A recent article showed how a simple moving average (SMA) could be employed as a screen to signal when you should be in or out of the market. It proved superior to a buy-and-hold strategy when applied to a mutual fund that represented the total US stock market. This simple test, although not thought of as such, was an example of a form of momentum investing.

Moving averages are rarely considered momentum investing. But moving averages are effective for the same reason that other momentum techniques are. Moving averages work because trends exist and generally do so for longer than some believe they should.

Efficient Market Theory (EMT) proponents are unable to explain this aberration in terms of their theory.

What Is Momentum Investing?

Momentum investing is predicated on corrections in stock or market pricing taking longer than the “instantaneous” adjustment predicted by EMT. If adjustment were immediate, excess returns would not be possible using a momentum strategy. However, the adjustment occurs more slowly than theory suggests it should. As a result, it has been shown that investors can use momentum to profit from buying stocks which are trending up and selling stocks which are trending down.

Tom Petruno of UCLA provides an overview of the momentum anomaly:

Conventional wisdom says that once any stock-picking strategy nears “sure thing” status it should be doomed. If everyone knows the secret to vast riches, how could the strategy possibly work anymore?

But there is a successful strategy that has been followed — and widely discussed — for decades, yet somehow persists as a relatively reliable money-maker: “momentum” investing, which is betting that the stock market’s recent winners will remain winners in the near term and, likewise, that the recent losers will remain losers. The strategy also is known as “relative strength” investing.

In the academic world, the pioneering research on momentum was a 1993 study published in the Journal of Finance by Narasimhan Jegadeesh and Sheridan Titman, both at UCLA Anderson at the time. They documented how strategies of buying recent stock winners and selling recent losers generated significantly higher near-term returns than the U.S. market overall from 1965 to 1989.

Regardless of why, momentum investing continues to offer unusual possibilities of profit.

The risk of following this strategy appears small. Even if the benefits were to disappear (a rational expectation) that does not mean losses would be incurred. Without the momentum effect one should expect normal market results from the selected securities.

How Momentum Is Calculated?

Momentum is a measure of how well a stock, ETF or mutual fund has performed over a recent period of time. All momentum investing uses ranking to select securities. Popular forms of ranking are the following:

• Percentage Returns
• Rank-order Percentage Returns
• Rank-order Percentage Returns and Volatility

All use recent security performance, generally gains or losses as a percentage of price, as a measure of momentum. Some brief comments on the types of ranking should demystify the matter:

1. Percentage Returns

Something as simple as the return on a security for the last month can be used as a momentum measure. Other time periods are also popular and could be used. Multiple time periods also are used.

In the one-period case, sort the returns of your population of candidates in descending order. Then, determine the number of candidates you want to invest in. If you are going long, choose the number from highest to lowest. If going short, choose the number from lowest to highest.

If you were interested in going long three stocks, you would buy equal amounts in the highest three ranked by return. (All comments from hereon assume that you are going long and not short.)

For multi-period momentum (say, one month and three months), calculate the return percentage for each period. If you were weighting each period equally, add the two percentages together and use that for the ranking order.

If you preferred to weight the periods unevenly, weight the first month accordingly and then weight the third month accordingly and the add them up to get your ranking order.

2. Rank-order Percentage Returns 

This method is similar to the first, except you choose by rank-order and not percentage. If you are only using one period (say a monthly return), either ranking will provide the same result. However if you are using multiple periods, say 1, 6 and 12 month returns, you could end up with different rankings using the two different methods.

Most pros and information users prefer the rankings method over the summation of percentages. That is my preference. It also is the method that ETFReplay provides.

If you choose rank-order (recommended), each security for each period is ranked from first to last. In the three period example mentioned above, you would generate three columns for each security. These columns would contain the returns for each period. Then you would create an additional three columns which represented the rankings (based on the returns for each period). Thus, a security would have three ranks and they would only by coincidence be the same for all columns.

Assume that security A, had the following three ranks for the three periods: 3, 1, 8. These ranks would be converted to a composite rank. If all ranks were to be weighted equally, then multiple each column by 33.3%. The composite rating for security A would be 4.0. This process would be repeated for all securities.

