Intra-industry bankruptcy contagion effects on financial
policy
Abstract
Bankruptcies can have severe consequences for more companies than just the bankrupted company, and it is increasingly argued that industry competitors are among the affected firms. This research contributes to this debate by investigating how a bankruptcy announcement affects the financial policy of industry competitors, through both debt issuance and investment. Using both quarterly as well as annual data for US firms, regressions show that industry competitors are indeed negatively affected in their debt issuance after the bankruptcy of an industry competitor, where no result is found for their investments. When including specifications for industry concentration, the negative effect for debt issuance is mostly found in competitive industries, while it is cancelled out by a positive effect in concentrated industries. These findings are in line with previous research
investigating contagion effects in market valuation and credit conditions, showing that a bankruptcy leads to real long-term effects for industry competitors.
University of Amsterdam, Amsterdam Business School
MSc Business Economics, Finance track
Master Thesis
Rik Stapersma
Juli 2016
2
Statement of Originality
This document is written by Rik Stapersma who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of Contents
I. Introduction... 4
II. Literature Review ... 6
a. Contagion effect ... 6 b. Competitive effect ... 8 c. Empirical evidence ... 9 d. Hypotheses ... 15 III. Methodology ... 17 IV. Results ... 20 a. Descriptive Statistics ... 20 b. Results ... 25 c. Robustness Checks ... 37 V. Conclusion ... 49 References ... 54 Appendix ... 57 Tables 1. Summary Literature Bankruptcy Contagion ... 10
2. Descriptive Statistics for Annual Sample ... 21
3. Descriptive Statistics for Quarterly Sample ... 22
4. Bankruptcy filings per industry ... 23
5. Bankruptcy filings per year ... 24
6. The annual effect of a bankruptcy on industry competitors’ financial policy ... 26
7. The quarterly effect of a bankruptcy on industry competitors’ financial policy ... 29
8. Industry concentration and the annual effect of a bankruptcy on industry competitors’ financial policy ... 31
9. Industry concentration and the quarterly effect of a bankruptcy on industry competitors’ financial policy ... 35
10. Time variation in the annual effect of a bankruptcy on industry competitors’ debt issuance ... 38
11. Time variation in the quarterly effect of a bankruptcy on industry competitors’ debt issuance ... 40
12. Bankruptcy waves and the effect of a bankruptcy on industry competitors’ financial policy ... 43
13. The effect of a large bankruptcy on industry competitors’ financial policy ... 45
14. The impact of CEO age on the effect of a bankruptcy on industry competitors’ financial policy ... 48
4
I. Introduction
In 2015, Vroom & Dreesman, a large chain of department stores in the Netherlands, went bankrupt. This signaled bad news about the industry, since this was a large company with great history. Therefore, competitors in the same industry might become more constrained in their ability to attract finance and in their investments due to this negative information for the industry coming out. This then induces lower levels of investment for these competitors for example. On the other hand, say that for example the Coca-Cola Company goes bankrupt. This can then have positive effects for the competitors of Coca-Cola in the soda production industry, such as PepsiCo. There aren’t many competitors in this industry, and they can reap profits since the demand for their products will now increase. To see which of these two effects dominates and on what this depends, I will investigate the effect of a bankruptcy on the industry competitors of the filing firm.
In general, bankruptcy is seen as bad news for competitors. It signals something about the industry that makes customers, suppliers and possibly also other stakeholders, such as debtholders and shareholders, wary of this industry. Therefore competitors might have to alter their financial policy due to this negative effect. For example, managers of these competitors might become more
conservative in their debt issuance, which can be forced or voluntarily. However, certain competitors might also profit from the bankruptcy. It can increase the competitive position of these competitors since it increases profitable opportunities available to them now that a competitor has gone
bankrupt. Because these effects represent real costs and benefits for firms, it is important to investigate them more deeply.
I will specifically look at the effect of the bankruptcy on certain aspects of financial policy of industry peers. These aspects are investment and debt issuance. The reason I focus on this is that it can show whether bankruptcies have significant real financial policy effects for industry competitors. Because a bankruptcy can be a shocking event in an industry, it can worsen for example credit conditions for industry rivals. I expect managers may act on this by altering their financial policy. There has been previous research on this topic, but this mostly investigated the relationship between a bankruptcy and market valuation, credit conditions and profitability for industry rivals (e.g. Lang and Stultz (1992) and Jorion and Zhang (2007)). These studies found a negative effect on the stock prices and cost of debt for industry competitors after a bankruptcy, something they call a contagion effect.
5 These aspects signal how markets and stakeholders act on the bankruptcy of an industry competitor. It is unclear however how this affects the decisions managers of competitors make, and this is where the contribution of my research lies. Hertzel and Officer (2012) mention that their results for the credit contagion effects of a bankruptcy may imply significant real investment effects for industry competitors. Previous research of e.g. Boone and Ivanov (2012) and Addoum et al. (2014) show financial policy contagion effects of a bankruptcy on other firms such as strategic alliance partners and local peer firms, but it is unknown whether these effects are also prevalent for industry competitors. This is important to investigate since it contributes to fully understanding the consequences of a bankruptcy.
As noted in Hertzel and Officer (2012), the effect of a bankruptcy on industry competitors may depend on industry characteristics. Firms in industries that are less competitive and more
concentrated are expected to act more aggressively upon the bankruptcy of an industry rival, since they want to capture part of the market share left behind by the bankrupted firm (Bolton and Scharfstein (1990)). The positive effect this generates is what previous literature describes as the competitive effect. This effect isn’t expected for less concentrated industries, so here negative contagion effects might dominate. It hasn’t previously been investigated whether companies react to either of these effects in terms of investment and debt issuance. Therefore, my research question is: Does a bankruptcy alter the financial policy of companies in the same industry, and how does this depend on industry characteristics?
I will use industry competitors of bankrupted firms from 2000 until 2007 in the US to estimate the effect the bankruptcy announcement had on the financial policy of these industry competitors. Quarterly as well as annual data will be used to capture the immediate and long-term effects of bankruptcy. A bankruptcy will be considered a firm filing for a chapter 11 bankruptcy, as was done in most previous literature (e.g. Lang and Stultz (1992)). A firm filing for chapter 11 bankruptcy goes into reorganization and possibly arises again after the reorganization. First I will test the overall effect of a bankruptcy announcement on the financial policy of industry competitors. Then I will include an interaction term with the industry concentration to investigate whether the level of industry
concentration matters. Results indicate that overall, less debt is issued by industry competitors after a bankruptcy, whereas no effect is found for investment. These effects are found in both the annual as well as the quarterly sample. When including interaction terms controlling for industry
6 concentration, it is found that the negative debt issuance effect is larger for competitive industries, whereas in concentrated industries there is a positive debt issuance effect that compensates this negative effect. Again this is found in both the annual as well as the quarterly sample, and no effects are found for investment. Further investigations show that the effects are prevalent up until two years after the bankruptcy announcement. This shows that a bankruptcy has significant long-term contagion effects, whereas other papers on contagion effects mostly found short term effects (e.g. Lang and Stultz (1992)). Furthermore, the contagion effects are magnified by the size of the bankrupted firm and are influenced by whether the bankruptcy takes place during a bankruptcy wave. This can indicate that not all bankruptcies hold the same information, and therefore affect industry competitors differently. Finally, it is shown that the contagion effects of a bankruptcy are enhanced when industry competitors are managed by young CEOs, signaling a behavioural aspect to these contagion effects.
