Brooks_2014_The Credit Risk–Return Puzzle_Impact of Credit Rating Announcements in Australia and...

Pacific-Basin Finance Journal xxx (2014) xxxxxx

PACFIN-00713; No of Pages 19

Contents lists available at ScienceDirect

Pacific-Basin Finance Journal

j ourna l homepage: www.e lsev ie r .com/ locate /pac f inThe credit riskreturn puzzle: Impact of creditrating announcements in Australia and JapanEmawtee Bissoondoyal-Bheenick a,, Robert Brooks b

a Department of Banking and Finance, Monash University, Po Box 197, Caulfield East, VIC 3145, Australiab Department of Econometrics and Business Statistics, Monash University, Australiaa r t i c l e i n f o Corresponding author at: Bld H, 900 Dandenong RE-mail addresses: banita.bissoondoyal-bheenick@m

http://dx.doi.org/10.1016/j.pacfin.2014.09.0010927-538X/ 2014 Elsevier B.V. All rights reserved.

Please cite this article as: Bissoondoyal-Bhcredit rating ..., Pacific-Basin Finance Journa b s t r a c tArticle history:Received 2 April 2014Accepted 13 September 2014Available online xxxxTraditional asset pricingmodels postulate that high risk investments areusually associated with higher returns. However, this does not hold inthe relationship between credit risk and return. There is a known creditriskreturn puzzle, which highlights a negative relationship betweencredit risk and the stock market returns. The objective of this study isto assess the puzzling credit riskreturn relationship of stocks; in partic-ular, comparing the stock returns of high versus low credit risk firms, asmeasured by credit ratings from Standard and Poor's in Australia andJapan for a period from January 1990 to June 2012. Our results indicatethat the credit riskreturn puzzle exists in both Japan and Australia.However, it seems that the credit riskreturn anomaly is explained bythe downgrade announcements in the market and hence we concludethat downgrade announcements of a firm have a significant impact onthe cross section of returns.

2014 Elsevier B.V. All rights reserved.JEL classification:G12G15

Keywords:Credit riskreturn puzzleCredit ratingsDowngradesEconomic cycles1. Introduction

If credit risk is systematic, onewould expect a positive association between credit risk and realised return.However, there is a known puzzle in the credit riskreturn relationship in that there is a negative relationshipbetween credit risk and stock market returns. Hence the literature suggests that firms with high credit risktend to have a lower return than firms with lower credit risk. Analysing the credit riskreturn puzzle, recentresearch by Campbell et al. (2008) shows evidence that the distress effect is stronger amongst small, illiquidstocks. Moreover, Dichev and Piotroski (2001) show that low credit quality firms perform poorly after down-grades, which they attribute to market underreaction. The relationship between credit rating and return hasbeen analysed by Avramov et al. (2009). However, it should be highlighted that most of the studies analysingoad, Caulfield East, VIC 3145, Australia.onash.edu (E. Bissoondoyal-Bheenick), [email protected] (R. Brooks)

eenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001

http://dx.doi.org/10.1016/j.pacfin.2014.09.001mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.pacfin.2014.09.001http://www.sciencedirect.com/science/journal/0927538Xhttp://dx.doi.org/10.1016/j.pacfin.2014.09.001

2 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxthis puzzle have been US centred. The key aim of this study is to provide evidence on this question by inves-tigating the empirical relation between credit risk and systematic risk by undertaking an analysis in Australiaand Japan.Wemeasure credit risk by using the long term foreign currency credit ratings provided by Standardand Poor's.

The rationale behind this study is, if we expect financial markets to be integrated the credit riskreturnpuzzle should hold in the developed markets. With the move toward a more financially integrated market,through stronger trade and financial linkages across countries, we would expect greater business cyclesynchronization. The huge spillover effect following the global financial crisis (GFC) further highlights thatworld markets are highly integrated. In their paper, Avramov et al. (2009) highlight that the premium be-tween high and low rated stocks persists even after returns are calculated using alternative pricing models,namely, the capital asset pricing model (CAPM) or Fama and French (1992, 1993) in the US stock market,at the firm level. Hence, the key research question that we consider in this paper is that if financial marketsare strongly related, this relationship should hold in other markets, in particular for other developed nations.In fact, the global financial crisis has revived the discussion on the financial markets integration. For instance,the emerging markets appear to have decoupled from developed nations for part of the GFC period, see forexample Dolley andHutchison (2009). In simple terms, the argument is that emergingmarkets have achievedstronger economic growth and hence emergingmarkets have decoupled from advanced economies. In 2007,the decoupling hypothesis held that Latin American and Asian economies, especially emerging ones, hadbroadened and deepened to the point that they no longer depend on the United States economy for growth,leaving them insulated from a slowdown there, even during the GFC. The beliefs in thefinancial marketswerethat these countries have generated strong outperformance for stocks outside the United States. Further,Avramov et al. (2012) study why global asset pricingmodels have failed to capture the cross section of coun-try equity returns. They indicate that the equities of countries in the high credit risk group, which are theemergingmarkets, outperform the equities of countries in the low credit risk group, which are the developednations. Hence, this suggests that at the sovereign level, we do not have a credit riskreturn puzzle, in contrastto the results at the individual firm level for the US market. Hence this brings us to our first key research con-tribution of this study. Avramov et al. (2009) assess the credit riskreturn puzzle in the USmarket and estab-lish that low credit risk firms realize higher returns than high credit risk firms and conclude that this ispuzzling as investors seem to pay a premium for bearing credit risk. Our contribution is different in thatwe analyse the credit riskreturn puzzle using two developed markets which have different characteristicsthan the US market, namely, Japan in Asia and Australia in the Pacific region, both being the largest marketsin the respective regions. The rationale for choosing Australia is that amongst the developed markets of theworld, Australia's experience in the global financial crisis has been different. Ali and Daly (2010) investigatethe macroeconomic determinants of aggregate credit risk and the implication for capital requirements byfocusing on the Australian and the US market. They conclude that Australia has been relatively immune tothe recent crisis and that the US markets is more susceptible to adverse macroeconomic shocks. Further,Brown and Davis (2010) analyse Australia's experience in the global financial crisis and conclude that of de-veloped economies around the world, Australia has emerged as amongst the least affected by the Global Fi-nancial Crisis. As such, extending the work of Avramov et al. (2009), we consider Australia to test if thecredit riskreturn puzzle holds in developed markets other than the US. The next country included in ouranalysis is Japan. We consider Japan given that Japan is a bank based system in contrast to the US market,which is a market based system. One of the key features of the Japanese market is that banks have playedan essential role in the financial system and have been the most dominant source of funding for businesses;see for example, Hoshi and Kashyap (2001). The debate on bank-based and market based system dateslong back in the literature, for example, Boyd and Smith (1998), Rajan and Zingales (1998) and Levine(2002) amongst others. As such our contribution will be to assess if the credit riskreturn puzzle holds in abank based system like Japan.

We use a panel regression approach for our modelling framework. We model the credit riskreturn byconsidering different states of the economy. Goldstein (2012) argues that the credit spread puzzle can be ad-dressed by taking account of such factors as the variability of the level of risk premium and the likelihood ofdefault over the course of the economic cycle. He argues that if default models incorporate economic states,like recessions, where defaults are more likely to happen, this can resolve the credit spread puzzle. Hence,we further differentiate our study from Avramov et al. (2009) and contribute to the literature by assessingwhether the credit riskreturn puzzle varies over different states of the economy as classified by a high,Please cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001

http://dx.doi.org/10.1016/j.pacfin.2014.09.001

3E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxmedium and low level. We initially classify the states of the economy by following the method used inMcQueeen and Roley (1993) and hence calculate the state of economy on amonthly basis, using the IndustrialProduction Index over the entire sample period. We also consider alternative classifications for the state ofeconomy based on market return quintiles and crisis dates. Third, by considering the credit risk-puzzle inthe global markets, we are equally contributing to the financial markets integration literature.

Our evidence suggests that the credit riskreturn puzzle does hold in both of the developed markets,Australia and Japan. Once we establish that the anomaly exists in both of the countries and is similar to theresults obtained in the USmarket, we explorewhat are the explanations for this puzzle. Both of thesemarketshave different institutional characteristics and hence we extend our analysis to assess the credit riskreturnpuzzle between different sub-samples including, investment grade firms, small versus large firms, and thecredit riskreturn relationship in different economic states. The results suggest that investment grade firmsdo have the anomaly whilst splitting the sample between small firms and large firms does not matter forJapan. The credit riskreturn puzzle exists in Japan, in particular when the state of economy is high andlow. When the economy is in the medium state, we do not have the credit riskreturn puzzle. We furtherfind that the Australian market has the credit riskreturn anomaly, but the results are mainly driven bylarge firms, which appear to be of investment grade rating. The results do not hold in the sample of smallfirms in the Australian market, that is the high credit risk firms. The credit riskreturn puzzle holds inAustralia when the states of economy are low and medium only which is consistent to the argument madeby Goldstein (2012). We equally confirm that Australia's experience in the crisis is different to other devel-oped markets. Further, we extend the analysis by considering rating changes of firms between upgradesand downgrades. A key finding in our paper is that the credit riskreturn anomaly exists mainly due to theannouncement of rating downgrades. This is the case for both Australia and Japan in that when we considerthe sample where we have firms with rating downgrades, we do have the credit riskreturn puzzle, but oncewe remove the downgrades events, we no longer have any credit riskreturn anomaly. Hence, the key conclu-sion is that the credit riskreturn puzzle in both of these developed markets can be explained by theannouncements of rating downgrades.

The remainder of the paper is organized as follows. Section 2 provides a summary of the literature.Section 3 presents the data and modelling framework. The results that are discussed in Sections 4 and 5provide some concluding remarks.

