Analysis of Moody s Bottom Rung Firms

Transcription

1 Analyss of Moody s Bottom Rung Frms Stoyu I. Ivanov * San Jose State Unversty Howard Turetsky San Jose State Unversty Abstract: Moody s publshed for the frst tme on March 10, 2009 a lst of Bottom Rung non-fnancal companes. Consderng the hgh stakes for both Moody s and the potental canddates for ncluson n the lst we examne what effects does the ncluson n the Bottom Rung lst have on the ncluded company s performance. Surprsngly, none of the orgnal publc Bottom Rung frms have been lqudated. We also attempt to dentfy what crtera exactly s Moody s utlzng to select the companes of the Bottom Rung lst. We fnd that frms ncluded n the Bottom Rung lst pror to the ncluson date experence lower cumulatve returns compared to a matched sample of control frms. We also fnd that the Bottom Rung frms do not underperform the control group after the publcaton of the lst. Keywords: Moody s Bottom Rung Lst, Company Performance JEL Classfcaton: G30, M40 Introducton The credt ratng agences have had to learn how to please both the nvestng communty and the clents requrng a credt ratng at the same tme. The nvestng communty requres accurate credt ratng, partcularly n tmes when the credt qualty of a frm deterorates. At the same tme, there s the agency conflct where the frmsclents of the credt ratng frms, rush to get the hghest possble credt ratng. * Correspondng author. Inqures & Perspectves 1 Volume 5 Number

2 Recently, the credt ratng companes have been blamed for the fnancal crss. Most often the credt ratng companes are crtczed for beng too frendly wth ther clents. They have also been accused of not provdng tmely sgnals ndcatng whch companes mght be headed for trouble. As Lucchett and Ng (2007) make t very clear n ther Wall Street Journal artcle: The credt-ratng frms are used to beng the whppng boys when thngs go badly n the markets. They were crtczed for beng late to alert nvestors to problems at Enron Corp. and other companes where major accountng msdeeds took place. Yet they also sometmes get chastsed when they downgrade a company s credt. In response to ths crtcsm Moody s publshed for the frst tme on March 10, 2009 and made publcly avalable a lst of Bottom Rung non-fnancal companes. Naturally, companes whch pay for a credt ratng would not be happy to be put on a lst wth other companes that mght be n trouble. Obvously, companes that are already n trouble would experence even more severe captal markets condtons and ncreased montorng after Moody s ponts a fnger at them. Moody s has always been n the forefront of nnovaton and thus has staked ts reputatonal captal before. As Poon (2003) ponts out Moody s has been the frst of the credt ratng agences to provde unsolcted credt ratngs, whch led to a lawsut by the Jefferson County (Colorado) whch Moody s subsequently won. Consderng the hgh stakes for both Moody s and the potental canddates for ncluson n the lst we examne what effects does the ncluson n the Bottom Rung lst have on the ncluded company s performance. We also attempt to dentfy what crtera exactly s Moody s utlzng to select the companes of the Bottom Rung lst. Most ratng companes admt that they assess both busness and fnancal rsk wth busness rsk very often based on ntervews wth managers of the company and ndustry. Ths arbtrarness however s not present n the assessment of fnancal rsk whch s most often based on fnancal parameters. Corporate managers can use the fndngs of ths study to dentfy corporate areas whch they need to concentrate on so that ther companes do not end up on the Moody s Bottom Rung lst. To the best of our knowledge ths s the frst study to address these ssues. We fnd that frms ncluded n the Bottom Rung lst pror to the ncluson date experence lower cumulatve returns compared to a matched sample of control frms. We also fnd that the Bottom Rung frms do not underperform the control group after the publcaton of the lst. Contrary to what mght be expected, none of the orgnal publc Bottom Rung frms have been lqudated. The multvarate analyss whch we performed suggests that consstently the most mportant factors sgnalng that a frm s a potental canddate for ncluson n the Bottom Rung lst are the ratos of EBITA and Average Assets, rato of EBITA and Interest Expense, rato of Debt and Book Captalzaton, and Earnngs Before Interest and Tax. Methodology We start wth all non-regulated frms on COMPUSTAT by excludng frms wth SIC codes of for regulated utltes and for fnancal servces companes. We Inqures & Perspectves 2 Volume 5 Number