The rest of this paper is organized as follows. Section II will summarize and discuss several papers related to the contagion effects of a bankruptcy announcement, and from this hypotheses will be derived. I will propose the methods and data used to investigate the research question in Section III. In Section IV the descriptive statistics, results and robustness checks will be presented and discussed. Finally, Section V concludes.
II. Literature Review
In this literature review I will discuss the two ways a bankruptcy can influence industry competitors. Firstly the theory behind the negative contagion effect is discussed. The positive competitive effect and the theory behind this effect is discussed secondly. Thirdly, I will furthermore discuss empirical evidence on these effects that is found in previous literature. Finally, following this I will derive the hypotheses for my research based on the theory and empirical evidence of the previous literature.
a. Contagion effect
The first paper that came up with the idea about a contagion effect of a bankruptcy on industry competitors was that of Lang and Stultz (1992). They distinguish two channels of the contagion effect. Firstly, they argue that a bankruptcy announcement holds information about the profitability
7 of the cash flows of the filing firm. It also holds information about firms with similar cash flows however. This newly available negative information about the firm is what they call the contagion effect.
Besides this cash flow channel of contagion, a second channel is also mentioned in Lang and Stultz (1992). This channel of the contagion effect works through dealings with a third party, such as customers and suppliers. These parties might see the bankruptcy announcement as a decrease in creditworthiness for all firms in that industry. Because of this, these third parties might require a compensation for the decreased creditworthiness. This can be in terms of higher risk premia for creditors, or lower prices for customers for example. Furthermore, these third parties might shy away from the industry all together, reducing the demand in this industry.
Another channel through which a bankruptcy causes contagion effects for industry peers is mentioned by Benmelech and Bergman (2011). They study the contagion effect through the collateral channel. This collateral channel indicates that a bankruptcy has negative externalities for industry peers through its impact on the value of collateral. This then affects debt financing since the costs of debt are raised. According to Benmelech and Bergman (2011), there are two reasons why an industry peer’s bankruptcy affects the value of collateral. Firstly, a bankruptcy can cause asset fire sales, which then depresses prices for similar assets. Secondly, the demand for industry specific assets is decreased due to the financial distress related to the bankruptcy. A decrease in demand for industry specific assets then again puts downward pressure on the price for these assets, which then reduces the value of collateral.
Based on the channels of contagion mentioned above, what follows is that the industry competitors of bankrupted firms have to pay higher costs of debt, which might lead to managers being more conservative in their debt policy. Furthermore, the decrease in demand together with the higher costs of debt might also lead to managers reducing their investment.
A final channel of bankruptcy contagion is through a negative sentiment among executives. As mentioned in Addoum et al. (2014), after a bankruptcy, managers of related firms are more likely to employ the ‘availability heuristic’ when making decisions. This means that they rely more on recent information in their decision making, and this can cause them to overreact to certain situations.
8 According to Addoum et al. (2014), this can lead to a negative sentiment among managers of related firms after a bankruptcy, which can then cause them to employ a more conservative financial policy.
b. Competitive effect
In contrast to the contagion effect, a bankruptcy can also have a positive effect on industry
competitors. This effect is what Lang and Stultz (1992) call the competitive effect. They describe this effect as follows. In an industry with imperfect competition, firms face an imperfectly elastic demand curve. One firm experiences a decrease in demand because its product is less attractive than that of its competitors. This arises from the bankruptcy. The competitors then experience an increase in demand for their product. Lang and Stultz (1992) thus argue that a bankruptcy announcement holds information about this shift in demand.
Furthermore, Lang and Stultz (1992) mention that the degree of competition matters for the
magnitude of the competitive effect. In perfectly competitive markets, firms will not be able to profit from the bankruptcy of a competitor. This is because these firms individually won’t experience an increase in demand after the bankruptcy announcement. In less competitive markets however firms can increase their prices due to the increase in demand and profit from this.
A second channel through which the competitive effect of a bankruptcy can positively influence competitors is through marginal costs (Lang and Stultz (1992)). The bankrupted firm might produce less efficiently due to the bankruptcy, and this then increases their marginal costs of production. This results in higher prices and lower output. What follows is that competitors now also can increase their prices, since their products are substitutes to the more expensive ones of the bankrupted firm. This way competitors can reap profits from the bankruptcy.
Lang and Stultz (1992) also come up with a final channel of how the competitive effect can work, which is through predation. Since the bankruptcy announcement reveals that the filing firm is weak, competitors can prey on this firm. Predation strategies are for example the lowering of output prices and the increase of investment (e.g. Telser (1966)). The bankrupted firm then has difficulty to respond to these aggressive tactics. Bolton and Scharfstein (1990) provide a model that shows rational predation by industry competitors, to cause a financing problem for the firm in distress. This
9 can then lead to an exit from the industry for these firms in distress. After this has happened, the remaining firms in the industry then have larger market power and can profit from this.
The channels of the competitive effect mentioned above can have several consequences for the financial policy of the industry competitors of the bankrupted firm. Firstly, due to the increased demand or generally due to the increased market power, firms are expected to grab more profits. This leads to an improved creditworthiness of these companies, which will lead to lower risk premia and thus lower loan spreads (Jorion and Zhang (2007)). What then follows is that companies can be less conservative in their debt policy, and increase their debt levels. Another reason why competitors might increase their debt levels is because of the increase in demand for the products of these firms. Furthermore, the fact that competitors of bankrupted firms in chapter 11 tend to prey on these firms will make them increase investment, which might also raise the needs for external financing. As mentioned before, the magnitude of these competitive effects depends on the degree of competition of the industry.
c. Empirical evidence
There are several empirical papers that investigated the contagion and competitive effects of a bankruptcy on industry competitors. Table 1 gives an overview of the most relevant literature on this topic. One of the first papers that studied this contagion effect is that of Lang and Stultz (1992). They studied the effect of a bankruptcy announcement on the stock price of industry competitors.
Whether the contagion or competitive effect dominates is unclear according to them. They also take industry characteristics into account. These characteristics are the industry leverage and the degree of competition, where the degree of competition is measured by industry concentration. They studied both effects by looking at the abnormal return of a value weighted portfolio of industry competitors after a firm filed for chapter 11 bankruptcy. These firms had liabilities larger than 120 million dollar, and the time period investigated is 1970-1986. Their results show that on average there was a negative effect on returns, indicating a contagion effect. However, they also found that for highly concentrated industries with low leverage, there was a competitive effect. The equity value of these firms is the most strongly affected, so therefore the largest competitive effect was found there. From this they derive that a bankruptcy announcement can be good news for competitors.
10
Table 1 – Summary Literature Bankruptcy Contagion
This table gives an overview of the findings of the literature about bankruptcy contagion. More specifically, it shows the author, year of publishing, measure investigated, type of companies investigated, findings and sample period of the most relevant literature on this topic.
