FTSE ActiveBeta Index Series: A New Approach to Equity Investing

Transcription

1 FTSE ActiveBeta Index Series: A New Approach to Equity Investing 2010: No 1 March 2010 Khalid Ghayur, CEO, Westpeak Global Advisors Patent Pending Abstract The ActiveBeta Framework asserts that a significant portion of active equity management returns (or alpha) is attributable to systematic sources of active returns, rather than managers stock selection skill. These systematic sources of active equity returns are shown to arise from the systematic behavior of short-term and long-term earnings expectations and discount rates. These systematic sources are closely linked to momentum and value and can be effectively captured through such strategies. The FTSE ActiveBeta Indices provide an efficient, transparent and cost-effective capture of these systematic sources of active equity returns. The ActiveBeta Framework, and the associated FTSE ActiveBeta Indices, provides a new approach to investing in equities. The proposed approach has important implications within the equity investment process, including: Structuring equity portfolios: Asset owners can create more efficient overall equity portfolios by directly allocating to Active Betas in addition to the traditional market beta and alpha allocations. A cost-effective exposure to Active Betas can be obtained easily through passive replication of FTSE ActiveBeta Indices. Benchmarking investment performance: Momentum and value better define the investment styles of active managers compared to growth and value. A combined capture of momentum and value better defines the investment style of core managers compared to the market. As such, FTSE ActiveBeta Momentum, Value, and combined Momentum and Value Indices represent more appropriate performance benchmarks for active growth, value, and core managers, respectively. Creating investment vehicles: FTSE ActiveBeta Indices provide a passive, efficient, and high-capacity alternative to traditional active management strategies. A New Approach to Equity Investing Page 1

2

3 A New Approach to Equity Investing Page 3 In this white paper, we present the case for the existence of systematic sources of active equity returns, which we refer to as Active Betas. The white paper is structured in two main sections. Section 1 is designed for the busy professional. It provides a broad overview of the ActiveBeta Framework, discusses the design and historical performance of the FTSE ActiveBeta Indices, and concludes with a discussion of the various applications of the new indices. Section 2 presents the research that supports the proposed ActiveBeta Framework. It discusses the statistical analysis documenting the systematic behavior of earnings growth and discount rates and establishes their links with price momentum and value strategies. (For a more detailed discussion of the ActiveBeta Approach and the underlying research, please refer to the full-length book titled ActiveBeta Indexes: Capturing Systematic Sources of Active Equity Returns, published by John Wiley & Sons.)

4 1. Overview 1.1 ActiveBeta Framework: Basic Principles The ActiveBeta Approach is based on the following investment principles and findings. 1. Price changes are driven by changes in earnings growth and discount rates. 2. Systematic sources of returns (price changes) arise from the systematic behavior of earnings growth and discount rates over time. We refer to these systematic sources of active returns as Active Betas. 3. A fundamental influence drives the behavior of earnings growth (and discount rates) in open, competitive free-market economies; that is, a large group of stocks in a given universe cannot sustain above-average earnings growth indefinitely. 4. This fundamental influence causes predictable patterns in the behavior of earnings growth over time. In particular, shortterm earnings growth trends in the short run (up to one year) and long-term earnings growth mean reverts in the long run (within three to five years). 5. Momentum strategies are closely linked to short-term earnings growth, and their returns are driven by the trending behavior of short-term earnings. Value strategies are closely linked to long-term earnings growth, and their returns are driven by the mean-reverting behavior of long-term earnings. 6. Momentum and value, therefore, represent systematic sources of active returns and should be viewed as additional forms of betas, or Active Betas. 7. Momentum and value independently provide positive active returns over time, but these active returns are also negatively correlated. As such, a combined capture of momentum and value should provide superior risk-adjusted performance compared to the capture of momentum or value alone. These principles and research findings provided the motivation for the creation of ActiveBeta Indices. 1.2 FTSE ActiveBeta Indices Rationale for a Momentum Index While value indices have been available for some time, no internally-consistent family of global market capitalizationweighted momentum indices and combinations of momentum and value indices were previously available in the marketplace. The absence of momentum indices is due partly to the lack of consensus on the rationale for explaining momentum active returns. The academic literature provides some explanations for the momentum effect (e.g., that momentum is a technical indicator or a measure that captures informational and/or behavioral inefficiencies). However, none appear to have gained wide scale acceptance in the investment community. The ActiveBeta Approach provides a simple, but robust, explanation for the nature of momentum and value returns. That is, these returns are driven by the systematic behavior of short-term and long-term earnings growth, respectively. We argue that broadly-diversified momentum strategies are an efficient way of capturing the tendency of short-term earnings to trend in the short run, and that investors should have a passive exposure to this systematic source of return. Momentum strategies are characterized by high frequency of rebalancing and high turnover, features considered undesirable from the perspective of passive, low-turnover, low-cost indices. However, this argument against creating momentum indices seems unreasonable in view of the benefits of this approach. As such, we have developed an index construction methodology to allow for an efficient capture of momentum strategies by limiting the turnover to a reasonable level through the use of rebalancing buffers. (Please refer to Methodology for the Management of the FTSE ActiveBeta Index Series for more details.) A New Approach to Equity Investing Page 4

5 Index Construction Process For a given universe within the FTSE All World Index Series, we use the following 3-step index construction process to create independent FTSE ActiveBeta Momentum and FTSE ActiveBeta Value Indices, as well as the combined FTSE ActiveBeta Momentum and Value Index (MVI). Figure 1. Index Construction Process Historical Performance of FTSE ActiveBeta Indices The active return, active risk, active return maximum drawdown, and other key characteristics of the FTSE ActiveBeta Indices relative to the underlying universe benchmark for various individual markets and market composites are shown in Table 1. (This table reports before-cost return statistics, which is customary for analyzing index performance. To review after-cost index performance statistics, please contact Westpeak Global Advisors.) Source: Westpeak. Stock Universe Momentum Ranking FTSE ActiveBeta Momentum Index FTSE ActiveBeta MVI Stock Universe Value Ranking FTSE ActiveBeta Value Index Active Returns In all countries and regions, the FTSE ActiveBeta Momentum and FTSE ActiveBeta Value Indices deliver positive active returns relative to the underlying market index, except for momentum in Japan (please see Section 2 for an explanation of momentum returns in Japan). In all instances, the FTSE ActiveBeta MVI also produces positive active returns. The active returns for the FTSE ActiveBeta MVIs range from 141 basis points per annum for the FTSE 100 universe to 192 basis points per annum for the FTSE Eurobloc universe. The performance statistics reported in Table 1 support the existence of the systematic sources of active equity returns, momentum and value, and the validity of the ActiveBeta Framework at a global level. Step 1: Rank the selection universe (a given FTSE market index, such as FTSE All World USA Index), from high to low, on momentum and value signals independently. Momentum is specified as past 12-month total return. Value is specified as the average of three valuation ratios: price-to-book value, price-to-sales, and priceto-cash flow (or price-to-earnings, where appropriate). Step 2: Create independent FTSE ActiveBeta Momentum and Value Indices targeting roughly 50% market cap coverage of the ranked selection universe for inclusion in each index, and market cap weighting the selected constituents in each index. Step 3: Combine the independent FTSE ActiveBeta Momentum and Value Indices, in a weighting scheme, to create the FTSE ActiveBeta MVI. Active Risk and Information Ratios Active risk represents the tracking error of the FTSE ActiveBeta Indices relative to the universe benchmark. The FTSE ActiveBeta Momentum and Value Indices generate relatively high active risk readings. In general, for single countries, the active risk of the FTSE ActiveBeta Momentum Indices is over 6%, while the active risk of the FTSE ActiveBeta Value Indices is around 5%. The active risk levels for the momentum and value indices are reduced when single countries are combined to form regional composites. In every market and region, the FTSE ActiveBeta MVI has significantly lower active risk, with most below 3%, than the component momentum and value series. The complementary nature (negative correlation) of momentum and value active returns, which range from to -0.50, drives this reduction in active risk. This diversification greatly improves the efficiency of the FTSE ActiveBeta MVIs, generally allowing them to yield information ratios superior to the momentum or value indices. The information ratios for the FTSE ActiveBeta MVIs range from 0.41 for Japan to 0.63 for Developed ex US. A New Approach to Equity Investing Page 5

