| |
 September 2004 From the Bottom Up
That's not to say it's not monitored on an ongoing basis -- its performance is discussed every two months on the Historical Performance page and its year-to-date returns are updated every week at the bottom of the Home Page. Its holdings are reviewed on a daily basis although turnover is quite low -- only 10% this year though August. What P1 hasn't had is a thorough analytical and statistical examination. We're constantly doing that with Portfolios 3 - 6 in an effort to gather the data necessary to tweak their quantitative models. (You can follow that on a regular basis in Work in Progress and in the Archives.) ApproachUnlike P3 - 6 that are based solely on top-down quantitative models -- some might say "black boxes" -- P1 is good old fashioned bottom-up stock picking. There's no regard for sector weighting (although we do try to avoid concentrating in only one or two). There's no set number of stocks that have to be in the portfolio and no specific times for rebalancing.As of mid-August 2004, only four sectors are represented: Technology (32.6%), Healthcare (31.6%), Industrials (19.8%), and Financials (16.0%). The average market cap is $105.5 billion vs. $89.5 billion for the S&P 500, and forward 4-quarter P/E is 14.9X vs. 15.8X for the index. Also unlike the quant portfolios, P1 is a value portfolio. It's not a "deep value" portfolio in that it's not comprised of stocks of broken companies scooped up on the cheap. Without a catalyst to return to fair value, such stocks can remain "cheap" for quite some time. Instead, in order to be considered for inclusion in P1 stocks must be large caps selling towards the low end of their historical P/E, Price-to-Book, and Price-to-Sales range and at the high end of their Return on Capital (ROC) range. We use ROC instead of Return on Equity (ROE) since unlike the latter, the former can't be improved by simply taking on more debt. The portfolio's ROC is about 5% ahead of that of the index.
It also doesn't hurt if there is a catalyst to move the stock in the foreseeable future. As the low turnover attests, this isn't a trading portfolio, but you still don't want to wait forever to see results. So if a stock is oversold or ignored because its industry has a black eye from missteps of its competitors, that's a good thing. There's no specific holding period once a stock is selected for inclusion. Like Bill Miller and his Legg Mason Value Trust, stocks may be purchased when they represent values, but can remain in the portfolio when things turn around and the growth guys come in. There's no rule that value investors have to sell as soon as they start making profits. The oldest holdings (Wachovia and Capital One Financial) have been around since inception back in July 1991. Actually it was their predecessors (First Union and Pioneer Financial, respectively) that were originally purchased, but it doesn't matter. General Dynamics is the largest single holding (19.8%) and Wachovia is the smallest (3.1%). Just as there's no attempt at sector weighting, there's no attempt to weight the holdings relative to one another, either. Despite being a value portfolio, no emphasis is placed on dividends in the stock selection process. The underlying presumption is that most equity investors -- at least in the 1990s -- weren't really seeking current income but rather capital appreciation. In fact, we haven't even tracked total return statistics for P1, only capital appreciation. Just for the record, as of mid-August, P1 was yielding 1.3% vs. 1.7% for the S&P 500 and 2.0% for the S&P/Barra 500 Value Index. ReturnsSo how did this approach work out? Chart 1 shows how $100 invested on June 30 1991 would grown by June 30, 2004.Since P1 is a large cap portfolio, the S&P 500 was a logical benchmark. To make the comparison more appropriate, Chart 1 illustrates only the capital appreciation component of the S&P 500's total return.
Actually, since P1 is a value portfolio, the S&P/Barra Value Index, the value component of the S&P 500, may be an even more appropriate comparison. Its capital appreciation is also included on Chart 1. Unlike quant portfolios 3 and 4 that had the misfortune to be introduced right at the beginning of the 3-year bear market, P1 was initiated just as the bull market of the 1990s was getting underway. That got it off to a good start and it has been able to build its lead over the benchmarks ever since. As you can see from Chart 1, $100 invested in P1 would have grown to $489.84 by June 30, 2004 while that same $100 in the S&P 500 would be worth $307.37, or $280.00 if invested in the S&P/Barra Value Index.
