September 2001
Just Momentum?
| "We didn’t lose the game, we just ran out of time." |
| -- Vince Lombardi |
UANTITATIVE MODELS FREQUENTLY SUFFER from the effects of momentum. If they are based upon historical factors, they often rely more heavily on those that were most influential in the recent past. In other words, they tend to favor what has been working most recently.
There’s nothing wrong with that, but most investors would prefer a model that is more
predictive; one that recommends stocks that will outperform in the future rather than in the past. After all, you can’t invest in the past.
What about our quantitative models, do they suffer from this flaw? To test, we looked at Portfolio 3. Unlike Portfolio 4 that is only reconstituted once a year, it is reformulated every other month. Momentum would have a greater opportunity to affect it and would be easier to detect.
Period Pieces
The model has only been around for thirteen months, so data is limited. We plan to test it for three years before making significant changes to its underlying methodology. Still, it’s not too early to take a reading on its performance.So far the model has gone through seven incarnations during that time: six two-month periods and one one-month (June - July 2001). The nearby chart shows how it performed relative to its
|
Chart 1: |
 Data Source: Baseline Portfolio 3 is reformulated every two months. From its inception effective July 1, 2000, it has bested the S&P 500 in three of the seven periods yet trails rather dramatically over the entire timeframe. |
Although the individual periods are as evenly split as possible, P3 significantly underperformed over the entire time frame. This indicates that when it outperforms, it does so by much smaller margins than when it trails the index. The chart supports this conclusion.
It also illustrates the portfolio’s growth bias. Like most growth portfolios, P3 performs best when the market is rising. It did just that in the June-August 2000, and April-June 2001 periods. Its other period of outperformance occurred in the August-October 2000 timeframe when the market showed relatively little movement.
On the other hand, in the periods when value prevailed (Oct-Dec. 2000 and Feb-April 2001) P3 faired poorly. In fact, it was these two periods when the lion’s share of the overall underperformance occurred.
|
Chart 2: |
 Data Source: Baseline This chart compares the actual return of P3 (green bar) and the S&P 500 (red bar) to the returns that would have been attained if each formulation of P3 had been held from its respective inception through July 31, 2001 (yellow bars). Performance clearly varied but that might be more a result of changing market conditions rather than specific portfolio composition. |
The growth bias is a result of using 10-years of performance data from the decade of the 90s. Growth dominated in almost every year, so it’s understandable that the model would favor growth characteristics. Even so, that doesn’t mean it’s biased towards near-term momentum.
Stick to It
In a volatile market like the one we’ve experienced over the life of P3, momentum bias would cause a lot of turnover. As market conditions change, stocks passing the model’s screen would change as well.For the 12-month period ending June 30th, turnover was 127%. That may sound high, but it’s right on par -- actually even a little lower -- than the average large-cap mutual fund. Of course that doesn’t mean that a momentum bias didn’t lead to that degree of turnover.
|
| |||||
|
If turnover was momentum induced, you’d expect the various incarnations of the portfolio to perform differently over time. To check this, we backtested to see what would have happened if the portfolio hadn’t been reconstituted.
To do this, we calculated what the returns would have been if each version of the portfolio had remained unchanged from its introduction through July 31, 2001. We then compared this to the actual return of P3 over the same time period. The results are shown in Chart 2.
Interestingly, the "static" portfolios led in each timeframe, but not by much. Biggest differences occurred in the periods ending December 2000 and February 2001 (6% and 10%, respectively). Given that the longer periods were roughly equivalent, these differences may also diminish with the passage of time.
To equalize the effects of time, we next compared each version of the portfolio over the entire 13 months to actual P3 performance and that of the S&P 500. This is shown on Chart 3.
Performance begins to improve with the portfolios created in December 2000 and later. These versions would have had returns quite similar to the S&P 500. This may be the effect of momentum.
|
Chart 3: |
 Data Source: Baseline This chart compares the actual return of P3 (yellow bar) to the returns that would have been attained if each formulation of P3 had been held from inception, July 1, 2000 through July 31, 2001 (green bars). In each instance, the static portfolios prevailed. |
Past and Future
Why would the more recent versions of the portfolio perform better over the 13 months than the earlier versions? We would suggest the first incarnations still reflected technology and growth stock leadership while the more recent versions were influenced by the market’s shift to value.This is supported by a sector comparison of the various versions. Chart 4 shows the weightings of each. The original portfolio had over 81% in Technology stocks. This was the hottest sector for the prior 24 months.
By the June 2001 portfolio, tech had fallen to 26%. At the same time, defensive sectors Healthcare and Consumer Staples showed a marked increase. These were some of the best performing sectors over the past 12 months.
|
Chart 4: |
 Portfolio 3's sector weightings have changed over the past 13 months. Initially it was heavily weighted in Technology Stocks, but that has fallen over time to the extent that Healthcare and Consumer Staples now account for much greater shares. |
Based on this rather limited data, one could argue that P3 reflects recent performing sectors rather than those that will outperform. This is momentum bias.
In a trending market, momentum based models can be helpful in stock selection. But in volatile or directionless markets, they will generally trail the overall market since previous winners will not repeat.
Perhaps the results will look different at the end of the full three years. If not, subsequent models will need to address the issue of momentum bias. Time will tell .
Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||
| Homepage | Return to Top |
