Choosing the Right ETF: Growth Versus Value in Asset Allocation

|
Includes: DIA, HHH, IUSG, IUSV, IVE, IVW, IWD, IWF, IYC, IYM, QQQ, SPYG, SPYV, XLB
by: Geoff Considine

An interesting and seemingly perpetual debate in the investing world is as to the relative merits of growth vs. value in terms of asset weighting. If, by some remote chance, you are not familiar with this seemingly perpetual discussion, just use a search engine on the term growth vs. value.

The basic definitions are that value firms tend to have low prices relative to fundamentals such as earnings per share or book value. Stocks designated as value stocks also often have high dividend yields. Growth stocks have much higher prices relative to fundamental measures of value, theoretically because the market is pricing in the potential for much higher earnings growth. There are abundant articles and research that suggests that value investing beats growth investing over time. Roger Ibbotson, for example, champions the idea that value trumps growth, with the caveat that that growth stocks can outperform value stocks over extended of time (link, .pdf).

Ibbotson’s analysis uses the underlying market indices and projects back to 1979. They show, for example, that the Russell 1000 Value Index generated 1% more in compound annual return than the Russell 1000 Growth Index from 1979-1997, and that the value index exhibited a standard deviation in annual return substantially less than the growth index. This is a comparison of large market capitalization firms, of course, but they found similar, but even more dramatic, results in other size classes. They also found consistent results using other indices.

In short, their results show that over this fairly long period of market history, ‘value’ indices beat ‘growth’ indices on an absolute basis and with less risk. Does this mean that investors should weight their portfolios towards ‘value’ funds rather than ‘growth’ funds? It would be wonderful if we could build market-beating strategies by simply buying more value-oriented funds and less growth-oriented funds, but the story is not yet complete.

John Bogle has examined the growth vs. value issue in depth in his investing classic, "Bogle On Mutual Funds" (excerpted here ). He looked at investments in two groups of mutual funds, growth-oriented and value-oriented, over the sixty-year period through 1997. His data showed compound annual returns of 11.7% per year for growth funds and 11.5% per year for value funds. This is a wash. Ibbotson’s paper and Bogle’s book both use data up to 1997. Why do two very credible sources come up with such different answers? If you read the previous paragraphs very carefully, you may even find the source of the disparity yourself. In Ibbotson’s work that suggests that value investing beats growth investing, as well as in other analysis that supports this claim, the analyses use index performance while Bogle uses performance of real value-oriented and growth-oriented mutual funds. John Bogle revisited this topic in 2002, and came to the conclusion that the apparent advantages in value vs. growth indices disappeared when you looked at real mutual fund data going back to 1937.

In other words, the marked advantage if you could invest in the reconstructed ‘value index’ as opposed to the ‘growth index’ simply disappeared in the real world of funds. So, now you may be somewhat confused. If not, read the papers by Ibbotson and Bogle for yourself.

My conclusions from these works is that the idea that you can simply invest in a fund that is oriented towards ‘value’ to achieve superior risk-adjusted returns is unconvincing. Ibbotson’s paper cites many of the potential problems with his statistical results that show that value beats growth. One of the effects that he cites is that certain industries tend to have the characteristics of ‘value,’ so that a value-index will be more heavily weighted in certain industries. An over-concentration to a specific industry can introduce risks by limiting portfolio diversification.

Now let’s move to the practical world of asset allocation. Let’s say that you have read some articles on value investing and you have seen that value-oriented stocks have creamed growth-oriented stocks over the past three to five years. You want to look at the impact of increasing your concentration in value-oriented stocks. The ETF market provides the investor with the ability to invest in a basket of stocks that is designed specifically to track growth and value indices, as well as to invest by Morningstar class. To create two proxy portfolios, I looked for ETF’s that are categorized as either ‘Large Value’ or ‘Large Growth’ and that had at least three years of history. I also wanted to make sure that I minimized the overall impact of HOLDR’s because they are not, in fact, ETF’s even though they have some of the same characteristics. I also limited this example to domestic equity funds. The selection of funds that I ended up with is shown below:

Considine # 1

Selection of growth-oriented and value-oriented ETF’s

As expected, the price-to-earnings ratios are much higher for growth funds than for value funds. If you built a portfolio that is equally allocated to the growth funds, how would you expect it to look relative to an equally-allocated portfolio of value funds? We know that value funds have greatly outperformed growth funds recently, but what do we think will happen in the future?

