The Effect of a Common Auditor on the Investment Behavior of
Restating Firms’ Competitors
Name: Jesse Walst
Student number: 10722491
Thesis supervisor: Pouyan Ghazizadeh Date: June 25, 2018
Word count: 13.251
MSc Accountancy & Control, specialization: Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Jesse Walst 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.
Abstract
This thesis investigates the effect of a common auditor on the investment decisions of restating firms’ competitors after a restatement announcement. Prior literature suggests that audit quality and audit reputation is perceived to be lower after a firm restates its financial statements. Therefore, the information upon which competitors base their investment decisions is perceived to be of lower quality as well, which in turn should have an increasing effect on the cost of capital competitors assign to their investment projects. Eventually, in theory, this should result in competitors lowering their investments. Using a sample of 576 restatements and 10.810 competitor firm years, I find enough evidence to support the hypothesis that competitors lower their investments significantly more in case they hire the same external auditor as restating firms do within the same industry. Moreover, I conclude that competitors do not learn about the payoffs of their investment projects. Rather, I show that competitors increase the cost of capital they assign to their own investment projects in case they hire the same external auditor as restating firms do, which explains the decrease in competitors’ investments.
Contents 1 Introduction ... 6 2 Theoretical background ... 8 2.1 Audit Quality ... 8 2.2 Accounting Restatements ... 9 3 Hypothesis Development ... 12 3.1 Restatement externalities ... 12
3.2 Restatements and Competitors’ Investments ... 13
3.3 The Effect of a Common Auditor ... 14
4 Research Design ... 16
4.1 Model ... 16
4.2 Regression Specifications ... 17
4.3 Measurement of Restatement News ... 18
4.3.1 Competitors’ Cumulated Abnormal Returns (CAR1) ... 18
4.4 Measurement of Competitors’ Investments ... 19
4.4.1 Scaled Changes in Competitors’ Investments (ΔINV) ... 19
4.5 Measurement of Control Variables ... 20
4.5.1 Scaled Changes in External Financing (ΔEXTFIN) ... 20
4.5.2 Scaled Changes in Cash (ΔCASH) ... 21
4.5.3 Scaled Changes in Tobin’s Q (ΔQ) ... 21
4.5.4 Scaled Changes in Size (ΔSIZE) ... 21
4.6 Measurement of a Common Auditor Effect ... 22
4.7 Sample ... 22
5 Results ... 28
5.1 Descriptive Statistics ... 28
5.3 Regression Results ... 32
5.3.1 Restatements and Competitors’ Investments ... 32
5.3.2 The Common Auditor Effect ... 34
5.4 Data assumptions ... 36
5.5 Additional Robustness Tests... 38
6 Conclusion ... 40
References ... 42
Appendix A Variable Definitions... 46
Appendix B Model Summaries... 49
Appendix C Normality test ... 50
1 Introduction
This thesis investigates the effect of a common auditor on the investment decisions of restating firms’ competitors after a restatement announcement. Restatement announcements convey new information about restating firms, since such announcements cause a significant decline in the stock price of these firms (Palmrose, Richardson, and Scholz, 2004). However, restating firms are not the only party affected by their own restatement announcements. Competitors of restating firms also experience significant drops in their abnormal returns right after restating firms make their announcement of a restatement, which is caused by a transfer of information from restating firms to their competitors (Durnev and Mangen, 2009; Gleason, Jenkins, and Johnson, 2008; Kravet and Shevlin, 2010). A prior study find that the nature of the information being transferred from restating firms to their competitors, right after restatement announcements, is linked to news about the investment projects of restating firms’ competitors, a so called learning effect (Durnev and Mangen, 2009). Another
explanation for this transfer of information is related to the contagion effect, which means that the quality of accounting information is perceived to be lower for the entire industry than previously thought, as a consequence of restatements (Gleason et al., 2008; Kravet and Shevlin, 2010).
I extend the existing knowledge about the information transfer between restating firms and their competitors, by proposing a different explanation regarding the nature of this information transfer. Specifically, I make an extension to the paper of Durnev and Mangen (2009), by testing whether the negative effect of restatements on competitors’ investments is rather caused by an increase in the cost of capital than by the learning effect. I expect that restatements cause an increase in the cost of capital competitors’ attach to their investment projects in case they hire the same auditor as restating firms do. Since auditor reputation and audit quality of financial statements is perceived to be impaired when auditors of competitors are involved in restatements in the past, uncertainty about the information upon which
competitors value their investment projects arises. This uncertainty would lead to competitors requiring a higher return from their investment projects as a compensation for this increase in uncertainty (Chen, Chen, Lobo, and Wang, 2011). This would be reflected by an increase in the cost of capital used by competitors in the valuation of investment projects. When the cost of capital increases, competitors assign a lower value to their investment projects, which, in turn, causes competitors to lower their investments.
Using a sample of 919 restatements, I conclude that the effect of restatements on competitors’ investment decisions is significantly larger when competitors hire the same auditor as restating firms do than when this is not the case. Specifically, I find that
competitors lower their investments significantly more when they employ the same auditor as restating firms do in the same industry than when they do not. I suggest that this effect is caused by the fact that competitors increase the cost of capital of their investment projects when their auditor is involved in restatements in the same industry rather than by the learning effect as described by Durnev and Mangen (2009).
This thesis makes an important contribution to the literature on accounting quality and investment behavior. By coming up with a novel explanation regarding the nature of
information being transferred from restating firms to their competitors, one is better able to understand the change in investment behavior of competitors as a consequence of
restatements made in the same industry. This is an important extension of the knowledge, since the results make clear that the quality and reputation of the auditor has a significant effect on the investments of competitors. This implies that competitors should carefully look at the history of the auditor they are going to hire in order to make proper investment
decisions based on reliable and well-audited information.
The remainder of this paper is organized as follows. Section two provides the theoretical background and concepts upon which this paper is based. Section three develops the hypotheses. Section four describes the sample selection procedures and the method used in this research. Section five discusses the main findings and section six contains a
2 Theoretical background
This paragraph outlines the theoretical background regarding audit quality and accounting restatements used in this thesis. The first subparagraph introduces the concept of accounting quality. The next subparagraph states when accounting restatements occur and what the implications are.
2.1 Audit Quality
The agency theory describes the relationship between a principal and an agent (Eisenhardt, 1989). When information is transferred from the agent to the principle, uncertainty arises regarding the reliability of this information, since there is a possibility of misalignment of their interests (Anderson and Yohn, 2002). This results in the presence of information asymmetry between the agent and the principle. The agency theory incorporates a third party in this relationship, the monitor, who’s role is to mitigate this information uncertainty (Eisenhardt, 1989). In reality, this conflict could arise between firms, being the agent, and investors, being the principle. Firms periodically dismiss information, in the form of financial statements, regarding their performance in order to inform stakeholders. However, stakeholders, such as investors, are not certain regarding the reliability of this information, since there could be a misalignment of interests between these two parties. Investors seek reliable information from firms in order to make proper investment decisions.
