� 2.57** ) � 1.602 (� 0.39)


Correlation � 3.470 (� 0.89) 5.539 (1.53) 5.051 (1.21)

F value 5.163*** 5.857*** 0.768

(continued on next page)


We perform regression analysis in order to investigate whether DROE can be explained by

Herfindahl index, stock return correlation, debt-to-assets ratio, and the bankruptcy- and

industry-related variables listed in Table 3. We first regress DROE on the Herfindahl index,

stock return correlation, and debt-to-assets ratio. Consistent with our previous findings in

Table 2, the results presented in Table 4 indicate that there is a positive association between

DROE and Herfindahl index in Years 0 and 1. The coefficient for the Herfindahl is 2.383

(t value = 3.19) in Year 0 and 2.888 (t value = 4.19) in Year 1. In Year 1, the coefficient of the

stock correlation variable is positive and significant (5.681, t value = 2.17), which again does

not provide support for the contagion effect.

In Table 5, we present the regression results by regressing DROE on the following

explanatory variables: Herfindahl index, stock return correlation, debt-to-assets ratio, number

of bankruptcy filings in an industry (coded 1 if multiple filings and coded 0 if single filing),

relative asset size of bankrupt firm (coded 1 if higher than industry median and coded 0 if less

than industry median), bankrupt firm’s earnings (coded 1 if negative earnings in either Year 1

or 0 and coded 0 if positive earnings in both Years 1 and 2), and bankrupt firm’s liability size

(coded 1 if higher than US$120 million and coded 0 if less than US$120 million).

As expected, the regression results in Table 5 show that Herfindahl index has a positive and

significant relationship with DROE. In Year 0, the coefficient is 2.502 (t value = 3.11), and in

Year 1, the coefficient is 2.724 (t value = 3.65). These results are consistent with the argument

Table 6 (continued )


Independent

variables

Year 0 coefficient

(t value)

Year 1 coefficient

(t value)

Year 2 coefficient

(t value)

R2 .192 .213 .034

<US$120 million (N= 76) Constant � 2.318 (� 0.59) � 4.583 (� 1.26) 1.375 (0.09)

Herfindahl index 1.691 (1.98** ) 2.071 (2.59** ) 0.583 (0.17)

Correlation 0.055 (0.01) 8.989 (1.75* ) 7.268 (0.35)

F value 1.463 4.131*** 0.092

R2 .057 .145 .004

Bankrupt firm’s earnings

Negative (N= 137) Constant � 3.921 (� 1.48) � 5.938 (� 2.40** ) � 4.771(� 0.59)


Correlation � 1.617 (� 0.50) 6.617 (2.20** ) 8.853 (0.91)

F value 4.931*** 8.520*** 0.285

R2 .099 .160 .006

Positive (N= 14) Constant � 4.248 (� 0.69) � 6.539 (� 1.28* ) 8.173 (1.68)

Herfindahl index 0.750 (1.23) 1.349 (2.68** ) � 0.365 (� 0.76)

Correlation � 2.368 (� 0.56) � 1.642 (� 0.47) 0.091 (0.03)

F value 1.082 4.298** 1.444

R2 .228 .540 .282

N is the number of industries in the regressions.





that the rival firms experience increase in earnings in concentrated industries than in less

concentrated industries. With regards to correlation variable, a significant and positive

coefficient of 7.079 (t value = 2.31) in Year 1 is inconsistent with the contagion effect argument.

To investigate whether the associations between DROE and Herfindahl index (i.e., the

competitive effect) observed earlier in our study depend on industry leverage, number of

bankruptcy filings in the industry, relative asset size of the bankrupt firm within the industry,

size of the bankrupt firm measured by total liabilities, and earnings performance of the

bankrupt firm, we estimate regression coefficients for subsamples categorized by these

attributes. The results in Table 6 show that the positive competitive effect of bankruptcy is

observed in high-leverage industries. For industries with debt-to-assets ratios greater than the

median, the Herfindahl index coefficients of 2.506 (t value = 3.56) in Year 0 and 3.094

(t value = 3.81) in Year 1 are significant. In Year 2, there is a decrease in ROE for industries

with debt-to-assets ratios less than the median. These results suggest that the positive

association between DROE and Herfindahl index exist in highly levered industries.

