Index Spotlight: Russell Style Indexes


Recently we looked at the performance disparity between two small-cap blend indexes: the S&P SmallCap 600 and the Russell 2000. We found that subtleties in index methodology can lead to different results in indexes designed to track similar market segments, especially over shorter time periods. Likewise, the style indexes from S&P and Russell rarely match up perfectly and often differ substantially, as shown in the table below.

1 mo
3 mo
1 yr
3 yr*
5 yr*
S&P 500/Barra Growth
Russell 1000 Growth
S&P 500/Barra Value
Russell 1000 Value
S&P SmallCap 600/BARRA Growth
Russell 2000 Growth
S&P SmallCap 600/BARRA Value
Russell 2000 Value

Source: Wiesenberger data as of 3/31/2002              *annualized returns

The goal of looking at differences between index provider methodologies isn't to find the "best" index, because that depends on the investor's unique individual needs and goals. For example, the index that posts the highest return for a market segment isn't the "best" index.

However, some major differences between Russell and S&P/Barra style indexes include:

  • Number of stocks. Russell has larger universe for both its large- and small-cap style indexes.
  • Style methodology. Russell uses 2 valuation measures: p/b and forward looking IBES forecasted earnings. S&P/Barra uses one, p/b.
  • Style weighting methodology. Russell uses non-linear weighting (more on this below) where some stocks can be in both the Russell growth and value indexes while S&P/Barra uses a "line in the sand" technique where stocks are pigeonholed into either the S&P/Barra growth or value index.
  • Russell indexes incorporate float adjustments in stock weights for unavailable capital like cross-ownership while S&P/Barra does not.
  • Russell indexes undergo comprehensive, objective annual reconstitution to reflect changing market reality. S&P continuously updates its samples of stocks using a committee framework to select representative stocks in various market segments.

All of these differences matter in terms of returns and portfolio characteristics at different times and in different market conditions. For more information on the S&P/Barra style indexes, click here to go to the Barra website. For more information on the Russell index methodology, click here.

We spoke with Jon Christopherson, a Research Fellow in Russell's Investment Policy and Research Group. He told us about some of the things that make the Russell indexes unique.

IF: In the Russell methodology, a stock can be in both the growth and value indices. Do you think the Russell indexes capture a more realistic way of looking at how investment decisions are made by active managers? For example, active managers don't move in a herd-like fashion when a stock crosses a fixed capitalization or valuation line.

JC: That is what inspired the development of the non-linear, or membership in both indexes, methodology. And it makes sense. As a stock moves from high price-to-book (P/B) and strong forecasted growth down toward market average levels and then down to below market levels, growth managers unload the stocks and value managers start thinking about buying them. We do not "count" how many characteristics we use because the number of characteristics used does not necessarily produce a better index. Russell used a combination of a forward looking metric in the IBES long term growth forecast, along with an historical metric, such as book/price adjusted for certain accounting write-offs, to come up with our style methodology. In developing the indexes many style models were tested along with many variables, but the non-linear probability model best approximated the behavior of style investment managers and the "ponds" in which they fished. We found that many stocks had a probability of being held by both growth and value manager portfolios and those stocks tended to be around the market medians. The Russell index style methodology attempts to reflect these more comprehensive opportunity sets and minimizes turnover in the process.

IF: Can you explain how the non-linear weighting works?


JC: The graph is essentially two crossing J-curves with the asymptotes at zero probability and 1.0 probability. They cross at 0.5. The original curve was based on cumulative frequency ogives of the amount of funds value, and then growth, managers invested in stocks with various levels of Price/Book. Since then, we have fitted a propriety function that we use. The one in the graph is based on a Logistics Curve function, but any non-linear function over the first to third quartiles of P/B would work and give the basic idea. (ogive - graph of a distribution function or a cumulative frequency distribution)

IF: Make the case for why Russell style indices are the best yardstick for measuring active managers.

JC: Intuitively the Russell style indices make more sense both in terms of the non-linear methodology and in the cross-ownership adjustment for weights. Even MSCI in a back handed compliment to Russell, is moving toward cross-ownership adjustments. Russell pioneered float adjustment back in 1984 when we created the Russell 3000. They also are moving from 65% capitalization coverage to 90%. Ours have always had a target of 98%. Why? Because that is the range of stocks that managers and mutual funds buy - it is their habitat. The reason Russell created the Russell 3000 and did not adopt the Wilshire 5000 was because managers did not often buy stocks below the 3,000th stock in a list ranked 1 to 7,000 in capitalization. I have heard that the Vanguard Total Stock Market Fund starts with the Russell 3000 and goes from there. Why? Because 2% of the market capitalization spread over 2,000 securities is a portfolio management headache no one wants to deal with. I doubt you will find many index funds, if any, that truly matches the Wilshire 5000 for this reason.

