Does default probability matter in Latin American Emerging markets?

Transcription

1 Does defaul probabiliy maer in Lain American Emerging markes? Isabel Abinzano, Luis Muga and Rafael Sanamaría January, 2013 Absrac: We analyse he impac of defaul probabiliy in four leading Lain American sock markes (Argenina, Brazil, Chile and Mexico). We find no posiive defaul risk premium excep in he case of Brazil, and in fac find a negaive risk premium for Argenina and Mexico. The laer effec ends o fade when he analysis accouns for size and BTM marke variables. Alhough we find no size effec in any of he markes considered, he BTM effec is very srong in all of hem, and our resuls reveal a consisen relaionship, analogous o ha found in more developed markes, beween defaul probabiliy and he size and book-o-marke variables. Keywords: Defaul probabiliy, size, book o marke, emerging markes JEL: G10, G14 ACKNOWLEDGEMENTS The auhors hank he edior and anonymous reviewers for heir helpful commens and suggesions. This paper has received financial suppor from he Spanish Minisry of Science and Innovaion (ECO ) and he Minisry of Economy and Compeiiveness (ECO C02-01). Isabel Abinzano paricularly acknowledges he financial suppor of he Andalusian Regional Governmen (P09-SEJ-4467). 1

2 Does defaul probabiliy maer in Lain American Emerging markes? 1.- Inroducion. Relaionships beween firms probabiliy of defaul and heir size and book-omarke have been under debae in he financial lieraure for he las weny years. Early sudies of his issue, underaken wihin he framework of he CAPM model, describe size and book-o-marke effecs as anomalies (see Sharpe, 1964 and Linner, 1965). Thus, Banz (1981) documens he size effec, while Chan, Hamao and Lakonishok (1991) and Fama and French (1992), among ohers, repor a booko-marke effec on sock reurns. These sudies show ha invesors can expec higher abnormal reurns from firms wih cerain characerisics. In he case of small firms, his is due o a higher deb load, which increases heir probabiliy of defaul (Fama and French 1992). In firms wih high book-o-marke raios, i is due parly o lower growh poenial and parly o higher defaul risk (Fama and French 1992). Based on his prior evidence, Fama and French (1993) add wo addiional facors o he CAPM by consrucing wo facor-mimicking porfolios o capure size (SMB) and book-o-marke (HML) effecs. In separae sudies, Fama and French (1995 and 1998) repor evidence of a defaul-risk effec on he reurns o hese mimicking porfolios. Neverheless, he Behavioural Finance approach argues ha he observed value premium for socks wih exreme characerisics is oo high o be explained by risk facors, and aribues more imporance o aspecs such as invesor overreacion o cerain ypes of announcemen (Lakonishok e al. 1994). Arguing from a differen perspecive, Daniel and Timan (1997) claim ha cerain level of premiums, as he observed in he case of Book o Marke, are no due o risk bu o sock characerisics. Several recen sudies have relaed size and book-o-marke effecs o he level of defaul risk associaed wih he socks. Using an approximaion based on risk facors and anoher based on sock characerisics, Vassalou and Xing (2004), invesigaing he US marke, show ha a firm s defaul probabiliy capures par of he size and book-o-marke effecs, bu ha he measure also incorporaes informaion ha is unconneced wih such effecs. Ferguson and Shockley (2003) reach similar conclusions, using an approximaion based solely on risk facormimicking porfolios. The above-menioned sudies use differen approaches for 2

3 measuring firms defaul probabiliy. Ferguson and Shockley s (2003) use measures of defaul risk based on accouning daa, such as Alman s Z-score or he leverage raio. However, his approximaion has several drawbacks. Accouning daa reflec a firm s pas performance and may offer lile indicaion of is fuure prospecs. Furhermore, since accouning-based models fail o consider sock volailiy, firms wih he same accouning raios will presen exacly he same degree of defaul probabiliy. Vassalou and Xing s (2004) approach, in conras, propose he use of a marke-based defaul probabiliy measure based on he Black-Scholes-Meron (BSM) opion-pricing model, which derives a firm s defaul-risk level from he marke price of is shares, hus overcoming some of he problems encounered when using accouning-based measures. The lieraure also offers disparae findings as o he impac of sock characerisics and he momenum effec in emerging markes. Gupa and Modise (2012) shows ha valuaion raios canno predic sock reurns in he Souh African Sock Exchange. Wang e al. (2012) finds no evidence of a momenum effec in he Taiwan Sock Exchange, while Muga and Sanamaría (2007) find a weaker effec in Lain American emerging markes han in hose of he developed counries. While some sudies (Fama and French 1998, Rouwenhors 1999, or Barry e al. 2002) have documened he presence of size and book-o-marke effecs in hese markes, ohers (Claessens e al or Serra 2003) have found, conrary o expecaions, negaive premiums for size and book-o-marke variables for some counries. Neverheless, he research has generally shown he value premium o be more robus han he size premium in emerging markes. Alhough here is some degree of consensus as o he imporance of size and book-o-marke effecs in his ype of markes, here is less consensus as o wheher hese effecs convey informaion abou defaul risk. Indeed, Esrada and Serra (2005) show ha Fama and French (1993) model, which incorporaes hese wo socks characerisics, do no yield saisfacory resuls for emerging markes. Our sudy, herefore, aims o provide an approximaion of he role played by firms defaul probabiliy in Lain American emerging markes and is relaionship wih he size and book-o-marke variables, using an approach based on sock characerisics. In his respec, we offer several conribuions o he lieraure. Firs, we es for he presence of a defaul risk premium in he four leading Lain American markes (Argenina, Brazil, Chile and Mexico), using a BSM-ype measure for he 3

