Impact of financial misreporting on syndicated loan structure G21

Lancaster University

Department of Accounting and Finance

Impact of financial misreporting on syndicated loan structure

September 2021

This dissertation is submitted in partial fulfilment of the degree of MSc Finance

Impact of financial misreporting on syndicated loan structure

Abstract

This paper examines the impact of financial misreporting on the syndicated loan structure with two aspects, including the number of non-lead banks in the syndicated loan structure and the proportion of total loan held by lead banks. An empirical analysis of a sample of 18,837 syndicated loan issued by 6,399 US listed firms between 1996 and 2017 yields a significant correlation between financial misreporting and the structure of syndicated loan. The results suggest that firms with financial misreporting attract fewer non-lead lenders to invest in their syndicated loan, and that lead banks hold a lower proportion of the loan. The implication of empirical analysis is consistent with the hypothesis proposed in advance.

I confirm that this is my own work and the use of all material from other sources has been properly and fully acknowledged.

Contents

1. Introduction.......................................................................................................................... 1

1. Literature Review................................................................................................................ 4
   
   
   • Financial misreporting......................................................................................................... 4
   
   
   • Syndicate loans and syndication structure........................................................................... 6
   
   
   

1. Hypothesis Development..................................................................................................... 9

1. Empirical Model................................................................................................................ 11
   
   
   • Variables Descriptions....................................................................................................... 11
   
   
   • Sample and Descriptive Statistics...................................................................................... 12
   
   
   • Descriptive Statistics.......................................................................................................... 12
   
   
   • Correlation Matrix............................................................................................................. 15
   
   
   

1. Empirical Results............................................................................................................... 17
   
   
   • Baseline results.................................................................................................................. 17
   
   
   • Additional Analysis........................................................................................................... 19
     
     
     • Past Relationship with Borrowers.................................................................................. 19
     
     
     • Auditors in the BigN...................................................................................................... 20
     
     
     
   
   
   • Robustness Check.............................................................................................................. 22
     
     
     • Including More Control Variable................................................................................... 22
     
     
     • Exclusion Financial Companies..................................................................................... 23
     
     
     
   
   
   

6. Conclusion.......................................................................................................................... 24

6. Reference............................................................................................................................ 25

List of Tables

Table 1 Descriptive Statistics of Variables 13

Table 2 Differences in Mean 14

Table 3 Pearson Correlation Matrix 16

Table 4 Relationship Between Financial Misreporting and Syndicated Loan Structure 17 Table 5 Prior Relationship with Borrowers 20

Table 6 Auditors in the BigN 21

Table 7 The Results of Robustness Check 22

1. Introduction

In the financial sector, financial reporting has always been the fundamental and important source of information, especially in the public market, where investors mainly apply the analysis of financial information to deepen their understanding of the investee firm and then make investment decisions. High-quality financial information provides a timely and accurate picture of companies' performance, thus reducing information discrepancies between internal and external stakeholders as well as agency costs. However, a number of incidents of firm's financial misreporting had dampened market confidence, some of the major ones including the Enron fraud in 2001, TESCO profit overstating in 2014, Carillion's overly aggressive accruals in 2017, etc. The Enron debacle is probably the most famous case of unsuccessful misreporting detection by the SEC. Significant accounting misreporting have emerged frequently in recent years, and studies have shown that disclosed restatements are only the tip of the iceberg (Soltes, 2019; Curtis et al., 2019; Dyck et al., 2021), but each financial misreporting is bound to result in serious business losses.

A firm admitting to a financial misreporting not only has to bear direct explicit costs such as legal fines, but also faces hidden costs such as the evaporation of materially market capitalisation. Dyck et al. (2021) predict that the total market value evaporated in 2020 as a result of corporate fraud is $744bn, so this is not only a devastating blow to the company, but also to the whole equity market. As shareholders are often the most direct bearers of losses, the impact of financial misreporting on equity markets and the structure of shareholdings, among other things, is a popular topic of research. In addition, management involved in distorting financial information can face penalties such as loss of job and jail time when the misconduct is detected, and the motivations and features of managers have also been studied in detail in order to fully understand why financial misreporting occur. Nevertheless, the losses associated with creditors, who are also important corporate stakeholders, have been less studied. Financial misreporting can seriously undermine the trustworthiness of accounting information, deteriorate information asymmetries, and increase the risk of corporate debt. Garmaise (2015) finds a strong correlation between misreporting and adverse loan performance such as serious delinquencies. Moreover, the reaction of creditors to financial misreporting has a direct and pronounced impact on a firm's continued access to financing markets and bargaining power to borrow on favourable terms, and additionally, the adequacy of financing can further affect investment and operational activities of the firm.

