Thesis MSc Finance University of Groningen
Smart beta portfolio strategies: A comparison with mutual funds and the market portfolio in the European stock market
Student number: S2578611
Name: Maickel Bonke
Study Program: MSc Finance
Supervisor: Prof. Dr. Theo Dijkstra
Field Key Words: Investment strategies, asset allocation, smart beta
Special Research Project: Smart beta portfolio strategies: Implementation and analysis
Master thesis version: 13-06-2016 Abstract
2
1. Introduction
The enormous rise in the amount of Exchange Traded Funds that implement Smart Beta strategies has made investing in different strategies more accessible than ever before. Currently there are over seventy ETF that are focussed on European equity alone. A large number of these ETF follow so called Smart Beta strategies. They break the link between the price of the stock and its weight in the
portfolio, explicitly deviating from the market capitalization weights. The ability of outperforming market cap-weighted indices combined with low expenses has attracted many investors into Smart Beta ETF at the expense of cash inflows to traditional passive and actively managed funds. Smart Beta ETF are seen as interesting assets to have in a portfolio. They can therefore force active managers to lower their fees, if not change their investment philosophy. The goal of this study is to evaluate the performance of several Smart Beta strategies compared to passive and active
management. The focus lies on European equity and active mutual equity funds during the period 1995 – 2015.
A universally accepted definition of Smart Beta is yet to be found. Arnott & Kose (2014) define it as ‘’A category of valuation-indifferent strategies that consciously and deliberately break the link between the price of an asset and its weight in the portfolio, seeking to earn excess returns over the cap-weighted benchmark by no longer weighting assets proportional to their popularity, while retaining most of the positive attributes of passive indexing’’ (p. 2). The positive attributes are transparency, rules-based investment style, low cost relative to active management, large capacity and good diversification. According to this definition Smart Beta possesses traits of passive and active investment management styles. However beating the market is hard enough for professional active managers, so how is a simple rules-based strategy able to do this?
According to Hsu (2006) cap-weighted portfolios have a high concentration in overpriced stocks and low in under-priced stocks. When the stock prices revert to their intrinsic value the cap-weighted portfolios lose value due to the price drop in the overpriced stocks and the gain in value of
3 Arnott, Hsu, Kalesnik and Tindall (2013) examined the performance of 36 different portfolio
strategies over the period 1964 – 2012. Although the main focus is on the U.S. equity market, they also present results for global developed markets, U.K. and French equity markets. The U.S. cap-weighted portfolio has the lowest mean return of all strategies and the second lowest Sharpe ratio. Globally the cap-weighted portfolio achieves the fourth lowest mean return and the lowest and second lowest mean return in France and the U.K. Many alternative strategies even produce a significant positive CAPM alpha. The fact that these results are found in several different equity markets around the world indicate that it is not just an anomaly. Unfortunately except for France and the U.K. Smart Beta strategies in European markets have not been analysed. The results that are available are supporting evidence for Hsu’s thesis, Hsu (2006).
Although Smart Beta strategies are not actively managed, they do attempt to beat the market in a rules-based investing approach. So they might be of interest to investors who currently aim to accomplish this trough expensive actively managed mutual funds. As is shown by Kahn and Lemmon (2015) investors are able to get factor exposure through cheaper Smart Beta funds. Therefore investments in actively managed funds could be revised. The study deals with global equity investment managers only and covers the period 2011 – 2014. To what extent the results are valuable for European equity managers is not discussed.
Research on Smart Beta performance in Europe is scarce and to what extent various Smart Beta strategies are able to outperform a market cap-weighted portfolio and actively managed funds is unknown. Therefore this study will attempt to answer the following question:
How do Smart Beta portfolio strategies perform in Europe in comparison to capitalization weighted and actively managed portfolios?
4
2. Literature review
The rising star of Smart Beta funds has attracted a lot of attention from academics and professionals in the investment world. Smart Beta strategies can be seen as a type of rules based investing and it has traits of both investment styles. I will first discuss the literature with respect to Smart Beta and passive management and follow up with active management. The investment styles will be
compared with each other and historical results will be discussed. I will end with a review on the various Smart Beta strategies.
2.1 Smart Beta and passive management
According to Hsu (2013), ‘’Smart Beta indices support investing in transparent, cost-efficient, easy-to-implement portfolios which encapsulate exposures across the full set of standard equity premiums.’’ The specific factors are the market, value, small cap, momentum and low volatility. These words are supported by Towers Watson (2013) who claims that Smart Beta gives the investor the opportunity to capture a wider spread of risk premiums than conventional market capitalization weighted strategies. However they only refer to the value, momentum and size factors. Kahn and Lemmon (2015) state that ‘’Smart Beta strategies offer some of the return advantages of active strategies with some of the implementation advantages of passive strategies’’ (p. 76). Smart Beta strategies are being recognized as a midway between passive and active investment strategies. Compared to a passive market cap-weighted strategy, the additional factor exposure offer investors the possibility of better future returns. Indeed, Fama and French (2012) show that over the period 1991 – 2010 the mean European market premium is 0.56% per month, the monthly mean value premium is 0.55%, the mean momentum premium is 0.92% and the mean size premium is -0.06%. Over this period additional factor exposure can increase returns, however the negative size premium shows that this is not by default for each factor.