4 esimaion of defaul risk in order o circumven problems arising from he use of accouning daa for his purpose. We hen es for an associaion beween he defaul risk measure and he size and book-o-marke variables in Lain American emerging markes, as Vassalou and Xing (2004) and Ferguson and Shockley (2003) have shown o be he case for he US marke. Finally, we perform calculaions o deermine wheher i is liquidiy, a key variable ha disinguishes developed from emerging markes, ha drives hese relaionships. To accomplish hese objecives, he paper comprises eigh secions. Secion wo presens he seleced defaul risk measure. Secion hree describes he daabase for he sudy. Secion four analyses he sock reurns sored by heir level of defaul probabiliy and size and book-o-marke characerisics. Secion five examines he relaionship beween defaul risk and firm size; and secion six he link beween defaul risk and he book-o-marke effec. Secion seven presens he robusness ess, where liquidiy is included as a conrol variable. Secion eigh summarizes he main conclusions. 2.- The defaul risk measure. Defaul risk, which can be defined as he uncerainy surrounding a firm s abiliy o service is debs and obligaions, has been approximaed in various ways hroughou he lieraure. The mos usual measures are hose based on accouning daa such as Alman s Z (1968) or Ohlson s O-score (1980), credi raings, deb spreads and marke-based measures, which use he BSM model. However, as noed by Hillegeis e al. (2004), here are various reasons o quesion he effeciveness of defaul risk measures calculaed using accouning daa. One is ha accoun saemens are inended o measure pas performance and may no ell us very much abou a firm s fuure prospecs. Furhermore, firms prepare accouning saemens under he going concern principle, which assumes ha he firm will no go bankrup. Anoher major drawback of hese measures is heir failure o consider asse volailiy, which leads hem o conclude ha firms wih similar raios will have exacly he same likelihood of going bankrup. However, volailiy is a crucial variable when aemping o predic defaul risk because i capures he probabiliy of a firm s asses being insufficien o cover is liabiliies. Ceeris paribus, he higher he volailiy of a firm s asse value, he greaer is defaul risk. 4

5 An alernaive o using he above-menioned approximaions of defaul risk is o consruc a measure using firms marke share prices, as in Vassalou and Xing (2004), Bysröm e al. (2005), and Bysröm (2006), among ohers. These sudies sar from Meron s (1974) proposal, which is o consider he firm s own equiy value as a European call opion on is asses value and use he Black and Scholes (1973) formula o calculae i. Defaul risk is approximaed by he following expression 1 : P def, V A, ln D = N 2 σ A, + µ 2 σ T A, ( T ) (1) where V A, is he value of he firm s asses a ime, µ is he expeced immediae rae of reurn on V A,, σ A, is asse reurn volailiy, D is he deb s face value, T is he mauriy period and N( ) is he cumulaive probabiliy of he Normal disribuion. To find he values of V A, and σ A, we use an ieraive process, similar o ha used by Vassalou and Xing (2004), beginning wih he marke price of he firm s shares. Thus, compared wih accouning-based models, he BSM measure has he advanage ha i no only akes ino accoun each firm s pas record, bu also manages o include invesors expecaions regarding fuure reurns by considering is curren share prices. In addiion, i considers he firm's sock reurn volailiy. Hillegeis e al. (2004) compare his measure, in his respec, o Alman s Z (1968) and Ohlson s O-Score (1980), finding BSM model o provide significanly more informaion regarding he firm s defaul risk han eiher of he oher models, which leads hem o recommend he use of he BSM measure o replace radiional accouning-based measures as a proxy for defaul probabiliy. Furhermore, general daabases, such as Thomson Financial, can provide scan daa for hese counries on some of he crucial variables needed o obain defaul risk esimaes. This fac could resul in a major sample bias 2. 1 See Vassalou and Xing (2004) for furher deails. 2 Daa on Earnings Before Ineres and Taxes, for example, are available for only 1.7% of socks in he Argenine sock marke, 5.6% of socks in he Brazilian marke, 3.8% of socks in he Chilean marke and 5.5% of socks in he Mexican marke. 5

6 Furhermore, compared wih he credi raing, as a defaul proxy, he BSM measure has he advanage of no lag beween variaion in credi worhiness and is incorporaion ino he risk measure, given ha, in he BSM measure marke prices are discouning expeced fuure cash flows. In addiion, i is a firm-specific measure in ha i provides a value for each firm based on is financial siuaion and is capializaion, which may differ from ha obained for anoher firm wih he same credi raing. In his way i provides more finely-uned rankings. Finally, requiring only a minimum amoun of informaion, i provides a measuremen value for every firm, no jus hose ha are credi-raed. The use of he BSM measure also overcomes some of he drawbacks associaed wih he use of credi spreads as a measure of defaul risk, such as he problem of muliple issues and he fac ha he firm needs o have raded bonds before a defaul risk measure can be obained. Noe ha i is usually easier o access a company s share price daa han is deb reurn daa. As shown by Vassalou and Xing (2004), he aggregae BSM defaul index, defined as he simple average of all firms in a marke, should signal periods of economic crisis, in he sense ha hey should be associaed wih higher aggregae marke levels of defaul probabiliy. Figure 1 shows aggregae defaul raes by counry. Thus, if he BSM index is a good defaul risk indicaor, i should show some correlaion wih he periods of economic crisis ha occurred in hese counries during he sample period. I can be seen ha his measure capures he main crises ha ook place in Lain America during he period analysed. In he case of he Argenine, he defaul rae reaches is peak wih he firs signs of economic disress owards he end of 2001, when exreme economic measures, including he noorious corralio, inroduced by he Miniser of Economy, Domingo Carvallo, and vehemenly proesed by he public, riggered a major economic crisis. This spread o oher Lain American counries in wha was ermed a ango effec, he sronges repercussions of which were fel in Brazil, Paraguay and Uruguay (which, ogeher wih he Argenine, make up MERCOSUR). Brazil s defaul risk rose beween 1998 and Inernal issues aside, he Souh Eas Asian financial crisis of 1997 and, especially, Russia s 1998 defaul on is deb may clearly have riggered he Brazilian financial crisis. The daa for Chile show defaul beginning o rise in 1998 hrough he knock-on effec of he Asian Crisis of 1997, due o Chile s geographical proximiy and economic ies wih Asian counries. Lasly, i is possible o see he effec of he equila crisis 6