Then, based on the weighted column sorted from low to high, the appropriate number of securities would be selected and purchased in equal dollar amounts.

3. Percentage Returns and Volatility

Some momentum traders include volatility (usually the standard deviation of a security) in their calculations. The rationale is that you are trying to achieve high returns but not at the expense of undue risk. In this case, the ranking method must be used because returns and volatility should not be added (the apples and oranges issue).

The same process used above is used to rank returns for the three periods. But, the volatility calculation is ranked from highest to lowest. That is, the lowest volatility would receive a rank of 1.

Now there would be four columns of rankings for each security. These would be weighted by whatever weighting desired to produce a composite column of ratings. These would be selected from low to high in order to obtain the number of securities necessary for investing.

The ETFReplay site provides the opportunity to utilize volatility in the rankings. Specifically, it allows two different return periods and one volatility measure. It does the rank-order work for you, based on your period inputs and weightings.

The complexity of the calculations overwhelms many investors/traders. Thus, the use of software or a service to provide the selection for you each month is recommended. Some provide limited capabilities for free while others provide pretty good capabilities for a fee.

Some Results of Momentum Investing

The best way to demonstrate the effectiveness of momentum investing is to review some basic results. Ten securities (mutual funds) were used in a pool for backtesting.  These securities included primarily mutual funds of stocks and bonds. Additionally, the price of gold was included.

The first test shows what buying only the highest ranked selection produced. Trades were made monthly at the end of each month. The tests were from Jan 1999 – Jan 2019, a period of just over ten years.

One Selected

ONE STOCK ONLY

The table above shows the one-stock as the timing portfolio. The equal weight portfolio represents what holding each of the ten stocks in equal dollar amounts produced. The Stock Index shows the results of a buy-and-hold strategy of the broad stock market.

The results are impressive. Over this period the Stock Index had a CAGR of 6.42% The timing portfolio had an 11.21% CAGR. All portfolios had maximum drawdowns that most would consider unacceptable.

Two Selected

The second test repeats the first except that it selects the two highest-ranked momentum securities. The results are dramatic in every respect.

TWO STOCKS ONLY

The drawdown was reduced to 19% and all other measures improved. The Stock index (3rd line from first table) is not repeated for this test or the subsequent one as it does not change.

Three Selected

The three-stock portfolio is shown below:

THREE STOCKS ONLY

The results are still dramatic compared to the market performance or the equal weight portfolio.

The four selected portfolio (not shown) is slightly worse than the last one shown. Additional selections, at least with only 10 choices to draw from, appear to diminish beyond those presented.

Charts of Two Stocks Only Outcome

The graph below shows the graphic outcome of the Two Stocks Only Momentum Outcome:

Summary of Tests

In the examples above, there were ten possible selections in the universe used in momentum selections.

The momentum algorithm dramatically outperformed a buy and hold market strategy in all cases shown. Selecting the two highest-ranked momentum securities produced the best results. Selecting only one created an improvement in both drawdown and returns, although the drawdown was higher than most investors would be comfortable with. At three and four selections (30 and 40% of the universe) momentum continued to outperform but at on a reduced scale.

The power of momentum investing seems pretty clear from these tests.

Conclusion

Dramatic improvements from using a simple momentum model were demonstrated. The model utilized was based on momentum only. More sophisticated models can be utilized.

1. One was referenced above (but not tested) and incorporates volatility into the rankings.
2. Another approach uses what is called adaptive momentum. Models of this type generally select a specified number of securities in the universe but do not weight them equally. Those passing the test are held in different proportions based on their momentum.
3. A third approach involves what is termed a dual momentum approach. This approach ranks based on relative momentum to select the securities. Then a moving average screen is applied to those chosen. If the screen is not met, cash is substituted for that chosen security.

Many other approaches are possible and have been used. For example, a modification of the third approach could apply the screen to the general market. If the general market failed the screen, then you could modify the selections by choosing fewer or choosing none at all.

Only your imagination and a reasonable grounding to economic reality should limit your designs. Whatever you come up with should produce good backtesting results.

That ends this discussion.