MindScope Use of the right factors can contribute to the best stock selection for a portfolio. But which factors are the right ones to use? And how can we most efficiently reap their rewards in factor portfolios? Multi-Factor Equity Investing www.nnip.com
Equity factor investing has become increasingly popular among a broad range of investors, including pension funds, mutual funds and sovereign wealth funds. Factor investing brings more transparency and lower costs, but the main driver of its rising popularity is that it can improve a portfolio s risk and return profile. It achieves this through the explicit allocation of risk in line with established drivers of return. Factors, in the context of stocks, are characteristics that explain the commonality in risks and returns. The most dominant factor is the equity market factor itself: all stocks tend to move in tandem with each other. In a single-factor model such as the Capital Asset Pricing Model (CAPM) 1, the market is the only factor driving the co-movement of stocks. The sensitivity of each stock is characterized by its market beta. All remaining risk and return expectations are idiosyncratic, or stock-specific, and can be diversified away by holding more stocks. Excess return can then only be added by superior skills, e.g. by selecting individual stocks that outperform the market after adjusting for beta. Numerous academic studies reveal that many factors other than the market drive risk and returns. The best-known and most established equity factors are value 2, momentum 3, low-vol 4 and quality 5. Use of the right factors can contribute to the best stock selection for a portfolio. Which factors are the right ones to use? And how can we most efficiently reap their rewards in factor portfolios? This Mindscope will answer these questions. Browsing through the factor jungle: factor selection The biggest pitfall in factor investing is selecting factors that are not true factors and do not carry a structural premium. A recent study shows that a large number of factors have been studied and are used in practice. Harvey, Liu and Zhu (2015) find no fewer than 316 factors documented in the academic literature, a number that is still growing! Figure 1 summarizes their results. Clearly we need some principles to guide us through this potential jungle of factors. Ultimately our assessment of a factor should be based on our expectation of its future efficacy, i.e. whether we think it will work in the future or not. We take several precautionary measures to make sure the factors we select and the way we construct them stand a good chance of delivering the anticipated benefits. First, we want stock selection factors to be supported by a clear and simple rationale. We should understand why the factor is expected to earn a premium. Otherwise it may be just a fluke that shows up in historical data and back-tests simply by chance. Given the massive increase of computational power available in the financial community, plus the fact that many studies are performed on similar data sets (typically US large cap stocks), this risk of data-mining is by no means an imaginary problem. But academic research and extensive in-house research guide us in the right direction. Second, we follow the evidence. The essence of factor investing is playing those strategies that have proven themselves and avoiding strategies that have not worked consistently over extended periods of time. Empirical and academic evidence supporting a factor must be very robust. This means that factors should work in multiple regions, sectors, periods and market environments. Obviously, no single factor will work everywhere all of the time. Individual factors generally have periods of underperformance. But in the long run (5+ years) factors tend to work. Moreover, a factor should not be subsumed by transaction costs. Figure 1: The number of documented factors has grown rapidly 350 300 250 200 150 100 50 0 1967 1970 1973 Dramatic growth in published factors in recent years 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 Idiosyncratic ( Characteristic ) factors Generalised ( Common ) factors Source: Harvey, Liu and Zhu, and the Cross-Section of Expected Returns, working paper version. Factors are from papers in finance, economics and accounting journals and SSRN working papers, collated by the authors. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2249314 Third, factors should be resilient to reasonable changes in their definition. For example, 11-month momentum should not exhibit very different behaviour from 12-month momentum. Similarly, price-to-earnings and price-to-cash flow should provide broadly comparable results. In our view it is rather meaningless, for example, to try to define the best value or quality factor. Although in broad terms it is clear what such factor entails, arguments on the exact definition are similar to discussions on the number of angels that can dance on the head of a pin. It is important that the concept of the factor can be reliably and robustly captured in a systematic manner. Value, momentum & quality All documented rewarding factors can be grouped into three comprehensive true factors. The three factors that we use to build a core multi-factor equity portfolio are value, momentum and quality. Each of these factors is well-established in its own right. They can also be efficiently combined together to construct a robust long-only equity portfolio. Value cheap stocks tend to have better performance 1 Capital Asset Pricing Model, Sharpe (1964) 2 Graham and Dodd (1923), Fama and French (1992) 3 Jegadeesh and Titman (1993) 4 Haugen and Heins (1972) 5 Novy-Marx (2013) Momentum Quality stocks that are doing well tend to continue to have better performance profitable companies tend to have better stock performance 2