5 equities that combine well with their domestic or global investments to get a mean-variance efficient portfolio. The investors will not build portfolios of European securities that is efficient relative to the European market. These investors have a different efficient set than investors who only invest in Europe. Consequently this influences equity prices and hence the weights and performance of the market cap-weighted portfolio. Hsu (2006) argues that a market cap-weighted portfolio is only mean-variance efficient or Sharpe ratio maximizing if the assumptions of the CAPM hold. However fully efficient markets are unlikely and a consequence of this is that stock prices do not fully reflect company fundamentals. This causes under-priced stocks to have smaller market capitalization than their true value would warrant, the opposite occurring for overpriced stocks. Hence severing the link between the price of a stock and its weight in a portfolio could be a good idea. Apart from
mathematical analysis Hsu (2006) also provides results for a backtest for the U.S. markets 1962 – 2003, and finds significant CAPM alphas. However, the results are not controlled for by the Carhart four factor model and thus possibly does not tell the whole story.
A more extensive backtest is conducted by Arnott et all (2013) for U.S. and global equity markets. The tested strategies base their portfolio weights on fundamentals and risk measures and twelve of them can be regarded as Smart Beta. Each portfolio achieved a better mean return in the U.S. 1962 – 2012 than the market cap portfolio and had a higher Sharpe ratio. In conflict with Baker and
Haugen’s (1991) prediction all the average returns of the alternatives are higher as well as most of the volatilities. Furthermore all of the average excess returns can be explained by the Carhart four factor model and no one has a significant alpha. For global markets 1991 – 2012 there are even better results. The book value weighted, five year average earnings weighted and fundamental weighted strategies do all have significant alphas. The results from this research are mixed, most returns can be explained by the factors, but internationally there are some significant alphas. Unfortunately the research only includes a few European countries and does not give an in-depth view of the European wide performance of Smart Beta strategies. Also one has to bear in mind that none of the strategies have been corrected for transaction costs or other fees. This could make a difference, especially when the alternatives were corrected for costs.
6 off in the future, ‘’It will depend on whether value and small-size tilts will still compensate investors for taking on the extra risk or whether they become more richly priced as more investors pour into these funds’’ (p. 4). This indeed could pose a risk, since when a profitable market anomaly has been observed it should in theory be arbitraged away. Hence the value and size premium could be significantly lower in the future. For the size premium this already seems to happen. Fama and French (2012) did not find size premiums in an international data set from 1990 – 2011.
2.2 Smart Beta and active management
Smart Beta strategies can be seen as an active strategy in the sense that it aims to improve risk-adjusted returns through exposure to more factors than just the market. They have some of the benefits of passive strategies in that portfolio construction is simple, rules-based, and transparent, which tend to lead to high capacity and low fees and costs, relative to traditional active strategies (Kahn and Lemmon, 2015). The extent of these benefits relative to the active strategies can differ substantially. Johnson, Bioy and Boyadzhiev (2016) show that the turnover of thirteen European equity Smart Beta ETF, rebalanced semi-annually, varies between 10% and 160%. Since higher turnover leads to higher costs, the benefits can indeed vary among Smart Beta strategies. According to Kahn and Lemmon (2015) investors are now able to get factor exposure at a lower cost than active management and thus investors should review their current allocations to active and index products. In the future active management should pursue alpha or returns investors cannot get by static exposure to factors. However this is no easy task. Carhart (1997) finds that only the top 10% of U.S. mutual funds earn their investment costs back. Research by Fama and French (2010) suggests that only the 97th, 98th and 99th percentile of active funds have managers with sufficient skill to produce
7 Producing a significant alpha is difficult, but Jacobs (2015) sees advantages of active management in contrast to passive management. He points out that the static rules-based investing with infrequent rebalancing suffers a serious flaw. Static strategies cannot respond to changes in stock fundamentals, markets and economic environment in the way that a dynamic, actively managed portfolio can. But then again most managers do not appear to have the skills to respond appropriately to these changes. From another point of view the static nature can be seen not as a flaw but as a strong trait, since Smart Beta strategies are less affected by short term optimism or pessimism and managers acting upon it. The competition of Smart Beta funds with actively managed funds is also discussed by Kahn and Lemmon (2016) who fear that Smart Beta will have a disruptive effect on the investment management industry. They argue that active managers whose returns originate largely from static exposure to factors will be forced to lower their fees or have to deliver pure alpha in the future. Static exposure is the long term average exposure to the factors value, momentum, size, quality and low volatility. The pure alpha component is the active return beyond static exposures to Smart Beta factors. Sources of pure alpha can be, but are not limited to, bottom-up security selection, top-down macroeconomic views, and the timing of Smart Beta factors. Another point is that the absence of an active manager when investing in Smart Beta. The investor himself will be responsible for the results of his investment choices and cannot put his trust in a professional manager. If one invests in an active mutual equity fund, the investor can hold the fund manager responsible for
underperformance. However, if one invests in a Smart Beta strategy it will follow that specific Smart Beta rules-based investment methodology. The investor himself is responsible for choosing his desired factor exposures and Smart Beta strategy. When the Smart Beta strategy underperforms the investor can only blame himself for his choices. This means that the investors need to be educated about Smart Beta and this can take time. Hence it can take some time for these changes to happen. Smart Beta appears to be a fierce competitor to actively managed funds. The lower fees, the simple and transparent way of investing appears to be a good substitute for the expensive active fund manager who is at least partly after the same factor premiums anyway. The fact that most active managers do not earn sufficient alpha to cover the investment costs is also beneficial for Smart Beta strategies.
