Updated February 2021
TACTICAL ASSET ALLOCATION
Abstract: Tactical Asset Allocation (TAA) is a subset of active investment strategies and is sometimes promoted for index funds. TAA shifts asset class weights based on various moving average crossovers and other statistical "technical" indicators on market statistics. Although in selected time samples TAA shows some value when larger samples are studied DFA has found that they do not produce returns higher than buy and hold strategies.
After large market declines like that which occurred in the 2008 financial crisis, active managers often promote the idea that their trading systems and the special knowledge they possess would have allowed investors to avoid the decline or minimize losses. Timing when to own and when not own a given industry, or market sector, or stock, or major asset class or index like the S&P 500 or Russell 2000 is certainly a tempting proposition. As noted elsewhere on this website in "Active vs. Passive Management" there is little empirical evidence to support the idea that active management strategies, including tactical asset allocation, are likely to outperform passive buy-and-hold strategies. If active management did work knowing when to own or not own a given asset class would certainly reduce losses and increase returns.
The term "Tactical Asset Allocation" has been given to a particular type of market timing and trading. It is a subset of active investment strategy and involves shifting money between various asset classes, equities, fixed income and cash based upon various statistical strategies and usually based on price changes and volume. It is sometimes done with passive and index funds. DFA advocates "buy-and-hold" strategies with highly diversified multiple asset classes in passive portfolios. This article examines asset class timing in general, a few tactical asset allocation systems which employ it, and evidence for their advantage, if any, over passive and index buy-and-hold strategies.
Extensive research has found that active managers in the aggregate fail to outperform passive buy-and-hold strategies. When DFA examined the percentage of active equity funds which failed to beat their respective indexes from July 2004 through June 2009 they found 63% of active managers underperformed US large cap growth indexes and 90% underperformed emerging markets. In bonds, active managers underperformed 100% of the time. This period includes a strong bull market, a huge bear market decline and subtantial shifts in interest rates, a challenge for active management but then that is where they claim an advantage.
Many studies indicate that talented or lucky managers and funds which outperformed indexes in the past show little or no ability to do so in the future. For example, the S&P Indices research department (June 2011) published research showing that only 19.2% of large-cap funds with a top quartile ranking over the five years ending in March 2011 maintained that rating over each of the five years. They conclude by that consistent top-quartile and top-half repeat rates have been at or below levels one should expect based solely on chance. In other words, highly motivated professionals with massive computer power and large research teams don't appear to possess trading systems and timing systems which consistently allow them to beat simple market indices. These results must be considered when looking at the claims of tactical asset allocators, close cousins to active managers and traders. Active managers employ hundreds of statistical trading and timing models with many variants or combinations of variants.
Unlike most active investment managers, who typically keep their trading systems and information proprietary and confidential, tactical index asset allocators often promote and publically define a specific statistical asset class timing strategy. Most popular tactical asset allocation models are based on price momentum, specifically moving price averages and their crossovers. These are often termed trend following systems. The 200-day moving average, for example, calculates the average price of a given security or asset class for the last 200 days and issues a sell signal when today's price moves below the average and a buy signal when today's price moves above the average. Often, a crossover of the 50-day moving average on the 200 day moving average is used to determine buy or sell signals. For selected time frames or markets, moving average crossovers, like all statistical trading indicators, work profitably some of the time.
One popular tactical trend following system claims to reduce drawdowns in market downturns and match or beat a buy-and-hold strategy in all asset classes. When the current price crosses the 10-month simple moving average on the upside, one buys, if it crosses on the downside, one sells. This timing system stays invested in equities about 70% of the time. For 1900 through 2008, it produced a return of 10.45% versus a return for the S&P 500 of 9.21%, reduced the maximum drawdown from -83.7% (during the Great Depression) to -50.31%, and reduced the standard deviation for equity portfolios from 17.87% to 12.02%. On the surface, this sounds great. However, this timing system has a 70% turnover rate versus around 8% for indexes, substantially increasing the realization of capital gains and commission costs. For investors using the system in taxable accounts, the after tax return since 1961 would have fallen to 6.72%, well below buy-and-hold strategies. Investors using this system must be able to follow the signals, ignore the financial media, and stay out of the market even when the timing system is underperforming while the market rises or falls. In addition, sideways markets oscillating relatively rapidly between gains and declines can whipsaw this trading system, causing it to move in and out of positions frequently.