Value The value factor relates to the old maxim of buy low, sell high : buy stocks when they are cheap, and sell stocks when they are expensive. There are several approaches to assess the cheapness of a stock. In general, valuation metrics boil down to comparing the market price of a stock to its fundamentals. The most frequently encountered fundamentals used to value a stock are book value, earnings, dividends, revenues and cash flows. When the valuation of a stock in is calculated using only one metric, for example dividend yield, the stock might appear to be selling at a steep discount. A look at several other valuation metrics, however, might reveal that the stock is rather expensive given its earnings and cash flow. The use of several value measures draws a more complete picture. For this reason, we combine multiple value characteristics to assess how attractive a stock is given the price at which it is trading and its historical and anticipated fundamentals. Momentum The momentum factor captures trends in the relative performance of stocks. This can be in terms of stock prices going up (winners outperforming losers, price momentum), but also in terms of earnings or positive changes in sell-side analyst estimates. A portfolio that every month buys stocks that had the best performance (or sells the worst performers) over the past 6 to 12 months outperforms market cap-weighted indexes. Behavioural finance experts explain this phenomenon by referring to the under-reaction of financial market participants to recent news. Similarly, stocks that experience upgrades in analysts consensus earnings estimates are more likely to outperform than recently downgraded stocks. Quality Unlike value and momentum, a quality factor has no generally accepted definition. The idea is that it is advantageous to buy stocks of a good or high-quality company, rather than those of a mediocre or lower-quality company. What distinguishes a highquality stock from a low-quality stock, however, is more subjective. A non-exhaustive list of characteristics that are often used as quality factors includes: high profitability, low accruals, low earnings and price volatility, stable dividend and earnings growth, high dividend coverage, low interest expense, high credit ratings, shareholder friendliness in the form of buybacks and low equity issuance, high workforce productivity and margin expansion. In order to systematically assess quality, we again look at a wide range of criteria that all link to certain facets of quality, such as profitability, financial robustness, productivity, earnings quality, growth at a reasonable price, and low-vol. The well-known low-vol factor is therefore subsumed in our quality factor. Ample academic and empirical evidence exists for the value, momentum and quality factors. We will further explore this evidence in the next section. Constructing portfolios with exposure to these factors makes intuitive sense, has a compelling investment rationale, and has demonstrated to add value in most market environments. We strongly believe in robust factor play. In defining a factor we combine several much more granular indicators. There are three reasons for doing so. First of all, the rationale for a factor typically does not imply a very narrow and specific definition. A single number does not tell the whole story. There are numerous ways to assess value, momentum and quality. The use of diverse methods to measure these attributes should result both in a more stable definition of the factor and a better translation of this factor into the portfolio. Selecting a specific metric (for example, price-to-book to assess value) is typically done in academic contexts to achieve exactness and reproducibility, and to avoid data-mining. In practice, though, it makes a lot more sense to refrain from a specific narrow definition, and instead use multiple equally valid definitions. Second, a broader definition has additional benefits when it comes to actual implementation. By construction, a diversified set of attributes will be more stable, and therefore warrant less turnover and transaction costs. Also there will be less overlap with other market participants, which could lead to everyone buying or selling the same stocks at the same time. Third, using a narrow definition is a bad idea because in certain sectors the use of a very specific factor is meaningless. For example, banks cannot be adequately compared using price-to-sales ratios as their business model is not driven by sales, but by interest rate margins. Price-to-book, on the other hand, is a good factor for banks but a poor metric for companies that have mostly intangible assets. The factor evidence Value, momentum and quality factors have delivered consistent returns and are expected to deliver attractive risk-adjusted returns going forward. In order to investigate the added value of the factors, we investigate the performance of portfolios that differ by their factor exposures. Every month we rank all stocks based on how they score on a particular factor. We then split the universe of all stocks into five quintile portfolios. The first quintile (Q1) contains the top 2 and the last quintile (Q5) contains the bottom 2. The performance of all these portfolios is depicted in Figure 2. Figure 2: Average annualized performance of five quintile portfolios for the factors value, momentum and quality 14% 14% 14% 12% 12% 12% 1 1 1 8% 8% 8% 6% 6% 6% 4% 4% 4% 2% 2% 2% Q5 Q4 Q3 Q2 Q1 Q5 Q4 Q3 Q2 Q1 Q5 Q4 Q3 Q2 Q1 Source (and for all subsequent figures and tables): NN IP Quantitative Research & Strategy, sample period 2002-2016, Global Developed Markets. 3