This agency conflict between firms and investors can be mitigated by firms hiring an external independent auditor. This auditor performs an independent audit over the financial statements of the firm by which the auditor is hired. The purpose of such an audit is to verify the information incorporated in firms financial statements and check whether such information is in line with applicable accounting regulations (Stein, 2003). When the auditor concludes that all the information is in line with the regulation, the firm is allowed to publicize its financial statements along with an unqualified audit opinion. This leads to an increase in the reliability of the information incorporated in the financial statements of audited firms, since investors now know that this information is verified by an independent external party (Chen et al., 2011). This means that the information asymmetry between firms and investors is decreased as a consequence of the performed audit. In the context of the agency model, an external auditor takes on the role of the monitor.
information asymmetry between firms and investors. In turn, when information asymmetry is lowered as a consequence of an audit, the cost of capital investors assign to firms is lowered as well (Chen et al., 2011). This means that investors are more eager to invest in such firms (Biddle and Hilary, 2006; Biddle, Hilary, and Verdi, 2009). From the perspective of a firm, it is easier to attract external financing from investors when the corresponding cost of capital is lower (Fazzari, Hubbard, Petersen, Blinder, and Poterba, 1988). However, the contrary holds as well.
Elaborating on audit quality, there are multiple ways to improve it. One way to improve audit quality is to increase the audit tenure. Especially from the view point of investors is audit tenure positively associated with audit quality (Ghosh and Moon, 2005). To put this in other words, the longer the audit tenure, the higher the audit quality. However, prior literature suggests that there is no consistency among the findings regarding the relationship between audit tenure and audit quality, since another stream of research advocates for a negative relation between audit tenure and audit quality (Francis, 2004). Second, industry specialization of the auditor is positively associated with audit quality (Francis, 2004). This implies that in case the auditor is a specialist in its client’s industry, the audit quality likely improves. Also, the bigger the audit firm, the higher the quality of the audit performed (DeAngelo, 1981). More specifically, Big 4 auditors perform audits of higher quality than non-Big 4 auditors (Francis, 2004). However, the degree of quality can differ across different Big 4 auditors and also across different offices belonging to the same audit firm.
However, not all audits are of high-quality. Prior research shows that earnings management is directly related to low audit quality (Becker, DeFond, Jiambalvo, and Subramanyam, 1998). When the auditor performs a low-quality audit, it is likely that earnings management or other omissions in the financial statements are not detected during the audit process. Eventually, this could result in audited financial statements not being in compliance with the applicable accounting regulations.
2.2 Accounting Restatements
Accounting restatements are necessary when previously issued financial statements are discovered to be false or misleading and therefore inconsistent with the applicable accounting principles, also known as GAAP (Durnev and Mangen, 2009; Palmrose et al., 2004). When this is the case, the misstatement or omission and the effect it has on the financial statements must be disclosed by management (Gleason et al., 2008). This effect on the financial
statements can be either income decreasing or income increasing. When the restatement is income decreasing, it appears that accounting techniques used for prior financial statements were aggressive. On the other hand, income increasing restatements are a consequence of conservative accounting practices in the past (Srinivasan, 2005). Accounting restatements are an indicator for a deficiency in the internal control system of restating firms (Kinney and McDaniel, 1987). Such restatements may cause legal liability for management of the restating firm and its external auditor, since this implies that previously issued financial statements are not consistent with the applicable accounting regulation (Francis, Philbrick, and Schipper, 1994).
Such restatements in financial reports convey new information relating to the restating firm itself, since its market value decreases significantly after a restatement announcement is made (Palmrose et al., 2004). This decrease in market value can be found in the stock prices of the restating firm. On average, restating firms’ stock prices fall by 6 to 10 percent right after the moment a restatement is announced (Dechow, Sloan, and Sweeney, 1996; Palmrose et al., 2004). One possible explanation for this is that the credibility of restating firms’ financial statements decreases as a consequence of the restatement (Gleason et al., 2008). Other factors causing this decrease in value include uncertainty regarding integrity of
management and their competence, perceptions about earnings quality in general, and the role of the external auditor in the underlying reasons for the restatement (Gleason et al., 2008; Hribar and Jenkins, 2004). Besides the direct effects causing a decrease in the value of restating firms, the cost of capital of those firms increases after a restatement announcement (Hribar and Jenkins, 2004). This increase in the cost of capital leads also, indirectly, in a loss of firm value since it would be more difficult for the firm to attract capital. However, not all misrepresentations are intentional. Sometimes, restatements are done in order to correct accounting errors not being made on purpose, even though they are material (Gleason et al., 2008). These kinds of restatements have usually no significant impact on restating firms’ financial statements and therefore the effect on their stock prices is also minimal to nihil (Anderson and Yohn, 2002).
However, restating firms are not the only party affected by their own restatements. Competitors of restating firms also experience significant drops in their abnormal returns right after restating firms make their announcement of a restatement (Durnev and Mangen, 2009; Gleason et al., 2008; Kravet and Shevlin, 2010). This is the case since there is a transfer of information from restating firms to their competitors in the same industry.
news regarding the payoffs of the investment projects of restating firms’ competitors, called the learning effect (Durnev and Mangen, 2009), and the other stream of literature advocates for the presence of a contagion effect (Gleason et al., 2008; Kravet and Shevlin, 2010).
3 Hypothesis Development
This paragraph contains the development of the hypothesis that are tested in this thesis. The hypothesis is developed from the theoretical background as mentioned in the previous paragraph as well as the findings by other researchers as outlined below in the following subparagraphs. Subparagraph one discusses the findings of prior studies regarding auditor reputation and audit quality. The second subparagraph outlines the relationship between restatements and competitors’ investments. Finally, the third subparagraph develops the hypothesis of this thesis, which tests the effect of a common auditor on the relationship between restatements and competitors’ investments.
3.1 Restatement externalities
As indicated earlier, restatements can have major consequences for both restating firms and their competitors. However, these parties are not the only ones affected by restatements. The auditor of a restating firm can also experience significant reactions due to the restatement. Chaney and Phillipich (2002) find that auditor reputation may be impaired when the auditor is involved in any kind of audit failure, of which a restatement is an example. They argue that auditors experience difficulties in attracting new clients and keeping existing ones when their reputation is impaired. Consistent with these findings, Irani, Tate and Xu (2002) find that non-restating firms are less likely to hire an auditor who is involved in multiple restatements in the past. They find that the bigger the number of restatements an auditor is associated with, the lower the likelihood that a non-restating firm will hire that auditor. They also conclude that firms are less likely to extend the relationship with their auditor in subsequent years when the auditor is associated with restatements in the past. On top of this, Francis and Michas (2012) find that an auditor involved in a restatement is likely to be involved in a restatement again in the next five years.
Skinner and Srinivasen (2012) find that reputable auditors are associated with high-quality audits. However, Billingsley and Schneller (2009) argue that auditor reputation is in fact rather associated with perceived audit quality than factual audit quality. They conclude that high-quality audits reduce the uncertainty investors have regarding the quality of information stated in financial statements. Moreover, they say that the better the auditor’s reputation, the less uncertainty is left for investors regarding such information. In addition, Chen et al. (2011) find that high-quality audits result in a significant decrease of the audited
firm’s cost of capital. Therefore, according to Barton (2005), firms seek to hire highly reputable auditors in order to enhance the perceived quality of their financial statements and foster the trust investors have in their firms.
