OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

- AN INTRADAY STUDY ON THE SWISS STOCK MARKET

1. Introduction In addition to fundamental analysis, technical analysis is a widely used method to valuate securities. A study of Allen and Taylor (1990) shows that 90% of the traders at the London foreign exchange market employ chart techniques intensively and that 60% of the traders consider technical analysis to be as valuable as fundamental analysis. However, the extent of academic acceptance of the instruments of technical analysis is small in relation to their practical application. Technical analysis is frequently called voodoo nance by fundamental analysts and confronted with the accusation of a self-fullling prophecy, as discussed for example in Achelis (2000). One of the main criticisms is the alleged subjectivity of the methods, especially of the chart techniques. In addition, technical analysis is supposed to be rather an accumulation of separate methods than a self-contained system. In particular, Malkiel (2004) criticizes technical analysis: I am not saying that technical strategies never make money. (...) The point is rather that a simple buy-and-hold strategy (...) typically makes as much or more money. However, several academic studies about technical analysis have disproved the Random Walk hypothesis and have shown that historical prices can be used to predict future stock developments as summarized further below. The present article aims to evaluate the performance of dierent methods of technical analysis. It deals with the question whether an active portfolio management using technical analysis achieves a signicantly better performance than a passive portfolio in an intraday context. To our knowledge it is the rst study that evaluates the performance of technical strategy in an intraday setting. While fundamental analysis determines the fundamental value of an asset and expects the market price to return to it, technical analysis deals with the forces of supply and demand as well as the sentiment in the market. Therefore, it should be especially useful in the short range. Another reason for the use of intrady data is that the
Key words and phrases. technical analysis, performance, intraday, stock market.
1

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

technical indicators are calculated very rapidly while a thorough fundamental valuation may take days to carry out. Ten dierent technical indicators are applied whereas chart techniques are not analyzed because of the above mentioned subjectivity. The portfolios consist of 18 Swiss blue chip stocks that are invested in according to the buy- and sell-signals of the indicators. An unconditioned portfolio comprising the same stocks is used as a benchmark. The sample includes 65 trading days and intraday data at 5-minutes intervals is used. Thus, one trading day equals 101 data points. Despite the small scientic acceptance of technical analysis, some research has proven that several instruments are able to derive useful information from historical prices. Lo and MacKinlay (1988) have shown that past stock prices can be used to predict future returns. Indirect support for a successful use of technical analysis methods is expressed in the studies by Tabell and Tabell (1964), Treynor and Ferguson (1985), Brown and Jennings (1989) and Chan, Jegadeesh, and Lakonishok (1996). Direct support is given by the studies by Pruitt and White (1988), Neftci (1991), Brock, Lakonishok, and LeBaron (1992) and Osler and Chang (1995). Treynor and Ferguson (1985) develop one of the rst theoretical models which tries to establish technical analysis. In this model, the investor receives a private information at a certain point in time. Subsequently, the investor chooses portfolio strategies which allow him to hold a security long or short in order to liquidate it later. The research indicated that the analysis of historical prices can be helpful, however only in combination with further information. Brown and Jennings (1989) show a superior performance of portfolio strategies using technical analysis in a market environment which does not contain all relevant information. This leads to dierent informed investors and makes it possible to gain prots above average by the use of private information. A huge number of trading strategies are investigated in Conrad and Kaul (1998) but less than 50% show statistically signicant prots. Interestingly, the momentum strategy proves to be successful at time horizons from 3 to 12 month while contrarian strategies are apt in the long run. Osler and Chang (1995) deal with the exchange market and examine the eect of Head-and-Shoulders formations. Their research claims that this kind of chart technique delivers partially useful predictions. Technical analysis is not only concerned about prices but also about trading volumes of the stocks. One of the rst studies regarding the information content of volume is Blume, Easley, and OHara (1994), where volume itself showed the ability to predict price changes as well as in Gervais, Kaniel, and Mingelgrin (2001). This in contrast to Campbell, Grossman, and Wang (1991) and Harris and Raviv (1993), which relied on the correlation between volume and other variables. Brock, Lakonishok, and LeBaron (1992) analyze two of the most popular indicators of technical analysis: the Moving Average and the Trading Range Break-Out. Their study illustrates that higher returns tend to be followed by buy-signals than by sell-signals. Furthermore, returns after buy-signals were less volatile than after sell-signals. Their research also showed that prots after sell-signals were negative. In general, all studies claim that methods of technical analysis can outperform a buy-and-hold strategy. But usually, the returns are only above those of a passive portfolio if transaction costs are not considered as for example Alexander (1961) or Fama and Blume (1966) show.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

