From the above analysis, we have evidence of a positive short-term announcement effect on distressed loan announcements, and Billett, Flannery & Garfinkel (2006) document that these positive loan announcements are followed by significantly negative long-term performance.
However, Rosenfeld (2007) documents that distressed loans issued by prior lenders have a significantly better long-term performance than those issued by new lenders. To determine whether there is a consistent story of long-term distressed firm performance following a loan announcement, I perform the same analysis as Billett, Flannery & Garfinkel (2006) and Rosenfeld (2007) on the sub-samples from the short-term analysis.
39
To verify that Rosenfeld’s holdings persist on this subset, I repeat the bivariate probit analysis that predicts future firm success while attempting to control for the possible endogeneity of future firm performance dictating the nature o f the lending relationship. This approach requires instruments to identify the relationship, including bank market concentration, a lagged relationship indicator, and analyst coverage, as well as an interaction o f leverage and analyst coverage to control for debt’s influence on analyst interest. Table 12 shows estimation results using this methodology. Columns I-VI and VIII each have significant instruments and significant rhos, so for these specifications, bivariate probit analysis is appropriate and is properly identified.
Referring to these columns on the first page of the bivariate probit analysis, it is clear that there is a positive effect of lending relationships on future firm performance. This only holds true for sub-samples that are not entirely composed o f firms that are severely financially distressed.
When only considering the most severely distressed firms, the only specifications with significant instruments have insignificant rhos, which means that there is no evidence o f endogeneity. In this case, we revert to results from probit analysis, found in the Appendix, where there is an insignificant impact o f lending relationships on future firm performance. Thus, the results from Rosenfeld (2007) hold for the sub-samples analyzed in this study, so long as the sub-sample is not solely composed o f severely distressed firms.
Next, I turn to Billett, Flannery & Garfinkel’s (2006) analysis. Since their findings are consistent across methodologies, I perform only part of their analysis: I compute buy and hold abnormal returns by matching firms based on size, size and book to market and size and industry.
Their methodology also requires that all sample and matched firms trade on the same financial exchange. Matches are performed as o f the fiscal year end prior to the loan announcement, which in the case o f Billett, Flannery & Garfinkel, is DealScan’s Deal Active Date11.
11 Results are qualitatively similar when follow ing Billett, Flannery & Garfinkel’s (2006) m ethodology o f using the Deal A ctive Date in place o f the loan announcement date found in Factiva. These results are listed in the Appendix. M y research show s that approximately h alf o f the matched Compustat/CRSP and
40
When sample firms delist prior to the end of three years following the loan announcement, I calculate abnormal returns at the time of delisting, as in Billett, Flannery &
Garfinkel (2006). Table 13 provides information on how many firms in each sample delist for reason other than going private or merging prior to the end o f the three-year study period. No distressed firms identified prior to 1996 fail by delisting, at which time the highest proportion of observations went on to fail by delisting. Overall, approximately 8% of the sample composed of the broadest range o f distressed firms delists, while about 14% of the most severely distressed firms delist. When the matched firm delists prior to the end of the sample period, I replace the matched firm with the next closest match, as in Billett, Flannery & Garfinkel (2006). This replacement continues until an appropriate match is found. Billett, Flannery & Garfinkel match based on two-digit SICs, but to omit as few sample firms as possible, I match based on one-digit SICs.
Table 14 shows the results for buy and hold abnormal returns. While there is some evidence that distressed firms incur significantly negative returns over the three years following the distressed loan announcement, the majority of the findings suggest that for each sub-sample as a whole, there is no evidence o f significant performance in either direction. In particular, when matching firms by size and book to market, the median buy and hold abnormal return is significantly negative for the two extreme sub-samples, as indicated by Wicoxon’s signed rank test. However, the means in each of these sub-samples, as well as for each of the other matching methods and sub-samples, there is no statistical evidence of long-run underperformance.
When evaluating only the relationship-backed financially distressed firms, there is no evidence, regardless of matching technique, of significant long-run performance. In contrast, when matching by either size and book to market or size and industry, there is evidence of significantly negative long-run performance for non-relationship funded distressed firms. For
DealScan tranches are announced in Factiva. O f those, approximately 45% have Factiva announcements in a different month from the Deal A ctive Date.
41
these two matching methods, t-tests of differences in means provide evidence that there is significantly different performance depending on whether a relationship lender funded the firm’s distressed loan. For the two sub-samples composed o f more severely distressed firms when matching by size and industry, the evidence is statistically weak but economically strong, with a difference in mean abnormal returns exceeding 30%.
To summarize, there is evidence that following financially distressed loan announcements, Billett, Flannery & Garfinkel’s (2006) robust findings o f negative long-run buy and hold abnormal returns do not hold. Further, there is evidence that firms obtaining distressed loans from prior lenders perform better in the three years following the loan than firms that obtain funds from a new lender.
These findings that relationship-backed firms experience better long-term performance, on average, than non-relationship-backed firms are consistent with the idea that with enough firm knowledge, the lender can tailor appropriate loan covenants to guide a firm towards better performance. Nini, Smith, & Sufi (2007) provide evidence that restrictive covenants are likely to follow negative performance, credit downgrades, and financial covenant violations. They find
“that firms obtaining contracts with a new restriction experience subsequent increases in their market value and operating performance” (p.l). Further, Demiroglu & James (2007) provide evidence that “tighter covenants are associated with improvements in performance,” and that “one reason borrowers choose restrictive loan covenants is to credibly convey private information about their credit quality” (Abstract). Thus, it is appropriate that financially distressed firms will experience restrictive covenants and that the decision to accept such a loan informs the market of the firm’s quality. It also follows that such firms will experience increased market value. These phenomena are consistent with the findings in this paper that the most severely distressed firms experience the highest abnormal announcement returns, and these are precisely those firms that are most likely to face the restrictive covenants that guide a firm towards improved performance.
42