3 examne the cumulatve returns one, two, three, sx and nne months before and after March 10, 2010 between the Bottom Rung and matchng frms. We exclude 7 days before and after March 10, 2010 to allow for assmlaton of nformaton. The date March 10, 2010 s selected because the lst has been made avalable to the publc by Moody s on ths date. Moody s metrcs used n ther credt ratng analyss s a source of relevant factors for the selecton of a frm to be ncluded n the Bottom Rung lst. On page 3 of ther Bottom Rung lst Moody s provdes the followng descrpton of the methodology used to create the lst: Buldng the Bottom Rung Moody's apples strct ratng crtera to assemble the Bottom Rung lst. Companes ncluded on the lst meet one of the followng crtera: 1. A Probablty of Default (PD) ratng of Caa1 or lower. 2. A PD of B3 and a negatve outlook. 3. A PD of B3 wth ratng under revew for downgrade. The exact crtera for ncludng a frm n the lst therefore are vague. Thus, n addton to the metrcs used n Moody s credt ratng analyss we survey the probablty of corporate falure lterature to dentfy possble factors for Bottom Rung ncluson. The ncluson n the Bottom Rung lst s to a certan extent smlar to gong bankrupt. The common characterstc between a company beng ncluded n the Bottom Rung lst and gong bankrupt s the degree of rarty of both events. Therefore, we utlze some of the probablty of corporate falure methodology n assessng the probablty of a frm endng on Moody s Bottom Rung lst. The decson whch fnancal characterstcs are mportant to the selecton of the frm for ncluson on the lst s based on analyss relatve to a control group of frms. The control group s selected based on a matchng exercse. The matchng frms are our control sample. Based on the comparatve analyss we dentfy whether the characterstcs are unque for the Bottom Rung frms or not. We dentfy matchng frms whch have the same two dgts SIC code, plus or mnus 75% of the Bottom Rung frm s total assets n 2007 and 2008 and the same sgn for net ncome. Based on these crtera the matchng frms are as good of canddates for ncluson n the Bottom Rung lst but are not added to the lst by Moody s. Logstc regresson analyss s employed to dentfy the factors whch mght get a company on the lst f not properly montored. The logstc analyss model s as follows: BRF 8, R8 R1 9, 1, R9 R2 10, 2 R10 R3 3, 11, NI R4 4, 12, AT R5 5, 13, EBIT R6 6, R7 7,, (1) where BRF s a dummy varable wth one assgned to the Bottom Rung frms and zero to the matchng frms. We use ten ratos from Moody s Fnancal Metrcs TM as possble factors. The ratos are R1 earnngs before nterest, taxes and amortzaton to average assets rato, R2 - earnngs before nterest, taxes and amortzaton to nterest expense rato, R3 earnngs before nterest, taxes and amortzaton margn, R4 funds from operaton and nterest expense to nterest expense rato, R5 funds from operaton to Inqures & Perspectves 3 Volume 5 Number

