Reconstitution And You Introduction This Monday marked the first day of trading following the 24 th reconstitution of the Russell Growth and Value Style Indexes, since their launch in 1987. For years the Russell Investments family of equity indices has allowed investors to track the performance of distinct market segments. By creating indices that are objectively constructed based upon transparent rules, Russell attempts to produce investible and replicable indices that are comprehensive representations of the market segments they are meant to embody. 1 While this is largely achieved, Westfield Capital has long held that Russell s process has fallen somewhat short in capturing what a growth-at-a-reasonable price (GARP) manager is attempting to buy: companies that have earnings growth not yet identified by consensus and that, as such, are trading at a discounted multiple to their peer group and/or their future earnings growth rate. It is our view that Russell adjusted their methodology with this year s reconstitution, in part, to account for this dynamic. The following is our attempt to shed some light on the recent enhancements. Methodology In an effort to ensure their indices remain accurate characterizations of the market segments they are designed to represent, Russell conducts an annual reconstitution every June, during which the indices are rebuilt to reflect changes in the fundamentals of their constituents. 2 Russell had historically used a two-variable construction methodology to distinguish between the growth and value style segments of the market in their index creation. This two-variable methodology utilized book-to-price (B/P) as the Value determinant and the Institutional Brokers Estimate System (I/B/E/S) long-term-growth (LTG) forecasts as the Growth determinant. 3 These two variables were then run through Russell s proprietary Composite Value Score (CVS) algorithm, with the resulting outputs determining the constituents of each style index. Russell had long claimed that this twovariable methodology resulted in the purest and most accurate representation of the style indices; however, this was changed for the 2011 reconstitution. As the Russell Indices have grown in popularity, they have become an increasingly powerful force in both the active and passive marketplaces. These indices serve not only as a performance standard by which active managers can be measured, but also as a proxy in the formation of asset allocations. Despite this, investors often do not take the time to understand the way the Russell Indices are constructed. While Russell s unbiased, rules-based construction methodology is generally successful at achieving its market proxy goal, we are still left with some individual securities that seem misplaced. 4 Westfield Capital contends that while the methodology is sound, Russell may classify certain securities as Growth simply because they are expensive, regardless of their actual growth prospects. Conversely, often securities might be classified as Value simply given
their high B/P, regardless of how truly robust their earnings prospects may be. Further clouding the matter, some stocks are constituents of both the Growth and Value indices. Examples Looking at historical changes within the Russell 1000 Growth and Value Indexes, and Russell 1000 Core Weight 100% 80% 60% 40% 20% 0% Exxon 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 XOM - Russell 1000 Growth Index XOM - Russell 1000 Value Index Source: FactSet isolating Exxon Mobil (XOM) and General Electric (GE), one can see some interesting implications. Over the past few years alone, we have seen a dramatic shift in how XOM has been classified: in June 2008, 17.5% of XOM s weight was in the Russell 1000 Growth Index ( Growth ) and 82.5% in the Russell 1000 Value Index ( Value ). That shifted to 92.0% in Growth and 8.0% in Value by June 2010. Digging into the numbers, we see that in 2008 XOM achieved historical earnings growth (3yr) of +22% and had a forward earnings growth forecast (3-5yr) of 12%, yet in June 2009 it still maintained 75.0% of its weight in Value. Meanwhile as of June 2011, XOM had projected earnings growth of 6% and actually displayed negative historical earnings growth of -11%, yet it still ended up with 74.0% of its weight in Growth and only 26.0% in Value following the reconstitution; clearly representative of potential distortions. Examining GE, we see its weight follow an almost completely opposite path than that