The contrary holds as well. When an auditor’s reputation is weak or shredded, the quality of the audit performed by this auditor is perceived to be low. Gleason et al. (2008) state that financial statement credibility may be diminished if a restatement announcement causes investors to question whether the external auditor’s functionality was appropriate or not. Also, Barton (2005) and Chaney and Philipich (2002) show that the shredded reputation of the well-known audit firm Arthur Andersen LLP caused them to lose multiple clients and eventually resulted in bankruptcy, since their performed work was perceived as being of low-quality.
Elaborating on the quality of financial reports, Biddle et. al (2009) find that firms with lower accounting quality tend to engage more frequently in over- and under-investment. They argue that the agency costs of excess free cash flows causes over-investment and capital rationing by external auditors causes under-investment. Another research by Biddle and Hilary (2006) find that information asymmetry is mitigated by higher financial reporting quality. They say that this reduces the chance of over- and under-investment.
3.2 Restatements and Competitors’ Investments
Prior research highlights several ways in which information can be transferred from restating firms to their competitors. Mitchell and Mulherin (1996) argue that competitors use the financial reports of restating firms to obtain information about industry specific conditions, such as demand and cost conditions. Also, Simmonds (1982) finds that competitors make use of restating firms’ financial reports in their own strategic decision-making processes regarding pricing. Another study by Simmonds (1986) indicates that these financial reports are also used to monitor the strategic choices of restating firms. In addition, studies by Elnathan and Kim (1995), Elnathan, Lin, and Young (1996), Cardinaels, Roodhooft, and Warlop (2004), and Maiga and Jacobs (2006) make clear that competitors use restating firms’ financial reports to benchmark themselves against these restating firms. Moreover, Simons (1990), Guilding (1999), and Guilding, Cravens, and Tayles (2000) state that competitors use the information obtained from restating firms’ financial reports to decide on their own investment projects. All of these studies provide evidence supporting the so called learning effect.
When competitors use the information obtained from restating firms’ financials reports to decide on their own investment projects, it is likely that restatements have an effect on the
information that competitors use in this decision-making process. In turn, this implies that a restatement likely effects the investment decisions of competitors, by means of the learning effect. Such expectations are confirmed by the research of Durnev and Mangen (2009). They find that competitors lower their investments right after a restatement is announced. Specifically, these researchers conclude that such decreases in competitors’ investments are significantly associated with competitor’s abnormal returns. Their results demonstrate a transfer of information from restating firms to their competitors regarding the investment projects of competitors, which occurs right after a restatement announcement.
3.3 The Effect of a Common Auditor
As indicated earlier, there are different explanations regarding the nature of information being transferred from restating firms to their competitors. Also, there are multiple factors affecting this transfer of information. Among such factors is the external auditor that is involved in the restatement.
Chaney and Philipich (2002) find that the auditor’s restating client is not the only party that experience the drawbacks caused by a restatement. They say that other non-restating clients of the auditor also experience negative market reactions as a consequence of the impaired reputation of this auditor. Such negative market reactions of non-restating clients suggest that investors are more skeptical about the quality of the audit performed by the auditor. Furthermore, Francis and Michas (2012) find that the quality of restating firms’ competitors’
financial reports is lower than for competitors of non-restating firms in the same fiscal year.
Moreover, Beatty (2013) find that fraudulent financial reporting reduces the investment efficiency of peer firms within the same industry.
Since Durnev and Mangen (2009) conclude that competitors use the financial statements of restating firms to retrieve information about their own investment projects, it is likely that competitors are less certain about the credibility of this information and the information from their own financial statements in case they hire the same auditor as the restating firm. As mentioned earlier, when firms are less certain about the information upon which they decide their investment projects, they require a higher rate of return from such projects in order to compensate for this uncertainty. In other words, competitors higher the cost of capital they assign to their investment projects, which leads to a decrease in the expected value of such projects This would result in competitors lowering their investments.
Based on the literature review as stated before, I expect that competitors lower their investments significantly more after a restatement announcement in case they hire the same external auditor as restating firms do. In order to test this expectation, I develop the following hypothesis:
H1: The positive relationship between competitors’ cumulated abnormal returns and
competitors’ scaled changes in investments is more profound in case competitors hire the same external auditor as restating firms do within the same industry.
.
4 Research Design
This paragraph gives an overview of the research design used in this thesis. The first subparagraph explains the model used to test the hypothesis. Next, the regression specifications are given in order to statistically test the hypothesis of this thesis. Then, the practical measurement of restatement announcements is described. Furthermore, the measurement of competitors’ investments is explained, followed by a brief explanation of its control variables. The fourth subparagraph introduces the measurement of the common auditor effect. This paragraph ends with a description of the sample selection procedures.
4.1 Model
Figure 1 depicts the model that is used to test the relationship between restatements and competitors’ investments. Consistent with the hypothesis, the dependent variable is competitors’ investments (INV). The independent variable is accounting restatements (CAR1). Control variables are competitors’ external financing (EXTFIN), competitors’ cash (CASH), Tobin’s Q of competitors (Q), and competitors’ size (SIZE). To test whether the relationship between restatements and competitors’ investments is more profound when restating firms and competitors hire the same auditor, a common auditor dummy variable (AUDITOR) is added to the model. This variable interacts with the independent variable (CAR1) and results in the interaction variable (CAR1AUDITOR).
FIGURE 1: Model Specification
Figure 1 gives an overview of the model used in this thesis. Definitions of the variables are given in Appendix A Variable Definitions. Measurement of the variables are explained in subparagraphs 4.2, 4.3, and 4.4.
