VALUING EMERGING MARKET EQUITIES – THE NEW EMPIRICAL EVIDENCE

By Niso Abuaf

Niso Abuaf is Clinical Professor of Finance and Economics at Pace University; and Chief Economist and Strategist at Ramirez and Co., in New York, New York.

I would like to thank Saichalee Bee Limaphichat for invaluable research assistance.

Though practitioners and academics rely on similar conceptual frameworks when valuing international equities in general and emerging market equities in particular, they emphasize different aspects of the framework. In contrast to academics, practitioners adjust discount rates as opposed to cash flows, and use the US as opposed to the global equity market risk premium.

Abuaf (2011) summarizes the arguments on the academic and practitioner sides, lets the data do the judging, and presents evidence that US dollar returns on emerging market equities (ADRs) primarily are a function of returns on the broad US equity market (e.g., the S&P 500) and on the corresponding country’s credit default swap (CDS) spreads. Because CDSs are standardized contracts that are far more liquid than dollar denominated emerging market bonds, in the current market environment they are better indicators of emerging-market risk. As such, I use them in my empirical work.

In this paper, I update the evidence presented in Abuaf (2011) with a slight variation. In particular, I regress returns on ADRs, versus returns on the S&P 500 (both represented as log differences) and changes in CDS spreads, which is the variation presented in this paper. As such, the coefficient of the S&P500 would be interpreted as the traditional β, and the coefficient of the CDS would be interpreted as the modified duration of equities (semi-elasticity) with respect to CDS. I mathematically demonstrate why the modified duration equals the P/E multiple. When presenting the empirical results, I visually test whether the modified duration of equities vs. CDS spreads approximate the P/E multiples.

When the worldwide privatization boom began in the late 1980s, sellers, buyers, and financial intermediaries realized that they needed a framework within which to price assets in disparate regions of the world. Unfortunately, standard international corporate finance theory could offer little assistance, primarily because it argued that when valuing, for example, telephone assets in Mexico, one should account for Mexican risk by adjusting the expected cash flows and then discounting these cash flows by applying a weighted average cost of capital (WACC). This approach was similar to valuing telephone assets residing in the US, but not applicable to international valuations as analysts had no rational way of adjusting cash flows to reflect country risk, such as Mexico’s.

As an alternative method, Abuaf and Chu (1991, 1994) and Abuaf, Chu, Czapla, Lawley and Thadani (1997) recommend a pragmatic strategy of adjusting the cost of equity, leading to a related adjustment to the WACC. Intuitively, in a virtually integrated global capital market, the risk associated with the Mexican telephone-asset example consists of two building blocks:

·  A US telephone-asset risk, and

·  The risk associated with an investment in Mexico.

The above statements are simplifications that practitioners have used and continue to use. In theory, the correct approach would be to model the risk of the telecommunications industry worldwide, which in all likelihood would be modeled as returns on the global capital markets, adjusted by a global telecom beta. A further adjustment might be needed in that Mexico’s telecom beta might be different than a global or US telecom beta, because of the life-cycle maturity, or other industry characteristics, of the Mexican telecom industry versus its global counterparts. The literature’s approach to these questions is not monolithic, as summarized below:

Literature Survey

This Journal has had a long history of publishing articles related to valuing international equities and investments, particularly as they relate to emerging market investments. For example, since the early 1990s various authors have made the following salient points:

·  Stulz (1995). “The increasing synchronization (or correlation) of both real international business activity and world financial markets is partly offsetting the benefits of global diversification.”

·  Lessard (1996). “How managers adjust for risk (whether by raising the discount rate or reducing expected cash flows) should depend primarily on whether the risks are ‘systematic’ or instead are ‘diversifiable’ by world capital markets.”

·  Godfrey and Espinoza (1996). “A credit spread is added to the US dollar risk-free rate to reflect transfer risks; and (2) a country-specific business volatility premium is used to reflect risks associated with the local business environment.”

·  Keck, Levengood, and Longfield (1998). “While partly segmented markets may in fact have some different types of risk, the primary driver of cost-of-capital differences is likely to be ‘risk-price differentials’: the same risk priced differently.”

·  Stulz (1999). “If the extra risk premium is used to compensate for country risks, then it must be demonstrated that those risks are not diversifiable and that shareholders charge a risk premium to hold those risks.”