DFA examined a 10-month moving average market timing strategy for the S&P 500 with a 2% buffer to prevent excessive trading due to whipsaws. For 1927 through 2008, they found an annual return of 10.75% for the moving average model with a standard deviation of 16.64% versus an 11.67% return and 20.69% standard deviation for a buy-and-hold strategy. They noted that the inclusion of 2008, a disastrous year for equities, hurt the buy-and-hold results and concluded that the timing model appeared to reduce volatility. They then examined the two systems in equity markets in 21 developed countries and found only one where the timing model outperformed buy-and-hold. Standard deviations were almost always higher with a buy-and-hold strategy. On a risk adjusted basis both systems were close to equal from 1975 through 2008. They concluded that the good performance of a simple moving average system in US data is sample specific and likely to be the result of data mining. Most moving average strategies focus on US markets to make their case and exclude global markets. DFA and EAM recommend global asset class diversification.
Mark Hulbert of the Hulbert Financial Digest, a newsletter which evaluates the performance of investment newsletters and a quant, periodically publishes statistical analyses of various timing models. In 2004 Hulbert back-tested the 200 day moving average on 106 years of data and found it beat buy-and-hold in the S&P 500 by 2.7% annualized and reduced volatility. His analysis did not include dividends, which a buy and hold strategy would increase, commissions for trading, and taxes on capital gains in taxable accounts. He notes, however, a stunning reversal in the timing model since 1990 when it began to lag buy-and-hold by about 5% per year. He suggests this underperformance might have occurred since so many people are trading with moving averages that they lost their advantage.
In a later article, Hulbert notes that the moving average timing model returned 6.5% annualized versus 16.6% annualized for a buy-and-hold strategy from October 2002 through September 2007. In a further article, 3-17-10, Hulbert notes that the moving average model was still underperforming and its proponents had been altering trading parameters to improve upon it. In June 2012 he back-tested the 200-day moving average on the Dow back to the late 1800's. He concludes, "Market timing systems based on the 200-day moving average have impressive long-term results. On the other hand, over the last couple of decades (they) have markedly lagged buy-and-hold." In March 2013 he concluded, "If the 200-day moving average is the best we can do in our search for our top-indentifying indicators, the rational thing to do would probably be to give up and resign ourselves to buying and holding."
Some timers recommend modifications of trend-following models. One promotes the use of a 12-month moving average, claiming it reduces whipsaws and trading, another promotes the advantage of a 10-month exponential moving average over a simple average, while a third advocates buying or selling when the 50-day moving average crosses over or under the 200-day moving average. Hulbert analyzed the DJIA using the 50/200 day crossover model on data going back to 1896 and found no advantage for this trading strategy and concluded that moving average crossovers shed no light on future stock market prices.
Hulbert periodically publishes articles examining other timing systems including those based on market breadth, number of days of gains versus losses, lagging secondary markets, Dogs of the Dow, and the January indicator. He finds little if any long-term statistical support for them. There are dozens of different technical timing systems showing good performance for selected time frames but these are likely the result of data mining. Data mining means that the conclusions are based upon limited time-specific samples or non-comparable assets and do not provide a high level of statistical confidence.
Hulbert (7-12-11) examined what he considers the best market-timing system of all time, Norman Fosback's Seasonality Trading System. Fosback's system calls for being 100% in stocks at the turn of each calendar month and prior to exchange holidays. At all other times it is 100% cash. Since the early 1980's Fosback's system has produced a 11.4% annualized return in contrast to the Wilshire's total stock market index of 11.1%. For 2000 through 2009 it came out ahead of buy-and-hold by 0.1% per year. Risk is certainly far lower if stocks are only held briefly each month. However, trading costs and, if in a taxable account, short-term capital gains, will more than offset its slight benefit. Hulbert also looked at whether past quarterly equity gains or losses predicted returns in the subsequent quarter going back to 1896 on the Dow. They had no predictive value.
DFA (2-28-13) published an article examining Marty Zweig's complex multivariate market timing system which includes momentum indicators, monetary indicators, sentiment indicators, seasonal indicators and other market and economic indicators. Zweig had become famous and rich from his prediction the market would crash one week before it's largest one day sell-off in October 1987. Annualized return for the Zweig fund from inception in October 1986 through January 2013 was 6.79% vs. 9.84% for the S&P 500 and 7.90% for a conservative static mix allocated 30% to the S&P 500 and 70% to the Barclays Aggregate Bond index.