Author (Year of publishing)
Measure Investigated Companies investigated
Findings Time period
Lang and Stultz (1992)
Stock returns Industry competitors Contagion effect and competitive effect
found
1970-1986
Ferris et al. (1997) Stock returns Industry competitors Contagion effect found; competitive effect found before
bankruptcy
1979-1989
Jorion and Zhang (2007)
CDS spread Industry competitors Contagion effect for Chapter 11, competitive effect for
Chapter 7
2001-2004
Hertzel and Officer (2012)
Loan spread Industry competitors Contagion effect and competitive effect
found
1981-2004
Benmelech and Bergman (2011)
Loan spread Competitors in airline industry
Contagion effect through collateral
channel
1990-2005
Kennedy (2000) Profitability Industry competitors of 51 largest bankruptcies Contagion effect before bankruptcy, competitive effect after 1982-1992
Iqbal (2002) Profitability Industry competitors Competitive effect found
1991-1996
Addoum et al. (2014) Investment and debt issuance
Local peers (based on geographic proximity)
Contagion effect found
1986-2006
Jorion and Zhang (2009)
Stock returns and CDS spread
Creditors Contagion effect found
1999-2005
Boone and Ivanov (2011) Stock returns, profitability and investment Strategic alliance partners Contagion effect in stock returns, profitability and investment 1989-2007
Ferris et al. (1997) also studied contagion effects on stock prices of competitors after a chapter 11 bankruptcy. They extended the work of Lang and Stultz (1992) by looking at small firm as well as large firm bankruptcies. Similar results as those of Lang and Stultz (1992) are found for large firms, whereas small firm bankruptcies show a significant contagion effect for competitors as well. These effects for small firm bankruptcies are smaller in size however. Furthermore they investigate the
11 contagion and competitive effect by dividing their sample into contagion candidates and competitive candidates. The contagion candidates are competitors that file for bankruptcy in the following three years, and therefore are mostly expected to have suffered from a contagion effect. The competitive candidates are competitors that didn’t file for bankruptcy in the next three years, and therefore Ferris et al. (1997) expect these firms are more likely to experience a competitive effect. For both groups they find a negative effect, where the effect is somewhat larger for the contagion candidates. According to Ferris et al. (1997) this indicates that the competitive effect is small, which they further confirm when performing additional tests for the competitive effect. They argue that this is because the competitive effect is already incorporated in the stock price before the bankruptcy
announcement of the competitor. This is also supported by evidence they find, namely a positive stock price reaction for competitors the hundred days prior to the bankruptcy announcement.
In addition to these papers, Hunsader et al. (2013) also find supportive evidence of the contagion effect in stock returns. They specify competitors as complements and substitutes based on a competitive strategy measure, and they find a negative effect for complements, while they do not find this effect for substitutes. This shows that besides industry concentration, other aspects also matter for the contagion effect. Cheng and McDonald (1996) find a contagion effect in the stock returns of the interdependent railroad industry, and a competitive effect in the airline industry, where firms exhibit market power. Furthermore, two case studies of Akhigbe et al. (2005) and Akhigbe et al. (2005), who investigated the bankruptcies of Enron and WorldCom respectively, show supportive evidence of the contagion effect for industry rivals. Haensly et al. (2001) on the other hand fail to find the same results as Lang and Stultz (1992) when they investigate a different time period, namely the period 1985-1993. Their explanation for their findings is a difference in legal regimes that has occurred over time. Most papers, however, thus tend to show a negative contagion effect in stock returns of competitors after a bankruptcy announcement, as well as a positive
competitive effect in less competitive, more concentrated markets.
Whereas the paper of Lang and Stultz (1992) mainly looks at the contagion effect in terms of stock returns, for my research it is also interesting to see whether there is an effect in terms of credit conditions as well. This is what Jorion and Zhang (2007) investigated. They looked at the effect of a bankruptcy announcement on the credit default swap (CDS) spread of industry competitors. A credit default swap is an instrument for insurance in the case a firm defaults on an obligation. The spread
12 on this instrument thus indicates the creditworthiness of a firm, with a higher spread indicating a lower creditworthiness.
More specifically, Jorion and Zhang (2007) studied the effect of chapter 7 and chapter 11
bankruptcies, as well as something they call a ‘jump event’. This is a sudden jump of the CDS spread of a firm, which is a signal for credit deterioration. The findings of Jorion and Zhang (2007) indicate a contagion effect for chapter 11 bankruptcies in that the CDS spread of industry competitors rose, signaling a lower experienced creditworthiness of these firms. Furthermore they find a competitive effect for chapter 7 bankruptcies and again a contagion effect for jump events. Their argument behind these findings is as follows. After a chapter 11 bankruptcy, a firm can get reorganized, its debt burden is alleviated and it (potentially) emerges again as a strong competitor. Therefore the
contagion effect dominates for industry competitors: these competitors are hurt by the emergence of the bankrupted firm with lower costs. After a chapter 7 bankruptcy however, the firm is liquidated and exits from the industry. Industry competitors can profit from this and a competitive effect is seen here. Finally, after a jump event a strong contagion effect is found because there are no competitive effects here since the firm doesn’t go bankrupt.
Contagion effects also affect the structure of loans to competitors of a bankrupted firm, as shown by Hertzel and Officer (2012). Their study tries to capture the price effects of contagion on loans, as well as several non-price dimensions. The measure they use as a proxy for the price effect is the loan spread. Furthermore, Hertzel and Officer (2012) distinguish between isolated bankruptcies and bankruptcies during a bankruptcy wave. While they find an increase in loan spread for both groups after the bankruptcy announcement of a competitor, the effect for isolated bankruptcies is smaller however. These effects are prevalent in loans extended in the two years surrounding the bankruptcy announcement. Similar to Lang and Stultz (1992), they also look at whether industry concentration is important for the contagion effect. This is because the same argument holds as before, some firms might be able to benefit from the bankruptcy of a competitor. Consistent with the findings from previous literature, Hertzel and Officer (2012) find an increase in loan spread in competitive markets and a decrease in concentrated markets, of such size that the competitive effect cancels out the contagion effect in these markets.
13 In addition to the effects of a bankruptcy announcement on the pricing of credit for industry
competitors, Hertzel and Officer (2012) also looked at the non-price components of credit
agreements. An example of these are the covenants in a loan contract. They find that more collateral is used after the bankruptcy of an industry competitor and an increase in the number of covenants in the loan contract. Furthermore, they also find that the loan is more likely to be extended by sole lenders instead of a syndicate. The argument behind this is that sole lenders can better and more extensively monitor the borrower. Hertzel and Officer (2012) conclude by stating that the higher costs of debt due to loan spread contagion potentially have long-term effects on decisions about financial policies, something my study will investigate.
