The Cost of Equity in Emerging Markets: A Downside Risk Approach JAVIER ESTRADA JAVIER ESTRADA very company evaluating an invest- most appropriate measure of risk. Few, how- is an associate ment project or an acquisition in an ever, call for discarding the CAPM outright.1 professor in the emerging market must estimate not The bulk of the evidence seems to indi- Department of Finance at the only future cash flows but also an cate that, although additional factors such as Eappropriate discount rate. Yet, there is no size and book-to-market ratios may be nec- IESE business school in Barcelona, Spain. widely accepted definition of risk in emerging essary to properly explain stock returns, beta markets, and therefore no standard way to esti- should not be discarded as a measure of risk. mate discount rates. This issue is critical for In other words, in developed markets, the companies investing in emerging markets. debate is not so much about whether beta is The aim of this article is to estimate a an appropriate measure of risk as it is about capital asset pricing model (CAPM)-based cost determining what additional variables (if any) of equity for a sample of 28 emerging markets, affect stock returns. and compare it to two alternative estimates In emerging markets, the debate is quite based on total risk and downside risk. The different. The international version of the downside risk methodology has several attrac- CAPM implicitly assumes fully integrated tive features. It is theoretically sound; it is very markets, thus implying that assets with the simple to implement; it can be applied both at same risk must have the same expected return the market level and at the company level; it is regardless of where they trade.2 Most of the not based on subjective measures of risk; it can evidence on the integration of emerging mar- be tailored to any desired benchmark return; kets is at odds with this assumption.3 and it captures the downsideDraft risk that investors The relevant question is, then, what want to avoid (as opposed to the upside risk to model should be used to estimate the cost of which investors want to be exposed). equity in these markets? WHAT’S AT STAKE Previous Approaches Estimating the cost of equity is very dif- Use of the CAPM to estimate the cost ferent in developed markets and emerging of equity in emerging markets has several markets. In developed markets, practitioners problems. From an empirical point of view, widely use a CAPM-based required return these problems are compounded by the fact on equity. This method, of course, is not free that, in emerging markets, betas and stock from controversy; over 30 years of academic returns are largely uncorrelated. Furthermore, debate have not settled whether beta is the early studies on the cost of equity in emerg- FALL 2000 EMERGING MARKETS QUARTERLY 1 ing markets, such as Harvey [1995], find that these mar- A Simple Approach kets have very low betas that, when used as an input in the CAPM equation, generate required returns typically The framework proposed in this article is very sim- considered “too low.” As a result, several alternative ways ple, and companies can apply it just as easily as the CAPM. of estimating the cost of equity in emerging markets have Furthermore, it is grounded in modern portfolio theory; been proposed. it can be applied both at the market level and at the com- Godfrey and Espinosa [1996] propose adjusting the pany level; it is not based on subjective measures of risk; CAPM in two ways. First, they add to the risk-free rate it can be fine-tuned to any desired benchmark return; and the spread between the yield of an emerging market it captures the downside risk that investors want to avoid. sovereign bond denominated in dollars and the yield of Any required return can be thought of as having two a comparable U.S. bond. Second, they are an “adjusted components: a risk-free rate, and a risk premium. The first beta,” defined as 60% of the ratio between an emerging component is compensation required for the expected loss market’s standard deviation of returns and the standard of purchasing power, which is demanded even for a risk- deviation of returns in the U.S. market. less asset. The second component is extra compensation There are at least two problems with this approach. for bearing risk, which depends on the asset considered.4 First and foremost, not all countries issue dollar-denom- I take the perspective of a U.S.