| |
 November 2011 Good May Not Be Enough P5 Has Exceeded Its Benchmark, But Has It Met Its Goal?
At the turn of the 21st Century, investors were fixated on "relative" performance. Wall Street and the advisors of the day had sold this concept, claiming to add alpha. In reality, those who did manage to beat the indexes were often relying on beta for an increased dose of market risk. That worked fine until stocks collapsed when the tech bubble popped and that beta came back to haunt them. Investors suddenly switched from worrying about beating an index to simply getting their money back. When we created P5, we not only wanted it to beat its benchmark, we also wanted it to be as efficient as possible in doing so. That meant not only producing leading returns to the upside but also providing downside protection, too. The past decade offered a good opportunity to evaluate this. Stocks began a five-year recovery shortly after P5 was launched in 2002, providing a good opportunity to gauge performance to the upside. The bear market induced by the 2007-2008 credit crisis was a chance to see what happened when seemingly all investments went down the drain. Most recently, the weak recovery and range-bound trading offered a glimpse of how the model would work in a stagnant market like that of the 1970s. At first glance, P5 seems to have done well (click here for a cumulative return graph). Despite changing market conditions, P5 jumped to a lead over the S&P 1500 Super Composite and constantly maintained it . But what about its overall risk and did it really always have the most efficient allocation of holdings? In other words, it obviously did well, but should it have done better?
The Standard Unlike prior models 1-4 (Portfolio 6 didn't come into existence until two years later), P5 was going to be a combination of both capitalizations and styles. The goal was to optimize frequently in order to capture the most efficient mix of holdings. If done properly, the stock selection and market timing would add value over and above that of the simple unmanaged index. The S&P 1500 Super Composite was chosen because as the combination of the S&P 500, 400, and 600, it covers all capitalizations and styles. In order to make P5 manageable and economically tradable, it only employs exchange traded funds (ETFs) rather than individual securities. This also aligns the potential holdings more closely with the proxies used in the optimization process while eliminating most diversifiable risk.
The concept behind the allocation is simple. Think of the Morningstar domestic equity stylebox which is nothing more than a 3x3 matrix. The three capitalizations (Large, Mid, and Small) have their own rows and the three styles (Growth, Blend, and Value) each have their own column. This allows the entire range of the domestic equity market to be modeled with only nine potential holdings. Once optimized, the corresponding Barclay's iShare ETFs are purchased (or sold) as necessary to reflect the desired weightings. Although no individual holding exceeded 50% since inception, there are no limits on maximum or minimum weights. The optimization process is completely unconstrained. Resampling does help spread the allocation a bit. It didn't take long for a pattern to emerge: P5 loves mid-cap stocks -- at least it has so far. We've commented on it several times over the past then years (most notably here and here) and have explored some of the causes. Regardless, mid caps served the model well given they were the leading share classes over the past ten years. Now that large caps appear to be reasserting their leadership, it will be interesting to see if P5 will move away from its beloved mid caps. There's one other aspect to P5's optimization process, the level of risk. As you probably already know, a mean-variance optimizer allows the user to pick the specific classes to use and then creates an efficient frontier of asset mixes generally running from the most conservative with the least expected return to the most aggressive with the greatest chance of loss (and return). Presumably any portfolio along the frontier is "efficient" in that it is expected to yield the greatest return for its specific level of risk. If it works as expected, the investor only needs to determine his or her acceptable level of risk and should then use the portfolio associated with it. It was tempting to simply use the midpoint of the risk range, which is exactly what we did when we built balanced Portfolio 6, but in that case we were trying to model the everyday buy-and-hold investor. The middle of the road was appropriate for that, but P5 didn't seem so constrained. So we turned to some earlier analysis we had previously started (and subsequently completed and posted posted) using historical risk and returns dating back to 1979. Based on risk-adjusted return as measured by the Sharpe Ratio, we found the optimal portfolio fell roughly 70% up the frontier extending from the least risk toward greatest risk. Historically, this was essentially a 75/25% split between stocks and bonds so it's also what became P5. One final thing about P5's risk: It isn't fixed. Every time the model is reoptimized, the allocation is determined to be the portfolio which has 70% of the maximum risk along the frontier. This value changes with overall market risk. The 2002 original P5 which came into being near the bottom of a three-year bear market, had less risk than today's version in the current volatile market. Had we decided to keep risk static at the 2002 level (12.27%) it would have been too low to appear on subsequent riskier frontiers that moved up the risk scale (see Chart 3, below). As a better alternative, it was decided to let P5's risk move with the market.
A Matter of Perspective Although P5 was based on analysis of historical asset class risk and return, the subsequent optimizations used forward-looking assumptions instead. After considering several different options, it was decided to use the Black-Litterman approach which incorporates estimates of returns for the current market and economic conditions as well as a degree of confidence. Using it allows P5's allocations to be more tactical -- an important point when the investment horizon is effectively three months.