Quantext Portfolio Planner [QPP] is a portfolio planning tool that combines fairly recent historical data on individual assets with long-term statistics on the risk-return balance of capital markets to create a projection of future portfolio performance. QPP is a Monte Carlo model which means that it calculates many different possible future market outcomes based upon the relative probability of each outcome. QPP accounts for correlations between assets and expenses.

These two features are especially important in light of Ibbotson’s comments regarding the potentially poor diversification effects of value stocks (because of industry concentration) and Bogle’s point regarding the importance of looking at real funds and their fees. QPP’s approach to simulating real funds is especially important given that the value-vs.-growth spread in performance appears in Bogle’s analysis when you look at indices but disappears when you look at real funds. For this analysis, I have used the recent three years for the near-term period from which to calculate certain parameters. QPP then adjusts the projected future performance of asset classes so that they are consistent with the long-term balance of risk and return in capital markets. These projections are also conditioned by the input assumption that the S&P500 will yield an average return of 8.3% per year, with a standard deviation of 15.07% per year. These estimates are consistent with the current consensus in the financial community.

When I run the pure growth portfolio through QPP, I get the following results:

Considine # 2

Equal allocation growth portfolio in QPP

There are a number of interesting features of the historical and projected future performance of this portfolio. Over the past three years (through May 2006), the S&P500 has yielded an average return of 9.41% with a standard deviation in annual return of 7.8% (see Historical Data above). The growth portfolio has generated an average annual return of 9.42%, but with a standard deviation of 11.48% (also shown in Historical Data above). This growth portfolio has generated the same average return as the S&P500 but with more risk. The simulated future results are shown in the table with the Annual Average Return for each asset and for the portfolio as a whole in Portfolio Stats. The projected future performance of this portfolio is an average annual return of 10.79% per year, but with a standard deviation of over 20%.

This is a high volatility portfolio. This portfolio also has Beta of 133% and a very low dividend yield of 0.7% per year. One of the unique tools that QPP provides is the ability to measure diversification effects among portfolio components that are not captured by Beta. This is measured by QPP’s Diversification Metric [DM]. The higher DM is, the less correlated (i.e. better diversified) the portfolio components are. DM is 15% for this portfolio, which we will compare to the results for the value portfolio.

When I run the equally-allocated value portfolio, the historical performance is very different. This portfolio generated an average annual return of 19.93% per year with a standard deviation in annual return of 9.95% per year over the past three years (see Historical Data below). This portfolio has generated substantially higher average return than the S&P500 over the past three years with less risk as measured by the standard deviation in return. This portfolio has a Beta of 118% and a dividend yield of 1.7% per year.

These results confirm, to some degree, what we already know: value stocks have been outperforming the growth stocks for some time now. The projected future performance (Portfolio Stats below) is for an average annual return of 9.47% per year with a standard deviation of 17.94% per year. This portfolio is projected to generate less average return but also to be less risky than the growth portfolio. When we look at the Diversification Metric (below), we find a value of 11% which is well below the 15% that we obtained for the equally weighted growth portfolio. This simply tells us that this portfolio is more concentrated than the growth portfolio, as we expected. This “concentration” effect is largely determined by the exact set of funds that we included in each portfolio, so I would not put too much interpretation into that value.

Considine # 3

Equal allocation value portfolio in QPP

Having seen that the value portfolio is both lower risk and lower average return than the growth portfolio, it is of interest to re-allocate the value portfolio so that its total risk level matches that in the growth portfolio. When we have risk-matched the portfolios, we can then compare the projected average annual return to see if QPP is projecting any ‘value premium.’ This is an important point. The equally-allocated portfolio of growth funds is projected to generate a higher average annual return than the equally-allocated portfolio of value funds. It is not terribly meaningful to compare these average returns without consideration for risk.

Considine # 4
Value portfolio with historical risk (i.e. Standard Deviation in return) matched to equally-allocated growth portfolio

To get the same level of historical risk (measured by the historical standard deviation in returns in Historical Data above), I had to increase the portfolio allocation to the higher return components of the portfolio (above). This portfolio has 41% of its total allocation in two ‘basic materials’ ETF’s (IYM and XLB). Even at this high correlation to one sector, the Diversification Metric has increased to 25%. This effect occurs because of the very high correlation (low diversification) between the other value-focused funds. This is not a portfolio that I would advocate for anyone—this is simply to explore the value vs. growth issue. This portfolio has identical historical standard deviation in return—total risk—as the equally-allocated growth portfolio. This portfolio has very slightly less projected future risk (standard deviation in annual return) and slightly lower projected annual return but given the uncertainties in all of the projected future variables, these portfolios are essentially identical in terms of risk and return. Note that the re-balanced value portfolio also now has a Beta of 130.8%, very close to the Beta for the equally-allocated growth portfolio.