-based, internation- inated debt. And second, it is difficult to assess how much ally diversified investor. Thus, the risk-free rate should risk is double-counted by the joint adjustment of the compensate this investor for the dollar’s expected loss of risk-free rate and the beta. Godfrey and Espinosa’s [1996] purchasing power, and the risk premium should com- ad-hoc adjustment is based on empirical results reported pensate the investor for the risk of investing in the world in Erb, Harvey, and Viskanta [1995]. market portfolio. Mathematically: An alternative approach is based on credit ratings, such as those provided by Institutional Investor or Polit- RRi = Rf + (RPW)(RMi) (1) ical Risk Services. Erb, Harvey, and Viskanta [1995, 1996a] report that country credit ratings are significantly where RRi is the required return, Rf is the (U.S.) risk- related to stock returns, and they propose a model to esti- free rate, RPW is the world market risk premium, RMi is mate the cost of equity in emerging markets based on a risk measure, and i indexes markets. these indices. Diamonte, Liew, and Stevens [1996] also There are several alternative risk variables. I propose report that ratings of political risk are significantly related to estimate required returns in emerging markets using a to stock returns. measure of downside risk. More precisely, I propose to use The problem with this approach is twofold. First, an RMi equal to the ratio between the semistandard devi- this methodology is designed to estimate a countrywide ation of returns with respect to the mean in market i and cost of equity and cannot be applied at the company the semistandard deviation of returns with respect to the level. Second, the numerical value of the ratings, critical mean in the world market. The semistandard deviation of for quantitative analysis, is highly subjective. In many returns, or semideviation for short, with respect to any ⌺ cases, the indices are based on qualitative variables and benchmark return B ( B) is given by arbitrary weights. Erb, Harvey, and Viskanta [1996b] pro- T 2 vide some details on the construction of these ratings. Â=Bt(/1 TRB ) Â=1 (t - ) for all Rt < B Another approach, proposed by Bekaert and Harvey (2) [1995], incorporates a time-varying measure of market integration. The cost of equity is plausibly allowed to where T is the number of observations in the sample, R change over time, depending on the degree of market denotes returns, and t indexes time. integration, and required returns are determined by a Several reasons support the plausibility of a down- time-varying weighted average of a global beta and a local side risk approach. First, Bawa and Lindenberg [1977] standard deviation. The main problem with this approach develop a model based on downside risk and show that, lies in the difficulty of the estimation procedure; it is far the CAPM being a special case of their model, the latter more complicated than the more straightforward methods must explain the data at least as well as the former. Sec- that companies typically seek to apply. ond, Harlow and Rao [1989] refine the Bawa-Lindenberg downside risk framework and find that the data support 2 THE COST OF EQUITY IN EMERGING MARKETS: A DOWNSIDE RISK APPROACH FALL 2000 their model. Third, Sortino and van der Meer [1991] Cross-Sectional Analysis show that a mean-downside variance optimizer outper- forms a mean-variance optimizer. Fourth, Markowitz The first step of the analysis consists of computing, [1959] does consider downside risk in his path-breaking over the whole sample period available for each country, analysis and states that “the semideviation produces effi- one statistic that summarizes the average (return) perfor- cient portfolios somewhat preferable to those of the stan- mance of each market, and another number that sum- dard deviation.”5 And finally, investors do not dislike marizes its risk under each of the definitions considered. volatility; they dislike downside volatility. Average returns over the whole sample period are sum- marized by mean monthly arithmetic returns. Several risk variables are considered. The first is sys- DATA AND RESULTS tematic risk (SR) measured by beta. The second is total risk The data used in this article consists of the entire (TR) measured by the standard deviation of returns. The database of Morgan Stanley Capital International (MSCI) third is idiosyncratic risk (IR) measured by the variability of of emerging markets, which covers 28 countries over returns not explained by beta.7 The fourth is size (Size) mea- different time periods. Summary statistics of the data in sured by the log of the average market capitalization over these markets, as well as for the Emerging Markets Free the sample period, and the fifth is downside risk measured (EMF) index, a capitalization-weighted index of emerg- by five different variables. ing markets, are provided in Exhibit A-1 in the appendix. I consider three downside risk variables based on the Returns used throughout are monthly returns, measured semideviation of returns, one for each of three different in dollars, accounting for both capital gains and divi- benchmarks.