4 debt rato, R6 retaned cash flow to net debt rato, R7 debt to earnngs before nterest, taxes, deprecaton and amortzaton rato, R8 debt to book captalzaton rato, R9 operatng margn rato and R10 captal expendture to deprecaton expense rato. Detaled descrpton of the ratos s provded n the Appendx of the paper. Also, to control for sze and proftablty, NI, AT and EBIT are employed as control varables n the logstc analyss: NI s net ncome, AT s log of total assets, EBIT s Earnngs Before Interest and Tax. Logstc regresson analyss has been used before n bankruptcy predcton. Ohlson (1980), Glbert, Menon and Schwartz (1990), Latnen and Teja Latnen (2000) and Barnv, Agarwal and Leach (2002) are just a few examples of studes employng ths model. The common element among these studes s that they dentfy possble factors whch mght be contrbutng to bankruptcy and use the logstc regresson modelng technque to dentfy the statstcally sgnfcant factors. The appealng characterstc of the logstc model s that the dentfcaton of statstcally sgnfcant factors can be drectly nterpreted as affectng the probablty of the frm gong bankrupt. Thus, the logstc regresson coeffcents, for the purposes of ths study, can also be nterpreted as the probablty of a frm beng ncluded n the Moody s Bottom Rung lst (Hej, De Boer, Franses, Kloek and Van Djk, 2004, p. 443). Analyss Moody s selects frms wth recent deteroraton n credt qualty to comple the lst. Ths lst s beng updated by Moody s every month. The lst s avalable at The frst Bottom Rung lst has been made publc on March 10, 2009 and conssted of 283 publc and prvate companes. In ths study we examne only the publc companes on ths lst because they have avalable nformaton n popular databases such as the Center for Research n Securty Prces (CRSP) and COMPUSTAT. We dentfy 109 publc companes from ths lst. Out of these 109 frms 60 are stll actve (have a CRSP delstng code of 100), 20 merged (have a CRSP delstng code n the 200s), one frm has been nvolved n an exchange (CRSP delstng code n the 300s) and 28 have been dropped (have CRSP delstng code n the 500s). There are no frms whch have been lqudated n ths tme perod (havng CRSP delstng code n the 400s). Only 49 of the 109 publc companes have complete data avalable one, two, three, sx and nne months before and after March 10, Therefore, the control sample of 49 matchng frms s dentfed as follows: the 49 matchng frms have the same two dgts SIC code, plus or mnus 75% of the Bottom Rung frm s total assets n 2007 and 2008 and the same sgn for net ncome. Based on these crtera the matchng frms are as good of canddates for ncluson n the Bottom Rung lst but are not added to the lst by Moody s. We perform matched sample unvarate analyss on the 49 Bottom Rung and 49 matchng frms fnancal metrcs. Inqures & Perspectves 4 Volume 5 Number

5 Table 1: Descrptve Statstcs Matched Dfferenc e value Varable N Mean Mnmum Maxmum at2008match at at2007match at nmatch NI ebtmatch EBIT revtmatch REVT ** r1match r r2match r ** r3match r r4match r *** r5match r *** r6match r *** r7match r r8match r *** r9match r r10match r Table 1 provdes descrptve statstcs of fnancal characterstcs of Bottom Rung and matched sample of control frms. The table also provdes unvarate tests on the dfferent varables used n the analyss. The table ndcates that the matchng exercse was successful n that total assets n 2007 and 2008 and net ncome are not statstcally dfferent between the sample of Bottom Rung Frms and matchng sample of frms. Revenues and some of the rato metrcs of Bottom Rung and match frms however are statstcally dfferent. Average revenues of the Bottom Rung frms are 2, mllon, Inqures & Perspectves 5 Volume 5 Number

6 whereas average revenues of matchng frms are less 1,854 mllon. Matchng frms have statstcally sgnfcant and hgher R2 - earnngs before nterest, taxes and amortzaton to nterest expense rato, R4 funds from operaton and nterest expense to nterest expense rato, R5 funds from operaton to debt rato, R6 retaned cash flow to net debt ratos, but statstcally sgnfcant and lower R8 debt to book captalzaton rato. Table 2: Unvarate Analyss on Cumulatve Returns of Bottom Rung and Control Group Frms Matched Varable N Mean Mnmum Maxmum Dfferenc e value 1m_beforeMatch m_before m_beforeMatch m_before m_beforeMatch m_before <.0001 *** 6m_beforeMatch m_before * 9m_beforeMatch m_before * 1m_afterMatch m_after m_afterMatch m_after m_afterMatch m_after *** 6m_afterMatch m_after m_afterMatch m_after Next, we analyze the market performance of the Bottom Rung frms pror and after ncluson to the lst relatve to the control group of frms. We compute the dfference n cumulatve returns based on matched sample methodology. We defne the dfference n returns as follows: Dfference n Returns = Return (Matchng Frm) - Return (Bottom Rung Frm) (2) The unvarate test results on the dfference n returns, presented n Table 2, ndcate that one and two months before and after the ncluson n the Bottom Rung lst the frms experence smlar to the matchng frms returns. However, when the longer perods are examned some patterns emerge. When we look at sx and nne months Inqures & Perspectves 6 Volume 5 Number