Another noteworthy finding in Table 6 indicate that the competitive effect is observed in

industries with multiple bankruptcy filings, but not in industries with single filing. The

remaining findings suggest that the positive relationship between DROE and Herfindahl holds

true for both large and small bankrupt firms categorized by relative asset size in the industry.

In addition, a positive relationship is detected for bankrupt firms with liabilities above

US$120 million (the sample selection criterion used by Lang & Stulz, 1992), as well as for

Table 7

Results of regressions where the dependent variable, DOR, is regressed on Herfindahl index, stock return

correlation, debt-to-assets ratio, number of bankruptcy filings in an industry (coded 1 if multiple filings and coded

0 if single filing), relative asset size of bankrupt firm (coded 1 if higher than industry median and coded 0 if less

than industry median), bankrupt firm’s liabilities (coded 1 if higher than US$120 million and coded 0 if less than

US$120 million), and bankrupt firm’s earnings (coded 1 if negative earnings in Year 1 or 0 and coded 0 if positive

earnings in both Years 1 and 2)


Independent variable

Year 0 coefficient

(t value)

Year 1 coefficient

(t value)

Year 2 coefficient

(t value)

Constant � 2.591 (� 0.84) � 4.763 (� 1.67* ) 1.043 (0.12)


Correlation � 2.245 (� 0.68) 7.079 (2.31 ** ) 4.414 (0.46)

Debt� to� assets ratio 0.763 (0.61) 4.104 (0.35) � 8.548 (� 0.63)

Number of bankruptcies (coded) 0.949 (0.78) � 2.242 (� 0.22) 4.798 (1.36)

Bankrupt firm asset size (coded) � 1.286 (� 0.99) � 0.969 (� 0.80) � 7.043 (� 1.86* )

Bankrupt firm liabilities (coded) 1.578 (1.14) � 1.136 (� 0.89) 5.650 (1.410)

Bankrupt firm earnings (coded) 0.380 (0.18) 0.197 (0.10) � 0.747 (� 0.12)

F value 1.851* 3.549*** 0.995

R2 .086 .153 .048

The sample size is 145 industries (20 industries are deleted due to missing data).





firms with liabilities less than US$120 million. DROE and Herfindahl are positively related

when the filing firms have negative earnings, as well as positive earnings at the time of the

bankruptcy filings. These results imply that the competitive effect is variant to the size and

earnings performance of the bankrupt firm.

Overall, the regression findings in Tables 4–6 provide evidence that the earnings of the

bankrupt firms’ competitors increase in concentrated industries than in less concentrated and

that these increases seem to depend on industry leverage and the number of bankruptcies in

the industry.

4.1. Intraindustry effect on operating earnings to sales

To verify the robustness of our results, we examine operating return (OR) measured as

operating income before extraordinary items divided by sales. We compute median values of

Table 8

Regression results for subsamples partitioned by (a) debt� to� assets ratio and (b) number of bankruptcies in

the industry

Dependent variables: drift� adjusted DROE

Independent variables

Year 0 coefficient

(t value)

Year 1 coefficient

(t value)

Year 2 coefficient

(t value)

Debt� to� assets

>Median (N= 72) Constant � 0.073 (� 0.35) � 0.146 (� 0.22) 0.747 (0.12)

Herfindahl index 0.539 (2.74***) 0.251 (0.64) � 0.167 (� 0.43)

Correlation 0.535 (0.56) 0.115 (0.06) 2.001 (1.06)

F value 3.961** 0.204 0.648

R2 .103 .066 .018

<Median (N= 81) Constant � 0.947 (� 1.46) � 0.550 (� 1.00) � 0.747 (0.17)

Herfindahl index � 0.320 (� 0.31) 0.626 (0.71) � 0.167 (� 0.43)

Correlation 0.812 (0.72) 1.777 (1.16) 2.001 (1.06)

F value 0.141 0.991 0.648

R2 .004 .025 .018

Number of bankruptcies in the industry

Multiple (N = 85) Constant � 0.645 (� 1.46) � 1.162 (� 2.43** ) � 0.482 (� 1.12)