The major research focus of this paper is syndicated loan, and the main reason is that in the debt market, an increasing number of debt deals are being syndicated. As of 2018, the issuance of syndicated loans in the US debt market exceeded other common types of credit, including corporate bonds (Faria-e-Castro and Bharadwaj, 2019). Firstly, the emergence of syndicated loan has facilitated the issuance of more jumbo loans, making business activities such as large corporate mergers and acquisitions more active. Secondly, syndicated loan has also allowed lending banks to participate in a more diverse range of industries with less capital outlay, contributing to a more diversified portfolio of holdings while avoiding the need to break capital and lending constraints. Therefore, it can be seen that a number of papers have examined the characteristics of syndicated loan.

The problem of information asymmetry that exists within syndicates has received increasing academic attention in order to optimise the syndicated structure and improve the efficiency of loan deals. The lead arrangers in syndicated loan, who are repeated traders in the market, have privileged access to information about the borrower, and their debt contract relationship with firms is in line with the characteristics of private debt market. However, for the other smaller participant lenders in the syndicate, although they are also at arm's length from the borrower through the same contract, they are dependent on the lead arrangers for credit information and the subsequent due diligence. Lee and Mullineaux (2004) show that a more concentrated syndication structure can enhance monitoring and reduce the cost of renegotiation and increase the efficiency of collective decision-making in the event of financial distress. Bae et al. (2014) find a diverse syndication structure that contradicts the common purely centralised syndication structure. They suggest that the proportion of the number of lead banks and the percentage of loan amounts retained by them decrease when borrowers risk increases, which is explained by the positive cooperation of participate banks to higher compensation from higher loan pricing and a greater degree of risk sharing. The syndication structure can therefore be seen as a composite reflection of information frictions and credit risk. This paper hopes to investigate whether misreporting has an appreciable incremental impact on the structure of syndicated loan by isolating confounding debt contractual characteristics and borrower factors.

Based on previous studies, this paper proposes the hypothesis that compared with firms without financial misreporting, the firms with financial misreporting can attract fewer non- lead banks in syndicated loan, and that lead banks would retain a higher proportion in the total loan deal. In order to test the hypothesis, a multiple regression model with syndicated loan structure as the dependent variable was developed, with the presence of accounting restatement as a dummy variable being the main explanatory variable of the model and controlling for the effects of variables related to debt characteristics and borrower characteristics. The model analysis is based on a sample of 188,337 syndicated loans issued by 6,399 public firms in the US market between 1996 and 2017. The empirical study was conducted using descriptive statistics, correlation tests, regression analysis and robustness testing. The empirical results show that financial misreporting are significantly negatively correlated with the number of participate banks and significantly positively associated with the percentage of lead bank share. The research hypotheses are supported by conclusive empirical evidence.

Overall, the main contribution of this paper is to demonstrate, on the one hand, that financial misreporting has an impact on syndicate structure through channels such as worsening information asymmetries and undermining the credibility of transactions and, on the other hand, that debt structure reflects the combined impact of multidimensional factors on syndicated loan risk.

2. Literature Review

The research question has been divided into two parts, financial missreporting and syndicated loan structures, and then a review of the relevant literature is provided on each. First, the motivation and consequences, especially in equity market, of financial misreporting have been widely examined in the past. Besides, the importance of financial statements and the essential role that the quality of accounting information plays in the contracting process is explored through prior research. From the perspective of financial misreporting as a counterexample of high- quality accounting information, I explored the impact of the existence of financial misreporting on firms and other stakeholders. In addition, a specific review of relevant studies on the influence of financial misreporting on debt contracts is presented. Secondly, with respect to the structure of syndicated loan, I learned about the prevalence of syndicated loan and the advantages of it over other types of credit. Moreover, it is found that the syndicated loan structure can comprehensively reflect the severity of information asymmetry between syndicates and the level of lenders' perceived credit risk. The literature review provides the basis for the hypothesis development in chapter 3.

2.1 Financial misreporting

A financial misreporting is defined by the SEC as a company's financial report containing errors that mislead investors and result in a monetary loss due to increased risk. Amiram et al. (2018) review previous studies of financial misconduct from different perspectives, and they clarify that the motives behind financial misreporting are of two types: capital market motives and contracting motives. The former advocates benefits derived from inflating common share prices, which further include profit from stock-based bonus or insider trading, and also centres on new financing on more favourable contract terms by embellishing accounting figures. The latter, on the other hand, refers to the manipulation of the accounting system to take advantages of contractual terms based on accounting information, for instance, to obtain additional benefits by means of a earning-matched compensation, or to avoid costly accounting-based covenants violation in debt contracts. Besides, Curtis et al. (2019) find that around 90% of the firms in their empirical sample intend to conceal financial misconduct unless they are identified and forced to restate by the SEC or other external enforcement institutions, and firms size, board independence, accounting accrual and other firm-level factors can affect restatement decision. Additionally, Dyck et al. (2021) update their projections of financial misconduct that exist but are not detected, and report that two-thirds of pre-existing corporate frauds may pass scrutiny, while an average of 11% of large publicly traded firms are suspected of securities fraud each year. Their findings demonstrate the difficulty of successfully detect all firms with financial misconducts and illustrate the pervasiveness of the issues.