4.2 Regression Specifications
The most important goal of this thesis is to examine the effect of a common external auditor on the relationship between restatements and competitors’ investments. By using the Ordinary Least Square (OLS) method (Keller, 2005), it is possible to investigate this effect and therefore answer the research question of this thesis. The following basic model tests the relationship between accounting restatements and competitors’ investments. The setup for this model is as follows:
𝛥𝐼𝑁𝑉 = 𝛽0+ 𝛽1∗ 𝐶𝐴𝑅1 + 𝜀𝑖
β0 is the intercept, and the error term is depicted by εi. To control for other factors influencing
the investments of competitors, the basic model is augmented with the earlier-mentioned control variables. Adding these control variables leads to the following equation:
𝛥𝐼𝑁𝑉 = 𝛽0+ 𝛽1∗ 𝐶𝐴𝑅1 + 𝛽2∗ 𝛥𝐸𝑋𝑇𝐹𝐼𝑁 + 𝛽3∗ 𝛥𝐶𝐴𝑆𝐻 + 𝛽4∗ 𝛥𝑄 + 𝛽5∗ 𝛥𝑆𝐼𝑍𝐸 + 𝜀𝑖
This model is used to examine the relationship between restatements and competitors’ investments. In order to test the hypothesis of this thesis, the abovementioned equation is extended by adding the common auditor dummy variable (AUDITOR) and the interaction variable (CAR1AUDITOR) as described earlier. This leads to the following equation:
𝛥𝐼𝑁𝑉 = 𝛽0+ + 𝛽1∗ 𝐶𝐴𝑅1 + 𝛽2 ∗ 𝛥𝐸𝑋𝑇𝐹𝐼𝑁 + 𝛽3∗ 𝛥𝐶𝐴𝑆𝐻 + 𝛽4∗ 𝛥𝑄 + 𝛽5∗ 𝛥𝑆𝐼𝑍𝐸 + 𝛽6∗ 𝐴𝑈𝐷𝐼𝑇𝑂𝑅 + 𝛽7∗ 𝐶𝐴𝑅1𝐴𝑈𝐷𝐼𝑇𝑂𝑅 + 𝜀𝑖
4.3 Measurement of Restatement News
Prior literature shows that the measurement of news conveyed in restatements is operationalized in roughly the same manner across various researches. Palmrose et al. (2004) use firms’ cumulative abnormal returns to measure restatement news, since they argue that restatements cause a significant decrease in the abnormal returns of restating firms themselves. They estimate abnormal returns by using a market-adjusted model. Hribar and Jenkins (2004) use the exact same approach. Both researches focus on the cumulative abnormal returns centered around the restatement announcement day. Kravet and Shevlin (2010) make use of a similar approach. However, they put emphasize on the mean cumulative abnormal returns of firms instead of the regular cumulative abnormal returns.
In contrast, Durnev and Mangen (2009) make also use of a different approach besides the cumulative abnormal returns they use to measure restatement news. They identify the restatement amounts as a measure for news conveyed in restatements, since they find that these amounts relate to changes in the investments of restating firms’ competitors. However, they conclude that restatement amounts are less significantly related to changes in competitors’ investments than cumulated abnormal returns are.
4.3.1 Competitors’ Cumulated Abnormal Returns (CAR1)
As described above, the most common measure of news from restatements is a firm’s cumulative abnormal returns. Since this thesis identifies both restating firms and their
competitors, there is a trade-off between the measurement of restatement news captured by restating firms’ cumulated abnormal returns and competitors’ cumulated abnormal returns. Durnev and Mangen (2009) conclude that competitors’ cumulated abnormal returns are a better proxy for restatement news than both restating firms’ cumulated abnormal returns and restatement amounts. They argue that the quality of these proxies depends on their interdependence. More specifically, the stronger the interdependence, the better the proxies. Therefore, the proxy of restatement news used in this thesis is the cumulative abnormal returns of competitors over the period of one day before to one day after the restatement announcement and is estimated using a market-adjusted model. The estimation quality of this model is comparable with abnormal return estimations of the CAPM (Kothari and Warner, 1997). The market-adjusted model estimates competitors’ cumulative abnormal returns (CAR1) as follows:
𝐶𝐴𝑅1 = ∑(𝑐𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑜𝑟𝑠′ 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 – 𝑚𝑎𝑟𝑘𝑒𝑡 𝑖𝑛𝑑𝑒𝑥 𝑟𝑒𝑡𝑢𝑟𝑛)
4.4 Measurement of Competitors’ Investments
In general, to measure any effect triggered by an event, one has to compare two different situations with each other. The first one is the situation before the effect triggering event occurs and the second one is the situation after the effect triggering event occurs (Keller, 2005). In the case of this thesis, a restatement announcement is perceived as the effect triggering event. In particular, comparing the investments of competitors before a restatement announcement with their investments after that same restatement announcement captures the effect such a restatement announcement has on competitors’ investments.
4.4.1 Scaled Changes in Competitors’ Investments (ΔINV)
To measure a firm’s investments, Baker, Stein and Wurgler (2003) scale the sum of that firm’s capital expenditures and research and development costs by the firm’s total assets of prior year. To examine whether restatements have an effect on competitors’ investments, a comparison is made between their investments before a restatement announcement and after a restatement announcement. As relative numbers are more meaningful in statistics than absolute numbers, scaling the changes in competitors’ investments by their investments before the restatement
announcement is deemed necessary in this thesis (Keller, 2005). Therefore, scaled changes in competitors’ investments (ΔINV) are formulized as follows:
𝛥𝐼𝑁𝑉 =𝐼𝑁𝑉𝑡− 𝐼𝑁𝑉𝑡−1 𝐼𝑁𝑉𝑡−1
Period t captures the average investments over the period of three years after the restatement announcement. Period t-1 captures the average investments over the period of three years before the restatement announcement. A more detailed explanation regarding variable calculations is given in Appendix A of this thesis.
4.5 Measurement of Control Variables
News from restatements is likely not the only factor influencing the choice of investment projects by competitors. There are plenty of other factors aiding and/or constraining in this decision-making process. To control for such factors, some additional variables are included in the model and described below, mainly based on the paper of Durnev and Mangen (2009).
4.5.1 Scaled Changes in External Financing (ΔEXTFIN)
Baker, Stein and Wurgler (2003) measure external financing as the sum of equity and debt issues scaled by the book value of assets in the prior period. Competitors with more external financed capital are expected to invest more than competitors with less external financing, since investments require a specific amount of money (Durnev and Mangen, 2009). Therefore the change in competitors’ external financing is expected to have a positive effect on its investments. In accordance with the paper of Durnev and Mangen (2009), the scaled changes in competitors’ external financing (ΔEXTFIN) are measured using the following formula:
𝛥𝐸𝑋𝑇𝐹𝐼𝑁 =𝐸𝑋𝑇𝐹𝐼𝑁𝑡− 𝐸𝑋𝑇𝐹𝐼𝑁𝑡−1 𝐸𝑋𝑇𝐹𝐼𝑁𝑡−1
4.5.2 Scaled Changes in Cash (ΔCASH)
In accordance with Fazzari et al. (1988), firms with a bigger increase in cash are expected to invest more. Durnev and Mangen (2009) define cash as the sum of net income, depreciation and amortization scaled by the book value of assets in the prior year. Following their paper, the scaled changes in competitors’ cash (ΔCASH) are measured as follows:
𝛥𝐶𝐴𝑆𝐻 =𝐶𝐴𝑆𝐻𝑡− 𝐶𝐴𝑆𝐻𝑡−1 𝐶𝐴𝑆𝐻𝑡−1
4.5.3 Scaled Changes in Tobin’s Q (ΔQ)
Tobin’s Q is believed to have an effect on the investments of firms, since stock prices reflect the marginal product of capital (Tobin, 1969). Tobin’s Q is defined as the sum of total assets and market value of equity, minus the book value of equity, all scaled by total assets. Following Durnev and Mangen (2009), the scaled changes in competitors’ Tobin’s Q (ΔQ) are measured as follows:
𝛥𝑄 = 𝑄𝑡− 𝑄𝑡−1 𝑄𝑡−1
4.5.4 Scaled Changes in Size (ΔSIZE)
Following Jensen (1986) and Stein (2003), changes in firm size should have a positive effect on its investments due to various factors. Consistent with Durnev and Mangen (2009), size is defined as the natural logarithm of total assets and its scaled changes (ΔSIZE) are measured as follows:
𝛥𝑆𝐼𝑍𝐸 =𝑆𝐼𝑍𝐸𝑡− 𝑆𝐼𝑍𝐸𝑡−1 𝑆𝐼𝑍𝐸𝑡−1
4.6 Measurement of a Common Auditor Effect
Consistent with the hypothesis of paragraph 3, a potential external auditor effect is examined in this thesis. More specifically, I test whether the effect of restatements on competitors’ investments is more profound when restating firms and their competitors hire the same external auditor. To investigate this potential effect, the external auditor of restating firms in the fiscal year right before the restatement, is compared with the external auditor of their competitors in the fiscal years of the restatement. Following Gleason et al. (2008), a common auditor variable (AUDITOR) is added to the regression as outlined in the next subparagraph. This common auditor variable is a dummy variable, which can either take a value of 0 or 1 (Keller, 2005). A value of 0 indicates that the restating firm and its competitor do not hire the same external auditor, whereas a value of 1 means that they do. This dummy variable interacts with the cumulative abnormal returns variable (CAR1) and results in the interaction variable (CAR1AUDITOR).