2. Literature review

2.1. Credit riskreturn puzzle

A key area of research that has been undertaken mainly in the US market is an assessment of the creditriskreturn anomaly. Analysing the credit riskreturn puzzle, Campbell et al. (2008) argue that since 1981, fi-nancially distressed stocks have low returns. One of the key issues they investigate is that, given an empiricalmeasure of distress, do the stock prices offinancial distressedfirmsmove together andwhat returns have theygenerated historically. Hence they investigate if there is evidence as to whether distress risk carries a premi-um. They consider credit rating as one of the measures of financial distress, using the default category of rat-ings, namely the rating category, D, as denoted by Standard and Poor's and conclude that stocks with a highrisk of failure tend to have low average returns and show evidence that the distress effect is stronger amongstsmall, illiquid stocks. Dichev and Piotroski (2001) provide a comprehensive study of long-run returns afterbond rating changes by using an extensive sample from 1970 to 1997. Their analysis is based on observations,which includes small, lower quality firms. They examine several time periods, including three-month, six-month, first-year, second-year, and third-year abnormal stock returns following bond ratings. The key findingof their paper is that low credit qualityfirms performpoorly after downgrades,which they attribute tomarketunderreaction. Griffin and Lemmon (2002) use a direct proxy of the likelihood of financial distress proposedby Ohlson (1980), the O-score, to examine the relationship between book-to-market equity, distress risk, andstock returns. Consistent with the mispricing arguments, they find that firms with high distress risk exhibitthe largest return reversals around earnings announcements and the book to market return premium islargest in small firms and hence conclude that poorly performing high credit risk firms also have low bookto market ratios. In fact their results are consistent with Dichev (1998). Dichev (1998) uses bankruptcy riskas a proxy for distress and argues that if bankruptcy risk is systematic, then there should be a positivePlease cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001


4 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxassociation between bankruptcy risk and realised returns. The main result from the study is that bankruptcyrisk is not rewarded by higher returns and hence confirms that firmswith high bankruptcy risk earn substan-tially lower returns.

Further, there are numerous papers which report evidence that distressed stocks underperform the mar-ket, but the results vary with the measure of financial distress that is used. For instance, Vassalou and Xing(2004) calculate distance to default; they find some evidence that distressed stocks with a low distance todefault have higher returns, but this evidence comes entirely from small value stocks. Shumway (1997)finds some evidence that firms with high distress risk do earn higher returns. Garlappi et al. (2008), on theother hand, do not find the negative credit riskreturn relation anomalous. They argue that, due to violationsof the absolute priority rule for claimants at bankruptcy, distressed stocks have lower betas and, therefore,command lower returns.

Whilst most papers in the literature use alternative distress measures, Avramov et al. (2009) find that thepremia between high and low rated stocks persists even after returns are adjusted by the capital asset pricingmodel (CAPM) or Fama and French's three factor model and consider credit ratings as themeasure of defaultrisk. Theyfind that the effect arisesmainly from thepoor performance of the lowest rated stocks around ratingdowngrades. They provide an institutional-selling explanation and suggest that the low-rating premium isdue to mispricing, rather than systematic risk.

Hence, most research undertaken in the literature has a US stock market focus. In this paper, we willcontribute by assessing whether the credit riskreturn anomaly exists in other developed markets in partic-ular by focussing onAustralia and Japan and analysingwhether the credit riskreturn puzzle holds as the stateof the economy varies.

2.2. Impact of rating announcements on the Australian and Japanese stock markets

Literature in this area has primarily assessed the impact of rating changes on the bond and stockmarkets,and ultimately led to the stylised finding that rating downgrades do have a significant wealth impact on themarket and rating upgrades do not carry the same informative value, see for example, Goh and Ederington(1993), Akhigbe et al. (1997), Dichev and Piotroski (2001). There are a few key papers which analyse the im-pact of rating announcements on the stockmarket returns usingAustralian data. The stock reaction during therating revision announcements in the Australianmarket is analysed by Matolcsy and Lianto (1995), based onrating revisions announced by Standard and Poor's (S&P) for the period 1982 to 1991. They undertook a studyto provide evidence on the information content of accounting income numbers and testing the incrementalinformation content of bond rating revision. The results of this study are consistent with other US studieswhereby bond downgrades have additional information content whilst upgrades do not have the same im-pact. Hence, the main finding by Matolcsy and Lianto (1995) is consistent with the view that rating agenciesonly add value to the already existing information set on downgrades. Thesefindings could be consistentwiththe propositions that good news travels fast compared to bad news, or that equity holders aremore concernedwith downgrades than upgrades, see for example, Holthausen and Leftwich (1986). Further, a comprehensivestudy on Australian bond rating revisionwas undertaken by Creighton et al. (2007). Based on rating revisionsannounced by both Standard and Poor's and Moody's from January 1990 to July 2003, they found a slightlydifferent result in comparison to Matolcsy and Lianto (1995). They found a significant positive movementin the stock price during upgrade announcements and negative stock price movement during downgradeannouncements. They also found that the stock price effect is larger for small companies and for bonds thatare downgraded from investment to speculative grade.

Themost recent research on Australia is by Chan et al. (2009) who focus onwhether the rating agency is aleading or lagging guide that influences the share price. By comparing the information content of a subscrip-tion rating agency (Corporate Scorecard Group) with the non-subscription based rating agencies (S&P andMoody's) in Australia and using the buy-and-hold abnormal returns, they found that the rating report provid-ed by Corporate Scorecard Group is beneficial to the subscribers and no abnormal share reaction is found to besignificant after the announcements of rating revision by non-subscriber rating agencies. The literature in theAustralian market is similar to the analysis in the US market where rating announcements do carry informa-tive value to the market.

One of the key analyses undertaken in the Japanesemarket is by Li et al. (2006). They used rating informa-tion for Japan from 1985 to 2003 and confirm that the market reactions are stronger for downgrades than forPlease cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001


5E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxupgrades, andmarket reactions are stronger for changes involving speculative grade debt rather than invest-ment grade debt for rating downgrades. For upgrades, the distinction between speculative and investmentgrade is insignificant. They equally find that the market reacts to negative refinements in ratings announcedby S&P and observe that the international rating agencies aremore substantial in their analysis in comparisonto local rating agencies.

3. Data and modelling framework

Our sample consists of monthly returns data and firm characteristics of common stocks for Australia andJapan for the period January 1990 to June 2012. The ratings data is obtained from rating Xpress from theWRDS database, which provides the rating history for each individual company for our sample period. Weinclude a number of firm characteristics as control variables in the model. The monthly returns and thefirm characteristics data are obtained from DataStream.

The ratings used in this study include Standard and Poor's issuer long term and issuer short term foreigncurrency ratings. Standard and Poor's define an issuer rating as1 Issuand foreglobal rRatings

2 Thedata.

Pleacredforward-looking opinion about an obligor's overall creditworthiness in order to pay its financialobligations. This opinion focuses on the obligor's capacity andwillingness tomeet its financial commit-ments as they come due. It does not apply to any specific financial obligation, as it does not take intoaccount the nature of and provisions of the obligation, its standing in bankruptcy or liquidation,statutory preferences, or the legality and enforceability of the obligation.The issuer ratings are hence at thefirm level and include both short termand long term ratings. A short termrating reflects the issuer that is firm's creditworthiness over a shorter time period.1 We include the long termforeign currency rating, the short term rating and the rating outlook. We calculate the rating changes2 for boththe long term and short term ratings.

Following previous studies in the ratings literature, we transpose the alphabetical grades to numericalgrades. The highest long term rating of AAA is mapped to a rating grade of 1. Hence if the numerical gradeincreases, this implies that there is a downgrade of the rating of the firm, whilst if the numerical grades de-creases, this implies an upgrade. We classify the ratings in 22 categories, as follows: AAA is 1, whilst AA+is 2, AA is 3 and AA is 4 and so forth. The lowest category of rating which represents default, D representsa numerical grade of 22.

The firm characteristics that are included are similar to Avramov et al. (2009) and are in line with thechoice of variables in Brennan et al. (1998). We collect the following firm characteristics: (1) Log (size-MV), measured by the market value, which is the share price multiplied by the number of ordinary sharesin issue. The market value is displayed in millions in local currency; (2) Log (BVMV), which is the ratio ofthe book value to market value; (3) Log(Turnover) which is the ratio of trading volume (VO) to the numberof shares outstanding (NOSH); (4) cumulative return, which is the return over the last 6 months; and (5) Log(price/earnings ratio). There is a wide literature supporting the choice of firm characteristics in explaining thevariability of returns. For example, several studies argue that thefirm size and the book tomarket ratio are themost significant variables in explainingfirm returns, see for example, Chanand Chen (1991), Fama and French(1992, 1993), and Chen et al. (1986). The turnover variablemeasures the ratio of trading volume to the num-ber of shares, which is a measure of liquidity. Whilst there are various measures of liquidity, for instance thebid-ask spread, see, Amihud and Mendelson (1986) and Brennan and Subrahmanyam (1996), we considervolume as suggested by Brennan et al. (1998) and Petersen and Fialkowski (1994) who conclude that thequoted spread is only loosely associated with the effective spread and suggest that it is therefore possiblethat trading volume provides a better measure of liquidity than the bid-ask spread. The cumulative returncaptures the momentum factor. Jegadeesh and Titman (1993) report that stocks with higher returns in theer ratings can be short term rating which are different to long term rating. Standard and Poor's provide both of the ratings in localign currency rating.While the local currency ratings are important, the foreign currency forms the basis whichmakes the rating aating. For more details on the rating definitions, refer to the following article: www.standardandpoors.com/spf/general/Direct_Commentary_979212_06_22_2012_12_42_54.pdf.ratings are analyzed for each firm and rating changes are calculated on a monthly basis from time t to t + 1 as they occur in the

se cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001

http://www.standardandpoors.com/spf/general/RatingsDirect_Commentary_979212_06_22_2012_12_42_54.pdfhttp://www.standardandpoors.com/spf/general/RatingsDirect_Commentary_979212_06_22_2012_12_42_54.pdfhttp://dx.doi.org/10.1016/j.pacfin.2014.09.001

6 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxprevious 12 months tend to have higher future returns than stock with lower returns in the previous12 months, that is themomentum factor. Similarly, Carhart (1997) develops what is known as the four factormodel which includesmomentum. As such in this study, we consider the cumulative return over a six monthperiod.