At the beginning of section 2 an overview of the ten technical indicators applied is presented. The denition of each indicator is given with a short explanation. Likewise, it will be shown under which market conditions the indicators will presumably achieve their best results. The explanations are mainly based on books for practitioners because of the lack of scientic literature. The selected indicators are grouped into trendfollowers, oscillators and indicators of the whole market. Section 2.2 gives a short introduction of the performance measures used, as well as the compilation of the portfolios. The calculated returns alone do not permit a performance evaluation since a higher return can be expected ex ante by taking a higher risk according to the capital market theory. The return of the portfolio must be related to a benchmark in order to consider the dierence between the prot of the portfolio and the prot of the benchmark at same risk. Hence, risk adjusted performance measures are required which will be illustrated here. The following section 2.3 presents the intraday data of the 18 stocks investigated. In section 3, the attained prots are presented and compared by the performance measures to a benchmark portfolio. Two subsamples are analyzed that conrm the results. The conclusion will answer the question whether an active portfolio management utilizing the analyzed technical indicators and intraday data is able to outperform a passive portfolio. 2. Methodology The following section 2.1 presents the technical indicators used in this article. Section 2.2 gives a brief overview of performance measurement. 2.1. Technical Indicators. The technical indicators may be roughly separated in trendfollowing indicators, oscillators and indicators of the whole market. Trendfollowing indicators are used to smooth out accidental movements of the stock prices in the short run which should make trends recognizable as Achelis (2000) points out. One has to keep in mind that trendfollowers always follow the trend which implies that they react with a time-lag. Therefore, trendfollowers are useful in market conditions with long-lasting trends while in volatile sidewardsmoving markets lots of erroneous signals are produced. The Simple Moving Average in time T is calculated according to the following formula: (2.1.1) SM AT = 1 n
T

Ps
s=T n+1

Ps denotes the stock price in time s which is interpolated from the previous tick. It is commonly calculated for 5 up to 200 trading days, as described in Achelis (2000). In our intraday context we estimate it over two trading days which means that n = 202. Whenever the stock price crosses the Simple Moving Average from below (above) a buy (sell) signal is generated. In contrast to the Simple Moving Average, the Weighted Moving Average considers more recent price movements as more relevant and gives them a higher weighting

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

as in Bauer and Dahlquist (1999):

(s T + n) Ps
n s=1

(2.1.2)

W M AT

s=T n+1

Again, n is set to 202, and buy- or sell-signals are generated whenever the stock price crosses the Weighted Moving Average. Since more weight is attributed to the most resent prices, the Weighted Moving Average is more volatile than the Simple Moving Average and yields, therefore, more buy- and sell-signals. The Exponential Moving Average gives the more recent prices also a higher weighting, but unlike the Weighted Moving Average it takes all available observations into account. In the present article it is calculated according to Achelis (2000): (2.1.3) EM AT = EM AT 1 + 2 (PT EM AT 1 ) n+1

As above, n equals 202, and the stock price crossing the Exponential Moving Average from below (above) is a signal to buy (sell). All the trendfollowing indicators produce lots of erroneous signals in a volatile sidewards moving market as Achelis (2000) points out. To delay the buy- or sellsignals and, therefore, to identify trends more accurately Envelopes of 1.5% are laid around the Simple Moving Average. Whenever the stock price crosses the upper envelope from below, a buy-signal is generated. If it crosses the lower envelope from above, it is a signal to sell. Oscillators are an alternative to trendfollowing indicators. They measure according to Colby and Meyers (1988) the momentum of changing prices. In contrast to the trendfollowers they perform best in volatile and sideward moving markets as Murphy (1999) states. The Rate of Change indicates the momentum and, therefore, the acceleration or slowing down of a trend. (2.1.4) RoCT = 100 PT PT n

According to Murphy (1999) the Rate of Change is usually calculated for ten days, but it is also used for 26 or 50 weeks. In the present article two trading days with n = 202 are used. Since the Rate of Change oscillates around 100, every crossing of 100 from below (above) generates a buy (sell) signal. The Volume Price Trend (also referred to as Price and Volume Trend) combines prices and trading volume:
T