7 from 1994 onwards, paricularly in Mexico, whose economy was hrown ino frenzy, following devaluaion of he peso. The siuaion spread o he whole of Lain America and is effecs persised unil The daabase. The daa, aken from he Thomson Financial daabase, refer o all sock lised in four Lain American Emerging Markes: Argenina, Brazil, Chile, and Mexico beween beween November and December In line wih Ince and Porer (2006), we have screened and correced he daabase. We have removed padded zero-reurn records a he end of he reurn ime series for delised socks. We have also removed all nonlocal firms, all lisings oher han hose on he primary exchange, and all lisings wih Type no equal o Equiy. We include only hose firms ha checked YES in he Primary quoe field. Banks, finance companies and insurance companies have been excluded from he analysis, because he peculiariies of heir capial srucure migh skew he defaul risk daa. We have also removed firms wih missing daa for some of he variables needed o calculae he seleced defaul risk measure. Finally, he availabiliy of daa on marke value and shor-erm and long-erm deb also had a noiceable impac on he sample, paricularly during he firs years of he daabase period. The average, maximum and minimum numbers of socks available for he sample and for he differen markes considered in he sudy are given in Table 1, which also shows he average values of he differen variables used in he analysis (reurn, BTM, size and BSM). Cross-counry dispariy is eviden, wih he Brazilian marke showing he highes average reurns ogeher wih higher BTM raios and higher BSM scores, while he Chilean marke has he lowes reurns ogeher wih a larger average firm size, lower BTM raios and lower BSM scores. Given he naure of he sudy, we use monhly daa for he differen variables wih he excepion of hose relaed o he Amihud (2002) illiquidiy measure since i requires daily daa. Following Vassalou and Xing (2004), we avoid problems relaed o reporing delays by using deb daa for he las hree quarers of he curren year and he firs quarer of he pas year. In line wih oher sudies 4, we 3 As noed in he previous secion, BSM esimaions require he use of pas daa o measure he volailiy boh of he firm s asses and is equiy, which means ha, alhough he sudy period sars in November 1992, he firs BSM esimaes are no obained unil January See, among ohers, Crouhy e al. (2000), Crosbie and Bohn (2003) and Vassalou and Xing (2004). 7

8 calculae he book value of deb as he sum of shor-erm deb and 50% of long-erm deb. As a proxy for he risk-free ineres rae we have used he 30-day prime rae for Argenina, he Selic rae for Brazil, 90-day adjusable rae noes from Chile s Cenral Bank for Chile and 28-day CETES for Mexico. 4.- Defaul risk, sock characerisics and sock reurns. We begin our analysis of he relaionship beween defaul risk and sock reurns by examining porfolios wih differen levels of defaul risk in order o deec poenially significan reurn differenials. We do his over a period running from January 1997 o December 2007 by soring he socks monhly ino five porfolios using quiniles of he firms defaul risk ranking. Finally, we calculae equally weighed reurns of he five porfolios and he reurn differenial beween he highes and lowes defaul risk porfolios for each of he four counries under consideraion. The resuls, repored in Panel A of Table 2, differ across he markes considered. The Argenine marke shows reurns decreasing monoonically from he low defaul risk porfolio o he high BSM porfolio. In oher words, in he Argenine sock marke, shares wih higher defaul risk show lower average reurns over he sudy period. Specifically, he monhly reurn differenial beween he high and low porfolios is -1.20% per monh, he difference is significan a 10% using Newey- Wes sandard errors. In he Mexican sock marke, alhough he reurn paern is no monoonic, he reurn differenial beween he high and low porfolios has he same direcion as in he Argenine marke, ha is, he reurns of socks wih high BSM scores are lower, in his case by -0.84% per monh, which is significan a 5%. In he Chilean sock marke, however, here is no clear reurn paern across he defaul risk porfolios, and he differenial beween he high and low BSM quiniles is no significan. Finally, he resuls for he Brazilian marke conras compleely wih hose repored so far for he res of he markes considered, in ha hey increase monoonically from he low BSM o he high BSM porfolio. The reurn differenial beween he high and low porfolios is 1.56% per monh, he highes reurns coming from he socks showing he highes defaul risk. Comparison of our resuls wih hose repored by Vassalou and Xing (2004) for he US sock marke reveals ha he only consisency, in erms of reurn paerns, beween heirs and ours is for he Brazilian sock marke, which is he only one 8