Given the firm characteristics data and ratings data are collected fromdifferent sources, thefinal sample offirms for which we have both ratings and firms characteristics data include 58 firms for Australia (hence wehave 58 270 monthly observations), and 245 firms for Japan (245 270 observations). To obtain a betterperspective of the data, we present Table 1 which provides some descriptive statistics on the firm character-istics as well as the ratings. The firm with the highest rating in Australia is Telstra with a numerical ratinggrade of 2, which is an equivalent of AA+. The firm with the highest average MV over the sample period isBHP Billiton, with an average market value of AUD 58,331 million. It should be noted that the maximumMV for BHP was AUD 157,400 million in November 2007. The average rating of BHP over the sample periodis a numerical grade of 5,which is a rating grade of A+. The firmwith the lowest rating in Australia isMirabelawith a rating grade of 16. However, this is not the firm with the lowest MV. The firm with the lowest MV isGoodman Group, with a lowest market value of AUD 0.23 million in January 1993. The larger firmstend to have a better credit quality as expected; however, size and credit ratings are not perfectly corre-lated. A key point to be noted in Australia is the long term rating changes. The rating changes are of sig-nificant magnitude. The mean for the long term change for downgrades is 7, which indicate that the firmsin Australia have been subject to significant downgrades. Whilst for upgrades, the maximum upgrade isby 2 notches only.

For the Japanesemarket, thefirmwith thehighest rating is AAA, (numerical grade 1) isHitachiMetalswithan averageMVover the sample period of JPY 316,919million. It should be noted that the firmwith the highestMV is NTT Docomo Inc. with a maximum of JPY 42,134,280million inMarch 2000, with an average long termrating of AA, over the sample period. The firm with the lowest MV in the sample for Japan is Yamada Denki,with a MV of JPY 3,970 million in August 1992. The firm with the lowest rating at C, which is a numericalgrade of 21, is Sojitz. It should be noted that this firm has the highest upgrade by 8 notches. However mostof the other upgrades in Japan are by 1 notch or 2 notches. Similar to the Australian market, there are someextreme downgrades, but this is not as frequent as the case of Australia. We have downgrades of up to 6notches in Japan.

In this study, we do not consider portfolio formation; we focus on whether credit risk is priced amongstindividual stocks. By considering the cross section of stocks, we can avoid datamining biases that are inherentin portfolio bases approaches as noted by Lo and Mackinlay (1990). In contrast to Avramov et al. (2009), weuse panel regression with both period and cross section fixed effects. Avramov et al. (2009) highlight that thecredit risk puzzle exists evenwhen alternative asset pricingmodels are employed that is the capital asset pric-ing model (CAPM) or Fama and French (1992, 1993). Hence, we model using panel equations using ordinaryleast squares methods, with fixed effects in both the cross-section (firms) and period (time) dimensions.Some of the firms have missing data from the database and the firms may be listed only in recent years andTable 1Summary statistics of firm characteristics and ratings.The MV is displayed in millions of local currency. Volume is expressed in thousands. NOSH represents the capital of the company. Thecumulative return is the cumulative for the last 6 month return.

MV Volume NOSH Cumulative return LOG_PE VOL_NOSH Long term rating

AustraliaMean 9441 73,314 1,218,880 0.0432 1.2241 0.0752 7Maximum 157,400 2,236,305 17,878,750 2.1092 3.4014 19.1825 16Minimum 0.2300 0.1000 14,876 2.4019 0.1549 0.0000 2.00Observations 12,193 12,139 12,193 11,788 10,696 12,142 7941

JapanMean 917,553 54,455 691,630 0 1.5038 0.7046 7Maximum 42,134,380 63,152,900 24,050,980 1.8746 4.2624 6.6667 21Minimum 3970 0.1000 190 2.0666 1.0000 33.3333 1Observations 62,593 60,201 62,592 60,930 55,407 62,081 29,732

Please cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001


7E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxhence we are left with an unbalanced panel. One of themajor benefits from using panel data is that it enablesus to control for firm level heterogeneity. Not controlling for these unobserved individual specific effects canlead to bias in the resulting estimates. Hence we estimate the following model for each of the countries.3 The4 For

releases

PleacredRit i i Longterm rating 2i Longterm rating Change 3i Short term rating 4iShort term rating change 5i Outlook 2i Firm characteristics 2 ei

1Where long term rating is the numerical scores, with a score of 1 represent an AAA rating and the lowestrating C represents 21 and a default rating, D, represents 22. In order to establish whether we have a creditriskreturn puzzle, where firms with higher credit quality enjoy higher returns rather than firms with lowcredit quality, we should expect a negative relationship between returns and ratings. This implies that is asthe numerical scores decrease (that is firms are being upgraded), the stock returns increases. If there is nosuch market anomaly, then we will expect that as firms are downgraded (as numerical scores increases),the stock return will be increasing and hence a positive relationship between long term rating andreturns. Short term rating has a maximum numerical score of 7, where the highest score is 1 being a rat-ing of A-1 and 7 represents default.3 A positive outlook is mapped as 1, a stable outlook is mapped as 0and a negative outlook is mapped as 1. The firm characteristics, as explained earlier include LOG(MV), Log (BVMV), Log (turnover), and cumulative return (for the last 6 months) and Log (P/E). FollowingBrennan et al. (1998), these variables are lagged by 2 months relative to the month the dependent variableis measured.

Further, one of the key contributions of this study is to assess whether the credit riskreturn puzzle variesover different states of the economy as classified by a high, medium and low level. As such, we classify thestates of economy for each of the countries, Australia and Japan, in these 3 states on amonthly basis from Jan-uary 1990 to June 2012. FollowingMcQueeen and Roley (1993), the state of the economy has been calculatedand classified as high, medium and low. In order to have the three states, we use the seasonally adjustedmonthly Industrial Production Index, obtained from the Federal Reserve Economic Data (FRED), to define eco-nomic states. These data are available from the FRED website for each country on a monthly basis.4 First, weestimate a trend in the log of Industrial Production Index by regressing the actual log of Industrial ProductionIndex on a constant and a time trend from the start of the sample date, January 1990. Then, we add and sub-tract the quartile from the trend, creating the upper and lower bounds. Hence, we add the lower quartile, Q1,to the fitted value to reach the lower bound, denoted as the low economic activity; we add the upper quartile,Q3, to the fitted value to get the upper bound, denoted as the high economic state. Medium economic activityis represented by the remaining observations between thebounds. Oncewe calculate these bounds,wemodelEq. (1) in each of the states and assess the credit riskreturn relationship in varying states of the economy aswell as partitioning the sample in various categories to establish if the credit riskreturn puzzle holds inAustralia and Japan.

There are alternativeways of classifying the state of the economy. In this study, in addition to the approachdescribed, we consider whether credit riskreturn puzzle holds in the bear and bull regimes by consideringvarious crisis periods. The global financial markets have suffered different crisis, (Mexican crisis, Argentineancrisis, Asian crisis, Russian crisis, Brazilian crisis etc.), causing a large amount of disruption and volatility in thestockmarkets. Themost recent crisis that is being considered as themost significant international crisis is theglobal financial crisis (GFC). Numerous papers examine the effect of the global financial crisis on both devel-oped and emerging markets see for example, Crotty (2009), Naoui et al. (2010), Neaime (2012). As such, weidentify 6 key crises, namely, Asian, Russian, Long Term Capital Management (LTCM), Brazilian, Dot-com,Argentinean and the Global Financial Crisis and hence classify the state of economy in a crisis and non-crisis period. Further, we consider the state of the economy based on market return quintiles. We dividethe market returns in quintiles and define the high state of economy as the top two quintiles, the mediumstate as the third quintile and the bottom two quintiles are classified as the low state of economy.mapping is as follows: A-1is 1, A-2 is 2; A-3 is 3, B is 4, C is 5, R is 6, D is 7 and D is 7.a detailed explanation of the Industrial Production Index, please refer to the following link: http://www.federalreserve.gov//g17/IpNotes.htm.

se cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001

http://www.federalreserve.gov/releases/g17/IpNotes.htmhttp://www.federalreserve.gov/releases/g17/IpNotes.htmhttp://dx.doi.org/10.1016/j.pacfin.2014.09.001

8 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxx4. Results

4.1. Australia

Wepartition the sample in differentways to establish if the credit riskreturn return anomaly holds in thevarious sub-samples. In particular,we assess the credit riskreturn relationship by considering sub-samples offirms by classifying the firms with an investment grade rating, that is firms with a rating of BBB and above,which is equivalent of a numerical score of 11 or lower. We split the sample into small firms and largefirms based on the averagemarket value of thefirms in the sample. Lastly, in order to assesswhether the cred-it riskreturn puzzle holds in different economic states, we run themodel for the three economic states, high,low andmedium. The results for Australia are reported in Table 2. Table 2 indicates that, similar to the USmar-ket, the credit riskreturn anomaly holds in the Australian market that is, firms with better credit quality doenjoy a higher return than firmswith lower credit quality asmeasured by ratings. Column1 of Table 2 shows anegative and significant relationship between the long term rating and returns. Hence in the Australian mar-ket, as firms have an upgrade the returns tend to improve rather than the alternative where we expect that asfirms are downgraded with a higher default probability, investors should be compensated with a higher re-turn. Considering the firm characteristics, our results confirm the literature on the size effects where firmswith lower market capitalisation tend to earn a higher return than firms with higher market capitalisation.However, we do not have the liquidity factors andmomentum factors included as control variables being sig-nificant for the Australianmarket. The results of the investment grade sample are presented in column two ofthis table. Investment grade firms aremostly large firmswhich enjoy a higher credit rating and are associatedwith a larger market capitalisation and are more likely to be subject to rating changes. The credit riskreturnanomaly persists for this category of firms as well. When we split the firms between small and large, in theAustralian sample of firms the average market value of the firms was at AUD 9,441million, hence we classifyTable 2Credit risk and return analysis Australia.The following table reports the results of the cross sectional analysis in Australia. We use panel regression with correction for fixed effectin both cross section and period. The sample for Australia includes 58 firms over the sample period January 1990 to June 2012. The valuesin parenthesis are p values.