Morningstar Advisor (February/March 2012) updated an earlier study of tactical asset allocation funds which found 70% of the funds underperformed VBINX, Vanguard's pure index fund with a standard 60% equity/40% fixed allocation. Only 9 of the 112 funds had lower volatility than VBINX for August 2010 through December 2011. Only 24% of the funds had lower drawdowns than VBINX. These results also suggest that tactical allocation funds are unlikely to outperform a standard passive strategy or help in reducing drawdowns and investor discomfort versus simple passive and index portfolios.
A later Morningstar study (Wall Street Journal, 11-2-12) reported that two-thirds of 42 mutual funds and ETF's with "tactical" in their name in 2012 didn't exist before the financial crisis erupted in 2008. Investors had poured money into them in an attempt to avoid big market downturns, swelling tactical fund assets by 20% in the first nine months of 2012. None of these funds are old enough to have been tested in both bull and bear markets. Some hold their positions for only weeks at a time and expense ratios average a steep 1.46%. The Vanguard Balanced Index Fund has an expense ratio around 0.10% for a portfolio 60% in US stocks and 40% in bonds. Vanguard's buy and hold portfolio produced returns an average of five percentage points greater than funds with "tactical" in their name for the 12 months through third quarter 2012 and over the next three years tactical funds trailed the Vanguard's balanced portfolio by 4.9% annually.
DFA has approached the issue of price momentum with its own unique models. This is significant since most tactical allocation models use momentum calculations to move into or out of asset classes. DFA finds that buying stocks that have had strong performance over the prior six months and selling short stocks that have weak performance appears to be a lucrative trading strategy with the best stocks outperforming the market by 0.5% to 1.5% per month for the following six months. The data also show that stocks with positive momentum tend to negative or return to average returns after six months. Research by Mark Carhart in the 1990's suggested momentum as a fourth market factor in addition to asset class, size, and value but noted that transaction costs from high portfolio turnover consumed the advantage.
DFA recommends viewing price momentum as a risk factor. They find that for the five years prior to early 2008, trading short-term price momentum resulted in a 6% annual underperformance to the Russell 2000 and produced high turnover, high costs, plenty of volatility, and many more taxes in taxable accounts. Each day, all stocks in DFA portfolios are sorted for momentum on a country-by-country basis with calculations based on their prior 6-month performance. Individual stocks are ranked into a number of categories from "momentum up" at one end to "momentum down" at the other. For stocks in the two polar positions, DFA avoids either buying or selling as the case may be. Stocks in the neutral category are bought or sold if necessary.
On September 20, 2013, the WSJ published a piece entitled, "Is Tactical a Good Strategy". They noted the number of "tactical allocation" funds was expanding again. Morningstar reported 41 mutual funds with "tactical" in their names, including 18 launched in 2012 and 2013. According to Morningstar, the latest tactical funds represent the third wave of such vehicles, each of which came after a market crash or downturn. Because the funds tend to debut after crashes, they typically sell investors protection at the wrong time. On average, tactical funds gained 6.9% over the 12 months ended August 31, 2013 and 7.7% annually over the three years ended August 31, 2013. A standard portfolio with 60% in the S&P 500 and 40% in the Barclays aggregate US bond index gained 10.2% and 12.1% over the same periods, significantly outperforming tactical strategies.
So, what's an investor to conclude regarding tactical asset allocation? When reduced to its elements, the term just applies to another set of active trading strategies amongst dozens. Dimson et al. note, "Active investors are often interested in anomalies and regularities in stock returns…however, posited relationships often break down or turn negative at just the time statistically meaningful evidence appears to have accumulated. Unfortunately, too many anomalies are temporary traits that self-destruct (pg.208-209)." They do, however, note that the value effect, the small cap effect, and the benefits of international diversification have been well-documented for over a century in almost all equity markets. These are the premiums DFA aims to capture. DFA also employs momentum and profitability screens in their portfolios. While it may have been possible to slightly beat a buy-and-hold strategy with tactical asset allocation during carefully selected historical periods, after transaction costs and taxes and management fees, that advantage disappears. In less opportune time samples, tactical allocation fails. Given that there are many market timing systems better performing timing systems are likely the result of chance data mining, limited time samples, and survivorship bias, not manager skill.