| IN THE ARCHIVES: January 2010 November 2008 September 2009 July 2009 May 2009 March 2009 More... |
 March 2010 Preliminary Performance An Early Look at P3's Results
It's a little early to draw any major conclusions about the success of the new algorithms, but we wanted to take a look anyway. It's not like P3 is a buy-and-hold portfolio, it has six scheduled reoptimizations each year. Forty-three trading days may not sound like much, but that's one entire cycle for P3 so why not take an early look?
What You Might Expect
The first factor (1-Year Price Return) rewards momentum. Thanks to the negative coefficient, the second factor (Debt/Capital) penalizes firms with high debt loads. The third factor gives a minor nod to stocks with high Price-to-Book ratios, but this isn't a very important factor given its miniscule coefficient. Finally, and this one's the head-scratcher -- the greatest weight is placed on stocks with low long-term Earnings Growth rates. Perhaps this is the contrarian play -- they may have poor historical long-term growth rates but they're due to pick up. So what fits this profile? How about financials? They certainly finished 2009 with some momentum. After being the most beaten down stocks early in the year, they came back with a flourish. Price-to-Book ratios also rose in the final months of 2009 as share prices climbed and book value remained depressed. Despite the price recovery, most analysts still weren't predicting overly impressive long-term earnings given the threat of increased government regulation as well as the fact that many Financials still carry "toxic" assets on their books. The Debt/Capital ratio can be an issue for some banks, particularly smaller ones, but this shouldn't' be an issue for the majority.
Well guess what? Financials actually fell by more than 10 percent: 42.79 - 32.62 percent. Technology was the big winner, jumping from 24.04 percent to 31.74 percent. This was somewhat surprising. While Techs certainly had 1-year momentum and low Debt/Capital ratios, their Price-to-Book ratios tend to be low and more importantly, their five-year Earnings Growth Rates are some of the best to be found. Of course the unexpected sector weightings may also be a function of above average turnover. This would be expected with the introduction of the new algorithm. In this case, seventeen of the thirty stocks was replaced yielding a turnover ratio of 57 percent. With so many stocks coming (and an equal number going) odd sector weightings are to be expected.
Ultimately, this version of P3 was a mixture of Tech (37.74 percent), Financials (32.62 percent), and consumer Discretionary (17.44 percent). Smaller allocations were included in Industrials (8.5 percent), Materials (5.91%), Healthcare (3.46 percent), and Telecommunications (0.27 percent). Consumer Staples, Energy, and Utilities were not represented.
What Actually Happened But that doesn't tell the entire story. To truly get a sense of how the new algorithm is working, you need to look back over the entire period when this particular composition was in effect. That doesn't start on January 1, 2010 but rather December 14, 2009 (Chart 2). That period looks a little different. In the final week of 2009, P3 jumped ahead of the benchmark by two percent. When that fast start is included, the model actually finishes the entire period with a 0.9 percent lead. This is understandable on a number of different levels. From a risk standpoint, the model performed exactly as would be expected given its composition. The lowest quality, riskiest stocks led the charge in 2009. While P3's holdings weren't particularly "junky", they weren't the bluest of the Blue Chips, either. Stocks such as Ford and M&T Bank Corp certainly had strong price momentum at the end of the year along with relatively high Price-to-Book ratios and questionable long-term earnings expectations. The price momentum was coming off an extremely low base as both had been pummeled in the first few months of 2009. It was holdings like this that gave P3 its fast start in December.