Whilst some prior papers have examine various misreporting indicators, such as internal control weakness as early signal (Costello and Wittenberg-Moerman, 2011), the vocal cues of CEO speech (Hobson et al., 2012), the textual content of labelled topic in annual report filling

(Brown et al., 2020), etc., in this study the financial misreporting is cautiously subject to a retroactive financial result change due to fraud, misrepresentation or an SECs investigation.

It is remarkable that the occurrence of financial misstatements has dealt a huge blow to the equity markets, with 1.7% of market capitalisation, worth $744bn, evaporating due to financial fraud in 2020. Equity holders, rather than corporate managers who directly engage in financial misconduct, are deceived by managers in a principal-agent conflict on the one hand, and suffer the penalties when the misconduct is revealed on the other. The penalties include not only a direct cost imposed by the regulator in accordance with the relevant lawsuits, but also a corresponding reputation loss and market loss. Amiram et al. (2018) interpret that revealed financial misconduct will undermine the fundamental trust mechanism in trading, which is maintained by various enforcements, thereby producing a reputation loss and a market loss of the firm. The reputation loss after a financial misreporting disclosure is also explained by Armstrong et al. (2010), who posit that firms financing advantage through high quality information as well as trading relationships will disappear due to creditors uncertainty about the company's disclosures and a increasing interest rate can be regarded as compensation for agency costs. Besides, market value loss indicates share price recovery from inflated level caused by exaggerated financial result. Graham et al. (2008) further suggest an alternative explanation for market loss that depressed demand for firms securities can relatively restrict investment activities, leading to a decline in market value. Given the severe consequences, victimised shareholders would expect the initial share purchase price to include the cost of a potential misconduct, and would thus have a greater incentive to deter financial misstatement through more intensive monitoring and corporate governance practices.

Notwithstanding the equity market, there is relatively less research on the impact of financial misreporting in the debt market. Since debt financing acts as an essential source of funding for firms, the reflection of financial misreporting by lending banks is of great significance for the continued development of the firms. This paper reviews the relevant previous literature from the perspective of financial misreporting as a reverse reference variable for the quality of accounting information. The potential impact of financial misstatements can be better examined by understanding the mechanisms of influence of financial information in debt contracts.

Financial statement, as a primary information resource, plays an important role when firm is seeking for external capital, including equity and debt. However, as opposed to equity holders, whose downside risk is also limited by the total amount of the investment, debt holders have a pre-determined return and are therefore more interested in whether the company provides reliable information to secure the risk of debt default and assess the quality of the firm's collateral. In addition, reported earnings are considered to make a very limited information contribution to market participants (Ball and Shivakumar, 2008), in contrast, Shivakumar (2013) documents that earning announcements can play an primary role in the debt or compensation contracting process, arguing that the main role of accounting information is reflected in contracting process.

In a loan contract, accounting information not only provides firms information for pre- contract credit assessment, reducing information discrepancies for creditors, but will also be used as a direct parameter determinant of the terms of the contract at the loan inception, including interest rate, covenants, maturity, convertibility, repayment plan, collateral, etc. Armstrong et al. (2010) explicate that if a firm provides opaque and unreliable financial information, hampering information users forecasting future cash flows or timely controlling riskiness, then it is difficult for that firm to cost effectively raise debt financing. Wittenberg- Moerman (2008) suggests that in the private debt deal, information asymmetry between creditors, which is measured by bid-ask spread, can be mitigated by high quality of financial reporting, thereby invigorating debt market transactions.

In addition, prior literature has provided evidence that largely and consistently supports the main conclusion that high quality accounting information is effective in attenuating information asymmetry and reducing the cost of debt, whereas some inconsistent findings reflect the difficulty of standardising the measurement of the quality of accounting information. Therefore, research on financial misstatement may provide more accurate results from the reverse direction.

Graham et al. (2008) were the first to investigate the banks reaction to financial misreporting, providing evidence that after a restatement, the cost of debt will significantly increase with higher loan spread, tighter non-price contract terms, additional covenants and also the lender structure will be more concentrated. These evidences are consistent with Costello and Wittenberg-Moerman (2011), who further examine the impact of restatement following internal control weakness and conjecture that stricter monitoring can be activated by more non-financial covenants and shorter loan maturity. They attribute financial misstatements to managerial culpability, so changes to contractual terms are designed to constrain management activity and force more frequently refinancing.

In this paper, financial misreporting, an event variable that has been frequent and influential in recent years, on the structure of syndicated loan. The presence of financial misstatements reflects, on the one hand, the unreliable accounting systems of borrowers, which exacerbate the opacity of participant lenders, and, on the other hand, the weakening of the credibility of borrowers and market regulatory mechanisms, which may ultimately results in fewer participants in syndication and a larger share of claims held by the lead bank.

