So far, so good, but the debate is not over…and may not be for a while
In late 2005, PowerShares launched an ETF (ticker symbol: PRF) based on a controversial new “index” known as the Research Affiliates Fundamental 1000 Index (RAFI 1000). We now have over 5 years of return data for this fund, and this seems to me to be enough data for us to take a preliminary look at the strategy’s performance.
A quick comparison of PRF against popular traditional index funds such as SPY and VTI is shown below. This plot shows total returns (including dividends).
This graph shows that PRF has outperformed both SPY and VTI over the past 5 years, but it also experienced a larger drop during the crisis of 2008 and 2009. To properly evaluate the performance of PRF, we need adjust the returns for risk using an asset pricing model such as the Fama-French Three Factor Model (FF3F). However, before I get to the detailed analysis, I’d like to review what Fundamental Indexing is and why it is controversial.
What is Fundamental Indexing?
Fundamental Indexing is an alternative method for constructing a stock index.
Traditional indexes weight stocks based on market capitalization. Stocks with greater market capitalization are given a greater weight in the index. For example, Exxon-Mobil has over 10 times the market capitalization of Target, so Exxon-Mobil has 10 times as much weight in capitalization weighted indexes such as the S&P500 or Wilshire5000.
Fundamental indexes weight the stocks in the index using size related factors which are not tied to market price. For example, metrics such as book value, revenue, dividends, etc. are used to determine the weights of the stocks in the index.
Proponents of Fundamental Indexing believe that their method of weighting stocks is superior because they believe that there is overpricing and underpricing in the market which investors correct over time. Traditional indexes, by construction, will overweight overvalued funds and underweight undervalued funds. When the mispricing of these stocks is eventually corrected, it creates a “structural drag” on the returns of the index. Fundamental Index advocates do not claim to be able to identify these underpriced and overpriced stocks in advance, but they do believe that their method of indexing will randomize the valuation errors in a way that capitalization weighting cannot.
The most well known proponents of Fundamental Indexing are Robert Arnott, Jason Hsu, and Phillip Moore who wrote one of the first papers on the strategy back in 2005. Arnott is the founder of Research Affiliates.
Why is Fundamental Indexing controversial?
Fundamental Indexing has been criticized by traditional index advocates who believe that it is not true indexing and that weighting stocks by fundamentals constitutes an “active” bet away from the market portfolio. These critics also note that Fundamental Indexing is likely to have higher costs than traditional indexing because more trading is required to keep portfolio weights in line with the model. Prominent critics from this school of thought are John Bogle and Burton Malkiel. Their 2006 WSJ critique of Fundamental Indexing can be read here.
Other critics of Fundamental Indexing believe that weighting stocks by fundamentals is really just value investing. They say that any outperformance of Fundamental Index portfolios is due to the “value tilt” that is inherent in a system that weights stocks by fundamental factors. Prominent critics from this school of thought include Eugene Fama, Kenneth French, and Cliff Asness. Fama and French discussed Fundamental Indexing in a interview which is available here.
How has Fundamental Indexing performed so far?
As shown above, Fundamental Indexing has outperformed the S&P500 and the Wilshire 5000 over the past 5 years. However, we need to adjust the returns for risk, and one common method for making this adjustment is the Fama-French 3 Factor model.
I ran the Fama-French regressions on a number of more traditionally constructed index funds, and I found a couple that were reasonably close matches to PRF in terms of FF3F loading. IWD is an ETF based on the Russell 1000 Value Index, and IWW is an ETF based on the Russell 3000 Value Index. Both of these funds have a slightly lower value loading than PRF, but they are generally quite similar.
The results of the FF3F regressions using 5 years of monthly return data for these three funds are shown here:
| Fama French Factors | PRF | IWD | IWW |
|---|---|---|---|
| FF - Monthly Alpha (%) | 0.064 | -0.17 | -0.20 |
| FF - Beta | 1.02 | 0.92 | 0.93 |
| FF - Small | -0.08 | -0.16 | -0.073 |
| FF - Value | 0.42 | 0.35 | 0.35 |
| R-squared | 0.95 | 0.98 | 0.98 |
Notice that PRF has a small, but statistically insignificant alpha of 0.064% per month, or 0.72% per year. The negative and small SMB loading indicates that this fund invests primarily in large cap stocks, and the HML loading indicates that, as Fama and French have argued, the fund has a significant “value” tilt.