The collateral channel of the contagion effect is studied by Benmelech and Bergman (2011). They investigate the airline industry, since the assets that are pledged as collateral there, aircraft, are highly industry specific. More specifically, they perform a regression analysis to test the effect of an airline’s bankruptcy on the credit spread of airlines that have great overlap in assets with the
bankrupted airline. Their results indeed indicate a contagion effect through the collateral channel, in that the cost of debt rises for firms that have great overlap in assets with the bankrupted firm. This shows that this channel of bankruptcy contagion has a real impact on industry competitors, and this can then also influence the financial policy of these competitors.
Carvalho (2015) studied the valuation effects of the collateral channel of the contagion effect. He looks at a different but related event to a bankruptcy however, namely contagion during industry downturns. More specifically, he studied the valuation of firms when an industry peer’s long-term debt matures at the time of a downturn for the industry. The argument to study this is that when firms have industry specific real assets, financial constraints of industry peers cause these assets to drop in prices. The financial constraints used in this study are long-term debt maturing during an industry downturn. The drop in real asset prices then reduces the ability of firms to raise external financing, which causes the lower valuation. When investigating this empirically through a regression using data on industry downturns from 1970 until 2010, Carvalho (2015) indeed shows that firms have significantly negative abnormal stock returns when the long-term debt of industry peers matures during an industry downturn. The effect is larger for competitive industries with specialized assets, since financial constraints of peers influence the prices of asset more in these industries.
14 These contagion effects can also arise in bankruptcies, when a bankruptcy depresses real assets prices for an industry.
Several studies also investigated the contagion and competitive effect of a bankruptcy on the profitability of the filing firm’s rivals. Kennedy (2000) finds that the operating performance of both the filing firm and its rivals decrease prior to and coincident with the announcement of the
bankruptcy. However, after the announcement, the operating performance recovers. He argues that the decrease in operating performance prior to and coincident with the bankruptcy filing might be due to predation. Furthermore, additional analysis on the industries show that the negative effect is slightly reduced in concentrated industries, which might indicate a competitive effect.
In addition to this, Iqbal (2002) also fails to find a contagion effect on the earnings of competitors after a bankruptcy announcement. This is done using an extended sample compared to Kennedy (2000). He does find a competitive effect in concentrated industries however, in that the competitors experience an increase in return on equity after a bankruptcy filing. Similar to the results of Hertzel and Officer (2012), the magnitude of the effect increases when there are multiple bankruptcies in the same industry.
Some research has also been done on the bankruptcy contagion effects in terms of financial policies. Addoum et al. (2014) focused on investigating the effect of a bankruptcy on the investment, capital structure and cash holdings of local peer firms. They argue that sentiment causes managers to become more conservative, and find that they indeed reduce investment and leverage, while increasing cash holdings. When controlling for default probability, they show that managers overreact in their financial policy. Whereas they focus on geographic proximity as an important driver of the contagion effect, I will look more at the intra-industry effects.
Besides the contagion effects for industry competitors mentioned before, there might be other channels through which a bankruptcy announcement affects other firms. For example, Jorion and Zhang (2009) also find a significant effect for creditors. They find that stock prices drop and CDS spreads increase for creditors after the bankruptcy announcement. These creditors include both industrial firms, where most of the credit takes the form of trade credit, as well as financial firms, where the credit mostly comprises of bonds and loans. Hertzel et al. (2008) find that customers
15 aren’t affected by the bankruptcy of the filing firm, but their findings for suppliers are similar to those of Jorion and Zhang (2009). They furthermore show that these effects also echo through in the industry of the supplier of the filing firm. Other parties that are affected by a bankruptcy are strategic alliance partners, as shown in Boone and Ivanov (2011). Their results show that the non-bankrupt firm experiences a negative stock price shock after the bankruptcy announcement, as well as lower profit margins and lower levels of investment in the subsequent years. These studies show that contagion effects are evident in many companies, and that bankruptcies thus cause severe disruptions.
As mentioned in Hertzel and Officer (2012), the contagion effects of a bankruptcy might have an impact on real investment decisions. This is what my research will investigate by looking at the impact of a bankruptcy on debt issuance and investment of industry competitors. It is important for this to know how managers react to changes in credit conditions due to contagion effects. Graham and Harvey (2001) studied the determinants of capital structure through a survey among 392 CFOs of American companies. They found that the interest rate and overall credit conditions are one of the most frequently mentioned factors that the CFOs take into account when determining debt issuance. Brounen et al. (2006) did a similar survey in Europe, and their results are comparable to the ones of Graham and Harvey (2001). These studies argue that the results indicate that managers time their debt issuance based on market conditions and thus credit conditions. However, it might also hold that following a potential change in credit conditions after a bankruptcy announcement, this then leads to managers timing their debt issuance by for example postponing it. So this suggests that bankruptcy contagion effects may also matter here for debt policy. This is something my research will investigate.
d. Hypotheses
Based on the literature mentioned previously, I will now derive the hypotheses I will test. This is done to provide an answer for my research question. There will be four hypotheses I will test. The first hypothesis is:
H1: a bankruptcy of a firm causes industry competitors to increase debt financing in concentrated markets.
16 The reason this effect is expected is because in concentrated markets, which are less competitive, firms will be able to capture a higher market share after a rival went bankrupt. They can also increase market power. Thus, a competitive effect dominates here, since firms can profit from the
bankruptcy. Because of this opportunity, they will get easier access to debt relative to competitive markets as shown in previous literature (Hertzel and Officer (2012)). This also ties in to the second hypothesis:
H2: a bankruptcy of a firm causes industry competitors to increase investment in concentrated markets.
This is expected since industry rivals will want to capture the market share the bankrupted firm has left behind (Hertzel and Officer (2012)), and therefore invest more to actually take this opportunity. They might compete more aggressively to ensure the exit of the bankrupted firm (Bolton and Scharfstein (1990)), which can be done by increasing investment. Furthermore, demand for these firms might increase (Lang and Stultz (1992)), which can increase investment opportunities.
The third and fourth hypotheses are opposite to the first two. This is because in less concentrated, more competitive markets, competitors won’t be able to reap profits from the bankruptcy. Therefore the third hypothesis is:
H3: a bankruptcy of a firm causes industry competitors to decrease debt financing in less concentrated markets.
Contagion effects dominate in these markets, and firms have less easy access to debt (Jorion and Zhang (2007)). This is because customers and suppliers might think the bankruptcy announcement holds information about the whole industry, and therefore the creditworthiness of the competitors decreases. Furthermore, managers might become more conservative in their debt policy. This is also a reason why the fourth hypothesis is expected:
H4: a bankruptcy of a firm causes industry competitors to decrease investment in less concentrated markets.
Due to the less easier access to external financing, firms might have trouble financing their
investments. Furthermore, since the creditworthiness of the competitors has decreased, it possibly becomes more difficult for them to find profitable investment opportunities. This is because
customers are less willing to do business with these firms after the bankruptcy announcement (Lang and Stultz (1992)).
17
III. Methodology
In this section I will discuss the methods used to test the hypotheses and answer the research question. Furthermore I will discuss the variables used in the models as well as the data used in this research. To investigate the research question, the hypotheses mentioned in the literature review will be tested.
