Industry Rotation. Using Historical Base Rates to Identify Patterns of Persistence vs. Reversion Among Industries

Transcription

1 INSTITUTIONAL PERSPECTIVES January 2015 Industry Rotation Using Historical Base Rates to Identify Patterns of Persistence vs. Reversion Among Industries Introduction...One seasonal analysis we do find useful comes from studying the top- and bottom-performing industries for the trailing one- and three-year periods. As the calendar turns each year, stock market strategies for the coming year arrive in abundance as research firms and trading desks publish all sorts of historical data. You will likely come across quant wonks re-evaluating well-known seasonal anomalies such as the January effect 1 and the January Barometer 2 and testing new theories regarding technical market behavior. As students of capital markets, we think such studies are interesting but find them too short term- and macro-focused to add value to our investment process with a three- to five-year horizon. However, as value investors with an affinity for contrarian bets, one seasonal analysis we do find useful comes from studying the top- and bottom-performing industries for the trailing one- and three-year periods. The underlying premise for this analysis is reversion to the mean, which means topperforming industries will underperform the market at some point in the future and the worst-performing industries will revert to a higher mean over some time period. This analysis helps us in several ways. First, we screen for stocks in the worst-performing industries for further research to take advantage of potential reversion to a higher mean. It is important to emphasize that an underperforming stock in a worst-performing industry is only considered for investment if our fundamental research shows a meaningful gap between its current price and our estimate of its intrinsic value, and if we have a variant perception about how that spread might close over time. Second, we revisit our investment cases for any stocks we own within the top-performing industries to make sure there is sufficient margin of safety if and when the industry stops being a top performer. Finally, we manage the portfolio s exposure to these industries to make sure we are comfortable with our industry-level bets. It is fairly easy to identify the industries in the top and bottom deciles of relative performance over different time periods, and a lot of research shops will provide this data, but it is unknowable how long they will stay at either tail (persistence) or revert to the mean. The purpose of this white paper is to apply quantitative rigor to this problem and provide statistical base rates for how industries have performed over the past few business cycles. To study this, we obtained monthly observations of three-year annualized returns for sub-industries in the S&P 1500 Index since 1994 and divided the industries into deciles based on their performance relative to the index. Then we counted how many times each decile appeared x months after each monthly observation. The Research Design and Methodology section on page 4 provides more detail on the topic. 1 The January effect refers to the tendency of small-cap stocks to outperform large-cap stocks in the month of January. In perhaps the most extensive study of this anomaly covering the period , Hirschey and Haug (2006) find that the more than 30 years after its discovery, the January effect continues to be an anomaly in stock market returns. However, Ziemba and Dzahabarov (2011) update the analysis through 2010 and show that monthly anomalies, including the January effect, do not seem very profitable in recent years. 2 The January Barometer refers to the tendency that the market s direction in January will determine its direction for the full calendar year. In other words, if the market ends up for January, then it will rise for the full year, and if it closes down for January, then it will be flat or down for the full year. Based on a survey of academic papers as well as their own research on this topic, Ziemba and Dzahabarov (2011) note that positive returns in January tend to result in positive returns for the full year but if January is negative, the rest of the year is noise.