2.2 Syndicate loans and syndication structure

The importance of syndicated loan as a corporate financing source is emphasized by previous studies. report that the syndicated loan, the largest source of corporate finance in the US, reached $1.04 trillion in issuance in 2006 and $1.8 trillion in 2011(Ivashina, 2009; Ivashina and Scharfstein, 2010). The Multi-funded syndicated loan not only allows borrowers to raise more funds on a larger scale, but also provides syndicate members an opportunity to participate with diversified credit risk. In addition, banks capital constraints and lending limits have contributed to the popularity of syndicated debt (Simon, 1993; Dennis and Mullineaux, 2000).

A syndicated loan contract is entered into by a single borrower with multiple lenders, and these lenders include lead banks as arrangers and non-lead banks as participants. An arranger of a syndication is responsible to originally negotiate and finalise the single credit agreement with a borrower, and also to sell part of the loan to other participants. Dennis and Mullineaux (2000) argue that in a lending syndication, in addition to agency conflicts in the bilateral relationship between borrower and lender, the relationship within the syndicates also leads to an information asymmetry issue. Sufi (2007) explains that the lead bank is internally well- informed and can access to more preliminary information of borrower, while the participant banks at an information disadvantage rely on the lead bank to collate and provide information for risk assessment ex ante, and to monitor and extant due diligence on borrowers ex post to conduct ongoing debt performance assessment and control the default risk (Sufi, 2007). In addition, Ivashina (2009) empirically examine that the within syndicates information asymmetry can lead to an information premium in loan spread of approximately 4% of the total cost of debt.

Furthermore, Chaudhry and Kleimeier (2015) argue that the problem of information asymmetry in loan syndication can further generate adverse selection and moral hazard, both of which undermine the interests of participant banks and ultimately increase the borrowers cost of debt. Moreover, adverse selection is caused by lead banks providing exaggerated credit information on borrowers in order to syndicate riskier loans, while moral hazard is prompted by arrangers decreasing incentive to conduct post-syndication monitoring of the loan performance on behalf of the syndicate, which is costly, unobservable and accordingly less rewarding for the arranger. As mentioned previously, the participant banks are in a vulnerable position in a situation of information asymmetry, whilst they are expected to take rational action to limit their exposure to risk. In consideration of the anticipant of lead lenders exploiting their information priority and ex post shirking from monitoring duty, participant lenders could rationally require the lead banks to retain a larger proportion of the debt or would not even participate in the first place. (Kim and Song, 2011). With respect to the optimal syndication, contacting parties need to consider that potential syndicate members rational anticipate adverse selection and moral hazard issues.

The two major issues arising from information asymmetry in syndicated loan have been examined in many previous studies in terms of their impact on syndicate size of lenders and lead lender retention of debt deal. Champagne and Coggins (2012) conclude that the way in which lead arranger manage the syndication structure reflects the level of information asymmetry perceived by various stakeholders. Sufi (2007) provides evidences that in the presence of moral hazard, arrangers are more likely to organise a concentrated syndicate structure and shift smaller share of loan to other lenders. With respect to rising possibility and uncertainty of financial distress in the face of asymmetric information, Graham et al. (2008) further illustrate that a syndicated structure containing less lenders can facilitate debt restructuring in financial distress through lower cost renegotiation and efficient collective decision-making, and act as an effective monitoring mechanism.

Furthermore, the role of lead arrangers strong reputation in interacting and combining to affect the way the syndicated loan is structured has been clarified. Since lead lenders are repeated parties in syndicate market, and may highly value the stable business relationship with other potential contracting parties, their reputation served as a discipline mechanism can attenuate asymmetric information problems. Therefore, the reputable lead banks would rather intend to retain a larger proportion of the loans in their own portfolios and then maintain a less diffuse loan structure in absence of information asymmetry (Simons, 1993; Sufi 2007; Chaudhry and Kleimeier, 2015). In addition, Dennis and Mullineaux (2000) support that the established reputation of arranger can improve the saleability of new loan deals. Notwithstanding the lead banks reputation, information asymmetry factors relating to third- party regulation, borrowers characteristics, and contractual features can also influence the syndicate organisation by weakening the principle-agent problem. Esty and Megginson (2003) show the unreliable legal enforcement would promote a loose syndicate structure, as the damped third-party monitoring may not avert borrowers financial evasion and protect participants claims, and the dispersed loan structure is applied against the strategic breach of repayment. Kim et al. (2011) note the syndicate loans issued by the firm with internal control weakness would intuitively engage fewer creditors. Dennis and Mullineaux (2000) document that when the firm is publicly listed or has a credit rating, indicating a high level of information transparency, more lenders will fund the firms new loan deal. Despite this, they also find the positive relationship between the syndicate member size and the loans maturity.