For all factors we observe a monotonic behaviour: the higher the portfolio scores on a factor, the better its return has been over this period. This clearly demonstrates that the factors bear significant risk premiums and that exposure to these factors is structurally rewarded. Although in the long run we observe strong outperformance of each of the factors, there can be prolonged periods of cumulative underperformance for any individual factor. The portfolio with the higher exposure to a factor (Q1) does not always outperform the portfolio with the lower exposure (Q5). This is shown in Figure 3a, which illustrates the outperformance of the top versus the bottom quintile portfolios (that is, Q1 Q5) for each of the value, momentum and quality factors, as well as their cumulative drawdowns (see Figure 3b). Corresponding performance and risk numbers are presented in Table 1. Figure 3a: Cumulative relative performance (Q1-Q5) of the factors 20 15 10 5 Table 1: Factor performance and risk statistics (based on Q1-Q5) Return (annualized) 7.2% 7. 4.5% Volatility 11.6% 9.9% 8.6% Information ratio 0.62 0.71 0.52 Max drawdown -34. -29. -21.2% Length (months) 64 32 72 The graphs also clearly show that different factors perform at different times. This should come as no surprise given their different rationale and background. The performances of the factors show low correlations (see Table 2), so we can expect most benefits if we successfully combine multiple factors. The lowest historical correlation is between value and momentum factor returns, again a feature for which we can have a clear intuition: the momentum factor favours stocks that have recently outperformed, but it is exactly this gain in relative return that often tends to make these stocks less attractive from a relative value perspective. Clearly momentum and value are not each other s opposites, but this interaction effect goes a long way in explaining the dependence between these two factors. Table 2: Monthly correlations of the factors (based on Q1-Q5) Momentum 1 Value -0.45 1 Quality 0.42-0.15 1-5 2002 2003 2004 2005 2006 2007 Figure 3b: Cumulative relative drawdowns (Q1-Q5) of the factors -5% -1-15% 2008 2009 2010 2011 2012 2013 2014 2015 2016 Next to the low correlations, the drawdown numbers and graphs highlight another important reason why it is beneficial to combine these factors: the underperformance of one factor may be compensated by the performance of one or more of the other factors. Constructing a long-only factor portfolio After having identified the factors that we want to have exposure to, the next step is to determine how to best incorporate these factors into long-only equity portfolios. As we already saw in the previous paragraph, we can anticipate most benefits if we succeed in combining multiple factors. -2-25% -3-35% -4 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 To gain insight into the alternative ways of putting the factors together, we compare the performance of the factor portfolios. Figure 4 shows the performance of the value, momentum and quality factor portfolios. These portfolios are long-only, and contain only those stocks that score in the top 2 (Q1) on the respective factor. In addition we include the Average portfolio, which simply splits its allocation equally between these three factor portfolios. Also shown is the Multi-Factor portfolio, which contains the 2 stocks that score best on the combination of factors. For reference we also include the equal-weighted market portfolio consisting of all stocks, irrespective of their factor exposures. 4
Figure 4a: Cumulative performance of long-only factor portfolios and the market 8 7 6 5 4 3 2 1 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Average Multi-Factor Market As can be immediately inferred from Figure 4a, the market factor is the dominant driver of risk. This can also be seen in the volatility numbers in Table 3, but this follows even more strikingly from the drawdown graphs in Figure 4b: for all portfolios the largest drawdowns occur simultaneously during the global financial crisis. There is no escape from this in a long-only equity portfolio, but we do see differential performances of the individual factor portfolios and the combinations thereof. Figure 4b: Cumulative drawdowns of long-only factor portfolios and the market -1-2 -3-4 -5-6 -7 Average Multi-Factor Market Table 3: Performance and risk statistics of long-only factor portfolios and the market Average Multi- Market Factor Return (annualized) 12. 11. 11.2% 11.5% 14.3% 8.2% Volatility 15.4% 18.1% 12.5% 14.9% 15. 