A variance equality test of the ICAPM on Philippine stocks: post-Asian financial crisis period

Abstract

The study examines whether Fama's discrete version of Merton's intertemporal CAPM (ICAPM) can explain the negative market risk premium and the cross-sectional variability of Philippine stock returns after the onset of the Asian financial crisis in July 1997. The change in foreign exchange rate, in addition to the change in market risk premium, is used as a state variable of hedging concern to investors. The relationship of Fama's multifactor minimum-variance (MMV) portfolio to the Markowitz minimum-variance (MV) portfolio is characterized in terms of the equality of the return variances for the same expected return. A test due to Basak et al. (2002) is then used to verify the equality of the return variance of a derived tangency portfolio along the MMV frontier to an MV portfolio with the same sample mean return. The results do not reject the ICAPM during the period covered by the study. Thus, the model provides a plausible explanation both for the cross-sectional variability of stock returns and the negative market risk premium within the framework of mean-variance optimizing investors.

Acknowledgements

The author is grateful to Amado P. Saguido and an anonymous referee for their valuable comments and suggestions. The usual disclaimer applies.

Notes

¹ The results of these studies are limited because of the use of an index which is probably a poor proxy for the market portfolio (see Aquino, 2002–2003). This is the 30-firm Phisix (to be described more fully later in the text).

² There were only 58 firms with continuous trading from 1992–1996 out of the 170–216 listed firms in the PSE during the same period.

³ The covariance matrix under the delta method is composed of covariances between returns and transformations of returns data. The covariance matrix is estimated using the Newey–West (1987) method. Since monthly returns are generally AR(1), a maximum lag of 3 is used.

⁴ Figure 1 has standard deviation as the horizontal axis instead of the variance, following traditional practice.

⁵ Since V is positive definite and the constraints are linear, the second order condition for minimization is met.