| |
 September 2006 Downside Up An Analysis of Portfolio 3 and 4's Upside and Downside Capture
Given these asymmetric expectations, investors should prefer investments with equally asymmetric return profiles. Beta, the traditional measure of an investment's sensitivity to market movements doesn't work this way. It's generally symmetrical: If an investment has a beta of 1.25, it will capture 125% of the market's return on the downside as well as the upside. That's not what most investors want. Instead, they'd prefer a high beta on the upside and a low -- or ideally negative -- beta on the downside.
Most investments don't naturally act this way, but a good manager can achieve this by simply increasing cash levels in negative markets. By dumping depreciating holdings in favor of cash, he or she can minimize or perhaps even eliminate downside capture. That's what you're paying them to do, right? It might be possible to include similar features into model portfolios, too. If their underlying equity selection process could somehow incorporate this feature, the resulting model would render more appealing results in both up and down markets. So how do quantitative Portfolios 3 and 4 stack up? Do they do a good job managing upside and downside capture? If not, why not?
Inputs The models were based on data from the prior ten years, the decade of the 1990s. As you’ll probably recall, that period was marked by a long-running equity bull market in which growth outperformed value. Large cap stocks were market leaders throughout the decade -- but were long overdue for a reversion to the mean. The models were created by running multiple regression analyses on seven fundamental characteristics and annual returns. P3’s algorithm is designed to locate what are expected to be the top performing stocks in the S&P 500 regardless of sector or index weighting. The portfolio is comprised of the top 30 and is reoptimized in the middle of each even-numbered month. P4 also looks for the top potential performers, but within each sector. Instead of having one regression for all stocks like P3, it has a different one for each sector. This comes from the assumption that the same fundamental factors affect the various sectors differently. P4's goal is to find what will ultimately be the top performing stocks in each sector. It's now reoptimized twice a year, in mid-June and mid-December. Before making any changes to these quantitative portfolios, we wanted to give them a chance to work through a full market cycle. Historically, that would take about three years, so that was the timeframe before the initial review. When we did revisit the portfolios in July 2003, we made two changes. Initially, stocks remained in P3 until their ranking fell below 40 in subsequent bimonthly optimizations. When that happened, the next highest rated stock not already in the portfolio was added. After the first triennial review, stocks were allowed to fall as low as 100 before being removed. P4 was originally only rebalanced on an annual basis. When the optimization was rerun, the top stocks in each sector were included in the portfolio. Prior holdings were removed whenever they were no longer top scorers. After the initial review, it was decided to reoptimize P4 more frequently, in June and December. These changes were designed to minimize adverse effects of frequent turnover (P3) and infrequent turnover (P4). Upside capture and downside capture weren't considered. However, when the portfolios got their second triennial review in July 2006, volatility was an issue. Now with a six-year history, the models had been through the equity bear market (2000-2002) and the subsequent bull market (2003-2005). Despite a strong market-beating recovery, both failed to pull themselves out of the hole caused by oversized bear market losses. No changes were made after the 2006 review, but it was clear that something should be done to control the models' volatility -- especially in down markets. The first step is to quantify its impact by looking back at upside and downside capture.
Outputs This indicates that both are more volatile than the market itself. That's why they fell so hard in the bear market and made substantial recoveries when stocks again turned up. There's nothing wrong with volatility per se, yet it's certainly more preferable on the upside than downside. In order to get a clearer picture of its impact, we calculated both upside and downside capture for each model. In addition to the overall figures for the entire six years running from July 2000 through June 2006, we also broke it down into the period before the first triennial's changes and the period following them. We also calculated an average capture rate for periods in which the market moved either up or down for three months or more. This was to gauge the effect of trending markets on capture. The results are summarized on the nearby chart. This is one of those cases where a picture is worth a thousand words. Both models' actual capture -- on both the upside and downside -- exceeded what would have been predicted by their beta alone. For example, with a average beta of 1.45, P3 should have had a downside capture of -145% of the index, but instead it was -246%. Similarly, P4's upside capture would have been predicted to be 115% of the index, yet it actually turned out to be 143%.
