國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2. Literature Review
There is no guarantee that investors are always mentally rational. Therefore, they try to avoid unfounded behaviors and adopt program trading to make their investments more reason-based. Over time, investors have built up many trading models to avoid the intervention of irrational trading decisions and have created new trading models with better performance. Therefore, technical analysis has become a commonly used investment tool in financial markets.
Is technical analysis effective? Fama (1970) addresses Efficient Market Hypothesis (EMH) which considers that all the market information is already reflected on the price and investors cannot obtain excess profits from predicting future prices. Shleifer (2000) also signifies some other points: 1. Investors are rational, so they can conduct rational price evaluation. 2. Even if some irrational investors exist, the trading are stochastic, price will go back to efficient at the end because of arbitrage and the offset of trading.
However, from the past to the present, there has been a growing number of empirical studies that find that some abnormal phenomenon cannot be explained, such as the January effect, week effect and the scale effect. Some scholars began to doubt the authenticity of efficient markets, and a lot of research verifies that technical analysis can indeed obtain excess profits in stock markets.
Alexander (1961) uses the ratio of filter rules (FR) between 0.5% and 50%
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
to test Dow Jones Industrial Average (DJIA) from 1897 to 1959 and Standard
& Poor’s 500 Index (S&P 500) from 1929 to 1959. Without considering transaction cost, the returns from FR are much better than buy-and-hold strategy. This study brought on the attention of several scholars. Mandelbrot (1963) points out the mistake in Alexander’s (1961) study: investors cannot trade on time when the signal shows up, so the bid price is always higher and the ask price always lower. Alexander (1964) revised the model based on Mandelbrot and found out FR could not beat buy-and-hold strategy.
Following Alexander’s (1964) study, Sweeney (1986) applies FR to foreign exchange markets and tests if the investors can obtain excess profits in Deutsche Mark. Although much literature points out FR can beat buy-and-hold strategy in many countries, those articles all neglect the risk and the statistical significance of the returns. Hence, Sweeney (1986) applies the terms of Capital Asset Price Model (CAPM) and compares buy-and-hold strategy and trading rules with considering risk premium. Sweeney (1986) found the smaller the ratio of FR, the more excess profits the trading rules can get related to buy-and-hold strategy, especially when the ratio of FR is 1%.
Besides FR, scholars also test other trading rules to see whether technical analysis is effective. Levy (1967) utilizes relative strength index (RSI) to test the stock price in New York Stock Exchange (NYSE) from 1960 to 1965. This method considers that past price is relative to future price, so if the stock performed strong in the past, it will perform relatively stronger in the future.
The result shows that RSI can beat buy-and-hold strategy and obtain excess
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
profits.
Jenson (1967) criticizes the view-point of Levy (1967). He argues the results of Levy (1967) are not strong and has selection biases. With consideration of more countries and transaction costs, these trading rules could not obtain excess profits, which shows technical analysis cannot beat the market.
In addition to simple trading rules, some other scholars utilize different trading rules combinations. Pruit and White (1988) created Cumulative Volume, Relative Strength, Moving Average (CRISMA) and conducted studies which resulted in that these strategies can obtain a better performance than buy-and-hold strategies regardless of the consideration of transaction cost.
Lo and MacKinlay (1990) quantify the biases of data snooping and focus on the evaluation of the relation between financial asset and data snooping. Data snooping biases exists because there is only one real history data. It is easy to find trading rules to obtain excess profits. However, the profits do not come from the effectiveness of the trading rules; instead they come from probabilities, which means people can find a trading rule obtaining excess profits when given a long enough period of time. But the trading rule can just prove that it was effective in the past and does not mean it will conduct the same results in the future. Therefore, many researchers began to have differing views concerning the effectiveness of these technical trading rules and utilized statistical test to prove if the trading rules can continuously work in the market.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Brock, Lakonishok and LeBaron (1992, BLL) used bootstrap to test two of the simplest and most popular trading rules - moving average and trading range break by utilizing the Dow Jones Index from 1897 to 1986. The overall purpose of this paper is to provide strong support for the technical strategies. The returns obtained from these strategies are not consistent with four popular null models: the random walk, the AR (1), the GARCH-M, and the Exponential GARCH.