(2.1.5)

V P TT

ln
s=T n+1

Ps Ps1

Vs

Vs denotes the number of shares traded between time s 1 and s. Since trading volume is always positive, the sign of the Volume Price Trend depends on the stock returns. A crossing of zero generates signals to trade. But as Achelis (2000) suggests, also the extremes of the Volume Price Trend may yield valuable information, since they show rising or declining momentum.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

The Relative Strength Index by Wilder (1979) is the most commonly used oscillator. 100 (2.1.6) RSIT = 100 1 + RST average of all gains RST = average of all losses RST denotes the relative strength in time T which is calculated as the average gains in periods with rising prices divided by the average losses in periods with declining markets. Originally, Wilder (1979) suggested to calculate the Relative Strength Index for 14 days while in the present article it comprises 202 ve minute time spaces. The limits for the buy- and sell-signals are set according to Murphy (1999) whenever the Relative Strength Index crosses 50. The following two indicators analyze the sentiment of the market as a whole. Therefore, they yield signals to invest in (or sell) the whole market and not a single stock. The Advance/Decline Ratio compares the number of rising and falling stock prices of a group of stocks. According to Achelis (2000) a broad trend is detected, whenever the majority of stocks follows a certain direction of an index. ADVT ADRT = (2.1.7) ADVT + DCLT As in Dukas and Park (1995) ADVT denotes the number of advancing stocks prices in the 202 ve-minute times spaces before time T . DCLT is the number of declining stock prices in the same time spaces. Usually, the Advance/Decline Ratio is used as a sort of contrarian indicator: Fosback (1991) considers values greater than 0.56 as bearish and values less than 0.43 as bullish. Achelis (2000) calculates the ADV Advance/Decline Ratio as DCLT and sets the limits at 1.25 and 0.9.1 T As well as the Advance/Decline Ratio, the Short-Term Trading Index analyzes the market as a whole. As Dukas and Park (1995) stress, it is the trading volume that is additionally taken into account: U P V OLT (2.1.8) ST T IT = U P V OLT + DN V OLT U P V OLT denotes the trading volume of all rising stocks while DN V OLT is the trading volume of the declining stocks. Buy- and sell-signals are generated whenever the Short-Term Trading Index crosse 0.5. 2.2. Performance Measurement. The buy- and sell-signals of every technical indicator are applied to each of the 18 stocks. Starting out from an equally weighted portfolio, the money is invested stocks when the technical indicator gives a signal to buy. After a sell-signal, the stock is sold and the money is redistributed in other stocks that currently yield a buy-signal. A cash position is created, whenever none of the 18 stocks shows a buy-signal. This is especially the case for the indicators which yield only signals for the market as a whole as the Advance/Decline Ratios and the Short-Term Trading Index. As transaction cost we add 0.1% and 0.2%, alternatively, for each trade. We measure the performance on a daily basis. Since two days are used to build up the technical indicators, the sample consists of 63
1Zweig (1990) states that current strength in the market implies future strength and treats the Advance/Decline Ratio as a momentum indicator.

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

trading days. The performance measurement is based on the CAPM, but we ignore the risk-free interest rate. This is justied because of the short time space covered and the very low interest rates, prevailing at the time. Whenever a technical indicator gives a signal to stay in cash, no interest is added which should, therefore, not distort the results of the performance measurement. We compare the return of the portfolios generated by the technical indicators to an equally weighted portfolio of the same 18 stocks. To take the risk into account, the volatilities are calculated out of the 63 daily returns and scaled to three month. To adjust returns for total risk, we calculate the Sharpe Ratio according to Sharpe (1966) and test whether it is signicantly dierent from the Sharpe Ratio of the benchmark portfolio according to Jobson and Korkie (1981). The test statistic is approximately standard normally distributed and calculated as (2.2.1) with = 1 2 2 [2 1 2 N
2 2 2 1 2 +