2 INDUSTRY ROTATION This exercise allowed us to assign base rates to the event that an industry will likely stay in the bottom or top over the following few months. It is important to underscore that these base rates are based on historical precedence and we are not attempting to create a predictive model of industry rotation. Therefore, while we use the terms base rate, chance, probability and likelihood interchangeably throughout this paper, these are consistently based on historical data. This analysis also highlighted important differences in the persistence versus reversion patterns among different sectors in the market. Specifically, industries in the staples, health care and industrials sectors revert away from the top and bottom deciles within a few months, while industries in the discretionary, energy, tech, materials and utilities sectors persisted in the top and bottom decile for one, and in some cases two, years. Key Findings Exhibit A shows the bottom- and top-10 industries in the S&P 1500 Index based on the past three years of relative returns, how long an industry has been in the bottom or top decile, and the base rate that the industry will stay in the bottom or top quintile (1 quintile = 2 deciles) over the next three and six months based on historical precedence. Exhibit A: Bottom- and Top-10 Industries Based on Three-Year Relative Performance and Base Rate of the Industry Staying in the Bottom/Top Quintile over the Next Few Months -5-44% -33% -26% -25% -19% -19% -19% -17% -17% Gold Coal & Consumable Fuel Oil & Gas Drilling Diversified Metals & Mining Education Services Steel Real Estate Development Oil & Gas Exploration & Production IT Consulting & Othr Srvcs Oil & Gas Equipment & Services Industry Ticker Sector # Consecutive Months in Bottom Decile Probability of Being in Bottom 2 Deciles in 3 Months Probability of Being in Bottom 2 Deciles in 6 Months S15GOLD Materials 22 57% 52% S15CCSF Energy 35 <13% <13% S15OILD Energy 15 64% 56% S15DIVM Materials 22 57% 52% S15EDSV Consumer Discretionary 39 <41% <41% S15STEL Materials 2 94% 88% S15RDVL Financials 1 92% 87% S15OILP Energy 2 88% 87% S15ITCS Information Technology 6 68% 6 S15OILE Energy 1 95% 85% Industry Ticker Sector # Consecutive Months in Top Decile Probability of Being in Top 2 Deciles in 3 Months Probability of Being in Top 2 Deciles in 6 Months Airlines Biotechnology Household Appliance Home Improvement Retail Oil & Gas Refining & Marketing Movies & Entertainment Building Products Distillers & Vintners Cable & Satellite Drug Retail 19% 16% 15% 14% 14% 13% 12% 28% 31% 36% S15AIRL Industrials 10 74% 68% S15BIOT Health Care 21 35% 3 S15HOAP Consumer Discretionary 4 85% 8 S15HOMI Consumer Discretionary 27 39% 32% S15OILR Energy 28 33% 3 S15MOVI Consumer Discretionary 9 76% 71% S15BUIL Industrials 6 77% 74% S15DSTL Consumer Staples 11 63% 54% S15CBST Consumer Discretionary 5 84% 78% S15DRUG Consumer Staples 2 87% 81% Notes: Analysis uses S&P 1500 as the benchmark and S&P 1500 GICS Level 4 Sub-Industry groups. Base rates are calculated for each industry relative to its sector rather than the entire market. Decile rankings are based on three-year annualized total return relative to the index from 1994 to Nov Source: ClearBridge analysis through Nov. 30,

3 INSTITUTIONAL PERSPECTIVES January 2015 The graphic above shows that the gold industry has underperformed the market by 5 over the past three years on an annualized basis, the industry has been in the bottom decile of performance for almost two years, and there is a 52% probability that it will remain in decile 9 or 10 for the next six months. Unless a manager has a variant perception for why gold stocks might outperform the market over the next six months, the high base rate for persistence would advocate avoiding / underweighting this group. On the other hand, the education services industry is an interesting source of ideas for further research because it has been in the dog house for over three years and has nearly a 6 probability of reverting to the mean over the next six months. Notably, four industries from the energy sector appear on the laggards list above but only coal & consumable fuel appears prime for reversion over the next six months, with an 87% chance of reverting to a higher quintile. Meanwhile, oil & gas exploration & production (E&P) and equipment & servicing companies will likely stay in the penalty box for a while longer based on the 85%-95% probability of remaining in the bottom quintile over the next six months. Value Equity s overweight in the E&P industry is an instructive example of how we use this analysis. While screening companies in this bottom-decile industry, we identified some trading at a discount to our estimate of their intrinsic value using our long-term fundamental valuation framework. This margin of safety gave us the confidence to invest in these companies at above-benchmark weights in anticipation of relative outperformance, even though the base rate shows that these industries will likely underperform over the next few months. The same logic applies in the lower half of the graphic. Airlines have outperformed the market by 36% over the past three years on an annualized basis, the industry has been in the top decile for the past 10 months and it will likely remain in the top two deciles of relative performance for Exhibit B: Probability of an Industry Staying in the Same Decile the next six months (68% probability). The over Time biotechnology industry, on the other hand, Energy has been in the top decile for close to two years and has a 7 chance of reverting to a lower quintile over the next six months. 8 8 Interestingly, oil & gas refinery & marketing appears on this list of recent winners where it 6 has been for the past 28 months, though this 6 blissful existence is getting long in the tooth, with a 7 probability of reversion to lower 4 4 deciles over the next six months Financials Notes: Analysis uses S&P 1500 as the benchmark and SP 1500 GICS Level 4 Sub-Industry groups. Probabilities are calculated for each industry relative to its sector rather than the market. Decile rankings are based on three-year annualized total return relative to the index from 1994 to Nov Source: ClearBridge analysis through Nov. 30, Data for the education services industry underscores an interesting takeaway from this research. Note that coal & consumable fuel has been in the worst decile for 35 months and has roughly less than 13% probability of staying in the bottom quintile over the next six months. Meanwhile, the education services industry has been in the bottom decile for longer, but has nearly three times the probability of staying in the bottom quintile over the same time period. While one would reasonably expect an industry with more time in detention to have a lower probability of extending their stay, we have found that persistence versus reversion patterns are very different among the 10 GICS sectors. This variation is useful as we evaluate our active exposures to different sectors and industries during the portfolio construction component of our investment process. To illustrate this point, Exhibit B shows the different probabilities of an industry staying in the same decile over time in the energy and financials sectors. Industries in the energy sector have a better-than-coin- 3