6 Table 1. FTSE ActiveBeta Index Performance Summary ( ) Universe Source: FTSE and Westpeak. Annualized Active Return (%) Active Risk (%) Information Ratio Active Return Drawdown Max. (log%) Names Held From Universe (%) Annual Turnover (%) Value/Momentum Active Return Correlation FTSE US (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE 100 (GBP) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE 350 (GBP) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE All-Share (GBP) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Japan (JPY) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed Europe ex UK (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed Europe (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Eurobloc (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed ex US (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed ex North America (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed ex Europe (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed ex UK (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Developed ex Japan (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE Emerging (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI FTSE All-World (USD) ActiveBeta Value Index ActiveBeta Momentum Index ActiveBeta MVI A New Approach to Equity Investing Page 6

7 Active Return Drawdown The advantage of combining the capture of momentum and value is more apparent when considering the concept of active return drawdown, which indicates the greatest magnitude by which the return of a portfolio trailed its benchmark during a distinct period of time. Looking at the individual FTSE ActiveBeta Momentum and Value Indices, one can see that maximum relative underperformance of about 40% is possible, and over 20% is common. The combined FTSE ActiveBeta MVI, however, rarely produces a drawdown in excess of 10% in any country or region during the analysis time period. The diversification benefits are even more significant in market composites. For instance, the FTSE All- World ActiveBeta MVI had an active return drawdown (6.37%) that was a fraction of the drawdown registered for value (17.29%) and momentum (24.39%). By mitigating underperformance relative to the core benchmark, the FTSE ActiveBeta MVIs achieve positive active returns with the consistency required by investors interested in alternatives to current style index and active management offerings. Market Exposure In addition to favorable return and risk characteristics, the FTSE ActiveBeta Indices also provide extensive market exposure, which may be a critical feature for many investors. The FTSE ActiveBeta Momentum and Value Indices target about 50% market coverage and tend to hold between 40% and 60% of the names in the underlying selection universe. Given the negative correlation of momentum and value, the FTSE ActiveBeta MVIs generally hold around 80% of the selection universe constituents. The FTSE ActiveBeta MVIs also provide between 75% and 85% market cap coverage of the selection universe. By investing in the FTSE ActiveBeta MVI for any market, an investor can be assured that the index will cover the vast majority of the target universe. Turnover The FTSE ActiveBeta Momentum and Value Indices have differing turnover characteristics. The FTSE ActiveBeta Value Indices typically turns over by 35% to 65% annualized, while the FTSE ActiveBeta Momentum Indices typically turns over by 85% to 105%. By combining these two indices into the FTSE ActiveBeta MVI, we generally hold turnover around 60% reasonable turnover for an index rebalanced on a monthly basis. The turnover numbers are kept within a reasonable range by the use of market cap-based buffers to rebalance the indices. Summary of Historical Performance The analysis presented in this section provides the following evidence: FTSE ActiveBeta Momentum and Value Indices independently outperform the universe benchmark in all markets, except Japan for momentum. FTSE ActiveBeta Momentum and Value Indices are negatively correlated in all markets and regions, thus providing significant diversification benefits when captured in a combined fashion through the FTSE ActiveBeta MVI. FTSE ActiveBeta MVI reliably lowers the risk and drawdown compared to the component momentum and value indices. FTSE ActiveBeta MVI outperforms its selection benchmark in all universes and generates a highly-respectable information ratio, especially from the perspective of a passive investment. FTSE ActiveBeta MVI outperforms the market benchmark nearly 60% of the time, on a monthly basis, in all universes. The preceding discussion highlights the universal validity and applicability of the ActiveBeta Framework. The risk and return characteristics of the FTSE ActiveBeta Indices provide a transparent, cost-effective, and passive investment alternative to institutional investors looking for ways to increase the expected return of their asset portfolios. A New Approach to Equity Investing Page 7

8 1.3 FTSE ActiveBeta Index Applications The existence of systematic sources of active equity returns, and their efficient capture through the FTSE ActiveBeta Indices, has many applications for style investing, performance attribution, performance benchmarking, portfolio structuring and asset allocation, and creation of investment vehicles within the equity asset class. Style Investing A New Framework Current Perspective The origins of style investing can be traced to the academic research conducted by Fama and French in 1992 to explain the cross-section of average stock returns. Their findings suggested that size, value, and market beta are the three factors that explain average stock returns. The Fama/French research, in turn, led to the categorization of active managers and market indices along the two dimensions of size (large, mid, and small) and style (value and growth). The categorization of value and growth as investment styles has emerged out of empirical research, as mentioned above. There appears to be no commonly-accepted conceptual or theoretical framework for explaining why value and growth represent the two investment styles. In particular, the choice of growth, as currently defined, as an investment style is extremely puzzling, as growth consistently provides negative active returns in the long run. Even for value, a well-documented source of positive active returns, there is little agreement on what drives its returns. Some researchers, such as Fama and French, argue that value represents a rational compensation for higher risk, while others, such as Lakonishok, Shleifer, and Vishny, argue that value returns arise from the behavioral biases of investors. ActiveBeta Perspective A style, by definition, represents a source of return that is common across a large number of active managers. As such, investment styles should represent systematic sources of active equity returns, especially if they persist over time. Value is clearly an investment style, as it captures the systematic source of active returns associated with the mean reversion of long-term earnings growth expectation and anticipated risk, as we have already discussed. But, which positive systematic source of active equity return does growth capture? Is Growth an Investment Style? The current definition of growth and construction of growth indices leads investors to buy high long-term actual and/or expected earnings growth companies. However, the systematic tendency of long-term earnings growth expectation to mean revert implies that a positive payoff to growth comes from buying current low-growth companies, not high-growth companies. If high long-term growth companies are bought, and long-term growth systematically mean reverts, as we find to be the case, then negative payoffs will be realized. Published historical performance of various value and growth indices for different geographical regions is shown in Table 2. Without exception, growth benchmarks provided negative active returns compared to the core index, while value benchmarks provided positive active returns. For example, in the case of the FTSE Developed universe, the FTSE Developed Value Index outperformed the underlying core index by 1.38% per annum from 1995 through Over the same period, the FTSE Developed Growth Index underperformed by 1.76% per annum. Furthermore, given the sign and magnitude of active returns and information ratios generated by the value and growth indices across all universes, it is clear that growth is simply the mirror image of value. A New Approach to Equity Investing Page 8