14.8% Sharpe 0.58 0.44 0.66 0.57 0.76 0.35 Max Drawdown -50. -60.7% -42.4% -50.8% -50.2% -51.8% Length (months) 61 70 37 46 44 67 There is substantial added performance compared to the market portfolio of all factor portfolios (and then naturally also for the Average), and in particular for the Multi-Factor portfolio. This should come as no surprise as the market portfolio is the sum of all stocks (so, the sum of Q1 through Q5) and we already established the additional performance of the top 2 factor portfolios (Q1) in Figure 2. The performance of the Average portfolio is basically the average performance of the three long-only factor portfolios, whereas the performance of the Multi-Factor portfolio is considerably higher. To facilitate a better comparison, we take a closer look at the relative performance of the different portfolios compared to the market portfolio, as shown in Figures 5a and 5b. The corresponding statistics are displayed in Table 4. Again we see strong outperformance of all factor portfolios, with the best performance for the Multi-Factor portfolio. In terms of relative risk or tracking error (TE) we clearly see the benefits of diversifying over multiple factors, with lower risk numbers for the Multi-Factor strategy compared to the single-factor portfolios, and an even lower tracking error for the Average singlefactor portfolio. Comparing information ratios (relative return per unit of relative risk, or outperformance divided by tracking error) the Average and Multi-Factor strategies show similar results, both considerably higher than the single-factor portfolios. Inspecting the relative drawdowns of the different strategies, we observe a much more benign behaviour of the Average and Multi- Factor strategy compared to the individual single-factor portfolios, both in terms of depth and length of drawdowns. This further strengthens the case for incorporating multiple factors in the portfolio. Figure 5a: Cumulative relative performance of long-only factor portfolios 14 12 10 8 6 4 2 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Average Multi-Factor 5
Figure 5b: Cumulative relative drawdowns of long-only factor portfolios -5% -1-15% -2-25% 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Average Multi-Factor Table 4: Relative performance and risk statistics of long-only factor portfolios Average Multi- Factor Outperformance (annualized) 3.5% 3.1% 2.5% 3.1% 5.7% Tracking error 5.6% 5.3% 3.6% 2.1% 3.9% Information ratio 0.62 0.59 0.68 1.45 1.47 Max drawdown -19.8% -16.8% -7.6% -2.9% -6.5% Length (months) 64 33 56 16 24 What is the best way to combine the factors? In the previous section we clearly established the diversification benefits of combining the factors, either in the Average strategy (which allocates equally across the three single-factor long-only portfolios) or in the Multi-Factor strategy (which allocates based on the score on the combined set of factors). We now further compare these two alternative approaches. The Average portfolio represents perhaps the simplest way to achieve this diversification: just split the portfolio three ways, and allocate a third of the portfolio to each of the factors. This simplicity is maybe why many asset owners employ this method, but it comes with a considerable cost. First of all, the exposure to the targeted factors can be substantially diluted: if we target three factors, on average only one-third of the portfolio will be used to gain exposure to any specific factor. Clearly this gets even worse when more factors are in scope. The combination of factors through the Multi-Factor strategy does a much better job than the Average strategy in providing strategic exposure to the factor premiums in a long-only setting. This means that we should expect a lower return of the Average single-factor portfolio compared to the Multi-Factor portfolio. Second, a specific single-factor portfolio can at times have negative exposure to the other factors. Adding up all single-factor portfolios may then even result in a negative exposure of the total portfolio to one of the individual factors! In the long run this effect more or less averages out, as can be deduced from the similar information ratios. In the short run, however, the Multi-Factor portfolio will be much better aligned with the objective of gaining diversified systematic exposures to all three factors. Finally, when one single-factor portfolio is buying a specific stock, another single-factor portfolio may well be selling the same stock (recall the example of value and momentum). At the overall portfolio level, nothing changes, and the transaction costs could be saved. We don t observe the impact of this effect in the above analysis, as we ignore transaction costs when we rebalance all portfolios. Clearly, though, the benefits of netting are present only in the Multi- Factor portfolio, and not in the Average single-factor portfolio. Within a Multi-Factor portfolio, the potential pitfalls described above can be avoided. The full portfolio can be used to gain exposure to all targeted factors simultaneously, and an outcome where the portfolio ends up with a negative exposure to any specific factor can simply be avoided. Also, when the portfolio is rebalanced, the contribution of each stock to the combination of factors is considered. The scenario described earlier will never occur, and a stock will only be bought or sold if its contribution to the total set of factors contributes or detracts significantly. The difference between the Average strategy and the Multi-Factor strategy is in essence driven by relative risk: the Multi-Factor strategy reaches a considerably higher active risk, with similar riskadjusted returns (information ratio) and drawdowns per unit of volatility. As a result it can provide a much more efficient exposure to the different factors. Also considerations about the optimal number of stocks to balance maximal exposure to the targeted factors with the amount of total or active risk can only be meaningfully addressed within the context of the full portfolio. Towards an actual portfolio: incorporating additional constraints The long-only portfolios that