4.7 Sample
The sample is retrieved from the GAO database, which contains 919 restatement announcements from the 1st of January, 1997 to 30th of June, 2002. The main reason for choosing this particular sample is for comparison purposes. Several prior studies focused on this sample, making prior results comparable with the results of this thesis (Durnev and Mangen, 2009; Gleason et al., 2008). Moreover, new knowledge gained from this thesis adds to the results of prior studies focusing on this specific sample.
Observation for which matching of information from the GAO database with information on Compustat or CRSP was not feasible or information on Compustat or CRSP was missing, are eliminated from the original sample. The remaining sample is controlled for firms changing their fiscal year-end in the year prior to the restatement announcement, as suggested by Durnev and Mangen (2009). Six of such changes are identified and therefore these observations are also eliminated from the sample. These procedures lead to a final sample of 573 restatement announcements.
Table 1 depicts the absolute and relative frequency of the number of restatement announcements per two-digit industry. The highest absolute, as well as relative number of restatements occur in the business service industry (108 and 18,85%, respectively), followed by the electronic and other electrical equipment manufacturing industry (59 and 10,30%,
(39 and 6,81%, respectively). Table 2 shows some descriptive statistics regarding the final restatement sample. Panel A shows an increase in the frequency of restatements over the years. Most of the restatements from this sample are occurred in the year 2001 (133 out of the 573 restatements, e.g. 23,31%). Panel B presents an overview of the prompters of the restatements from this sample. It shows that 37,87% of the restatements is initiated by the restating firm itself, 15,36% by an agency, 6,63% by auditors, and 0,52% by some external party. The remaining 39,62% is unknown. The most frequently involved accounting issues are shown in panel C. Issues regarding revenue recognition appear most frequently (232 times, 40,99%), followed by restructuring, assets, or inventory (79 times, 13,96%), and cost or expenses (74 times, 13,07%). Finally, table 3 presents absolute and relative frequency of different auditors hired by the restating firms in the year prior to the restatement. The auditor involved in most of the restatements is PricewaterhouseCoopers (139 and 24,26%, respectively), followed by Ernst & Young (126 and 21,99%, respectively) and Arthur Andersen, who ceased operations in 2002 (95 and 16,58%, respectively).
TABLE 1
Absolute and Relative Frequency of Final Sample Restatements by Two-Digit SIC Industry 2-digit
SIC Industry Frequency Relative frequency
10 Metal Mining 3 0,52%
12 Coal Mining 1 0,17%
13 Oil & Gas Extraction 13 2,27%
15 Building Construction-gen Contractors 1 0,17%
16 Heavy Construction Except Building 3 0,52%
20 Food & Kindred Products Mfrs 16 2,79%
21 Tobacco Products Mfrs 2 0,35%
22 Textile Mill Products Mfrs 2 0,35%
23 Apparel & Other Finished Products-Mfrs 5 0,87%
25 Furniture & Fixtures Mfrs 3 0,52%
26 Paper & Allied Products Mfrs 6 1,05%
27 Printing Publishing & Allied Industries 6 1,05%
28 Chemicals & Allied Products Mfrs 32 5,58%
30 Rubber & Miscellaneous Plastics Mfrs 6 1,05%
31 Leather & Leather Products Mfrs 1 0,17%
32 Stone Clay Glass & Concrete Prods Mfrs 5 0,87%
33 Primary Metal Industries Mfrs 5 0,87%
34 Fabricated Metal Products Mfrs 4 0,70%
35 Industrial & Commercial Machinery Mfrs 39 6,81% 36 Electronic & Other Electrical Equip Mfr 59 10,30%
37 Transportation Equipment Mfrs 15 2,62%
38 Measuring & Analyzing Instruments-Mfrs 28 4,89%
39 Miscellaneous Manufacturing Inds Mfrs 7 1,22%
42 Motor Freight Transportation/warehouse 1 0,17%
44 Water Transportation 1 0,17%
45 Transportation By Air 3 0,52%
47 Transportation Services 1 0,17%
48 Communications 20 3,49%
49 Electric Gas & Sanitary Services 16 2,79%
50 Wholesale Trade-durable Goods 13 2,27%
51 Wholesale Trade-nondurable Goods 10 1,75%
52 Building Materials & Hardware 1 0,17%
53 General Merchandise Stores 6 1,05%
54 Food Stores 2 0,35%
56 Apparel & Accessory Stores 9 1,57%
57 Home Furniture & Furnishings Stores 4 0,70%
58 Eating & Drinking Places 6 1,05%
59 Miscellaneous Retail 10 1,75%
61 Nondepository Credit Institutions 8 1,40%
62 Security & Commodity Brokers 8 1,40%
65 Real Estate 4 0,70%
66 Combined Real Estate, Insurance, Etc. 2 0,35%
67 Holding & Other Investment Offices 14 2,44%
70 Hotels Rooming Houses & Camps 4 0,70%
72 Personal Services 2 0,35%
73 Business Services 108 18,85%
78 Motion Pictures 4 0,70%
79 Amusement & Recreation Services 7 1,22%
80 Health Services 12 2,09%
82 Educational Services 4 0,70%
83 Social Services 3 0,52%
87 Engineering & Accounting & Mgmt Svcs 12 2,09%
89 Miscellaneous Services Nec 3 0,52%
TABLE 2
Descriptive Statistics for Sample Restatements
PANEL A Fiscal years Frequency
Relative frequency 1997 53 9,25% 1998 58 10,12% 1999 103 17,98% 2000 113 19,72% 2001 133 23,21% 2002 113 19,72% Total 573 100% PANEL B Prompter Auditor 38 6,63% ? 227 39,62% Company 217 37,87% SEC/FASB/CDFI/FDIC 88 15,36% External 3 0,52% Total 573 100,00% PANEL C Issue Revenue recognition 234 40,84% Securities related 31 5,41% Cost or expense 77 13,44% Unspecified 16 2,79%
Acquisitions and mergers 34 5,93%
Restructuring, assets, or inventory 79 13,79% Reclassification 20 3,49% Other 27 4,71% Tax related 6 1,05% Related-party transactions 18 3,14% Loan-loss 2 0,35% IPR&D 29 5,06% Total 573 100%
TABLE 3
Absolute and Relative Frequency of Involvement of Auditors in the Final Sample of Restatements
Code Audit firm Frequency Relative frequency
1 Arthur Andersen 95 16,58%
2 Arthur Young 0 0,00%
3 Coopers & Lybrand 18 3,14%
4 Ernst & Young 126 21,99%
5 Deloitte & Touche 78 13,61%
6 KPMG 75 13,09%
7 PricewaterhouseCoopers 139 24,26%
8 Touche Ross 0 0,00%
9 Other 22 3,84%
10 Altschuler, Melvoin and Glasser 0 0,00%
11 Binder, Dijker, Otte 8 1,40%
12 Baird, Kurtz and Dobson 0 0,00%
13 Cherry, Bekaert and Holland 0 0,00%
14 Clarkson, Gordon 0 0,00% 15 Clifton, Gunderson 0 0,00% 16 Crowe Chizek 0 0,00% 17 Grant Thornton 9 1,57% 18 J H Cohn 0 0,00% 19 Kenneth Leventhal 0 0,00%
