| |
 January 2011 Incrementally Better Model Portfolios 3 & 4 are Getting Better at Beating Their Benchmark
Two of our quantitative models, Portfolios 3 and 4, were created just as the Tech boom was coming to an end. They weren't intended to be index models, but they were designed to beat their benchmark, the S&P 500. The goal was to do so incrementally rather than dramatically each and every year. This more subdued approach was expected to restrain risk yet still allow the models to shine in the long-term.
Almost immediately out of the gate, the market moved against them. P3 and P4 were quickly in the hole as Tech stocks -- a major holding in P3 -- led the all stocks into a two-year bear market that didn't end until the final quarter of 2002. By then, both models were well behind the index. Today both indexes continue to trail the benchmark on a cumulative basis from inception. The poor start put them in the worst predicament imaginable: They were forced to make up a serious amount of ground, but were only designed to do so incrementally. No wonder they still trail ten years later. But that doesn't mean they're ultimate failures. In fact, as you'll see below, the evidence actually points in the opposite direction. When viewed in a broader context, not only are P3 and P4 doing what they were supposed to do, they're doing better as time goes by.
The Key So nothing could be done to cushion what happened over the following two years. When Tech stocks first imploded, they represented over half of P3. Fortunately for P4, its algorithm required it to at least start each year with a market weighting of all ten S&P market sectors. Of course Tech was the the biggest allocation in the index, too, so P4 still shared the pain.
The initial downdraft hit both portfolios hard. Their algorithms are designed to pick stocks that performed well over the previous five years. As the market heated up in the late 1990s, the stocks that did best were the high-beta stocks, those that moved faster than the market -- primarily because they were carrying above-average risk. When stocks abruptly turned downward, these high-flyers became fast crashers -- and P3 and P4 were full of them. Again, nothing could be done to help them other than wait until the next scheduled rebalancing. Initially, P4 was rebalanced only once at the end of June while P3 was optimized every two months. Starting in 2003, P3 is now rebalanced every six months and results have improved. It's easy to see why if you look back at what happened in 2000-2001. Those high-beta stocks that had done so well in the prior years fell hard. P3 could at least dump some in less than two months, but poor P4 had to retain them for a full twelve.. Of course P3's more frequent rebalancing didn't really help it initially, either. Although it wasn't stuck in the same portfolio for an entire year, it was whipsawed by false signs of recovery. When the market went south, it didn't do so on a straight path. There were a number of false starts which sent more volatile stocks up sharply. This fooled P3's radar forcing it to swap out losing stocks for shares of other companies which were poised for a fall. Looking back, it might have helped P3 to have had a longer period between rebalancings. So both models got off to a lousy start, but that's not the complete measure of their success or failure. It's the old "mountain chart" fallacy. You certainly know what a mountain chart looks like; you've probably seen more than enough of them. The basic shape shows a strong upward gain and then there may be a pause or even a few inconsequential losses before moving back up. Brokers and mutual fund companies like to use mountain charts to tout their performance. But have you ever stopped to realize that you've probably never been shown a mountain chart that starts off with a steep loss? There's got to be a lot of investments out there that started off like that, but you'll probably never see their mountain chart unless you chart it yourself. We've consistently shown you the charts for P3 and P4. A bad start is a bad start and again, it's hard to eradicate if you only beat the index incrementally. But that's also the key to evaluating the two models: Are they behaving as expected? In other words, are they beating the benchmark incrementally? The answer is yes. In fact they're not only beating the index incrementally, evidence suggests they're doing so at an ever expanding pace.
