Managing Sector Risk in Factor Investing
1. Implicit sector risks in smart factor indices
Common smart factor indices give explicit exposures to priced risk factors that should provide good long-term risk-adjusted performance. But they are also known to expose investors to a number of hidden or implicit risks as it has been documented in a recent Scientific Beta white paper (Shirbini, 2018). In particular, investors expose themselves to an implicit bet on market beta, since most smart factor indices have a market beta below one. Other implicit risks are macroeconomics risks and sector or country risks. In this article, we will focus on the implicit sector risk taken by smart factor indices and will try to understand the implications on their short and long-term risk-adjusted performance. We will also discuss possibilities to avoid sector risks through appropriate risk control options, in particular the sector neutrality constraints introduced in Amenc and Goltz (2013)1 and available on the Scientific Beta index platform since its launch in 2013. Our analysis in this article extends the earlier analysis of Amenc et al. (2015), which analysed the benefits of sector-neutrality constraints in value factor indices.
Smart factor indices have time-varying sector relative allocations compared to the cap-weighted (CW) index. It is well known that, for instance, the Low Volatility factor tends to overweight sectors like utilities and non-cyclical consumer goods and underweights riskier sectors. In the case of the Low Volatility factor, these relative allocations are fairly stable through time, as seen in Exhibit 1. But we also see that for the High Momentum factor, these relative allocations are much more varying through time. It is therefore evident that sectors, even if they are not priced in the cross-section of expected returns, can nevertheless have an important impact on the short-term performance of smart factor indices that might negatively affect investors. This is why sector neutrality is very often used in the asset management industry, especially for investors who need tracking error control, which is the case for most institutional investors.
Exhibit 1: Historical relative sector allocation of EDHEC-Risk Long-Term United States HFI Low Volatility and High Momentum Diversified Multi-Strategy (4-Strategy) smart factor indices
The data are based on quarterly sector allocations from 31-Dec-1976 to 31-Dec-2016
In the rest of this article, we focus on a comparison between standard smart factor indices and their sector neutral counterparts. Our key findings are that sector neutrality adds value in terms of reducing tracking error and short-term underperformance of the cap-weighted reference index, but also comes with costs in the form of higher volatility and lower factor intensity. Moreover, the sector-neutrality objective naturally reduces the distance of strategy weights to market capitalisation weights, which may not be suitable for investors who are looking for a pronounced difference with cap-weighted indices. Given these trade-offs, the most appropriate solution depends on investor preferences. It is precisely the objective of Scientific Beta's Smart Beta 2.0 framework to allow investors to make such explicit choices, not just on the targeted factors but also on additional risk control options. The remainder of this article is organised as follows. In section 2, we explain the two complementary approaches to control sector risk in factor investing. In section 3, we perform an analysis between standard smart factor indices (with no sector-neutral objective) and sector-neutral smart factor indices on absolute risk-adjusted performance. In section 4, we analyse their differences in terms of factor exposures. In section 5, we analyse their relative risk-adjusted performance and show the short-term risk that can arise from sector exposures. Finally, we draw the conclusions of this article. 2. Two complementary approaches to control sector risk There are two important aspects that investors should take into account to control sector risk in factor investing, which are i) stock selection to maintain proper sector diversity and ii) sector-neutral allocation as a risk management tool. We will discuss these in more detail in this section. 2.1 Stock selection In the academic literature, stock selection is carried out on the whole universe with sometimes the exclusion of financial companies, as in Fama and French (1993) and Novy-Marx (2013), or treated separately in two different selections (see Novy-Marx (2012)). It is well known that some risk factors have persistent biases towards some sectors, like the Low Volatility or Value factors, or like Momentum, which can be very concentrated when a few sectors outperform or underperform the stock market. A first approach to control sector risk is to avoid concentration in a few sectors by maintaining a proper level of sector diversity through an appropriate stock selection process. There are two very similar approaches used in the academic literature to tackle this issue. The first one consists in selecting stocks on the whole universe but where stock scores are adjusted by the mean of the score of the sector, as in Novy-Marx (2012). The second approach consists in selecting stocks directly in sectors as in Moskowitz et al. (2002), Banko et al. (2006) and Asness et al. (2014). These two approaches are very similar to maintain proper sector diversity, but we favour the latter one, which is more direct and guarantees the same relative number of stocks in each sector. Another important argument for carrying out stock selection by taking into account sectors is linked to the comparability of accounting quantities. Indeed, accounting quantities can be affected by sector specificities that can make them difficult to compare across sectors. This, in turn, might lead to concentration in a few stocks and thus to a lack of sector diversity in the stock selection. This is especially the case for the Value, Low Investment and High Profitability risk factors, which are all based on accounting measures. Financial leverage, intangibles and off-balance sheet items are the three main elements that can make accounting quantities difficult to compare across sectors. Directly adjusting the accounting measures is an alternative to sector-relative stock selection, but we will argue that accounting adjustments are not desirable, based on a range of examples. A first example is financial leverage. Financial leverage, i.e. the financing of assets through debt, can make stocks difficult to compare, especially with regard to financial stocks versus other stocks. It is well known that value firms are typically high-leverage firms. Many practitioners adjust the book-to-market for leverage and include only those value firms that are not highly levered. However, the relevance of such adjustments lacks academic consensus, and the proposed adjustments will lead to greater dependency on accounting information needed to make the adjustments. A second example is that it is often suggested that the accounting treatment of certain off-balance sheet quantities, like leases, R&D expenditure, or intangible assets such as 'goodwill,' might not capture the true value of the assets of the firm if not taken into account, leading to less robust value stock selection. Given the dematerialisation of the economy, this poor integration can lead to poor representation of information technology or biotechnology companies in factor selections, for which the factor proxy could be strongly impacted by these accounting discrepancies. Concerning intangible assets, the academic literature questions the validity of several well-known adjustments and suggests that the simple book-to-market variable is a more robust measure of value. For off-balance sheet items like leases, it has been shown by Damodaran (1999) that the book value of equity remains unaffected by leases as both assets and liabilities change by the same amount. Finally, R&D expenditure is known to capture intangible asset creation (see Gu (2016), Franzen et al. (2007), Chan, Lakonishok and Sougiannis (2001) and Lev and Sougiannis (1996, 1999)). The value of R&D capital is hard to determine and depreciation rates are hard to justify. Moreover, the major industry that is affected is 'Technology' where typically R&D expenditure is significant and their accounting treatment is different compared to other sectors. From both the academic and practitioner literature on the adjustment of accounting quantities, we can conclude that there is no consensus on their nature and that the results relating to these adjustments are highly sample-dependent. It seems therefore obvious that stock selection within sectors is a more robust approach and is more consistent with the academic literature. At Scientific Beta, we already carry out the stock selection for the Value, Low Investment and Profitability factors on our standard indices (with no control of sector risk) with a stock selection that takes into account three mega sectors (non-financial/non-tech, non-financial/tech, and financial) to deal with the comparability of accounting measures. The sector-neutrality option will lead to an even more stringent approach, where the stock selection is conducted separately in each sector. 