It is notable that these factors produce the mechanism of direct or indirect influence are reciprocal and combined (Champagne and Coggins, 2011). This finding is also supported by Ball et al. (2008), they report that the impact of the quality of accounting information on the lead lenders retained loan share could be conditional by the arrangers reputation, the existence of repeated syndicate contracts between arranger and borrower as well as the borrowers credit rating.

It can be concluded from the previous studies mentioned above that many different types of factors identified through empirical research can significantly influence the syndicating process of new debt deal, and it is also worth noting that these influencing mechanisms can be explained to some extent by the efficiency of the ex post monitoring mechanisms arising from different syndication structures. Building on the literature review, this study examine whether and how the lead arrangers retained fraction of syndicated loan and the participating lenders size are affected the presence of borrowers financial misreporting.

3. Hypothesis Development

Previous research has demonstrated a boom in the syndicated debt market, which can facilitate larger scale financing for borrowers and opportunities for smaller banks to diversify their portfolio through a unique syndicated funding model. However, this form of syndication raises information asymmetry issue between syndicate members, where the motivation for lead banks to potentially exploit their information superiority over other participant banks can further lead to pre-contract adverse selection and post-contract moral hazard. Kim and Song (2011) assume that when there is a significant information asymmetry problem, existing and potential creditors will be rationally risk averse and protect their own interests and may therefore choose not to join the syndicate in the beginning or demand a larger claim committed by the lead arranger. Alternatively, Simon (1993) argues that from the perspective of preserving credibility in long- term transactions, lead lenders will actively hold a larger share of information-problematic syndicated debt to avoid damaging future trading opportunities. It can be seen that the syndication structure, i.e. the number of participants and lead lenders retained share, is indicative of the extent of information asymmetry within syndicate.

As the review of past literature in section 2.1 shows, financial information has also been the subject of extensive research, and many empirical studies have demonstrated that high quality accounting information can reduce the cost of debt. Accounting information imposes considerable impact on the loan market through two main channels. The first is that accounting numbers play a direct role in the contracting process, where the output of accounting system can be an input to the debt contract, and these accounting-based terms can timely and accurately capture the debt performance, reducing the cost of renegotiation. The other is that the high quality accounting information can ameliorate the problem of information asymmetry, reduce the participants' reliance on the arrangers and diminish the possibility of adverse selection and moral hazard. Moreover, Ball et al. (2008) employ the debt-contracting value to reveal the timeliness and accuracy of firms accounting systems, and then document that the higher debt-contracting value possessed by the borrowers financial information can ameliorate information asymmetry problem within syndicate lenders, and shape the explicit loan ownership structure.

The motivations for financial misreporting have been comprehensively explained through a literature review, from the perspective of managers, insider traders and other stakeholders. However the frequency of corporate misconduct, especially financial fraud, has a deleterious effect on the market and participants. Dyck et al. (2021) report $72bn of market value loss due to corporate fraud in 2020, representing 1.7% of total market capitalisation. Losses to firms and share investors from financial misreporting have been a popular topic of research in the past, however, less research has explored the impact of misreporting on debt markets.

These empirical evidences are consistent with the previous research findings that financial misreporting change credit risk and aggravate information asymmetry problem.

Existing research has argued that financial restatement substantially undermine trust in the financial report itself (Graham et al., 2008), exacerbate information asymmetries and raise questions about a company's management (Costello and Wittenberg-Moerman, 2011). It could be inferred that the two main influence channels of the high-quality accounting information in debt contracting process discussed above would be inherently cancelled out. The risk perceived by existing and potential lenders could be materially increased due to the destruction of trust. For participating lenders in syndicated loan in particular, they may be more cautious and reluctant in credit decision due to the lack of other reliable information sources to further assess debt issued by firms with financial misreporting. The participating banks' lending choices are then directly reflected in the syndication structure, leading to the first hypothesis of this paper's empirical analysis.

Hypothesis 1: Borrowers with financial misreporting revealed can attract significantly fewer participant lenders for their syndicated loan compared with borrowers without financial misreporting revealed.

As mentioned in the previous section, financial misreporting can be used as a counterpart sample where harmonising the measurement of the quality of accounting information is difficult. It has been also known that high quality accounting information can ultimately manifest itself as an effective reduction in the cost of debt by reducing asymmetric information problems and generating input data for contractual terms. The presence of financial restatement directly negates the quality of accounting information in the first place, which may specifically be reflected in more stringent terms, shorter loan maturity, etc. The efficient monitoring mechanisms that were built on high quality accounting information fail, thus placing an overall higher demand on due diligence responsibilities of the lead arranger as a whole. In this case, lead arranger may choose to retain a larger share in order to endorse the intensity of monitoring. The second hypothesis of the empirical analysis is accordingly formulated.

Hypothesis 2: Proportion of claims held by lead lender in syndicated debt issued by borrowers with financial misreporting revealed is larger compared than by borrowers without financial misreporting revealed.

