ارزش سرمایه گذاری و تجزیه و تحلیل فنی در بازار سهام تایوان

Unlike the U.S. and most developed countries, Taiwan stock market has been widely documented to have no value premium. Prior studies on the value premium typically adopt a conventional approach proposed by Fama and French (1992), which suggests a buy-and-hold strategy with annual rebalancing. We argue that a sophisticated investor can do better (obtain higher returns) than a simple buy-and-hold strategy by timing the market with the help of some technical analysis. Specifically, we show that an application of a moving average timing strategy to portfolios sorted by book-to-market (BM) ratios could generate higher returns than the buy-and-hold strategy. Using common stocks listed on the Taiwan Stock Exchange (TWSE), we confirm that the moving average timing strategy does substantially outperform the buy-and-hold strategy. Taking advantage of this observation, we propose a zero-cost portfolio constructed by buying the highest BM portfolio, and short-selling the lowest BM portfolio based on trading signals issued by the moving average rule, and demonstrate that such a new investment strategy can produce significantly positive returns. Robustness of results obtained in this paper is further verified and consolidated by extending the empirical study with a different currency, alternative lag lengths, transaction cost, subperiod analysis, business cycles and market timing.

The value premium, which states that stocks with high book-to-market (BM) ratios yield higher returns than those with low BM ratios, has been extensively documented in both the U.S. and international stock markets over the past two decades. Fama and French, 1992 and Fama and French, 1996, Lakonishok, Shleifer, and Vishny (1994) and Loughran (1997) all show that the value premium is a prevalent phenomenon in the U.S. market. As for the international evidence, Chan, Hamao, and Lakonishok (1991), Roll (1995), Mukherji, Dhatt, and Kim (1997), Bauman, Conover, and Miller (1998), Fama and French (1998), and Daniel, Titman, and Wei (2001) all demonstrate the existence of the value premium, either in a specific country outside the U.S. or on an aggregate basis across countries. The evidence of the value premium in emerging markets or Asian markets, however, is somehow mixed and less pronounced. For example, out of the 20 emerging countries being examined, Rouwenhorst (1999) finds that high BM stocks have higher returns than low BM stocks in 16 of them, but significantly positive premiums only exist in Brazil, Korea, Malaysia and Zimbabwe. Furthermore, some countries such as Colombia, Pakistan, Portugal and Thailand even carry negative premium. For Asian markets, Mukherji et al. (1997), Chen and Zhang (1998), Chui and Wei (1998) and Ding, Chua, and Fetherston (2005) show that high BM stocks outperform growth stocks in Japan, Hong Kong, Korea, Malaysia and Singapore, but not in Indonesia, Taiwan and Thailand. Brown, Rhee, and Zhang (2008) show that, after controlling for size and liquidity effects, there exist significant value premiums in Hong Kong, Korea and Singapore, but value discounts in Taiwan. One thing in common among the aforementioned studies is the adoption of a conventional approach proposed by Fama and French (1992), which suggests a buy-and-hold strategy with annual rebalancing. Specifically, at the beginning of July in a given year, an individual firm's BM ratio is evaluated as its book value of equity divided by its market value of equity at the end of December in previous fiscal year. After assigning individual stocks into a particular portfolio according to their BM ratios, investors are assumed to sell short the low BM portfolio, and to use the proceeds to invest in the high BM portfolio simultaneously from July of year t to June of year t + 1 The value premium is then defined as the return of such a zero-cost value-minus-growth portfolio, which is assumed to be held for one year with annual rebalancing. But, is the buy-and-hold strategy necessarily good for investors? What if the value premium exhibits some predictable patterns? Indeed, Cohen, Polk, and Vuolteenaho (2003) document statistically strong results concerning the predictability of returns on the value-minus-growth strategy. They show that the expected return on the value-minus-growth strategy is high when the spread in BM ratios is wide. In addition, if some sophisticated investors are able to “time the market”, a better trading strategy for investors might be to invest in the value-minus-growth strategy only when the value premium is positive and significant. Therefore, the main question we address in this paper is whether trading signals generated by some technical analyses can contribute incremental value to the value-investing strategy, even in a market absent of the BM effect. Prior studies have already demonstrated the usefulness of technical analysis on some investment strategies. For example, Brock, Lakonishok, and LeBaron (1992), Brown and Jennings (1989), Lo, Mamaysky, and Wang (2000) and Neely, Rapach, Tu, and Zhou (2011) all find that technical analysis adds value in investing stock or market returns. Zhu and Zhou (2009) find that a trading rule based on the moving average (MA) provides additional information for the fixed-proportion strategy that follows Markowitz's (1952) modern portfolio theory and Tobin's (1958) two-fund separation theorem. Han, Yang, and Zhou (forthcoming) show that an application of the MA rule to portfolios sorted by volatility can generate reliably higher returns than the traditional buy-and-hold strategy. However, the usefulness of technical analysis documented in these studies is theoretically or empirically conducted based on particular strategies that have been documented to be profitable with prior data. But, if a given style investing strategy has been verified to produce no premium in one country, does such investment strategy combined with technical analysis become profitable? We answer this question by linking value strategies based on BM portfolios and MA signals in Taiwan stock market, which is chosen for two main reasons. First, Taiwan has been widely documented to have no value premium (Brown et al., 2008, Chen and Zhang, 1998, Chui and Wei, 1998, Ding et al., 2005 and Hung et al., 2012), hence it serves as a natural experimental environment to examine the pure effect of technical analysis on value investing. Second, as argued by Zhu and Zhou (2009) and Han et al. (forthcoming), incomplete information on the fundamentals is a key for investors to use technical analysis. Since the degree of information incompleteness in an emerging market like Taiwan would be more severe than those in developed markets, technical analysis is more important for investors in Taiwan. With a sample containing all common stocks listed on the TWSE from January 1980 to December 2010, we first confirm that the standard long-short BM portfolio based on buy-and-hold strategy does not produce significantly positive premium. After considering the buy or sell signals implied by the MA rule as suggested by Han et al. (forthcoming), we show that the 10-day MA timing strategies outperform the buy-and-hold strategies for all of the ten BM decile portfolios. Abnormal returns between the MA timing strategies and the buy-and-hold strategies cannot be explained by either the capital asset pricing model (CAPM) or Fama and French's (1993) three-factor model. Given the fact that MA signals are able to enhance the profitability of all BM portfolios, a simple difference between the highest and lowest BM portfolios under MA timing strategies still fails to produce significant premium. We argue that the simple difference between the highest and lowest BM timing portfolios implies that investors sell short the lowest BM portfolio when the index price is higher than its MA indicator, which is in contradiction to the spirit of the MA rule. Hence, we propose a new strategy (TLS) to construct the zero-cost long-short portfolio by buying the highest BM portfolio when its index price is higher than its MA indicator, and by short-selling the lowest BM portfolio when its index price is lower than its MA indicator. We document a significantly positive premium of 21.953 (16.277) basis points per day for the equally-weighted (value-weighted) TLS strategy, which amounts to an annual return of 54% (40%) per annum. High premiums are not explained by the CAPM or Fama and French's (1993) three-factor model. For example, the TLS strategy with equal weights has a CAPM-adjusted return of 22.500 basis points per day, and a Fama–French risk-adjusted return of 22.864 basis points per day. We test the robustness of our results in several ways. First, the main context of this paper is performed based on prices and returns in the local currency. We examine whether our results are affected by exchange rates by converting them into U.S. dollars, and show that our results remain unchanged. Second, we consider alternative lag lengths, including 5, 20, 50, 100, and 200 days, for the moving averages. We find that the abnormal BM premiums under MA timing strategies tend to be higher in short-horizon, with decreasing magnitude over the lag lengths. Nevertheless, the BM premiums are still significantly positive with the long lag lengths. Third, we examine the trading behavior, trading frequencies and break-even transaction costs of these MA timing portfolios. We show that investors do not have to trade these portfolios very often implied by MA signals and that the break-even transaction costs are reasonably large. Fourth, we show that our results sustain in subperiods, and are thus free from the data mining problem. We further show that our results are not driven by business cycles. Finally, we analyze the source of the superior performance of the TLS strategy by examining whether the TLS strategy reveals any market timing ability. By applying two approaches proposed by Treynor and Mazuy (1966) and Henriksson and Merton (1981), we show supporting evidence for the market timing ability of the TLS strategy. However, market timing alone does not explain abnormal returns of the TLS strategy. The rest of the paper is organized as follows. Section 2 describes the investment timing strategy using the MA as the timing signal, and the data used in this paper. In Section 3, we document the evidence for the profitability of the MA timing strategy, as well as the significance of the BM premium conditioning on the MA timing strategy. The robustness of profitability in a number of dimensions is examined with results presented in Section 4. The last section gives concluding remarks.

