By Catherine Shalen, Senior Economist, Chicago Board of Trade
There is an intriguing revival of academic and popular interest in trading theories based on the Dow Jones Industrial Average.1 Part of this interest was undoubtedly spurred by the increased visibility of the Dow. The Dow reached the 10,000 milestone in mid-March of 1999 after a period of unusual volatility and dramatic price swings.2 A second factor is that, during the last few years, financial economists have begun to reassess the long-standing "efficient markets" paradigm and to test the validity of various technical trading rules. Third, the introduction of low-cost Dow derivativessCBOT. DJIA futures and options on futures and CBOE DJIA optionsshas facilitated implementation of these trading rules. This article reviews this recent empirical evidence for two well-known Dow trading strategies. The first is the Dow Theory, perhaps the oldest technical trading strategy, and the second is the newer Dow-10 strategy, also called the "Dogs of the Dow" strategy.
The Dow Theory
The Dow Theory is a market timing strategy outlined in a series of Wall Street Journal editorials written by William Hamilton from 1902 to 1929. It is described in greater detail in Hamilton's book, The Stock Market Barometer (1922). The essential elements of the theory, as interpreted by successive generations of Dow theorists, are that (1) the stock market alternates between bull and bear trends, and these trends are persistent; (2) bull and bear trends develop with a change in the fundamental outlook of investors, which is soon reflected in stock price changes; (3) these trends ultimately peak after a period of market overreaction to the initial news, and most critically; (4) bull and bear trends can be forecast from patterns of the past values of the Dow Jones Industrial and Transportation indexes. Specifically, large price moves of the industrial and transportation indexes that occur after an extended period of variation within a trading range signal a switch from a bull to a bear regime or vice versa.
Alfred Cowles (1934) performed the first formal test of the Dow Theory but concluded the theory did not work. The return on a portfolio based on Hamilton's market calls lagged the return on a portfolio fully invested in Cowles' market index, a predecessor of the S&P 500. Cowles' study laid the foundation for the random walk hypothesis and the efficient market theory that dominated the academic scene until about 1990. Brown, Goetzmann, and Kumar, Journal of Finance, August 1998 (BGK), have reexamined Cowles' test. Unlike Cowles, BGK adjust returns for risk using two standard measuressSharpe's performance ratio (the ratio of excess return relative to a short-term, nearly riskless rate of interest and of volatility) and Jensen's alpha (the excess return relative to the expected return given the risk of the security, where risk is measured by sensitivity to a broad-based market proxy). They apply a variety of parametric and nonparametric tests to Cowles' data and conclude that the Dow Theory indeed appears to work, especially up to 1930. A statistical analysis of Hamilton's market calls (bear, bull, or neutral) suggests he had the ability to time bear markets.
BGK estimate that a portfolio of stocks (long or short) and Treasury bills rebalanced monthly according to Hamilton's market calls has a higher Sharpe ratio than a portfolio indexed to Cowles' index. Furthermore, the alpha of the Hamilton portfolio is positive and equal to 4.04%, and the portfolio return is less volatile than the return on a portfolio fully invested in the Cowles index.
Since Hamilton's editorials only provide imprecise clues about his timing rules, BGK search for models that most closely fit these rules using stepwise regressions of Hamilton's bear calls on lagged values of the industrial and transportation averages and on products of these lagged values. They also search for all possible "topological" patterns such as head and shoulder patterns and resistance levels that might predict Hamilton's calls. The types of predictive patterns that emerge from this analysis are recent peaks, which predict sell signals, and recent increases, which predict buy signals, as do recoveries from recent declines.
BGK's simulations confirm standard interpretations of the Dow Theory as a momentum theory. For example, negative 60-day returns of the industrial and transportation averages predict bear calls, and there are systematic links between his calls and return patterns that exhibit some form of persistence. The out-of-sample performance (9/1/30-12/1/97) of BGK's reconstructed Dow trading rules models is not consistent from decade to decade but tends to do better in severe bear markets.
Beyond its historical interest, BGK's fresh look at the Dow Theory provides interesting evidence of the effectiveness of technical approaches, in this case, timing strategies based on persistence in bull and bear trends. Numerous studies of timing ability have appeared in financial journals since Cowles' original article, and the majority found only weak evidence of successful timing, particularly after taking into account risk exposure and transaction costs. The issue is far from settled, however. While several recent academic studies also find returns on market timing strategies to lag returns on passive holding strategies, others find evidence in favor of timing. For example, an exhaustive analysis by Brock, Lakonishok, and LeBaron of technical rules applied to the DJIA yielded 26 rules that outperformed a passive cash holding strategy. Furthermore, their results stood up to data snooping screens conducted by Sullivan, Timmermann, and White. The latter expanded the universe of trading rules to 8000 parametrizations, which they applied to the DJIA over the period 1897-1986. Even after controlling for data snooping biases, certain trading rules outperformed the benchmark.