9 Table 2. Historical Returns of Published Core, Value, and Growth Indices ( ) Index Annualized Active Return (%) (vs. Parent Index) Active Risk (%) Information Ratio Total Return (%) Total Risk (%) Sharpe Ratio Russell Russell 1000 Value Russell 1000 Growth FTSE Japan FTSE Japan Value FTSE Japan Growth FTSE Developed Europe FTSE Developed Europe Value FTSE Developed Europe Growth MSCI EAFE MSCI EAFE Value MSCI EAFE Growth FTSE Developed FTSE Developed Value FTSE Developed Growth Notes: FTSE Developed Europe Index uses data from FTSE World Europe Index prior to 3 January All returns in USD terms. Source: Westpeak, based on data from FTSE, MSCI Barra, and Russell. Given this evidence, how can growth investing, as currently defined by index providers, be an investment style that a large number of managers willingly follow? Growth Style Redefined as Momentum Growth benchmarks do not accurately specify and capture the investment process of so-called growth managers. These managers typically develop strategies to predict short-term changes in earnings expectation, not the level of long-term earnings expectation. In other words, growth managers follow momentum strategies. Table 3 shows the correlation of US large cap growth managers active returns against the active returns of a long-term expected earnings growth index and a price momentum index. Growth manager active returns are obtained from the PSN manager database and correspond to a US large cap growth universe. The IBES 5-year projected earnings growth estimate is a proxy for long-term expected earnings growth. Price momentum refers to 12-month relative price change. Table 3. Correlation of US Large Cap Growth Manager Active Returns versus Active Returns for Long-term Growth Expectation and 12-Month Price Momentum Indices ( ) Manager Performance Quintile Long-Term Growth Expectation Source: Westpeak, based on data from IBES and PSN. 12-Month Price Momentum Q Q Q Q Q ALL Indices for long-term growth and momentum are created using the ActiveBeta Index Methodology. The Russell 1000 Growth Index return is subtracted from all manager, growth index, and momentum index returns each month. The large cap growth manager universe is divided into quintiles based on performance over the 10-year period from 1999 through Quintile 1 corresponds to the best-performing managers and Quintile 5 to the worst-performing managers. For each manager, we calculate the correlation of its active monthly return and the active monthly return of the long-term growth index. We repeat the procedure for the momentum index. The last row of Table 3 shows the average correlation of all growth managers against the projected growth index (column 2) and the momentum index (column 3). A New Approach to Equity Investing Page 9

10 We find that growth managers, as a group, have a 0.28 correlation with high momentum and a correlation with projected growth. The negative correlation with projected growth is probably due to growth managers tilting toward value within the growth universe. The other rows in the table show the average correlations for managers within the specified quintile. Figure 2. ActiveBeta Style Boxes Large Value Large Blend Large Momentum The best-performing managers (Quintile 1) depict an even more pronounced correlation with momentum active returns. Their active returns have a negative correlation of with long-term growth active returns and a highly positive correlation of 0.37 with the active returns of the momentum index. These relationships hold across all manager performance quintiles. The evidence presented argues that momentum appears to be a superior representation of the so-called growth investment style. Redefining Investment Styles and Style Boxes The ActiveBeta Framework would suggest that value and momentum more accurately describe active management styles than value and growth. In essence, value and growth capture just one systematic source of active equity returns, that arising from the systematic behavior of long-term earnings growth. Value delivers the positive payoff associated with the tendency for long-term earnings growth to mean revert, while growth provides the negative payoff. What is missing from the current style framework is the capture of the second (independent) systematic source of active equity returns, that arising from the systematic behavior of change in short-term earnings expectation to trend. Momentum provides an efficient capture of this systematic source. As such, momentum and value represent two independent systematic sources of active equity returns and provide a better framework for defining investment styles. This conclusion leads to the potential development of new style boxes to classify active managers, as shown in Figure 2. Size Medium Value Small Value Source: Westpeak. Medium Blend Small Blend Value/Momentum Medium Momentum Small Momentum Performance Attribution Decomposing Active Manager Returns If value and growth do not appropriately or fully reflect the investment styles of active managers, then performance attribution analysis based only on value and growth factors or indices could lead to misleading conclusions. Incorporating momentum, the missing systematic source, or style, into the analysis may provide a better and more complete performance attribution. The FTSE ActiveBeta Indices can be used to conduct a return decomposition analysis for active managers. We performed such an analysis for US large cap core, value and growth managers, over the 10-year period ending in 2008, using manager return data obtained from the PSN US Large Cap universe. (The results are discussed in detail in the book referenced above.) From this analysis, the following conclusions emerge: Core managers investment processes have a substantial value and momentum tilt. Growth managers are essentially momentum players with a small value tilt. Value managers are essentially value players with no meaningful bias to momentum or growth. The systematic tilts or exposures to momentum and value can then be used to determine what proportion of active manager returns constitutes pure alpha. This analysis shows that almost all active returns, especially for core and growth managers, are essentially attributable to systematic momentum and value tilts. A New Approach to Equity Investing Page 10

11 These conclusions paint a somewhat troubling picture for active managers, especially when we consider the existence of significant selection and survivorship biases in manager returns. This analysis implies that active managers, as a group and on average, do not provide active returns over and beyond those provided by the systematic sources (momentum and value tilts). Asset owners can use the FTSE ActiveBeta Indices to conduct an active return decomposition analysis of the active managers they employ. This analysis could help to quantify the pure alpha contribution of each manager. Performance Benchmarking As we have shown above, momentum and value better represent the systematic sources and the styles of active managers, compared to the traditional choices of growth and value. Therefore, the FTSE ActiveBeta Momentum and Value Indices represent more appropriate performance benchmarks for active style managers. In addition, in the current style-investing paradigm, the performance benchmark of core managers is typically the market index. But, should the market index be the performance benchmark for active core managers? By making the market index the neutral portfolio of core managers, the assumption is that core managers are style neutral. But, are they? The style decomposition analysis highlights that core managers have substantial momentum and value tilts. That is, they have a much higher exposure to high value and high momentum securities than they do to low value and low momentum securities. An appropriate performance benchmark for these managers could be one that limits the exposure to low value and low momentum securities. This is what the FTSE ActiveBeta MVI provides, and as such, it is perhaps a better performance benchmark for core managers, compared to the market index. Asset Allocation Allocating to Active Betas Current industry practice is to structure global equity portfolios in terms of a market beta component and an alpha component. The market beta component is implemented through a passive replication of a given market index. The alpha component aims to capture manager-specific skill and is implemented through active managers. However, to the extent that manager active returns largely originate from systematic sources, the decomposition of portfolio returns and structuring of portfolios into only market beta and alpha is incomplete and potentially misleading. Indeed, if alpha and market beta are the only two choices in decomposing portfolio returns, clear difficulties arise in distinguishing between systematic sources of active equity returns and a manager s true investment skill. In the alpha-beta return separation debate, therefore, there is a third source that should be considered explicitly Active Betas. This implies that asset owners should structure their overall equity portfolios in terms of three components, as depicted in Figure 3: Market beta, implemented through replication of market indices; Active Betas, implemented through replication of ActiveBeta Index-like vehicles; and Pure alpha, implemented through skilled active managers. Figure 3. Equity Management Return Sources Alpha Market Beta Traditional Perspective Source: Westpeak. Pure Alpha Active Betas Market Beta ActiveBeta Perspective Asset owners should realize that they are already exposed to systematic sources of active equity returns through the active managers they employ. Therefore, as a starting point, asset owners should view Active Betas as an efficient and costeffective alternative to capturing that portion of active management returns attributable to systematic sources. A New Approach to Equity Investing Page 11