Z=

r1 2 r2 1 1 2 2 1 2 2 r + r 2 1 2 2 2 1

r1 r2 2 2 2 2 2 1 2 1 2 ]. 2 1 2

ri is the average return of the portfolio i, i denotes the volatility of portfolio i, the correlation between the returns of the two portfolios and N the number of observations. Additionally, the Jensens Alpha according to Jensen (1968), the Beta, and the Treynor Ratio which adjusts returns for systematic risk according to Treynor (1965) are calculated. The Tracking Error as a measure for unsystematic risk is the standard deviation of the residuals from the CAPM regression. Finally, the Treynor/Black Ratio as the adjustment of the Jensens Alpha for unsystematic risk according to Treynor and Black (1973) is presented. 2.3. Data. We use intraday data of 18 blue chip stocks that are part of the Swiss Market Index (SMI).2 The data consists of so-called order history reports which cover every single order how it entered the order book and if it was cleared or partially or fully traded. This sample consists of more than 2.7 million datapoints. Due to the huge amount of data, the sample size has to be limited to 65 trading days from May 2 until July 31, 2002. Furthermore, it was impossible to extract the order history reports of the largest Swiss stocks from the data storage system of the Swiss Exchange. Nevertheless, the choice of the sample ensures that analyst coverage of the stocks is given and that no small cap eects should distort the results. For every ve-minute time space, the last traded price and the volume is calculated. This time space is used in order to generate homogenous time series out of the randomly arriving orders. A shorter time space may not leave the trader enough time to calculate his technical indicators and to react.
2The 18 stocks are: Adecco, Julius Baer, Ciba Specialty Chemicals, Clariant, Givaudan, Hol-

cim, Kudelski, Lonza Group, Richemont, Serono, Sulzer, Surveillance, Swatch registered share, Swatch bearer share, Swiss Re, Swisscom, Syngenta and Unaxis.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

3. Empirical Results The next section presents the results for the whole sample period of 63 trading days. Section 3.2 shows the performance of the technical indicators in two subsamples of about equal length. 3.1. Whole Sample Period. Table 1 gives the returns of the long-only portfolios using the ten technical indicators. Starting out with an equally weighted portfolio, the money is invested in those stocks where the indicator yields a buy-signal. If none of the 18 stocks should be long, the money is invested in cash yielding no interest. In the rst column no transaction costs are added, in the second and third column 0.1% and 0.2%, respectively are added. Additionally, the returns of the Swiss Market Index (SMI) and an equally weighted portfolio of the 18 stocks are presented. The three month investigated showed a huge bear market with a return of the SMI of -23.88%. The benchmark portfolio yielded an even slightly lower return of -25.64%. Four of the ten indicators lead to a higher return than the benchmark portfolio without transaction costs. 0.1% Transaction Costs 0.2% Transaction Costs Return Return Volatility Return Volatility SMA -51.75% -176.05% 26.91% -300.20% 33.57% WMA -65.06% -220.35% 28.09% -375.45% 36.45% EMA -64.12% -204.31% 26.29% -344.32% 32.21% Envelopes -30.25% -41.49% 19.62% -52.73% 19.77% RoC -33.19% -148.81% 22.18% -264.26% 28.58% VPT -22.10% -61.70% 18.33% -101.24% 19.47% RSI -32.43% -153.23% 22.45% -273.88% 28.26% ADR Fosback 12.96% 12.86% 8.36% 12.76% 8.34% ADR Achelis -24.86% -24.96% 16.96% -25.06% 16.95% STTI -0.42% -8.42% 10.31% -16.42% 10.95% Benchmark -25.64% 17.20% SMI -23.88% 18.22% Table 1. Returns of the long-only portfolios without transaction costs, returns and volatilities with 0.1% and 0.2% transaction costs. Trades 239 322 284 21 110 84 221 1 1 80

The returns worsen considerably when transaction costs are added. But still the Envelopes, the Advance/Decline Ratios and the Short-Term Trading Index yield higher returns than the benchmark. A particularly bad return show the trendfollowing indicators without envelope, the Rate of Change, and the Relative Strength Index due to the excessive number of trades. The three month investigated were a period with a very high volatility of 18.22%. The volatilities of the rst seven portfolios are all above the volatility of the benchmark. Because of the investments in cash, the indicators for the market as a whole exhibit considerably lower volatilities than the benchmark. In general, higher trading costs augment the volatility. The last column in table 1 presents the average number of trades that has to be executed in each stock. Interestingly, the Advance/Decline Ratios trade only once and switch from cash to equity. The Weighted Moving Average shows the highest