Although BLL tried to solve data snooping biases, they still could not find a strong statistical explanation. White (2000) public Reality Check (RC), provides a more strictly statistical method to test the effectiveness of trading rules. Sullivan, Timmermann and White (1999, STW) utilized White’s Reality Check to evaluate five simple technical trading rules while quantifying the data-snooping biases and fully adjusting for its effect in the context of the full universe from which the trading rules were drawn. Hence, for the first time, STW(1999) expanded BLL(1992)’s research to universe of 26 kinds of technical trading rules, applied the rules to 100 years of daily data on the DJIA, and determined the effects of data-snooping. STW(1999) indicated that the performance of technical analysis are much better than buy-and-hold strategy, but they also found that adding more trading rules could not make the effectiveness more significant. Consequently, STW (1999) offered the possibility that all the excess profits from technical trading rules were just a coincidence. They had chosen exactly the best trading rules suitable for that period of time.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Although White’s Reality Check test can solve the problems of data snooping, Hansen (2005) points out two drawbacks of White’s Reality Check test. First, the average return does not standardize. Secondly, White’s RC is based on least favorable configuration, so the expected returns from the trading rules are all equal to the returns of the benchmark model. Therefore, White’s RC conducts a worse result because the test contains some worse trading rules.
In other words, although all the testing trading rules can beat the benchmark model, the p-value still goes up because the test also covers some of the trading rules with less explanation power, which makes the probability of rejecting null hypothesis decline.
As to foreign exchange markets, economic theories believe that exchange rates depend on the fundamental factors of countries, such as price, interest rate, currency supplies and national real income. However, Meese and Rogoff (1983) indicate that these traditional exchange rate models (Frenkel-Bilson Model, Dornbusch-Frankel Model and Hooper-Morton Model) have significant predictability only in the sample data but not outside the sample data, which even Random Walk Model can conduct a better prediction. Hooper (1997) demonstrates that the traditional economic model and fundamental factors cannot explain the changes of historical exchange rates. Engel and West (2005) even denote that exchange rates can affect microeconomics by Granger Causality Test instead. As a result, the predictability of the foreign exchange rate is still a puzzle.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the empirical results show trading rules make excess profits, then the weak form efficient market does not exist. Lin (2007) conducts back testing through program trading systems from 1986 to 2005. For the daily exchange rate data of 11 currencies, 8 frequently used technical analysis indicators are selected:
MA, MACD, KD, SAR, MTM, SAR, DMI and Channel. The empirical analysis indicates: 1. The trend-following trading systems is superior to the range trading systems. 2. Technical analysis cannot obtain excess returns.
Therefore, the weak form foreign exchange market efficiency stands, which means it is unprofitable applying those technical analysis indicators to foreign exchange markets.
Also, Chuang, Lin and Kuo (2011) utilize Hansen’s Superior Predictive Ability test (2005) to exam the performance of technical trading applied to NTD/USD foreign exchange market. This paper indicates that it is profitable using four popular technical rules in inter-day trading but not in intra-day trading with the exchange rate of NTD/USD. In other words, the profitability does not perform better when people use less time to make trading decisions, but technical analysis is still effective in some markets.
As a result, the main purpose of this paper is to extend and to enrich the earlier research on technical trading rules by applying CCI and WMS Index to Taiwan’s foreign exchange market. Consequently, this thesis demonstrates the empirical tests on using two commonly used indicators in Taiwan’s foreign exchange market and investors can obtain new technical trading rules from this research and increase the profitability of these two trading indicators.
‧
國立 政 治 大 學