of XOM. In 1999, 2001, and 2002 GE held 100.0% of its weight in the Russell 1000 Growth Index and zero percent in the Russell 1000 Core Weight 100% 80% 60% 40% 20% 0% General Electric 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 GE - Russell 1000 Growth Index GE - Russell 1000 Value Index Souce: Fact Set Russell 1000 Value Index. From 2003 to 2005, GE essentially bounced back and forth; in June 2003, 100.0% of its weighting was in Growth, which then swung to 88.0% in Value in June 2004, only to return to 76.0% in Growth by June 2005. Surprisingly, GE ended up with zero representation in the Russell 1000 Growth Index after this year s reconstitution, and yet its projected earnings growth of +12% is double that of Exxon s! Analyzing both companies B/P as of June 2011, perhaps GE s B/P of 0.6, compared to that of XOM s at 0.4, is the main driver of its positioning within the Russell 1000 Value Index. 2
June-03 June-04 June-05 June-07 Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr XOM 0.35 8.5 9.7 0.32 8.7 7.0 0.29 8.7 27.2 0.25 6.8 29.8 GE 0.25 10.5 11.9 0.29 8.7 6.9 0.31 11.8 3.9 0.30 10.2 7.4 June-08 June-09 June-10 June-11 Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Book/ Est. EPS Hist EPS Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr Price Growth - LT Growth 3yr XOM 0.27 15.5 22.5 0.32 7.0 14.5 0.48 7.7-12.6 0.37 6.0-11.8 GE 0.45 10.6 13.1 0.90 8.8 5.5 0.74 10.0-19.6 0.58 12.5-22.1 Source: FactSet Methodology Changes As mentioned, Russell enhanced their methodology for the 2011 reconstitution. The new construction methodology saw the two-variable approach retired and resulted in a material change to the Growth variables being applied. Separately, the Value metric remains unchanged and continues to be book-to-price, with an increasing B/P ratio resulting in a higher probability of a larger weighting in the Value benchmark. The former growth variable of I/B/E/S LTG forecasts is now replaced by a two-variable combination: 1) Medium-term-growth (MTG) rate forecast, using I/B/E/S two-year earnings forecasts. 2) A sales-growth variable, using reported five-year historical sales growth per share. 5 The MTG earnings forecast is converted into a growth ratio by dividing a 2-year forward EPS estimate (FY2) by the last reported annual earnings (FY0). The historical sales growth factor is slightly more complex, with Russell calculating an average of annual revenue growth rates over the past 5 years. Also of note, while the previous EPS growth determinant used the mean of analysts forecasts, the new MTG forecasts uses the median of analysts estimates, which is expected to be less sensitive to outliers. 6 The Growth and Value variables are equally weighted in the construction methodology, with the Value variable carrying a 50% weighting and the two Growth variables accounting for 25% each. At this point, it would be fair to ask what sparked the need to change a methodology that had been generally viewed as accurate and acceptable for years. While there are a variety of reasons that could be debated, we will focus on the one provided by Russell. To determine the LTG earnings input, Russell relied upon the voluntarily contributed earnings forecasts of Wall Street analysts. For many, the five-year earnings forecasts have often been considered a loose approximation and the accuracy and diligence of the estimate has been questioned; it is, after all, predicting five years into the future. As a result, the number of LTG forecasts provided by the Street has been in steady decline. As of May 31, the average number of analysts contributing LTG forecasts has dropped from 5.8 in 2006 to 3.6 in 2010. Meanwhile, the average number of analysts contributing a MTG forecast has increased from 12.2 in 2006 to 14.4 in 2010. 7 It is possible that recent market turmoil has made these longer-term forecasts more difficult to project and as a result, fewer analysts attempt them. Perhaps the reduced number of outliers in the MTG forecasts is also evidence of the less liberal approach to these estimates. Regardless of the 3