Dependent variable: return State of economy

Independent variables All firms Investmentgrade firms

Small firmsMV b 9441

Large firmsMV N 9441

High Medium Low

C 0.1856 0.196 0.2219 0.3537 0.1558 0.2814 0.22690.0000 0.0000 0.0008 0.0000 0.0859 0.0000 0.0309

LONG_TERM 0.0035 0.0043 0.0029 0.0056 0.0018 0.0059 0.008(0.0462) (0.0180) (0.3540) (0.0145) (0.6790) (0.0199) (0.0982)

LT_CHANGE 0.0056 0.0045 0.0017 0.0124 0.0088 0.0092 0.0042(0.4937) (0.5794) (0.8944) (0.2320) (0.6014) (0.4424) (0.8088)

SHORT_TERM 0.0008 0.0006 0.003 0.0026 0.0052 0.0009 0.0021(0.4896) (0.6371) (0.0897) (0.3374) (0.1764) (0.6091) (0.7255)

ST_CHANGE 0.0106 0.0136 0.0071 0.0128 0.0026 0.0089 0.0256(0.1194) (0.0650) (0.4461) (0.1923) (0.9416) (0.2197) (0.4483)

OUTLOOK 0.0077 0.0082 0.0066 0.0085 0.0034 0.0071 0.0163(0.0002) (0.0001) (0.0339) (0.0024) (0.4466) (0.0125) (0.0032)

LOGMV(2) 0.0346 0.0356 0.0487 0.0624 0.0281 0.0518 0.0511(0.0000) (0.0000) (0.0002) (0.0000) (0.0159) (0.0000) (0.0307)

LOGBVMV(2) 0.0068 0.0042 0.0063 0.0278 0.0429 0.0068 0.0219(0.4018) (0.6046) (0.5613) (0.0694) (0.0341) (0.5910) (0.3950)

LOG_TURNOVER(2) 0.0084 0.0094 0.0092 0.0149 0.0131 0.0137 0.0212(0.1539) (0.1218) (0.2194) (0.2929) (0.2996) (0.1091) (0.2744)

CUMRETURN(2) 0.0060 0.0062 0.0105 0.0062 0.0086 0.0015 0.0476(0.3936) (0.3801) (0.2787) (0.5864) (0.5658) (0.8833) (0.0012)

LOG_PE(2) 0.0057 0.0068 0.0026 0.0096 0.0088 0.0142 0.0170(0.2054) (0.1528) (0.6307) (0.2690) (0.4612) (0.0226) (0.0703)

R-squared 32.14% 32.47% 33.61% 44.27% 25.34% 33.28% 41.74%No of observations 4750 4634 2975 1775 1137 2463 1150



9E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxfirms below thismean to be small andfirms above thismean to be large. In contrast towhat is suggested in theliterature where the credit riskreturn puzzle is mainly for small firms (see for example, Dichev andPiotroski's (2001)), the results obtained for the Australian sample of small firms suggests that we do nothave the credit riskreturn anomaly.

Avramov et al. (2009) argue that the impact of rating downgrades cannot be solely attributed to a partic-ular state of economy. They consider the NBER definition of recession and expansion. However, this methodclassifies the direction of the economy rather than the level of activity in the economy and hence we use theapproach suggested by McQueeen and Roley (1993). Our results in the Australian market do not suggest thesame finding. Analysing the results of the different states of economy, when themarket is on a high economicstate, we do not find evidence of any credit riskreturn anomaly. The credit riskreturn anomaly is significantwhen the state of the economy ismediumandweakly significant during a low economic state. This is not verysurprising as themarket will expect to have firms at a lower credit quality when the economy is at a low leveland hence this could lead tomispricing in themarket.We equally find that in theAustralianmarket, one of theindependent variables which is important in explaining the distribution of returns is outlook changes. Theoutlook of the firms is significant across all the different sub-samples that are being considered. Hence ourkey results from Table 2 indicate that similar to the results obtained by Avramov et al. (2009), overall thecredit riskreturn anomaly does exist in the Australian market.

Avramov et al. (2009) argue that the evidence highlights mispricing by retail investors sustained byilliquidity and short-selling constraints. Whilst we establish the same credit riskreturn relationship, theAustralian market is different from the US market. We hence analyse what are the possible reason(s) forthis anomaly in the Australian market. As highlighted in the previous section, the key characteristics of theratings data of the Australian firms is that the long term rating changes are quite significant. Analysing thedistribution of long term rating changes, we have amean of 7. This implies that the rating downgrades shouldhave a significant impact on the returns. We, hence analyse the credit riskreturn anomaly by considering anTable 3Credit risk and return analysis Australia excluding downgrade announcements.The following table reports the results of the cross sectional analysis in Australia excluding returns around the downgrades that is thesample of firmswith upgrades and firmswhich has had no change in ratings. High implies high economic state,medium impliesmediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent variable: return Downgrades excluded

Independent variables All Firms Investmentgrade


Large FirmsMV N 9441

High Medium Low

C 0.1702 0.18 0.2343 0.3166 0.1471 0.2447 0.2291(0.0000) (0.0000) (0.0004) (0.0000) (0.1070) (0.0000) (0.0303)

LONG_TERM 0.0027 0.0034 0.0033 0.0035 0.0016 0.0041 0.0079(0.1255) (0.0614) (0.2969) (0.1336) (0.7049) (0.1116) (0.1059)

LT_CHANGE 0.0057 0.0058 0.0016 0.0074 0.0082 0.0001 0.0192(0.6947) (0.6851) (0.9436) (0.6862) (0.7859) (0.9950) (0.5598)

SHORT_TERM 0.0009 0.0006 0.0033 0.0027 0.0051 0.0010 0.0025(0.4661) (0.6179) (0.0618) (0.3148) (0.1908) (0.5510) (0.6761)

ST_CHANGE 0.0080 0.0107 0.0064 0.0075 0.0062 0.0050 0.0671(0.2555) (0.1609) (0.5106) (0.4506) (0.9115) (0.4946) (0.1156)

OUTLOOK 0.007 0.0075 0.0063 0.0066 0.0028 0.0057 0.0165(0.0007) (0.0003) (0.0446) (0.0169) (0.5365) (0.0417) (0.0031)

LOGMV(2) 0.0322 0.0331 0.0512 0.0589 0.0262 0.0457 0.0514(0.0000) (0.0000) (0.0001) (0.0000) (0.1930) (0.0000) (0.0311)

LOGBVMV(2) 0.0060 0.0035 0.0068 0.0219 0.0489 0.0099 0.0244(0.4590) (0.6639) (0.5339) (0.1526) (0.0171) (0.4305) (0.3449)

LOG_TURNOVER(2) 0.0075 0.0086 0.0089 0.0068 0.0114 0.0129 0.0205(0.2020) (0.1548) (0.2365) (0.6268) (0.3657) (0.1278) (0.2959)

CUMRETURN(2) 0.0069 0.0072 0.0089 0.0026 0.0071 0.0015 0.0472(0.3294) (0.3107) (0.3630) (0.8166) (0.6385) (0.8797) (0.0015)

LOG_PE(2) 0.0061 0.0072 0.0028 0.0109 0.0086 0.0151 0.0174(0.1690) (0.1273) (0.6131) (0.2036) (0.4747) (0.0145) (0.0658)




10 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxupgrade versus downgrade analysis. We undertake the analysis by excluding the returns around the down-grade announcements and by considering a sample of firms where we have an upgrade or firms with nolong term rating changes. The results are reported in Table 3. Comparing the results of Tables 2 and 3, it isevident that once we remove the effects of downgrade announcements, we no longer have the creditriskreturn anomaly in the Australian market. The other control variables are more or less the same withthe outlook and size effect being the same. Across all the sub-samples in Table 3, we do not establish any cred-it riskreturn anomaly. Hence,we argue that the key driver of the credit riskreturn anomaly in theAustralianmarket are rating announcements, in particular, downgrades do have a significant impact on the distributionof stock market returns. The results obtained are in fact consistent with the literature on the impact of ratingchanges on the Australian stock market, as highlighted in the previous section. The studies undertaken byMatolcsy and Lianto (1995), Creighton et al. (2007) and Chan et al. (2009) establish the impact and impor-tance of rating announcement on the stock or bond market in Australia. In our analysis, we contribute tothis literature by arguing that given the magnitude of rating downgrades is substantial, the key driver as towhy we have a credit riskreturn anomaly in the Australian market, is mainly because of the rating down-grades. Hence, we confirm the impact of rating change announcements on the stock market returns.