An panel data regression will be used to test the hypotheses. Firstly, I will run a regression of a dummy that indicates the bankruptcy of an industry competitor in the previous period on the financial policy of a firm (1). This is to see the overall effect of a bankruptcy on industry competitors’ financial policy. As mentioned before, the financial policy variables that will be investigated are investment and debt issuance. The second equation I will estimate includes the industry
concentration and an interaction between industry concentration and the dummy indicating the bankruptcy of an industry peer (2). By including this dummy, I can investigate whether the contagion effect depends on industry concentration. The equations I will estimate are the following:
𝑌
𝑖𝑗𝑡= 𝛽
0+ 𝛽
1𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡
−𝑖𝑗𝑡−1+ 𝛽
2𝑋
𝑖𝑗𝑡+ 𝜀
𝑖𝑗𝑡(1)
𝑌
𝑖𝑗𝑡= 𝛾
0+ 𝛾
1𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡
−𝑖𝑗𝑡−1+ 𝛾
2𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡
−𝑖𝑗𝑡−1∗ 𝐻𝐻𝐼
𝑗𝑡+ 𝛾
3𝐻𝐻𝐼
𝑗𝑡+ 𝛾
4𝑋
𝑖𝑗𝑡+ 𝜀
𝑖𝑗𝑡(2)
The indices i, j and t indicate firm, industry and period respectively, while –i is an indication for peer firms. Peer firms are defined as firms with the same 4-digit industry SIC-code, similar to other studies on this topic such as Lang and Stultz (1992). 𝑌𝑖𝑗𝑡 is a measure of financial policy, and this can be the
level of investment or the amount of debt issued, 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡−𝑖𝑗𝑡−1 is a dummy indicating a chapter 11 bankruptcy of a peer in the previous period and 𝐻𝐻𝐼𝑗𝑡 is the Herfindahl-Hirschman Index, a
measure of market concentration for an industry. 𝑋𝑖𝑗𝑡 are control variables, which are firm
characteristics that I will discuss more broadly later on, as well as industry and year fixed effects. 𝜀𝑖𝑗𝑡
is the error term. These equations are based on the ones used in Hertzel and Officer (2012). However, where they used the loan spread as dependent variable, I will use a measure for financial policy as the dependent variable.
18 The dependent variables, the level of investment and the amount of debt issued, are defined as follows. The level of investment is calculated as the capital expenditures plus research and development expense, normalized by total assets. This is similar to the definition of previous research such as Boone and Ivanov (2012). Debt issuance is defined as the change in total assets minus the change in book equity normalized by total assets. Book equity here is calculated as total assets minus total liabilities, minus preferred shares plus deferred taxes. This definition of debt issuance has also been used in for example Addoum et al. (2014).
The coefficient of interest in the first regression equation is 𝛽1, since this will give information about
the effect of an industry competitor’s bankruptcy on a firm’s financial policy. In the second equation 𝛾1 and 𝛾2 are the coefficients of interest. This is because the bankruptcy contagion effect for firms
with low industry concentration will be measured by 𝛾1, while the effect for firms with high industry
concentration will be measured by 𝛾2. The first and second hypothesis will thus be tested by 𝛾2, the
third and fourth hypothesis by 𝛾1. Based on the hypotheses mentioned before, a negative coefficient
is expected for 𝛾1, while a positive coefficient is expected for 𝛾2.
I will look at other factors that influence investment and debt issuance as control variables. This is done to partially control for endogeneity and to make sure that the true effect of the bankruptcy is captured. It might be the case that firms actually get more or less easy access to debt due to a change in the firm’s creditworthiness. The effect I measure will then not only be caused by the bankruptcy of an industry rival. To control for this, I will include a variable that captures the expected default probability of a firm, similar to Hertzel and Officer (2012). The Altman Z-score will be used for this.
Furthermore, other control variables are also added that affect financial policy. For this I will look at the control variables previous literature used, as well as specific control variables for the dependent variables I will use. Firstly, size is added as a control variable. Size will be measured as the natural logarithm of total assets. This is expected to have both a positive impact on debt issuance as well as on investment, based on previous literature (Kadapakkam et al. (1998) and Rajan and Zingales (1995)). The argument behind this is that larger firms are better known in the market and thus can more easily get access to debt financing. Furthermore, since these larger firms have more easy access to external financing, they can also invest more. A second control variable that I will include is
19 profitability, which will be measured by operating earnings. The expected effect of profitability on debt issuance is negative. This is based on a Pecking order argument: firms with higher profitability have more internally generated cash. Therefore these firms have less need for external financing (Frank and Goyal (2007)). In contrast, the expected effect of profitability on investment is positive: more profitable firms have more cash available to invest (Chava and Roberts (2008)). However, because of the definition I use for investment, including research and development expenses that decrease profitability, a negative effect is expected.
A third control variable I will add specifically for the debt issuance regressions is the current
debt/assets ratio. This can be informative about debt issuance, since firms with a higher debt/assets ratio are expected to also issue more debt in the future, as was argued in for example Addoum et al. (2014) and Hovakimian et al. (2004). Debt issuance will be lagged one period compared to the debt/assets ratio, since otherwise there would be simultaneous causality between the two variables. Market to book ratio is also added as a control variable in all regressions. This is because market to book ratio tells something about the market valuation and growth opportunities of a firm. I expect a positive effect on debt issuance, since firms with high market valuation tend to have higher levels of leverage (Dell’Acqua et al. (2013)). The effect of market to book ratio on investment is expected to be positive. The argument behind this is that firms in their growth stage are expected to invest more (Anderson and Gracia-Feijóo (2006)). Tangibility is a final control variable I will add to the regression with debt issuance as the dependent variable. The reason for this is that firm with more tangible assets can more easily issue debt due to the higher collateral value (Rajan and Zingales (1995)). Besides these control variables, industry and year fixed effects are included, as done in previous literature such as Hertzel and Officer (2012).
I will use data of American listed firms, because of availability and since previous research about contagion effects mostly also used this data. Accounting data will be retrieved from Compustat, while data on bankruptcies is retrieved from the website of the LoPucki Bankruptcy Research Database. This data source for bankruptcies has detailed information for bankruptcies of companies with assets larger than 100 million dollar in 1980 dollars. This ensures that the bankruptcies are large enough to have an impact, as argued by Lang and Stultz (1992). I will investigate the period 2000 – 2007, since this way more recent data than in for example Hertzel and Officer (2012) is used. Furthermore, using this time period I will not have to deal with the recent financial crisis, since this crisis might bias my
20 results due to the large number of bankruptcies (Geiger et al (2014)). As shown in previous literature (Jorion and Zhang (2007)), credit contagion already takes places immediately after the bankruptcy filing. Therefore quarterly data will be used since this captures the immediate effect of an industry rival’s bankruptcy. I will use up to 4 lags of quarterly data, as well as 2 lags of annual data. This is done because previous literature (e.g. Hertzel and Officer (2012)) has shown that contagion and competitive effects of a bankruptcy can be prevalent for a significant period of time.