2.2 Sector-neutral allocation The objective of the sector-neutral allocation is to have direct control over sector risk by rescaling the weights of the smart factor to obtain the same sector allocation as the CW index. This is clearly a risk management approach since the goal is to reduce the distance between the sector weights of the smart factor index and the CW sector weights in order to be less dependent on specific short-term sector shocks. Novy-Marx (2013), in a long/short framework, uses industry indices to make sure that each long stock position is hedged out for industry exposure by taking an offsetting position in the corresponding stock’s value-weighted industry portfolio. In a long-only setting, the equivalent way of hedging out sector risk is to ensure a sector-neutral allocation between the smart factor sector weights and the corresponding CW weights by rescaling the stock weights of the smart factor index. Asness et al. (2014) construct an industry-neutral low-risk factor (BAB) in a long/short framework where each industry is market-beta-neutral ex-ante. This is an alternative to sector-neutral allocation that makes sense for the low-risk factor because the long and short leg have significantly different level of risks. But the problem of this approach is that it involves the use of leverage, which is difficult to implement for many institutional investors in a long-only framework, therefore we do not favour it. 2.3 Sector neutral indices Overall, stock selection within sectors and sector-neutral allocation put together allow the construction of sector-neutral smart factor indices that maintain proper sector diversity and hedge out most of the sector risk. It nevertheless raises the question of whether these sector-neutral risk factors can still deliver a significant risk premium over the long run. We found evidence in the academic literature that sector-neutral risk factors do still provide significant long-term risk-adjusted performance as seen in Banko et al. (2006), Novy-Marx (2013) and Asness et al. (2014). At Scientific Beta, we offer sector-neutral indices that incorporate both approaches discussed above. Indeed, we carry out the stock selection within each of the 10 sectors defined as the Economic Sectors of the TRBC classification as well as the rescaling of weights versus the CW index in order to achieve sector-neutral allocation. Note that sector-neutral allocation is not perfectly achieved since we apply turnover and liquidity rules after the weight rescaling, which are applied to maintain a high level of replicability of our indices. In the next sections, the sector-neutral smart factor indices we present are based on this technology. 3. Impact of sector risk on absolute performance Sector neutrality reduces the distance between the allocation of the smart factors and the CW index. This is in total opposition to the objective of smart factor investing, which consists in having the highest distance to the CW index allocation to benefit from two sources of added value, which are long-term exposures to rewarded risk factors that are not present in the CW index, as well as good diversification of specific risks. This implies allocations that are very different to those proposed by the CW index, which is concentrated in stocks with the highest capitalisations and does not take stock correlations into account. Therefore we expect that sector-neutral smart factors will generate lower Sharpe ratios than their standard counterparts, with an increase in volatility. Exhibit 2: Comparative annual statistics and sector deviation between standard and sector-neutral EDHEC-Risk Long-Term US HFI Diversified Multi-Strategy (4 Strategy) Smart Factor indices The analysis is based on daily total returns in USD from 31-Dec-1976 to 31-Dec-2016 In Exhibit 2, we show the absolute statistics of the EDHEC-Risk Long-Term US HFI Diversified Multi-Strategy (4-Strategy) Smart Factor indices for the six well-known academic risk factors as well as the multi-beta index that incorporates all risk factors on an equally-weighted basis. We also show the same smart factor indices, but with the sector-neutral-control option. Standard smart factor indices exhibit strong outperformance over the CW with lower volatilities and higher returns resulting in Sharpe ratios being from 60% to 80% higher. On the other hand, sector-neutral smart factor indices, as expected, suffer from higher volatilities compared to their standard counterparts, resulting in lowered Sharpe ratios, with the exception of Mid-Cap, which exhibits a slightly higher Sharpe ratio with sector-neutrality control. Volatilities are increased by an average of 6%, ranging from 1% for High Profitability to 10% for Low Volatility and are statistically different from standard indices. Returns are fairly similar. We observe that they are higher for Mid-Cap and Value with sector neutrality but slightly lower for the other smart factors. Overall, sector neutrality strongly decreases the Distance from CW measure by an average of 60%. This reduction is an indicator of the potential loss of the added value that a smart factor index can provide in terms of improved risk-adjusted performance. Indeed, the closer the Distance from CW measure is to zero, the lower the potential to improve risk-adjusted performance compared to the CW index. Thus investors who seek high value-add potential with their smart beta strategies, and thus require a clear deviation or "active share," might not find sector neutrality constraints suitable. 4. Impact of sector risk on factor exposures Because it forces their allocations to be closer to the CW index, the factor exposures of smart factors are affected in the same way. Since the sector-neutral smart factor has closer allocations to the CW index, it will have a smaller loading on its desired tilt. This implies that sector neutrality will automatically decrease factor intensity. On the positive side, the market beta exposure should be improved for the same reason and we should expect sector-neutral smart factors to have market beta closer to 1, which provides them with better conditional performance and a reduction in relative risk associated with the market beta gap that exists between the CW index and smart factor indices. Exhibit 3: Factor exposures of standard and sector-neutral EDHEC-Risk Long-Term US HFI Diversified Multi-Strategy (4 Strategy) Smart Factor indices The analysis is based on daily total returns in USD from 31-Dec-1976 to 31-Dec-2016 In Exhibit 3, we show the factor exposures of the standard and sector-neutral versions of EDHEC-Risk Long-Term US HFI Multi-Strategy Smart Factor indices. Several comments can be drawn from this exhibit. First, we note that the market beta exposures of standard smart factor indices are lower than one. This is a typical defensive characteristic of smart factor indices and is one of the implicit risks documented by Shirbini (2018), but they are increased when using the sector-neutrality option, with a notable difference for Low Volatility. Next, we examine the exposure of smart factors to their desired factor tilt. For standard indices, Mid-Cap has the lowest exposure with 0.17, whereas Value as the highest with 0.27. Sector-neutral smart factors exhibit lower factor exposure to their desired factor tilt with the exception of Mid-Cap. Low Investment exposure is decreased by 20%, from 0.19 to 0.15. Value exposure falls by 25%, from 0.27 to 0.2. For Momentum, its exposure is decreased by 30% from 0.21 to 0.14. Finally, for Low Volatility and High Profitability, their exposures to their desired tilts fall by 50%. Remember that these two smart factors have the highest Distance to CW measures in their standard version. The case of the Mid-Cap smart factor is interesting, since its exposure to its desired tilt is slightly higher with sector neutrality, but we note a considerable reduction in exposures to the other factors, leading to lower factor intensity. There is also another notable difference in the overall factor intensity between the standard and sector-neutral versions of the smart factor indices. The factor intensity of the standard smart factor indices is the highest for Value with 0.73 and lowest for High Profitability with 0.49. In the case of the sector-neutral versions, factor intensity is reduced for all smart factors by an average of 28%. The reduction is stronger for High Momentum, with a decrease of 38%. Low Volatility and High Profitability have reductions that are close to the average despite the considerable decrease in the exposure to their desired tilts. This reduction in factor intensity is explained by the closer allocation of the sector neutral smart factors to the CW index, which implies that returns are more explained by the market beta than risk factors.. 5. Impact of sector risk on relative risk and performance For now, we focused on the impact of sector neutrality on smart factor indices in terms of absolute risk-adjusted returns and conclude that they were negatively impacted because of the loss of factor intensity. But as mentioned before, sector neutrality is also a way of controlling for relative risk. Sector-neutral smart factors should therefore have lower tracking errors. In Exhibit 5, we show the relative statistics of the EDHEC-Risk Long-Term US HFI Diversified Multi-Strategy (4-Strategy) standard and sector-neutral smart factor indices. We first observe that standard smart factor indices have relative returns that are fairly similar, in the range of 2.5% to 3.5%, whereas the tracking errors are much more scattered, ranging from 4.9% for High Profitability to 7% for Low Volatility. Information ratios are therefore strongly positive, with 0.67 for High Momentum and 0.41 for Low Volatility. Exhibit 5: Comparative relative statistics between standard and sector-neutral EDHEC-Risk Long-Term US HFI Smart Factor indices The analysis is based on daily total returns in USD from 30-Jun-2004 to 30-Apr-2018 As expected, sector-neutral smart factor indices exhibit lower tracking errors. The reduction ranges from -4% for the Mid-Cap to -32% for the Low Volatility smart factors compared to their standard counterparts. Moreover, we show that these reductions are statistically significant at the 1% level, which clearly indicates that sector neutrality is definitely helpful to decrease relative risks. We highlight that the Maximum Relative Drawdown measure is also reduced with sector neutrality by an average of 30%, with 48% for Mid-Cap and 17% for High Momentum. An interesting measure of extreme risk is the 10-year rolling worst 5% tracking error, which is also reduced by an average of 23% across the different smart factors. Relative returns are very similar but most importantly we observe that information ratios are higher compared to standard smart factors, but these differences are not statistically significant. The probability of outperformance at a 1, 3 and 5-year horizon is increased for Mid-Cap, Value, High Momentum and the Multi-Beta Multi-Strategy indices. Overall, we can conclude that sector neutrality is important for controlling relative risks, since the decreases in tracking errors are statistically significant, but when we look at relative risk-adjusted performance, there is no real impact and