·  Schramm and Wang (1999). An important judgment call relates to “ what the base portfolio should be – the home-country market (assuming markets are segmented) or the global market (assuming integrated markets).”

·  O’Brien (1999) suggests: “a formula for converting a firm’s cost of equity from US dollars into another currency.”

·  Sabal (2004). “This paper argues that the traditional practitioners’ approach of incorporating a country risk premium is not appropriate, mainly because country risk is neither the same for all projects nor totally systematic, and there is no reason for it to be closely related to the spread on the government bonds of the country concerned.”

·  Estrada (2007). “Although much has been published about discount rates in emerging markets, there is probably a long way to go until a convergence of opinions finally arises. “

·  Soenen and Johnson (2008). “ When valuing projects in emerging economies, we recommend use of the CAPM adjusted for political risk and a measure of co-movement (country beta) between foreign and US stock markets. In the long run, increased international capital market integration can be expected to move country betas toward unity. But in the meantime, corporate planners should consider making the necessary adjustments to the CAPM.”

·  Garcia-Sachez, Preve, and Sarria-Allende (2010). “Current methodologies based on adjustments of the discount rate present several problems. On the one hand, the practice of adding country risk to CAPM estimates violates the spirit of the model, according to which discount rates should reflect only ‘symmetric’ (or two-sided), non-diversifiable risks. Country risk, however, is not symmetric and may be at least partially diversifiable (though in respect to the latter, some models of dynamic correlations across countries give some support to the alternative view). On the other hand, popular techniques usually view the impact of emerging market country risk as the same for all firms and industries. It is easy, however, to find differences in business fundamentals or strategies that justify the opposite view: namely, that the effect of country risk on corporate values should depend on company- or industry-specific characteristics.”

·  Pereiro (2010). “The choice of a cost of equity model for an emerging-market firm is very personal: it depends on how conceptually sound the model looks to the analyst, and on her view on which risks can – and which cannot – be diversified away by the investor.”

·  Anshuman, Martin, and Titman (2011). “But in practice, the political risk associated with this type of investment [emerging markets] is typically accounted for implicitly by adjusting the investment’s required rate of return or the discount rate.”

·  My bottom line. As stated in the Introduction, my bias has been to adjust the discount rate primarily because I have believed that adjusting cash flows is ad hoc and also because I have never believed that country risk is fully diversifiable. In light of the diversity of opinion reflected above, I will let the data do the judging and determine whether:

o  Industry betas in emerging markets are similar to industry betas in the US.

o  CDS spreads are statistically significantly correlated with ADR returns.

o  Industries have differing CDS sensitivities, possibly due to differential political and country risk exposures of differing industries.

I. Theory

To reiterate, I take a pragmatic approach and move towards letting the data do the talking. My own intuition is that country risk is not fully diversifiable, and should be incorporated in the discount rate. However, this intuition is not written in stone, and as Keynes famously stated “if the data change, I will change my mind.” I also agree with the above authors who opine that different industries might bear different amounts of country risk. Again, as we shall see later on, I will let the data judge that.

As stated in Abuaf (2011), the practical approach to estimating the cost of an emerging market equity is similar to estimating extended Capital Asset Pricing Models (CAPM) models or frameworks that academics have introduced such as the arbitrage pricing theory (APT), the initial, or the extended Fama-French models. Stated differently, the introduction of a political risk premium variable (CDS) is in the same spirit as the above extensions of the CAPM. In this spirit, I postulate that the cost of equity, ke , can be estimated by an extended CAPM:

Ke =dlog (ADR) = c + β(dlogS&P500) + γ(ΔCDS), (1)

where c is a constant (a purist would argue that c equals the risk-free rate minus β times the risk-free rate), dlog represents log differences, approximating percentage changes, ADR represents the levels of emerging market stock prices being investigated, β is the traditional CAPM constant, S&P500 is the level of the broad US market stock index, γ is the modified duration of ADRs vs. CDS (aka semi elasticity), Δ represents first differences, and CDS is represented as a real number.