IV. Results
Firstly, the descriptive statistics of both the annual and the quarterly sample will be presented and discussed, as well as the distribution of the bankruptcies. Then the results for the regression
equations presented in the methodology will be given and discussed. Finally, robustness checks will be performed on the results.
a. Descriptive statistics
Table 2 shows the annual descriptive statistics for the 9,795 industry competitors of firms that filed for chapter 11 bankruptcy (see Appendix A for the list of 398 filing firms), where an industry competitor is defined by the 4-digit SIC code. The time period investigated is 2000-2007. The bankrupted firms aren’t taken into account in this sample, since their summary statistics don’t hold much information for the contagion effects besides the size of the bankrupted firm. The size of the bankrupted firms will be taken into account in the robustness checks. The variables that are shown are investment, debt issuance, the Altman Z-score, size, profitability, market-to-book ratio,
debt/assets ratio, tangible assets and the Herfindahl-Hirschman index. On average, the amount firms invest in a year is equal to 14% of total assets. This is higher than the amount of debt issued by firms on average in a year, which is 3% of total assets. The reason the average debt issuance is close to zero is that the debt issuance can be negative as well, which can be seen by looking at the 25th percentile of debt issuance. This is the case when a firm reduces its amount of debt. The Altman Z-score is on average 2.28 with a median of 2.26. In general it is considered that firms with an Altman Z-score below 1.8 have an increased chance of bankruptcy, so on average the firms in the sample are healthy firms. When looking at the descriptive statistics of the other variables, the results don’t hold much surprise. It is interesting to note that the mean for profitability is negative, which indicates that
21
Table 2 – Descriptive Statistics for Annual Sample
This table gives the mean, median, standard deviation, 25th percentile, 75th percentile and number of observations for annual data in the time period 2000-2007 for industry competitors of the firms that went bankrupt that same time period. Industry competitors are defined as companies in the same 4-digits SIC group. Investment is defined as capital expenditures plus research and development expense, normalized by total assets. Debt Issuance is the change in total assets minus the change in book equity normalized by total assets, where book equity is calculated as total assets minus total liabilities minus preferred shares plus deferred taxes. Altman Z-score is a control variable calculated in the following way:1.2A + 1.4B + 3.3C + 0.6D + 1.0E. A is here current assets minus current liabilities divided by total assets, B is retained earnings divided by total assets, C is net income plus interest expense plus taxes divided by total assets, D is common shares outstanding times closing price divided by total liabilities and E is sales divided by total assets. Size is a control variable defined as the natural logarithm of total assets. Profitability is a control variable calculated as the operating income before depreciation plus interest expense, divided by total assets. Market-to-book is a control variable defined as (total assets + (common shares outstanding * closing price) – book equity)/total assets. Debt/Assets is a control variable which is long term debt plus debt in current liabilities, divided by total assets. Tangible is a control variable calculated as the net property, plant and equipment normalized by total assets. HHI is the Herfindahl-Hirschman index, which is calculated as the sum of the squared company’s sales as a fraction of total sales in an industry per quarter
Mean Median SD 25th percentile 75th percentile Observations Investment 0.14 0.08 0.20 0.03 0.17 43,485 Debt Issuance 0.03 0.02 0.65 -0.05 0.11 37,056 Altman Z- score 2.28 2.26 7.38 0.48 4.65 43,904 Size 4.74 4.78 2.79 2.97 6.60 43,904 Profitability -0.02 0.10 0.36 -0.06 0.17 43,745 Market-to-Book 2.62 1.47 2.85 1.03 2.66 42,514 Debt/Assets 0.27 0.21 0.27 0.02 0.40 43,904 Tangible 0.30 0.21 0.27 0.07 0.48 43,877 HHI 0.31 0.24 0.24 0.15 0.40 2,077
firms on average had a negative EBITDA. When looking at the median however, it is positive again, which might indicate that the negative mean is caused by a few relatively large negative
observations. The number of observations of the Hirfindahl-Hirschman index is lower than of the other variables. This is because this variable is calculated per quarter for each industry, whereas other variables are calculated for each company.
Table 3 gives similar descriptive statistics but now for the quarterly sample of industry competitors of bankrupted firms for the time period 2000-2007. As expected, the average investment is now lower with a mean of 4% of total assets and a median of 2% of total assets. This is about one-fourth of the level of investment in the annual sample, which seems logical. Debt issuance however is on average 2% of total assets, which is somewhat similar to the value in the annual sample. This is because both the negative and positive values of debt issuance are expected to become smaller, which cancels each other out. This can furthermore be seen in the fact that the standard deviation for the quarterly sample is smaller than in the annual sample. The Altman Z-score has a mean of 1.71 and a median of
22
Table 3 – Descriptive Statistics for Quarterly Sample
This table gives the mean, median, standard deviation, 25th percentile, 75th percentile and number of observations for quarterly data in the time period 2000-2007 for industry competitors of the firms that went bankrupt that same time period. Industry competitors are defined as companies in the same 4-digits SIC group. Investment is defined as capital expenditures plus research and development expense, normalized by total assets. Debt Issuance is the change in total assets minus the change in book equity normalized by total assets, where book equity is calculated as total assets minus total liabilities minus preferred shares plus deferred taxes. Altman Z-score is a control variable calculated in the following way:1.2A + 1.4B + 3.3C + 0.6D + 1.0E. A is here current assets minus current liabilities divided by total assets, B is retained earnings divided by total assets, C is net income plus interest expense plus taxes divided by total assets, D is common shares outstanding times closing price divided by total liabilities and E is sales divided by total assets. Size is a control variable defined as the natural logarithm of total assets. Profitability is a control variable calculated as the operating income before depreciation plus interest expense, divided by total assets. Market-to-book is a control variable defined as (total assets + (common shares outstanding * closing price) – book equity)/total assets. Debt/Assets is a control variable which is long term debt plus debt in current liabilities, divided by total assets. Tangible is a control variable calculated as the net property, plant and equipment normalized by total assets. HHI is the Herfindahl-Hirschman index, which is calculated as the sum of the squared company’s sales as a fraction of total sales in an industry per quarter
Mean Median SD 25th percentile 75th percentile Observations Investment 0.04 0.02 0.06 0.01 0.04 138,967 Debt Issuance 0.02 0.00 0.23 -0.02 0.04 133,296 Altman Z- score 1.71 1.37 7.45 0.04 3.34 145,316 Size 4.65 4.72 2.77 2.89 6.53 145,315 Profitability -0.04 0.03 0.27 -0.01 0.05 136,272 Market-to-Book 2.67 1.50 2.92 1.06 2.69 139,026 Debt/Assets 0.29 0.23 0.29 0.03 0.42 145,316 Tangible 0.31 0.21 0.28 0.07 0.50 145,111 HHI 0.30 0.24 0.21 0.15 0.38 7,702
1.37, which is lower than in the annual sample and also lower than the critical value of 1.8. It is however not the case that most companies in the sample are in distress. The low average value of the Altman Z-score arises simply due to the fact that quarterly data are used. This lowers certain components of the Z-score such as operating income and sales compared to their annual counterparts. Besides this, the other variables have values according to the expectations. The descriptive statistics for variables such as size are comparable to the ones using the annual sample, because those variables are balance sheet data that don’t change much when looking at quarterly or annual data. Other variables that depend on whether annual or quarterly data are used, such as profitability, do become smaller when using quarterly data. Furthermore it can be seen that there are roughly 4 times more observations for the quarterly dataset than for the annual dataset, which is according to expectations.