It was also the riskier holdings that penalized P3 in early 2010. As fourth quarter earnings reports came in, investors rethought the prospects for 2010. In many instances, the effects of improving markets was already reflected in current share prices, diminishing investors' risk appetite. As a result, the riskier shares that had put P3 in the lead worked against it in January. The year-to-date results in Chart 1 only focus on this second phase without the benefit of the initial positive one. Performance over the entire period is much more encouraging. The results also makes sense from a sector standpoint. Financials, Technology, and Consumer Discretionary were leading sectors when the market turned up in the final nine months of 2009. These are generally considered "riskier" sectors which entered December with strong momentum. These were also the sectors that were shunned in January when investors sought the safety of traditionally "safer" sectors such as Utilities and Consumer Staples. Both of these latter sectors weren't even represented in P3. Finally, P3's performance makes sense from a relative risk standpoint as well. The December/January version of the model had a beta of 1.29 vs. the benchmark S&P 500. As you probably know, beta measures the market risk of the model in terms of the benchmark. A beta of 1.29 means for every 1 percent the market moves (in either direction), P3 will move 1.29 percent. A relatively high beta is what you'd expect when a portfolio is comprised of riskier stocks. It's also why P3 moved ahead of the benchmark in December when the market was moving up and why it so quickly fell behind in January when they reversed course. Despite the introduction of P3's new algorithm, this level of beta is actually right in line with the previous versions. Ever since its inception, P3 has always had the highest beta of all our model portfolios. Not only that, its average beta has always been in the 1.25 - 1.20 range. In this one sample case, the new algorithm did not subject the model to more (or less) market risk than in the past. One characteristic that did change was how closely the model tracked the index. In the past, P3's r-square to the S&P 500 had been over 0.90. R-square, also called the "coefficient of determination", measures how closely one series tracks another. It can be interpreted as the percentage of the dependent series' movement (in this case P3's return) is explained by that of the independent series (the S&P 500's return). The value can range from +1 (100 percent) to 0 (0 percent). Over this particular period, the r-square was 0.80, suggesting 80 percent of P3's return was explainable by that of the benchmark. This was considerably lower than the model's historical average, but may have been attributable to the volatile nature of the market at year-end.
Looking Ahead
First, the algorithm's heavy weighting on low Long-Term Earnings expectations suggest this may be more of a contrarian approach than the original formula. The influence of this factor was evident in the "riskier" stocks selected in the December reoptimization. As Chart 2 illustrates, this served the model well when investors embraced risk. It also helps to account for the models' relatively high beta. Second, the momentum factor also appears to have a major impact. When reformulated on February 12, Financials -- the sector which started the year with a high degree of momentum yet suffered the most in the mid-January selloff -- fell from 32.62 percent to 7.83 percent. This is a dramatic period-to-period change rarely seen in this model. On the other hand, Healthcare -- a sector that struggled throughout 2009 under the cloud of potential nationalized healthcare -- saw its weight jump from 3.46 percent in December to 27.18 percent in February. This is most likely the effect of the heavy weighting placed on lower Long-Term Earnings projections and again, this increase was much more pronounced than had been typical in P3. Third, early indications are the algorithm may be well suited for today's volatile market. The move into Financials was well-timed in December, allowing P3 to build a lead over the index able to withstand the selloff in January. The move into Healthcare came at a time when political threats were waning possibly allowing them to make up last year's lost ground against the market. It's still too early to tell how this will work out, but at this point, the move appears to be well-timed. Fourth, turnover (see Chart 3) in February (23.3 percent) was near the high end of the P3's historical range and well above the historical average (14.5 percent). This comes on the heels of December's 27.7 percent turnover. As suggested above, it's conceivable the latter's above average reading was due to the switch to a new algorithm. February's slightly lower score could be the result of the volatile markets and the major sector weighting change. Regardless, this is something to watch closely because one of the goals of the model portfolios was, and remains, to create trading models that won't break the average investor's bank with transaction costs. One of the reasons for the changes in procedures in June 2003 was to reign in turnover. Finally, and most importantly, the new algorithm itself is still something of a concern. It's really puzzling why it would reward stocks with lower Long-Term Earnings growth rates over those that would seem to be better positioned. As suggested above, this may be a way to add a contrarian twist or perhaps it's a vote of no-confidence for the analysts that follow the stocks, but regardless, it's still not what you would expect. All else being equal, this wouldn't be such a concern at this point, but the relatively low r-square (there's that concept again) for the algorithm's regression makes it more of an issue. As pointed out when the new algorithm was introduced, the r-square is a lot lower for this regression (0.0576) than for the original (.6574)... In this case, the original regression explained over 65% of the movement while the new is roughly ten times lower. Doesn't give you a lot of confidence, does it? That could go a long way to explaining why the r-square between P3 and the benchmark in the December/January period was so much less than what it had been in the past. It could also explain the weird coefficient for Long-Term Earnings Growth Rate, too. As time passes and more data comes in, this is definitely something to watch. You can always get a regression between two series, whether they're actually related or not. Although it's too early to draw any definitive conclusions about 2009's changes to P3, here's hoping it's still running with a meaningful algorithm. If not, this will be a rough three years before it's revisited again. Search this site! Just enter you key word or words: Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||||||||||||||||||||||||||||
UANTITATIVE PORTFOLIOS 3 AND 4 received updated algorithms in December 2009. Both were rebalanced then, too. P4 won't be revisited again until late June, but P3 was already rebalanced on February 16.
<