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

number of trades since it gives much weight to the more recent stock prices, which makes it rather volatile. Table 2 presents the results of the performance measures of the portfolios with transaction costs of 0.1%. Only the Advance/Decline Ratio according to Fosback reaches with its positive return a positive Sharpe Ratio that is in addition signicantly higher than the one of the benchmark. The Sharpe Ratios of the Envelopes, the Advance/Decline Ratio according to Achelis and the Short-Term Trading Index are not signicantly dierent from the value of -1.46 of the benchmark portfolio. All other Sharpe Ratios are signicantly worse than the benchmarks Sharpe Ratio. Sharpe Beta Jensens Treynor Tracking Treynor/ Ratio Alpha Ratio Error Black Ratio SMA 6.54 1.26 144.37% -1.40 15.90% -9.08 WMA 7.84 1.44 184.12% -1.53 13.14% -14.01 EMA 7.77 1.37 169.97% -1.49 11.71% -14.52 Envelopes 2.12 0.98 16.78% -0.42 9.90% -1.70 RoC 6.71 0.91 125.95% -1.63 15.71% -8.02 VPT 3.37 0.92 38.62% -0.67 9.26% -4.17 RSI 6.83 0.99 128.32% -1.54 14.58% -8.80 ADR Fosback 1.54 0.22 18.44% 0.58 7.43% 2.48 ADR Achelis 1.47 0.97 0.59% -0.26 2.94% -0.20 STTI 0.82 0.34 0.23% -0.24 8.43% 0.03 Benchmark 1.46 1 0.00% -0.25 0.00% Table 2. Performance measures of the long-only portfolios with transaction costs of 0.1%. Results for the full sample. */**/*** denotes signicance on a level of 10%/5%/1%.

Due to the cash positions, the Betas of the Advance/Decline Ratio according to Fosback and the Short-Term Trading Index are considerably smaller than one. The Advance/Decline Ratio according to Fosback shows the lowest with 0.22 while the Advance/Decline Ratio according to Achelis behaves similar to the benchmark. A large Beta of 1.44 shows the Weighted Moving Average. This may be due to its volatility mentioned above. The Jensens Alpha of the Advance/Decline Ratio according to Fosback and the Short-Term Trading Index are the only positive ones; the value of 18.44% is statistically signicant on a level of 5%. The other eight technical indicators show all negative Alphas seven of them are signicant, at least at a level of 10%. Since the Advance/Decline Ratio after Fosback is the only indicator with a positive return, it is also the only one with a positive Treynor Ratio. The Advance/Decline Ratio according to Achelis shows the lowest Tracking Error because this strategy switches early from cash into the stock market where it remains. High Tracking Errors yield the trendfollowing indicators without the envelopes, the Rate of Change and the Relative Strength Index. In line with the Jensens Alphas, there are only the Advance/Decline Ratio according to Fosback and the Short-Term Trading Index that lead to a positive Treynor/Black Ratio.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

In general, only two of the the ten technical indicators are able to beat the benchmark portfolio on a risk-adjusted basis and after transaction costs of 0.1%: the Advance/Decline Ratio according to Fosback and the Short-Term Trading Index. The results worsen if the transaction costs are augmented to 0.2% as the results in table 3 show. The strategies that imply a lot of trading lead to considerably lower performance measures. Sharpe Beta Jensens Treynor Tracking Treynor/ Ratio Alpha Ratio Error Black Ratio SMA 8.94 1.44 263.95% -2.08 22.59% -11.69 WMA 10.30 1.76 331.29% -2.13 20.31% -16.31 EMA 10.69 1.57 304.92% -2.19 17.57% -17.35 Envelopes 2.67 0.99 27.89% -0.53 10.05% -2.78 RoC 9.25 0.97 240.01% -2.74 23.25% -10.32 VPT 5.20 0.93 77.86% -1.09 11.07% -7.04 RSI 9.69 1.03 247.99% -2.66 21.99% -11.28 ADR Fosback 1.53 0.22 18.37% 0.57 7.40% 2.48 ADR Achelis 1.48 0.97 0.69% -0.26 2.93% -0.24 STTI 1.50 0.32 8.36% -0.51 9.45% -0.88 Benchmark 1.46 1 0.00% -0.25 0.00% Table 3. Performance measures of the long-only portfolios with transaction costs of 0.2%. Results for the full sample. **/*** denotes signicance on a level of 5%/1%.