2.3 Smart Beta strategies
8 Beta approaches. The first is thematic Smart Beta which tries to take advantage of temporal
mispricing issues due to the excessive focus on the short term of investors. Themes on which this strategy tries to capitalize are demographic development, emerging markets and future resource scarcity. An example of a thematic Smart Beta strategy is a fund that attempts to exploit the trend that people are living longer lives by investing in health care related companies. Although in the short term companies might underperform due to economic circumstances, it seems to be inevitable that future demand of health care products will increase. The strategy tries to profit from the
opportunities that are longer term in nature. The second category is systematic Smart Beta which aims to capture risk premiums or exploit inefficiencies in market cap investing. The systematic Smart Beta is subdivided into four approaches:
2.3.1 The equally weighted approach
This is the most basic approach of forming a Smart Beta portfolio. The portfolio weights are all equal and periodically rebalanced to restore the proper portfolio weights. In Table 1 (Arnott, Hsu, Kalesnik and Tindall, 2013, p. 93) it can be observed that over the past 50 years the equal weighted approach has outperformed the U.S. cap-weighted approach with a higher mean return and Sharpe ratio. The same results hold for a global developed markets portfolio over the last 20 years. These findings do not take costs into account, however these can make a difference for investors. The source of these higher returns can be found in exposures to small cap and value stocks and in the U.S. and globally a small and insignificant alpha remains. Another source could well be the phenomenon that equally weighted strategies work well in very volatile mean reverting environments. In order to restore the equal weights the strategy buys stocks whose price decrease below the mean and sells stocks whose value increase above the mean. Once the prices mean revert the strategy realizes good returns.
2.3.2 Economically weighted approach
9 multiple fundamentals and earnings growth outperform the U.S. market cap-weighted benchmark ranging from 1.52% to 2.76% per year. All the Sharpe Ratios are between 31% and 41% higher. With the exception of the earnings growth strategy all strategies indeed show a value tilt. The alphas are substantially larger than the equally weighted alpha, however still insignificant.
Table 1
Performance summary strategies United States (1964 – 2012)
2.3.3 Risk weighted approach
The risk weighted approach aims to improve the risk-return trade-off by making assumptions about future risk measurements based upon historical observations. According to Towers Watson (2013) there are three main risk weighted approaches:
2.3.3.1 Volatility based weighting
This approach bases the portfolio weights upon the inverses of the stock its standard deviation. This strategy aims for a good risk-return trade-off by minimizing the risk based on historical observations. An opposite strategy is also possible. Generally it is assumed that higher risk equals higher expected returns. Thus one can assign higher portfolio weights to more volatile stocks and reap the rewards. However the low risk anomaly is not speaking in favour of this strategy. According to Ang (2014) ‘’The risk anomaly is that risk, measured by market beta or volatility, is negatively related to returns’’ (p. 332). The low risk anomaly is confirmed in markets worldwide. Ang, Hodrick, Xing and Zhang (2009)
Strategy Annual return Standard deviation Sharpe ratio Annual alpha Alpha t-statistic Market exposure Size exposure Value exposure Momentum exposure U.S. cap-weighted 9.66% 15.29% 0.29 0.00% 0.00 1.00 0.00 0.00 0.00 Equal weighted 11.46% 17.37% 0.36 0.15% 0.38 1.05 0.38 0.12 -0.02 Book value weighted 11.23% 15.66% 0.38 0.54% 1.56 1.03 0.03 0.34 -0.10 Earnings 5 yr avg weighted 11.18% 15.08% 0.40 0.64% 1.92 1.00 0.00 0.31 -0.08 Fundamental weighted 11.60% 15.45% 0.41 0.64% 1.83 1.01 0.05 0.37 -0.09 Earnings growth weighted 12.42% 19.03% 0.38 0.96% 1.34 1.09 0.47 0.04 0.00 Volatility weighted 12.15% 19.13% 0.36 0.23% 0.46 1.10 0.55 0.16 -0.04 Inverse volatility 12.53% 15.64% 0.47 0.58% 1.13 0.97 0.28 0.33 -0.03 Market beta weighted 11.89% 19.76% 0.34 0.56% 1.01 1.13 0.54 0.13 -0.09 Inverse market beta 13.48% 16.40% 0.55 0.86% 1.07 0.91 0.25 0.43 0.03 Minimum variance 11.75% 11.69% 0.56 1.05% 1.39 0.70 0.13 0.34 0.00 Maximum diversification 11.99% 13.96% 0.48 0.40% 0.54 0.83 0.26 0.26 0.04
10 find for 23 developed markets that stocks with high past idiosyncratic volatility – after controlling for value, market and size - have lower future returns than stocks with low past idiosyncratic volatility. Baker and Haugen (2012) find similar results for stock volatility in 21 developed and 12 emerging markets for the period 1990 - 2012. They rank stocks by volatility and form deciles and find that the return and Sharpe ratio of the lowest decile is higher than the highest decile. When looking at Table 1 and the results of the volatility weighted and inverse volatility weighted strategies, one can see a hint of the existence of the risk anomaly. Although research by Arnott et all (2013) focusses on portfolio strategies, all the inverse volatility strategies outperform their counterparts.