There is also evidence that certain mutual funds display timing ability (Bello and Janjigian), and that some investment advisors possess "hot hands" (Harvey and Graham). Furthermore, new timing strategies are still being proposed. Thus, in "Market Timing: Style and Size Rotation Using the VIX," M.M. Copeland and T.E. Copeland report that changes in the CBOE Market Volatility Index (VIX) predict the performance of the spread between large capitalization and small capitalization stocks and also the performance of the spread between portfolios of value and portfolios of growth stocks. Last, while BGK suggested that transaction costs for stocks offset the excess return of their portfolio, the transaction costs of trading futures on the DJIA would not.
Dogs of the Dow
A newer, and quite popular, theory of Dow performance is the Dow-10 strategy. Michael O'Higgins described this trading strategy in 1992 in "Beating the Dow." The central premise is that an equally weighted portfolio of the 10 DJIA stocks with the highest dividend yieldssthe so-called Dogs of the Dow (Dogs)swill outperform the DJIA. The strategy continues to appeal to investors, and a number of web sites, mutual funds, and investment unit trusts follow Dow-10 strategies. In January 1999, CBOE launched a new option contract on the MUT, an equally-weighted index of the Dow-10 stocks. The MUT index is reconstituted at the beginning of every year from the 10 DJIA stocks with the highest dividend yield in the previous quarter.
Since dividend yields are ratios of dividends to share prices, a possible rationale for a Dow-10 strategy is that high dividend yields signal that the stocks are underpriced. Higher returns are presumably generated once this mispricing is corrected. The remarkable continuity in the set of stocks classified as Dogs implies such corrections might take a long time. Only two to three stocks change each year, and Dogs remain Dogs for multiple years. In the last 10 years, the Dogs have included only 17 different stocks. A Dow-10 portfolio would therefore have to be held for a long time to generate an excess return from pricing corrections. A second possible explanation for greater returns on Dow-10 stocks is that high dividend yields are correlated with a risk factor, such as return volatility, or exposure to market risk.
Review of Empirical Evidence
Either of these explanations may be academic because the empirical evidence in favor of the Dow-10 strategy has become less convincing. McQueen, in Financial Analysts Journal, July-August 1997, calculated that from 1946 to 1995, the average annual return of the Dow-10 was 16.77% (standard deviation 19.10%), while the average return of the DJIA was 13.71% (standard deviation 16.64%); the excess return of a Dow-10 portfolio over the DJIA was statistically significant in 32 of these 50 years. On the other hand, McQueen determined that the difference in returns was not economically significant over the whole period; it was only significant in two subperiods, from 1966 to 1975 and 1976 to 1985. Overall, the premium of the Dogs was offset by risk differences, transaction costs, and the higher tax rate on dividends. Variations of the strategy based on alternative dividend yield calculations, rebalancing the portfolio at different times of the year, and using four, five, or 15 instead of 10 stocks fared no better.3 McQueen also warned against the possibility that the outperformance of the Dogs from 1946 to 1995 might be the result of data mining or that it would disappear once investors attempted to profit from it.
TABLE 1
NOV 88 - MAR 99 DJIA MUTS MUT - DJIA
Mean Monthly 1.53% 1.33% -0.19%
STD. DEV. 3.90% 3.62% 1.99%
T-Stat 0.39 0.37 -0.10
Median 1.93% 1.50% -0.80%
MIN -14.88% -12.07% -5.60%
MAX 10.28% 8.09% 4.21%
1989 - 1998
Mean Annual Return -1.72%
STD. DEV.
8.96%
T-Stat -0.19
Median -1.01%
MIN -13.27
MAX 13.90%
Chart 1 shows that, since 1989, the Dow-10 strategy perceptibly outperformed the DJIA in only two years, 1991 and 1993. The spread between MUT and DJIA returns was positive in five out of 10 years, and it was close to zero in four years. The graph also shows that the Dow-10 strategy was not much more successful for shorter holding periods. A comparison of monthly and yearly total returns (Table 1) of the MUT and DJIA from November 1988 through March 1999 confirms that Dow-10 stocks have not outperformed the DJIA in the last decade. Over this period, the mean monthly return of the MUT was 18 basis points lower than the mean monthly return of the DJIA; the mean yearly return of the MUT was 1.72% lower than that of the DJIA. Margins in both directions were slim, and, statistically, mean monthly and mean yearly returns of the MUT and DJIA were equivalent.