7 before the ncluson on the lst the Bottom Rung frms underperform relatve to the match frms but mprove ther performance three months before the ncluson date. The sx month underperformance s % and the nne month underperformance s %. The three month extra performance by Bottom Rung frms s %. Even though both types of frms have negatve returns the Bottom Rung frms have even lower negatve returns sx and nne months before the ncluson event. When we examne the perods after, the sx and nne months after the ncluson event the Bottom Rung frms perform n statstcally smlar fashon to the frms whch are good canddates for the lst but are not ncluded n the lst. However, the economc performance of the Bottom Rung frms s superor to the match frms economc performance. The Bottom Rung frms cumulatve return s hgher by % after sx months and by % nne months after the event relatve to the match frms cumulatve returns. However, the Bottom Rung frms underperform three moths after the ncluson event relatve to the match frms. The three months returns of the Bottom Rung frms are lower by % relatve to the match frms returns. Both types of frms have postve cumulatve returns after the ncluson n the Bottom Rung lst event. The fact that the frms are ncluded n the lst however does not seem to have an effect on ther returns after the ncluson date. We fnd that the cumulatve returns of the Bottom Rung frms and matchng frms are not statstcally dfferent. The table also suggests that pror to the ncluson date there are meanngful sgnals suggestng that the Bottom Rung frms are antcpated by the market to underperform ndcated by the lower cumulatve returns. We next examne the queston what other factors besdes the lower returns mght be used by Moody s to choose these frms and not the matchng frms by usng a multvarate analyss. Before we conduct the multvarate analyss we examne the correlaton among the varables. Table 3 provdes correlaton coeffcents among the dfferent varables used n the logstc regresson analyss. The table suggests that cauton needs to be exercsed when varables R1, R2, R3, and R4, R5, R6 and NI, AT are combned n the multvarate analyss because of the hgh correlaton coeffcents among these sets of varables. Table 4 shows the logstc regresson results. The dependent varable s defned as havng values of one and zero wth one assgned to the Bottom Rung frms and zero to the matchng frms. The ndependent varables n the analyss are the ratos used most often by Moody s n ther analyss and control varables used n the bankruptcy lterature. The multvarate analyss suggests that consstently the most mportant factors for the dentfcaton of potental frms to be ncluded n the Bottom Rung lst are varables R1, R2, R8 and earnngs before nterest and tax. Where R1 s defned as earnngs before nterest tax and amortzaton dvded by average assets, R2 s defned as earnngs before nterest tax and amortzaton dvded by nterest expense, R8 s defned as debt dvded by book captalzaton. These varables are consstently statstcally sgnfcant n three model specfcatons desgned to exclude varables whch mght be hghly correlated among each other. Inqures & Perspectves 7 Volume 5 Number

8 Table 3: Correlaton Table r2 r3 r4 r5 r6 r7 r8 r9 r10 at n ebt r r r r r r r r r r at n Table 4: Logstc Regresson Results Panel A Panel B Panel C Estmate value Estmate value Estmate value Const ** * * r *** ** * r ** * r ** r r r r r * ** ** r r n at ebt ** ** ** n Test Ch-Sq value Ch-Sq value Ch-Sq value LR *** *** *** Score ** ** *** Wald * * Inqures & Perspectves 8 Volume 5 Number