explicit reason for the decline, it afforded Russell the opportunity to re-examine their style methodology. Russell s goal in analyzing their style methodology was to settle on a suitable combination of variables that would produce indexes that best represented style portfolios held by investment managers. 8 Russell then examined a host of different variable combinations to try to produce the most accurate representation. Their analysis suggested that the incumbent two-variable approach resulted in the most accurate representation of style segments, which could make one question the need to make a change. However, conceding that the reduced availability of LTG forecasts necessitated a variable change, Russell s analysis then confirmed that maintaining the B/P variable and swapping out the LTG forecast for the MTG forecast variable resulted in the next best representation. 9 Here again, a fair question arises: if the new two-variable methodology tested by Russell produced the 2 nd best result, then why even bother introducing a 3 rd variable? Russell s reasoning is that forecasts over shorter time horizons are understandably more volatile, thus switching to MTG forecasts resulted in greater volatility. 10 In contrast to forecasts, historical growth measures are typically smoothed by construction. Therefore, adding a historical growth variable, such as sales growth per share, serves as a smoothing component and reduces turnover. This could call into question whether or not the historical sales growth variable carries any predictive qualities or if it just a stabilizer. To a growth manager, this would appear to be the essence of investing: be ahead of consensus in forecasting potential earnings growth, then move on from the position once that potential has been achieved and before it has decelerated. The premium valuations often ascribed to growth stocks are based on expectations of faster-than market earnings growth, with the avoidance of negative surprises being highly correlated with success in growth investing. It is our view that a historical sales growth element includes a lookback component, almost double-counting the credit for sales that have already occurred and that should theoretically already be priced into the stocks. Additionally, many companies that are trading at discounted valuations can exhibit earnings growth that is not yet fully understood by investors; we believe our focus on earnings growth allows us to objectively source companies that have high appreciation potential across the broad investment horizon. It is for these reasons that, while we acknowledge Russell s methodology as being solid and achieving its purpose, we believe it still leaves open the opportunity for talented active managers to outperform over a complete market cycle. In an effort to reduce turnover resulting from the methodology enhancements, Russell implemented a CVS-banding method, which should help dampen some of the swings we have seen. In order for a newly calculated CVS to be adopted, the style banding requires a stock s new CVS to differ from its current CVS by at least +/- 0.15. 11 Thus, the new CVS must breach this implemented band to force a stock out of one style index and into another or even to increase/decrease a constituent s weight in a style index. The banding should eliminate any small changes and reduce turnover, which at times has been difficult to respond to for index-driven management styles. A few years ago, Russell introduced banding for just their size indexes and it substantially lowered overall turnover; we would expect a similar impact for the style indices and applaud this enhancement. 4
Recent Reconstitution With the methodology changes having been discussed, it is now important to review the changes to the indices that resulted from the reconstitution on June 24 th. There are material changes across all four Russell Growth Indexes and in a variety of economic sectors: Russell 1000 Growth New Pre-recon % Change Russell Midcap Growth New Pre-recon % Change Consumer Staples 12.0 10.1 1.9 Energy 9.8 5.7 4.1 Health Care 11.1 10.1 1.0 Materials 9.0 7.1 1.9 Materials 5.9 5.0 0.9 Consumer Staples 5.8 5.7 0.1 Telecommunication 1.2 0.8 0.4 Utilities 0.3 0.3 0.0 Industrials 13.4 13.4 0.0 Health Care 13.7 13.8-0.1 Utilities 0.1 0.1 0.0 Telecommunication 1.6 1.8-0.2 Energy 11.3 11.4-0.1 Financials 6.7 7.1-0.4 Consumer Discretionary 14.0 14.5-0.5 Industrials 14.7 15.5-0.8 Financials 4.0 4.8-0.8 Consumer Discretionary 19.2 20.5-1.3 Information Technology 27.0 29.8-2.8 Information Technology 19.3 22.6-3.3 Russell 2500 Growth New Pre-recon % Change Russell 2000 Growth New Pre-recon % Change Energy 8.3 5.1 3.2 Energy 8.3 5.2 3.1 Materials 7.9 6.3 1.6 Financials 7.4 5.0 2.4 Financials 8.0 6.8 1.2 Consumer Staples 3.9 3.2 0.7 Utilities 0.4 0.3 0.1 Telecommunication 1.2 1.1 0.1 Health Care 16.0 16.0 0.0 Utilities 0.1 0.1 