Changes in the performance of the Dow-10 relative to DJIA stocks appear to mirror shifts in the relative performance of value and growth stocks. This is understandable since Dow-10 stocks tend to be value stockssnine of the 1999 batch are so classified, the exception being Philip Morrisswhile DJIA stocks are half value and half growth stocks. Empirical studies find that from 1960 to 1992 returns on value stocks were higher than those on growth stocks. This "value" factor contributed to the higher return of the Dow-10 during that period.4 In the last few years, however, the trend has reversed and value stocks have lagged growth stocks. The spread between their returns widened considerably starting in 1997. To illustrate, the S&P 500 Growth Index had a total return of 36.33% in 1997 and 41.85% in 1998. The S&P 500 Value Index had a total return of 29.63% in 1997 and 14.51% in 1998.
A comparable spread has emerged between the DJIA and Dow-10 returns. Chart 2, a graph of compounded returns of the DJIA and the MUT since the end of 1988, shows that until June 1997, compounded returns on the DJIA and MUT were nearly indistinguishable. Since then, their paths have parted. Furthermore, the spread has been widening since August 1998. While this bodes poorly for the near-term success of the classical Dow-10 strategy, it suggests alternative Dow-10-based spread strategies.
An Alternative Dow-10 Strategy
Chart 3 provides some insight concerning the spread between the DJIA and Dow-10 stocks. The chart is a scatter plot of monthly returns of the spread difference between the MUT against monthly returns of the DJIA from November 1988 through March 1999. It shows an inverse relationship between the relative performance of the MUT and the direction of the DJIA. The MUT tends to do better than the DJIA when the DJIA has very low returns and vice versa. Equivalently, the MUT is less sensitive than the DJIA to extreme market movements. This lower sensitivity is also seen in the narrower range of MUT returns in Table 1 and in smaller S&P 500-based betas of the MUT. By these measures, Dow-10 stocks are less risky than the DJIA .
The inverse relationship between the DJIA and the MUT-DJIA spread has strengthened over time. In Chart 3, this is reflected in the steeper negative slope of the trend line since 1994. Regressions of monthly returns of the MUT-DJIA spread on the DJIA from November 1988 to December 1993 and from January 1994 through March 1999 point to the same effect. Coefficients of the spread return are negative in both regressions, but the coefficient in the later period is more significant and the regression has a higher squared correlation.
This pattern implies that spreading Dow-10 stocks short (long) against long (short) CBOT. DJIA futures is a relatively low-risk play on extreme bear (bull) market moves. The spread is implemented fairly easily with CBOT.DJIA futures and either a basket of the Dow-10 stocks or a synthetic MUT futures. This spread is well adapted to the current market, in which there is widespread apprehension that the perceived U.S. stock price bubble will burst.
Conclusion
Another look at seemingly discredited Dow theories reveals these theories capture persistent regularities in stock returns. While financial economists are still elucidating what factors might be generating such regularities, the low transaction costs of DJIA derivatives enable investors to implement a variety of trading strategies inspired by the original Dow theories.
Two of the latest articles bearing on the Dow published in the October 1999 issue of the Journal of Finance are "Data-Snooping, Technical Trading Rule Performance and the Bootstrap," Sullivan, Timmermann, and White, "What is the Intrinsic Value of the Dow?" Lee, Myers, and Swaminathan.
The Dow has been fluctuating around 11,000 for several months. Many market participants fear a crash but Ibbotson, who predicted that it would reach 10,000 by November 1999, is now forecasting a level of 100,000 by early 2024.
O'Higgins originally proposed that the Dogs portfolio should be rebalanced annually, but the timing provisions of the Taxpayer Relief Act of 1997 suggest that increasing the rebalancing period from 12 to 18 months might improve the after-tax return of the Dogs.
Two alternative explanations were proposed for the premium of value over growth stocks. The first attributed the premium to risk differences between the two classes of stocks (Fama and French, 1992). The second attributed the premium to market inefficiencies (Lakonishok and Chan, 1992). Lakonishok and Chan pinpointed the source of the difference in returns of growth and value stocks to different market reactions to their earnings announcements. Investors revised expectations upwards (downwards) following earnings announcements of value (growth) stocks, and value stocks therefore outperformed growth stocks after such announcements.