information ratios are fairly similar. Now we turn to a very important aspect of sector neutrality, i.e. the better control of short-term risk. We will show through three different examples that sector risks can lead to significant short-term underperformance that sector neutrality can help reduce. For the purpose of illustration, we will focus on two smart factors that are known to have persistent over/under exposures to some specific sectors. For the first example, we show the performance of the Low Volatility smart factor index, which is known to be underexposed to the technology sector. In 1999, the sector outperformed the CW index by more than 77%, as seen in Exhibit 6, whereas the underexposure of the smart factor was close to -19% on average. The standard Low Volatility smart factor posted a negative performance of -2.8% and underperformed the CW index in 1999 by -26%, whereas it sector-neutral version posted a positive performance of 6.4% and underperformed the CW index by -16%. It is also important to highlight that the maximum relative drawdown, which measures the maximum relative loss of a strategy compared to its benchmark, is reduced by more than 50%. The annual tracking error of the sector-neutral smart factor is also reduced compared to the standard version. Exhibit 6: Year 1999 absolute and relative statistics of EDHEC-Risk Long-Term US Broad and Technology CW index and EDHEC-Risk Long-Term US HFI Low Volatility Diversified Multi-Strategy (4 Strategy) standard and sector-neutral Smart Factor indices The analysis is based on daily total returns in USD from 31-Dec-1998 to 31-Dec-1999 For the second example, we show the performance of the Value smart factor index, which has a persistent overexposure to the utilities sector. In 2015, the sector underperformed the CW index by -6.7%, as seen in Exhibit 7, whereas the overexposure of the smart factor index was close to +7%. The standard Value smart factor underperformed the CW index in 2012 by -3.5%, whereas its sector-neutral version underperformed the CW index by -2.4%, a reduction of more than 30%. The maximum relative drawdown and the annual tracking error of the sector-neutral smart factor index are also reduced compared to the standard version. Exhibit 7: Year 2015 absolute and relative statistics of EDHEC-Risk Long-Term US Broad and Utilities CW index and EDHEC-Risk Long-Term US HFI Value standard and sector-neutral Smart Factor indices The analysis is based on daily total returns in USD from 31-Dec-2014 to 31-Dec-2015. For the last example, we show that the use of the sector-neutrality option in 2017 and 2018 would have been beneficial to investors, because those two years were marked by outstanding performance of the technology sector. We observe in Exhibit 8 that the technology sector index outperformed the SciBeta US CW index by 19% from the beginning of 2017 to the end of June 2018. The standard SciBeta US HFI Diversified MBMA 6-Factor 4-Strategy EW, which was underexposed to the sector by an average of -13%, underperformed the CW index by -2%, whereas its sector-neutral version outperformed it by +0.5%. Exhibit 8: Relative performance of the standard and sector neutral SciBeta US HFI Diversified Multi-Beta Multi-Strategy (MBMS) 6-Factor 4-Strategy EW The analysis is based on daily total returns in USD from 31-Dec-2016 to 30-Jun-2018 It is clear from the different examples we provide that sectors can have a strong impact on the short-term performance and relative risks of smart factor indices, even if these effects should not be persistent over the long term, since sector risk is not priced in the cross-section of expected returns. It is important however for investors to understand these implicit risks and the different ways of dealing with them and their implications. 5. Conclusion The objective of smart factor investing is to obtain the highest distance to the CW index allocation to benefit from two sources of added value, which are long-term exposures to rewarded risk factors and good diversification of specific risks. This is efficiently achieved through our standard smart factor offering which deliver significant long-term risk-adjusted performance. Nevertheless, we showed in this article that sector risk, which is one of the several implicit risks taken when investing in smart factor indices, can be very high. Indeed, deviations from the CW sector weights can be important and very persistent through time and this can lead to short-term underperformance for investors that might be undesirable. Investors looking to manage these short-term risks can use the sector-neutral risk control option offered on Scientific Beta indices. Using the sector-neutral risk option has a clear advantage in terms of relative risk-adjusted performance since information ratios are increased. We also showed that it is an effective approach to reducing short-term relative losses since it significantly reduces tracking error as well as extreme relative statistics such as maximum relative drawdown and extreme tracking error. Nevertheless, sector neutrality comes with costs. Indeed the reduction of the distance from CW weights implies two very important consequences for long-term risk-adjusted performance. These consequences are shown in Exhibit 9, which reports the percentage change of key metrics when moving from the standard to the sector-neutral version of a multi smart factor index. First, there is a clear reduction in factor intensity. Since exposures to risk factors are the key drivers of performance in smart factor investing, this may be undesirable for some investors. Second, there is a considerable increase in the level of absolute risk (volatility). These two elements lead to a reduction in the long-term absolute risk-adjusted performance (Sharpe ratio) of sector-neutral smart factor indices compared to their standard counterparts. These trade-offs imply that the choice of the most suitable version of the index depends on investor preferences. Exhibit 9: Costs of sector-neutral smart factor indices compared to their standard counterparts in terms of factor intensity, returns, relative returns, volatility and Sharpe ratio The analysis is based on total returns in USD from 31-Dec-1976 to 31-Dec-2016 To provide further analysis of the difference between the standard index and its sector-neutral version, we conduct an analysis of the return difference over time. Exhibit 10 shows the 3-year and 10-year rolling outperformance of the standard EDHEC-Risk Long-Term US HFI Diversified Multi-Beta Multi-Strategy 6-Factor 4-Strategy EW index versus its sector-neutral counterparts. The probability of outperformance of the standard index is 64% for the 3-year rolling window and 60% for the 10-year rolling window. This is consistent with the considerable reduction in factor intensity observed with sector neutrality and shows the superiority of the standard smart factors over sector-neutral ones in producing higher returns over the long run. It is also clear from Exhibit 6 that return differences between the two versions are relatively minor over a ten-year horizon, but can be pronounced over a three-year horizon. We can see for example that the period of the late 1990s was marked by pronounced underperformance of the standard version over its sector-neutral counterpart, as sector deviations were a drag on performance. However, in the following years, sector neutrality became a drag on performance with the standard index posting its most pronounced outperformance over the sector-neutral version. For such shorter horizons, the impact of a choice concerning sector neutrality is considerable. Exhibit 10: 3-year and 10-year rolling outperformance of standard versus sector-neutral EDHEC-Risk Long-Term US HFI Diversified Multi-Beta Multi-Strategy 6-Factor 4-Strategy EW Smart Factor indices The analysis is based on 3-year and 10-year rolling weekly total returns in USD from 31-Dec-1976 to 31-Dec-2016 Overall, it is a fiduciary choice for investors to decide to use the sector-neutral option or not. This choice is a trade-off between their aversion to short-term risks generated by sector risk embedded in standard smart factor indices, which can lead to short-term losses relative to the CW index, and their willingness to harvest factor risk premia in the most efficient way and to achieve the highest risk-adjusted performance over the long run. In terms of risk budgeting, it also involves investors knowing whether they want to minimise absolute or relative risks and this choice should probably be considered from a broader perspective than a simple investment in equity factors, i.e. in a risk management framework for all of their asset classes. For ERI Scientific Beta, in line with its status of index provider, its single and multi-factor index offering, including the sector-neutrality option or not, offers investors and their asset managers the possibility to exercise this fiduciary option. Footnotes: 1See in particular Exhibit 5 in Amenc and Goltz (2013) which analyses the performance of a smart beta index with and without use of the sector neutrality option.
(RI/USD)
Momentum
Volatility
Profitability
Investment
(RI/USD)
Momentum
Volatility
Profitability
Investment
(RI/USD)
Momentum
Volatility
Profitability
Investment
P-value
Information Ratio difference
Diversification within an Equity Factor-Based Framework
This article by Silvio Corgiat Mecio, Senior Portfolio Manager – Portfolio Solutions, Aniket Das, Senior Investment Strategist – Index & Factor-Based Investing and Andrzej Pioch, Fund Manager – Asset Allocation at Legal & General Investment Management, points out that an understanding of the design choices underlying multi-factor products is crucial if investors are to avoid outcomes that may ultimately disappoint them. These design choices include: factor selection, starting universe, multi-factor construction approach, stock weighting scheme, factor weights, regional allocation and currency exposure. Using evidence and beliefs, the authors outline a 'blank-sheet-of-paper' approach to design a particular strategy that places a heavy emphasis on diversification at the factor, region, sector and stock level. This leads to considered objectives for portfolio return, risk and diversification which are able to be clearly messaged to investors.