Looking a little closer at the risk factors, you'll notice P5 spent four more months on the downside than its benchmark. Pay particular attention to the semi-standard deviation below 0 (next to last row on Chart 1). This is what most investors are worried about, the risk and size of loss. In this case, the S&P 1500 comes out on top. Putting all this together, P5 appears to be a slightly above-average long-term performer with increased volatility on the downside. Most investors would embrace the former, but not so much the latter. So could it be considered the "best" allocation/process over the period? The answer to that is, of course, dependent on how one defines benchmarks and expectations. For example, if the benchmark is simply the S&P 1500 Super Composite, P5 certainly did exceed expectations on both a return and risk-adjusted return basis. On the other hand, other dynamic portfolios may have accomplished that or even more. One such portfolio would have been the 70% Maximum Standard Deviation portfolio which served as its model. Chart 2 shows the efficient frontier using historical data from January 1, 2002 through September 30, 2011 -- the lifespan of P5. By the way, the axes don't cross at the 0,0 point, we've truncated the x-axis (standard deviation) to bring it closer to the y-axis (expected return) in an effort to make it easier to read. Notice how the efficient frontier for the entire measurement period has moved further to the right. That's why we needed to let that average portfolio risk drift up from the 12.55% of the original benchmark portfolio. If it didn't float, it wouldn't have been possible to create efficient tactical portfolios as the market grew riskier.
The Missing Factor
Back on Chart 2, P5 and the S&P 1500 are highlighted, and so is the 70% MSD. You'll notice that the latter is the only one of the three on the efficient frontier. By definition, the 70% MSD has to be on the frontier, but in comparison, P5 appears to be much less efficient. There's another way to gauge P5 and it's also shown on Chart 2: the holdings based approach. Unlike a static asset allocation, P5 can -- and in most cases did -- change from quarter to quarter. The point labeled "P5 Holdings Based" represents the risk and return of the time-weighed asset class holdings. It's much more efficient than P5's actual results and its also much closer to the 70% MSD portfolio. The composition of this portfolio is shown in Chart 3. This raises the question, "Why are P5's actual results so different than its target portfolio (70% MSD) and the average weighted portfolio?" The former was used to create the dynamic P5 while the latter is the average weighted mix. Clearly there are other factors involved. One thing to bear in mind is the fact that P5 is actually the result of optimizing historical stock and bond performance with various forward-looking economic statistics. There's no rocket science here, just a basic optimization based on what we've come to expect from various market and economic conditions. The forward-looking figures offer higher expected risk and lower expected return than they have historically presented. There's little wonder then that actual P5 risk and return would consistently fall below the historical valuations. Just as the the policy portfolio would be expected to have a major impact to relative returns, selection would not seem to be very important. However, because it's not possible to directly invest in any index itself, P5 has to invest in proxies, in this case the S&P style and capitalization ETF series. These funds closely track the respective indexes but they also carry expense ratios ranging from 0.10% - 0.27%. As small as they are, these fees cut into the funds' return and send them below that of the unmanaged index.
It's also important to note that even these are the maximum expenses of the average iShares index ETF, there are periods when the actual charges are a little lower. This can happen if the fund boosts returns by lending holdings in exchange for short-term interest payments or if due to cash flows and accounting rules, returns briefly outperform the index. This is what happened with the ETFs of P5. Instead of causing the model to trail annual returns by 0.26% (P5's average weighted expense ratio), security selection only cost the fund 0.11% per year. This is a plus on a net basis. The third contributor to performance, market timing, appears to have played little if any role in the results. While it's true P5 didn't suffer from wild allocation swings, it was rebalanced once every two months from inception. Nevertheless, attribution data suggests timing had zero effect on the ultimate return. These three components of return are summed up on Chart 4, where the result is a positive 0.76%. The negative 0.11% effect of security selection partially offsets the +87% stemming from P5's policy portfolio. It's hard to say if P5 was, or ever had been, the "best asset allocation". What's more obvious are the facts that P5's capital appreciation has exceeded that of the benchmark portfolio in the the past few years when domestic large caps have struggled. Also, mid caps have always comprised a major portion of P5, which in hindsight, was optimal dynamic asset allocation. These are two things P5 definitely did consistently and well.
However, consistent relative performance is not identical to overall outperformance. While it's encouraging to see P5 out in front if its two benchmarks, that doesn't necessarily crown it as king of all allocations. If anything it shows true superior performance will have to come from all sectors and virtually each individual security, not which is destined to be the next top performer. Arguably, a static approach with the right mix could have performed as well as P5, the major difference would be prospectively finding the correct initial static asset allocation. As you've probably already gathered, however you come down on P5's performance will be based on your personal perspective. Considering only risk and return, P5 has been quite a success. Also, focusing on the portfolio's stated asset allocation, it's has been relatively efficient. What more could you ask for? Clearly other points along the actual historical frontier would bear less risk while possibly delivering greater risk-adjusted return. If this is your goal, P5 probably isn't your portfolio. Then again, that's not P5's goal -- it was designed to beat the S&P 1500 which, in fact, it did. The consistent performance demonstrated by P5 goes over and above this. It's highly likely P5 wasn't the absolutely best asset allocation, but the chances of finding that prospectively are right around zero. Does that really matter when P5 regularly exceeds the return of the benchmark S&P 1500 Super Composite with similar levels of risk? Probably not. Search this site! Just enter you key word or words: Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||||||||||||||||||||||||||||||
T'S HAD PLENTY OF UPS AND DOWNS along the way, but domestic multi-cap Portfolio 5 has steadily outpaced its benchmark, the S&P 1500. Now with its ten-year anniversary just around the corner on January 1, it's time to look back to analyze these results. In a world where the average mutual fund manager (and many separate account managers as well) can't beat their respective benchmark on a consistent basis, P5 looks like a sure winner. But it was designed to do more than simply exceed a benchmark, it was supposed to be the "best" allocation of the available resources. You have to look deeper than simple return to determine this.