4 INDUSTRY ROTATION flip chance of staying in the top decile (blue line in the top graphic) for 13 months and in the worst decile (teal line) for 19 months. In the financials sector, however, a top-decile industry has greater than 5 chance of staying at the top for only a couple of months, while a bottom-decile industry remains there with a probability above 5 for 21 months! This shows that the persistence versus reversion context is very different depending on which sector an industry belongs to. This makes sense as different sectors have very different business model drivers, sources of return and investment needs. For instance, leverage is an important element of a financial company s business, which can punish a financials industry a lot longer than it supports the industry at the top of the league tables. Exhibit E shows the number of consecutive months an industry will likely stay in a decile with greater-than-5 probability and Exhibit F shows the rate of persistence for an industry in the context of the overall market and in the context of each sector. The base rates shown in Exhibit A should serve as a useful barometer for portfolio managers. If one is interested in sourcing ideas and/or tilting the portfolio towards/away from a certain industry, it is useful to know how long it is expected to stay at the bottom/top. As active managers, we are paid to invest differently than the market and our competitors. Therefore, active portfolio managers should absolutely bet against these base rates if they have a variant perception for why the future will unfold in a different way than it has on multiple occasions over the past 20 years. After all, active managers are paid to bet against the market and take risks for which they expect to get paid. But if they don t have a variant perception about an industry, or a group of stocks within an industry, these base rates should be a useful probabilistic approach to handicapping portfolio tilts. Background and Review of Academic Literature Before we delve into the research methodology and findings, let us spend a few minutes on the principles of reversion to the mean and persistence of a trend in stock markets. There is extensive academic literature on the momentum and reversal of stock price patterns over different time periods. For the purposes of this paper, momentum is analogous to persistence and reversal is analogous to reversion to the mean. The primary reason we use a different terminology is to distinguish between our study of base rates versus academic studies of price trends. DeBondt and Thaler (1985) examined patterns in individual stocks prices between 1933 and 1980 and found strong evidence that recent good-performing stocks became poor performers over three- and five-year holding periods, and vice versa. Using annual data from 1871 to 1985 and monthly data from 1926 to 1985, Poterba and Summers (1988) found evidence of momentum over short time horizons and reversals over longer periods. Using data from 1965 to 1989, Jegadeesh and Titman (1993) also found 4 evidence of short-term momentum. In perhaps the most influential paper on the topic, Fama and French (1996) show that their now-famous three-factor risk model fully explains the long-term reversals in stock prices documented in previous studies but it cannot fully account for mediumterm momentum. In other words, Fama and French claim that long-term reversal is fully explained by market returns and exposure to size and value, but momentum strategies remain profitable on a risk-adjusted basis. Finally, and in a nod to this paper s subject of research, Moskowitz and Grinblatt (1999) and O Neal (2000) suggest that intermediate-term momentum in individual stocks is actually driven by the performance of its industry. So why do momentum strategies work? Some researchers suggest that momentum is a proxy for higher-risk investments but, as noted above, Fama and French (1996) show that medium-term momentum persists even after accounting for standard risk factors. The other, and more commonly accepted, explanation for momentum is investor behavior, whereby investors underreact to new information and therefore it takes some time for the market to fully incorporate new information concerning a stock s value. While momentum and other technical analytical tools do not drive our investment process which focuses on a long-term investment horizon, these are a consideration in our stock selection and portfolio construction process because of their implications for near-term performance. Research Design and Methodology The primary unit of analysis for this research is a subindustry and the primary unit of measurement is the relative return of a sub-industry to its benchmark. We used the S&P 1500 Composite Index as proxy for the all-cap universe of stocks and level-4 sub-industry groups as defined by the Global Industry Classification Standard (GICS), although we will use the term industry throughout this paper to refer to these 156 sub-industries. We obtained total-return data for the industries from January 1994 to November 2014 from Bloomberg. We ranked the industries into deciles (1-10) each month using one- and three-year annualized returns relative to the benchmark. This gave us over 200 monthly observations for the 156 industries over the time period. Once we had a table of each industry s monthly observations by decile, we used a proprietary software program to count the number of times an industry appeared in a decile x months after observing its appearance in each of the other deciles. For instance, if we walk through this exercise for decile 1 and a one-month period in Exhibit C, decile 1 appears eight times in the sample, where an observation of decile 1 is followed by decile 1 on six occasions and decile 2 on two occasions. Therefore, based on this sample, there is 75% probability (6/8) that a decile 1 observation is followed by decile 1 in one month, and 25% probability (2/8) that a decile 1 observation is followed by decile 2 in one month. Similarly, if we repeat this exercise for decile 10, we would count that decile 10 appears five times,