There is an intriguing revival of academic and popular interest in trading theories based on the Dow Jones Industrial Average.1 Part of this interest was undoubtedly spurred by the increased visibility of the Dow. The Dow reached the 10,000 milestone in mid-March of 1999 after a period of unusual volatility and dramatic price swings.2 A second factor is that, during the last few years, financial economists have begun to reassess the long-standing "efficient markets" paradigm and to test the validity of various technical trading rules. Third, the introduction of low-cost Dow derivativessCBOT. DJIA futures and options on futures and CBOE DJIA optionsshas facilitated implementation of these trading rules. This article reviews this recent empirical evidence for two well-known Dow trading strategies. The first is the Dow Theory, perhaps the oldest technical trading strategy, and the second is the newer Dow-10 strategy, also called the "Dogs of the Dow" strategy.
The Dow Theory
The Dow Theory is a market timing strategy outlined in a series of Wall Street Journal editorials written by William Hamilton from 1902 to 1929. It is described in greater detail in Hamilton's book, The Stock Market Barometer (1922). The essential elements of the theory, as interpreted by successive generations of Dow theorists, are that (1) the stock market alternates between bull and bear trends, and these trends are persistent; (2) bull and bear trends develop with a change in the fundamental outlook of investors, which is soon reflected in stock price changes; (3) these trends ultimately peak after a period of market overreaction to the initial news, and most critically; (4) bull and bear trends can be forecast from patterns of the past values of the Dow Jones Industrial and Transportation indexes. Specifically, large price moves of the industrial and transportation indexes that occur after an extended period of variation within a trading range signal a switch from a bull to a bear regime or vice versa.
Alfred Cowles (1934) performed the first formal test of the Dow Theory but concluded the theory did not work. The return on a portfolio based on Hamilton's market calls lagged the return on a portfolio fully invested in Cowles' market index, a predecessor of the S&P 500. Cowles' study laid the foundation for the random walk hypothesis and the efficient market theory that dominated the academic scene until about 1990. Brown, Goetzmann, and Kumar, Journal of Finance, August 1998 (BGK), have reexamined Cowles' test. Unlike Cowles, BGK adjust returns for risk using two standard measuressSharpe's performance ratio (the ratio of excess return relative to a short-term, nearly riskless rate of interest and of volatility) and Jensen's alpha (the excess return relative to the expected return given the risk of the security, where risk is measured by sensitivity to a broad-based market proxy). They apply a variety of parametric and nonparametric tests to Cowles' data and conclude that the Dow Theory indeed appears to work, especially up to 1930. A statistical analysis of Hamilton's market calls (bear, bull, or neutral) suggests he had the ability to time bear markets.
BGK estimate that a portfolio of stocks (long or short) and Treasury bills rebalanced monthly according to Hamilton's market calls has a higher Sharpe ratio than a portfolio indexed to Cowles' index. Furthermore, the alpha of the Hamilton portfolio is positive and equal to 4.04%, and the portfolio return is less volatile than the return on a portfolio fully invested in the Cowles index.
Since Hamilton's editorials only provide imprecise clues about his timing rules, BGK search for models that most closely fit these rules using stepwise regressions of Hamilton's bear calls on lagged values of the industrial and transportation averages and on products of these lagged values. They also search for all possible "topological" patterns such as head and shoulder patterns and resistance levels that might predict Hamilton's calls. The types of predictive patterns that emerge from this analysis are recent peaks, which predict sell signals, and recent increases, which predict buy signals, as do recoveries from recent declines.
BGK's simulations confirm standard interpretations of the Dow Theory as a momentum theory. For example, negative 60-day returns of the industrial and transportation averages predict bear calls, and there are systematic links between his calls and return patterns that exhibit some form of persistence. The out-of-sample performance (9/1/30-12/1/97) of BGK's reconstructed Dow trading rules models is not consistent from decade to decade but tends to do better in severe bear markets.
Beyond its historical interest, BGK's fresh look at the Dow Theory provides interesting evidence of the effectiveness of technical approaches, in this case, timing strategies based on persistence in bull and bear trends. Numerous studies of timing ability have appeared in financial journals since Cowles' original article, and the majority found only weak evidence of successful timing, particularly after taking into account risk exposure and transaction costs. The issue is far from settled, however. While several recent academic studies also find returns on market timing strategies to lag returns on passive holding strategies, others find evidence in favor of timing. For example, an exhaustive analysis by Brock, Lakonishok, and LeBaron of technical rules applied to the DJIA yielded 26 rules that outperformed a passive cash holding strategy. Furthermore, their results stood up to data snooping screens conducted by Sullivan, Timmermann, and White. The latter expanded the universe of trading rules to 8000 parametrizations, which they applied to the DJIA over the period 1897-1986. Even after controlling for data snooping biases, certain trading rules outperformed the benchmark.