23
Table 4 – Bankruptcy filings per industry
This table shows the industry distribution of the filed bankruptcies for the sample of all bankruptcies in 2000-2007. The industries are defined by their 2-digit SIC Code.
Industry 2-digit SIC # %
Communications 48 63 15,83%
Business Services 73 26 6,53%
Primary Metal Industries 33 17 4,27%
Electronic And Other Electrical Equipment And Components
36 16 4,02%
Electric, Gas, And Sanitary Services 49 16 4,02%
Health Services 80 15 3,77%
Transportation By Air 45 14 3,52%
Fabricated Metal Products Except Machinery and Transportation Equipment
34 13 3,27%
Transportation Equipment 37 13 3,27%
Industrial and Commercial Machinery and Computer Equipment
35 12 3,02%
Chemicals and Allied Products 28 10 2,51%
Wholesale Trade-durable Goods 50 10 2,51%
Miscellaneous Retail 59 10 2,51%
Textile Mill Products 22 9 2,26%
Paper and Allied Products 26 8 2,01%
Stone Clay Glass And Concrete Products 32 8 2,01%
Wholesale Trade-non-durable Goods 51 8 2,01%
Non-depository Credit Institutions 61 8 2,01%
Insurance Carriers 63 8 2,01%
Food and Kindred Products 20 7 1,76%
Rubber and Miscellaneous Plastics Products 30 7 1,76%
Holding And Other Investment Offices 67 7 1,76%
Other - 93 23,37%
Total 398 100%
The distribution of the number of bankruptcy filings across industries is shown in table 4. These industries are defined as 2-digits SIC groups, as done previously in Boone and Ivanov (2011). The total number of bankruptcies in the 2000-2007 time period included in the sample are 398. These
bankruptcies are distributed across 51 2-digit SIC codes and 184 4-digit SIC codes. To see whether the bankruptcy distribution is similar across industries to the one used in Boone and Ivanov
(2011), I will use the 2-digit SIC codes in table 4. It can be seen that most bankruptcies arose in the Communications, Business Services and Primary Metal industries, with 62, 26 and 17 bankruptcies respectively, or 15.83%, 6.53% and 4.27% of total bankruptcies respectively. Furthermore, the 22 industries with the most bankruptcies account for approximately 76.63% of the total bankruptcies in
24
Table 5 – Bankruptcy filings per year
This table shows the distribution of the filed bankruptcies per year for the sample of all bankruptcies in 2000-2007.
Year # % 2000 80 20,10% 2001 97 24,37% 2002 83 20,85% 2003 57 14,32% 2004 29 7,29% 2005 25 6,28% 2006 14 3,52% 2007 13 3,27% Total 398 100%
the sample. This large number of bankruptcies in certain industries might indicate the possibility of a bankruptcy wave in these industries, which can then amplify the contagion effects of a bankruptcy as shown in Hertzel and Officer (2012). However, it is also necessary to look at the yearly distribution of the bankruptcy filings, since the bankruptcies might be clustered in a certain year. The influence of bankruptcy waves on the contagion and competitive effect of a bankruptcy announcement is something I will investigate more broadly when performing robustness checks on the results. When comparing the industry distribution of the bankruptcy filings with the distribution of Boone and Ivanov (2011), the industries that have the most bankruptcy filings are quite similar. This strengthens the belief that these industries do indeed tend to have more bankruptcies than others.
The distribution of the bankruptcy filings per year is shown in table 5. It can be seen that most bankruptcies were filed in 2001 with 97 bankruptcies, or 24.37% of all bankruptcies in the sample. In general, there is a pattern visible with most bankruptcies arising in the early 2000s, with a decreasing number of bankruptcies in the following years. This might be due to the collapse of the internet bubble in the beginning of the sample period and the subsequent crisis, which possibly caused many firms to file for bankruptcy. This would also explain why there are fewer bankruptcies in the
following years. However, as Hertzel and Officer (2012) show, the amount of bankruptcies in the early 2000s isn’t much higher than in the period before, so these years aren’t considered
extraordinary in the sample, similar to Hertzel and Officer (2012). Therefore I don’t make a distinction between the beginning and the end of the sample period. The pattern of bankruptcies that is shown in table 5 is also visible in the papers by Hertzel and Officer (2012) and Jorion and Zhang (2007).
25 b. Results
Next I will discuss the results of the first regression mentioned in the methodology.
𝑌
𝑖𝑗𝑡= 𝛽
0+ 𝛽
1𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡
−𝑖𝑗𝑡−1+ 𝛽
2𝑋
𝑖𝑗𝑡+ 𝜀
𝑖𝑗𝑡This model estimates the effect of a bankruptcy on the financial policy of industry competitors of the bankrupted firm, where financial policy can be either debt issuance or investment. A dummy variable indicates whether there has been a bankruptcy in the previous period or not. The results are shown in table 6. Firstly, the coefficient of interest, 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦𝑡−1, has a negative sign in the debt issuance
regressions, which means that industry competitors issue less debt the year following the bankruptcy announcement. The expectation for this coefficient was unclear, since either the positive competitive effect or the negative contagion effect might dominate. However, other papers that investigated the effect of a bankruptcy on industry competitors also found a negative effect when not distinguishing between degrees of industry competition (e.g. Lang and Stultz (1992) and Hertzel and Officer (2012)). The coefficient of -0.0349 in specification (1) indicates that firms who experienced an industry competitor going bankrupt in the previous year issued almost 3.5% less debt as a fraction of total assets. When comparing this effect to that found by Addoum et al. (2014) for the contagion effect of a bankruptcy of a local peer, the effect on industry competitors is larger in magnitude. Whereas Addoum et al. (2014) found a decrease in debt levels after a local peer’s bankruptcy of about 0.6% of total assets, the effect of an industry competitor is larger with 3.5% of total assets less debt issued. This indicates that firms react more to the bankruptcy of an industry competitor than to that of a local peer.