In addition, rather than being symmetrical, downside capture is greater for both models. This is especially noticeable for P3. This can't be the result of beta alone because it would lead to roughly the same percentage results on both the upside and downside. Finally, capture increases on both the upside and downside when the market is trending. Again this is more noticeable for P3, but holds for P4 as well. This would not be expected based solely upon beta. So what is it about the models that leads to these results? More importantly, is it something that can be corrected?
A Possible Explanation Although both models' algorithms are based on the fundamentals of the best performing stocks of the 1990s, each seeks to find the stocks that have those characteristics now. These factors were derived from ten-year's worth of data, but the models don't look for historically good stocks but rather those that are expected to do well in the future. Generally these will be stocks that have been doing well recently; in other words, those that have a high degree of price momentum. This is no new discovery. In fact, as early as September 2001 we explored the effects of price momentum on P3. At that point, the models had only been out of sample for less than a year, not enough time to draw any meaningful conclusions. Five years later, we have a little more data. By its very nature, P3 chases price momentum. Being reoptimized every two months without regard to sector or index weighting, it has the opportunity to add shares with the highest price momentum while dropping those that taper off. The relatively rapid reoptimization allows it to always hold shares with the greatest momentum. On the other hand, those high-momentum shares have often experienced a run up in price before being added to the portfolio. In many instances, they're overvalued. While they come into the portfolio with strong momentum, they also have further to fall when market conditions change or they lose that momentum. This could go a long way in explaining why P3 has always had more downside than upside capture. It also explains why downside capture rises so precipitously in downward trending markets. It also fits with the differences resulting from the 2003 changes, too. By allowing stocks to remain in the model when their ratings fell as low as 100 rather the original limit of 40, the 2003 changes dramatically reduced portfolio turnover. In other words, stocks now stay in P3 much longer than they did in the first three years. This had a predictable effect on both upside and downside capture.
As shown on Chart 1, downside capture increased from 178% to 204% while upside capture declined from 202% to 181%. The increased holding period means stocks are now retained for a longer time as their momentum fades. As a result upside potential is diminished and downside is enhanced when declining shares are retained. P4 reacted differently to the 2003 changes. In this case, upside capture remained relatively unchanged (142% vs. 145%) while downside capture decreased (163% vs. 122%). Momentum can explain this, too. In its first three years, P4 was only reoptimized once a year, allowing fading momentum stocks to remain in the portfolio up to twice as long as they would following the 2003 changes. Again, assuming many stocks entering the model were already run up in price, the longer holding period would have a greater impact on the downside than the upside. By doubling the reoptimization frequency, the 2003 changes limited the downside impact while leaving the upside potential virtually unchanged. Over the entire six-year period, P4's greater diversification and lower beta relative to P3 tempered momentum's effect on both the upside and downside. If this analysis is correct, beta and momentum are the major drivers of both models.
What's Next? At this point, no changes will be incorporated until the next triennial review in July 2009. In the meantime, however, it will be informative to closely observe how incoming data fit with the theory. Further confirmation will only strengthen the justification for future changes. There's also plenty of time to consider potential modifications that could minimize the effects of momentum -- or at least turn them in the models' favor. There's also time to test potential solutions before actually implementing them. Momentum in and of itself is not a bad thing. When it skews upside and downside capture as it appears to with regard to P3 and P4, it becomes a problem. Other factors may be at work, but for now, momentum appears to be the main culprit. Its impact and how to potentially neutralize it -- at least on the downside -- will certainly be a focus over the next three years. Search this site! Just enter you key word or words: Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||||||||||
NVESTORS HAVE DIFFERENT EXPECTATIONS when the market's going up and when it's going down. When stocks are rising, they want to beat the market and come out ahead in relative terms. But when stocks decline, relative performance is little solace for investors who simply lose less than the market. In that case, a relative win is still an absolute loss while they're seeking positive absolute returns.