4. Empirical Model

In order to examine the impact of financial misreporting on syndicated loan structure, the following multivariate regression model is accepted:

Loanstructureijt = a + b1 Misreportijt + b2 lnmaturityijt + b3 lnspreadijt + b4 securedijt+b5 fcovijt + b6 ratedijt + b7 lnsizeijt + b8 leverageijt + b9 roaijt + m j + lt + eijt

(1)

where i indicates company, j indicates industry, t represents year; Loanstructure is the dependent variable, including participant lender and lead lender. Misreport is the independent variable. To segregate the effect of Misreport from other factors that have a potential influence on the syndicated loan structure, different control variables are introduced into the regression model(1), namely lnmaturity, lnspread, secured, fcov, rated, lnsize, leverage, roa. mj stands for the industry fixed effects; lt is the year fixed effects; eijt is the error term. To alleviate possible heteroscedasticity and autocorrelation, standard errors are clustered at the firm level.

The estimated coefficient b1 captures the influence of financial misreporting on syndicated loan structure. b1 is expected to be significantly negative when the dependent variable is the number of participants in the loan syndicate, that is, borrowers with a financial misreporting revealed will attract a smaller number of participant lenders in syndicated loans compared to borrowers without a financial misreporting revealed. b1 is significantly positive when the dependent variable is the percentage of a syndicated loan retained by the lead lender(s), that is, the proportion of a syndicated loan held by the lead arranger is larger for borrowers with a financial misreporting revealed compared to borrowers without a financial misreporting revealed. In order to examine the soundness of the specification, a series of robustness tests is conducted, including alternative measurement of firm size, including more control variables, exclusion the borrowers who belongs to the finance industry etc.

4.1 Variables Descriptions

• The explained variable of this manuscript is syndicated loan structure (Loanstructure). participant and lead are used to jointly measure Loanstructure, where participant is the number of participant lenders in the loan syndicate and lead is the proportion of syndicated loan contract that is retained by the lead lenders. If there is more than one lead bank in a loan deal, it is the average percentage retained.

• The core explanatory variable in this paper is financial misreporting (Misreport). Misreport is measured by whether the firms financial results have been restated. Misreport is defined as a dummy variable. A dummy variable that equals one if any of the firms financial results were subsequently restated (due to fraud, misrepresentation or an investigation by the SEC), and zero otherwise.

• In this study, eight control variables have been adopted. Firstly, the natural log of maturity is applied to measure the maturity period of each syndicated loan from the date of signing to the expiration in months (lnmaturity). Secondly, this study utilises the logarithm of the all-in drawn, which measures the loan pricing in basis points (lnspread). The spread is equal to 1% if the loan pricing in basis points is 100 basis Thirdly, the control variable secured is an indicator variable that equates to 1 if the syndicated loan is backed by collateral, and 0 if otherwise. Fourthly, gcov represents the number of general covenants included in the loan agreement. General covenants include the percentage of net proceeds from an asset sale that must be used to pay down an outstanding loan balance, whether a borrower is allowed to pay dividends, and the percentage of excess cash flow a borrower is allowed to use towards dividends. Fifthly, rated is an indicator variable that is equal to 1 if the borrower owns a Standard & Poors (S&P) Domestic Long-Term Issuer Credit Rating, and 0 if otherwise. Sixth, the logarithm of the total assets of the borrowers is used to measure firms scale (lnsize). While leverage is the ratio of total debts to total assets. Finally, the ratio of income before extraordinary items to the total assets is adopted to measure firms return on assets (roa).

4.2 Sample and Descriptive Statistics

The sample is constructed using financial data collected from Compustat and loan data collected from Loan Pricing Corporations DealScan database for the period from 1996 to 2017, comprising of 6,399 unique listed companies in the USA with syndicated loan information and 18,837 loan deals. All observations with missing data required for the empirical model are dropped. Moreover, for eliminating the influence of extreme values, the analysis is performed at deal level instead of facility level and all continuous variables are winsorised at the 1^st and 99^th percentiles.

4.3 Descriptive Statistics

Table 1 reports the descriptive statistics including mean, standard deviation, maximum, minimum and the percentiles 25, 50 and 75 for the explained variable, core explanatory variable and control variables.

Firstly, it is important to note that that 12.1% of the firms which commit a financial misreporting prior to the syndicated loan issuance. Conversely, the mean of 0.121 indicates that 87.9% of observations with absence of financial restatement. The relatively small fraction of observations with financial misreporting might weaken the application of the statistical result in a broader context. The estimation by Dyck et al. (2021) of the average probability of financial fraud predictions for large listed companies is 11%, so 12% of observations with financial misreporting in this analysis can reasonably be considered to be close to market reality.