9 The factor loadngs can drectly be nterpreted as contrbutng to the probablty of the frm beng ncluded n the Bottom Rung lst whch s one of the appealng characterstcs of logstc regresson analyss. Thus, the negatve coeffcents of R1 and R8 ratos can be nterpreted as follows: the hgher the R1 and R8 ratos, the lower the probablty of the frm beng ncluded n the lst. The nfluence of R2 and EBIT can be nterpreted as follows: the hgher R2 and EBIT the hgher the probablty of the frm to be n the lst, because these factors have postve coeffcents. The nfluence of R1, R2 and R8 on the probablty of ncluson n the lst makes economc sense: the proftable and less ndebted frms relatve to the sze of the frm should be less lkely to be ncluded n the lst. Concluson Moody s selects frms wth recent deteroraton n credt qualty to comple a lst and updates the lst every month. The frst Bottom Rung lst was ssued on March 10, 2009 and conssted of 283 publc and prvate companes. Naturally, n ths study we examne only the publc companes on ths lst because they have avalable nformaton n popular databases such as CRSP and COMPUSTAT. Surprsngly, none of the orgnal publc Bottom Rung frms have been lqudated. We fnd that frms ncluded n the Bottom Rung lst pror to the ncluson date experence lower cumulatve returns compared to a matched sample of control frms. We also fnd that the Bottom Rung frms do not underperform the control group after the publcaton of the lst. The multvarate analyss suggests that consstently the most mportant factors for the dentfcaton of potental frms to be ncluded n the Bottom Rung lst are the ratos of earnngs before nterest tax and amortzaton and average assets, rato of earnngs before nterest tax and amortzaton and nterest expense, rato of debt and book captalzaton, and earnngs before nterest and tax as a separate varable. In future research t would be nterestng to examne the performance of the frms n the recent Bottom Rung lsts. In ths study we are lmted to examnng only the publc frms and ther performance. It mght be nterestng to study the performance of the prvate frms on the lst to complete the assessment of Moody s methodology to create the Bottom Rung lst. References Barnv, Ran, Anurag Agarwal and Robert Leach. (2002) Predctng Bankruptcy Resoluton. Journal of Busness Fnance & Accountng, Vol. 29, No. 3/4, pp Glbert, Lsa R., Krshnagopal Menon and Kenneth B. Schwartz. (1990) Predctng Bankruptcy for Frms n Fnancal Dstress. Journal of Busness Fnance & Accountng, Vol. 17, No. 1, pp Hej, Chrstaan, Paul De Boer, Phlp H. Franses, Teun Kloek and Herman K. Van Djk. (2004) Econometrc Methods wth Applcatons n Busness and Economcs. Oxford Unversty Press, New York. Latnen, Erkk K. and Teja Latnen. (2000) Bankruptcy predcton: Applcaton of the Taylor's expanson n logstc regresson. Internatonal Revew of Fnancal Analyss, Vol. 9, No. 4, pp Lucchett, Aaron and Serena Ng. (2007) How Ratng Frms' Calls Fueled Subprme Mess. Wall Street Journal, August 15, 2007, page A1. Inqures & Perspectves 9 Volume 5 Number

10 Ohlson, James A. (1980) Fnancal Ratos and the Probablstc Predcton of Bankruptcy. Journal of Accountng Research, Vol. 18, No. 1, pp Poon, Wnne P. H. (2003) Are unsolcted credt ratngs based downward? Journal of Bankng & Fnance, Vol. 27, No. 4, pp Appendx We use the followng ratos from Moody s Fnancal Metrcs TM as possble factors: R1 =EBITA / Average Assets, R2 =EBITA / Interest Expense, R3 =EBITA Margn = EBITA / Net Sales, R4 =(FFO + Interest Expense ) / Interest Expense, R5 =FFO / Debt, R6 =RCF / Net Debt, R7 =Debt / EBITDA, R8 =Debt / Book Captalzaton, R9 =Operatng Margn = Net Income / Net Sales, R10 =Captal Expendture / Deprecaton Expense, where EBIT s Earnngs Before Interest and Tax, EBITDA s Earnngs Before Interest, Tax, Deprecaton and Amortzaton, RCF s retaned cash flow, and FFO s funds from operaton. Inqures & Perspectves 10 Volume 5 Number