In order to further provide evidence on this explanation, we repeat this analysis, to exclude the firm withupgrade announcement to ensure that the credit riskreturn anomaly is explained by the impact of down-grade announcements. Hence, we run the model using a sample of firms with downgrade announcementsand firms with no changes that is we exclude the observations where we have upgrade announcements.The results are reported in Table 4. The results in this table are quite similar to the results reported inTable 2, in that we find a credit riskreturn anomaly for the different sub-samples that we consider. Theseresults hence further highlight that the credit riskreturn anomaly is in fact explained by the announcementsof downgrades. Comparing Tables 3 and 4, that is, the sample with upgrade announcements included andsample with downgrade announcements included, we have the credit riskreturn anomaly only when weTable 4Credit risk and return analysis Australia excluding upgrade announcements.The following table reports the results of the cross sectional analysis in Australia excluding returns around the upgrades that is the sampleof firms with downgrades and firms which has had no change in ratings. High implies high economic state, medium implies mediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent variable: return Upgrades excluded

Independent variables All Firms Investmentgrade Firms

Small FirmsMV b 9441


High Medium Low

C 0.1855 0.192337 0.21903 0.349453 0.1497 0.2750 0.2268(0.0000) (0.0000) (0.0010) (0.0000) (0.1007) (0.0000) (0.0315)

LONG_TERM 0.0035 0.0041 0.0028 0.0053 0.0014 0.0057 0.008(0.0434) (0.0237) (0.3783) (0.0225) (0.7461) (0.0255) (0.0978)

LT_CHANGE 0.0097 0.0016 0.0211 0.0160 0.0145 0.0033 (0.3298) (0.9166) (0.0899) (0.4398) (0.3267) (0.8724)

SHORT_TERM 0.0008 0.0006 0.0031 0.0027 0.0051 0.0008 0.0020(0.5149) (0.6506) (0.0905) (0.3225) (0.1878) (0.6492) (0.7336)

ST_CHANGE 0.0131 0.015233 0.0080 0.0149 0.0062 0.0097 0.0398(0.0546) (0.0440) (0.4004) (0.1369) (0.8963) (0.1854) (0.2858)

OUTLOOK 0.0077 0.0081 0.0064 0.0083 0.0030 0.0069 0.0164(0.0002) (0.0001) (0.0392) (0.0029) (0.4971) (0.0152) (0.0031)

LOGMV(2) 0.0346 0.035 0.0481 0.0627 0.0279 0.0506 0.0512(0.0000) (0.0000) (0.0002) (0.0000) (0.1652) (0.0000) (0.0307)

LOGBVMV(2) 0.0070 0.0046 0.0066 0.0295 0.043091 0.0072 0.0224(0.3914) (0.5743) (0.5465) (0.0558) (0.0341) (0.5681) (0.3854)

LOG_TURNOVER(2) 0.0084 0.0091 0.0093 0.0122 0.0125 0.0136 0.0221(0.1547) (0.1331) (0.2151) (0.3943) (0.3229) (0.1135) (0.2551)

CUMRETURN(2) 0.0054 0.0060 0.0105 0.0075 0.0090 0.0021 0.0476(0.4416) (0.4049) (0.2802) (0.5117) (0.5497) (0.8397) (0.0013)

LOG_PE(2) 0.0054 0.0065 0.0026 0.0083 0.0103 0.0144 0.0168(0.2275) (0.1724) (0.6315) (0.3428) (0.3954) (0.0213) (0.0755)




11E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxinclude downgrades. The control variables are similar given we have outlook as well as the size variableswhich are significant. We contribute to the literature by providing an explanation for the credit riskreturnpuzzle and our results are equally in line with the impact of rating changes on stock markets. Our resultssuggest that downgrade announcements have an impact on the market whilst upgrades do not carry thesame informational value, which is similar to the US, see for example, Goh and Ederington (1993); Handet al. (1992) and Australia, see Matolcsy and Lianto (1995).

As a further robustness check, we run the analysis across firms that have no change in their rating an-nouncements that is we exclude firms with both rating upgrades and downgrades. If our view that the creditriskreturn puzzle in Australia is in fact explained by the downgrade announcements, we should expect thatas we run the model with firmwith no rating changes, we should not establish any credit riskreturn puzzle.The results for the sample of firms with no rating changes for Australia are reported in Table 5. The results inthis table clearly support our argument that in fact the credit riskreturn puzzle holds only whenwe have thefirms with the downgrade announcement in the sample. The results in this table are similar to Table 3, whichis the sample of upgrades and firms with no change where we have significant results only in the sub-sampleof investment grade firms.

4.1.1. State of economy analysis AustraliaOne of the key contributions of this study is to assess the credit riskreturn puzzle across various economic

states. The initial analysis is undertaken by followingMcQueeen and Roley (1993). As highlighted previously,we consider some alternativemeasures of classifying the state of the economy. The first alternativemeasure isto test if the credit riskreturn puzzle holds during a crisis or non-crisis period. There is a wide literaturewhich explores the dates to be considered as significant crises, see for example, Dungey et al. (2010) andFry et al. (2011). In our analysis, we consider the definitions provided by Fry et al. (2011) and Yiu et al.(2010) and account for the six crises in the model: Asian, Russian, LTCM, Brazilian, Dot-com, ArgentineanTable 5Credit risk and return analysis Australia excluding both upgrade and downgrade announcements.The following table reports the results of the cross sectional analysis in Australia excluding returns around the upgrades as well as thedowngrades that is, the sample of firms which has had no change in ratings. High implies high economic state, medium implies mediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent Variable: Return Sample with no rating change

Independent variables All Firms Investmentgrade Firms



High Medium Low

C 0.1674 0.1776 0.2314 0.3201 0.1491 0.1183 0.2330(0.0000) (0.0000) (0.0005) (0.0000) (0.1034) (0.0203) (0.0278)

LONG_TERM 0.0026 0.0033 0.0032 0.0033 0.0021 0.0018 0.0079(0.1460) (0.0719) (0.3179) (0.1532) (0.6272) (0.4756) (0.1052)

LT_CHANGE

SHORT_TERM 0.0009 0.0006 0.0034 0.0028 0.0045 0.0017 0.0026(0.4721) (0.6287) (0.0626) (0.2934) (0.2427) (0.3325) (0.6574)

ST_CHANGE 0.0094 0.0124 0.0074 0.0098 0.1199 0.1122(0.1911) (0.1126) (0.4563) (0.3349) (0.0138) (0.0266)

OUTLOOK 0.007 0.0075 0.0062 0.0065 0.0030 0.0089 0.0166(0.0008) (0.0003) (0.0510) (0.0193) (0.5086) (0.0059) (0.0030)

LOGMV(2) 0.0318 0.0328 0.0507 0.0606 0.0266 0.0522 0.0522(0.0000) (0.0000) (0.0001) (0.0000) (0.1869) (0.0134) (0.0288)

LOGBVMV(2) 0.0063 0.0038 0.0071 0.0231 0.0491 0.0158 0.0248(0.4391) (0.6382) (0.5180) (0.1331) (0.0170) (0.1526) (0.3376)

LOGTURNOVER(2) 0.0074 0.0085 0.0090 0.0049 0.0102 0.0018 0.0210(0.2106) (0.1613) (0.2314) (0.7302) (0.4213) (0.8372) (0.2848)

CUMRETURN(2) 0.0065 0.0068 0.0089 0.0041 0.0077 0.0225 0.0473(0.3582) (0.3416) (0.3636) (0.7147) (0.6103) (0.0268) (0.0015)

LOG_PE(2) 0.0059 0.0070 0.0028 0.0101 0.0085 0.0077 0.0167(0.1846) (0.1418) (0.6137) (0.2415) (0.4815) (0.2650) (0.0760)

R-squared 31.80% 32.12% 33.38% 43.72% 25.34% 32.98% 41.63%No Of observations 4688 4573 2942 1746 1121 2253 1133



Table 6Crisis period dates.The following table summarises the start and end dates of the crisis periods that have been examined in the paper. The data sources arealso provided.

Crisis Start Date End Date Source

Asia 2-Jul-97 31-Dec-98 Yiu et at. (2010)Russia 17-Aug-98 31-Dec-98 Fry et al. (2011)LTCM 23-Sep-98 15-Oct-98 Fry et al. (2011)Brazil 7-Jan-99 25-Feb-99 Fry et al. (2011)Dot.Com 28-Feb-00 7-Jun-00 Fry et al. (2011)Argentina 11-Oct-01 3-Mar-05 Fry et al. (2011)GFC 26-Jul-07 31-Dec-10 Fry et al. (2011)

12 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxand the GFC. A summary of the dates is presented in Table 6. The next measure of the states of the economyconsidered is the market returns based on quintiles over the sample period, with the top two quintiles beingthehigh state, themedium state as the third quintile and thebottom twoquintiles as the low state.We run themodel using these sub samples and the results for Australia are reported in Table 7.

The first two columns of Table 7 report the results of the sample during the crisis and non-crisis phase. Thecredit riskreturn puzzle holds only in a non-crisis phase. This particular result on the crisis and non-crisisphase further strengthens the motivation of this paper. We consider Australia in this study as it is suggestedin the literature that the Australianmarket reacted differently to crisis, in particular the global financial crisis.Brown and Davis (2010) suggest that the March quarter 2009 GDP growth of 0.4% suggested that Australiahad largely escaped the world-wide recession. Australia's banking sector was also less adversely affectedthan elsewhere, with no failures and profitability remaining strong, although down somewhat from previouslevels andwith increased bad debt levels. Hence, we equally contribute to the literature assessing the impactTable 7Credit risk and return analysis Australia alternative measures of economic states.The following table reports the results of the cross sectional analysis in Australia. The sample for Australia includes 58 firms over thesample period January 1990 to June 2012. The values in parenthesis are p values. The first two columns report the results of the sampleduring crisis and non-crisis phase. The next three columns reports the results of the economic state classified using the market returns.

Dependent variable: return Market return

Independent variables Crisis period No Crisis Period High Medium Low

C 0.2219 0.1446 0.1508 0.2487 0.2425(0.0002) (0.0013) (0.0055) (0.0001) (0.0002)

LONG_TERM 0.0021 0.0047 0.0065 0.0028 0.0024(0.4607) (0.0327) (0.0123) (0.3588) (0.4347)

LT_CHANGE 0.0041 0.0031 0.0173 0.0394 0.0417(0.7613) (0.7267) (0.1291) (0.0184) (0.0150)

SHORT_TERM 0.0001 0.0001 0.0036 0.0049 0.0049(0.9537) (0.9374) (0.0355) (0.0206) (0.0243)

ST_CHANGE 0.0188 0.0010 0.0063 0.0131 0.0141(0.0626) (0.9085) (0.4576) (0.2689) (0.2388)

OUTLOOK 0.0097 0.0031 0.0064 0.0095 0.0098(0.0021) (0.2654) (0.0286) (0.0109) (0.0094)

LOGMV(2) 0.0451 0.0228 0.0188 0.0551 0.0541(0.0002) (0.0120) (0.0844) (0.0000) (0.0000)

LOGBVMV(2) 0.0270 0.0076 0.0227 0.0055 0.0075(0.0487) (0.4408) (0.0563) (0.7003) (0.6079)

LOG_TURNOVER(2) 0.0088 0.0126 0.0080 0.0190 0.0184(0.3751) (0.0656) (0.3540) (0.0702) (0.0867)

CUMRETURN(2) 0.0120 0.0053 0.0201 0.0092 0.0107(0.2436) (0.5866) (0.0345) (0.4762) (0.4150)