As factor-based investing has increased in popularity since the financial crisis, so has the number of products available for investors to choose from. Underlying each of these products is a set of design choices whether they are explicitly or implicitly made. For investors we believe it is critically important that they understand these design choices in order to assess how a strategy is likely to perform in the different environments it will invariably face.
We outline a multi-factor equity strategy where a 'blank-sheet-of paper' approach is taken to the following design choices which are explicitly considered and incorporated within the end strategy:
- Factor selection;
- Starting universe;
- Multi-factor construction approach;
- Stock weighting scheme;
- Factor weights; and
- Regional weights and currency exposure.
Having explicit consideration of these areas facilitates clear messaging to investors with respect to the relevant objectives for the strategy and its characteristics. Underlying our strategy's philosophy is a belief in the power of diversification which can be shown not only to reduce risk but to improve geometric returns1. Diversification can occur at different levels and is pivotal to the strategy's construction.
Below we highlight the design choices that we make in designing the strategy and elaborate on the evidence and beliefs that underpin these choices.
Factor Selection
Through our research and investment experience we have developed beliefs on the merits of different factors. While a vast number of factors are documented in academic literature, with over 300 found in one study2, there are relatively few that have an established body of academic research associated with them. Those that we find to be covered more consistently include: value, low volatility, quality, momentum and size.
These correspond closely to the ERI Scientific Beta range of factors available. The key distinction to be made is with regard to the quality factor. ERI Scientific Beta considers quality to be composed of two separate and distinct factors, namely high profitability and low investment, which is in line with Fama and French (2014). We agree with this assessment though we are cautious in giving these two factors as much weight as more established factors. High profitability and low investment have only been published in the academic literature in this century while factors such as value, momentum, size and low volatility all have papers associated with them from the previous century. As such, our confidence in high profitability and low investment is reflected through an adjustment such that each receives half-weight. In essence, these two factors equally-weighted combine to form a single 'quality' factor.
Additionally, we have a prior belief that cross-sectional momentum (ie, momentum at the stock level) is difficult to capture through regularly-rebalanced indices and may induce additional turnover without significant additional benefit, particularly within a multi-factor framework. Published papers by Koraczyk and Sadka (2004) and more recently Novy-Marx and Velikov (2016) support the belief of limited capacity for momentum strategies prior to alpha erosion though a working paper by Frazzini, Israel and Moskowitz (2015) challenges this wisdom. However, this latter paper uses a proprietary dataset which cannot be scrutinised. Where momentum is to be used in portfolios, our general preference is to target time-series momentum involving futures contracts rather than cross-sectional momentum involving individual stocks with the aim to reduce the transaction costs of trading momentum (and indeed Pedersen, Moskowitz and Ooi [2012] present evidence of time-series momentum's ability to completely explain cross-sectional momentum in equities). In our testing we retain momentum as a possible factor for consideration though it faces a higher hurdle for inclusion based on the prior belief.
As we will note later, the Scientific Beta High Factor Intensity (HFI) indices, which we have chosen to use, incorporate a filter which removes stocks with poor multi-factor scores. Momentum is an input into the multi-factor score which means that stocks that score poorly on momentum, all else equal, are more likely to be filtered out. We feel that by removing stocks with poor momentum rather than focusing on stocks with good momentum, this enables us to incorporate the factor in an efficient way.
Our aim in factor selection is to have enough factors such that factor diversification is effective though crucially we must have a high level of belief in these factors.
Starting Universe
Given our aim is to construct a global multi-factor equity strategy, the key question with respect to the starting universe is whether to include emerging markets or restrict the choice to developed markets where there is already a significant body of research on the existence of factors. We find evidence of all the main factors above working in emerging markets as listed in figure 13 and as such we include this region within our universe. This improves our ability to diversify across the markets of more countries, many of which are less co-integrated with developed markets.
Figure 1: Factor Premiums
Factor (long/short)
|
Sample
|
Period
|
Premium
|
Source
|
Value |
Emerging Markets |
1990-2011 |
1.15% (Monthly Mean) |
Cakici, Fabozzi and Tan (2013) |
Momentum |
Emerging Markets |
1990-2011 |
0.86% (Monthly Mean) |
Cakici, Fabozzi and Tan (2013) |
Size |
Emerging Markets |
1990-2011 |
0.28% (Monthly Mean) |
Cakici, Fabozzi and Tan (2013) |
Low Volatility |
Emerging Markets |
1999-2012 |
2.10% (Annual Mean) |
Blitz, Pang and Van Vliet (2013) |
Low Investment |
Emerging and Developed Markets |
1982-2010 |
6.18% (Annual Mean) |
Watanabe, Xu, Yao and Yu (2013) |
High Profitability |
Emerging Markets (Europe) |
2002-2014 |
0.71% (Monthly Mean) |
Zaremba (2014) |
The table presents the factor premium identified for each of six factors from four studies. The factor premium in each case is defined as the difference in returns between a portfolio that contains the top quintile of stocks exposed to the factor and a portfolio that contains the bottom quintile (ie, the least exposed to the factor). These are effectively long/short factor premiums. Note that for the Watanabe, Xu, Yao and Yu study, portfolios based off top and bottom terciles, quintiles and deciles are used depending on whether there are 30, 50 or 100 stocks respectively for each country-year.
Multi-Factor Construction Approach
A key debate going on within factor investing circles surrounds the issue of multi-factor portfolio construction: whether to go 'top-down' or 'bottom-up' with the factor exposures4. The top-down approach allocates to factors as individual building blocks. For example, a top-down multi-factor strategy might have allocations to a value portfolio, a quality portfolio and a low volatility portfolio (where each of these portfolios contains stocks that score strongly on their respective characteristics).
In contrast, the bottom-up approach to multi-factor investing gives each stock in the universe a score on each of the desired factors. These individual factor scores are then combined into an overall multi-factor score for each security in the universe. This composite score is then used to derive a weight in the multi-factor portfolio. There is variation in methodology across different bottom-up strategies.
Additionally, ERI Scientific Beta in 2017 introduced its new range of High Factor Intensity (HFI) multi-factor indices which retain the overall top-down structure used within its original Multi-Beta Multi-Strategy range of multi-factor indices though it adds a bottom-up filtering process applied within each factor sleeve, as described in Amenc et al (2017). We see this as a way of effectively synthesising the top-down and bottom-up approaches. It preserves the simplicity and transparency of the top-down approach but accounts for the cross-factor interaction which, until now, has only been captured by bottom-up approaches. The key difference from other providers we find is in its use of the bottom-up part of the process. Here ERI Scientific Beta focuses on eliminating stocks with poor multi-factor scores rather than adding weight to stocks with strong multi-factor scores – a method which is prevalent among most pure bottom-up approaches.
In order to understand the difference between the bottom-up approach, the top-down approach and the two step-filtering approach present in the HFI methodology (from here on referred to as the 'combined' approach), we conduct our own independent empirical research to understand strategy characteristics. One of the key challenges in comparing approaches across index providers is due to differences in factor definitions, stock weighting schemes or stock universes, among others. Hence it is important to construct strategies using a uniform set of inputs apart from the multi-factor construction approach in order to create a true 'apples-to-apples' comparison.
Figures 2, 3 and 4: Risk/Return Charts
The risk/return charts above show the results of beta-adjusted active return and beta-adjusted tracking error for various multi-factor portfolios using monthly returns. These portfolios are composed of: (1) different multi-factor construction approaches, (2) different stock weighting schemes and (3) different sector/region neutrality constraints. The multi-factor construction approaches include: (i) a top-down approach based on top x% selection within each factor sleeve where x = 15, 30 and 50, (ii) three different bottom-up approaches including geometric S-score multi-factor scores, arithmetic S-score multi-factor scores and average factor rank multi-factor scores*, all based on top x% selection for the overall portfolio where x = 15, 30 and 50 and (iii) the combined approach which uses a top 50% initial selection per the top-down approach in (i) followed by a top 60% selection based on average factor rank multi-factor scores within each factor sleeve. The two stock weighting schemes tested are capitalisation-weighting and equal-weighting. Portfolios incorporating region-neutrality, sector-neutrality as well as region and sector-neutrality are examined alongside the unconstrained version. This leads to 104 different multi-factor portfolios being formed (13 multi-factor construction approaches × two weighting schemes × four portfolio constraint options). We note though that our list of approaches tested is far from exhaustive and indeed only scratches the surface with some of the most common found within the industry. Our dataset uses information for a global universe of stocks (including developed and emerging market stocks) between March 2002 and December 2016. The benchmark is a cap-weighted portfolio including all stocks in the universe. The factors we include are value (book-to-price), low volatility (based on oneyear daily returns), momentum (last 12 months return omitting the most recent month) and quality (which is an equally-weighted combination of high profitability – gross profits-to-assets definition - and low investment – based on three-year asset growth). These four factors (value, low volatility, momentum and quality) are given equal weight. We form portfolios that are semi-annually rebalanced at the end of February and the end of August.