12 One way of achieving this would be to reduce or eliminate investments in those managers who lack the skill to deliver pure alpha. Thus, the FTSE ActiveBeta Indices provide a framework and a vehicle for asset owners to streamline their overall equity portfolio structure by alleviating the burden of maintaining numerous manager relationships, concentrating investments in truly skilled managers, and allocating active risk and management fee budgets more efficiently. An allocation to Active Betas, however, could also come from the passive component of the overall equity portfolio. The FTSE ActiveBeta Index Methodology decomposes the overall selection universe into distinct quadrants, represented in Figure 4. FTSE ActiveBeta Value Index comprises quadrants 1 and 2; FTSE ActiveBeta Momentum Index comprises quadrants 2 and 3; and FTSE ActiveBeta Momentum and Value Index comprises quadrants 1, 2, and 3. Figure 4. FTSE ActiveBeta Style Quadrants In this index structure, as the securities in quadrant 4 are excluded from the FTSE ActiveBeta Indices, the combination of FTSE ActiveBeta Value and Momentum Indices do not simply add up to the entire selection universe, as is the case with most traditional value and growth indices. These securities have low momentum and low value characteristics. That is, they have low relative change in short-term expectation and high relative longterm growth expectation. These characteristics deliver the negative payoffs associated with the systematic behavior of earnings growth expectations. Table 4. Historical Performance of Quadrant 4 for US Large Cap, Europe ex-uk, and Japan Universes ( ) Universe Source: Westpeak. Annualized Active Return (%) Active Risk (%) Active Return Drawdown Max. (log%) US Large Cap Europe ex-uk Japan High Quadrant 1 High Value Low Momentum Quadrant 2 High Value High Momentum The historical performance of quadrant 4 securities for three universes, US Large Cap, Europe ex-uk, and Japan, for the 1992 through 2008 time period, is shown in Table 4. On average, these securities underperform the market universe by over 300 basis points per annum and have a very large active return maximum drawdown. Value Quadrant 4 Low Value Low Momentum Quadrant 3 Low Value High Momentum Low Low Source: Westpeak. Momentum High A New Approach to Equity Investing Page 12

13 Figure 5. Historical Performance of Quadrant 4 for US Large Cap ( ) Cumulative Active Return (log%) Active Returns ActiveBeta Q Source: Westpeak. The long-term and year-by-year active returns of quadrant 4 securities for the US Large Cap universe is shown in Figure 5. Not only do these securities underperform the market, on average, by a significant margin, they do so on a consistent basis. In the case of US Large Cap, these securities underperform in 12 of the 17 years under study. In Europe ex-uk, they also underperform in 12 out of 17 years. In Japan, they have negative performance in 14 out of 17 years. Some researchers have argued that market cap weighting introduces a large performance drag on market indices, by overweighting expensive securities and underweighting cheap securities. We argue, instead, that the performance drag largely comes from the quadrant 4 securities, which represent approximately 25% of the securities in the universe. Therefore, the FTSE ActiveBeta MVIs could constitute a more efficient way of capturing both the market beta and Active Beta components for those investors who are willing to bear the risk of not holding about 25% of the securities in the universe and the risk that quadrant 4 securities outperform the market on an occasional basis. If not holding 25% of the securities is considered too high a risk, then the number of securities not held can be reduced to below 10% through a patent-pending custom index methodology. (Please contact Westpeak Global Advisors for more details on the various custom indices and other options to capture Active Betas in implementation alternatives designed to better suit an investor s specific needs.) Investment Vehicles The FTSE ActiveBeta Indices are designed to be investable and replicable. Therefore, they are well-suited for the creation of a wide variety of investment vehicles, including mutual funds, ETFs, and derivatives. Investment products based on the FTSE ActiveBeta Indices would provide investors an efficient, transparent, and costeffective capture of systematic sources of active equity returns. In particular, the FTSE ActiveBeta MVIs provide a vehicle, in a single investment, for the combined capture of the two systematic sources in an internally-consistent family of global indices. The FTSE ActiveBeta MVI takes advantage of the significant diversification benefits offered by the counter-cyclical active return patterns of momentum and value. This efficient capture of Active Betas significantly lowers risk compared to the independent capture of either momentum or value. Given the amount of assets currently invested in pure value and growth strategies or linked to value and growth indices, this is potentially a very meaningful insight provided by the ActiveBeta Framework. A New Approach to Equity Investing Page 13

14 2. Research Findings 2.1 Why Active Betas? Reconciling Investment Beliefs and Observed Facts Many investors believe that markets are highly, though not perfectly, efficient and highly adaptive. In a highly efficient market, alpha does exist but it is hard to find, and requires significant investment skill. Further, the adaptive nature of markets implies that a particular source of alpha, once discovered, should lose its effectiveness over time as it becomes commoditized or is arbitraged away. Yet, sources of positive active returns, which do not require a high level of investment skill to capture, have been identified. These include: Size (market capitalization) Value (e.g., price-to-book value) Momentum (relative returns) From 1927 through 2008, momentum and value have provided significant excess returns. Many studies have shown the persistence of size, value, and momentum strategies. These sources of active returns were discovered decades ago, yet they persist. For instance, the excess returns over cash generated by the Fama/French 12-month price momentum factor and price-to-book value factor for US securities since 1927 is shown in Figure 6 below. From 1927 through 2008, momentum and value have provided significant excess returns, and this has been equally true since the public documentation of these sources of returns. How can these excess returns exist and persist if one accepts the highly efficient, adaptive nature of equity markets? Our research provides some fresh insights: some well-known sources of active returns are in fact systematic sources of active returns. Once these systematic sources of active returns are considered as additional forms of beta, as they should be, pure alpha does become much harder to find. This is consistent with the notion of a highly efficient and adaptive market. In order to understand what influences might give rise to systematic sources of active equity returns, we have to identify and study the behavior of the drivers of equity returns. A New Approach to Equity Investing Page 14

15 Figure 6. Fama/French Factors ( ) Cumulative Excess Return to Momentum 800% 700% 600% 500% 400% 300% 200% 100% 0% -100% Cumulative Excess Return to Value 500% 400% 300% 200% 100% 0% -100% Jan-27 Jan-30 Jan-33 Jan-36 Jan-39 Jan-42 Jan-45 Jan-48 Jan-51 Jan-54 Jan-57 Jan-60 Jan-63 Jan-66 Jan-69 Jan-72 Jan-73 Jan-76 Jan-79 Jan-81 Jan-84 Jan-87 Jan-90 Jan-93 Jan-96 Jan-99 Jan-02 Jan-05 Jan-08 Jan-27 Jan-30 Jan-33 Jan-36 Jan-39 Jan-42 Jan-45 Jan-48 Jan-51 Jan-54 Jan-57 Jan-60 Jan-63 Jan-66 Jan-69 Jan-72 Jan-73 Jan-76 Jan-79 Jan-81 Jan-84 Jan-87 Jan-90 Jan-93 Jan-96 Jan-99 Jan-02 Jan-05 Jan-08 Source: Westpeak, based on data from Kenneth French. ( A New Approach to Equity Investing Page 15

16 2.2 Drivers of Equity Returns A simple constant-growth earnings discount model can be used to identify the two primary drivers of price returns, as shown in Equation 4 in Figure 7 below. Figure 7. Drivers of Equity Returns (1) P t = E t / (k t g t ) And, (2) E t / P t = (k t g t) Where: P t = Price of a security at time t E t = Expected short-term earnings per share for next fiscal year at time t k t = Required rate of return for the security given the risk associated with it, at time t g t = Long-term (steady-state) expected growth in earnings at time t In terms of price change, the variable of interest, Equation 1 provides the following approximation: (3) ΔP t,t+1 ΔE t,t+1 Δ(k t g t ) t,t+1 And, (4) ΔP t,t+1 ΔE t,t+1 Δ(E t / P t ) t,t+1 Where: ΔP t,t+1 = Change in price between time t and t+1 ΔE t,t+1 = Change in short-term earnings expectation for next fiscal year between time t and t+1 Δ(E t /P t ) t,t+1 = Change in valuation between time t and t+1, which is a function of the change in k g between time t and t+1 Source: Westpeak. Driver #1: Change in Short-Term Earnings Expectation Driver #2: Change in Valuation The second driver of security returns is the change in valuation (the E/P or P/E ratio). The level of the E/P ratio is driven by the difference between the consensus estimate for the discount rate and the consensus long-term growth expectation (k g), according to Equation 2. The systematic sources of active equity returns arise from the systematic behavior of these two drivers of equity returns, as discussed in the next sections. 2.3 Short-Term Earnings Expectation and the Link with Price Momentum In this section, we study the behavior of the first driver of price returns, namely, change in expectation, and its strong link to price momentum. The various relationships explored in this section are graphically represented in Figure 8. These are: The relationship between price momentum and change in analyst expectation; The relationship between past change in analyst expectation and future change in analyst expectation; and The relationship between past price momentum and future price momentum. Figure 8. Relationships Studied Trend 2 Equation 4 in Figure 7 highlights that the first driver of security returns is the change in short-term consensus earnings expectation not the growth in short-term realized earnings. Price changes are driven primarily by the change in consensus earnings estimates over a given short-term future earnings horizon, rather than by growth in actual earnings. Analyst Expectation Change 1 t-1 t1 t Price Momentum Analyst Expectation Change t t+1 Price Momentum Let s assume that, at time t, the last 12-month realized earnings are $10 per share and the consensus expectation for next fiscal year earnings is $11 per share. At t+1, the last 12-month realized earnings are $10 and next year s consensus expectation is $12. Growth in 12-month realized earnings is 0% (10/10); but, the change in consensus expectation for next year, for the same look-ahead period, is 9% (12/11) between time t and t+1. It follows from Equation 4 that this change in consensus earnings expectation is an important driver of price change. Source: Westpeak. 3 Trend A New Approach to Equity Investing Page 16