9 where we find higher reurns o be associaed wih higher defaul risk, approximaed by he BSM measure. Since Fama and French (1992), he research has suggesed ha a firm s defaul risk level could be linked in some way o is size and book-o-marke characerisics. Thus, in he second par of our analysis, we use hese wo variables o characerize he BSM-based porfolios. Vassalou and Xing (2004) repor ha in heir defaul risk porfolios higher defaul risk is associaed wih higher book-omarke raios and, overall, smaller firm size. The resuls of his characerizaion for each marke of ineres are also given in Table 2 (Panel A). In line wih he findings of Vassalou and Xing (2004), defaul probabiliy appears o be linked boh wih size and book-o-marke variables. Thus, in all he markes considered, he socks locaed in he high defaul risk porfolio show higher book-o-marke raios and smaller firm size. This suggess ha his is a reasonably accurae approximaion of defaul risk, despie limiaions deriving from he use of marke variables in financial markes of his ype. The observed relaionship wih reurns canno herefore be aribued o inaccurae esimaion of defaul risk, bu may derive from he characerisics of he asses raded on hese markes. These wo variables do no presen he same paern, however. Firsly, he booko-marke variable behaves in exacly he same manner in all four markes, increasing monoonically excep in he lowes defaul risk quinile. Secondly, size shows a less clear paern han book-o-marke, allowing us o repor nohing beyond wha has already been saed, namely, ha he high defaul risk quinile is associaed wih smaller firm size in all four of he markes under consideraion. These relaionships beween he defaul risk porfolios and he size and book-omarke variables could be he underlying facor for he observed variaion in reurn paerns, given ha here is no prior evidence of robus reurn paerns associaed wih hese wo variables in emerging markes, as already poined ou in he inroducion o his paper. 5.- The size effec. Vassalou and Xing (2004) have shown ha defaul probabiliy is more closely associaed wih size han wih book-o-marke in he US sock marke. This moivaes he second par of our analysis, which is o examine he relaionship beween BSM scores and firm size in Lain American markes. Neverheless, a key 9

10 prior o he analysis of hese markes is ha he lised firms do no adequaely represen heir respecive economies. This, ogeher wih he fac ha small firms are under-represened, may seriously affec analyses including his variable, paricularly in view of he fac ha, in developed markes, significan defaul risk premiums have been deeced only wihin he small size porfolio The size effec and sock characerisics. In a Fama and French (1993) asse-pricing framework, he smaller he firm, he higher he risk, and, as a resul, he higher he expeced reurns. In a Behavioural Finance framework, we would expec o find higher reurns in he lowes size quinile, because small firm size can serve as a proxy for socks ha are hard o value or o arbirage, or ha have a more uncerain informaion environmen (see Baker and Wurgler, 2006 or Jiang, Lee and Zhang, 2005). The resuls for he size effec are shown in Panel B of Table 2. As wih he defaul risk resuls, he reurn paern for he size effec is also couner o expecaions across he porfolios, in boh, a Behavioral Finance or a Fama French model framework. Furhermore, in hree of he markes considered, here is no significan difference in reurns beween large and small socks. In he Chilean sock marke, he average monhly reurn differenial beween large and small socks is 0.48% per monh, versus -0.20% for he Argenine marke, and -0.63% for he Brazilian marke, all hese differences being non-significan according o he saisic, as already noed. In he Argenine and Brazilian markes, however, he small size porfolios yield he highes reurns, as expeced. Finally, we do find significan reurn differences beween he large and small size porfolios in he Mexican sock marke. Specifically, here is a difference of 0.97% beween he average monhly reurns from he larges and smalles size porfolios, alhough he sign is couner o expecaions. Again, as found for he porfolios based on he BSM measure, despie he divergence of he reurn paerns across he markes considered, a number of commonaliies do emerge when he size porfolios are characerized by oher variables. Thus, in all four sock markes, he book-o-marke variable declines monoonically across he size porfolios from small o large, such ha he smalles socks presen he highes average book-o-marke raios. In erms of he BSM measure, in he Chilean, Mexican and Brazilian markes, defaul probabiliy is higher in he porfolios conaining he smalles socks, decreasing monoonically 10

11 owards hose conaining he larges. The Argenine sock marke, despie presening a very similar overall paern o he oher hree, differs in some ways, because, alhough he lowes average levels of defaul probabiliy appear in he large size porfolio, he highes level does no appear in he small size porfolio. The findings on firm size in relaion o book-o-marke raios and defaul probabiliy are consisen wih hose repored for developed markes, (see Vassalou and Xing 2004). Tha is, overall, small socks show higher book-o-marke raios and higher defaul probabiliy levels. Neverheless, i is imporan o emphasize ha porfolio reurn paerns vary across he differen markes considered The size effec and defaul probabiliy In ligh of he undoubed relaionship ha exiss beween defaul probabiliy and firm size, he observed porfolio reurns migh be explained by a grouping of socks sharing a specific characerisic. In his respec, Vassalou and Xing (2004) show ha he defaul probabiliy effec appears only in small socks, and ha he size effec is found only in socks wih high defaul probabiliy. In order o examine he effec of defaul probabiliy across firm size groupings, we proceeded as follows. Firs, all he socks in a given marke were sored ino eriles 5 according o size, afer which he socks in each of he resuling groupings were again sored ino eriles based on heir defaul probabiliy ranking, which is he variable of ineres. Panel A of Table 3 gives he reurns of he defaul probabiliy porfolios for he differen firm-size groupings. In he Argenine sock marke, no significan defaul probabiliy effec is found in eiher large or small socks. For he Chilean sock marke, as occurred in he analysis of he porfolios based on he BSM measure, no significan reurn differenials appear among he differen firm-size groupings. In he Mexican sock marke, meanwhile, alhough he low defaul risk socks presen he highes reurns, when hey are re-sored by size, he difference proves significan only for he middle firm size grouping, which shows an average monhly reurn differenial of 0.91%. Finally, he resuls for he Brazilian sock marke also reveal a significan average monhly reurn differenial of 1.10% and 1.16% beween he high and low defaul probabiliy porfolios in he large and small size groups, respecively, as indicaed by he saisic. 5 In he formaion of he wo-way soring porfolios, he firs soring was done by eriles in he case of he Mexican and Chilean sock markes and by he median value in ha of he Brazilian and Argenine sock markes, where he number of lised firms available was smaller. 11

