國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
5. The Announcement Effects
In the previous chapter, we examine the impact of corporate conviction and some specific criminal conditions (fines, compliance program, the country where convicted firm is incorporated, the severity and types of violation) on the cost of debt. However, with both year and industry effects controlled, most of our results show that there’s no significant evidence to reject our null hypothesis. Among the results including interaction terms shown in Table 4, only when the severity and the types of crime is taken into account does the conviction effect appear to be significant. Even so, the coefficient indicates that the conviction effect is inversely associated with the cost of loan. These findings are quite on the contrary to both our expectation and the related studies on the effect of corporate violation.
Among related studies of criminal effects, Karpoff et al.(2014) identify late initial
revelation dates as one of the challenges frequently faced when conducting research on
financial misconduct. That is, the initial public revelations of financial misconduct may have occurred months before the initial coverage in the database. Even though Garrett’s database is not discussed in Karpoff et al.(2014), by cross-referencing news on LexisNexis, we discovered that the convictions dates recognized in Garrett’s database, like the databases pointed out in Karpoff et al. (2014), also face the problem of late
initial revelation dates. In our sample, some related news covering the corporate
convictions has been announced months or years before the agreement date of the conviction logged in US Department of Justice.
Taking this condition into consideration, there’s possibility that the lending banks have adjusted their evaluation toward convicted firms after related news releases (instead of after the agreement date of the conviction). This may help us explain why the results based on agreement dates would derail from relevant research. Hence, in this
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
section, we intend to catch a glimpse of the relationship between the announcement effect of corporate conviction and the cost of loan based on sketchy adjustments on the turning point recognized.
In the previous chapter, we base our turning point in years. That is, we distinguish facilities initiated before and after convictions based on the year difference between facility start year and agreement year. Therefore, the lag in the recognition of the release of corporate conviction make little difference to our previous results if it’s within 1 year.
In our sample, there’s around 1/3 of the agreement date of convictions that are set one year later than the release of its news. To gain a glimpse of the potential difference in results comparing previous chapter, we shift each agreement year one year backward as the proxy of announcement year of the news. For example, if the agreement year of the conviction is 2001, here we assume year 2000 is the year when the news of prosecution releases, and so on. Based on this sketchy adjustment of turning point recognized, the following shows the results of our empirical model.
5.1 Effect of Corporate Conviction on the Cost of Bank Debt
Tables 5 and 6 presents the summary statistics results of loan contract terms for
convicted firms. Number of observations, mean and standard deviation of debt contract terms are reported for loans in the full sample, facilities started before and facilities started after the proxy year for announcement. Table 6 presents the means of the differences between the variables representing bank contract terms before and after proxy year for criminal news. The results show that after conviction news releases, loan spread, loan maturity and the number of covenants increase at significance level of 1%;On the other hand, the appliance of performance pricing and the number of lenders decrease at significance level of 1%. The results show that the coefficients of loan size, number of security and lead bank share of ex post facilities increase, but not significantly different from that of ex ante facilities.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
27
The regression results are reported in Table 7. Column 1 analyze the cost of debt with post-news dummy as the only independent variable. The estimated coefficient for post-conviction news is 13.09, with t-value of 1.33. Column 2 and column 3 shows results with firm characteristics and additional loan characteristics taken into concerns.
The coefficients of Ln(asset), Market-to-Book Ratio, Profitability, Loan Size and
Performance Pricing are negative, while Leverage Ratio and Loan Maturity are
positively associated with the spread of facility, which, on the whole, is consistent with previous studies in loan contracting. From column 2 and column 3 of Table 7 we can see that the coefficient of Post-Conviction News is positive under both models, significant at 1% and 5% level, respectively. This results show that when a company is involved in the news of conviction or prosecution, they tend to face a higher price of bank debt than before.
5.2 Specific Conviction Terms on the Cost of Bank Debt
Similar to chapter 4.2, in this section, we further look into some specific terms obtained by or condition of the convicted companies in order to see if there are significant differences in the results where the change of loan spread can be explained by particular conviction characteristics.