17 Figure 9. Relationship between Price Momentum and Short-Term Earnings Expectations US Large Cap ( ) t-1 t P t-1 Momentum P t FY2 t-1 Estimate Change FY2 t Momentum (Returns) Past Year vs Correlation t-stat Earnings Estimate Change Past Year Time Series Correlation: Momentum (Returns) Past Year vs. Earnings Estimate Change Past Year Quintile Momentum Past Year (%) Earnings Estimate Change Past Year (%) Hi-Lo Spread Source: Westpeak, based on data from IBES. A New Approach to Equity Investing Page 17

18 Change in Short-Term Analyst Earnings Expectation and Price Momentum In order to study the behavior of short-term earnings expectations and its link with price momentum, we use Fiscal Year 2 (FY2) projections from the Institutional Brokers Estimate System (IBES) global database as a proxy for true market expectations. For the US large cap segment, we study the time period from 1985 through 2008 using the top 500 stocks by market cap. For international markets, we analyze the time period from 1992 through We do not present the results for the international markets in this paper, though we do discuss them briefly. (Detailed analysis of results for all markets and market segments appears in the book ActiveBeta Indexes, referenced above.) We investigate the relationship between past-year change in short-term analyst expectation and past-year price change (price momentum). The strength of this contemporaneous relationship provides verification that change in expectation is, indeed, a driver of price returns. Figure 9 provides evidence that a significant positive relationship exists in the US large cap market between 1-year price momentum (relative returns) and 1-year change in FY2 earnings estimates, on a contemporaneous basis. The rank correlation coefficient is an amazing 0.54 and is highly significant. For a single variable to explain price variation to this extent is truly remarkable. The graph in the middle of the figure plots the rank correlations for each year and highlights the stability of the relationship. The bottom portion of Figure 9 shows a quintile analysis. In this analysis, the top 500 companies are first ranked by past-year price momentum, from low to high, and then divided into quintiles containing about 100 stocks each (Column 1). The values reported for each quintile are the averages of the yearly median values. The other column contains the median values of the dependent variable (estimate change) within each price momentum quintile. The bar graphs next to the median values graphically depict the linear or non-linear nature of the relationship. The bar for each quintile is calculated relative to the median value of the quintiles. This analysis shows that sorting the universe on past-year price momentum is equivalent to sorting on FY2 estimate change. That is, the lowest past-year price momentum companies (Quintile 1) are also the lowest past-year estimate change companies, and this relationship is linear across all quintiles. A similarly strong relationship also holds for all major markets outside of the US. Fundamentally, price momentum (relative return) is contemporaneously driven by changes in short-term analyst expectation. This relationship implies that investing in high momentum stocks is substantially similar to investing in high change in short-term expectation stocks. Change in Short-Term Analyst Earnings Expectation Trends Is past 1-year change in FY2 estimates correlated with future 1-year and 2-year change in FY2 estimates? Figure 10 provides results for the US Large Cap universe. It shows that past 1-year change in FY2 estimates has a highly significant positive correlation of 0.21 with 1-year forward change in FY2 estimates. For ease of comprehension, we term this positive serial correlation of change in expectation as a trending behavior. The correlation of past 1-year change with 2-year forward change, on the other hand, drops to an insignificant level of The time series correlation graph indicates that, apart from two years, 2000 and 2001, the relationship with 1-year forward change stays positively correlated and is fairly stable over time. The quintile analysis corroborates the correlation statistics. That is, companies with low change in FY2 estimates in the last year (Quintile 1) tend to have lower change in FY2 estimates 1-year forward, compared to Quintile 5 companies. This relationship is quite linear across quintiles for 1-year forward change in FY2 estimates, but not for 2-year forward change. A New Approach to Equity Investing Page 18

19 Figure 10. Behavior of Short-Term Earnings Growth US Large Cap ( ) t-1 t t+1 t+2 FY2 t-1 FY2 t FY2 t+1 FY2 t+2 Past-Year 1-Year Forward 2-Year Forward Change Change Change FY2 Earnings Estimate Change Past Year vs Correlation t-stat Earnings Estimate Change 1-Year Forward Earnings Estimate Change 2-Years Forward Time Series Correlation: FY2 Earnings Estimate Change Past Year vs. 1-Year Forward Change Quintile Earnings Estimate Change Past Year (%) Earnings Estimate Change 1-Year Forward (%) Earnings Estimate Change 2-Years Forward (%) Hi-Lo Spread Source: Westpeak, based on data from IBES. A New Approach to Equity Investing Page 19

20 These results clearly indicate that change in short-term analyst expectation exhibits an average tendency to trend in the short run (up to one year). That is, companies that have experienced high change in earnings expectation in one period, relatively speaking, tend to have high change in earnings expectation in the next period. These results hold for all international markets, except for Japan, where the correlation of past 1-year change in FY2 estimates with 1-year and 2-year forward change is insignificant, implying that in the deflationary environment experienced by Japan during this time period, companies have been unable to maintain high relative earnings growth from one year to the next. Price Momentum Trends In the previous section, we reported a significant positive serial correlation in analyst growth expectations. We used analyst expectations as a proxy for true market expectations. If market expectations, which drive market pricing, behave like analyst expectations, then we should observe a similar positive serial correlation (or trend) in stock returns. We study this relationship by correlating past 1-year return (price momentum) with 1-year forward and 2-year forward return, or price momentum. Companies that have experienced high change in earnings expectation in one period, relatively speaking, tend to have high change in earnings expectation in the next period. A New Approach to Equity Investing Page 20

21 Figure 11. Investment Horizon of Momentum Strategies US Large Cap ( ) t-1 t t+1 t+2 P t-1 P t P t+1 P t+2 Momentum Momentum Momentum Past Year 1-Year Forward 2-Years Forward Price Momentum Past Year vs Correlation t-stat Price Momentum 1-Year Forward Price Momentum 2-Years Forward Time Series Correlation: Price Momentum Past Year vs. Price Momentum 1-Year Forward Quintile Price Momentum Past Year (%) Price Momentum 1-Year Forward (%) Price Momentum 2-Years Forward (%) Hi-Lo Spread Source: Westpeak. A New Approach to Equity Investing Page 21

