This article explains why and how I created Buy and Sell (and Hold) "signals" to help investors make important action choices within their ETF and fund portfolios. Since the empirical groundwork for these signals was first completed back in July, 2008, the results in applying these classifications thus far have been so outstanding, it is important for readers to consider what the signals are showing now.
It is a well-established axiom that past performance should not be taken as indicative of future returns. However, while this suggests that strong past fund returns cannot guarantee good future performance, and vice versa, past performance data, when considered on broad investment classes and over the long term, can often provide some of the best data an investor has available to make decisions on the relative strengths and potential weaknesses of different asset classes.
Thus, whenever a broad asset class has performed poorly for a considerably long stretch, say 5 or even 10 years, investment experts have recognized that the investment class is more likely than not to do an about face. This is referred to as regression to the mean. Likewise, an investment class on a "tear," such as large cap US stocks between 1995 and 2000, are more likely than not to fall back to earth, as they did over the following 5 and even 10 year periods.
Having observed this happen so often, and given my background as a former research-oriented psychologist, I decided to construct an empirically-based scale which would try to weigh the likelihood of an asset class performing well or poorly based on the above premise. My assumption of an inverse long-term relation between returns and subsequent performance also assumed that an asset class doing far better than expected over a long period meant that the class had become overvalued. Likewise, poor long-term performance of an asset class tended to suggest it may now be undervalued.
Beyond this, however, we also know that momentum of an asset class seems to play an important role in that assets that are doing well seem to continue doing well. That is, if small-cap stocks for example, have been doing above average for say 6 months or a year, they tend to keep doing well significantly longer. Otherwise, established long-lasting patterns of upward growth would be hard to discern on a graph of a representative fund's net asset value (NAV, or price). But to the contrary, such trends are quite commonplace.
Given the importance of both over- and under-valuation and momentum, my challenge was to attempt to determine how important differing degrees of each of these two factors were to subsequent investment performance. A skeptic might assume that it would be next to impossible to develop a rating scale based on past performance of fund asset categories that could be useful in predicting how the same asset categories would perform in the future. They might say too many unpredictable variables influence stock fund prices, especially such things as interest rates, recessions, political considerations, etc., suggesting that future fund performance cannot be successfully predicted from a past performance composite no matter how one might try.
I looked at fund past performance patterns for a variety of broad stock asset categories going back several decades. I then looked at how the categories subsequently performed. This led me to create a numerical score for all possible combinations of long- and short-term performances, to reflect the sum of valuation and momentum attributes within a single numerical "score."
Based on past performances, ranges for scores were determined depending upon whether they tended to lead to good vs. poor subsequent total returns for the asset class. A "high" score meant that the asset category typically performed well; such scores were defined to be "Buy" signals. "Low" scores mean the category was mostly associated with poor subsequent performance and were presumably a poor choice for investors looking forward, that is, they were "Sells." Scores between the two extremes seemed to suggest from my study of prior returns that the investment class would do moderately well, but not as well as scores the data showed had been associated with Buy signals; these in-between scoring ratings were called "Holds".
While the data was all constructed from an analysis of past performance, and therefore only suggestive, I concluded that the only way to argue that the ratings could truly be considered valid is to apply them to more recent returns and future asset class performances. So, starting in Aug. 2008, I did just that. That is, I assigned scores, and associated Buy, Sell, or Hold designations, to 10 broad categories of stock funds based on their past long- and shorter-term performances and then waited to see if the results for average category performance were commensurate with the 3 designations given. The following two tables show the results.
Table 1: Performance for 10 Vanguard ETFs
57.69 / 46.43
57.04 / 44.88
55.66 / 42.71
Mid-Cap Growth (VOT)
56.63 / 41.31
66.88 / 51.02
Mid-Cap Value (VOE)
47.21 / 37.67
Sm-Cap Growth (VBK)
66.36 / 52.24
63.62 / 50.17
Small-Cap Value (VBR)
60.60 / 47.76
50.13 / 39.54
Note: Dividends are not included in percent changes.
Table 2: Performance for 10 Vanguard ETFs
49.76 / 70.30
49.50 / 64.19
47.04 / 57.79
45.94 / 68.92
57.25 / 81.71
42.65 / 57.62
39.42 / 86.93
54.18 / 78.77
52.03 / 70.48
43.94 / 44.60
Notes: 1. Dividends are not included in percent changes.
2. Small Cap Growth BUY Alert from Jan. 31, 2009.
3. Large-Cap Growth BUY Alert from Oct. 8, 2009.
4. All other categories BUY Alert from Nov. 12, 2009.
5. Another BUY Alert for all categories issued on July 18, 2011.
(VOO NAV up about 7%, not annualized, since the signal.)
6. All above categories are now Hold.
The tables results are dramatic and speak for themselves: Sell signals were followed by drastic declines in average asset class prices over the following year. Then, when signals turned highly positive, in 9 out of 10 index funds serving as proxies for an entire fund category, longer-term returns were quite positive to varying degrees over the following 28 to 37 months, or roughly, two to three years.
In conclusion, although three and a half years is not a long time as far as investment returns go, the results presented here, based on two decades of prior studies of returns, are highly suggestive that my effort to design a tool to help investors determine whether an investment category might be considered Buy, Sell, or Hold, has been highly successful thus far in an up and down market environment. The results suggest that the signals should be considered for current and future use by investors in making important relatively long-term investment decisions, or even deciding where to rebalance within stock fund asset categories.
While current signals for all categories are no longer Buys, but rather Holds, my prior several decades of data show that returns associated with past Hold ratings are also quite positive, possibly reaching within the range of about 11 to 13% per year over the next several years. Past and current signals are discussed further within my website and updated ratings are reported whenever they change, which tends to be perhaps only once or twice a year.