5.2.1 The Impact of Legal Penalty－Fined or not
Column 1 of Table 8 presents the results of our main empirical model with an additional interaction term PCN*Fined included in the regression. The result, where the coefficient of PCN*Fined is 20.67 at significance level of 10%, shows that when convicted, the companies sanctioned with fine tend to face higher spread in loan than those not-fined. While the coefficient of post-conviction news is still positive but not statistically significant, the addition of this variable slightly increases the adjusted R square of the model, from 72.08% to 72.15%.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
28
5.2.2 Compliance Program
Column 2 of Table 8 shows the results of our model with an interaction term PCN*withCompliance included in the regression. The coefficient of the interaction
term is not insignificant enough to reject our null hypothesis that there’s no difference between the costs of loan faced by companies adopting compliance program at the time of conviction and those don’t. While the conviction news is still positively related to loan spread (here, at significance level of 5%), the adjusted R square slightly decreases by 0.03%.
5.2.3 State of Registry
In this section, we examine whether there’s difference in the conviction effect between domestic-incorporated and foreign-incorporated criminal firms. Same as section 4.2.3., we set a dummy variable FCC (which is equal to one if the convicted firm is incorporated outside of US, and zero otherwise), and focus on its interaction term with conviction effect, PCN*FCC. The results are shown in column 3 of Table 8.
While the coefficient of Post-Conviction News is 25.43, significant at 1% level, the coefficient of interaction term PC*FCC is -80.81, at the significance level of 5%. The result indicates that, compared to those incorporated in the US, firms incorporated outside of US tend to face lower cost of spread after the conviction.
5.2.4 Conviction Type and the cost of Bank Debt
Same as section 4.2.4., we divide our sample into two groups based on (1).the severity of crime and (2).whether the violation is fraud-related. We assign 2 dummy variables indicating whether it is Type1 conviction and whether it is fraud-related.
Column 4 and column 5 of Table 8 presents the results with these two additional
interaction term included in the model. Under both models, the coefficients ofPost-Conviction News are positive. In column 4, Post-Post-Conviction News is at significance
level of 5%, however, its interaction with CrimeSeverity is not significant. Column 5
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
29
shows the result of regression model including the interaction term PCN*FRC. At significance level of 5%, the coefficient of PCN*FRC is 24.29, positively associated with the cost of bank debt. This indicates that firms involved in fraud-related convictions, compared to other types of violation, tend to face higher spread of loan after conviction news releases.
5.3 Suggestions on Further Studies
The results based on adjustment of turning point (shifting one year back as proxy for release time of conviction news) are more consistent with our expectations. Not only does it indicates that conviction effect exists in loan market and is reflected in the spread of bank debt, but it also implies that the news of corporate conviction is more influential, not only in equity market but in loan market as well, than the actual time of court conviction. However, since this results are based on sketchy adjustments on the turning point recognized, further accurate information from LexisNexis and Factiva database is required so as to update and confirm each of the release time of related news.
Even though the sorting process for precise time of related news has yet to be completed, we regress loan spreads on variables mentioned earlier using 56 criminal companies preliminarily sorted with precise news release year along with 890 facilities. The results are presented in table 9 and 10. Although the effect of Post-conviction news are similar to the results presented in table 8, the effects are less significant with accurate criminal news. Nonetheless, the results shows that the coefficient of PCN*FRC is positively significant, implying that the type of crimes (fraud-related crime) lays certain degree of impact on loan spreads between criminal corporations.
In this paper, we differentiate ex ante and post facilities based on facility starting years and conviction (or conviction-related news released years) instead of starting dates (news-released dates). And we consider the first conviction only, if there’re more
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
30
than one. Moreover, we focus on the conviction effect on loan spread. These may explain why some of the results derail from our expectations. If the criminal companies are convicted of more serious crimes after the first one, by excluding the effects of subsequent convictions, the results based on the first conviction may be misinterpreted.
Apart from that, there’re possibilities that after the convictions, the convicted companies are not able to, or not welling to, obtain syndicated loan. In this situation, the impact of conviction is more serious and decisive in loan market, but by focusing on the change of loan spread, this effect cannot be captured since the companies in question have no ex post facilities for comparison. For potential further studies, apart from resolving limitations mentioned above, by using linkage between the databases in this paper, we hope to expand research upon other loan contract terms to see if banks stretch their influence in contract terms other than loan spread. This may help us understand the cross effect of loan contracts when facing borrowers with conviction history. In the conviction and agreement database constructed by Professor Garrett, more details on the crime and convicted companies are listed for further studies.
Potentially, by combining the details with existing prestigious corporate database such as AAER, GAO, LexisNexis, Factiva etc., we expect to understand different aspects of conviction effect on loan markets in more angles.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31