There is also evidence that certain mutual funds display timing ability (Bello and Janjigian), and that some investment advisors possess "hot hands" (Harvey and Graham). Furthermore, new timing strategies are still being proposed. Thus, in "Market Timing: Style and Size Rotation Using the VIX," M.M. Copeland and T.E. Copeland report that changes in the CBOE Market Volatility Index (VIX) predict the performance of the spread between large capitalization and small capitalization stocks and also the performance of the spread between portfolios of value and portfolios of growth stocks. Last, while BGK suggested that transaction costs for stocks offset the excess return of their portfolio, the transaction costs of trading futures on the DJIA would not.
Dogs of the Dow
A newer, and quite popular, theory of Dow performance is the Dow-10 strategy. Michael O'Higgins described this trading strategy in 1992 in "Beating the Dow." The central premise is that an equally weighted portfolio of the 10 DJIA stocks with the highest dividend yieldssthe so-called Dogs of the Dow (Dogs)swill outperform the DJIA. The strategy continues to appeal to investors, and a number of web sites, mutual funds, and investment unit trusts follow Dow-10 strategies. In January 1999, CBOE launched a new option contract on the MUT, an equally-weighted index of the Dow-10 stocks. The MUT index is reconstituted at the beginning of every year from the 10 DJIA stocks with the highest dividend yield in the previous quarter.
Since dividend yields are ratios of dividends to share prices, a possible rationale for a Dow-10 strategy is that high dividend yields signal that the stocks are underpriced. Higher returns are presumably generated once this mispricing is corrected. The remarkable continuity in the set of stocks classified as Dogs implies such corrections might take a long time. Only two to three stocks change each year, and Dogs remain Dogs for multiple years. In the last 10 years, the Dogs have included only 17 different stocks. A Dow-10 portfolio would therefore have to be held for a long time to generate an excess return from pricing corrections. A second possible explanation for greater returns on Dow-10 stocks is that high dividend yields are correlated with a risk factor, such as return volatility, or exposure to market risk.
Review of Empirical Evidence
Either of these explanations may be academic because the empirical evidence in favor of the Dow-10 strategy has become less convincing. McQueen, in Financial Analysts Journal, July-August 1997, calculated that from 1946 to 1995, the average annual return of the Dow-10 was 16.77% (standard deviation 19.10%), while the average return of the DJIA was 13.71% (standard deviation 16.64%); the excess return of a Dow-10 portfolio over the DJIA was statistically significant in 32 of these 50 years. On the other hand, McQueen determined that the difference in returns was not economically significant over the whole period; it was only significant in two subperiods, from 1966 to 1975 and 1976 to 1985. Overall, the premium of the Dogs was offset by risk differences, transaction costs, and the higher tax rate on dividends. Variations of the strategy based on alternative dividend yield calculations, rebalancing the portfolio at different times of the year, and using four, five, or 15 instead of 10 stocks fared no better.3 McQueen also warned against the possibility that the outperformance of the Dogs from 1946 to 1995 might be the result of data mining or that it would disappear once investors attempted to profit from it.
TABLE 1
NOV 88 - MAR 99 DJIA MUTS MUT - DJIA
Mean Monthly 1.53% 1.33% -0.19%
STD. DEV. 3.90% 3.62% 1.99%
T-Stat 0.39 0.37 -0.10
Median 1.93% 1.50% -0.80%
MIN -14.88% -12.07% -5.60%
MAX 10.28% 8.09% 4.21%
1989 - 1998
Mean Annual Return -1.72%
STD. DEV.
8.96%
T-Stat -0.19
Median -1.01%
MIN -13.27
MAX 13.90%
Chart 1 shows that, since 1989, the Dow-10 strategy perceptibly outperformed the DJIA in only two years, 1991 and 1993. The spread between MUT and DJIA returns was positive in five out of 10 years, and it was close to zero in four years. The graph also shows that the Dow-10 strategy was not much more successful for shorter holding periods. A comparison of monthly and yearly total returns (Table 1) of the MUT and DJIA from November 1988 through March 1999 confirms that Dow-10 stocks have not outperformed the DJIA in the last decade. Over this period, the mean monthly return of the MUT was 18 basis points lower than the mean monthly return of the DJIA; the mean yearly return of the MUT was 1.72% lower than that of the DJIA. Margins in both directions were slim, and, statistically, mean monthly and mean yearly returns of the MUT and DJIA were equivalent.
