WHITE PAPER July 2015 For professional investors

The beauty and the beast of value and momentum investing

David Blitz, PhD Matthias Hanauer, PhD, CFA

The beauty and the beast of value and momentum investing | 1

Capturing factor premiums We usually focus on how to make a good investment strategy even better, but in this note we look at how to make it worse. In a previous note we already found a 6% return difference between good and bad low-volatility strategies. We now extend these results to value and momentum strategies and find similar performance gaps. Clearly, not all value stocks and not all momentum stocks are equally attractive. Our results argue against generic approaches towards capturing factor premiums, and in favor of a more selective approach. 1

The rise of factor investing Factor investing is inspired by the overwhelming evidence for the existence of various factor premiums, such as the well-known value and momentum premiums. The key idea behind factor investing is to allocate strategically to such factor premiums next to traditional asset classes. In our previous work we have argued that the three key factor premiums in the equity market are value, momentum and low-volatility, and that each of them deserves a sizable allocation.2 Our real-life strategies are designed to harvest these factors most efficiently. To that end we eliminate unrewarded risks, prevent going against other established factors, maximize diversification benefits, and avoid unnecessary portfolio turnover.

The best versus the worst factor strategies Our research and live track records show that it is indeed possible to create factor strategies that are superior to generic approaches. This made us wonder whether we could also construct inferior factor strategies. In other words, factor strategies that are not designed to outperform, but to underperform a generic approach. This started out as a kind of joke, but the results turned out to be quite interesting. We focus on value and momentum factor strategies in this note, having already considered low-volatility strategies in a previous note. The comparison between good and bad factor strategies is relevant for various reasons. For one, the bigger the gap between good and bad strategies, the more important it is to be selective when the objective is to capture factor premiums. Also, our analysis helps to better understand what differentiates a successful factor investing approach from an unsuccessful one. Finally, distinguishing between good and bad factor strategies may shed new light on the performance of generic factor strategies. These basically contain both types of stocks, although not necessarily in the same, constant proportion.

1

Blitz, Hanauer and van Vliet, “Beauty and the beast of low-volatility investing”, Robeco research paper, February 2015. See “Efficient factor investing strategies”, Robeco client research note, July 2014 and Blitz, “Strategic allocation to premiums in the equity market”, Journal of Index Investing, Spring 2012, Vol. 2, No. 4, pp. 42-49. 2

The beauty and the beast of value and momentum investing | 2

Mind the gap Methodology We simulate generic, good and bad global value strategies. For our generic value strategy we consider a model which ranks stocks on their price-to-book (P/B) ratio, price-toearnings (P/E) ratio and dividend yield (DY). The starting point for our good value strategy is a combination of more sophisticated value measures aimed at identifying stocks that are cheap but also have sound future earnings power. In addition, we take the momentum and risk characteristics of each stock into consideration. By doing so we aim to avoid value stocks that have a weak recent performance or that are very risky. For the bad value strategy we consider a model that takes the generic value factors as the starting point, but adds momentum and risk factors with the opposite sign. In other words, here we intentionally select cheap stocks with weak recent performance or high risk. The factor weights are chosen such that the P/B, P/E and DY characteristics of the good and bad value strategies are similar. We simulate generic, good and bad global momentum strategies in a very similar fashion. For our generic momentum strategy we consider a model which ranks stocks on their past twelve-month return excluding the most recent month. The starting point for our good momentum strategy is a combination of more sophisticated momentum measures, including not only price momentum factors but also analyst revisions. In addition, we take the value and risk characteristics of each stock into consideration. By doing so we aim to avoid good momentum stocks that are expensive or that are very risky. For the bad momentum strategy we consider a model that takes the generic momentum factor as the starting point, but adds value and risk factors with the opposite sign. In other words, here we intentionally select momentum stocks that are expensive or risky. The factor weights are chosen such that the 12-1 month momentum of the good and bad momentum strategies is similar. For all models we then simulate portfolios using the same buy-sell rules, conservative transaction costs and concentration limits on individual stocks, countries and sectors. The sample period is from January 1986 until December 2014. Figure 1 shows the full-sample performance characteristics of the three value strategies, and Figure 2 shows the same for the three momentum strategies.

Comparing different value and momentum strategies The generic strategies in Figures 1 and 2 confirm the existence of a value premium and a momentum premium. Over this sample period the value premium was particularly strong. Compared to the generic strategies, the good value strategy added 1.5% to return, while the good momentum strategy added 2.1%. Both good strategies also reduce volatility, implying that their added value is even larger on a risk-adjusted basis. Of particular interest though is the performance of the bad value and momentum strategies. We find that the bad value strategy lags the generic value strategy by 5.1%, while the bad momentum strategy lags the generic momentum strategy by 2.6%. Both bad strategies even fall short of the market return. They also exhibit a higher volatility than their generic counterparts, implying that they are even worse on a risk-adjusted basis. Adjusting for market beta we find alpha gaps between the good and bad strategies of 57%. These figures are closely in line with the results we found previously for good versus bad low-volatility strategies.

The beauty and the beast of value and momentum investing | 3

Figure 1 | Long-term performance of different value strategies

Source: Robeco Quantitative Research. Historical simulations, January 1986 – December 2014.

Figure 2 | Long-term performance of different momentum strategies

Source: Robeco Quantitative Research. Historical simulations, January 1986 – December 2014.