22 Figure 11 provides the results for the US large cap market. The rank correlation coefficient of past 1-year price momentum with 1-year forward price momentum is a positive This level of correlation may seem low at first glance, but it is quite meaningful in the context of predicting future returns. It represents an economically exploitable relationship, as shown by the quintile analysis. The spread in 1-year forward price momentum between high past 1-year price momentum companies (Quintile 5) and low past 1-year price momentum companies (Quintile 1) is 440 basis points per annum, and the relationship is linear across quintiles. This relationship loses its linear characteristics for 2-year forward returns and the spread between Quintile 5 and Quintile 1 returns drops to only 30 basis points per annum. Across all markets, price momentum appears to exhibit behavior similar to analyst expectations. For all markets, except Japan, positive serial correlation in price momentum exists at the 1-year forward horizon. Beyond the first year, there is little evidence of any influence of past momentum on future returns. In other words, the horizon of momentum as an investment strategy is consistent with the horizon over which short-term earnings growth trends. Japan shows no evidence of serial correlation in momentum, whether looking at 1-year forward or 2-year forward returns. The results, however, are consistent with the serial correlation of analyst expectations. The results provide fresh insight into why momentum, as a strategy, has not worked in Japan over the period studied. Researchers who advocate behavioral inefficiencies for explaining momentum active returns characterize this finding as puzzling. Our research suggests that the failure of momentum in Japan can be simply explained by the lack of positive serial correlation in change in short-term earnings expectation. Summary The main findings reported in this section can be summarized (and generalized) as follows: Momentum is contemporaneously linked to change in analyst short-term earnings expectation. This relationship is highly significant in all markets. Change in short-term earnings expectation trends in the short run (up to one year). True market expectations embedded in market prices are likely to follow a similar pattern. Momentum exhibits a behavior identical to that of change in analyst expectation. That is, it trends in the short run (up to one year). This is why momentum is a short investment horizon strategy. The failure of momentum in Japan over the recent past is explained partly by the lack of positive serial correlation in change in short-term expectation. In summary, momentum active returns appear to be driven, and explained, partly by the average tendency of change in expectation to trend in the short term. Therefore, momentum is an investment strategy that provides an effective capture of the systematic portion of the change in expectation component of future price changes. 2.4 Long-Term Growth Expectation and the Link with Value In this section, we study the behavior of the second driver of price returns, namely, change in valuation, and its link with analyst long- term growth expectation. We have already documented in Figure 9 the strong contemporaneous relationship between price change (momentum) and change in expectation. As further verification, we document the strong contemporaneous relationship between price change and change in valuation (e.g., P/E ratio) in Figure 12. As can be seen in this figure, change in valuation has a highly significant relationship with price change, and the two drivers of security returns have a similar influence on price change over a 1-year change horizon. Figure 12. Relationship of Price Change with Change in Analyst Expectation and Change in Valuation US Large Cap ( ) t-1 t P t-1 P t Price Change FY2 t-1 P/E t-1 Expectation Change Valuation Change Price Change Past Year vs Correlation t-stat FY2 Estimate Change Past Year P/E Change Past Year Source: Westpeak, based on data from IBES. The various relationships explored in this section are graphically represented in Figure 13. These are: The relationship between current valuation and analyst longterm growth expectation. The relationship between current level of analyst long-term growth expectation and future change in that growth expectation. The relationship between current valuation level and future changes in valuation. FY2 t P/E t A New Approach to Equity Investing Page 22

23 Figure 13. Relationships Studied 1 Analyst Long-Term Growth Expectation Level t Source: Westpeak. P/E Level Value Reflects Long-Term Growth Expectations and Anticipated Risk The E/P ratio was specified in Equation 2 as: E t / P t = (k t g t ) Mean Reversion 2 3 Mean Reversion Analyst Long-Term Growth Expectation Change t t+1 P/E Change Price Change In this equation, k represents the required rate of return for a security, given its risk, and g represents the long-term expected earnings growth. Therefore, value can be defined and termed as a long-term growth-adjusted stock risk premium. Since true market expectations of g and k are unknown, to verify the relationship of Equation 2, we use analyst long-term expected growth estimates provided by IBES as a proxy for market expectations of long-term growth. Further, we consider a few commonly-used risk factors, such as leverage, earnings variability, and volatility of returns, as a proxy for stock-specific risk. As theory would suggest, and practitioners commonly know, we do find strong relationships between valuation ratios and growth prospects, and between valuation ratios and stock-specific risk, as shown in Figure 14. This figure analyzes the relationship between the current level of the P/E ratio and IBES long-term growth estimates, and between the current level of the P/E ratio and the three risk factors, for the US large cap market segment. The univariate rank correlation analysis shows that these variables independently have an influence on the level of the P/E ratio, with the strongest influence coming from the analyst longterm growth expectation. The univariate correlations can, however, be somewhat misleading if the various variables are highly correlated amongst themselves. Therefore, we also conduct a multiple regression analysis to identify the true influences. This analysis shows that the significance of the analyst long-term expected growth remains stable, with a rank correlation coefficient of 0.40, while the significance of leverage and return volatility drops substantially. Figure 14 also shows a quintile analysis, in which the universe is sorted by the P/E ratio, from low to high, and then divided into quintiles containing around 100 stocks each. This analysis highlights that when a universe is sorted by P/E quintiles, the most linear relationship is obtained with long-term growth expectation quintiles. Thus, value reflects current long-term growth expectation, and investing in low P/E stocks is similar to investing in low long-term expected growth stocks. The influence of risk factors, on the other hand, appears to be driven by the extremes, that is, Quintile 1 for leverage and earnings variability and Quintile 5 for volatility of returns. These conclusions hold for all the major markets and market segments we have studied. They clearly indicate that valuation ratios are a proxy for current estimates of long-term growth prospects. That is, low P/E companies have low long-term growth expectations, and vice versa. Thus, cross-sectional differences in valuation ratios across stocks are explained primarily by crosssectional differences in long-term growth prospects. A New Approach to Equity Investing Page 23

24 Figure 14. Relationship of P/E Ratio with Analyst Long-Term Growth Expectation and Various Risk Factors US Large Cap ( ) P/E Current vs Univariate Correlation t-stat Multiple Regression t-stat Long-Term Growth Expectation Leverage Earnings Variability Volatility Quintile P/E Current (%) Long-Term Growth Expectation (%) Leverage Earnings Variability Volatility Hi-Lo Spread Source: Westpeak, based on data from IBES. Change in Long-Term Growth Expectation Mean Reverts Having established the link between analyst long-term growth expectation and value, we now study the behavior of analyst long-term growth expectation over time. The relationship studied is whether the current level of analyst long-term expected growth (g) is correlated with future 1-year, 3-year, and 5-year percentage change in analyst long-term expected growth estimates. These results provide evidence of what we call mean reversion in analyst long-term expected growth estimates. The declining rate of increase in the correlation coefficient also suggests that the bulk of the mean reversion takes place within a three- to fiveyear period. These results also hold for the international markets studied. Figure 15 shows that the current level of analyst long-term expected growth has a highly significant negative correlation of -0.24, -0.38, and with 1-year forward, 3-year forward, and 5-year forward change in analyst long-term expected growth, respectively. These relationships are remarkably stable on a yearby-year basis. The quintile analysis further shows that companies with low (high) current analyst long-term expected growth, that is Quintile 1 (Quintile 5) companies, tend to experience increases (decreases) in their long-term growth forecasts over the next one year, three years, and five years. This relationship is linear across quintiles for the three forward periods studied. A New Approach to Equity Investing Page 24

25 Figure 15. Relationship between Current Analyst Long-Term Growth Expectation and Future Changes in Analyst Long-Term Growth Expectation US Large Cap ( ) t t+1 t+3 t+5 G t G t G t+1 G t+3 G t+5 1-Year Change 3-Year Change 5-Year Change Estimated Long-Term Growth Current vs Estimated Long-Term Growth 1-Year Change Estimated Long-Term Growth 3-Year Change Estimated Long-Term Growth 5-Year Change Correlation t-stat Time Series Correlation: Estimated Long-Term Growth Current vs. Estimated Long-Term Growth 1-Year, 3-Year, and 5-Year Change Quintile Year Change Year Change 1990 Estimated Long-Term Growth Current (%) Year Change Estimated Long-Term Growth Price Momentum Estimated Long-Term Growth 1-Year Change (%) 1-Year Forward 3-Year (%) Change (%) Price Estimated Momentum Long-Term Growth 2-Years 5-Year Forward Change (%) (%) Hi-Lo Spread Source: Westpeak, based on data from IBES. A New Approach to Equity Investing Page 25