The Advance/Decline Ratio according to Fosback with only one trade shows still a signicantly positive Sharpe Ratio. All other Sharpe Ratios are smaller due to the augmented transaction costs. Higher transaction costs lead in general to slightly higher Betas as the second column in table 3 shows. The Advance/Decline Ratio according to Fosback is the only indicator that reaches a positive Jensens Alpha. It remains signicant. Seven out of the ten technical indicators lead to signicantly negative Alphas. In line with the returns, only the Advance/Decline Ratio according to Fosback generates a positive Treynor Ratio. The Tracking Errors are, in general, higher with the additional transaction costs. Only the indicators of the whole market that invest partially in cash lead to Tracking Errors below 10%. A positive Treynor-Black Ratio shows again the Advance/Decline Ratio according to Fosback due to its positive Jensens Alpha. 3.2. Subsamples. To conrm the results above, the sample is divided in two. The rst subperiod covers 31, the second one 32 trading days. The results are scaled to the full sample size of 63 trading days, in order to make comparison easier. Table 4 shows the returns, volatilities and number of trades of the technical indicators for the two subsamples. Transaction costs of 0.1% are taken into account. The SMI performed in the second subsample signicantly worse with a higher volatility than in the rst as the benchmark portfolio conrms. In the rst subsample, three of the ten indicators the indicators of the whole market where

10

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

Subsample 1 Subsample 2 Return Volatility Trades Return Volatility Trades SMA -169.99% 17.18% 136 -187.80% 34.08% 104 WMA -209.56% 18.61% 178 -238.25% 35.21% 144 EMA -197.65% 16.53% 160 -217.56% 33.41% 124 Envelopes -43.05% 7.47% 9 -41.27% 26.75% 12 RoC -161.26% 13.59% 62 -141.15% 28.30% 48 VPT -66.80% 8.92% 48 -58.58% 24.37% 36 RSI -167.82% 13.87% 126 -143.59% 28.55% 95 26.13% 11.60% 1 ADR Fosback 0.00% 0.00% 0 ADR Achelis -13.39% 7.39% 1 -37.33% 22.76% 0 STTI -12.28% 5.90% 50 -4.83% 13.35% 30 Benchmark -13.69% 8.49% -37.33% 22.76% Table 4. Returns, volatilities and number of trades of the longonly portfolios with transaction costs of 0.1%. Results for the two subsamples.

able to beat the benchmark, while in the second subsample the Advance/Decline Ratio according to Fosback and the Short-Term Trading index were able to do so. During the second subsample the Advance/Decline Ratio after Achelis was always invested in the stock market. Therefore, the gures of this portfolios are the same as for the Benchmark. The Advance/Decline Ratio according to Fosback remained during the rst 31 trading days always in cash which leads to zero return and zero volatility. All of the technical indicators yield higher volatilities in the second subsample. The indicators of the whole market show volatilities well below the one of the benchmark in both subperiods. Portfolios with high volatilities are generated by the trendfollowing indicators except the Envelopes, the Rate of Change and the Relative Strength Index. The Advance/Decline Ratios switch only once from cash to stock. The measure with the bounds according to Achelis switches rst in the rst subperiod while the Advance/Decline Ratio after Fosback gives the signal later capturing an uptrend in the market during the last six trading days. In general, the average number of trades is reduced in the second subsample which is due to the clearer market trend. The following tables 5 and 6 give the results of the performance measurement for the two subsamples. In the rst subsample, there are no positive Sharpe Ratios since no return is positive. Seven of them are signicant. The Advance/Decline Ratio according to Fosback remains always in cash which makes performance measurement impossible. The Betas range from 0.73 of the Envelopes to 1.64 of the Weighted Moving Average. All Jensens Alphas are negative with the same signicance as the Sharpe Ratios. There are only negative Treynor Ratios due to the negative returns. The largest Tracking Error shows with 13.61% the Simple Moving Average. Since all the Jensens Alphas are negative, also the Treynor/Black Ratios are negative. As a result of the clearer market trend with the reduced number of trades the performance in the second subsample is better than in the rst.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