5 INSTITUTIONAL PERSPECTIVES January 2015 Exhibit C: Hypothetical Example of Establishing the Base Rate Month Industry A Industry B Source: ClearBridge Investments. Exhibit D: Probability of an Industry Staying in the Same Decile over Time Using One-Year Returns Using Three-Year Annualized Returns where an observation of decile 10 is followed by the same decile three times and decile 9 two times. Therefore, based on this sample, there is 6 probability (3/5) that a sub-industry would stay in decile 10 (persistence), and 4 probability (2/5) that a sub-industry would move to decile 9 (reversion to the mean) in one month s time. We repeated this process for months 1 to 36 to obtain base rates for how an industry changes deciles each month up to three years. The primary motivation to analyze decile changes each month was to allow us to update the exercise monthly, while a secondary motivation was to gather a sufficiently large sample to derive statistically significant results. We performed this analysis for the market as a whole and then we repeated the analysis for the industries in each of the 10 GICS sectors, which allowed us to observe different base rates of persistence and reversion to the mean among the different sectors. Research Findings Notes: Analysis uses S&P 1500 as the benchmark and SP 1500 GICS Level 4 Sub-Industry groups. Decile rankings are based on three-year annualized total return relative to the index from 1994 to Nov Source: ClearBridge analysis through Nov. 30, We tabulated the results to calculate the probability that an industry stays in the same decile (persistence) or moves to another decile (reversion) over time. Given our interest in the top- and bottom-performing industries, we will limit the discussion to industries in deciles 1 and 10, which can only move toward the mean (reversion to the mean). However, in the final analysis, we looked at the probability of an industry in the tail decile persisting in the bottom-2 and top-2 deciles, or the bottom and top quintiles, because while the tail deciles can be interesting source of ideas, it makes sense for portfolio managers to wait for an industry to change quintiles and confirm a reversion to the mean before pulling the trigger to either buy or sell. Exhibit D shows the base rate that an industry remains in the same decile relative to the other industries in the market over a three-year period following an observation using one- and three-year returns. Using one-year returns, industries in deciles 1 and 10 still have very high probability of staying in that decile one month later (around 75%) but reversion sets in 5