2.3.3.2 Diversification strategies
These strategies aim to achieve superior risk adjusted returns, under the assumption that correlation is mispriced, by constructing a portfolio with very low correlations or maximum diversification. According to Choueifaty and Coignard (2008) the most diversified portfolio is obtained by maximizing the diversification ratio, which is the weighted average of individual volatilities divided by the
portfolio its volatility. The results in Table 1 show that an insignificant alpha of 0.4% is obtained. The strategy is somewhat similar to the minimum variance strategy. The portfolio construction is also more complex in comparison to the volatility based method.
2.3.3.3 Volatility minimization strategies
These strategies aim to minimize a portfolio its volatility by combining individual volatilities and correlations in an optimal way. The minimum variance portfolio is first developed by Markowitz in 1952. If one takes the low risk anomaly discussed before previously into account, a minimum variance portfolio should perform well. In Table 1 it becomes clear that it indeed has the highest Sharpe ratio of all strategies. Also an alpha of 1.05% is obtained, however it is insignificant. Constructing a minimum variance portfolio is also complex and requires many variance and correlation estimates, which makes it vulnerable for errors.
2.3.4 Factor tilts
11 premium, but it is rather a consequence of the weighting method. In order to form a pure factor portfolio the investor has to take long and short positions simultaneously so that the market portfolio is netted out (Ang, 2014).
The literature is divided about Smart Beta strategies. Market cap-weighted portfolios appear to be suboptimal in comparison to alternative strategies, however the results in Table 1 show that the higher mean returns of other strategies are mostly explained by exposure to other factors. A critical point is that these results are not corrected for costs. Hence they might be too optimistic. Still there is global evidence of significant alphas being generated by some Smart Beta strategies. For active management Smart Beta ETF can give static factor exposure for less costs. If it appears to be that mutual funds get their returns largely from factor exposure investors are better off buying cheaper Smart Beta ETF. There are many Smart Beta strategies that can be tested. I will analyse the
performance of the equity book value weighted, 5 year average earnings weighted and low volatility weighted strategies. The first two showed significant alphas globally and the third is interesting because of the low risk anomaly. The results will be analysed and compared with the performance of market cap portfolios and actively managed mutual funds.
3. Data and methodology
In this section the methodology of constructing and comparing the Smart Beta portfolios with the market cap portfolio and the actively managed portfolio will be discussed. The portfolio formation will use the methodology as is used by Arnott, Hsu and Moore (2005). I will form twenty portfolios over the period 1995 – 2015 for each strategy. First the dataset will be discussed, followed by the portfolio methodology and performance evaluation.
3.1 Data
12 In order to measure the returns for each stock the annual total return index (RI) of Datastream is used. The RI assumes that dividends are reinvested in the stock. The standard deviation of returns is calculated using the monthly RI. In order to get the equity book value I multiplied the common equity book value per share with the amount of common outstanding shares. For the five year earnings portfolio the net income is used. All of the used metrics are in line with the metrics used by Arnott et all (2013).
This Stoxx 600 data is also subject to survivorship bias, since there is no information with respect to the firms that left the index throughout the period. The bias can be solved by reconstructing the index with the old constituents. However this data is not available and hence one should take into account that returns are biased upwards. Another issue is that the net income and equity book value is reported in the domestic currency of the firms. This biases the portfolio weighting process towards firms with a domestic currency that is cheap relative to the Euro. The issue is solved by transforming all data noted in domestic currencies to Euros at the appropriate exchange rate. The exchange rate for a specific year is the average rate of that year.
The dataset for actively managed portfolios contain mutual funds located in Germany, Italy, France, Belgium and the Netherlands. The dataset contains 12,555 funds that focus on many different asset classes, funds that focus on European equity are filtered out. This results in a dataset containing 51 mutual funds over the period 1995 - 2015. Once more the data is subject to survivorship bias since only data on the mutual funds active end of sample is provided by Datastream. Blake, Elton and Gruber (1996) find that survivorship bias in mutual fund performance studies is a serious issue. For a performance study covering twenty years the bias is estimated to be 0.966% per year. Otten and Bams (2002) find a lower estimate of survivorship bias when comparing returns of mutual fund portfolios with and without dead funds. They find the bias to be 0.12% for Germany, 0.45% for Italy, 0.11% for the Netherlands and 0.15% for the U.K. over the period 1991 – 1998.