12 In summary, afer conrolling for size, he negaive impac on reurns of defaul probabiliy fades in he Argenine sock marke, while, in he Mexican marke, is impac is found o be significan only in hose firms locaed in he middle firm-size grouping. Meanwhile, he posiive impac of defaul probabiliy found in he Brazilian marke holds for boh he large and small size groupings. In a similar manner, we examined he reurn paerns of he firm size porfolios across differen levels of defaul probabiliy. The resuls of his analysis, given in Panel B of Table 3, show ha, across all four of he sock markes considered, size has no significan effec in any of he defaul probabiliy groupings. Prior research neverheless shows ha he book-o-marke effec is more robus han he size effec in emerging markes, as advanced in he inroducion o his paper. The resuls of his analysis allow us o conclude ha here is a close associaion beween firm size and defaul probabiliy in he Lain American emerging markes. Firm size does no appear o have a significan impac on sock reurns, however, and is effec is even less significan when considered in relaion o defaul probabiliy levels. As a final poin, and as already noed, he under-represenaion of small firms in hese markes may be a possible explanaion for he resuls obained. 6.- The book-o-marke effec. The second defaul risk-relaed variable radiionally repored in he lieraure is he book-o-marke raio. Fama and French (1992, 1995 and 1998) sugges ha he higher reurns associaed wih firms having higher book-o-marke raios migh be relaed o defaul probabiliy. Our analysis herefore proceeds by examining he relaionships beween defaul probabiliy and book-o-marke raios in companies lised in he four Lain American sock markes under consideraion The book-o-marke effec and sock characerisics. We sar, as in he analysis of he relaionship beween defaul risk and size, by forming porfolios based on BTM raio quiniles for he four markes of ineres and observing he sock reurns and characerisics. In conras o our findings for he analysis of size in relaion o defaul risk, repored above, he book-o-marke based porfolios show a uniform reurn paern across he four sock markes considered, which is consisen wih he radiional inerpreaion for boh developed and 12

13 emerging markes. Tha is, in all four markes, we find higher reurns in he high BTM porfolios and lower reurns in he low BTM porfolios. The resuls are presened in Table 4, where i can be seen ha, in he Brazilian sock marke, he average monhly reurn differenial beween he high and low BTM porfolios is 2.44%, versus 0.91% in he Mexican marke, and 1.53% in he Chilean marke. All hese values are significan a 5% as indicaed by he saisic. In he Argenine marke, he average monhly reurn differenial reaches 1.19% and i is significan a 10%. Wih respec o he porfolio characerisics, as previous analyses sugges, he BTM porfolios correlae closely wih boh size and defaul risk. Thus, in all four of he markes considered, he high BTM porfolio conains he smalles firms. Furhermore, firm size decreases monoonically across he porfolios from low o high BTM socks. Meanwhile, he porfolios conaining he high BTM socks also presens he highes defaul risk levels in all four sock markes. The defaul risk paern is monoonically increasing from low o high BTM socks in he Argenine and Chilean sock markes. In he Mexican and Brazilian markes, he paern is no monoonic, bu he high BTM porfolio presens lower levels of defaul risk. The resuls displayed in his able reveal a srong impac of book-o-marke on sock reurns, which appears, a he same ime, o be relaed o defaul risk and firm size The book-o-marke effec and defaul probabiliy. The reurn differenials among he book-o-marke porfolios could be due o he relaionships of he laer wih oher variables such as size or defaul probabiliy. The differences may also be locaed in a specific group of socks sharing a cerain characerisic. We herefore repeaed he analysis performed on he size-based porfolios, in order o examine he ineracion beween he book-o-marke and defaul probabiliy effecs. We sar wih an analysis of he effec of defaul probabiliy on sock reurns by book-o-marke raio levels. The resuls, which appear in Panel A of Table 5, reveal no defaul risk effec in he Chilean sock marke. The Argenine sock marke, likewise, shows no significan reurn differenials beween high and low defaul risk socks in any of he book-o-marke groupings, alhough he analysis repored in Table 2 (Panel A) reveals an average monhly reurn differenial of -1.20% when no oher variable is aken ino consideraion. 13

14 The mos sriking of he resuls shown in Table 5 appear in he Mexican and Brazilian sock markes. The Mexican marke coninues o show differences in reurns beween he high and low defaul probabiliy groupings. The specific monhly average reurn differenials are -0.71% for he low BTM grouping, -0.99% for he middle BTM grouping, and -1.04% for he high BTM grouping. In oher words, in he Mexican sock marke, defaul risk sill has a negaive impac on a leas one group of socks when considered in relaion o he size or book-o-marke variables. Finally, he Brazilian sock marke shows no significan effec of defaul risk on reurns in eiher high or low BTM socks, even hough he previous analysis revealed an average monhly reurn differenial of 1.56% among he defaul risk porfolios. In summary, afer aking ino accoun he relaionships wih he book-o-marke variable, while he defaul risk effec fades for all sock groupings in he Argenine, Chilean and Brazilian sock markes, i sill has a significan impac in he Mexican sock marke, albei only in socks locaed in he middle size grouping. The second par of he sudy is an analysis of he effec of book-o-marke on reurns in he various defaul risk groupings. The resuls, shown in Panel B of Table 5, show ha in his case, he significan reurn differenials among he booko-marke porfolios hold o some exen across all he sock markes considered. Thus, he Argenine sock marke shows reurn differenials ha are significan a 10% boh in he high and low defaul risk groupings. In he Chilean and he Brazilian sock markes, a significan book-o-marke effec is found in all defaul risk groupings. Finally, he Mexican sock marke shows a significan book-omarke effec in he low and middle defaul risk groupings. In oher words, he book-o-marke effec proves o be consisen across all counries when analysed in relaion o BSM scores The book-o-marke effec and size. Having observed an imporan book-o-marke effec in all he markes considered, and having found ha he reurns of he book-o-marke porfolios are significanly associaed wih firm size, we decided o run a similar analysis, his ime consrucing porfolios based on hese wo variables. Table 6 shows he reurn paerns of he BTM porfolios by size groupings. As in he case of he analysis by defaul risk levels, all markes coninue o show a significanly posiive book-o-marke effec in a leas one size grouping. The effec 14