When examining results, two measures of risk-adjusted return we consider are beta-adjusted information ratio (ie, beta-adjusted active return over beta-adjusted tracking error5) and Sharpe ratio. We find that the combined approach which features in the HFI methodology stacks up well against the various other methodologies. The combined approach with region-neutral formation and equal-weighting, which maps closest to the methodology in HFI indices, has a beta-adjusted information ratio in the top decile and a Sharpe ratio in the third decile. However we would de-emphasise the importance of this empirical testing. We see the testing as validating our belief that the combined approach is an efficient implementation rather than it driving our decision.
Bottom-up strategies are seen to carry higher beta-adjusted tracking errors which are generally commensurate with higher beta-adjusted returns though this link weakens for concentrated approaches (ie, top 15% selection6). Overall on this risk-adjusted return measure we find that there is generally a linear relationship between risk and return for all strategies except those with particularly high beta-adjusted tracking errors (which mostly correspond with concentrated portfolios).
In terms of stock weighting schemes, we find that equal-weighted strategies dominate cap-weighted strategies with this being robust to examining time periods when the size factor produced a zero return. This would indicate that the performance of the size factor may not have been the only driver of the performance differential but could be due to the effects of better diversification and lower stock-specific risk for equal-weighted strategies.
Also, we notice that bottom-up strategies have tended to carry a low beta bias over the time period. This finding is similar to that of Jivraj et al (2016) who also look at multi-factor construction approaches that include the low volatility factor for a US universe between January 2003 and July 2016.
Figure 5: Portfolio CAPM Beta
The figure above shows the portfolio CAPM beta for the 104 different portfolios referred to in figures 2, 3 and 4 split by multi-factor construction approach.. The CAPM beta is defined as the slope of the regression of portfolio returns on the market factor.
Additionally, we find that while region-neutrality (formed using 11 regional building blocks akin to the approach taken by ERI Scientific Beta in its methodology) leads typically to improvements in beta-adjusted information ratio relative to a global approach to stock selection, this also typically leads to small declines in Sharpe ratio. We find little support for either sector neutrality or region-and-sector neutrality on a performance basis where sector neutrality is achieved through a re-scaling process back to market-cap sector weights (at the global level for sector-neutral and within region for region-and-sector neutral). We also note the higher turnover of these strategies, particularly for region-and-sector neutrality.
Figure 6: Beta-Adjusted Information Ratio: Equally-Weighted Portfolios
The figure above plots the beta-adjusted information ratio for the 52 equally-weighted portfolios referred to in figures 2, 3 and 4 across 13 different multi-factor construction approaches and four portfolio constraint options.
While we do not believe that the time period used in our research is long enough to make definitive conclusions, we would argue that it still provides some level of insight. We acknowledge the results in Amenc et al, who use US stock data over the period 1975 to 2015 to confirm the robustness of the Scientific Beta HFI approach relative to a concentrated bottom-up approach. Similarly, Leippold and Rüegg (2017), who use US stock data from 1963 onward, find a similar pattern as us with regard to the low beta bias of bottom-up strategies that include the low volatility factor while also finding similar levels of risk-adjusted return between top-down and bottom-up approaches.
Overall, having undertaken the independent research above, our results seem to favour the combined approach which aligns with the methodology within ERI Scientific Beta HFI indices. This validates our belief that the combined approach is an efficient way of integrating bottom-up and top-down approaches. As such, we decide to use indices within this range to implement our multi-factor strategy.
Stock Weighting Scheme
When considering how to weight individual stocks after the stock selection process, our goal is to seek diversification such that stock-specific risk is reduced. When deciding what weight to give assets, we like to consider both capital weights as well as risk weights. Diversifying by capital weights corresponds to the ERI Scientific Beta maximum deconcentration stock weighting scheme while diversifying by risk weights corresponds to the ERI Scientific Beta diversified risk weighted stock weighting scheme. As such we use an equal-weighted combination of these two weighting schemes. This leads to a significant reduction in stock-level concentration relative to cap-weighted indices7, which is a common aim for many investors.
By seeking this diversified stock weighting scheme, we note that we also by proxy achieve more diversified sector weights. This has the effect of reducing the influence of the largest sectors which could be susceptible to periods of over-valuation8.
Factor Weights
Our primary objective when setting factor weights is to seek diversified factor exposure. This means that we want to ensure that we are carrying significantly positive and relatively balanced exposures to the factors we are targeting through the economic cycle. Our starting point is to test equal factor weights and if then there is a need to deviate from this position we would do so. However, as the ERI Scientific Beta HFI methodology explicitly accounts for cross-factor interactions, we expect factor balance to naturally occur.
Additionally, as we are opting for a diversified stock weighting scheme, we recognise that this introduces a significant amount of size factor exposure itself and an explicit allocation to the factor would lead to an imbalance that would go against our objective. We test different factor weights over two time periods: for a US stock universe from 1975 to 2015 and for a global stock universe from 2002 to 2016. We see broadly similar overall return and risk statistics across different combinations of factors including those that held momentum and size factors though critically there are differences when it comes to factor exposures.
We notice a heavy imbalance in factor exposures when including the size factor which matches our initial intuition. When looking at the momentum factor exposure over the US long-term backtest, while still being statistically significant, it is markedly lower in magnitude relative to the other factors (considering high profitability and low investment as a single factor). This means that possibly momentum has indeed been more difficult to capture over the long-term. This gives us sufficient reason to exclude momentum as we have not seen much evidence to challenge our prior belief.
As a result of excluding explicit allocations to momentum and size, we are left with equal weighting value, low volatility and quality factors (where quality itself is an equal-weighted combination of high profitability and low investment). We note that we achieve very good factor balance across the factors targeted in the US long-term backtest. The balance is not as good in the shorter-term global backtest, though we feel this to be still quite reasonable. All factor exposures across the two backtests are significant at the 1% level except for momentum in the US long-term backtest. While momentum is positive and significant in the shorter-term global backtest (even though it is not explicitly targeted), we would not expect this to be the case over all time periods given the result of the longer-term US test.
This leaves us with diversified factor exposure, which added to diversification at stock and sector level, all enable us to reduce the risks of particular factors, stocks or sectors performing poorly.
Figure 7: Factor Exposure
The charts above represent a measure of factor exposure for our chosen strategy. These factor exposures are defined as the factor regression coefficients from a seven-factor model that includes the market factor alongside six long/short factor portfolios (value, high profitability, low investment, low volatility, size and momentum). The results for these six factors only are shown. The first chart uses a US universe of stocks between the period December 1975 and December 2015 while the second chart uses a global universe of stocks (including emerging markets) between June 2002 and December 2016. Weekly returns are used in each case. The calculation has been performed by ERI Scientific Beta.
Regional Weights and Currency Exposure
The final layer of diversification we seek is with regard to the strategy's regional weights and currency exposures. Often a multi-factor strategy's regional weights are an artefact of the stock selection process. We believe this risks the introduction of unintended regional bets. An explicit regional allocation process can alleviate the issue.
When deciding upon regional allocations, we split the global equity universe into six distinct regions: UK, developed Europe ex UK, North America, Japan, developed Asia Pacific ex Japan and emerging markets9.
Individual regional weights can be chosen in many alternative ways. In addition to an equally-weighted allocation, two of the most intuitive alternatives include market-cap weighting and GDP weighting.