Chart 2 compares periodic returns over the life of P1. Although it's solidly leads the indexes over the entire period, it's not always on top. In fact, over the past three years -- two of which were bear market years -- P1 was the poorest performer, losing almost 9% while the benchmarks were off less than 3%. At first this may seem odd since value portfolios tend to hold up better in down markets, but upon closer inspection it does make sense. P1 is not deep value and as a result, it's holdings have more to lose than stocks that are already selling at rock bottom prices. In such a situation it's understandable that the S&P/Barra Value Index would win out, but the overall S&P 500's outperformance isn't so clear. Indeed, over the past five years -- three of which were bear market years -- P1 and the S&P 500 were off 3.54% and 3.63%, respectively, both trailing the S&P/Barra Value Index (-1.78%). This is more of the pattern you'd expect so perhaps the three year period is more of an exception than a rule. Aside from the past year when it was overweighted in Healthcare issues that underperformed the market, P1 has tended to exceed the S&P/Barra Value Index in up years. Again this would be expected since P1's stocks are quicker to recover than some of the deeper values in the Value Index. Even so, P1 wasn't able to keep up with the overall S&P 500 over the ten year period which included the growth-led bull market of the 1990s. Again, that's what you'd expect.
Turning to the short-term, P1's monthly returns certainly vary -- from a high of 21.96% to a low of -20.39% -- yet fall into a normal bell-shaped pattern. They're illustrated in Chart 3. The high and low extremes are really outliers, with just over 50% (89 of 156) of the monthly returns falling between -4% and +6%. The greatest frequency of returns is 2-4% with 27 months (17%) falling in this range. Frequencies fall off on either side of the averages. Statisticians call this a "normal distribution". From a quantitative perspective this is a good thing since returns fall into a regular pattern, enhancing predictability. Speaking of predictability, a great deal of the variation in return of a diversified portfolio can often be explained by that of its benchmark index. If the portfolio is compared to the correct benchmark, both have at least some common holdings. External and market factors affecting the benchmark should have a similar effect on the portfolio. This is an issue of correlation. An index fund, explicitly designed to track a particular index, will obviously have a high correlation with its benchmark. Many actually duplicate the holdings and weightings of the index, so for them correlation is close to a perfect +1.000. Other large and diversified funds may not hold all the stocks in the index, but the more they do, the greater the correlation. P1, while somewhat diversified, holds far fewer stocks than either of its benchmarks. Nevertheless, the correlations since July 1991 have still been relatively high: +.802 with the S&P 500 Cap App and +.791 with the S&P/Barra Value Cap App. Taking this a step further, we ran linear regressions with both. Chart 4 shows the results for the S&P 500 Cap App. There each box is derived from the returns of the index (measured on the horizontal axis) and P1 (measured on the vertical axis). The line on the chart is the so-called "best fit", which is the line with minimal distances from each box.
If P1 and the S&P 500 Cap App were perfectly correlated, all the boxes would fall on the best fit line. Clearly they don't, but they are closely bunched around it illustrating the high degree of correlation. According to this analysis, 64% of P1's return is explained by that of the S&P 500 Cap App. It's not an index fund, but it's still at least partially influenced by its benchmark. This relation is mathematically captured by the equation at the bottom of Chart 4. By substituting in a given return for the S&P 500 Cap App ('X' in the equation), you can derive the "predicted" value for P1 ('Y' in the equation). It's statistically significant, but don't be the farm on it. As expected, the linear regression with the S&P/Barra Value Cap App yielded similar results. Since correlation was a little less, the regression wasn't as precise. Approximately 62.5% of P1's return is explained by that of the Value Index. RiskTo this point, we've focused on return, but what about risk? As you probably know, there's a direct relation between risk and return: The more risk you assume, the more return you can expect. How risky is P1?It's no great feat for P1 to meet or exceed the returns of its benchmarks if it's inherently riskier -- in that case it's supposed to. Under the best possible scenario, it would beat the benchmarks while having less risk. Now that would be remarkable. Chart 5 is a graphical representation of the risk and return relations between P1 and its benchmarks. Annual standard deviation, a measure of risk, is plotted on the horizontal axis and annual return is measured on the vertical axis. The dashed line is the Security Market Line. It's the line that passes through both the short-term risk-free rate (here the 30-Day T-Bill rate) and the S&P 500 return. Efficient portfolios can be created along the line by combining different weightings of these two series. These mixes are said to be "efficient" since they produce the highest return for every level of risk.