20 Laventhol and Horwath 0 0,00%
21 McGladrey and Pullen 2 0,35%
22 Moore Stephens 0 0,00%
23 Moss Adams 0 0,00%
24 Pannell Kerr Foster 0 0,00%
25 Plante & Moran 0 0,00%
26 Richard A. Eisner 1 0,17%
27 Spicer & Oppenheim 0 0,00%
5 Results
This chapter shows the main results of this thesis. The first subparagraph gives some descriptive statistics regarding the variables used in the models. Thereafter, the correlation between the different variables is analyzed. Next, the regression results are shown and interpreted. Subsequently, some assumptions regarding the dataset are tested. Finally, some additional robustness tests are performed.
5.1 Descriptive Statistics
Table 4 presents some descriptive statistics regarding the variables from the final competitor sample used in this thesis.
TABLE 4
Descriptive Statistics for the Final Competitor Sample
Variable N Minimum Maximum Mean Std. Deviation
ΔINV 1354 -0,995 2,070 -0,209 0,419 CAR1 1354 -0,447 0,575 -0,001 0,078 ΔEXTFIN 1354 -3,830 3,900 -0,412 1,195 ΔCASH 1354 -3,771 3,893 -0,343 1,027 ΔQ 1354 -0,906 2,689 -0,004 0,459 ΔSIZE 1354 -0,764 2,520 0,097 0,212 AUDITOR 1354 0,000 1,000 0,257 0,437 CAR1AUDITOR 1354 -0,351 0,575 0,001 0,041
This table presents some descriptive statistics regarding the final sample consisting of competitors in the same four-digit industry as restating firms in the year of the restatement announcement. Competitors with missing data in one of the eight years are excluded from the sample. This sample consists of 10.832 firm-years resulting in 1.354 variable observations. All variables, except for the cumulated abnormal returns (CAR1), the common auditor dummy variable (AUDITOR), and the interaction variable (CAR1AUDITOR) are scaled changes between the three years before the restatement announcement and the three years after the restatement announcement. The variable cumulated abnormal returns (CAR1) is captured by the cumulated abnormal returns for a competitor between one day before and one day after the restatement announcement. A list of all the variable definitions is included in appendix A
The mean of competitors’ cumulated abnormal returns (CAR1) is equal to -0,001, which represents an average decrease of 0,1% in competitors’ cumulated abnormal returns over the 3 days period around a restatement announcement. This finding is quite similar to the reported results by Durnev and Mangen (2009). Based on their sample, they find an average decrease of 0,3% in competitors’ cumulated abnormal returns. Even though this
sample is significantly smaller than the one of Durnev and Mangen (2009), the result is quite similar, indicating a reliable measure of news conveyed in restatement announcements.
The scaled changes in investments (ΔINV) decrease by 20,9% on average over the period of 3 years before and three years after the restatement announcement. This decrease in investments could be related to the restatement announcement of restating firms as explained in the previous sections of this thesis. Exact relationships are discussed in the next
subparagraphs. However, a decrease in competitors’ investments of such a magnitude are extraordinary large. Therefore it is more plausible to think of other drivers for this large decrease than solely restatement announcements. Even though restatement announcements could be partly a driver of such a decrease, it is more likely that an external event, such as an economic crisis, is the main driver. For example, the dot-com bubble occurred in the period 1997 to 2002, which equals the sample period used in this thesis. This bubble caused a world-wide decline in the stock prices of different firms. This could be the main driver for the found decrease in competitors’ investments. Durnev and Mangen (2009) report an average decrease of 14,4% over the same period. Even though 14,4% is substantially lower than the found 20,9% in this thesis, the magnitude of this decrease is still perceived to be large. Therefore I question the statement by Durnev and Mangen (2009) that this decrease in competitors’ investments is mainly explained by restatement announcements. In other words, I question whether this decrease in competitors’ investments is mainly driven by the learning effect as described by Durnev and Mangen (2009). Differences in the results are probably related to the difference in sample size.
Another interesting result from table 4 is the mean of the common auditor dummy variable (AUDITOR), which can be interpreted as the proportion of competitors hiring the same external auditor as restating firms do within the same industry. This mean equals 0,257 indicating that 25,7% of the competitors in this sample hire the same external auditor as restating firms do in the same industry.
Further results indicate an average decrease of 41,2% in competitors’ external financing (ΔEXTFIN), 34,4% in competitors’ cash balance (ΔCASH) and 0,4% in the Tobin’s Q of competitors (ΔQ) over the period of three years before the restatement announcement and three years after the restatement announcement of restating firms in the same four-digit industry. Table 4 also indicates an average increase of 9,7% in competitors’ size (ΔSIZE) over the same period. Durnev and Mangen (2009) also report a decrease in the Tobin’s Q and cash balance of competitors as well as an increase in competitor size. In
contrast to the reported results in table 4, they find an increase in competitors’ external financing rather than a decrease. Therefore, it is assumable that at least three out of the four control variables are reliably estimated in this thesis.
Even after controlling for outliers, some variable minima and maxima, as presented in table 4, are not meaningful in the context of this thesis. Additional robustness tests are
performed later in this chapter, which deals with this problem and checks whether the results still hold after some modifications of the data.
5.2 Correlations among Variables
Table 5 presents Pearson correlation coefficients among the variables from the final competitor sample used in this thesis. An asterisk denotes a significant Pearson correlation coefficient at the 5 percent significance level.