The results from QPP are consistent with John Bogle’s analysis of the relative performance of growth funds vs. value funds. QPP suggests that simply investing in ‘value’ funds will confer no advantage. This does not mean that there are not value-oriented funds that might be highly attractive in a portfolio. Energy utilities (typically a low P/E sector) such as IDU are considered as ‘value’ investments and these typically have very positive portfolio impacts. In other words, the best approach to portfolio building is to choose specific stocks or sectors that have lower prices relative to earnings, book value, or dividend yield rather than attempting to purchase a ‘value’ fund. I find that the entire process by which broad ‘value’ and ‘growth’ funds choose their components seems to muddy the asset allocation problem with no apparent advantage.

In looking at an arbitrary selection of growth and value funds, part of my purpose was to test QPP. The results from QPP are what you would hope to find if you believe John Bogle’s analysis, and I would put myself squarely in that camp. The historical evidence for value funds (actual funds rather than reconstructed historical indices) to provide any long-term advantage is not compelling. I find it reassuring that QPP supports this view of the world. Unfortunately, many Monte Carlo and related portfolio planning tools rely too heavily on recent market history.

In the current environment in which ‘value’ funds have recently been the out-performers, these models will tend to suggest that being over-weighted in value funds will provide additional return. QPP automatically discounts part of recent performance relative to the long-term market balance between risk and return for individual assets and entire asset classes. While the ‘value’ funds in this sample have dramatically outperformed the ‘growth’ funds over the past few years, QPP’s projections suggest no difference in long-term performance once we account for differences in risk between the two portfolios. Many portfolio planning and Monte Carlo tools model portfolios in terms of the underlying indices (as with the Style Analysis used at FinancialEngines.com) and may thereby tend to support the apparent value-premium that is observed in studies of index performance (as opposed to performance of the actual funds).

This is a particularly interesting issue for portfolio planning models because these models are typically not capable of modeling highly-concentrated portfolio very well, and value-focused portfolio often end up being highly concentrated (as noted earlier and in the Ibbotson paper). QPP simulates the specific individual funds and has been proven to be able to account for highly-concentrated portfolios. Both of these capabilities contribute to QPP’s results in showing that there is no real value to simply buying ‘value oriented’ funds.

[Note: see Figure 7 in http://www.stanford.edu/~wfsharpe/art/sa/sa.htm to see how Style Analysis, the approach used for portfolio simulation in a number of other tools, fails to resolve the performance of a highly concentrated fund—a utility fund in this case. This paper was written by Nobel Laureate Dr. Bill Sharpe and provides an outstanding explanation of Style Analysis. ]

When our results from QPP show that there is no inherent advantage to investing in ‘value’ funds more heavily that ‘growth’ funds, does this mean that we are saying that fundamental measures of the price of a stock such as price-to-earnings or price-to-book don’t matter? The answer to this question is resoundingly to the negative. The price at which your purchase a stock matters. In fact, I am more and more convinced over time of the general superiority of using dividend yield as a measure of value in a stock. There is a substantial difference between timing an increase or decrease in an allocation in an asset using fundamentals and simply investing in a ‘value fund.’ The change in the price of a stock relative to its earnings, dividends, or book value over time is a relative measure of the changes in its price through time as compared to growth in the underlying business. Investing in value or growth funds simply weights your investment relative to the rest of the market at a given instant in time. This is a slightly subtle but very important concept.

To summarize, long-term historical analysis of the performance of actual ‘value funds’ vs. ‘growth funds’ by John Bogle show that there is no real advantage in investing any more heavily in value-oriented funds than growth-oriented funds. I find his analysis and results compelling and reasonable. When I analyze an arbitrary portfolio of growth ETFs in Quantext Portfolio Planner and compare these to a selection of value ETFs, the projected future performance of these portfolios is the same once we control for portfolio risk effects.

This result is especially notable given that value-oriented funds have dramatically out-performed in the last three to five years. Bogle’s common-sense approach to testing the ‘value vs. growth’ dilemma suggests that it makes very little sense for investors to think in terms of the fairly arbitrary ‘value’ and ‘growth’ distinctions and QPP supports this conclusion.

More information on Quantext Portfolio Planner and a free trial are available at http://www.quantext.com/gpage3.html