15 is apparen in boh he large and small size groupings for he Argenine marke, in all size groupings for he Chilean marke, in he middle and small size groupings for he Mexican marke, and only in he small size groupings for he Brazilian marke. These resuls are consisen wih hose repored for developed markes in ha he sronges book-o-marke effec on reurns occurs in small firms in all four of he markes considered in our sudy. Thus, in conras o wha was found for he size variable, he BTM variable responds as expeced and does no herefore appear o be significanly affeced by he daa limiaions of hese markes. 7.- Robusness Checks. The resuls obained by means of sock porfolio analysis allow us o draw a number of conclusions. Firsly, defaul risk did no have a uniform impac on reurns in all he markes considered during he sample period. Defaul premium is significan only in he Brazilian sock marke and is in fac negaive in he Argenine and Mexican markes. Secondly, as previous research on developed markes has repored, he observed defaul risk effec is associaed wih he size and book-o-marke variables, and, excep in he case of he Mexican marke, he effec fades when eiher of hese wo variables is aken ino consideraion. Finally, he resuls illusrae he imporan role played by he book-o-marke effec, which is presen in all four of he markes considered, even when aking ino consideraion ineracions wih sock characerisics relaing o defaul risk or firm size Fama-Macbeh regression on individual sock reurns. The robusness of he above resuls has been checked using an alernaive mehodology o porfolio analysis. We have used Fama-Macbeh (1973) regressions in which sock reurns for each marke in a given monh is explained in erms of firm size (Size), book-o-marke raios (BTM), level of defaul risk (DLI), and he ineracions beween hese variables. The esimaed equaions for each marke considered were as follows: α β (2) R i, = +. Size 1 + ui, α β (3) R i, = +. BTM 1 + ui, α β (4) R i, = +. DLI 1 + ui, α β γ λ (5) R i, = +. Size 1 +. BTM 1 +. DLI 1 + ui, 15

16 R α β. γ λ δ δ + u (6) i, = + Size 1 +. BTM 1 +. DLI DLI 1. Size DLI 1. BTM 1 The resuls, shown in Table 7, are consisen in overall erms wih hose obained from he porfolio analysis. In oher words, he book-o-marke raio, wheher considered in isolaion or in conjuncion wih oher variables, has explanaory power wih respec o sock reurns, in ha higher BTM raios are associaed wih higher sock reurns. In he Brazilian and Argenine markes, moreover, he BTM raio is he only variable ha has predicive power when considered in conjuncion wih all oher sock characerisics. Defaul risk, as in he porfolio analysis, shows significan explanaory power wih he expeced sign only in he Brazilian sock marke, and loses is significance when considered in conjuncion wih all he oher variables. These resuls confirm he findings obained from he porfolio analysis wih respec o he relaionship beween he book-o-marke, size and defaul risk variables and he relaive role of he book-o-marke raio in explaining sock reurns in emerging markes. The defaul risk effec appears o be weak overall, fading alogeher in he Brazilian marke once he book-o-marke effec is aken ino accoun. This is consisen wih findings for developed markes, where, according o several sudies, size and book-o-marke raios migh be serving as a proxy for firms defaul risk. i, Defaul probabiliy and liquidiy. Since one of he main disinguishing feaures beween developed and emerging markes is he lower liquidiy of he laer, his secion incorporaes liquidiy ino he analysis in order o check is impac on he relaionship beween defaul probabiliy and sock reurns in Lain American sock markes. Liquidiy has a wo-way causal relaionship wih defaul probabiliy (Vassalou e al. 2006) and may explain fuure sock reurns, since illiquid socks should yield higher fuure reurns han liquid socks. Repeaing he analysis using Amihud s (2002) raio as a liquidiy proxy, we find ha, in conras o he resuls for he variables examined in he previous analyses, in he case of liquidiy, he impac of he probabiliy of defaul on sock reurns is concenraed in one specific group of socks, he less liquid, regardless of he sign of he effec in he previous analyses 6. This shows ha he illiquidiy of emerging markes has no role in deermining he poenial presence of a defaul risk premium. However he defaul risk effec, whaever is 6 These resuls, no included in he paper o save space, are available from auhors upon reques. 16

17 direcion, is concenraed in less liquid socks. Finally, in ligh of he negaive relaionship beween size and liquidiy, we acknowledge ha hese findings migh be somewha differen if here were a greaer presence of small firms in hese markes. 8.- Conclusions. This sudy has examined he probabiliy of defaul in relaion o a series of variables in he conex of he Lain American emerging markes. The iniial finding is ha defaul probabiliy has a heerogeneous effec across he sock markes considered, alhough in all of hem i is concenraed in less liquid socks. Despie hese differences, here are several common feaures. Specifically, he resuls indicae ha here is an associaion beween probabiliy of defaul, size, book-o-marke and liquidiy. Thus, he socks wih he highes defaul probabiliy are also he smalles, and have he lowes liquidiy and he highes book-o-marke raios, jus as occurs in more developed markes (Vassalou and Xing 2004). Analysis of he impac of defaul probabiliy on reurns in differen sock groupings shows ha is effec ends o fade when he model accouns for size or book-o-marke variables. This allows us o conclude ha hese wo characerisics carry informaion relaing o firms probabiliy of defaul, as is he case in more developed markes (Vassalou and Xing 2004 or Ferguson and Shockley 2003). These resuls hold for eiher porfolio analysis or Fama-MacBeh regressions. The low represenaion of small firms in hese markes migh explain some of he resuls obained and accoun for heir heerogeneiy, in as far as defaul risk premiums in he US marke are found mainly in small firms, as shown by Vassalou and Xing (2004). Finally, his sudy shows ha book-o-marke plays an imporan role in predicing sock reurns in hese markes and hus does more han simply provide informaion relaing o firms defaul risk. Furher research is herefore required regarding he informaion conen of his variable and he poenial heerogeneiy of is effecs beween developed and emerging markes. REFERENCES Alman, E. I Financial raios, discriminan analysis and he predicion of corporae bankrupcy. Journal of Finance 23, no. 4 (Sepember):