IF: The Russell indices rebalance once a year. Is this placing undue pressure on index fund managers?

JC: This subject has been the source of endless debate around Russell since we made the decision to do annual reconstitution. The annual decision was made to accommodate clients who wanted to build index funds but were frustrated by the high transaction costs of quarterly or semi-annual rebalancing. My colleague Mahesh Pritamani feels that there are enough providers of liquidity in the market around reconstitution, hedge funds for example, that there should not be undue pressure on index fund managers and we do not hear passive fund managers saying that they want us to move to semi-annual or quarterly reconstitution.

I did a study that showed transaction cost go down as reconstitution frequency is lengthened. The main problems with annual reconstitution is that the indexes get stale so that many of the growth stocks just before reconstitution have not been growthy as far as the market is concerned for quite a while. However, Russell's Pritimani, Gardner and Kondra did a study on reconstitution frequency, the return differences generated, and transaction costs. They found that in terms of index returns, for most periods during the years 1982 through 2000, the returns of the simulated indexes are very similar for annual, semi-annual and quarterly reconstitution. These results suggest that an index with annual reconstitution gives a representation of the market that is very similar to that of indexes with more frequent reconstitution in "typical" market environments. However, this similarity breaks down in volatile market periods such as 1999 and 2000. The simulated indexes differ substantially during these years. This is especially true of the style indexes.

When turnover is measured either by the number of name changes or the amount of an index portfolio that must be traded, the simulations suggest that increasing the frequency of reconstitution substantially increases the turnover of all the indexes. Quarterly recon roughly doubles the transaction costs. This result holds in all market environments.

Russell believes that these results support the continued use of annual reconstitution. A move to more frequent reconstitution would make little difference in index returns during most periods. Consequently, such a move would not meaningfully improve the indexes in terms of accurately reflecting the market. However, for most of the indexes, more frequent reconstitution would likely increase turnover and significantly raise the cost of managing an index fund based on the indexes. You can see the entire article on Russell's website: http://www.russell.com/US/Indexes/US/membership/Recon_Frequency.asp.

A final point made by Lori Richards, Manager of Russell Indexes, is that the lack of complete reconstitution in market indexes can result in capitalization, sector, and style biases creeping into index returns. The universe of small, mid, and large cap stocks are different between the Russell and S&P indexes to begin with given disparate construction rules. The core indexes are different because the Russell family is an objective, capitalization-ranked, broad, float-adjusted, completely recalibrated market reflection. The S&P family is a subjective or committee-chosen, significantly reduced sampling of companies from the universe. We should expect the core indexes to perform disparately, and especially in the small cap market where float adjustment and reconstitution have significant impact on weight and membership. We can expect the resulting style index performance to be more dramatically different given the combination of influences including the use of different core index methodology, different variables to describe style, and different means of reflecting style opportunity sets.

IF: Index providers are always talking about "style purity." How do you possibly measure that?

JC: The non-linear methodology I created is a move toward addressing this issue. However, "purity" like beauty is somewhat in the eye of the beholder. Just about any growth or value manager will say that our style indexes do not do a "good enough" job of measuring growth and value but at the same time they will say that ours are the best that are available given what they do. The problem lies in what exactly is growth and value? We thought the use of Price/Book and Forecasted Earning metrics were particularly valuable, and we disregarded a few other seemingly obvious ones for different reasons. Any list of variables the critics come up with lead to problems with implementation. For example, dividend yield as a measure of value fails as a good variable because deep value stocks in a turn-around situation that have suspended paying dividends are no different than zero dividend stocks such as Microsoft. Another example: 5-year earning growth as a measure of growth fails because it requires 5 years of history, which means that new IPO's cannot be evaluated in your index for 5 years. This discriminates against small cap stocks, some of which are the growthiest of growth stocks.

Finally, "purity" implies a scale from most pure to least pure, which brings us right back to the problems of what variables (and their values) are more or less growthy than others. The degree to which a variable measures growth or value is confounded with its metric. For example, is a stock with minus one sigma value on price/sales more or less valuey than a stock with a minus one sigma value on price/book or price/free cash flow? The answer certainly does not jump out at you. We at Russell have made choices. We believe they are reasonable choices given behavior in the marketplace and the needs of users and our customers, but the choices are certainly open to criticism, which we welcome. Russell is always looking for ways to improve our indexes and our other products. The indexes are not locked in concrete.