All control variables have the expected sign except for the control variable for the Altman Z-score. A positive coefficient is expected here, but the results show a negative coefficient. This is possibly due to the definitions of the dependent variables, since now all types of debt are taken into account. Some of these types of debt may actually increase when a firms is in distress. This is something that will be further investigated later on. However, using other control variables for the probability of default in specifications (2) and (3), the coefficient of interest doesn’t change significantly. In specification (2), the S&P Quality Ranking is used as a control variable instead of the Altman Z-score for the probability of default. The S&P Quality Ranking, formerly known as the S&P Earnings and Dividend Ranking, is a letter grade assigned to companies based on several factors, indicating the quality of that company (Stovall (2015)). Dummies for each ranking are included, where the lower
26
Table 6 – The annual effect of a bankruptcy on industry competitors’ financial policy
This table gives the coefficients and t-statistics for the panel data regressions of annual debt issuance as well as annual investment on 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦𝑡−1, a dummy indicating a bankruptcy of an industry competitor, and control variables. Industry
competitors are defined as companies in the same 4-digits SIC group. Investment is defined as capital expenditures plus research and development expense, normalized by total assets. Debt Issuance is the change in total assets minus the change in book equity normalized by total assets, where book equity is calculated as total assets minus total liabilities minus preferred shares plus deferred taxes. In column 1-3, Debt Issuance is used as the dependent variable, in column 4-6 Investment is used as the dependent variable. Columns 1 and 4 use the Altman Z-score as a control variable for probability of default, columns 2 and 5 use dummies indicating the S&P Quality Ranking for this. In columns 3 and 6, Distress is used as a control variable for the probability of default. Distress is a dummy indicating either the EBITDA being less than interest expense in the previous two years, or EBITDA being less than 80% of interest expense in the previous year. All columns use the control variables as defined in table 2, as well as time and industry fixed effects. *, **, *** indicate statistical significance at a 10%, 5% and 1% level respectively.
(1) (2) (3) (4) (5) (6)
Debt Issuance Debt Issuance Debt Issuance Investment Investment Investment Bankruptcyt−1 -0.0349*** -0.0378*** -0.0374*** 0.00230 0.00242 0.00229 (-3.25) (-3.66) (-3.62) (0.88) (1.02) (0.96) Size 0.0134*** 0.0108*** 0.00969*** 0.00427*** 0.00235*** 0.00142*** (6.71) (5.24) (4.94) (9.38) (5.41) (3.40) Profitability -0.0541* -0.0839*** -0.106*** -0.175*** -0.192*** -0.205*** (-1.86) (-3.35) (-3.50) (-28.07) (-32.61) (-30.34) Market-to-Book 0.0305*** 0.0260*** 0.0262*** 0.0109*** 0.00928*** 0.00909*** (9.11) (9.55) (9.59) (15.99) (14.58) (14.41) Debt/Assets 0.270*** 0.327*** 0.328*** (9.30) (14.04) (14.16) Tangible 0.125*** 0.127*** 0.123*** (4.15) (4.52) (4.38) Altman Z-score -0.00417*** -0.00257*** (-3.38) (-11.68) Distress -0.0385*** -0.0184*** (-2.97) (-6.68) S&P Ranking Dummies? No Yes No No Yes No Time Fixed Effects?
Yes Yes Yes Yes Yes Yes
Industry Fixed Effects?
Yes Yes Yes Yes Yes Yes
Constant -0.291*** -0.276*** -0.270*** 0.0527*** 0.0522*** 0.0618*** (-4.88) (-5.06) (-4.95) (4.95) (3.22) (4.11) Observations 36933 38993 38993 37181 40850 40850 Adjusted R2 0.045 0.043 0.043 0.302 0.312 0.312
rankings imply a higher probability of default. As can be seen in specification (2), the coefficients don’t change significantly. A dummy indicating whether a firm is in distress is included instead of the
27 Altman Z-score in specification (3). This dummy equals 1 if either the earnings before interest, taxes, depreciation and amortization is less than the interest expense in the previous 2 years or if the EBITDA is lower than 80% of the previous year’s interest expense. This is the same definition used as in Addoum et al. (2014) for a firm being in distress. As can be seen in specification (3), the
coefficients of the other variables again don’t change significantly. Furthermore, the coefficient for the distress variable is negative and significant, which is according to expectation since firms in distress will tend to issue less debt. Specifications (2) and (3) thus further confirm that a bankruptcy of an industry competitor leads to a lower level of debt issuance for other firms in the year following the bankruptcy announcement.
When looking at the investment regressions, the coefficient of interest is positive but insignificant however. This would mean that the year following a bankruptcy of an industry competitor, firms do not show a significant increase or decrease in their investment patterns. Neither a competitive nor a contagion effect thus dominates. This contrasts the contagion effects found for both the bankruptcy of a local peer (Addoum et al. (2014)) and the bankruptcy of a strategic alliance partner (Boone and Ivanov (2012)). These studies namely showed that firms will cut on investment after such a
bankruptcy, something that isn’t found in the results for industry competitors. However, it might be the case that for industry competitors, the contagion and the competitive effect cancel each other out, which would mean that the results found don’t indicate that there is no effect at all. To see whether this is the case, the results will be more specifically investigated. The coefficients of the Altman Z-score are also contrasting expectations in the investment regressions. Whether this is because of the definition of the dependent variable will be investigated later. When including other control variables for the probability of default in specifications (5) and (6), the coefficients of interest as well as the control variables again don’t change significantly. Dummies indicating the S&P Quality Ranking are included in specification (5), and the distress dummy is included in specification (6). The distress variable is again negative, indicating that firms in distress invest less. These results confirm the findings of specification (4), namely that firms don’t significantly invest more or less after an industry competitor went bankrupt in the previous year. For both the debt issuance as well as the investment regressions, the adjusted R-squared is comparable to the ones of similar regressions in other papers (e.g. Addoum et al. (2014)).
28 Table 7 also shows the results for the same regression equation mentioned above, but now quarterly data is used. The reason this is done is because several papers such as that of Jorion and Zhang (2007) showed that credit conditions for industry competitors of bankrupted firms are immediately affected after the bankruptcy announcement. This may cause the financial policy to also be affected immediately. As can be seen by looking at the debt issuance regressions, this is indeed the case. The coefficient of interest, 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦𝑡−1, is statistically significant and negative, which indicates that
industry competitors of bankrupted firms issue less debt the quarter following the bankruptcy announcement than other quarters. More specifically, the coefficient is -0.00498 in specification (1). This means that firms of which an industry competitor has filed for bankruptcy in the previous quarter issue almost 0.5% of total assets less than when a competitor didn’t file for bankruptcy in the previous quarter. As expected, this is less than the 3.5% found in the annual sample. This is because the annual effect comprises of both the immediate effect as well as the long-term effect. Whether the long-term effect per quarter is larger than the immediate effect will be further investigated. The result that a contagion effect dominates a potential competitive effect when not distinguishing between degrees of industry concentration is again similar to that of for example Lang and Stultz (1992).
The control variables for the debt issuance regression are according to expectations, again with exception of the Altman Z-score. However, the results are all in line with the ones of the annual regressions. Again, as before, the coefficient of the Altman Z-score might be influenced by the way debt issuance is defined, something that will be investigated later on. When using another control variable for the probability of default in specification (2), namely dummies indicating the S&P Quality Ranking, the coefficient of interest barely changes. The distress variable used as a proxy for
probability of default in the annual sample cannot be calculated using quarterly data, so therefore it isn’t included. However, based on the fact that the inclusion didn’t change the coefficients of interest significantly in table 6, it is assumed that the exclusion here won’t influence the results.
When looking at the investment regressions in specification (3) and (4), it is visible that the
coefficient of interest is slightly negative in the specification (3) and slightly positive in specification (4). However, in both cases the coefficient isn’t statistically significant. This is consistent with the findings for the annual sample. It would mean that in the quarter after a bankruptcy filing, industry competitors of the bankrupted firms do not invest more or less than other firms. So these firms don’t