0.0 Telecommunication 1.5 1.6-0.1 Materials 4.6 4.9-0.3 Information Technology 23.9 24.0-0.1 Health Care 19.5 19.9-0.4 Consumer Staples 3.4 3.6-0.2 Industrials 15.5 16.2-0.7 Industrials 16.2 17.3-1.1 Consumer Discretionary 14.8 16.6-1.8 Consumer Discretionary 14.4 18.9-4.5 Information Technology 24.7 27.7-3.0 Source: FactSet The Energy sector saw the largest increase in weight across the Russell Growth Indexes, with an average boost of 2.6%, and the largest overall increase in each index except for the Russell 1000 Growth Index. Speaking from the perspective of a growth manager that has seen real growth in the Energy sector for years, Westfield Capital views this weight expansion, in some ways, as an affirmation that the earnings growth we have forecasted is coming to fruition. The Materials sector saw the 2 nd largest increase across the Growth Indexes, seeing an average increase of 1.0%, even though its weight in the Russell 2000 Growth Index actually decreased. The Financials sector s changes are even more varied, with increases of 2.4% and 1.2% in the Russell 2000 Growth Index and Russell 2500 Growth Index respectively, while the sector sees its weight decrease by an average of 0.6% in the Russell 1000 Growth and Russell Midcap Growth Indexes. The Information Technology sector saw the largest decrease in weight across the Russell Growth Indexes, with an average decline of 2.3%, and the largest overall reduction in each index except for the Russell 2500 Growth Index. The Consumer Discretionary sector experienced the 2 nd largest decrease in weight across the four Russell Growth indexes, resulting in an average drop-off of 2.0%. It also saw the largest change of any sector, seeing a decrease of 4.5% within the Russell 2500 Growth Index alone. The annual Russell reconstitution is not only a focus of active and passive investment managers, but also for trading desks. In what has become known as the Russell Trade, desks have become increasingly interested in monitoring and projecting potential 5
additions and deletions to the Russell core and style indexes, in an attempt to profit on the rebalance trade. At one point heading into the reconstitution date, a basket of predicted small cap additions had outperformed the market (the Russell 2000) by as much as 450 basis points, according to one sell-side firm. This outperformance, which was greater than last years, suggests an increased popularity and evidence of more pre-trading taking place. Summary As an active domestic growth equity manager, with a research process driven by fundamental, bottom-up analysis, the Russell Reconstitution and any changes to its methodology have no impact on the way Westfield Capital manages money on behalf of our clients. While Westfield is aware of the index weights, they do not drive investment decisions in our strategies. That being said, we are keenly interested in any and all changes to our clients benchmarks as we continue to present our portfolios relative to current benchmark weights, and appreciate that other investors, as well as clients, may use these metrics to gauge relative views on market segments. In conclusion, we hope this paper can serve as a resource to facilitate a deeper understanding of the construction methodology Russell uses in developing benchmarks that have become the gold standard of market segment proxies. Westfield Capital commends Russell Investments on the job they continue to do in creating investible indices, just as we commend those managers who can successfully outperform them. I hope everyone enjoys their 4 th of July weekend! Regards, Kevin D. Robertson Marketing and Client Service Manager Westfield Capital Management Company, LP One Financial Center, 24th Floor Boston, MA 02111 617.428.7135 direct 617.428.7190 fax krobertson@wcmgmt.com 1 Russell Investments. Russell U.S. Equity Indexes Construction and Methodology. www.russell.com/indexes. March 2011. p. 2. 2 Ibid., 2. 3 Ibid., 12. 4 Ibid., I. 5 Russell Investments. Russell U.S. Equity Indexes Construction and Methodology. www.russell.com/indexes. March 2011. p. 3. 6 Ibid., 6. 7 Ibid., 2. 8 Ibid., 4. 9 Ibid., 8. 10 Ibid., 10. 11 Ibid., 13. 6
BIBLIOGRAPHY Russell Investments. Russell U.S. Style Index Methodology. www.russell.com. March 2011. Russell Investments. Russell U.S. Equity Indexes Construction and Methodology. www.russell.com/indexes. March 2011. All data was provided by FactSet Research Systems Inc. ( FactSet ), a non-affiliated data provider. 7