Act Like the Market
To avoid this problem and truly test the models' algorithm against the unmanaged index, we required both to limit holdings to the 500 S&P stocks. In other words the only potential were stocks that could also be in the S&P 500. This means any market beating performance would have to come from the stocks selection. P4 had it even a little more difficult being forced to maintain market weights in each sector. Because of these requirements, you'd expect the models to behave similar to the index. Looking back over the past 10+ years that assumption has proven correct. Chart 1 is a scatter plot with the green diamonds representing P3 and S&P 500 returns for the same time periods. The blue dots show the same, but for P4. Because P4 was originally rebalanced only once a year and now semi-annually, it has only 17 points, while P3 has 60 from its bi-monthly rebalancing. The green line is the regression line (least-squares fit) for P3 and the S&P 500 while the blue line shows the same for P4. The regression equations are also shown on Chart 1 in their respective colors. Although there are some outlying points, the tight pattern of both the diamonds and circles shows the models acted quite similarly to the index. The outlying points say more about the volatility of the market over the period than they do about the models themselves. The regression lines both fall extremely close to the origin (0,0) with P3 missing by 0.0010 and P4 by 0.0046. These values are the "intercept" in the regression equations for each model, respectively. The slope of the regression equations are also indicative of the model's betas. P3 has always been our highest beta model, typically around 1.4. That's precisely the slope of its regression line (1.4019). P4's beta has always been above that of the market (1.00), but not nearly as high as P3's. This is also borne out by the slope of its regression line, 1.0506. So the models have acted quite similarly to the index, just as designed. That's a good thing. Of course the scatter chart doesn't tell anything about relative performance and that's what we're really interested in. To see that we need to look closer at the measurement periods.
Behave Better than the Market As discussed above, the risk profiles are similar. To measure the relative performance, we looked at each of the performance periods from the models' inception. Because it's rebalanced every two months, P3 had considerably more periods (60) than annual/semi-annually rebalanced P4 (17). The larger number made P3 more meaningful in this review. As a measure of consistency, we broke relative performance into 0.2% bands. We then measured the frequency of each. The results are shown in Chart 2. Just looking at the chart you'll notice it's skewed negatively with most periods clustered toward the right and a long tail to the left. This means P3 more frequently exceeded the index than not. Looking closer, it actually did so 75% of the time (45 of 60) by a weighted average of 0.37%. That's not big difference but certainly qualifies as incremental. That's exactly what it's designed to do. But recall, the measure of frequency doesn't give much insight into the long-term trend. For example, it's possible that most of the outperformance occurred back in the early years and is now decaying. If so, P3 still wouldn't be meeting its goal. Obviously, the trend is important, too. To measure this, we mapped P3's relative return for each of the sixty measurement measurement periods. This is shown by the green line on Chart 3. As Chart 2 suggests, more are positive than negative and are distributed thoroughly over the past nine years. But that's not the entire story. Chart 3 clearly shows the worst relative performance occurred in the first three periods in 2001, during the height of the Tech meltdown. Since then there have been negative periods, but not nearly so severe. On the other hand, positive relative returns have increased and more are falling in the 2-4% range. As a result, the 10-month rolling trend (red line) has a definite positive slope suggesting a positive long-term trend.
The pattern is precisely the same for P4, but with fewer points (measurement periods), results aren't as statistically significant. Nevertheless, the early evidence suggests it, too, is improving in its quest to beat the index.
The Future Looks Better than the Past But taking a longer-term view and focusing on the future rather than the first few months of the two models, everything looks a lot better. Without the initial poor performance, both P3 and P4 would be well ahead of the benchmark. Even more encouraging is the fact that they are performing as expected by generally outpacing the index and doing it with increasing frequency and arguably, magnitude. This is precisely what they were intended to do. Of course, choosing to focus solely on the periods after they fell behind so severely is not unlike companies relying on "pro forma" earnings which are essentially results before all the bad stuff. A full evaluation can't simply cherry pick the best results. Those first few months must be considered when evaluating P3 and P4. So no, neither P3 nor P4 has lived up to all their expectations, yet we know why they suffered in 2000 and 2001. In the case of P4, more frequent rebalancing may actually help to minimize similar effects in future markets. As P3 and P4 suggest, the long-term trends are improving. Not only that, the fact that P4 actually caught and surpassed the S&P 500 means there's hope on the cumulative basis as well. As mentioned earlier, incremental outperformance makes it difficult, if not impossible, to rapidly make up lost ground. But increasing frequency and magnitude of relative performance also bodes well for P3 and P4's future. The first ten years weren't all we'd hoped for, but recovery began and the future looks even brighter. Search this site! Just enter you key word or words: Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||||||||||||||||
HE TECH STOCK COLLAPSE A DECADE ago launched the "passive" vs. "active" investing debate. Despite the heady gains of speculative stocks in the 1990s, a quick market decline left old stodgy indexes back in front. If professional money managers choosing from the hottest of the hot stocks couldn't beat an unmanaged index over 5-10 years, why would anyone want to pay them to
fall behind? That was the start of the indexing movement and thanks to the likes of Vanguard, it's still going strong today.