3.2 Portfolio methodology
13 used for the portfolio. The specific selection and weighting procedure for the market cap, equity book value, five year average earnings and low volatility portfolios are as follows.
The market cap portfolio contains all companies in the investable universe for a specific year. The weight of each stock in the portfolio is calculated by dividing the company its market capitalization by the sum of all market capitalizations. For construction of the equity book value weighted portfolio the same methodology will be used, only here the equity book value of the fiscal year ending two years prior to index construction will be used to determine the weight. The average earnings strategy bases the weights on the five years earnings average. The averaging period covers the five years ending two years prior to index construction. For the low volatility strategy the weights are based on the inverses of the standard deviation of monthly returns over the previous five years. The formula for determining the portfolio weight, w, of the ith firm for the market cap, equity book value and five
year average earnings strategies is as follows.
𝑤
𝑖=
𝑀𝑒𝑡𝑟𝑖𝑐𝑖
∑𝑁𝑖=1𝑀𝑒𝑡𝑟𝑖𝑐
(1)
The formula for determining the weight for the low volatility strategy is given below. In both formulas N=50.
𝑤
𝑖=
1 𝜎𝑖 ∑ 1 𝜎 𝑁 𝑖=1 (2) The first portfolio will be formed on 29 December 1995 and will be rebalanced annually on the last trading day in December. The annual rebalancing is in line with Arnott et all (2005) who find that monthly, quarterly and semi-annually rebalancing only increases turnover, but does not offer increased returns over annual rebalancing.In addition to the Smart Beta and market cap portfolios I will form one mutual funds portfolio. There are two main approaches to determine the weights, the equally weighted and value weighted approach. The value weighted approach is less sensitive to survivorship bias than the equally weighted approach and hence is a good option (Rohleder, Scholz and Wilkens, 2007). This weighting method is also used by Otten en Bams (2002). Despite these arguments in favor of the value
14
3.3 Portfolio evaluation
To analyse and compare the returns generated by the strategies I will use the performance measures also employed by Arnott et all (2013), in particular the Sharpe Ratio, the information ratio and the Carhart four factor alpha and factor exposures. But the main performance evaluation of the mutual funds will be the Carhart four factor model.
Mean returns alone do not provide an investor with enough information to evaluate an investment strategy. A more informative measure to compare and evaluate strategies is the Sharpe Ratio developed by Sharpe (1994). This ratio measures the return per unit of risk taken and is as follows;
𝐴𝑛𝑛𝑢𝑎𝑙𝑖𝑧𝑒𝑑 𝑆ℎ𝑎𝑟𝑝𝑒 𝑟𝑎𝑡𝑖𝑜 = (
𝜎𝐷̅𝐷
) ∗ √12
(3)where the average historical excess return is,
𝐷
̅
= 1𝑇
∑
𝐷𝑡𝑇
𝑡=1 (4)
where the excess return per month is,
𝐷𝑡 = 𝑅𝑃𝑡− 𝑅𝐹𝑡 (5)
In formula 3 the σD is the standard deviation of the monthly excess returns over the entire 1995 –
2015 period. In formula 5 the monthly portfolio return is RPt and the monthly risk free rate is RFt. For
the risk free rate the one month Euribor rate is used.
The information ratio is a measure to evaluate portfolio managers against a previously determined market index. If the information ratio exceeds 1.96 it can be inferred that the manager has a probability of 95% of beating the index in any period (Plantinga, 2012). The formula is as follows;
𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑖𝑜 =
𝑅𝑝−𝑅𝑚𝜎𝑡𝑒
(6)
The market index, Rm, is the market cap -weighted portfolio. The tracking error, σte, is calculated
separately and measures the standard deviation of the difference between portfolio returns and benchmark returns. The annualized tracking error is calculated with the following formula;
𝑇𝑟𝑎𝑐𝑘𝑖𝑛𝑔 𝑒𝑟𝑟𝑜𝑟 = √
12𝑇
∑
(𝑅
𝑝,𝑡− 𝑅
𝑚,𝑡)
2 𝑇
𝑡=1 (7)
where Rp, t is the return of the portfolio at time t, Rm,t is the return of the market index at time t and T
15 Finally, I use the Carhart four factor model, originally developed by Fama and French (1996) and extended by Carhart (1997), for an analysis of the returns. The model allows one to find the factor exposure of a portfolio and explains the origins of the returns. The returns that the model cannot explain are subsumed by alpha, α, and can be a sign of investor selection skill or a superior strategy. The model captures exposure to the market, value, size and momentum factor and is as follows;
𝑅𝑖− 𝑅𝑓 = 𝛼𝑖+ 𝑏𝑖(𝑅𝑚− 𝑅𝑓) + 𝑠𝑖𝑆𝑀𝐵 + ℎ𝑖𝐻𝑀𝐿 + 𝑚𝑖𝑀𝑂𝑀 + 𝜀𝑖 (8)
where Ri is the portfolio or mutual funds return, Rf is the risk free rate, αis the excess return, Rm is
the market return and bi is the exposure to the market, Si is the exposure to the size factor and SMB is the size premium, hi is the exposure to the value factor and HML is the value premium, mi is the
exposure to the momentum factor and MOM is the momentum premium, finally εi is the error term.