Changes in the performance of the Dow-10 relative to DJIA stocks appear to mirror shifts in the relative performance of value and growth stocks. This is understandable since Dow-10 stocks tend to be value stockssnine of the 1999 batch are so classified, the exception being Philip Morrisswhile DJIA stocks are half value and half growth stocks. Empirical studies find that from 1960 to 1992 returns on value stocks were higher than those on growth stocks. This "value" factor contributed to the higher return of the Dow-10 during that period.4 In the last few years, however, the trend has reversed and value stocks have lagged growth stocks. The spread between their returns widened considerably starting in 1997. To illustrate, the S&P 500 Growth Index had a total return of 36.33% in 1997 and 41.85% in 1998. The S&P 500 Value Index had a total return of 29.63% in 1997 and 14.51% in 1998.
A comparable spread has emerged between the DJIA and Dow-10 returns. Chart 2, a graph of compounded returns of the DJIA and the MUT since the end of 1988, shows that until June 1997, compounded returns on the DJIA and MUT were nearly indistinguishable. Since then, their paths have parted. Furthermore, the spread has been widening since August 1998. While this bodes poorly for the near-term success of the classical Dow-10 strategy, it suggests alternative Dow-10-based spread strategies.
An Alternative Dow-10 Strategy
Chart 3 provides some insight concerning the spread between the DJIA and Dow-10 stocks. The chart is a scatter plot of monthly returns of the spread difference between the MUT against monthly returns of the DJIA from November 1988 through March 1999. It shows an inverse relationship between the relative performance of the MUT and the direction of the DJIA. The MUT tends to do better than the DJIA when the DJIA has very low returns and vice versa. Equivalently, the MUT is less sensitive than the DJIA to extreme market movements. This lower sensitivity is also seen in the narrower range of MUT returns in Table 1 and in smaller S&P 500-based betas of the MUT. By these measures, Dow-10 stocks are less risky than the DJIA .
The inverse relationship between the DJIA and the MUT-DJIA spread has strengthened over time. In Chart 3, this is reflected in the steeper negative slope of the trend line since 1994. Regressions of monthly returns of the MUT-DJIA spread on the DJIA from November 1988 to December 1993 and from January 1994 through March 1999 point to the same effect. Coefficients of the spread return are negative in both regressions, but the coefficient in the later period is more significant and the regression has a higher squared correlation.
This pattern implies that spreading Dow-10 stocks short (long) against long (short) CBOT. DJIA futures is a relatively low-risk play on extreme bear (bull) market moves. The spread is implemented fairly easily with CBOT.DJIA futures and either a basket of the Dow-10 stocks or a synthetic MUT futures. This spread is well adapted to the current market, in which there is widespread apprehension that the perceived U.S. stock price bubble will burst.
Conclusion
Another look at seemingly discredited Dow theories reveals these theories capture persistent regularities in stock returns. While financial economists are still elucidating what factors might be generating such regularities, the low transaction costs of DJIA derivatives enable investors to implement a variety of trading strategies inspired by the original Dow theories.
Two of the latest articles bearing on the Dow published in the October 1999 issue of the Journal of Finance are "Data-Snooping, Technical Trading Rule Performance and the Bootstrap," Sullivan, Timmermann, and White, "What is the Intrinsic Value of the Dow?" Lee, Myers, and Swaminathan.
The Dow has been fluctuating around 11,000 for several months. Many market participants fear a crash but Ibbotson, who predicted that it would reach 10,000 by November 1999, is now forecasting a level of 100,000 by early 2024.
O'Higgins originally proposed that the Dogs portfolio should be rebalanced annually, but the timing provisions of the Taxpayer Relief Act of 1997 suggest that increasing the rebalancing period from 12 to 18 months might improve the after-tax return of the Dogs.
Two alternative explanations were proposed for the premium of value over growth stocks. The first attributed the premium to risk differences between the two classes of stocks (Fama and French, 1992). The second attributed the premium to market inefficiencies (Lakonishok and Chan, 1992). Lakonishok and Chan pinpointed the source of the difference in returns of growth and value stocks to different market reactions to their earnings announcements. Investors revised expectations upwards (downwards) following earnings announcements of value (growth) stocks, and value stocks therefore outperformed growth stocks after such announcements.