LOG_PE(2) 0.0077 0.0030 0.0042 0.0116 0.0138(0.2458) (0.6449) (0.4863) (0.1420) (0.0936)

R-squared 32.97% 33.12% 21.25% 25.73% 25.61%No of observations 2513 2237 1921 1923 1875



13E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxof the global financial crisis on Australia and highlight that the market did not seem to react to crisis periodsand hence we conclude that the credit riskreturn puzzle does not hold in the crisis period in the Australiancontext. We also consider the market returns to classify the states of economy as high, medium and low. Theresults are reported in the last three columns of Table 7. The results indicate that in contrast to the results ob-tained by using the log of Industrial Production Index, we do not have the credit riskreturn anomaly whenwe use the market returns as a benchmark to classify the states of economy in the medium and low sates.We do have the puzzle in the high state of economy. For the medium and low economic states, we havethe puzzle only if we consider the long term rating change.Whilst the results suggest that most possibly, clas-sifying the states of economy provides different results with regard to the credit riskreturn puzzle, anotherpossible explanation is that the composition of the market returns in Australia is different from the marketreturns of other developed markets, in particular the US. The Australian market index is concentrated in thebanking (mainly the big 4 banks), and the resources sector in contrast to othermarkets where the index com-position is more diversified. The concentration level is well known to investors in that 44% of the total marketcapitalisation and 49% of the risk is from the 4 banks, 1 mining and 1 utility firm, see, Engel (2014) and hencethis possibly explains the difference in the results obtained for the credit riskreturn puzzle using the In-dustrial Production Index and the market returns.

4.2. Japan results

Yonezawea and Miyake (1998) argue that the Japanese stock market is characterized by two prominentfeatures. First, stock prices have been extremely volatile over the past ten years. Second, the market is dom-inated by cross-shareholdings and stagnant individual stock ownership and they highlight several similaritiesbetween the USmarket and the Japanesemarket. However a key difference between the Japanese and the USmarket is that the US is a market based system and that of Japan is a bank based system. Hence because Japanhas one of the world's largest financial centres and some unique features, we assess the credit riskreturnTable 8Credit risk and return analysis Japan.The following table reports the results of the cross sectional analysis in Japan. The sample for Japan includes 245 firms and hence we havean unbalanced panel of 245 270 observations (January 1990 to June 2012). The values in parenthesis are p values.

Dependent variable: return State of economy




High Medium Low

C 0.2705 0.2725 0.8227 0.1766 0.6709 0.138238 0.076453(0.0008) (0.0010) (0.0000) (0.1356) (0.0001) (0.2238) (0.6709)

LONG_TERM 0.0023 0.002 0.0019 0.0002 0.0049 0.00148 0.0047(0.0502) (0.1074) (0.3202) (0.9022) (0.0774) (0.3915) (0.0751)

LT_CHANGE 0.0177 0.0223 0.0054 0.0082 0.0013 0.0091 0.02711(0.0013) (0.0002) (0.5051) (0.3325) (0.8949) (0.3437) (0.0071)

SHORT_TERM 0.0001 0.0001 0.0003 0.0022 0.0016 0.0000934 0.0030(0.9314) (0.9087) (0.8169) (0.6378) (0.7863) (0.9510) (0.1905)

ST_CHANGE 0.0013 0.0020 0.0018 0.0086 0.0091 0.0042 0.0018(0.8281) (0.7457) (0.7778) (0.6161) (0.6270) (0.6897) (0.8377)

OUTLOOK 0.0057 0.0053 0.0027 0.0056 0.0013 0.0033 0.0136(0.0008) (0.0021) (0.3013) (0.0151) (0.6674) (0.2336) (0.0002)

LOGMV(2) 0.0398 0.0403 0.1478 0.0201 0.0963 0.0193 0.0104(0.0023) (0.0027) (0.0000) (0.2573) (0.0003) (0.2961) (0.7255)

LOGBVMV(2) 0.0788 0.0801 0.0131 0.1032 0.07235 0.0731 0.1232(0.0000) (0.0000) (0.5093) (0.0000) (0.0049) (0.0001) (0.0000)

LOGTURNOVER(2) 0.0084 0.00918 0.0038 0.0191 0.0386 0.0096 0.0028(0.0101) (0.0051) (0.4312) (0.0000) (0.0000) (0.0324) (0.6798)

CUMRETURN(2) 0.0065 0.0048 0.0126 0.0118 0.0073 0.0069 0.0463(0.2284) (0.3717) (0.1055) (0.1385) (0.4621) (0.4103) (0.0000)

LOG_PE(2) 0.0001 0.0004 0.0003 0.0033 0.0052 0.0044 0.0075(0.9752) (0.8890) (0.9433) (0.4400) (0.4440) (0.3251) (0.2423)

R-squared 32.57% 32.61% 37.83% 40.34% 37.68% 38.57% 37.17%No of Observations 8619 8552 4325 4294 2900 3579 2190



Table 9Credit risk and return analysis Japan excluding downgrade announcements.The following table reports the results of the cross sectional analysis in Japan excluding returns around the downgrades that is the sampleof firms with upgrades and firms which has had no change in ratings. High implies high economic state, medium implies mediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent variable: return Downgrades excluded

Independent variables All Firms Investmentgrade firms



High Medium Low

C 0.2376 0.2245 0.8247 0.1201 0.6141 0.1198 0.0844(0.0027) (0.0041) (0.0000) (0.2871) (0.0002) (0.2941) (0.6200)

LONG_TERM 0.0015 0.0007 0.0055 0.0007 0.0041 0.0016 0.0039(0.2230) (0.5667) (0.0020) (0.6777) (0.1415) (0.3587) (0.1173)

LT_CHANGE 0.0136 0.0125 0.0098 0.0048 0.0083 0.0088 0.0464(0.1193) (0.1443) (0.5027) (0.6808) (0.4263) (0.5983) (0.4072)

SHORT_TERM 0.0001 0.0001 0.0006 0.0025 0.0012 0.0002 0.0031(0.9185) (0.9350) (0.6242) (0.5722) (0.8401) (0.9027) (0.1482)

ST_CHANGE 0.0016 0.0005 0.0026 0.0209 0.0137 0.0063 0.0026(0.7954) (0.9339) (0.6729) (0.3070) (0.4815) (0.5669) (0.7573)

OUTLOOK 0.0049 0.004 0.0018 0.0063 0.0019 0.0035 0.0103(0.0036) (0.0135) (0.4451) (0.0044) (0.5220) (0.2128) (0.0030)

LOGMV(2) 0.0353 0.0339 0.1459 0.0121 0.0885 0.0163 0.0130(0.0059) (0.0077) (0.0000) (0.4729) (0.0009) (0.3799) (0.6422)

LOGBVMV(2) 0.0775 0.0769 0.0132 0.1058 0.0798 0.0768 0.109(0.0000) (0.0000) (0.4804) (0.0000) (0.0019) (0.0001) (0.0001)

LOGTURNOVER(2) 0.0079 0.008 0.0001 0.0185 0.0352 0.0096 0.0028(0.0128) (0.0086) (0.9831) (0.0000) (0.0002) (0.0327) (0.6601)

CUMRETURN(2) 0.0068 0.0059 0.0075 0.0137 0.0109 0.0074 0.0487(0.1961) (0.2573) (0.3155) (0.0754) (0.2717) (0.3754) (0.0000)

LOG_PE(2) 0.0002 0.0005 0.0015 0.0025 0.0037 0.0042 0.0075(0.9384) (0.8588) (0.7222) (0.5359) (0.5866) (0.3494) (0.2112)


14 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxpuzzle in the Japanesemarket. There is a wide literature undertaking empirical analysis on the Japanese stockmarket, see for example Gunaratne and Yonesawa (1997), Iihara et al. (2004) and Chou and Chung (2007).Whilst there are studies addressing other market anomalies, for example, Sasaki and Miyazaki (2012) analy-ses the relationship between credit rating and the contrarian strategy, the credit riskreturn anomaly does notseem to have been examined in the literature for Japan. Hencewe contribute to the literature on the Japanesemarkets by investigating the credit riskreturn puzzle in the Japanese market.

We have a final sample of 245 firms that are included in the analysis.We initially run ourmodel by includ-ing all the rating announcements and different sub-samples, including investment grade firms, small firmsand large firms,5 in additionwe consider themodels in the different state of economy. The results are reportedin Table 8. The results of Table 8 indicate that we do have the credit riskreturn anomaly in the Japanesemar-ket. The credit riskreturn relationship exists in the all firms sample aswell as the sample of investment gradefirms. The long term rating has a negative coefficient indicating that as firms enjoy better credit quality, theyequally enjoy better returns. However, the split between the firm size does not reveal that there is a creditriskreturn anomaly. Considering the last three columns of this table, whenwe analyse the credit riskreturnrelationship in different economic states, there is aweak evidence that the Japanesemarket tends to have thisanomaly in extreme conditions that iswhen the state of the economy is high andwhen the state of economy islow, where onewill expect that there will be more rating changes and the stock returnswill be more volatile.The control variables for size, liquidity, value and momentum seem to be consistent with previous findingsdocumented in the literature. Hence similar to the Australian market, our results in the Japanese marketcomplement the research of Avramov et al. (2009) for the USmarket that is low credit risk firms enjoy higherreturns than high credit risk firms.5 The averagemarket value of thefirms in the sample is JPY 917,533million.Hence smallfirms are less than this average value and largefirms are above this average value.