The traditional index funds both have a relatively large negative alpha. It is not quite statistically significant in the 5-year regressions, but it is large enough to be meaningful if it persists over the long term. The alpha is -0.17% per month or -2% per year for IWD, and it is -0.20% per month or -2.4% per year for IWW.
Based on these regressions results, it seems that Fundamental Indexing has done a better job of capturing the theoretical returns predicted by the Fama-French model than comparable value indexes. Does this suggest that the Research Affiliates Fundamental Index is really a better mousetrap for capturing the value effect? I’m not convinced, and the reason can be seen in the following plot.
All three funds track each other very closely until the low point of the financial crisis in 2009, but in the first few months of the stock market recovery PRF begins to pull away. I don’t know to what extent this is due to the different stock weighting strategy and to what extent it is due to the index reconstitution timing. However, it is very interesting that PRF is reconstituted in late March based on data from the end of February, and the Russell indexes are reconstituted in June based on data from the end of May. Here are the statements from the RAFI and Russell websites:
RAFI: The FTSE RAFI Index Series will be reviewed annually based on data as at the close of business on the last trading day of February, taking into account any additions and deletions planned in the underlying indices. Changes arising from the annual review will be implemented after the close of the index calculation on the third Friday of March each year.
Russell: On the last trading day in May (this year May 31), all eligible securities globally are ranked by their total market capitalization. Companies whose stocks are listed on eligible stock exchanges in eligible countries and who pass minimum liquidity and other investability rules are considered for inclusion in the indexes. Beginning on June 11, preliminary lists for the additions and deletions to the indexes are communicated to the marketplace. These changes go into effect after the close on Friday, June 25. All Russell indexes are subindexes of the Global Index.
In my opinion, the difference in performance between the RAFI index and the Russell-based value indexes is likely to be largely, or even entirely, due to this difference in reconstitution timing. In other words, the RAFI index rebalanced into the most beaten down stocks at the optimal time, and the Russell based indexes did not reconstitute until several months later after these stocks had already experienced a significant recovery. Thanks to ‘caklim00’ on the bogleheads.org forum for first pointing out that the divergence corresponds to reconstitution timing differences.
What can we conclude?
Unfortunately, financial data is very noisy, and it often takes a very long time to build up a reasonable confidence in any new theory or idea. The data so far shows that Fundamental Indexing has performed well, but the divergence from funds with similar factor loading may be due to the “luck” of reconstitution timing. Much more data is needed to establish with confidence that Fundamental Indexing is significantly different from more established methodologies.
Supplemental Info (Regression Results):
| Fund: PRF | Factor Estimate | Std Err | t-stat | p-factor |
|---|---|---|---|---|
| FF - alpha | 0.00064 | 0.0018 | 0.368 | 0.7140 |
| FF - beta | 1.023 | 0.038 | 27.204 | 0.0000 |
| FF - size | -0.08 | 0.078 | -1.033 | 0.3060 |
| FF - value | 0.42 | 0.066 | 6.356 | 0.0000 |
| Fund: IWD | Factor Estimate | Std Err | t-stat | p-factor |
|---|---|---|---|---|
| FF - alpha | -0.0017 | 0.0011 | -1.60 | 0.1100 |
| FF - beta | 0.92 | 0.023 | 39.92 | 0.0000 |
| FF - size | -0.16 | 0.048 | -3.39 | 0.0013 |
| FF - value | 0.35 | 0.041 | 8.50 | 0.0000 |
| Fund: IWW | Factor Estimate | Std. Err | t-stat | p-factor |
|---|---|---|---|---|
| FF - alpha | -0.002 | 0.001 | -1.98 | 0.0520 |
| FF - beta | 0.93 | 0.022 | 42.04 | 0.0000 |
| FF - size | -0.073 | 0.046 | -1.60 | 0.1200 |
| FF - value | 0.35 | 0.039 | 9.06 | 0.0000 |
The R code used for these regressions was a modified version of the regression code used for previous ETF evaluation.