A positive loading on the size factor indicates a small cap tilt and a negative loading a large cap tilt. For the value factor a positive loading indicates a value tilt and a negative loading a growth tilt. A positive loading for the momentum factor inclines that the portfolio consists of winner stocks and with a negative loading of loser stocks. Arnott et all (2013) use U.S. and global factor returns
16 French, since this is the most reliable source and makes the results comparable with other papers. In addition I will run the Fama French model with the market cap-weighted portfolio as a proxy for the market factor. The Fama and French size, value and momentum data will be used for the other factors. By using a market factor proxy that represents the actual market returns in the investable universe, it is possible to analyse the difference between using more representative factor data and less representative factor data for the Fama French model.
Arnott et all (2013) acknowledged that their results are not corrected for transaction costs and fees. To give more realistic results I will correct the returns for their expense ratios. Currently the cheapest way of investing in the Stoxx Europe 600 index is to buy an ETF. The same holds for the Smart Beta strategies of which several are being offered as ETF products. For both the expense ratio can be observed and deducted from the returns. All mutual funds used in this research are active today and the expense ratios are provided by Morningstar. The expense ratio includes all professional fees, management fees, audit fees and custody fees. It should be noted that buy/sales charges and performance fees are excluded from the costs, hence the true costs could even be higher. The average of these expense ratios will be used for the mutual funds costs.
4. Results
In this section I will present and discuss the performance of the European Smart Beta portfolios in comparison with the market cap portfolio and the actively managed mutual funds.
17 deviate in return in either direction from the market cap portfolio. Arnott et all (2013) find similar results, although they do not find negative information ratios.
Table 2
Performance summary market cap, Smart Beta and mutual funds uncorrected for costs (1995 – 2015). Standard deviation and tracking error calculated with monthly portfolio returns. Returns are uncorrected for costs.
Portfolio Annual return Standard deviation Sharpe ratio Value added Tracking error Information ratio Market cap 9.03% 16.53% 0.46 0.00% 0.00% 0.00 Low volatility 13.03% 10.56% 0.98 4.00% 9.63% 0.44
Book value equity 7.49% 18.59% 0.35 -1.54% 4.78% -0.32
Five year average income 7.95% 17.44% 0.39 -1.08% 3.98% -0.27
Mutual funds 7.50% 16.89% 0.37 n/a n/a n/a
The main suspect for the large difference in results when compared to Arnott et all (2013) is the survivorship bias in the data. Since no company goes bankrupt or disappears from the dataset certain strategies profit from this due to the weighting process. Although not presented, a high volatility strategy achieves a mean annual return of 23%. This is simply the result of selecting stocks with large, presumably negative, price swings. However these stocks will not go bankrupt or disappear out of the Stoxx 600 index, so over time as the prices mean-revert the stock has a large weight and the portfolio profits. The specific effect of the survivorship bias on the other strategies is more ambiguous and it is unknown to what extent a strategy profits from the bias. The effect can be analysed by comparing these results with the results of a survivorship bias free sample over the same period. Over the period 2001 – 2015 the Stoxx Europe 600 index realized a mean annual return of 3.32% against 4.04% for the market cap portfolio. For this portfolio the mean annual returns are 0.72% higher, which lays between the survivorship bias estimates for mutual funds by Otten and Bams (2002) and Blake et all (1996). It appears to be that the returns of the market cap and mutual fund portfolio increase with roughly the same percentage due to the bias.
18 Table 3
Performance summary market cap, Smart Beta and mutual funds corrected for costs (1995 – 2015). Standard deviation and tracking error calculated with monthly portfolio returns. Returns are corrected for costs by subtracting the monthly cost
percentage from the monthly returns.
Portfolio Annual return Difference net-gross return Standard deviation Sharpe ratio Value added Tracking error Information ratio Market cap 8.74% -0.29% 16.53% 0.44 0.00% 0.00% 0.00 Low volatility 12.60% -0.44% 10.56% 0.95 3.86% 9.63% 0.40
Book value equity 7.07% -0.42% 18.59% 0.33 -1.67% 4.78% -0.35
Five year average income 7.53% -0.42% 17.44% 0.36 -1.21% 3.98% -0.30
Mutual funds 5.76% -1.74% 16.89% 0.27 n/a n/a n/a
After the correction for costs the mutual fund portfolio has the lowest mean return and Sharpe ratio. These results show that the average mutual fund underperforms all portfolios and that the costs correction further weakens the case for active portfolio management. The low volatility strategy remains the best performing strategy by far, although the difference with the market cap portfolio decreases. Furthermore the Sharpe ratio of the book value equity and income portfolio are nearly identical to those in Arnott et all (2013). The results support the findings of Baker and Haugen (1991) and Hsu (2006) that a market cap portfolio is suboptimal. However only one portfolio outperforms the market cap portfolio and overall performance is good. The information ratios do not change to a significant level.