Exceptions do occur Although good value and momentum perform much better than bad value and momentum in the long run, success in the short run is not guaranteed. Bad value sometimes benefits from a recovery of stocks which have become cheap after a period of very poor and volatile performance, notably financials in 2009 (following the credit crisis in 2008) and technology stocks in 2003 (following the tech crash in 2000-2002). There are also several years in our sample during which good momentum significantly lags bad momentum, with 1999 standing out in particular. During that year bad momentum benefited strongly from the bubble in expensive, high-risk technology stocks.

Take-away Our results imply that value stocks and also momentum stocks are highly heterogeneous. Portfolios of stocks with the same value or momentum characteristics can have very different returns (and also a different risk), not only in the short run, but also in the long run. This highlights the importance of being selective when constructing a value or a momentum portfolio, in particular by avoiding stocks which are not attractive on other proven factors.

The beauty and the beast of value and momentum investing | 4

Don’t ignore other factors The key difference between the good and bad strategies is their exposure towards other factors. The good and bad value strategies have a similar value exposure, but one prefers value stocks with positive momentum and low risk, while the other prefers value stocks with negative momentum and high risk. Similarly, the good and bad momentum strategies have similar momentum exposure, but one prefers momentum stocks that are cheap and have low risk, while the other prefers momentum stocks that are expensive and have high risk. In the figures below we provide more color on this by showing key portfolio characteristics, measured relative to the market, over time. The left-hand side figures show results for the good and bad value strategies, and the right-hand side figures show results for the good and bad momentum strategies. Figure 3 | Portfolio characteristics relative to the market

Earnings To Price good value

Earnings To Price

bad value

good mom 4% 2% 0% -2% -4% -6% -8% -10%

10% 8% 6% 4% 2% 0% -2% -4% -6% -8%

Momentum 12-1M

Momentum 12-1M

good value

good mom

bad value

40%

0% -20% -40% -60%

Volatility 156W good value 15%

5% 0% -5% -10%

bad mom

140% 120% 100% 80% 60% 40% 20% 0%

20%

10%

bad mom

Volatility 156W

bad value

good mom

bad mom

30% 25% 20% 15% 10% 5% 0% -5%

Source: Robeco Quantitative Research. Historical simulations, January 1986 – December 2014.

The beauty and the beast of value and momentum investing | 5

The figures confirm that the good and bad value strategies have a very similar value exposure, here measured in terms of E/P, but markedly different momentum and risk characteristics. The good value strategy has a better momentum score over 99% of the time, and a lower past volatility all the time. On average, the holdings of the good value portfolio have a 12% higher momentum than the holdings of the bad value portfolio, and a 4% lower volatility, measured by the 156-week historical volatility. Similarly, the good and bad momentum strategies have a very similar momentum exposure, measured in terms of past 12-1 month momentum, but markedly different value and risk characteristics. The good momentum strategy has better value and risk characteristics than the bad momentum strategy throughout the entire sample period. On average, the holdings of the good momentum portfolio have a 2.4% higher earnings yield (E/P) than the holdings of the bad momentum portfolio, and a 12% lower volatility. Another interesting observation is that in absolute terms the good value strategy is, on average, about neutral on momentum and volatility, relative to the market. Also the good momentum strategy is, on average, about neutral on value and volatility relative to the market. In other words, the good value and momentum strategies are not ‘contaminated’ by the inclusion of other factors, but prevent the natural tendency of their generic (and bad) counterparts to go directly against other established factor premiums.

Generic ≈ Good + Bad The good and bad value strategies represent two extremes of the value spectrum, and the good and bad momentum strategies represent two extremes of the momentum spectrum. A generic strategy can also be interpreted as the approach that simply selects both types of stocks, i.e. good ones as well as bad ones. Importantly though, the proportion of good versus bad stocks is not constant over time. For instance, at some points in time the majority of value stocks may have favorable momentum and risk characteristics, while at other moments the majority of value stocks may have unfavorable characteristics. This implies that the expected return of a generic strategy is not always half-way in between the good and bad strategies, but sometimes closer to good, and at other times closer to bad.

The beauty and the beast of value and momentum investing | 6

Be selective In this note we have shown that different types of value and momentum stocks exhibit very different performance characteristics. A value approach which incorporates some momentum and risk considerations tends to outperform a generic approach, but a value approach which intentionally goes against these factors significantly underperforms. Similarly, a momentum approach which incorporates some value and risk considerations tends to outperform a generic approach, while a momentum approach which intentionally goes against these factors significantly underperforms. Over the past decades, the two approaches show a difference in risk-adjusted return amounting to an astounding 5-7% per annum. Based on these results it is clear that investors should not buy stocks merely based on one particular factor, but also take into account other factors that are known to have a large impact on returns. Generic factor strategies tend to contain a time-varying mix of stocks that are attractive and stocks that are unattractive on other factors. On balance, the exposures of a generic factor strategy to other factors are not necessarily neutral, but can also be unattractive. The Robeco approach towards factor investing effectively prevents going against other established factors.

David Blitz, PhD

Matthias Hanauer, PhD, CFA

Head Quant Equity Research

Quantitative Researcher

Important information This publication is intended for professional investors. Robeco Institutional Asset Management B.V. (trade register number: 24123167) has a license as manager of UCITS and AIFs from the Netherlands Authority for the Financial Markets in Amsterdam. This document is intended to provide general information on Robeco’s specific capabilities, but does not constitute a recommendation or an advice to buy or sell certain securities or investment products. The prospectus and the Key Investor Information Document for the Robeco Funds can all be obtained free of charge at www.robeco.com.

The beauty and the beast of value and momentum investing | 7