15E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxIn the next segment of the analysis,we explore the reasons for the credit riskreturn anomaly.We split oursample and analyse the firms associated with upgrade announcements and downgrade announcements.Table 9 reports the results of the sample of firms with upgrade announcements and firms with no ratingchanges that is we exclude the observations which have a downgrade announcement. The long term ratingchange for Japanese firms seems to be quite large, which is similar to the Australianmarkets. We have down-grades of up to 6 notches; however it should be noted that the frequency of downgrades is lower than theAustralian market. The upgrade announcements happen mostly up to a maximum of 2 notches. The resultsreported in Table 9 are similar to the results obtained for the Australianmarket, that is aswe remove the effectof rating downgrade announcements from the sample; we no longer have the credit riskreturn anomaly inthe Japanese market. This confirms our previous analysis and leads to the conclusion that the leading factorwhich contributes to the credit riskreturn anomaly is the credit rating agencies downgrade announcements.Our results are equally in line with the established literature on the impact of rating changes on the Japanesestockmarket, see for example, Li et al. (2006). It should be highlighted that the credit riskreturn anomaly ap-pears in the sample of small firms whenwe exclude the downgrade announcements. In fact, this is somehowexpected given that only those firms where the credit rating is low will be expected to be upgraded in thelonger term. The control variables included in the model indicate similar results to those obtained for thefull sample of firms' results in Table 8.

Next, we run the analysis by excluding the upgrade announcements that is we run our model with thesample of firms with downgrade announcements and firms with no rating announcements to confirm ourconclusion from Table 9. The results of the model excluding the upgrades are reported in Table 10. The con-clusion that can be drawn when analysing Table 10 is that once we have the downgrade announcements inthe sample, we do have the credit riskreturn puzzle in the Japanese market including the sub-sample ofinvestment grade as well as large firms, which will be the firms likely to be subject to downgrades as wellin the low state of economy where the market confidence will be affected and the expectation will be inTable 10Credit risk and return analysis Japan excluding upgrade announcements.The following table reports the results of the cross sectional analysis in Japan excluding returns around the upgrades that is the sample offirms with downgrades and firms which has had no change in ratings. High implies high economic state, medium implies mediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent variable: return Upgrades excluded

Independent variables All Firms Investmentgrade firms



High Medium Low

C 0.2656 0.2700 0.8067 0.1504 0.6788 0.1373 0.0753(0.0007) (0.0008) (0.0000) (0.1834) (0.0001) (0.2281) (0.6759)

LONG_TERM 0.0022 0.0018 0.0020 0.0004 0.0045 0.0016 0.0047(0.0581) (0.0121) (0.2786) (0.0792) (0.1127) (0.3581) (0.0756)

LT_CHANGE 0.021 0.0302 0.0014 0.0172 0.0411 0.0114 0.0265(0.0013) (0.0001) (0.8830) (0.1261) (0.1173) (0.3095) (0.0090)

SHORT_TERM 0.0002 0.0000 0.0003 0.0027 0.0008 0.0000 0.0030(0.8457) (0.9959) (0.8162) (0.5394) (0.8991) (0.9843) (0.1930)

ST_CHANGE 0.0003 0.0013 0.0018 0.0267 0.1238 0.0107 0.0022(0.9656) (0.8324) (0.7757) (0.3369) (0.0718) (0.3391) (0.8076)

OUTLOOK 0.0054 0.0049 0.0027 0.0058 0.0022 0.0032 0.0136(0.0011) (0.0030) (0.2736) (0.0085) (0.4541) (0.2500) (0.0002)

LOGMV(2) 0.0395 0.0403 0.1452 0.0167 0.0981 0.0191 0.0102(0.0018) (0.0020) (0.0000) (0.3243) (0.0003) (0.3021) (0.7306)

LOGBVMV(2) 0.0766 0.0775 0.0109 0.1028 0.0744 0.0744 0.1233***(0.0000) (0.0000) (0.5751) (0.0000) (0.0040) (0.0001) (0.0000)

LOGTURNOVER(2) 0.0077 0.0084 0.0036 0.0186 0.0394 0.0099 0.0029(0.0143) (0.0075) (0.4360) (0.0000) (0.0000) (0.0282) (0.6762)

CUMRETURN(2) 0.0060 0.0045 0.0131 0.0132 0.0114 0.0083 0.0463(0.2580) (0.4020) (0.0891) (0.0884) (0.2511) (0.3193) (0.0000)

LOG_PE(2) 0.0000 0.0002 0.0001 0.0028 0.0041 0.0044 0.0074(0.9891) (0.9359) (0.9830) (0.4878) (0.5440) (0.3226) (0.2453)




16 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxgeneral that firms have a low credit quality. The control variables are similar to the results obtained in Table 8,where on average we do have a size, liquidity, and value and momentum effect in the Japanese market.

Similar to the Australian analysis, as robustness check to confirm our explanation of the credit riskreturnpuzzle in the Japanese market, we run the model with the sample of firms which do not have any ratingchanges that is by excluding all upgrade and downgrade announcements. The results are reported inTable 11. The results indicate that across the various subsamples, with the exception of small firms, we donot establish the credit riskreturn puzzle in the Japanese markets. The results are similar to the results ofTable 9 that is as we exclude the downgrades, the credit riskreturn puzzle does not hold. As such thisconfirms our view, that the key explanation for the existence of the credit riskreturn puzzle in Japan is theannouncement of rating downgrades.

4.2.1. State of economy analysis JapanWe run themodel for the Japanesemarket using the alternative definitions of the states of economy, sim-

ilar to the Australian analysis, that is by considering a crisis and a non-crisis phase and using the marketreturns. The results are reported in Table 12. In contrast to the results obtained for the Australian market,the credit riskreturn puzzle holds in the sample of crisis periods rather than the non-crisis periods. The re-sults obtained for the Japanesemarket can be explained by the fact that the Japanesemarket had experienceda deep and prolonged period of banking crisis. There is a wide literature which supports the fact that firms,including non-financial firms were largely affected by the banking crisis. For example, Miyajima and Yafeh(2007), use an event study to assess the extent to which non-financial firms were affected by the crisis andthey conclude that whilst the magnitude to which were affected is different, there is a large number offirms which were largely affected, in particular small, leveraged, low-tech with low credit ratings and lowmarket to book ratio firms. Further, Hoshi and Kashyap (2001) highlight that the estimates of the magnitudeof the crisis vary, but there is no doubt that the banking crisis had a severe impact on the Japanesemarket. Assuch, with regard to the crisis and non-crisis phases, we conclude that the credit riskreturn puzzle holds inTable 11Credit risk and return analysis Japan excluding both upgrade and downgrade announcements.The following table reports the results of the cross sectional analysis in Japan excluding returns around the upgrades as well as thedowngrades that are the sample of firmswhich has had no change in ratings. High implies high economic state, medium impliesmediumeconomic state and low implies low economic state. The values in parenthesis are p values.

Dependent variable: return Sample with no rating change




High Medium Low

C 0.2346 0.2321 0.8294 0.1273 0.6331 0.1184 0.0835(0.0022) (0.0031) (0.0000) (0.2605) (0.0002) (0.3009) (0.6239)

LONG_TERM 0.0014 0.0006 0.0056 0.0008 0.0040 0.0017 0.0039(0.2382) (0.5919) (0.0021) (0.6708) (0.1507) (0.3238) (0.1198)

LT_CHANGESHORT_TERM 0.0002 0.0000 0.0006 0.0026 0.0006 0.0001 0.0031

(0.8344) (0.9911) (0.6160) (0.5586) (0.9213) (0.9374) (0.1517)ST_CHANGE 0.0036 0.0033 0.0031 0.0145 0.0022

(0.5705) (0.5986) (0.6172) (0.2228) (0.8006)OUTLOOK 0.0047 0.0041 0.0019 0.0063 0.0022 0.0034 0.0104

(0.0039) (0.0136) (0.4247) (0.0049) (0.4465) (0.2280) (0.0028)LOGMV(2) 0.0352 0.0353 0.1467 0.0134 0.0917 0.0160 0.0129

(0.0047) (0.0059) (0.0000) (0.4306) (0.0006) (0.3887) (0.6456)LOGBVMV(2) 0.0758 0.0766 0.0124 0.1049 0.0781 0.0783 0.1088

(0.0000) (0.0000) (0.5116) (0.0000) (0.0025) (0.0000) (0.0001)LOGTURNOVER(2) 0.0073 0.0081 0.0001 0.0185 0.0358 0.0099 0.0028

(0.0173) (0.0087) (0.9853) (0.0000) (0.0002) (0.0286) (0.6582)CUMRETURN(2) 0.0060 0.0048 0.0075 0.0120 0.0136 0.0089 0.0488

(0.2456) (0.3620) (0.3194) (0.1207) (0.1728) (0.2897) (0.0000)LOG_PE(2) 0.0001 0.0006 0.0015 0.0024 0.0029 0.0043 0.0075

(0.9678) (0.8431) (0.7243) (0.5482) (0.6724) (0.3413) (0.2126)R-squared 37.75% 37.83% 40.89% 46.70% 38.00% 38.84% 39.12%No of observations 8522 8458 4296 4231 2847 3532 2147



Table 12Credit risk and return analysis Japan alternative measures of economic states.The following table reports the results of the cross sectional analysis in Japan. The sample for Japan includes 245 firms over the sampleperiod January 1990 to June 2012. The values in parenthesis are p values. The first two columns report the results of the sample duringcrisis and non-crisis phases. The next three columns reports the results of the economic state classified using the market returns.