A.  Why γ is Related to (or Approaches) the P/E Multiple

In a traditional perpetuity growth model:

P/E = 1/(ke-g+cds) (2)

where g represents the perpetual growth rate. Taking derivatives on both sides:

d (P/E) = - (ke-g+cds)-2 cds, and (3)

d (P/E)/(P/E)/dcds= -1/( ke-g+cds). (4)

I have therefore shown that γ should theoretically equal the negative P/E multiple in the absence of a political risk premium. That is, in theory, γ equals the negative of the P/E multiple for an equity that has no country risk. In principle:

ke = risk-free rate + β(equity market risk premium) +α (CDS), (5)

where α represents the proportion of CDS risk borne by the specific industry.

II. The Empirical Results

Tables I-X summarize my empirical results. In choosing the countries, I wanted to make sure that various regions of the world such as Latin America, the Asia Pacific region, and Europe were well represented. My results include two European countries, Italy and Spain, which strictly speaking should not be counted as emerging market countries. I have, nonetheless, included them because the European financial crisis has underscored the political and country risk associated with investing in these countries. Other criteria were that each country should have at least ten ADRs, and that the results should be meaningful. In this regard, I rejected:

A.  Rejected Countries

Argentina. I would characterize the results for Argentina as weak with 6/13 “γ”s as statistically significant. None of the Argentine “γ”s approach their theoretically expected P/E multiples. Moreover, the Argentine R2s are by and large weak with only 5/13 exceeding 25%. These weak results may be due to Argentina’s troubled political economy and its declining importance in the global landscape.

Greece. Having only five observations, and respectively median γ, R2, and P/E’s of 0.85, 36%, and 7.44x, I would characterize the Greek results as weak, possibly due to the small sample size and everything else that has been happening in Greece.

Indonesia. Having only six observations, and respectively median γ, R2, and P/E’s of 7.41, 21%, and 14.24x, I would characterize the Indonesian results as weak, possibly due to the small sample size. Nonetheless, a typical Indonesian “γ”s is statistically indistinguishable from its corresponding P/E’s, as predicted by theory.

Israel. Having only eleven observations, and respectively median γ, R2, and P/E’s of 5.96, 12%, and 10.24x, I would characterize the Israeli results as weak, possibly due to the small sample size. Nonetheless, a typical Israeli “γ” is still statistically indistinguishable from its corresponding P/E, as predicted by theory.

Japan. With 20 observations, has respectively median γ, R2, and P/E’s of 3.61, 28%, and 12.32x, I would characterize the Japanese results as weak. Most of th“γ”s (17/20) for Japan are statistically insignificant. Nonetheless, the “γ” of a global brand such as Nissan is statistically indistinguishable from its corresponding P/E, as predicted by theory.

Malaysia. With only two observations, the results are extremely weak with R2s approaching zero.

Philippines. With only two observations, the results are extremely weak with one R2 at 22%, and the other approaching zero. Nonetheless, Philippine “γ”s are statistically indistinguishable from their corresponding P/E’s, as predicted by theory.

Portugal. Having only five observations, and respectively median γ (only two statistically significant), R2, and P/E’s of 0.79, 19%, and 23.83x, I would characterize the Portuguese results as weak, possibly due to the small sample size and everything else that has been happening in Portugal.

Thailand. With only five observations, the results are extremely weak with R2’s approaching zero.

Turkey. Having only five observations, and respectively median γ (only three statistically significant), R2, and P/E’s of 11.59, 28%, and 12.81x, I would characterize the Turkish results as modest, primarily because the median γ is virtually equal to the median P/E.

Regarding data frequency and the use of daily versus weekly data, I find that, with two and a half years-long data, the bottom line does not change appreciably whether we use daily or weekly data.

B. Analysis of Results

1.  Brazil. I would characterize the results for Brazil as strong with all “γ”s as statistically significant. The median Brazilian “γ”s approach their theoretically expected median P/E multiples. Specifically, the median Brazilian γ is 13.37 while the median Brazilian forward P/E is 9.31. Given a typical standard error of a Brazilian γ of around four, I can confidently say that Brazilian “γ”s are statistically indistinguishable from their corresponding P/E’s, as predicted by theory. Moreover, the Brazilian R2s are by and large strong with all exceeding 24%, and 7/17 approaching, or exceeding 50%. These strong results may be due to Brazil’s growing importance in the global landscape.