Robert Arnott

A proposed index improvement became a live test of value and rebalancing.

The proposition began as an index experiment

Robert Arnott founded Research Affiliates in 2002 and, with Jason Hsu and Philip Moore, published "Fundamental Indexation" in 2005. The paper asked a narrow question with large consequences: could a broad stock index retain the low-turnover and transparent character of indexing while assigning company weights without using current market prices?

The proposed alternative used accounting measures of company size. Revenue, sales, cash flow, book value, dividends, and employment each produced a different simulated index. A composite combined several measures. The design was not an estimate of each company's intrinsic value. It was a set of rules for deciding how much of each company to hold.

This distinction matters because Arnott's work spans several roles. Research Affiliates currently identifies him as partner and chair, active in research, portfolio management, product design, and business strategy. Fundamental indexation is both a research proposition and licensed investment technology. Evidence for the proposition cannot be inferred from the firm's commercial reach.

What capitalization weighting actually does

A capitalization-weighted index assigns a company a weight proportional to the market value of its eligible shares. If its share price rises while everything else remains unchanged, its index weight rises automatically. A fund tracking the index normally does not need to trade merely because of that price movement.

Arnott, Hsu, and Moore argued that this convenience creates a problem when prices contain errors. Relative to an unobserved fair-value portfolio, securities priced too high receive too much weight and securities priced too low receive too little. A weighting anchor independent of price would not magnify the same error, and later rebalancing could benefit if the error reversed.

A cap-weighted index also has a different purpose from an active estimate of fair value. It represents the investable market in proportion to quoted value. It can be a useful benchmark even if it is not the portfolio with the highest future risk-adjusted return. Fundamental indexing changes that objective: it seeks a transparent rule that may earn more than the cap-weighted benchmark.

How the published test was constructed

The 2005 paper built US portfolios for 1962 through 2004. Each year it selected 1,000 large companies according to a non-price measure and reset the weights. Five-year averages reduced the volatility of several accounting measures. The composite portfolio produced a 12.47 percent geometric annual return in the simulation, compared with 10.35 percent for the paper's capitalization-weighted reference portfolio, a difference of 2.12 percentage points a year.

That number is historical model output, not an investor return. The period was selected before the first commercial RAFI index launched, the portfolios were reconstructed from past data, and returns were not reduced by actual trading costs. The authors separately modeled costs and reported that their assumed two percent round-trip cost reduced the average single-measure excess return from 2.15 to 2.01 percentage points. That sensitivity test shows the arithmetic under one cost assumption; it does not record prices that a live fund received.

The paper also contained contrary periods. Its composite lagged the reference portfolio by about one percentage point a year during the 1990s, when high-valued growth companies led the market. The full-period average concealed that decade-long difference.

The rebalance is the observable mechanism

The current FTSE RAFI methodology makes the process concrete. The US 1000 index ranks companies using book value, cash flow, sales, and dividends. It averages five years of data for three of those measures, combines the resulting company shares, selects the 1,000 largest fundamental values, and reviews the index annually.

Suppose a company's price rises much faster than its recorded fundamentals. Its cap weight increases immediately, but its RAFI target changes little. At the annual review, a RAFI portfolio is likely to sell some of the relative winner. If another company's price falls while its fundamental weight is stable, the portfolio is likely to buy. The method therefore contra-trades price changes rather than following them.

That mechanism can add return when relative valuations mean-revert. It can subtract return when the rising company continues to improve or when the falling price correctly anticipates permanent deterioration. The rule observes a gap between price and a slow accounting anchor; it does not diagnose why the gap exists.

Is this a better index or a value strategy?

The original paper acknowledged that fundamental weights introduce a value bias. A later paper by David Blitz and Laurens Swinkels, "Fundamental Indexation: An Active Value Strategy in Disguise," made that point the competing explanation. A stock's fundamental weight differs from its cap weight when its ratio of price to the selected fundamental differs from the market. The portfolio is therefore overweight lower-valued companies and underweight higher-valued ones relative to the cap-weighted market.

On that account, the result need not show a unique flaw in indexing. It can reflect the known value premium, the timing of value exposure, and periodic rebalancing. The choice of book value, sales, cash flow, and dividends; their averaging periods; the company universe; and the review date are active design decisions even when the resulting trades are automatic.

The two explanations can overlap. Fundamental indexing is a real, transparent way to implement diversified value exposure without ranking companies directly on a single valuation ratio. Its rebalance is mechanically different from a buy-and-hold cap-weighted index. Neither fact establishes that it is the most efficient value strategy or that the premium will appear in every period.

A live product behaved differently from the backtest

FTSE Russell dates the launch of its RAFI US 1000 index to November 28, 2005 and labels results before launch hypothetical. An investable product followed soon after. The September 2025 Invesco factsheet reported a 12.83 percent annualized net asset value return for the fund over ten years, versus 15.04 percent for the Russell 1000. From the fund's December 2005 inception, the corresponding figures were 10.04 and 10.89 percent.

Those numbers do not overturn the earlier paper. They test one fund in a different period, include a 0.34 percent expense ratio, and reflect tracking and implementation. The fund also changed to a related Research Affiliates index in March 2025, so its record does not represent one unaltered specification through the final date. It nevertheless demonstrates an essential property that a full-sample backtest can obscure: a value-tilted rebalance can lag a broad cap-weighted market for many years.

The fundamentals are records with their own lag

Book value, cash flow, sales, and dividends are not interchangeable measures of economic size. Accounting standards treat internally developed intangible assets differently from factories or acquired businesses. A bank's balance sheet and a software company's balance sheet describe different operating processes. Leverage can expand sales or assets without creating a comparable claim for shareholders.

Five-year averaging reduces turnover and the effect of one unusual year, but it also slows recognition of a genuine change. A company whose business has deteriorated can retain a large historical fundamental weight. A young company creating valuable intangible systems may have a small weight because its reported book value or distributions have not caught up.

These are not errors hidden from the method; they are the cost of using stable, observable records rather than current price. The index provider must also decide how to handle missing data, zero dividends, multiple share classes, free float, mergers, and new listings. Transparency makes those choices inspectable, not neutral.

What Arnott's contribution establishes

Arnott and his coauthors converted a critique of price-weighted portfolios into a rule that could be simulated, challenged, licensed, and implemented. That is more concrete than a general appeal to buy unpopular securities. The target weight is public, the annual trade can be observed, factor exposure can be measured, and a live vehicle can be compared with its benchmark.

The record supports a conditional mechanism: a price-independent anchor plus rebalancing creates value exposure and can harvest relative mean reversion. It does not establish that market prices are systematically wrong by a knowable amount, that accounting size is fair value, or that the original simulated premium will persist. Fundamental indexation is useful precisely because it can be tested as a quantitative proposition - including during the long intervals when its result differs from the one in the founding paper.

Inside CompanyGraph

The deep-value configuration is observable: companies priced well below book value while current assets exceed current liabilities by a wide margin and the equity ratio sits high for the industry.