Table 1 Descriptive Statistics of Variables

Secondly, the mean and median number of participant lenders in the loan syndicate are approximately 8 and 6 respectively, and the number of participant lenders in sample set varies from 1 to 290. Considering the percentile 75 of 12, the maximum number of syndicate participants of 240 may indicate the existence of a few extremely large syndicate structures. The proportion of the syndicated loan retained by lead banks has a mean of 12.91 percent, and the median is infinitely close to zero, implying that more than half of the observations are below the average in terms of the lead banks share retention.

Finally, regarding control variables, the mean and median maturity of the loans in the sample data set is around 46 months and 48 months. The mean (median) of all-in drawn loan spread are about 204 basis points (174 basis points), whereas the impractical large standard deviation of the loan spread can indicate a skewed and dispersed distribution and a broad difference in the minimum and maximum values. On average, approximately 60.2 percent of the syndicated loans are secured. Regarding the number of financial covenants within this data set, the mean number of financial covenants included in loan documentation (where applicable) is 2.469. This suggests that, on average, lenders issue loans containing approximately 3 financial covenants imposed in each syndicated loan contract. Only 33.3 percent of the borrowers in the sample has been rated by S&P. The total assets, which has a mean and median of $6,282.947 million and $953.347 million. It is notable that the total assets variable is expected to possess an excessively dispersed and skewed distribution, based on its large standard deviation, which also makes it stand out among the variables in Table 1. The mean (median) of leverage are 32.5 percent (30.2 percent). The mean (median) of roa is -0.2 percent (3.2 percent).

To explore the differences between the variables of the borrowing firms with the presence of financial misreporting and without it, the sample is divided into two sub-samples, borrowers without financial misreporting revealed (n=18,868) and borrowers with financial misreporting (n = 2,606). Table 2 presents the mean of each sub-sample and the differences for each variable.

Table 2 Differences in Mean

Note: ***, **, and * represent the significance level at 1%, 5%, 10%, respectively. The standard errors are reported in parentheses and clustered at company level.

As illustrated in Table 2, the financial restatement leads to a difference of 0.277 in terms of the mean number of participants, which is not significant at 10 percent level. This implies that there is no statistically significant difference of the mean number of participants for borrowers with and without financial misreporting. However, it remains to be tested, whether the financial misreporting may affect the number of participant banks in the syndicated loan. The difference of the mean proportion of the syndicated loan retained by a lead arranger for between borrowers with and without financial misreporting is 2.679 percent. It is statistically significant at 1 percent level. This suggests that lead arranger owns a larger ratio of the syndicated loan for borrowers with correctly financial reporting compared to that with financial misreporting. Nonetheless, it is opposite with the second hypotheses which expect that lead arranger will retain a larger proportion of the syndicated loan for borrowers with financial misreporting revealed. As discussed in the literature review, on the one hand, the lead bank's claim-holding motive will be weighed against the benefits of risk diversification, credit risk screening and costly monitoring responsibilities (e.g. Panyagometh and Roberts, 2010; Ivashina, 2009), and thus arranger share retention may be affected in diverse directions by various factors, including contractual characteristics, borrower characteristics, lead arranger characteristics, etc. The effects of the factors on the contract processing may offset each other to some extent in facilitating monitoring as well as resolving principal agency conflicts, which may lead to a situation where financial restatement is revealed and the lead banks holding decision is affected by an overall impact and reflect a different result from that assumed in the previous section. On the other hand, the small sub-sample size of borrowers with financial restatement in this paper may cause statistical distortion, thereby leading to this disparity. However, it is still worth examining the incremental impact that financial restatement can have on the lead bank share retention after controlling for other relevant variables.

Overall, the differences of firms characteristics (such as asset, leverage and roa) are not significant at 10 percent level, whilst the differences of most loan contractual characteristics (such as maturity, spread and rated) are significant at 5 percent level between borrowers with and without financial misreporting. Compared to borrowers without financial misreporting, the other sub-sample shows a higher syndicated loan contractual characteristics level, which could generally contribute a more unfavourable loan contract, and this implies the financial misreporting may raise the cost of debt financing for borrowers.

4.4 Correlation Matrix

Before performing the regression, correlation analysis is a necessary step to identify the relationship between all variables used in the estimation equation (1). As shown in Table 3, from the correlation between the explained variable and the explanatory variable in this paper, Misreporting is negatively correlated with participant and lead with a magnitude of -0.010 and -0.036, and both coefficients may suggest a weak relationship between the dependent variables and the independent variable. Furthermore, the negative relationship of Misreporting and lead is significant at 1 percent level, whilst the relationship of Misreporting and participant is unsignificant at 10 percent level. This indicates that the participation decisions of potential non- lead banks are not significantly influenced by the presence or absence of financial restatements by the borrowers issuing syndicated loan, whereas lead banks would hold considerably fewer loan share if the borrowers financial statement were restated. It is evident that this result is inconsistent with the hypothesis presented in this paper, but concurred with the implications of the descriptive statistical results in the previous section.