The market-cap weighting approach is widely adopted in index-based investing due to its straightforward implementation, removing the need to rebalance among regions10. However, we recognise that one of the key aims of investors when considering investments in this area is to avoid links to market-cap weighting as it can re-introduce the sensitivity to company valuations. As valuations of an individual stock or a group of stocks within a region increase, this would drive higher the weight of the region where these stocks are listed. Furthermore, such an approach results in a large concentration in North America, with nearly 60% weight in the region at present.
Nevertheless, while acknowledging its limitations, the market-cap weights of individual regions represent their importance in financial markets and as such they remain a dimension worthy of consideration. However, we believe it should be accounted for in conjunction with the regions' economic significance that is reflected in their GDP. The GDP weighted approach breaks the link between country weightings and market-cap size, hence reducing the sensitivity of regional exposures to changes in market sentiment. Consequently, the weight of larger emerging market economies like China will be higher and the overall regional exposure could be significantly different from the conventional market-cap benchmark.
The regional allocation we choose for our global multi-factor equity strategy aims to reflect both the economic and the financial significance of individual regions to provide a more diversified exposure that is not overly reliant on any single region. To enhance that diversification even further we marginally increase the weight of those regions that are less correlated with the home market. For a UK investor, this would normally mean a positive adjustment to Asia Pacific, including Japan, and emerging markets on a stand-alone basis. For investors based elsewhere, for example in the eurozone, these adjustments would be different. We also ensure that the chosen allocation does not result in a concentrated exposure to the politics of a specific country. That on a stand-alone basis would lead to a reduction in weight of regions such as the UK and Japan, and to a slightly lesser degree, North America. Finally we also consider governance standards to determine whether investors get the returns they earn for taking the equity risk in a particular region. Overall, our approach results in a more balanced regional allocation which further reduces the strategy’s concentration risk (figure 8).
Figure 8: Regional Allocations
The chart above shows the weights of different regions based on: (i) market capitalisation using data from ERI Scientific Beta, (ii) GDP based on data from the World Bank and (iii) our target allocation.
Interestingly we note that by lowering the weight of the North American region we are able to increase stock-level diversification as measured by the effective number of stocks11. Our multi-factor strategy using market-cap regional weights results in an effective number of stocks figure of 657 while the same strategy using our regional weights has a figure of 852 as of December 2016. This increase is possible due to the high average stock weight for the North American universe relative to other regions. By moving weight to other regions, this leads to greater diversification at the stock level as measured by effective number of stocks.
In addition to the regional allocation, for a UK investor, we would hold currency exposures which hedge 50% of the overseas (ie, non-GBP) developed markets currency risk within the strategy. The currency hedge for investors in other regions will depend on the correlation of the home currency with global equity markets. We believe the currency hedge reduces volatility over the long-term while not sacrificing return (see Joiner and Mollan [2017]).
While there is an initial allocation suggested, we believe it is necessary to have an ongoing monitoring process in place to change elements of the strategy in order to be able to continue to deliver on its objectives. Nevertheless, we would expect any changes, including the regional allocation, to be gradually made over time reflecting the strategic rather than tactical nature of the process.
Conclusion
We have demonstrated above the evidence and beliefs that underpin the design choices we make for our multi-factor strategy. Choices were made with regard to the selection of factors, the starting universe, the multi-factor construction approach, factor weights and the stock weighting scheme as well as the regional allocation and currency exposures. Explicit consideration was given to many different elements that can influence outcomes.
This level of understanding also allows us to create well-informed objectives for return and risk that can be messaged to investors. For our strategy, based on its exposure to the targeted factors, we aim to, over the long term, outperform a blend of market-cap indices with a similar regional weighting to ours, at a lower level of volatility.
By seeking diversification at multiple layers including at factor, region, sector and stock level, we have designed a solution which we believe will meet the objectives of many investors who are looking for a strategic, long-only exposure to equity factors delivered in a diversified manner.
The views expressed in this article are those of the authors and do not necessarily represent those of their firm.
This article was published in the Spring 2018 issue of the IPE/EDHEC Research Insights supplement.
Additional articles in the supplement covered the topics of
"Misconceptions and mis-selling in smart beta: Improving the risk conversation in the smart beta space", "Assessing the investability of smart beta indices" and "Debating the merits of 'top-down' and 'bottom-up' approaches to multi-factor index construction".
Appendix: Geometric S-score and Arithmetic S-score
In order to calculate S-Scores for stocks, first we calculate z-scores.
Fj,i is stock i’s attribute value for factor j
μj is the cross-sectional mean (i.e. the average across all stocks) for factor j
σj is the cross-sectional standard deviation (i.e. the average across all stocks) for factor j
zj,i is stock i’s z-score for factor j
We then apply a winsoring process to the z-scores such that values above 3 are set to 3 and that values below -3 are set to -3. The z-score formula above is re-run with the new values and the winsoring process is applied repeatedly until all z-scores in the universe fall between -3 and 3. We use these winsorised z-scores to calculate the S-scores.
Sj,i is stock i’s S-score for factor j
The winsorised z-score is mapped to an S-Score using the cumulative distribution function of the standard normal such that it lies between 0 and 1.
For stock i its geometric S-score (GS) multi-factor score (MFS) across k factors is:
And its arithmetic S-score (AS) multi-factor score across k factors is:
Average Factor Rank
For a universe with n stocks, the attribute rank for stock i on factor j is defined by:
Then the average factor rank (AFR) multi-factor score for stock i with k factors is simply given by:
Footnotes:
1Humble and Southall (2014).
2Harvey et al (2016).
3Adapted from Shirbini (2016).
4See Fitzgibbon et al (2016) and Bender and Wang (2016).
5Beta-adjusted active return is what is known otherwise as 'Jensen's alpha' or 'ex-post alpha' as described in Jensen (1967) using the Capital Asset Pricing Model (CAPM) beta to adjust active returns. Beta-adjusted tracking error is a related statistic equal to the volatility of the beta-adjusted active returns. We prefer the beta-adjusted information ratio over the standard information ratio as it does not automatically penalise strategies with beta less than 1 (which is a desirable characteristic for some investors).
6Note that the combined approach leads to the equivalent of a top 30% selection (ie, a top 50% selection followed by a top 60% selection).
7As an indication of the reduction, the percentage weight in the 20 largest stocks in our strategy in June 2017 was about 5% while this was about 15% for a representative global market-cap equity index.
8For example, the information technology sector during the 'dot-com' bubble came to represent about 30% of the S&P 500 at one point.
9Note that this is similar though different to the 11 ERI Scientific Beta regions (eight developed, three emerging) that were used in our multi-factor construction approach research detailed previously.
10Although rebalancing would still occur within regions as factor data changes.
11The effective number of stocks is defined as the reciprocal of the Herfindahl index, which is a commonly used measure of portfolio concentration:
where N is the number of constituent stocks in the index and wi is the weight of stock i in the index. In brief, the effective number of stocks in a portfolio indicates how many stocks would be needed in an equal-weighted portfolio to obtain the same level of concentration (as measured by the Herfindahl index). Equal-weighting stocks in a portfolio will lead to the maximum effective number of stocks.
"Design Choices in Multi-Factor Investing" Webinar
Legal & General Investment Management / ERI Scientific Beta
A joint webinar featuring Eric Shirbini, Global Research and Investment Solutions Director at ERI Scientific Beta, and Adam Willis, Head of Index and Multi-Asset Distribution, and Aniket Das, Senior Investment Strategist, at Legal & General Investment Management.
This webinar on design choices in multi-factor investing, co-presented by Legal & General Investment Management and ERI Scientific Beta, examines how factor-based investing can best be implemented in an institutional investor's portfolio in a context of increasing investor sophistication, with the principle underlying the webinar being that investors should consider design choices before considering products. Discussing a risk-based approach to multi-factor index construction, the speakers address crucial issues such as factor balance, factor interaction and risk premia diversification.
Interview
Tim Taylor, Senior Investment Officer (SIO), Global Equity, Florida State Board of Administration
Tim Taylor, Senior Investment Officer (SIO), Global Equity, Florida State Board of Administration
In this interview, Tim Taylor, Senior Investment Officer (SIO), Global Equity at Florida State Board of Administration, explains the choice to replicate Scientific Beta indexes, presents the objectives of the allocation and discusses the benefits and future of smart beta investing.
Why did the Florida State Board of Administration decide to replicate the ERI Scientific Beta developed smart factor and multi smart factor indexes?