11

Sharpe Beta Jensens Treynor Tracking Treynor/ Ratio Alpha Ratio Error Black Ratio SMA 9.89 1.23 153.08% -1.38 13.61% -11.25 WMA 11.26 1.64 187.18% -1.28 12.39% -15.10 EMA 11.96 1.41 178.39% -1.41 11.42% -15.62 Envelopes 5.77 0.73 33.06% -0.59 4.17% -7.93 RoC 11.87 0.86 149.50% -1.88 11.46% -13.04 VPT 7.49 0.92 54.16% -0.72 4.25% -12.73 RSI 12.10 1.01 154.05% -1.67 10.93% -14.09 ADR Fosback ADR Achelis 1.81 0.75 3.11% -0.18 3.74% -0.83 STTI 2.08 0.41 6.62% -0.30 4.74% -1.39 Benchmark 1.61 1 0.00% -0.14 0.00% Table 5. Performance measures of the long-only portfolios with transaction costs of 0.1%. Results for the rst subsample. *** denotes signicance on a level of 1%.

Sharpe Beta Jensens Treynor Tracking Treynor/ Ratio Alpha Ratio Error Black Ratio SMA 5.51 1.27 135.96% -1.48 18.01% -7.55 WMA 6.77 1.42 179.41% -1.68 13.89% -12.91 EMA 6.51 1.37 161.27% -1.59 12.06% -13.38 Envelopes 1.54 1.03 2.83% -0.40 12.98% -0.22 RoC 4.99 0.93 103.00% -1.51 18.70% -5.51 VPT 2.40 0.93 23.15% -0.63 12.10% -1.91 RSI 5.03 1.01 102.65% -1.42 16.98% -6.04 ADR Fosback 2.25 0.26 34.81% 0.99 9.94% 3.50 ADR Achelis 1.64 1 0.00% -0.37 0.00% STTI 0.36 0.34 7.65% -0.14 10.87% 0.70 Benchmark 1.64 1 0.00% -0.37 0.00% Table 6. Performance measures of the long-only portfolios with transaction costs of 0.1%. Results for the second subsample. **/*** denotes signicance on a level of 5%/1%.

The Advance/Decline Ratio according to Fosback is the only indicator with a positive Sharpe Ratio. It is signicant on a level of 1%. The Betas range from 1.42 of the Weighted Moving Average to 0.26 of the Advance/Decline Ratio according to Fosback. The Advance/Decline according to Achelis Ratio was during the second sample period always invested in the stock market, which leads to the same performance measures as the benchmark portfolio. Two of the Jensens Alphas are positive, but only the one of the Advance/Decline Ratio according to Fosback is signicant. This is also the only measure to reach a positive Treynor Ratio. The Tracking Errors in the second subsample are much higher: The Rate of Change strategy leads to a value of 18.70%. The positive Jensens Alphas of the Advance/Decline Ratios according to Fosback and the Short-Term Trading Index lead to positive Treynor/Black Ratios.

12

MANUEL AMMANN, MATTHIAS REKATE AND RICO VON WYSS

4. Summary and Conclusion Ten dierent technical indicators are used to form portfolios out of 18 stocks of the SMI using intraday data. The performance of the strategy shows a huge variance. Due to the excessive number of trades, yielding too many erroneous signals, the Moving Averages are never able to beat the benchmark portfolio on a transaction cost adjusted basis. The introduction of Envelopes around the Simple Moving Average improves its performance and leads in the second subsample to results similar to the benchmark. Rate of Change, Volume Price Trend and the Relative Strength Index always show a worse performance than the benchmark. The returns of the Short-Term Trading Index are above the benchmark in the whole sample period as well as in the subperiods. Its performance is about equal to the benchmark and lays in the second subperiod slightly above it, although not signicantly. The most promising results show the Advance/Decline Ratios. These two indicators give each only one buy-signal which keeps transaction costs low. The bounds according to Fosback lead to a superior performance with respect to the benchmark either in the whole sample or in the subperiods. It is the only indicator to reach a signicantly positive Jensens Alpha. The period investigated was subject to a huge bear market and the technical indicators may yield better results in other market conditions as the result in the second subsample with a clearer trend suggest. Sideward moving market as in the rst half of the sample lead to too many trades which makes a superior performance after transaction costs impossible. Acknowledgment. We thank the Swiss Exchange for providing the data. References
Achelis, S. B. (2000): Technical Analysis from A to Z. McGraw Hill, New York. Alexander, S. S. (1961): Price Movements in Speculative Markets: Trends or Random Walks, Industrial Management Review, 2(2), 726. Allen, H., and M. P. Taylor (1990): Charts, Noise and Fundamentals in the London Foreign Exchange Market, Economic Journal, 100(400), 4959. Bauer, R. J., and J. R. Dahlquist (1999): Technical Market Indicators. Wiley, New York. Blume, L. E., D. Easley, and M. OHara (1994): Market Statistics and Technical analysis: The role of volume, Journal of Finance, 49(1), 153181. Brock, W., J. Lakonishok, and B. LeBaron (1992): Simple Technical Trading Rules and the Stochastic Properties of Stock Returns, Journal of Finance, 47(5), 17311764. Brown, D., and R. Jennings (1989): On Technical Analysis, Review of Financial Studies, 2(4), 527551. Campbell, J. S., S. Grossman, and J. Wang (1991): Trading Volume and Serial Correlation in Stock Returns. Princeton University Press, Princeton. Chan, L. K., N. Jegadeesh, and J. Lakonishok (1996): Momentum Strategies, Journal of Finance, 51(5), 16811713. Colby, R. W., and T. A. Meyers (1988): The Encyclopedia of Technical Market Indicators. McGraw-Hill, New York. Conrad, J., and G. Kaul (1998): An Anatomy of Trading Strategies, Review of Financial Studies, 11(3), 489519. Dukas, S., and J. Park (1995): Trading Activity Indicators and Market Timing, Applied Financial Economics, 5, 337344.