we considered so far do not obey any further constraints. In practice there are several considerations that warrant imposing constraints on the actual portfolio. The most high-impact constraints are related to risk. The factor portfolios considered so far could at times be skewed towards certain regions or sectors. As an example, if US stock markets have gone up compared to the rest of the world, the momentum factor portfolio is likely to have a large tilt towards the US. Using multiple factors reduces these biases somewhat, but in practice, additional risk limitations and constraints are often imposed to avoid such biases. Most often these risk constraints are applied versus a benchmark, either in terms of active risk or tracking error, but also in terms of limits on relative sector or regional exposures. Additional constraints can be motivated by actual implementation considerations. These typically scale with the size of the portfolio, and are imposed to mitigate market impact at times of rebalancing. 6
This can lead to thresholds on market capitalization of a company, position sizing or on minimal trading volumes. An important set of constraints is implemented to exclude investments in stocks with highly controversial profiles from an Environmental, Social or Governance (ESG) perspective. In addition it is straightforward in a Multi-Factor strategy to also add ESG inclusion through the use of an ESG factor or low-carbon factor if the asset owner desires such a profile. All of these constraints are more efficiently applied at the final stage of portfolio construction. This is yet another reason why the Multi- Factor approach is superior to the Average single-factor approach. Summary We have presented a factor-based approach to core equity portfolios that should achieve consistent outperformance over traditional core equity portfolios. Our approach hinges on three factors momentum, value and quality, which all have outperformed market indexes over the long-run. The explicit allocation of risk in line with these established drivers of return leads to an improved risk and return profile. Additional benefits of the systematic use of these factors to make investment decisions are increased transparency and low cost. Bas Peeters Head of Quantitative Research & Strategy Karim Bannouh Senior Quantitative Strategist For more information about NN IP s Equity Factor Investing capabilities and solutions, or about the role Equity Factor Investing can play in your portfolio, please contact: Karim Bannouh (Karim.Bannouh@nnip.com) or Bas Peeters (Bas.Peeters@nnip.com). Factor robustness of our momentum, value and quality factors is supported by a solid economic rationale and academic evidence. There is extensive and strong empirical evidence across various definitions, sample periods and the performance of these factors has been found to be strong in several market environments and investment universes. Diversification over multiple types of factors increases returns and reduces risk in the long term. There are several ways to combine individual factors into a Multi-Factor strategy. Constructing single-factor portfolios and equal-weighting these results in higher risk-adjusted returns and smaller drawdowns than investing only in one factor portfolio. A more efficient approach of combining factors can be implemented through favouring stocks that score well on a combination of targeted factors. This approach increases efficiency as it avoids selecting stocks that score well on one factor but poorly on one or more other factors. In addition, this Multi-Factor strategy provides a better platform for translating this approach into an actual long-only equity portfolio. 7
Disclaimer The elements contained in this document have been prepared solely for the purpose of information and do not constitute an offer, in particular a prospectus or any invitation to treat, buy or sell any security or to participate in any trading strategy. This document is intended only for MiFID professional investors. While particular attention has been paid to the contents of this document, no guarantee, warranty or representation, express or implied, is given to the accuracy, correctness or completeness thereof. Any information given in this document may be subject to change or update without notice. Neither NN Investment Partners Holdings B.V., NN Investment Partners Holdings N.V. nor any other company or unit belonging to the NN Group, nor any of its officers, directors or employees can be held direct or indirect liable or responsible with respect to the information and/or recommendations of any kind expressed herein. The information contained in this document cannot be understood as provision of investment services. If you wish to obtain investment services please contact our office for advice. Use of the information contained in this document is solely at your risk. Investment sustains risk. Please note that the value of your investment may rise or fall and also that past performance is not indicative of future results and shall in no event be deemed as such. This document and information contained herein must not be copied, reproduced, distributed or passed to any person at any time without our prior written consent. Any claims arising out of or in connection with the terms and conditions of this disclaimer are governed by Dutch law. This document is not intended and may not be used to solicit sales of investments or subscription of securities in countries where this is prohibited by the relevant authorities or legislation. 8