We believe that securities possessing certain factors have historically outperformed the market, and that going forward investors will continue to reward these factors. By replicating ERI Scientific Beta's factor indexes we obtain cost-efficient exposure to these potential sources of positive alpha. And we further enhance the cost benefits, and increase the alpha potential, by internally managing our factor index strategy.
Could you tell us a little about the objectives of this allocation?
The objective is to outperform the MSCI World Index. The allocation is consistent with our desire to identify value-adding solutions that can be internally managed. The majority of the funds used to launch the strategy were sourced from traditional long-only active managers. We believe that we do a good job negotiating asset management fees with investment managers, however internally managed portfolios may have even greater cost advantages.
What do you consider to be the main advantages of smart beta indexes for investors?
One advantage is that investors have another way to evaluate active management skill. Managers must be able to demonstrate that they are generating positive alpha above and beyond a naively constructed benchmark, and not just the capitalization-weighted benchmark. This may be a difficult task particularly after accounting for the asset management fee. Managers must also be prepared to qualitatively discuss why their approach is preferable to the factor beta index. What value does their judgement bring to the relationship, and how does it distinguish their services from the factor index?
What makes a good factor index in your view?
It must provide you with reproducible and transparent exposure to the source of potential alpha that you are targeting. It should be constructed in a way that is consistent with the theory and evidence that supports building an index around that factor in the first place. It should be clearly defined, investable and cognizant of transaction costs when establishing its rebalancing methodology.
Scientific Beta's smart factor index philosophy involves combining exposure to factors with diversification of specific risk. What is your view of this approach?
Diversification across a large number of stocks is consistent with the evidence related to factor based investing. Scientific Beta's focus on reducing specific risk makes sense from an absolute return perspective and over the long run leads to better performance than the benchmark. However, the factor index methodologies do not take client benchmarks into account and therefore do not address stock specific active risk relative to those client benchmarks. Over short periods, methods that reduce stock specific risk on an absolute basis could increase or decrease tracking error relative to the benchmark.
How do you see the future of smart beta investing?
I believe that investors will increase their allocations to factor index strategies, and that investment firms will continue to create alternatives. As the number of factor indexes proliferate it will become even harder for investors to evaluate their options. Transparency and simplicity will be compromised as options become cloudier and more complex, which will make it more difficult for investors to select an appropriate solution. However many investors will find that they are able to achieve, and even improve, their returns by incorporating factor investing into their investment process.
Timothy E. Taylor: Tim is currently Senior Investment Officer (SIO), Global Equity at the Florida State Board of Administration (SBA). The SBA primarily oversees the investments of the $150 billion Florida Retirement System Trust Fund. He, along with Global Equity's other SIO, is responsible for all matters related to the asset class including personnel, investment structure, and oversight of internal and external portfolio management. Tim graduated from Florida State University (FSU) with a Bachelor of Science in Finance. He began his career in the banking industry in Tampa. Tim relocated to Tallahassee after accepting an opportunity with Florida's 457 Deferred Compensation Program. He helped create materials so that employees could compare investment providers and products, and made educational presentations throughout the state. While working in Deferred Compensation Tim began his post-graduate work at FSU, ultimately graduating with a Master of Business Administration (MBA) degree. Tim joined the SBA as a Portfolio Manager in the International Equity Asset Class and was later promoted to Senior Portfolio Manager. He has been involved in the oversight of external investment managers, and has spearheaded multiple manager searches. Tim was responsible for coverage of Japan in support of an actively managed internal portfolio, for which he also served as head trader. He has evaluated opportunities in the China A-Share market, led transition manager searches and restructuring events, completed research on currency management, and produced equity and foreign currency transaction analysis. Tim was part of the team that implemented a restructuring of the investments of the Global Equity asset class, a $12 billion transition over several months. Tim was promoted to Deputy SIO in February 2015 and subsequently appointed to his current position on January 1, 2016.
Partner News
News of the partnerships and licences established with industry participants to replicate ERI Scientific Beta indices.
ERI Scientific Beta has forged partnerships with asset managers who not only replicate its indices but also propose open funds enabling investors to readily invest in the strategies proposed by ERI Scientific Beta that they have selected. In this issue, we focus on CoreShares, Desjardins Global Asset Management and Legal & General Investment Management.
CoreShares
On 21 June, 2018 in Bryanston, Johannesburg, CoreShares held a "Think Index Investing" convention with a focus on smart beta. The event provided participants with strategic insights and practical advice on issues pertaining to the passive investment management market and ETFs in particular from a panel of both global and local industry leaders. Erik Christiansen, Senior Business Development Director at ERI Scientific Beta, participated as a panelist at the event in the session on "Pulling it all Together".
In October, 2017, CoreShares, a leading passive investment manager in South Africa, launched the CoreShares Scientific Beta Multi-factor Index Fund based on the SciBeta CoreShares South-Africa Multi-Beta Multi-Strategy Six-Factor EW index. The fund's aim is to capture academically-proven investment styles in the market within one holistic index approach. The solution is suitable for investors looking to incorporate a low cost, rules-based, smart beta equity solution into their investment framework with an additional level of diversification by combining multiple factors and therefore smoothing the cyclicality that is associated with individual factors. The key advantages of this smart multi-factor approach are its diversification, not only of factor risk but also of specific risk, and its top-down approach to risk allocation, thus meeting investors' concern to keep risk management at the forefront of their investments.
The Scientific Beta Multi-Beta Multi-Strategy Six-Factor EW Indices correspond to an equal-weighted combination of the most popular choices of factor considered to be well-rewarded in the financial literature (Value, Size, Momentum, Low Volatility, High Profitability and Low Investment). These indices benefit from a double diversification – allocation across different factors (multi-beta allocation) along with diversification of the specific risks of each smart factor index (diversified multi-strategy weighting scheme).
Desjardins Global Asset Management
Desjardins Global Asset Management was a privileged sponsor of the inaugural Inside ETFs Canada conference in Montreal on 21-22 June, 2018, where Eric Shirbini, Global Research and Investment Solutions Director at Scientific Beta also participated in a panel session on the theme "The Hottest Topic in Smart Beta: Single vs Multifactor Strategies".
In 2017, the Desjardins Group launched a multifactor-controlled volatility suite of exchange-traded funds:
- Desjardins Canada Multifactor-Controlled Volatility ETF (DFC)
replicating the Scientific Beta Canada Multi-Factor Controlled Volatility Index - Desjardins USA Multifactor-Controlled Volatility ETF (DFU)
replicating the Scientific Beta United States Multi-Factor Controlled Volatility Index - Desjardins Developed ex-USA ex-Canada Multifactor-Controlled Volatility ETF (DFD)
replicating the Scientific Beta Developed ex-US ex-CA Multi-Factor Controlled Volatility Index - Desjardins Emerging Markets Multifactor-Controlled Volatility ETF (DFE)
replicating the Scientific Beta Emerging Multi-Factor Controlled Volatility Index
The ETFs, listed on the Toronto Stock Exchange, have been designed to select securities with an exposure to six factors documented to add value over the long-term, with an expected level of risk that is lower than the traditional index and to mitigate losses in market downturns and participate in market recoveries.
The Scientific Beta Multi-Factor – Controlled Volatility solutions are custom indices that use a new type of defensive strategy approach based on dynamic allocation between the components of the six smart factor Diversified Multi-Strategy indices (Value, Momentum, Size, Low Volatility, High Profitability, and Low Investment). The aim of this dynamic allocation is to maximise the deconcentration of the defensive index and hence avoid concentration in the Low Volatility factor alone, while at the same time setting a constraint that respects a maximum of 90% of the volatility of the cap-weighted benchmark. This provides protection against strong market volatility whilst avoiding an excessive defensive bias in bull markets and results in improved risk-adjusted performance and better performance conditionality.
Legal & General Investment Management
Legal & General Investment Management and ERI Scientific Beta have co-produced a webinar on the subject of design choices in multi-factor investing. In the webinar, Eric Shirbini, Global Research and Investment Solutions Director at ERI Scientific Beta, joins Adam Willis, Head of Index and Multi-Asset Distribution, and Aniket Das, Senior Investment Strategist, at Legal & General Investment Management, to discuss how factor-based investing can best be implemented in an investor's portfolio in a context of increasing investor sophistication, with the principle underlying the webinar being that investors should consider design choices before considering products. Discussing a risk-based approach to multi-factor index construction, the speakers address crucial issues such as factor balance, factor interaction and risk premia diversification.
