Uniform testing and portfolio strategies for single and multifactor asset pricing models in the Pacific Basin markets

Abstract

We test and implement portfolio strategies for three major asset pricing models, under uniform diagnostic measures using the PACAP data set containing all current listing and de-listing of firms for the local stock exchange in several Pacific Basin countries. Compared to the often used MSCI database that include only a subset of the (large) firms in the local markets, the more complete coverage of our database allows for more robust testing of current multifactor asset pricing models since the possible effects of additional factors such as size and book to market may not show up correctly using less comprehensive data sets. Our data set also provides a natural packet of nonUS data for addressing the issue of whether the results of recent asset pricing research are sample specific. Our overall results provide multi-country (sample nonspecific) support for the additional asset pricing risk factors of the Fama-French three-factor model but not for the momentum factor of the Carhart model. We additionally find that the size risk factor is more prominent than value risk factor in the Pacific Basin markets. Finally, we find strong evidence that portfolio strategies implemented to capture value and size effects are profitable in the Pacific Basin stock markets.

Notes

¹ For example, Dowen (1988).

² We include all the common stocks listed in the main board and the second board of the sample countries, but exclude those common stocks listed in the alien board of Malaysia and Thailand, and those traded by foreign currency but listed in the main board of Singapore.

³ For example, when we compute book-to-market equity ratio (BE/ME), we transform the book equity (denominated as foreign currency) to the local currency through the exchange rate of the reporting month and then divide by market equity (denominated as local currency).

⁴ Though there are some other financial ratios, such as cash flow yield (C/P) and earning yield (E/P), that can be used to identify value and growth stocks, studies such as Fama and French (1998) and Lakonishok et al. (1994), for example, have shown that the book-to-market equity ratio is no worse than the other measures. Moreover, Fama and French (1992) found size and BE/ME to jointly absorb the roles of earning yield (E/P) in average returns.

⁵ In the PACAP database, if any item is unavailable, the corresponding monthly returns will be coded as missing.

⁶ In their paper Fama and French (1993) also described forming 25 size-BE/ME portfolios using the same construction process.

⁷ Chan et al.(1991), for example, investigate the relationship of fundamental variables and the cross-section of stock returns in Japan market. They form the portfolios of three sorting variables divided into four equal parts (4 × 4 × 4). In order to overcome the problem of ‘too few stocks in a portfolio’, they use a ‘three-way grouping’ procedure similar to the sequential sorting approach of this article.

⁸ Technically, this means, for example, if we let H/S/L denote the portfolio of high book-to-market equity ratio conditional on small market capitalization and past losers (low past returns), then this portfolio can also be interpreted as small size portfolio conditional on high book-to-market (HBM) equity ratio and past losers, or low past performance portfolio conditional on high book-to-market equity ratio and small size.

⁹ The data series code of IFS is, Japan 15860B … ZF … , China, P. R.: Hong Kong 53260B … ZF … , Indonesia 53660B … ZF … , Korea 54260B … ZF … , Malaysia 54860C … ZF … , Thailand 57860C … ZF ….

¹⁰ The series of Hong Kong, known as HIBOR (Hong Kong Interbank Overnight Rate), is unavailable before December 1993 in IFS; their complete historical records are accessible online in information center of the Hong Kong Monetary Authority (HKMA).

¹¹ CAPM, FF, Carhart

¹² We use the following target portfolios from the PACAP Monthly Index File for the trading tests: Hang Seng Index (HSI) of Hong Kong, Straits Time Industrial Index (STII) of Singapore, Kuala Lumpur Composite Index (KLCI) of Bursa Malaysia, Tokyo Stock Price Index (TOPIX) and First Section Stock Index of Korea Stock Exchange (KSE 1). For Thailand we use the average of 12 industrial indexes as our target portfolio. Indonesia was excluded from the trading tests because we could not find a suitable target portfolio from our database that does not overlap excessively with the market portfolio.

¹³ Our results do not change substantially when they are converted to US dollar returns.

¹⁴ Only the SMB case for RH = 4Q in market of Indonesia remains insignificant under the looser 90% confidence level.

¹⁵ It should be noted that Rouwenhorst (1999)'s conclusions are based on using a sample of all emerging markets as a whole. Even his results show that there does not exist momentum effects in the Asian emerging markets of Indonesia, South Korea, Malaysia and Thailand.

¹⁶ Momentum implies positive return difference between past winner and past loser. Only the cases of RH = 2Q in Singapore and RH = 3Q in Malaysia produce significant positive profit.

¹⁷ The time-series regression approach is based on the methodology discussed in Black et al. (1972) and Fama and French (1993).

¹⁸ Unbiased, well-behaved, CAPM based models should, in theory, have the coefficient ‘a’ near zero and ‘b’ near 1. Our results show that Malaysia was the only country that yielded S-a that was too high to fit the CAPM.