I am grateful to Goldman Sachs, the International Finance Corporation, and Morgan Stanley Capital International for generously providing data. I also thank Brian Aitken, John Anderson, Jahangir Aziz, Steven Dunaway, Shogo Ishii, Takatoshi Ito, John McDermott, and Mark Stone for helpful discussions. The views expressed here are my own and do not necessarily reflect those of the Fund. The usual disclaimer applies.
All returns are measured in excess of the riskless rate of return.
Assuming zero transactions or informational costs and that assets are infinitely divisible, a representative investor would hold each equity in proportion to its share in the world portfolio, such that his or her portfolio would be identical to the world portfolio in composition.
All returns are calculated in excess of the holding yield on a constant one-month maturity U.S. Treasury bill intended to represent the riskless rate of return.
By shifting from a portfolio composed entirely of the Financial Times-Actuaries (FT-A) World index to a portfolio composed of 50 percent of the IFC Emerging Markets Data Base (EMDB) Investible Composite index and 50 percent the FT-A World index, investors would have increased the return on their portfolios from roughly 7 to 17 percent, while slightly decreasing portfolio variance from about 17 to 16 percent.
Bekaert and Urias’s (1995) analysis of the diversification benefits from emerging market closed-end funds versus from the IFC EMDB indices suggest that the high returns on the indices may overstate the effective return once transactions costs and other barriers to investment are considered. Bekaert and Urias assume that returns on closed-end funds approximate achievable returns inclusive of all transactions and other costs. They find statistically significant diversification benefits from U.K. emerging market funds, but not from comparable U.S. funds.
Benartzi and Thaler cite 60 percent stocks and 40 percent bonds and treasury bills as typical asset-allocation proportions. Leibowitz and Langetieg indicate that for a twenty-year horizon, the stock to bond ratio should exceed 100 percent most of the time.
Benartzi and Thaler also present myopic loss aversion as an explanation for the equity premium puzzle–the return premium, typically estimated at 6.5 percent, which investors demand on stock relative to bonds. Benartzi and Thaler attribute the premium to the fact that stocks have both upside and downside risk, whereas a bond is often treated as having no downside risk if held to maturity. However, even if held to maturity, bonds have downside risk in the form of default risk.
Loss-averse investors will demand a return premium for accepting additional variance risk, regardless of the asset’s covariance with the market portfolio and, due to the utility function’s convexity over losses, the magnitude of the premium will increase more than one-for-one with increases in an asset’s variance. Hence, loss aversion is in contradiction to the Capital Asset Pricing Model.
The size of the return premium that loss-averse investors will demand is greater if the investor evaluates his portfolio over short-time horizons, as do institutional investors. Benartzi and Thaler observe that institutional investors typically operate with relatively short horizons tied to the length of time they expect to remain in their job; in addition, investment managers’ portfolio performance is evaluated annually, and their bonus set accordingly. Similarly, De Bondt and Thaler (1994) observe that institutional investors have higher turnover ratios than individual investors.
The IFC EMDB Global Composite portfolio is used to proxy a true capitalization-weighted portfolio of developing country equities.
This is the case even if expected returns are correlated with one another within the asset class–any two developing country assets which are correlated with one another should affect expected returns only through their impact on the world portfolio. When using actual returns as a proxy for expected returns, however, a non-zero coefficient on the developing country portfolio can reflect two possibilities: (i) the traditional one-factor CAPM is insufficient to explain investor behavior or (ii) unanticipated shocks to asset returns are shared across developing countries. If an unanticipated shock to one country spreads to others because investors, who treat developing country assets as a class, shift their sentiment regarding other markets, a non-zero coefficient would imply that the strict one-factor CAPM did not hold.
The investible indices include only those stocks accessible to foreign investors.
The IFC indices are available only weekly, whereas local indices are available on every trading day and, in some cases, continuously throughout the day.
Using an F test, the CAPM can be rejected with 95 percent, confidence in seven markets (Brazil, Chile, Korea, Mexico, Malaysia, Taiwan Province of China, and Thailand) and with 90 percent confidence for India.
Tests on a split sample (presented below) indicate that the rejection may be due to substantial structural change in the Argentine market’s behavior. The- CAPM model is rejected in favor of the two-factor model for 1992-95, but not for 1989-91.
The estimates of βem may be biased upward somewhat because the return on market j, rj t, may have a significant weight in determining rem,t.
The exception is Jordan.
The estimates of βem and the regional betas may be biased upward somewhat because the return on market j, rj, t, may have a significant weight in determining rem, t or the regional return.