26 Value Mean Reverts In the two previous sections, we reported 1) a significant relationship between value and analyst long-term growth expectation and 2) a significant negative serial correlation in analyst long-term growth expectation. If true market expectations embedded in market prices also behave like analyst expectations, then we should observe a similar negative serial correlation (or mean reversion) in valuation ratios. We investigate this relationship by correlating the current level of the P/E ratio with 1-year forward, 3-year forward, and 5-year forward change in the P/E ratio. In the case of the US large cap market, as shown in Figure 16, a strong tendency exists for P/E ratios to mean revert. This behavior of P/E ratios is highlighted by significant, highlynegative average rank correlation coefficients, and is evidenced in all markets. As with analyst long-term growth expectation, the bulk of the mean reversion occurs within three to five years. These relationships are remarkably consistent across time and within the various markets. The quintile analysis confirms the strength of the relationship, as well as its linear nature. Low P/E stocks systematically see their valuation ratios expand on average over time. Meanwhile, high P/E stocks witness a contraction in their ratios, as mean reversion takes effect. The mean reversion increases over time, as confirmed by the spreads between the low P/E (Quintile 1) stocks and the high P/E (Quintile 5) stocks. As with analyst long-term growth expectation, the bulk of the mean reversion of P/E ratios occurs within three to five years. A New Approach to Equity Investing Page 26

27 Figure 16. Relationship between Current Level of Valuation and Future Changes in Valuation US Large Cap ( ) t t+1 t+3 t+5 P/E t P/E t P/E t+1 P/E t+3 P/E t+5 1-Year Change 3-Year Change 5-Year Change P/E Current vs Correlation t-stat P/E 1-Year Change P/E 3-Year Change P/E 5-Year Change Time Series Correlation: P/E Current vs. P/E 1-Year, 3-Year, and 5-Year Change Quintile Year Change Year Change 1990 P/E Current (%) Year Change P/E 1-Year Change (%) Price Momentum P/E 1-Year Forward 3-Year (%) Change (%) Price Momentum P/E 2-Years 5-Year Forward Change (%) (%) Hi-Lo Spread Source: Westpeak. A New Approach to Equity Investing Page 27

28 Investment Horizon of Value Strategies As previously shown, change in valuation is an important component of price change. If, as suggested above, current value predicts change in valuation, then current value should also predict future change in price. Does current value predict future change in price, and if so, how far into the future? Figure 17 answers these questions by analyzing the relationship between current level of P/E ratio and 1-year, 3-year, and 5-year forward price change. The correlations between the current level of P/E ratio and forward price returns are consistently negative. This result is similar to the relationship between current P/E and forward change in P/E. However, and as expected, the connection between P/E and price change is lesser in magnitude. While these correlation coefficients may appear low, they represent economically exploitable relationships when considered in the context of predicting future returns. The correlations between the current level of P/E ratio and forward price returns are consistently negative. The average correlation coefficients become stronger, and more significant, as the horizon is lengthened. The correlation coefficient is for 1-year forward return, for 2-year forward return, and for 5-year forward return. The enhanced connection when moving from 1-year to 5-year forward price returns indicates that the effect accumulates over longer stretches of time. The quintile analysis also depicts that the forecasting ability of value increases with longer investment horizons, as evidenced by the spread between Quintile 1 and Quintile 5 returns for the 1- year, 3-year, and 5-year forward change horizons. In the case of US Large Cap, the quintile returns also highlight that the returns to value (or quintile spread differences) were primarily driven by high P/E stocks (Quintile 5) experiencing significantly lower returns than stocks in other quintiles. This result is due largely to the Technology Bubble, which produced much more pronounced swings in P/E ratios in certain sectors. In general, the investment horizon of a simple P/E ratio-based value strategy appears to be consistent with the mean reversion horizon of long-term earnings expectations and valuations. A New Approach to Equity Investing Page 28

29 Figure 17. Investment Horizon of Value Strategies US Large Cap ( ) t t+1 t+3 t+5 P/E t P t P t+1 P t+3 P t+5 1-Year Forward Return 3-Year Forward Return 5-Year Forward Return P/E Current vs Correlation t-stat Return 1-Year Forward Return 3-Years Forward Return 5-Years Forward Time Series Correlation: P/E Current vs. Return 1-Year, 3-Years, and 5-Years Forward Year Forward Quintile Year Forward P/E Current (%) Year Forward Return Price Momentum Return 1-Year Forward (%) 1-Year Forward 3-Years (%) Forward (%) Price Momentum Return 2-Years 5-Years Forward Forward (%) (%) Hi-Lo Spread Source: Westpeak. A New Approach to Equity Investing Page 29

30 Summary The main findings reported in this section can be summarized (and generalized) as follows: Value is contemporaneously linked to analyst long-term growth expectation. Analyst long-term growth expectation mean reverts in the long run (three to five years). The tendency to mean revert is strong in every market studied. True market expectations embedded in market prices are likely to follow a similar pattern. Valuations exhibit a pattern similar to long-term growth expectations and also mean revert in the long run (three to five years). Value is a long investment horizon strategy (three to five years). Its horizon is consistent with the horizon over which long-term growth expectations mean revert. Therefore, value active returns appear to be driven, and explained, partly by the average tendency of long-term growth expectation to mean revert in the long run (three to five years). The strong mean reversion of long-term growth expectation results in the mean reversion of the growth-adjusted stock risk premium (k g), which, in turn, leads to the mean reversion in valuation. As such, value is an investment strategy that provides an effective capture of the systematic portion of the change in valuation component of future price changes. 2.5 ActiveBeta Framework In previous sections, we have established strong links between momentum and change in expectation, and value and change in valuation. These relationships would imply that a significant portion of momentum future returns should come from the change in expectation component, while value future returns should be driven by the change in valuation component. We can verify this expectation by decomposing momentum and value future returns into the change in expectation and change in valuation components. Decomposing Momentum and Value Returns Table 5 decomposes the 1-year forward returns of high momentum and high value stocks in terms of the change in expectation and change in valuation components for the US large cap market. In this table, high momentum stocks are defined as the highest quintile of stocks based on a 12-month momentum ranking of the selection universe. The high value stocks represent the lowest quintile of stocks based on a P/E ranking of the universe. Change in expectation is defined as the change in IBES FY2 estimates over the last year. The P/E ratio is calculated using the IBES FY2 earnings estimates at each point in time. At the start of June of each year, the High Momentum and High Value Quintiles are identified and their median performance over the next year is decomposed in terms of a median change in expectation (change in FY2 estimates) contribution and a median change in valuation (change in P/E ratio) contribution to total returns. This process is repeated every year. Table 5 reports the average median returns coming from change in expectation and change in valuation. As Table 5 indicates, the High Momentum Quintile experienced a 16% increase in FY2 estimates and only a -5.7% change in the P/E ratio, leading to a 9.4% total return for this quintile. On the other hand, the High Value Quintile registers large positive increases in P/E ratios. Thus, a simple decomposition of momentum and value returns provides further evidence that momentum returns indeed are driven by higher short-term expectation change, caused by trends in short-term earnings expectation, while value returns are driven by expansions in valuation ratios, caused by the mean reversion of long-term earnings growth and the growthadjusted stock risk premium (k g). Table 5. Decomposing Momentum and Value Returns ( ) Top Quintile US Large Cap FY2 Estimate Change 1-Year Forward (%) P/E Change 1-Year Forward (%) High Momentum High Value Source: Westpeak, based on data from IBES. A New Approach to Equity Investing Page 30