6 INDUSTRY ROTATION quickly and the probability of staying in that decile falls below 5 (worse than a coin flip) within four months. Using threeyear returns, industries in deciles 1 and 10 also have over 8 probability of staying in the same decile one month after observation and this probability stays above 5 (better than a coin flip) up to 10 months for decile-1 industries and up to 13 months for decile-10 industries. It is worth making a few observations here: The analysis supports the academic findings of short- to medium-term momentum and longer-term reversal trends, albeit in the performance of industries rather than single stocks. This means that portfolio managers should not rush to source ideas from the worst-performing industries or overweight those industries, unless they have a variant perception and are willing to bet against the historical base rate. The period of evaluating initial performance matters in the tension between persistence and reversion to the mean. Analysis using one-year returns, which is commonly seen at year-ends, shows reversion to the mean within a few months, while analysis using three-year returns shows persistence for almost a year before reverting to the mean. This observation underscores the rather obvious point of knowing what sample you are using in data analysis as different time frames can have significant influence on the analysis. Given our three- to five-year investment horizon, we use the analysis based on three-year annualized returns as our primary tool going forward. This is because the probability of an industry staying in any decile drops very quickly with one-year returns and while this lends some credibility to the year-end reports from numerous research firms, the short-term reaction function is likely inappropriate for most of our strategies. We already discussed the other useful lesson from the project: it is inappropriate to treat all industries the same because the persistence versus reversion pattern varies significantly among industries in different sectors. Exhibit B demonstrates this graphically for industries in the energy and financials sectors and Exhibit F shows the charts for the market and each of the ten GICS sectors in the S&P 1500 Index. Exhibits E tabulates the data a little differently, showing the number of consecutive months an industry will likely stay in a decile with greater-than-5 probability. The 5 probability is useful for the tail deciles (1 and 10) because in any given month, an observation would either be in the same decile or not. As the exhibit shows, and consistent with the discussion accompanying Exhibit B, an industry in the Energy sector has greater-than-5 probability of staying in decile 1 for 13 months and in decile 10 for 19 months, while an industry in the Financials sector has greater-than-5 probability of staying in decile 1 for 2 months and in decile 10 for 21 months. It is interesting to add that industries in consumer staples, health care and industrials sectors revert away from deciles 1 and 10 within a few months, while industries in the consumer discretionary, energy, tech, materials and utilities sectors generally persist in that decile for one, and in some cases two, years. Exhibit E: Number of Months an Industry Will Likely Stay in a Decile (>5 Probability) All Sectors Consumer Discretionary Consumer Staples Energy Financials Health Care Industrials Information Technology Materials Telecommunication Services Utilities Notes: Analysis uses S&P 1500 as the benchmark and SP 1500 GICS Level 4 Sub-Industry groups. Base rates are calculated for each industry relative to its sector as well as relative to the market. Decile rankings are based on three-year annualized total return relative to the index from 1994 to Nov Source: ClearBridge analysis through Nov. 30,

7 INSTITUTIONAL PERSPECTIVES January 2015 Exhibit F: Base Rate of an Industry Staying in the Same Decile Ranking Using Three-Year Annualized Returns Market Consumer Discretionary Consumer Staples Energy Financials Health Care Industrials Information Technology Materials Telecommunication Services Utilities

8 INDUSTRY ROTATION About the Author Farhan Mustafa, CFA Director, Senior Research Analyst 11 years of investment industry experience Joined ClearBridge, LLC (f/k/a Legg Mason Capital Management, LLC) in 2003 Member of the CFA Institute BA in Economics and Computer Science from Washington and Lee University MBA from the Robert H. Smith School of Business at the University of Maryland References (in order of appearance): Ziemba, William T. and Dzahabarov, Constantine N. Seasonal Anomalies (February 12, 2011). Available at SSRN: Dunham, Lee. Momentum: The Technical Analysis Anomaly (2011). The Handbook of Equity Market Anomalies, Translating Market Inefficiencies into Effective Investment Strategies, Copyright: John Wiley & Sons, In., Hoboken, NJ. Haug, Mark and Hirschey, Mark. The January effect (February 2006). Financial Analysts Journal, Volume 62, The views expressed are those of ClearBridge Investments, LLC, are current as of January 2015, and are subject to change without notice. The opinions should not be construed as investment advice or recommendations on behalf of any ClearBridge or Legg Mason product. This material should not be used as the sole basis for an investment decision. Factual information relating to the topics covered was obtained from sources believed to be reliable, but there can be no guarantee as to their accuracy. To the extent that specific securities are mentioned in the commentary, they have been selected to illustrate views expressed in the commentary and do not represent securities purchased, sold or recommended for clients of ClearBridge Investments, LLC, and it should not be assumed that investments in such securities have been or will be profitable. Employees of ClearBridge Investments, LLC may own securities referenced herein. Past performance is not a guarantee of future results ClearBridge Investments, LLC. ClearBridge Investments 100 International Drive, Baltimore, MD ClearBridge.com Issue 5, DeBondt, Werner and Thaler, Richard. Does the Stock Market Overreact (July 1985). Journal of Finance, Volume 40, Issue 3, Poterba, James and Summers, Lawrence. Mean Reversion in Stock Prices: Evidence and Implications (August 1987). National Bureau of Economic Research (NBER) Working Paper No Available at Jegadeesh, Narasimhan and Titman, Sheridan. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency (March 1993). Journal of Finance, Volume 48, Issue 1, Fama, Eugene and French, Kenneth. Multifactor Explanations of Asset Pricing Anomalies (March 1996). Journal of Finance, Volume 51, Issue 1, Moskowitz, Tobias and Grinblatt, Mark. Do Industries Explain Momentum (August 1999). Journal of Finance, Volume 54, Issue 4, O Neal, Edward. Industry Momentum and Sector Mutual Funds (July/August 2000). Financial Analysts Journal, Volume 56, Issue 4,