OUTPERFORMANCE WITH TECHNICAL ANALYSIS?

13

Fama, E. F., and M. E. Blume (1966): Filter Rules and Stock-Market Trading, Journal of Business, 39(1, Part 2), 226241. Fosback, N. G. (1991): Stock Market Logic: A Sophisticated Approach to Prots on Wall Street. Dearborn Trade Publishing, Chicago. Gervais, S., R. Kaniel, and D. H. Mingelgrin (2001): The High-Volume Return Premium, Journal of Finance, 56(3), 877919. Harris, M., and A. Raviv (1993): Dierences of Opinion Make a Horse Race, Review of Financial Studies, 6(3), 473506. Jensen, M. C. (1968): The Performance of Mutual Funds in the Period 1945-1964, Journal of Finance, 23(2), 389416. Jobson, J., and B. Korkie (1981): Performance Hypothesis Testing with the Sharpe and Treynor Measures, Journal of Finance, 36(4), 889908. Lo, A. W., and A. C. MacKinlay (1988): Stock Market Prices Do Not Follow Random Walks: Evidence from a Simple Specication Test, Review of Financial Studies, 1(1), 4166. Malkiel, B. G. (2004): A Random Walk down Wall Street. W W Norton & Company, New York. Murphy, J. J. (1999): Technical Analysis of the Financial Markets: A Comprehensive Guide to Trading Methods and Applications. Paramus, New York. Neftci, S. N. (1991): Naive Trading Rules in Financial Markets and Wiener-Kolmogorov Prediction Theory: A Study of Technical Analysis, Journal of Business, 64(4), 549571. Osler, C., and K. Chang (1995): Head and Shoulders: Not Just a Flaky Pattern, Sta Report No. 4, Federal Reserve Bank of New York. Pruitt, S. W., and R. E. White (1988): The CRISMA Trading System: Who Says Technical Analysis Cant Beat the Market?, Journal of Portfolio Management, 14(3), 5558. Sharpe, W. F. (1966): Mutual Fund Performance, Journal of Business, 39(1), 119138. Tabell, E. W., and A. W. Tabell (1964): The Case for Technical Analysis, Financial Analysts Journal, 20(2), 6776. Treynor, J. (1965): How to Rate Management of Investment Funds, Harvard Business Review, 43(1), 6375. Treynor, J., and F. Black (1973): How to Use Security Analysis to Improve Portfolio Selection, Journal of Business, 46(1), 6686. Treynor, J., and R. Ferguson (1985): In Defense of Technical Analysis, Journal of Finance, 40(3), 757773. Wilder, J. W. (1979): New Concepts in Technical Trading Systems. Trend Research, Greensboro. Zweig, M. (1990): Winning on Wall Street. Warner Books, New York. Swiss Institute for Banking and Finance, Rosenbergstrasse 52, CH-9000 St. Gallen E-mail address: [email protected]