31 Persistence of Momentum and Value Active Returns Understanding the source of momentum and value active returns is essential to determining whether these active returns are likely to persist in the future. Unfortunately, there is still little or no agreement on the source of such active returns in the academic or practitioner community. Broadly speaking, two somewhat diametrically opposed explanations are offered. On the one hand, the advocates of efficient markets, such as Fama and French, argue that value securities represent higher risk investments, and, therefore, value active returns merely represent a compensation for bearing higher risk. On the other hand, the proponents of inefficient markets, such as Lakonishok, Shleifer, and Vishny, argue that behavioral biases of investors and various agency issues better explain the persistence of these anomalous active returns. The ActiveBeta Framework provides some useful insights into the source of momentum and value returns. It highlights that momentum and value returns are driven by the systematic behavior of earnings growth and, as such, constitute systematic sources of active equity returns (or Active Betas). It further highlights that momentum active returns arise from the fact that positive change in expectation contributes more to future returns than the negative change in valuation. Value active returns, on the other hand, come from large positive change in valuation, which more than offsets the negative change in expectation. Why do momentum and value active returns persist over time? They persist because 1) the systematic behavior of earnings persists over time, and 2) the systematic behavior of earnings is difficult to fully incorporate in current expectations and prices. The systematic behavior of earnings to trend and then mean revert is the outcome of the fundamental way in which an open, competitive economic system works. In such a system, an individual company, such as Microsoft, may be able to sustain above-average earnings growth for a relatively long period of time. But, a large group of stocks, say 200 to 300, may be able to sustain above-average growth for only a short period of time. It is highly unlikely that above-average earnings growth for such a large group of stocks will continue indefinitely. Competitive forces will cause above-average growth rates to fall over time, on average, as new entrants appear to take advantage of above-average return on capital opportunities or existing competitors move to develop and bring to market competing products. Similarly, below-average growth rates tend to rise over time, on average, as companies restructure to become more efficient and profitable or exit from unattractive lines of businesses, thus raising their own profitability as well as the profitability of the companies that remain in those lines of businesses. Thus, competitive forces, a fundamental driver of an open economic system, cause earnings for large groups of stocks to behave in a trending, and then mean-reverting, fashion. The existence and persistence of the systematic behavior of earnings and earnings expectations is the outcome of the fundamental way in which open and competitive systems work in free-market economies. The systematic behavior of earnings is difficult to incorporate fully in current expectations and prices for two main reasons. First, there is uncertainty related to the magnitude and timing of the systematic behavior at a given point in time. As a result, investors adopt a conservative approach in the formation of their expectations. They discount the systematic tendency only slowly over time as new information becomes available and provides more clarity on the magnitude and timing of the systematic behavior. In the face of uncertainty, a reasonable argument can be made that this conservatism represents rational behavior and that blindly discounting the observed historical average tendencies in current expectations and prices may entail more hazards than benefits. Second, it is difficult to incorporate the average tendency of earnings to trend in the short run and mean revert in the long run in the pricing of individual securities. However, this average tendency is still effectively captured in broadly-diversified momentum and value portfolios. (The book ActiveBeta Indexes provides a more detailed discussion and statistical analysis on the pricing and persistence of systematic sources of active returns.) A New Approach to Equity Investing Page 31

32 Momentum, Value, and Investor Risk Aversion A Diversification Free Lunch? Our research highlights that value and momentum relative returns are influenced by the risk aversion of investors. That is, when investor risk aversion is high (down markets), value stocks underperform momentum stocks. When investor risk aversion is low (up markets), value stocks outperform momentum stocks. The link of momentum and value relative returns with the risk aversion of investors explains both the cyclicality and the counter-cyclicality (negative correlation) of momentum and value active returns. The risk aversion of investors is linked to the overall economic environment and its impact on the predictability of future earnings. When the visibility of future earnings shortens and earnings predictability risk increases, such as during a period of economic contraction, investors tend to focus on companies that have high current relative short-term earnings growth. In times like these, when risk aversion is high, investors tend to favor short-horizon momentum strategies over long-horizon value strategies, as was the case in 2007 and On the other hand, when the visibility of future earnings lengthens and earnings predictability risk decreases, such as during a period of economic expansion or toward the end of an economic contraction, investors tend to focus on companies where the likely mean reversion of current relatively low longterm growth will cause significant expansion in P/E multiples. During these times, when risk aversion is low, investors tend to favor long-horizon value strategies over short-horizon momentum strategies, as was the case during 2001 through 2006 and in the second quarter of Our research findings discussed thus far suggest that the marketplace provides two levels of diversification opportunities, of which investors do not currently take full advantage. First, simply holding broadly-diversified momentum and value portfolios significantly increases the probability of capturing the active returns associated with the average tendency of earnings growth to trend in the short run and mean revert in the long run, respectively. Concentrated momentum and value portfolios do not provide an efficient capture of this average tendency and rely more on a manager s stock selection skill to generate active returns. Second, because of the link of momentum and value with investor risk aversion, the independently positive momentum and value active returns also become negatively correlated. The negative correlation of active returns implies that by combining momentum and value into one strategy, investors can significantly improve the risk/return profile of the combined strategy, compared to independent momentum and value strategies. FTSE ActiveBeta Indices, therefore, have been developed to take advantage of the two levels of diversification benefits mentioned above. They provide 1) a broad and independent capture of the momentum and value systematic active returns, and 2) a combined capture of momentum and value active returns to improve the efficiency of capture. A New Approach to Equity Investing Page 32

33 2.6 Conclusion The various relationships discussed in Section 2 are summarized in Figure 18. Figure 18. Summary of Relationships Change in Price Tendency: Horizon: Capture: Source: Westpeak. Change in Expectation Δ P Δ E Trend [of expectations (E)] Short-Term Momentum Change in Valuation Δ E/P Long-Term Value Δ (k g) Mean Reversion [of growth (g)] The Main Tenets of the ActiveBeta Framework 1. Change in expectation and change in valuation are the two primary drivers of price change. 2. Change in valuation is driven by the change in the difference between the discount rate and the long-term growth expectation. 3. The average tendency for change in analyst expectation is to trend (positive serial correlation) in the short run (up to one year). 4. The average tendency for P/E ratios is to mean revert (negative serial correlation) in the long run (within three to five years). 5. This mean reversion in E/P ratios is caused by the mean reversion of long-term growth expectation, which causes the mean reversion of the growth-adjusted stock risk premium (k g). 6. Price momentum is contemporaneously linked to change in analyst short-term expectation. Momentum does not fully discount future systematic changes in analyst short-term expectation. Momentum strategies, therefore, represent one simple investment process whose active returns are driven by the undiscounted systematic behavior of change in shortterm expectation to trend in the short run (up to one year). As such, momentum is a short investment horizon strategy. 7. Value is contemporaneously linked to the level of analyst long-term growth expectation. Value does not fully discount future systematic changes in analyst long-term growth expectation. Value strategies, therefore, represent an investment process whose active returns are driven by the undiscounted systematic behavior of long-term growth expectation to mean revert (within three to five years) and the associated mean reversion of k g and valuation multiples. As such, value is a long investment horizon strategy. 8. Broadly-diversified momentum and value portfolios increase the probability of capturing the average systematic behavior of earnings expectations at a reasonable level of risk. 9. Momentum and value generate positive active returns, independently. Momentum and value active returns are also negatively correlated because of their link to investor risk aversion. This implies that a combined capture of momentum and value active returns should provide superior risk-adjusted performance compared to an independent capture of either momentum or value. A New Approach to Equity Investing Page 33