Herfindahl index 0.521 (1.98**) 0.237 (0.81) 0.046 (0.17)

Correlation 0.602 (0.48) 3.173 (2.35** ) 1.553 (1.27)

F value 1.989 3.092** 0.827

R2 .046 .069 .020

Single (N= 66) Constant � 0.805 (� 1.32) � 0.759 (� 0.19) 0.059 (0.07)

Herfindahl index 0.711 (0.73) � 0.725 (� 0.30) � 0.829 (� 0.62)

Correlation 0.735 (0.42) 6.309 (1.37) 2.807 (1.11)

F value 0.377 0.678 0.746

R2 .012 .031 .023

N is the number of industries in the regressions.

* Significantly different from 0 at the 10% level using a two� tailed t test.

** Significantly different from 0 at the 5% level using a two� tailed t test.



5-year drift-adjusted changes in OR (DOR) for the bankrupt firms’ industries. In Table 7, we

present results of the regression model that regresses DOR on all of the explanatory variables

examined previously. Consistent with the previous findings, the Herfindahl index coefficient

is positive and significant for Year 0. The index coefficients in Years 1 and 2 are not

statistically significant indicating that the competitive effect when measured by OR exists in

the year of bankruptcy filing only.

Finally, we investigate whether the relationship between DOR and Herfindahl index

depends on industry leverage and number of bankruptcy in the industry as previously

observed. We estimate regressions coefficients for subsamples divided by the median value of

debt-to-assets ratio and number of filings (multiple vs. single filings). According to Table 8,

the Herfindahl index coefficient in Year 0 is positive and significant when debt-to-assets ratio

is higher than the median and when multiple bankruptcies are filed in the industry. These

results are consistent with the ROE findings in Section 4.

5. Conclusions

Following prior evidence of valuation effect of bankruptcy on the competitors, we

investigate whether similar effect is observed on the earnings of the bankrupt firms’

competitors. Our empirical analysis is centered around contagion and competitive effects

of bankruptcy. A contagion effect is associated with negative performance of the rival firms,

whereas a competitive effect indicates improvements in rival firms’ performance.

We detect insignificant results for the total sample, but significant results for sub-

samples partitioned by industry concentration and stock return correlations. We observe

that changes in competitors’ ROE are higher in concentrated (less competitive) industries

when compared to less concentrated (competitive) industries. Furthermore, this compet-

itive effect on earnings is observed for high-leverage industries and when more than one

bankruptcy filing occur within the industry. Our findings, therefore, provide support for

the competitive effect of bankruptcy on competitors’ equity value observed by Lang and

Stulz(1992). Our data, however, fail to detect a contagion effect of bankruptcy on

competitors’ earnings.

References

Benartzi, S., Michaely, R., & Thaler, R. (1997). Do changes in dividend signal the future or the past? Journal of

Finance, 52, 1007–1034.

Cheng, L. T. W., & McDonald, J. E. (1996). Industry structure and ripple effects of bankruptcy announcements.

The Financial Review, 31, 783–807.

Ferris, S. P., Jayaraman, N., & Makhija, A. K. (1997). The response of competitors to announcements of bank-

ruptcy: an empirical examination of contagion and competitive effects. Journal of Corporate Finance, 3,

367–395.

Gay, G. D., Timme, S. G., & Yung, K. (1991). Bank failure and contagion effects: evidence from Hong Kong.

Journal of Financial Research, 14, 153–168.


Kennedy, R. E. (2000). The effect of bankruptcy filings on rivals’ operating performance: evidence from 51 large

bankruptcies. International Journal of the Economics and Business, 7, 5–25.

Lamy, R., & Thompson, G. R. (1986). Penn Square, problem loans, and insolvency risk. Journal of Financial

Research, 9, 103–111.

Lang, L., & Stulz, R. (1992). Contagion and competitive intraindustry effects of bankruptcy announcements: an

empirical analysis. Journal of Financial Economics, 32, 45–60.


the effects of bankruptcy filings on the competitors' earnings

Documents