Dependent variable: return Market return

Independent variables Crisis period No crisis Period High Medium Low

C 0.4300 0.1481 0.1872 0.5176 0.5153(0.0005) (0.1711) (0.1117) (0.0001) (0.0001)

LONG_TERM 0.0037 0.0015 0.0004 0.0096 0.0097(0.0505) (0.3364) (0.8005) (0.0000) (0.0000)

LT_CHANGE 0.0014 0.0306 0.0006 0.0586 0.0585(0.8675) (0.0000) (0.9381) (0.0000) (0.0000)

SHORT_TERM 0.0012 0.0002 0.0002 0.0002 0.0004(0.5033) (0.9102) (0.9184) (0.9022) (0.8279)

ST_CHANGE 0.0003 0.0025 0.0049 0.0011 0.0009(0.9655) (0.7931) (0.4732) (0.9265) (0.9375)

OUTLOOK 0.0060 0.0059 0.0058 0.0070 0.0075(0.0173) (0.0100) (0.0200) (0.0117) (0.0066)

LOGMV(2) 0.0664 0.0187 0.0225 0.0835 0.0831(0.0008) (0.2885) (0.2365) (0.0001) (0.0001)

LOGBVMV(2) 0.0698 0.0836 0.0492 0.0655 0.0656(0.0002) (0.0000) (0.0079) (0.0016) (0.0017)

LOGTURNOVER(2) 0.0133 0.0061 0.0152 0.0019 0.0025(0.0077) (0.1459) (0.0006) (0.7274) (0.6534)

CUMRETURN(2) 0.0205 0.0104 0.0365 0.0092 0.0083(0.0051) (0.1937) (0.0000) (0.2898) (0.3458)

LOG_PE(2) 0.0030 0.0037 0.0060 0.0052 0.0059(0.4619) (0.4437) (0.1765) (0.2905) (0.2358)

R-squared 36.89% 37.75% 25.05% 31.37% 30.96%No of observations 4605 4064 3560 3426 3399

17E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxthe crisis period in Japan. The last three columns of the table report the results of the model using themarketreturns as a benchmark to classify the states of the economy. Similar to the results obtained in the Australianstudy, we conclude that using alternative measures to classify the states of economy lead to a different set ofresults. When the states of economy were classified using the log of Industrial Production Index, the creditriskreturn puzzle holds in the high and low state and when we use the market returns the puzzle seem toexist in the medium and low states of economy.

5. Summary and conclusion

We assess the credit riskreturn puzzle in two developedmarkets, namely Australia and Japan to establishwhether markets other than the US face a similar puzzle as documented in the literature, in particular, byAvramov et al. (2009). The objective of the paper is to equally assess whether this relationship is affectedby different states of economy and hence we calculate and classify the state of the economy as high, mediumand low.

Ourfindings indicate that similar to the results obtained in the literature establishing the credit riskreturnrelationship in the US market, we find that both the Australian and Japanese markets face the same creditriskreturn anomaly. We conclude that given the strength of the global markets, the stock market returnfor high and low credit quality firms do have a similar pattern in the developed markets that is, firms withhigh credit ratings seems to enjoy a higher return rather than firms with low credit rating that should haveenjoyed a higher return to compensate investors for the extra risk they are holding. As far as the states ofeconomy are concerned, we conclude that in Japan, there is a weak evidence that the credit riskreturnrelationship holds in extreme market conditions that is in the high and low states of the economy whilst inthe Australian market, this does not seem to be the case in the high state of economy. We have the creditriskreturn puzzle in the medium and low states of the economy where there are expectations that firmsmight be subject to rating changes. Further, we conclude that using different measures to classify the statePlease cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001


18 E. Bissoondoyal-Bheenick, R. Brooks / Pacific-Basin Finance Journal xxx (2014) xxxxxxof economy does matter. We equally confirm that Australia is one of the countries which has had a differentexperience to crises in contrast to other developed markets.

A key contribution of our study is that we explain why we have the credit riskreturn anomaly in thesemarkets. Our key finding is that the credit riskreturn anomaly is driven by the announcement of ratingdowngrades. Our results therefore suggest that rating downgrades do have a significant impact in thesetwo markets, whilst this is not the case for rating upgrade announcements. We equally establish whilstthere are different methods established to assess default risk and bankruptcy risk in the literature to assessthe returns behaviour, credit ratings from the rating agencies does provide a good measure of default risk,given that the downgrades are the key driver of the stock market returns behaviour.

References

Akhigbe, A., Madura, J., Whyte, A.M., 1997. Intra-industry effects of bond rating adjustments. J. Financ. Res. 20, 545561.Ali, A., Daly, K., 2010. Macroeconomic determinants of credit risk: Recent evidence from a cross country study. Int. Rev. Financ. Anal. 19,

165171.Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. J. Financ. Econ. 17, 223249.Avramov, D., Chordia, T., Jostova, G., Philipov, A., 2009. Credit ratings and the cross-section of stock returns. J. Financ. Mark. 12, 469499.Avramov, D., Chordia, T., Jostova, G., Philipov, A., 2012. World price of credit risk. Rev. Asset Pricing Stud. 2 (2), 112152.Boyd, J.H., Smith, B.D., 1998. The evolution of debt and equity markets in economic development. Economic Theory 12, 519560.Brennan, M.J., Subrahmanyam, A., 1996. Market microstructure and asset pricing: on the compensation for illiquidity in stock returns. J.

Financ. Econ. 41, 341364.Brennan, M.J., Chordia, T., Subrahmanyam, A., 1998. Alternative factor specifications, security characteristics, and the cross-section of

expected stock returns. J. Financ. Econ. 49, 345373.Brown, C., Davis, K., 2010. Australia's experience in the global financial crisis. In: Kolb, R. (Ed.), Lessons from the Financial Crisis: Insights

and Analysis from Today's Leading Minds. John Wiley & Sons, Hoboken, New Jersey.Campbell, J.Y., Hilscher, J., Szilagyi, J., 2008. In search of distress risk. J. Financ. 63, 28992939.Carhart, M., 1997. On persistence of mutual fund performance. J. Financ. 52, 5782.Chan, K.C., Chen, N., 1991. Structural and return characteristics of small and large firms. J. Financ. 46, 14671484.Chan, P.T., Edwards, V., Walter, T., 2009. The information content of Australian credit ratings: a comparison between subscription and

non-subscription-based credit rating agencies. Econ. Syst. 33, 2244.Chen, N., Roll, R., Ross, S., 1986. Economic forces and the stock market. J. Bus. 59, 383403.Chou, K.C.J. Wei, Chung, H., 2007. Sources of contrarian profits in the Japanese stock market. J. Empir. Financ. 14, 261286.Creighton, A., Gower, L., Richards, A.J., 2007. The impact of rating changes in Australian financial markets. Pac. Basin Financ. J. 15, 117.Crotty, J., 2009. Structural causes of the global financial crisis: a critical assessment of the new financial architecture. Camb. J. Econ. 33,

563580.Dichev, I.D., 1998. Is the risk of bankruptcy a systematic risk? J. Financ. 53, 11311147.Dichev, I.D., Piotroski, J.D., 2001. The long-run stock returns following bond rating changes. J. Financ. 56, 5584.Dolley, M., Hutchison, M., 2009. Transmission of the US subprime crisis to emerging markets: evidence on the decouplingrecoupling

hypothesis. J. Int. Money Financ. 28, 13311349.Dungey, M., Fry, R., Martin, V.L., Tang, C., Gonzlez-Hermosillo, B., 2010. Are financial crises alike? IMF Working Papers.Engel, O., 2014. Banks and resources are likely to deliver below market returns. The Australian. , (March 12th).Fama, E.F., French, K.R., 1992. The cross-section of expected stock returns. J. Financ. 47, 427465.Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. J. Financ. Econ. 33, 356.Fama, E., French, K., 1998. Value vs. growth: the international evidence. J. Financ. 53, 19751999.FRED, d. data: http://www.federalreserve.gov/releases/g17/IpNotes.htm.Fry, R., Hsiao, Y.L., Tang, C., 2011. Actually this time is different Centre for AppliedMacroeconomic Analysis. Working paperThe Australian

National University.Garlappi, L., Shu, T., Yan, H., 2008. Default risk, shareholder advantage, and stock returns. Rev. Financ. Stud. 21, 27432778.Goh, J.C., Ederington, L.H., 1993. Is a bond rating downgrade bad news, good news, or no new for stockholders? J. Financ. 48, 20012008.Goldstein, R., 2012. Can Structural Models of Default Explain the Credit Spread Puzzle? Federal Reserve Bank Economic Letters.Griffin, J.M., Lemmon, M.L., 2002. Book-to-market equity, distress risk, and stock returns. J. Financ. 57, 23172336.Gunaratne, J., Yonesawa, Y., 1997. Return reversals in the Tokyo stock exchange: a test of stock market overreaction. Jpn. World Econ. 9,

363384.Hand, J., Holthausen, R.W., Leftwich, R.W., 1992. The effect of bond rating agency announcements on bond and stock prices. J. Financ. 47,

733752.Holthausen, R.W., Leftwich, R.W., 1986. The effect of bond rating changes on common stock prices. J. Bus. Res. 24, 5789.Hoshi, T., Kashyap, A.K., 2001. Corporate Financing and Governance in Japan: The Road to the Future. MIT Press, Cambridge and London.Iihara, Y., Kato, H., Tokunaga, T., 2004. The winnerloser effect in Japanese stock return. Jpn. World Econ. 16, 471485.Jegadeesh, N., Titman, S., 1993. Returns to buyingwinners and selling losers: implications for stockmarket efficiency. J. Financ. 48, 6592.Levine, R., 2002. Bank based or market based financial system: which is better? J. Financ. Intermed. 11, 130.Li, J., Shin, Y.S., Moore, W.T., 2006. Reactions of Japanese markets to changes in credit ratings by global and local agencies. J. Bank. Financ.

30, 10071021.Lo, A.W., Mackinlay, M.C., 1990. Data snooping biases in test of financial asset pricing models. Rev. Financ. Stud. 3, 431468.Matolcsy, Z.P., Lianto, T., 1995. The incremental information content of bond rating revisions: the Australian evidence. J. Bank. Financ. 19,

891902.McQueeen, G., Roley, V., 1993. Stock prices, news and business conditions. Rev. Financ. Stud. 6 (3), 683707.Miyajima, H., Yafeh, Y., 2007. Japan's banking crisis: an event study perspective. J. Bank. Financ. 31, 28662885.Please cite this article as: Bissoondoyal-Bheenick, E., Brooks, R., The credit riskreturn puzzle: Impact ofcredit rating ..., Pacific-Basin Finance Journal (2014), http://dx.doi.org/10.1016/j.pacfin.2014.09.001

http://refhub.elsevier.com/S0927-538X(14)00079-1/rf0010http://refhub.elsevier.com/S0927-538X(14)00079-1/rf0015http://refhub.elsevier

Brooks_2014_The Credit Risk–Return Puzzle_Impact of Credit Rating Announcements in Australia and...

Documents

Transcript of Brooks_2014_The Credit Risk–Return Puzzle_Impact of Credit Rating Announcements in Australia and...