The performance analysis is continued with the Carhart four factor model. In the regression the risk free rate and factor data is provided by Fama and French. In order to correctly interpret the results from the Carhart four factor model, it is important that four ordinary least squares assumptions hold. Violation of these assumptions can result in wrong coefficient estimates or wrong standard errors. The first assumption is that the average value of the errors is zero. This assumption will never be violated if a constant term is included in the regression equation, which is. The second assumption is that the errors have a constant and finite variance, also known as the assumption of
19 Heteroscedasticity is detected in all regressions except the mutual funds. Autocorrelation is only detected in the market cap-weighted portfolio regression were the same portfolio is used as a proxy for the market factor. Therefore Newey-West corrected standard errors are used in all regressions, except for mutual funds.
The results in Table 4 show that all portfolios produce a positive alpha, yet only the market cap and low volatility alphas are highly significant even after costs. All portfolios show a profound tilt towards large cap companies. Although the negative momentum exposure is similar to the findings of Arnott et all (2013), the other factor exposures are very different. These differences are most likely the result of the survivorship bias in the data. Large firms tend to survive and hence are overrepresented in the portfolios. This also explains why the market cap portfolio is not only exposed to the market, but also to other factors. These unusual outcomes make it difficult to adequately discuss the performance. It is unclear to what extent the bias influences the results. Hence the performance cannot be completely attributed to a certain strategy. To decrease the effects of the survivorship bias one could compute the monthly factor returns with the stocks in the dataset. Whilst this will give more accurate factor data specifically tailored to the investable universe of the dataset, it will also deviate from actual market data. Hence the outcome of the regression can differ from outcomes of the model being applied to a survivorship free dataset using real market data. Due to a lack of data it is only possible to compute factor returns for the market factor. The results of the Carhart four factor model using the market cap portfolio as a proxy for the market factor is presented in Table 5.
Table 4
Carhart four factor model results (1995 – 2015). Factor data and risk free rate retrieved from Fama and French. Newey-West corrected standard errors used for all regressions, except for mutual funds. Monthly FF alpha net is the alpha
corrected for costs.
Portfolio Monthly alpha gross in % Monthly alpha net in % Market exposure Value exposure Size exposure Momentum exposure Market cap 0.498*** 0.476*** 0.71*** -0.08 -0.42*** -0.18*** Low volatility 0.639*** 0.606*** 0.38*** 0.11 -0.18** 0.01
Book value equity 0.260 0.228 0.77*** 0.21*** -0.45*** -0.17***
Five year average income 0.313* 0.280 0.72*** 0.14** -0.46*** -0.15***
Mutual funds 0.267 0.129 0.76*** -0.26*** -0.15* -0.04
Note. * is significant at 10% level, ** is significant at 5% level, *** is significant at 1% level.
20 overall. When compared to the results of Arnott et all (2013) the factor exposures are quite similar, except for the significant negative alphas. These alphas could disappear when I were to use not just specific factor data for the market factor but also for the value, size and momentum factors.
Table 5
Carhart four factor model results with market cap portfolio as market factor (1995 – 2015). Value, size, momentum and risk free rate retrieved from Fama and French. Newey-West corrected standard errors used for all regressions.
Portfolio Monthly alpha gross in % Monthly alpha net in % Market exposure Value exposure Size exposure Momentum exposure Market cap 0.000 -0.023*** 1.00*** 0.00 0.00 0.00 Low volatility 0.327** 0.295** 0.58*** 0.15** 0.07 0.12***
Book value equity -0.259*** -0.292*** 1.07*** 0.30*** 0.00 0.02
Five year average income -0.182** -0.215*** 1.01*** 0.22*** -0.03 0.03 Note. * is significant at 10% level, ** is significant at 5% level, *** is significant at 1% level.
21
5. Conclusion
This study aimed to analyse the performance of Smart Beta strategies in European equity markets. Outside of Europe transparent, low cost, rules-based, large capacity and diversified portfolio strategies performed significantly better than standard cap-weighted portfolios and thus can be a true added value for investors. The goal of this study was to answer the following question:
How do Smart Beta portfolio strategies perform in Europe in comparison to capitalization weighted and actively managed portfolios?
After back testing three Smart Beta strategies over the period 1995-2015 the results show that the book value equity and five year average income strategies underperform the market cap-weighted portfolio by more than one percent per annum. On the other hand the performance of the low volatility strategy is superior to all other portfolios and displays a substantial added value over the market cap-weighted portfolio. After expenses the mutual funds realize the lowest mean return and Sharpe ratio of all portfolios. The Carhart four factor model regression results show a large market exposure of the book value equity and five year average income strategies. All the Smart Beta strategies are value tilted where the mutual funds are growth tilted. Only the low volatility strategy captures a significant positive alpha in both four factor models.