There is ample evidence suggesting that the BM effect does not exist in Taiwan stock market. A common feature among the vast literature is the adoption of a buy-and-hold strategy with annual rebalancing, which is proposed by Fama and French (1992). In this paper, we examine the role of technical analysis on the value investing, and whether MAs signals provide additional information in predicting the return difference between high- and low-BM portfolios. We follow Han et al. (forthcoming) in constructing the MA timing strategy across BM portfolios, and document superior performance for the MA timing strategy over the buy-and-hold strategy. We further contribute to the finance literature by proposing a long–short portfolio with zero cost conditioning on MA signals. The new strategy suggests a buying signal when the index price of the highest BM portfolio is higher than its MA indicator, and a short-selling signal when the index price of the lowest BM portfolio is lower than its MA indicator. Under such a construct, we show that the new strategy yields significantly positive return, and provides higher returns than the standard buy-and-hold strategy in Taiwan. The abnormal return is both economically and statistically significant, and cannot be explained by either the CAPM or Fama and French's (1993) three-factor model. The significance of the premium under the newly proposed zero-cost strategy is quite robust in several aspects. First, the overall results are robust to exchange rates. Second, the premium is significant regardless of lag lengths of MA signals. Third, it turns out that investors adopting MA signals do not have to trade BM portfolios too often, and that the break-even transaction costs are reasonably large to ensure profitability of such strategies. Fourth, the results sustain in subsamples and are not driven by business cycles. Finally, we document successful market-timing ability of the new strategy, but market timing alone does not explain abnormal returns.