TABLE 5 Pearson Correlations
CAR1-
ΔEXT- AUD- AUD-
ΔINV CAR1 FIN ΔCASH ΔQ ΔSIZE ITOR ITOR
ΔINV 1,000 CAR1 0,014 1,000 ΔEXTFIN 0,226* 0,009 1,000 ΔCASH 0,191* 0,080* 0,123* 1,0000 ΔQ 0,214* -0,039 0,017 -0,003 1,000 ΔSIZE -0,199* -0,012 0,137* -0,039 -0,337* 1,000 AUDITOR -0,002 0,041 0,005 0,007 -0,043 0,049* 1,000 CAR1AUDITOR 0,082* 0,524* 0,018 0,053* 0,021 -0,028 0,046* 1,000
This table presents the Pearson correlation coefficient for the variables from the final sample consisting of competitors in the same four-digit industry as restating firms in the year of the restatement announcement. Competitors with missing data in one of the eight years are excluded from the sample. This sample consists of 10.832 firm-years resulting in 1.354 variable observations. * denotes significant correlation coefficients defined at the 5 percent significance level. All variables, except for the cumulated abnormal returns (CAR1), the common auditor dummy variable (AUDITOR), and the interaction variable (CAR1AUDITOR) are scaled changes between the three years before the restatement announcement and the three years after the restatement announcement. The variable cumulated abnormal returns (CAR1) is captured by the cumulated abnormal returns for a competitor between one day before and one day after the restatement announcement. A list of all the variable definitions is included in appendix A.
Although not significant, table 5 shows that there is a positive correlation at 0,014 between the scaled changes in competitors’ investments (ΔINV) and their cumulated abnormal returns (CAR1). This correlation indicates that positive cumulated abnormal returns result in
positive changes in investments. In the context of this thesis, it is more interesting to put this in other words. Negative cumulated abnormal returns result in negative changes in investments. However, this correlation is relatively small and not significant at the 5 percent significance level. This finding is somewhat in line with the correlation reported by Durnev and Mangen (2009). They also find a positive correlation between competitors’ cumulated abnormal returns and scaled changes in investments. However, the correlation they report is significant at the 1 percent significance level and amounts to 0,031.
On the other hand, table 5 shows a positive correlation between the scaled changes in competitors investments (ΔINV) and their cumulated abnormal returns in case they hire the same external auditor as restating firms do within the same industry (CAR1AUDITOR). In contrast to the previous finding, this correlation is positive and significant at the 5 percent significance level. In other words, positive cumulated abnormal returns result in positive changes in investments. In the context of this thesis, it is more interesting to look at the opposite. In case competitors hire the same external auditor as restating firms do within the same industry, their negative cumulated abnormal returns result in negative changes in their investments. This correlation is larger and more significant in comparison to the correlation between competitors cumulated abnormal returns and their investments in case there is no control for an external auditor effect. This finding is in line with the literature and hypothesis as described in the previous sections.
With regards to the correlation of control variables with the scaled changes of competitors’ investments (ΔINV), again, there are three out of four control variables for which expectations regarding correlations are satisfied. Table 5 shows that the scaled changes in competitors’ cash (ΔCASH) and Tobin’s Q (ΔQ) are positively correlated to competitors’ scaled changes in investments (ΔINV). Also, a negative correlation between the scaled change in competitors’ size (ΔSIZE) and investments (ΔINV) is derived from table 5. These results are in line with the results of Durnev and Mangen (2009) regarding the correlation coefficients. However, table 5 shows a positive and significant correlation between scaled changes in competitors’ external financing and investments, whereas Durnev and Mangen (2009) find a negative and significant correlation. This difference in results could be caused by a different sample size or measurement errors regarding the scaled changes in competitors’ external financing in this thesis.
Overall, table 5 shows no Pearson correlation coefficients exceeding the threshold of 0,7, indicating that multicollinearity is not an issue in the models used in this thesis.
5.3 Regression Results
As indicated earlier, this thesis runs three regressions. First, regression 1 tests the relationship between restatement announcements and competitors’ investments without taking other control variables into consideration. Then, the second regression tests the relationship between restatement announcements and competitors’ investments taking other control variables into consideration. Third, regressions 2 is augmented by inserting a common auditor dummy variable (AUDITOR) and let it interact with the variable for news conveyed in restatements (CAR1), which results in regressions number 3. The results of all regressions are shown and interpreted below.
5.3.1 Restatements and Competitors’ Investments
Table 6 presents the results for all the regressions in this thesis. Coefficients and corresponding t-statistics are shown for each variable. Asterisks denote a significant coefficient at the 5 percent significance level. Standard errors are tested for their robustness to heteroscedasticity, which does not raise any problems in this thesis as explained later on.
Column 1 in table 6 represents the first basic regression. By using the OLS-method, this regression tests the relationship between competitors’ cumulated abnormal returns and the scaled changes in investments. Control variables are not taken into consideration for this very first regression. This column shows a weak, though positive relationship between competitors’ cumulated abnormal returns and the scaled changes in their investments. Durnev and Mangen (2009) also find a positive relationship. However, the results from the regression in this thesis are not significant at the 5 percent significance level, whereas their findings are highly significant. Even though this model shows a positive relationship between competitors’ abnormal returns and the scaled changes in their investments, the explanatory power is nihil as is depicted by the low adjusted R-square of 0,0%. This means that competitors’ cumulated abnormal returns do not explain the changes in their investments. In other words, there is not enough statistical evidence to support the conclusion of Durnev and Mangen (2009) regarding a learning effect. In order to control for other factors that might have an effect on the scaled changes in investments of competitors, regression 1 is extended with four control variables.
TABLE 6 Regression Results Coefficient Independent Variable 1 2 3 β0 CONSTANT -0,209* -0,120* -0,124* (-18,335) (-9,384) (-8,762) β1 CAR1 0,074 0,017 -0,219 (0,505) (0,127) (-1,376) β2 ΔEXTFIN 0,080* 0,079* (8,876) (8,852) β3 ΔCASH 0,064* 0,063* (6,135) (6,117) β4 ΔQ 0,139* 0,136* (5,688) (5,584) β5 ΔSIZE -0,341* -0,340* (-6,368) (-6,357) β6 AUDITOR 0,008 (0,320) β7 CAR1AUDITOR 0,853* (2,817) Adjusted R2 0,0% 14,5% 14,9% N 1354 1354 1354
This table presents the coefficients for the variables from the final sample consisting of competitors in the same four-digit industry as restating firms in the fiscal year of the restatement announcement. Competitors with missing data in one of the eight years are excluded from the sample. This sample consists of 10.832 firm-years resulting in 1.354 variable observations. * denotes significant coefficients defined at the 5 percent significance level. T-statistics are given between the brackets. Standard errors are tested and concluded is that there are robust to heteroscedasticity. The variables regarding competitors’ changes in investments (INV), external financing (EXTFIN), cash (CASH), Tobin’s Q (Q), and size (SIZE) are scaled changes between the three years before the restatement announcement and the three years after the restatement announcement. The variable cumulated abnormal returns (CAR1) is captured by the cumulated abnormal returns for a competitor between one day before and one day after the restatement announcement. A common auditor is captured by a dummy variable (AUDITOR), taking either a value of 0 or 1. This dummy variable also interacts with the variable for competitors’ cumulated abnormal returns (CAR1) and results in the following variable captured by the multiplication of both variables (CAR1AUDITOR). A list of all the variable definitions is included in appendix A.
