| |
 September 2009 Mid Cap Surprise Why P5's Mid Cap Reliance Worked
Still, that doesn't diminish the results. While the overall positive performance may be a function of the inception date, P5's ability to stay ahead of its benchmark, the broad-based S&P Super Composite 1500, is definitely worthy of note. With so many active approaches struggling to stay up with unmanaged indexes, the question is, "How does P5 do it?" Interestingly enough, it does so by relying on unmanaged indexes. P5 is an ETF-based model with potential holdings limited to the iShares representing the nine Morningstar domestic style boxes. These are index ETFs that closely track the nine style indexes based on capitalization (Large, Mid, Small) and style (Growth, Value, Blend). The model is rebalanced in the first month of each calendar quarter using historical results and forward-looking estimates in a Black-Litterman forecast model. Because it is always composed of a combination of indexes, any outperformance must stem from market timing and the weighting of specific styles.
Less than two years into P5's existence, we noted (Mid Cap Preference) that it had a major bias towards mid caps. Even at that point and going back an additional five years in its backtest, mid caps have always played a major role in its makeup. This is even more remarkable given that P5's algorithm does not require any specific representation of any style index. Conceivably, the model could be 100 percent invested in just one. That's never happened with any style, not even mid caps. Nevertheless, mid caps have often represented 50 percent or more. (It is worth noting, however, that styles have been more frequently represented since 2007 when we started using a resampling technique in rebalancing.)
There's nothing in P5's algorithm that would make it favor mid caps over other capitalizations, but this approach has certainly worked. It's enough to make you wonder if there's something unique about mid caps that makes them the key to outperformance. Consider what has happened since January 1, 2002, P5's inception.
Historical Winners and Losers As you can see from Chart 1, small caps were actually the leading capitalization from January 1, 2002 through July 2009. Mid caps were close behind, and P5 was right there, too. You'd expect the model would closely follow mid caps given their steady heavy weight. Even so, P5 was far ahead of its benchmark, the S&P Super Composite 1500. But this is more a function of the relatively poor performance of large caps than P5's astute timing. All of the S&P style indexes are capitalization weighted. This means the higher the stock's capitalization, the more impact its return has on the overall index. Again as you can see from Chart 1, the S&P 1500 closely tracks the S&P 500. This is true despite the fact that the Mid Cap 400 and Small Cap 600 are also included in the 1500. In essence, they're outweighed by the stocks of the S&P 500. As a result, when large caps are underperforming other capitalizations, it's easier to beat the broad market S&P 1500 by simply avoiding large caps. P5 did this by focusing on mid caps, but arguably, small caps would have been even better.
But it's not that simple. Since P5's inception, small caps were the top monthly performers 43 percent of the time. Yet large caps led the way 32 percent of the time. That only leaves 27 percent of the months for mid caps. They didn't build their cumulative performance based on frequency and P5's success isn't just the result of riding the dominant capitalization. Consistency is another way to outperform over time. Rather than always beating the competitors, a consistent performance can, in the long run, provide greater cumulative results by avoiding large swings of good and bad months. It's actually possible to never be the top monthly performer, yet still lead the pack through consistently good results. But that's not what happened with P5 or the mid caps of the S&P 400. Chart 2 shows the number of positive months for each series measured across the 91 months extending from January 2002 - July 2009. Oddly enough, consistency was correlated with return, but inversely. Comparing Charts 1 and 2 you'll notice that the S&P 500 and 1500 had the lowest returns, but had the greatest number of positive months. At the other end of the spectrum, the mid cap S&P 400 and the S&P 600 had the highest cumulative results, but the fewest positive months. In this instance, consistency counted for something, just not what you would have expected. So what P5 lacked in frequency and consistency, it must have made up with magnitude of outperformance. In those months when it fared better than all or most of the unmanaged indexes, it must have done so by a wide margin. This would allow it to build a cumulative lead over time.
Chart 3 shows the scatter plot of risk and return for all five series. The overall shape of the distribution is what you'd expect: The larger, less volatile stocks of the S&P 500 experience less risk, but commensurately less return. The riskier small and mid caps offer higher return along with higher risk. In light of P5's reliance on midcaps, it's no surprise that the two plot very closely. The slightly negative returns for both the S&P 500 and 1500 attest to the impact of the bear markets at the start and end of the measurement period. Large caps bore the brunt of both declines so it's understandable that P5 would experience better returns with disproportionately less additional risk. Of course volatility works in both directions -- up and down. If symmetrical, P5's higher risk could help boost returns in an up market, but would also work to increase losses during downturns. But portfolio risk is not always symmetrical. If managed properly, risk should be higher on the upside than on the downside. Everyone likes upside risk, but absolutely loathes downside. A good manager -- or in this case a good model -- will manage risk by increasing it during up trends and lowering it downturns. That appears to be exactly what P5's algorithm did. Chart 4 shows the downside semi-standard deviations. P5's is noticeably lower than its overall standard deviation as illustrated on Chart 3. This enabled it to hold onto more of its gains in down markets than would have occurred if its risk was more symmetrical. This is an example of a model working correctly.
Putting it All Together But that's only true with 20-20 hindsight and for this particular period. Chart 1 indicates that had the measurement period ended in late 2002 or early 2003, even holding a high percentage of large caps would have been preferable. The overall assessment of the success of the model should not be dependent on the specific time period. The goal is not to create a quantitative model that will beat all other possible alternatives, but rather to outperform the relevant index. In this case, that index isn't the S&P 400 or even 500, but rather the broader-based S&P 1500. Throughout it's existence, it's done that, so on that count, P5 has been entirely successful. However there's one other thing to consider: The high correlation between the S&P 500 and 1500. After leading the market up into the tech bubble in the late 1990s, large caps have remained out of favor. Until the summer rally of 2009, large caps had been the lagging capitalization. It's quite possible P5's ability to outperform the S&P 1500 came not directly from its reliance on mid caps but from its avoidance of large caps. If 33 percent of your universe of potential holdings are laggards and you can avoid them, you will generally do quite well. P5 can be credited with this if indeed its algorithm steered it away from prospective underperformers. But if it has a natural aversion to large caps, the tide can quickly turn if large caps continue to lead the market for a prolonged period of time. P5 is a trading portfolio, not one designed to be buy and hold. Capitalization leaders remained fairly unchanged since 2002, so looking back, very little trading (other than to get out of large caps) was needed. P5 already missed part of the large cap run-up by maintaining roughly 12% in large caps so far in 2009. In fact, the last change was to reduce the large cap exposure in October 2008. In light of what's transpired in the first eight months of 2009, P5's trading algorithm may be starting to the boat. The next year or so will clarify this assessment. Larger allocations of large caps would be expected if they continue their run. If not, P5's early success may be no more than avoiding the right thing at the right time, not the result of timely tactical allocation. Search this site! Just enter you key word or words: Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||||||||
N THE SURFACE, PORTFOLIO 5 is the most successful of all six models. Compared to the others, it spent very little time in negative territory and exceeded its benchmark for almost all of the period. Of course it did have the benefit of a late start (January 1, 2002) which allowed it to miss the bulk of the declines in the aftermath of the tech bubble.
