Rob Arnott caused shock-waves in the index fund and ETF world with a 2004 paper which claimed that the indexes used by most index funds and ETFs are inefficient and lead to underperformance. Instead, he argued, market-cap weighted indexes should be replaced by indexes weighted by companies’ fundamentals, such as sales, revenues, book value, earnings, or number of employees. Mr Arnott's work was discussed by Mark Hulbert and Jonathan Clements, and PowerShares subsequently announced an ETF based on a fundamental index, the PowerShares FTSE RAFI US 1000 Portfolio (PRF). However, a reader of ETF Investor submitted a brief but important critique of fundamental indexes, which we published as Fundamental Indexes Don't Outperform. Now, in this article written for ETF Investor, Mr Arnott addresses the issues raised:
Fundamental Indexes have received a tremendous amount of attention and scrutiny in the finance community recently. Much of the debate is centered around the attribution of their out-performance relative to standard capitalization-weighted indexes. The U.S. RA Fundamental Index (OTCPK:RAFI) outperforms the S&P500 by more than 2% per annum over the last 40 years, and internationally, the Developed Countries x-US RAFI outperforms the MSCI EAFE by 3.5% per annum over the last 20 years.
Investment pundits have correctly pointed out that the observed excess returns over standard indexes can be accounted for by the Fama-French factors. This is unsurprising as most known passive investment portfolios do not exhibit positive alpha net of Fama-French exposures. If they did, the FF model would be fundamentally flawed. From both the academic and practitioner's literature, there has been no evidence that the FF attribution does not completely capture sources of returns in passive indexes. Therefore, using the FF model to assess the "goodness" of a model is, in fact, quite appropriate. Particularly, well-constructed indexes should exhibit FF alpha that are non-negative.
The Fundamental Indexes exhibit modestly positive FF alphas, which are not statistically significant. Bad indexes, by comparison, shows statistically and economically significant negative FF alphas. S&P Equal Weight has a -1.5% alpha in the Fama-French return decomposition (FF factors say it should add 2.5% in excess return relative to the broad equity market return, but it only adds 1%). Russell 1000 Value has a -1.8% (FF factors say it should add 2.4%, but it only adds 0.6%). And the list goes on. I think the surprise is that most indexes underperform net of FF factors.
From a conventional value-added point of view (relative to standard benchmark), no known indexes add as much value as Fundamental Indexes, without taking on massive tracking error relative to the benchmark. RAFI adds 2.1% per annum over the last 43 years in the US and adds value averaging 2.7% per annum in 23 out of 23 EAFE countries, with no exceptions (see second attachment). If beating conventional benchmarks is the objective, RAFI delivers out-performance, without a question. Certainly, one may try to replicate RAFI by adding Fama-French tilts to a cap-weighted benchmark portfolio (through complicated long-short equity overlay programs derived from active Fama-French value and size portfolio), but that's a whole lot messier (and more expensive) than just investing in a Fundamental Index portfolio!
Chairman, Research Affiliates, LLC