As discussed in the hypothesis development section, the expectation that the syndicated debt structure differs between the two scenarios of whether or not the borrower has a financial restatement is mainly based on the theory of information asymmetry within syndicates and the inference of the combined effect of various contracting party-related factors on the monitoring mechanism. When considering the impact of external market conditions, it can be noted that the sample observations are all from the US market, where rigorous legal enforcement provides additional strong protection for lenders' claims. In this case, lenders may have a further incentive to obtain lower contract negotiation costs and more intensive monitoring by reducing the size of the syndicate, and this conjecture can be supported by the research of Esty and Megginson (2003) mentioned in 2.1 section.

In order to provide an alternative explanation of the disparity between the hypothesis to be tested in the paper and the results of the correlation analysis, the characteristics and subjective motivations of the lead bank are further considered. A review of the past literature suggests that adverse selection due to information asymmetry between syndicates is specifically due to the fact that the lead bank, as the organiser of the syndicate, can attempt to share the risk of problem debt to a greater extent by taking its information priority. It can then be speculated instead that in the absence of financial restatements, i.e. in the presence of relatively low credit risk, the lead bank is more likely to retain a larger share of the loan before seeking other potential participate banks, thereby securing a larger predetermined rate return at low risk. In addition, since a larger share of lead banks also means that less of the share is allocated to other non-lead lenders, it may intuitively lead to a reduction in the number of non-lead banks that the syndicate can reasonably accommodate. This may provide a partial explanation for the weak correlation between the two variables of number of participants and financial misreporting.

The results in Table 3 also shows that maturity and rated is significantly positively related to participant while it is significantly negatively related to lead. For rest of variables (spread, secured, fcov) used to capture loan contractual characteristics, they are significantly negatively related to participant, and positively related to lead. Moreover, as for borrowers characteristic variables (asset, leverage, roa), there are significantly positive relationships between borrowers characteristic variables and participant, while there are significantly negative relationships between borrowers characteristic variables and lead.

In general, it is worth pointing out that all correlation coefficients indicate a weak relationship between the variables. In addition, a problem of multi-collinearity exists between the dependent variables only when the correlation coefficients exceed the order of 0.7. However, the correlation matrix shows that there is no such problem.

Table 3 Pearson Correlation Matrix

5. Empirical Results

5.1 Baseline results

According to the empirical method, the fixed effect model is employed to estimate the effect of financial misreporting (Misreporting) on the number of participants in one single loan syndicate (participant) reported in column (1) of Table 4, and the percentage of a syndicated loan retained by the lead banks (lead) shown in column (2) of Table 4. The statistical significance of two-tailed test at 10 percent, 5 percent and 1 percent level is indicated by *, ** and ***, respectively.

6. Conclusion

By examining 18,837 syndicated loan transactions proceeded by 6,399 US-listed firms during the period of 1996 to 2017, this paper explores the impact of corporate financial misreporting on the structure of subsequent syndicated loan. The research model combines pertinent control variables to segregate the impact of loan and firm characteristics on the syndication structure, and provides conclusive evidence that the presence of financial misreporting of borrowers leads to a more concentrated lender structure in syndicated loan. Specifically, the empirical results of this paper imply that firms with financial misreporting are more likely to recruit a smaller number of participate lenders and the lead bank that initiates and organises the syndicate will hold a larger proportion of the debt than firms without financial misreporting. It can be noted that the empirical results support the research hypotheses presented through the literature review. Therefore, it can be argued that the research in this paper provides additional evidence on the influence mechanisms such as financial misreporting increasing credit risk (Amiram et al., 2018) and exacerbating information asymmetry between syndicates (Graham et al., 2008).

In this paper, when studying banks' responses to financial misreporting, the main consideration is the offsetting effect of financial misreporting as counterparts to high-quality financial statement on the intrinsic impact channel of accounting information, as reliable financial information has been shown in much of the literature to have a significant impact on syndicated debt (Ball et al., 2008; Armstrong et al., 2010). The paper further investigates the significant effects of two factors, previous lead bank-borrower trading relationships and external audit quality, as substitutes for accounting information on the syndication structure through additional regression analysis. Robust test take into account the impact of these additional factors and the results, which are consistent with the baseline regression results, indicate a substantial relationship between financial misreporting and syndicated loan structure, further supporting the research hypothesis of this paper.

Overall, the paper obtains the expected findings, but there could be further enhancement in the future on this research question. Constrained by the limited number of observations, the interpretation of the syndicated debt structure focuses on the impact on the parties involved in the debt contract, while the impact of the macro market environment could be more fully considered. Furthermore, it is evident from the empirical analysis that the lending behaviour of the lead bank is the result of a complex trade-off, and thus how the lead bank organises a syndicate can be influenced by various factors in different directions. In addition to the reaction to financial misreporting, future research could provide a deeper understanding of how debt structure reflects the effects of financial misreporting in aggregate.