22 The results should be interpreted with the knowledge that the data is flawed by a survivorship bias. This applies to the market cap-weighted and Smart Beta portfolios, but also to the mutual funds. A consequence is that the returns possibly are overestimated compared to a more realistic survivorship bias free data set. For further research this is one of the main issues to tackle when assessing the performance of Smart Beta strategies. Research with a survivorship bias free data set can assess the true performance of the Smart Beta strategies in Europe. Furthermore there are many more Smart Beta strategies of which the performance can be analysed.
All in all this study has shown that Smart Beta strategies can outperform the market and are an attractive alternative to the expensive mutual funds, especially when costs are taken into account. The performance of the strategies do differ amongst each other and hence it is not self-evident that Smart Beta strategies outperform the suboptimal market cap-weighted portfolio.
23
References
Ang, A., 2014. Asset management: A systematic approach to factor investing. Oxford university press, Oxford.
Ang, A., Hodrick, R., Xing, Y., & Zhang, X., 2009. High idiosyncratic volatility and low returns: International and further U.S. evidence. Journal of Financial Economics, 91(1), 1-23.
Arnott, R., Hsu, J., & Moore, P., 2005. Fundamental indexation. Financial analysts journal, 61(2), 83-99.
Arnott, R., Hsu, J., Kalesnik, V., & Tindall, P., 2013. The surprising alpha from Malkiel's Monkey and upside-down strategies. Journal of Portfolio Management, 39(4), 1-16.
Arnott, R., & Kose, E., 2014. What smart beta means to us. Retrieved from
https://www.researchaffiliates.com/Our%20Ideas/Insights/Fundamentals/Pages/292_What_Smart_ Beta_Means_to_Us.aspx
Baker, N., & Haugen, R., 1991. The efficient market inefficiency of capitalization-weighted stock portfolios. The journal of portfolio management, 17(3), 35-40.
Baker, N., & Haugen, R., 2012. Low risk stocks outperform within all observable markets of the world. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2055431
Blake, C., Elton, E., & Gruber, M., 1996. Survivorship bias and mutual fund performance. The Review of Financial Studies, 9(4), 1097-1120.
Carhart, M., 1997. On persistence in mutual fund performance. European financial management, 52(1), 57-82.
Choueifaty, Y., & Coignard, Y., 2008. Towards maximum diversification. The journal of portfolio management, 35(1), 40-51.
Fama, E. F. & K. R. French., 1996. Multifactor explanations of asset pricing anomalies. Journal of Finance, 51(1), 55-84.
Fama, E. F. & K. R. French., 2010. Luck versus skill in the cross-section of mutual fund returns. Journal of Finance, 65(5), 1915 – 1947.
24 Hsu, J., 2006. Cap-weighted portfolios are sub-optimal portfolios. Journal of Investment
Management, 4(3), 1-10.
Hsu, J., 2013. The genesis of smart beta Investing. Retrieved from
http://www.researchaffiliates.com/Our%20Ideas/Insights/Fundamentals/Pages/SBS/1-The-Genesis-of-Smart-Beta-Investing.aspx
Jacobs, B., 2015. Is smart beta state of the art? The Journal of Portfolio Management, 41(4), 1-3. Johnson, B., Bioy, H., & Boyadzhiev, D. 2016. Assessing the true cost of strategic-beta ETFs. Retrieved from http://media.morningstar.com/uk/MEDIA/ETF/AssessingtheTrueCostofStrategicBetaETFs.pdf Kahn, R., & Lemmon, M., 2015. Smart Beta: The owner’s manual. The Journal of Portfolio
Management, 41(2), 76-83.
Kahn, R., & Lemmon, M., 2016. The asset manager’s dilemma: How smart beta is disrupting the investment management industry. Financial analysts journal, 72(1), 15-20.
Malkiel, B., 2014. Is smart beta really smart? Journal of Portfolio Management, 40(5), 127-134 Otten, R., & Bams, D., 2002. European mutual fund performance. European financial
management, 8(1), 75-101.
Plantinga, A., 2012. Reader portfolio management. Unpublished reader, Rijksuniversiteit Groningen, Groningen.
Rohleder, M., Scholz, H., Wilkens, M., 2007. Survivorship bias and mutual fund performance:
Relevance, significance, and methodical differences. Unpublished working paper. Catholic University of Eichstaett-Ingostadt, Ingostadt.
Sharpe, W., 1994. The Sharpe ratio. The journal of portfolio management, 21(1), 60-89. Towers Watson., 2013. Understanding smart beta. Retrieved from
25
Appendix A
The number of firms in the investable universe and the number of firms that have required data available for a specific metric in a certain year. In order to be part of the investable universe a firm needs to have the following data available;
- The return index one year prior to portfolio construction - The equity book value two year prior to portfolio construction - The net income six year prior to portfolio construction
- The monthly volatility over the previous five years prior to index construction
26
Appendix B
The annual returns of the market cap, Smart Beta and mutual funds portfolio.