The second column in table 6 represents the first regression extended with four control variables. Again, regression 2 shows a positive relationship between competitors’ abnormal returns and the scaled changes in their investments, which means that the results from regression 1 still hold. However, this time, the coefficient for competitors’ abnormal returns is weaker than this coefficient as is shown by means of regression 1. Moreover, this coefficient is still not significant at the 5 percent significance level. This would mean that there is still not
enough statistical evidence to support the findings by Durnev and Mangen (2009) regarding the learning effect. Further, regression 2 shows a significant effect for all added control variables at the this specific significance level. Since the explanatory power of model 2 increased by 14,5% in comparison to model 1, it is assumable that this part of the model is fully explained by the control variables instead of competitors’ cumulated abnormal returns. Scaled changes in competitors’ external financing, cash, and Tobin’s Q are positively related to the scaled changes in their investments. These findings are consistent with the findings of Durnev and Mangen (2009). On the other hand, regression 2 shows a negative relationship between competitors’ size and the scaled changes in their investments, which is not in line with their findings. They report a positive relationship between these two variables instead of a negative one in case they use the scaled changes of competitors investments as the proxy for news conveyed in restatement announcements. Another regression by Durnev and Mangen (2009) shows a negative relationship between competitors’ size and the scaled changes in their investments. However, this result only holds in case they use restating firm’s abnormal returns as a proxy for news conveyed in restatement announcements. Therefore, it is assumable that the difference in findings between regression 2 of this thesis and the regression in the paper of Durnev and Mangen (2009) is caused by a difference in sample size and by the fact that the sample in this thesis includes less firms.
All in all, the findings in this section do not support the findings by Durnev and Mangen (2009). This would indicate that competitors do not learn about the payoffs of their own investment projects via the information conveyed in restatement announcements. Although the findings regarding control variables are quite similar.
5.3.2 The Common Auditor Effect
The results of the third regression in this thesis are shown in column 3 of table 6. By augmenting the second regression with a dummy variable (AUDITOR) to indicate whether competitors hire the same auditor as restating firms, and let this dummy variable interact with the variable for competitors’ cumulated abnormal returns (CAR1), it is possible to capture the effect a common auditor has on the relationship between restatements and competitors’ investments, indicated by the new interaction variable (CAR1AUDITOR).
As shown in table 6, the coefficient of competitors’ cumulated abnormal returns (CAR1) takes on a value of -0,219, which means that competitors’ cumulated abnormal returns have a negative effect on the scaled changes of competitors’ investments. However, the
interpretation of this variable is drastically different in model 3 in comparison with the interpretation of this same variable in model 2. Whereas in model 2 this variable captures the total effect of competitors’ cumulated abnormal returns on the scaled changes in their investments, this variable only captures the total effect in model 3 in case the common auditor dummy variable (AUDITOR) takes on a value of 0. More specifically, in case competitors do not hire the same external auditor as restating firms do, the variable (CAR1) captures the total effect of competitors’ cumulated abnormal returns on the scaled changes in competitors’ investments. On the other hand, in case competitors and restating firms do hire the same external auditor, the variable (CAR1) only captures a partial effect of competitors’ cumulated abnormal returns on the scaled changes in competitors’ investments. In that case, the other part of the effect of competitors’ cumulated abnormal returns on the scaled changes in competitors’ investments in captured by the interaction variable (CAR1AUDITOR), which can take on a meaningful value as the dummy variable equals 1 in this situation. This would mean that for competitors not hiring the same external auditor as restating firms do, the effect competitors’ cumulated abnormal returns on the scaled changes of competitors’ investments is captured solely by the variable (CAR1) taking on a value of -0,219. For competitors that do hire the same external auditor as restating firms do, the effect of competitors’ cumulated abnormal returns on the scaled changes of competitors’ investments is captured by cumulating the effect of the variable (CAR1) taking on a value of -0,219 and the interaction variable (CAR1AUDITOR) taking on a value of 0,853, resulting in a combined value of 0,634.
By taking a closer look at the results from column 3 of table 6, one can see that there is a negative and statistically insignificant relationship between the variable (CAR1) taking on a value of -0,219, and the scaled changes in competitors’ investments. This can be interpreted as follows. When competitors do not hire the same external auditor as restating firms do, the cumulated abnormal returns of competitors have a negative effect on the scaled changes of competitors’ investments, although statistically not significant at the 5 percent significance level. This particular finding contradicts the findings of Durnev and Mangen (2009). They claim that competitors learn about the payoffs of their own investment projects following restatement announcements. According to Durnev and Mangen (2009), this learning effect would be the main driver of the decrease in competitors’ investments. However, since table 6 shows a negative relationship between the variable (CAR1) and the scaled changes in competitors’ investments, rather than a positive relationship, I question the findings of Durnev
and Mangen (2009) regarding this learning effect being the main driver of the decrease in competitors’ investments.
Another plausible explanation for this decrease in competitors’ investments is also shown in table 6. This table shows a positive and statistically significant relationship between the interaction variable (CAR1AUDITOR) taking on a value of 0,853, and the scaled changes in competitors’ investments. This can be interpreted as follows. When competitors do hire the same external auditor as restating firms do, the cumulated abnormal returns of competitors have a positive effect on the scaled changes in competitors’ investment, captured by the combined effect of the variable (CAR1) and the interaction variable (CAR1AUDITOR) taking on a combined value of 0,634. This would mean that a common auditor has such an effect on the relationship between competitors’ cumulated abnormal returns and competitors’ scaled changes in investments that it changes both the direction and the magnitude of the relationship.
These findings support the hypothesis as developed in section 3 of this thesis. As table 6 shows, competitors do not learn about the payoffs of their own investment projects following accounting restatements. However, competitors lower their investments when they hire the same external auditor as restating firms do. This is exactly what is expected to be the case as described in section 3 of this thesis. As described in that section, competitors seem to learn about the cost of capital they assign to their own investment projects in case they hire the same external auditor as restating firms do. This causes competitors to increase this cost of capital which results in a decrease in their investments.
All in all, the results from the three regressions indicate that the learning effect regarding the payoffs of competitors’ investment projects, as described by Durnev and Mangen (2009), is not a plausible explanation for the decrease in competitors’ investments. More specifically, the results of this thesis reject the conclusion Durnev and Mangen (2009) draw. Instead, the regression models in this thesis suggest a more plausible driver for this decrease in competitors’ investments. The models show that competitors increase the cost of capital they assign to their investment projects in case they hire the same external auditor as restating firms do within the same industry. Eventually, this leads to a decrease in competitors’ investments.
5.4 Data assumptions
In order to test whether the data used in the thesis is actually suitable for a moderating analysis, eight assumptions need to be satisfied. First, The dependent variable is measured on a