Chart 5 doesn't really have any surprises. You knew from Chart 2 that P1 had a higher annualized return than either of the benchmarks, so it's not unexpected to see it plotted higher up the vertical axis. (You might have also noticed that its expected return (15.88%) is higher than that on Chart 2 (13.00%). This is the difference between the arithmetic mean (Chart 5) and geometric mean (Chart 2).) P1's excess return comes at a price -- more risk. Notice how much further to the right it plots than the indexes. Specifically, P1's standard deviation is 26.08% while the benchmarks are identical at 15.98%. So more risk provided more return, but is that a good thing? All three series fall below the Security Market Line so none are as efficient at Modern Portfolio Theory would predict. The trick then is to find a common ground to compare them. The Sharpe Ratio is the way to do it. It's a measure of risk-adjusted return calculated by subtracting the risk-free rate from a portfolio's return, and then dividing that result by the portfolio's standard deviation. The greater the standard deviation (risk), the lower the ratio at any given level of return. There's that direct relation between risk and return again. In this case, P1 has a Sharpe Ratio of .1899, the S&P 500 Cap App's is .1930, and the S&P/Barra Value Cap App's is .1830. The differences aren't that dramatic, but they do show P1 slightly behind the broader index yet ahead of the value index. After 13 years, it's remarkable how close they are. ConsistencyOne last thing to consider is P1's consistency. Not in regard to periodic returns this time but rather investment style. Remember P1 uses a value approach in selecting stocks, but has no specific sell discipline when its holdings start to appreciate.Chart 6 is a graphical depiction of style over time. It's divided into four quadrants with deep value on the far left and aggressive growth on the far right right. The center represents the "core" approach. On the vertical scale, large cap is represented in the top half and small cap at the bottom. Midcaps would fall around the midpoint. P1, the S&P 500 Cap App, and S&P/Barra Value are represented by the red squares, blue triangles, and green pyramids, respectively. Each icon represents the series' style for a 12 month period with the smallest representing the earliest and the largest the most recent. The average for each series for the entire 13-year period is represented by the darker icons for each.
Several things jump out from this chart. First both of the benchmarks have been remarkably consistent -- just what you'd expect. Since the largest companies have the greatest influence over the S&P 500, it stays at the upper end of the large caps and its average is right in the middle between growth and value. The S&P/Barra Value index is also near the top end of the capitalization scale and its average falls in the middle of the value segment. P1, on the other hand, is all over the map. The majority of its boxes fall in the large cap area, yet a number fall below the median in the small cap area. Quite a number fall on the growth side of the scale with the average falling slightly in the growth area. None of this is too surprising. While it is true that all the stocks ever included in P1 came from the S&P 500, there has never been (and still isn't) any requirement that they be from the large end of the capitalization scale. Currently 5 of the 9 companies represented in the portfolio have market caps below the S&P 500 average with the smallest, AmeriSourceBergen, being only $5.97 billion placing it in the middle of the midcap range. So it's no wonder P1 has moved around in the capitalization range. The movement between value and growth is understandable, too. Many of the current holdings can now be viewed as growth stocks although they were values when added. Unlike the S&P/Barra Value Index that periodically removes stocks that become too growthy, there's no such requirement here. As long as P1's holdings remain fundamentally sound, there's no need to sell them simply because they may have changed their value/growth classification. That's why the S&P 500 is a more appropriate benchmark than a value index. The broader index has always been the measure and as you can see from Charts 1 and 2, P1 has fared well against it. And that's probably the most important thing to take away from this. When suggesting how to structure your portfolio we offered the following: [H]ow many stocks do you need for proper diversification? Probably less than you think. Studies conducted in the early 90's showed that there is little additional benefit once you surpass 10 - 15 diversified holdings...Look for stocks in different industries...Also consider buying stocks of different size companies. The statistics show that this is precisely how P1 has been constructed. As Chart 2 illustrates, it didn't outperform in every short-term period, but overall it has kept up with and exceeded the return of the benchmark index. P1 provides solid evidence that it is possible to build a successful portfolio from the bottom up. You don't need hundreds of holdings or expensive turnover. You don't need to throw in the towel and settle for mere indexing. You do need to carefully pick and structure the selections and perhaps most of all, you need time to get a true picture of the results. Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||||||||||||||
O MATTER HOW YOU DEFINE LONG-TERM and short-term, thirteen years has to be considered long-term. That's how long Portfolio 1 has been around. After 13 years, it's probably overdue for a thorough review.