18 Amihud, Y., Illiquidiy and sock reurns: cross-secion and ime-series effecs. Journal of Financial Markes 5, no.1 (January): Banz, R The Relaionship beween Reurn and Marke Value of Common Socks. Journal of Financial Economics 9, no. 1 (March): Baker, M. and Wurgler, J Invesor Senimen and he Cross-Secion of Sock Reurns. Journal of Finance 61, no.4: Barry, C., Goldreyer, E. Lockwood, L. and Rodriguez, M Robusness of size and value effecs in emerging equiy markes, Emerging Markes Review 3, (March):1 30. Black, F. and Scholes, M The pricing of opions and corporae liabiliies. Journal of Poliical Economy 81, no.3 (May-June): Bysröm, H Meron unraveled: A flexible way of modeling defaul risk. Journal of Alernaive Invesmens 8, no.4 (Spring): Bysröm, H., Worasinchai, L. and Chongsihipol, S Defaul risk, sysemaic risk and Thai firms before, during and afer he Asian crisis. Research in Inernaional Business and Finance 19, no.1 (March): Chan, L. K., Hamao, Y. and Lakonishok, J Fundamenals and sock reurns in Japan. Journal of Finance 46, no.5 (December): Claessens, S., Dasgupa, S, Glen, J Reurn Behavior In Emerging Sock Markes. The World Bank Economic Review, 9, No.1 (January): Crosbie, P. and Bohn, J Modeling defaul risk. Moody s KMV. Crouhy, M., Galai, D. and Mark, R A comparaive analysis of curren credi risk models. Journal of Banking and Finance 24, no.1-2: Daniel K. and Timan, S Evidence on he characerisics of cross secional variaion in sock reurns. Journal of Finance 52, no.1 (March): Esrada J. and Serra P Risk and reurn in emerging markes: Family maers. Journal of Mulinaional Financial Managemen 15, no.3 (July): Fama, E. F. and French, K. R The cross-secion of expeced sock reurns. Journal of Finance 47, no.2 (June): Fama, E.F, and French, K. R Common Risk Facors in he Reurns of Socks and Bonds. Journal of Financial Economics 33, no. 1 (February): Fama, E. F. and French, K. R Size and Book o Marke facors in earnings and reurns. Journal of Finance 50, no.1 (March): Fama, E. F. and French, K. R Value versus growh: The inernaional evidence. Journal of Finance 53, no.6 (December):

19 Fama, E.F. and Macbeh J Risk and reurn: some empirical ess. Journal of Poliical Economy 81, no.3 (March): Ferguson, M. and Shockley R Equilibrium Anomalies. Journal of Finance 58, no.6 (December): Gupa, R., and Modise, M.P Valuaion Raios and Sock Reurn Predicabiliy in Souh Africa: Is I There?. Emerging Markes Finance and Trade 48, no.1 (January-February): Hillegeis, S. A., Keaing, E. K., Cram, D. P. and Lundsed. K.G Assessing he Probabiliy of Bankrupcy. Review of Accouning Sudies 9, no.1: Ince, O. S., and Porer, R. B Individual equiy reurn daa from Thomson Daasream: Handle wih care!. Journal of Financial Research, 29, no.4 (Winer): Jiang, G., Lee, Ch.M. and Zhang, Y Informaion uncerainy and expeced reurns., Review of Accouning Sudies 10, no.2-3 (Sepember): Lakonishok, J., Shleifer, A. and Vishny, R. W Conrarian invesmen, exrapolaion and risk. Journal of Finance 49, no.5 (December): Linner, J The Valuaion of Risk Asses and he Selecion of Risky Invesmens in Sock Porfolios and Capial Budges. Review of Economics and Saisics 47, no.1 (February): Meron, R. C On he pricing of corporae deb: The risk srucure of ineres raes. Journal of Finance 29, no.2 (May): Muga, L. and Sanamaría, R The Momenum Effec in Lain American Emerging Markes. Emerging Markes Finance and Trade, 43, no.4 (July-Augus): Newey, W.K. and Wes, K.D A simple, posiive semi-definie, heeroskedasiciy and auocorrelaion consisen covariance marix. Economerica 55, no.3 (May): Ohlson, J Financial raios and he probabilisic predicion of bankrupcy. Journal of Accouning Research 18, no.1 (Spring): Rouwenhors, R Local reurn facors and urnover in emerging sock markes. Journal of Finance, 54, no.4 (Augus): Serra, A. P The cross secional deerminans of reurns: Evidence from emerging markes socks. Journal of Emerging Markes Finance 2, no.2:

20 Sharpe, W Capial Asse Prices: A Theory of Marke Equilibrium under Condiions of Risk. Journal of Finance 19, no.3 (Sepember): Vassalou, M. and Xing Y Defaul risk in equiy reurns. Journal of Finance 59, no.2 (April): Vassalou, M., Chen, J., and Zhou, L., The Relaion beween Liquidiy Risk and Defaul Risk in Equiy Reurns. EFA 2006 Zurich Meeings. Available a SSRN: hp://ssrn.com/absrac= Wang, C.P., Huang, H.H., and Huang C.H Momenum and Conrarian Profis Corresponding o he Coinciden Economic Indicaor on he Taiwan Sock Marke. Emerging Markes Finance and Trade, 48, Supplemen 1:

21 FIGURE 1: Average value of he BSM measure aggregaed by counry This figure shows he average value of he defaul measure given by he Black-Scholes-Meron model for Argenina, Brazil, Chile and Mexico, using he available daa described in Secion 3. 21

22 TABLE 1: Descripive saisics of he sample This able shows he average values of he sock characerisic variables used in he analysis: Reurn, Size, BTM and Defaul risk (BSM), and he average, maximum and minimum number of socks. Reurn Size BTM BSM Avg. Min. Max. Argenine Brazil Chile Mexico

23 TABLE 2: Reurn and Characerisics of BSM and Size Porfolios This able presens he average reurn, Size and BTM characerisics of he BSM porfolios (Panel A) and he average reurn, BSM and BTM characerisics of he Size porfolios (Panel B) in each of he markes considered for he period January 1997 January * and # denoe reurn differenials (H-L, High minus Low, and B-S, Big minus Small) ha are significan a he 5% and 10% levels, respecively, using he Newey-Wes (1987) sandard errors. Panel A: BSM Porfolios ARGENTINE Low High H-L Reurn # BTM SIZE BRAZIL Low High H-L Reurn * BTM SIZE CHILE Low High H-L Reurn BTM SIZE MEXICO Low High H-L Reurn * BTM SIZE Panel B: Size Porfolios ARGENTINE Small Big B-S Reurn BTM BSM BRAZIL Small Big B-S Reurn BTM BSM CHILE Small Big B-S Reurn BTM BSM MEXICO Small Big B-S Reurn * BTM BSM

24 TABLE 3: Dependen sor by size and BSM and by BSM and Size This able presens he January January 2007 reurns from he dependen sor by size and BSM defaul risk levels (Panel A) and by BSM defaul risk levels and size (Panel B). Size is sored ino eriles for Chile and Mexico and for Argenine and Brazil, he median is used o caegorize he socks as small or big. BSM is sored ino eriles for Chile and Mexico and for Argenine and Brazil, he median is used o caegorize he socks as low or high defaul risk. * and # denoe reurn differenials (H-L, High minus Low, and B-S, Big minus Small) ha are significan a he 5% and 10% levels, respecively, using he Newey-Wes (1987) sandard errors. Panel A: Dependen sor by Size and BSM ARGENTINE BSM SIZE SMALL BIG BRAZIL BSM SIZE SMALL * BIG # CHILE BSM SMALL SIZE BIG MEXICO BSM SMALL SIZE * BIG Panel B: Dependen sor by BSM and Size ARGENTINE SIZE SMALL 2 BIG B-S BSM LOW HIGH BRAZIL SIZE SMALL 2 BIG B-S BSM LOW HIGH CHILE SIZE SMALL 2 BIG B-S LOW BSM HIGH MEXICO SIZE SMALL 2 BIG B-S LOW BSM HIGH

25 TABLE 4: Reurn and Characerisics of BTM Porfolios This able presens he average reurn, Size and BSM characerisics of he BTM porfolios in each of he markes considered for he period January 1997 January * and # denoe reurn differenials (H-L, High minus Low) ha are significan a he 5% and 10% levels, respecively, using he Newey-Wes (1987) sandard errors. ARGENTINE Low High H-L Reurn # Size BSM BRAZIL Low High H-L Reurn * Size BSM CHILE Low High H-L Reurn * Size BSM MEXICO Low High H-L Reurn * Size BSM

26 TABLE 5: Dependen sor by BTM and BSM and by BSM and BTM This able presens he January January 2007 reurns from he dependen sor by BTM and BSM defaul risk levels (Panel A) and by BSM defaul risk levels and BTM (Panel B). BTM is sored ino eriles for Chile and Mexico and for Argenine and Brazil, he median is used o caegorize he socks as low or high book o marke. BSM is sored ino eriles for Chile and Mexico and for Argenine and Brazil, he median is used o caegorize he socks as low or high defaul risk. * and # denoe reurn differenials (H-L, High minus Low) ha are significan a he 5% and 10% levels, respecively, using he Newey-Wes (1987) sandard errors. Panel A: Dependen sor by BTM and BSM ARGENTINE BSM BTM LOW HIGH BRAZIL BSM BTM LOW HIGH CHILE BSM LOW BTM HIGH MEXICO BSM LOW # BTM * HIGH # Panel B: Dependen sor by BSM and BTM ARGENTINE BTM BSM LOW # HIGH # BRAZIL BTM BSM LOW # HIGH * CHILE BTM LOW * BSM * HIGH * MEXICO BTM LOW * BSM * HIGH

27 TABLE 6: Dependen sor by size and BTM This able presens he January January 2007 reurns from he dependen sor by size and BTM. Size is sored ino eriles for Chile and Mexico and for Argenine and Brazil, he median is used o caegorize he socks as small or big. * and # denoe reurn differenials (H-L, High minus Low) ha are significan a he 5% and 10% levels, respecively, using he Newey-Wes (1987) sandard errors. ARGENTINE BTM SIZE SMALL * BIG * BRAZIL BTM SIZE SMALL * BIG CHILE BTM SMALL * SIZE * BIG # MEXICO BTM SMALL * SIZE * BIG