In July 2017, Legal & General Investment Management launched a commingled life fund for UK institutional clients based on ERI Scientific Beta’s multi-beta multi-strategy indices. The LGIM Diversified Multi-Factor Equity Fund allocates between the following Scientific Beta indices according to regional weights determined by Legal & General Investment Management:
- SciBeta Developed America High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
- SciBeta United Kingdom High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
- SciBeta Developed EMEA ex-UK High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
- SciBeta Japan High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
- SciBeta Developed Asia Pacific ex-Japan High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
- SciBeta Emerging High-Factor-Intensity Multi-Beta (val, vol, pro/inv) Multi-Strategy (Max Deconc, DRW)
The Scientific Beta High-Factor-Intensity Multi-Beta (vol, val, pro/inv) Multi-Strategy (Max Deconc, DRW) EW Indices are custom indices that provide strong exposure to the rewarded risk factors (Low Volatility, Value, Low Investment and High Profitability) and a good level of diversification using an equal-weighted combination of the Maximum Deconcentration and Diversified Risk Weighted approaches.
Performance
Performance data for ERI Scientific Beta smart factor and multi smart factor indices over different time periods as of 30 June, 2018.
Scientific Beta Smart Factor Indices Performance
ERI Scientific Beta offers smart factor indices that provide exposure to the six well known rewarded factors (Mid-Cap, Value, High Momentum, Low Volatility, High Profitability, and Low Investment) and which are also well diversified in order to reduce the specific risks. These indices are available in a wide range of versions, notably enabling broad and narrow indices to be distinguished that correspond to more or less pronounced choices of factor exposure. In addition, these single smart factor indices can be used in a multi-factor allocation by taking into account the interactions between the indices in order to guarantee very high factor intensity. In this report, we have chosen to present the smart factors represented by the Scientific Beta Narrow High Factor Intensity Diversified Multi-Strategy (4-Strategy) indices.
Performance Overview
The following table displays the short-term, mid-term and long-term performance of Narrow High Factor Intensity Diversified Multi-Strategies (4-Strategy) by factor tilt in the Developed equity universe, which is the global universe managed by Scientific Beta. The six rewarded tilts selected (Mid-Cap, Value, High Momentum, Low Volatility, High Profitability, and Low Investment) are part of the common tilts documented in the literature as liable to produce outperformance compared to cap-weighted indices.
Narrow High Factor Intensity Diversified Multi-Strategy (4-Strategy) Index |
Past Quarter |
Year-to-Date |
1 Year |
5 Years |
10 Years
|
Long-Term US Track Records 31/12/1976 to 31/12/2016 (40 years) |
Annualised Relative Return Compared to Broad Cap-Weighted as of 30/06/2018 |
||||||
Mid-Cap |
-0.77% |
1.23% |
-0.24% |
2.40% |
3.44% |
3.54% |
Value |
-0.52% |
-0.63% |
0.32% |
1.71% |
2.64% |
2.63% |
High Momentum |
-1.36% |
0.50% |
2.97% |
2.65% |
1.37% |
3.02% |
Low Volatility |
-1.38% |
-1.46% |
-4.54% |
0.05% |
2.56% |
2.35% |
High Profitability |
1.37% |
2.58% |
2.85% |
3.16% |
4.52% |
3.27% |
Low Investment |
0.22% |
0.48% |
0.43% |
1.91% |
3.17% |
3.06% |
The history of Scientific Beta Index returns begins on 21/06/2002. The statistics are based on daily total returns (with dividends reinvested). All statistics are annualised, except for the past quarter relative returns and year-to-date relative returns which are non-annualised. The performance ratios that involve the average returns are based on the geometric average, which reliably reflects multiple holding period returns for investors. ERI Scientific Beta uses the yield on Secondary Market US Treasury Bills (3M) as a proxy for the risk-free rate in US Dollars. All results are in USD.
During the second quarter of 2018, the performance of Scientific Beta smart factor indices ranged from -1.38% for the SciBeta Developed Narrow High-Factor-Intensity Low-Volatility Diversified Multi-Strategy (4-Strategy) index to 1.37% for the SciBeta Developed Narrow High-Factor-Intensity High-Profitability Diversified Multi-Strategy (4-Strategy) index compared to broad cap-weighted indices.
Over the past ten years, all strategies posted positive relative returns in relation to broad cap-weighted indices, with values ranging from 1.37% for the SciBeta Developed Narrow High-Factor-Intensity High-Momentum Diversified Multi-Strategy (4-Strategy) index to 4.52% for the SciBeta Developed Narrow High-Factor-Intensity High-Profitability Diversified Multi-Strategy (4-Strategy) index.
Scientific Beta Multi Smart Factor Indices Performance
Scientific Beta Multi-Beta Multi-Strategy (MBMS) indices provide an allocation to well-rewarded smart factor indices. Here again, Scientific Beta proposes a wide range of Multi-Beta Multi-Strategy indices based on the same Smart Beta 2.0 investment philosophy. This section presents those indices that enable the diversification of factor and specific risks to be reconciled. Among these indices, we have chosen to present some of the more popular ones, namely the strategy with the longest live track record – the Scientific Beta Multi-Beta Multi-Strategy Four-Factor Equal-Weight index and a strategy created more recently which takes into account the interactions between single-factor indices in order to provide higher factor intensity at a multi-factor level – represented by the Scientific Beta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy Equal-Weight index and its Sector Neutral and Market Beta Adjusted (Leverage) versions, the latter corresponding to a risk control option where the indices are leveraged by borrowing cash at the overnight rate with an objective of achieving a market beta of one.
Performance Overview
The table below displays an overview of the relative and absolute performance of Scientific Beta Multi-Beta Multi-Strategy indices for the Developed, United States and Global regions over different time periods.
Region |
Multi-Beta
Multi-Strategy Index |
Nº Constituents |
Relative Return Compared to Cap-Weighted |
Information Ratio |
Absolute Return |
Volatility |
Sharpe Ratio |
||
Past Quarter |
YTD |
10 Years |
10 Years |
10 Years |
10 Years |
10 Years |
|||
Developed |
4-Factor EW |
1784 |
-0.92% |
-0.11% |
1.58% |
0.60 |
8.20% |
15.42% |
0.51 |
HFI 6-Factor 4-Strategy EW |
1199 |
-0.54% |
-0.41% |
3.17% |
0.91 |
9.80% |
14.52% |
0.65 |
|
HFI 6-Factor 4-Strategy EW Sector Neutral |
1220 |
-0.05% |
0.83% |
2.78% |
0.97 |
9.40% |
14.99% |
0.61 |
|
HFI 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) |
1199 |
-0.48% |
-0.48% |
4.72% |
1.73 |
11.34% |
17.21% |
0.64 |
|
SciBeta Developed CW |
1896 |
6.62% |
16.94% |
0.37 |
|||||
United States |
4-Factor EW |
474 |
-1.26% |
-0.36% |
0.59% |
0.19 |
10.46% |
18.91% |
0.54 |
HFI 6-Factor 4-Strategy EW |
318 |
-1.01% |
-1.47% |
2.03% |
0.48 |
11.91% |
17.61% |
0.66 |
|
HFI 6-Factor 4-Strategy EW Sector Neutral |
320 |
0.01% |
0.76% |
1.97% |
0.58 |
11.84% |
18.45% |
0.62 |
|
HFI 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) |
318 |
-0.90% |
-1.48% |
3.73% |
0.96 |
13.61% |
20.68% |
0.64 |
|
SciBeta United States CW |
500 |
9.87% |
20.04% |
0.48 |
|||||
Global |
4-Factor EW |
2431 |
-0.77% |
-0.20% |
1.63% |
0.60 |
7.88% |
15.07% |
0.50 |
HFI 6-Factor 4-Strategy EW |
1633 |
-0.41% |
-0.44% |
3.28% |
0.93 |
9.54% |
14.25% |
0.65 |
|
HFI 6-Factor 4-Strategy EW Sector Neutral |
1658 |
-0.03% |
0.69% |
2.86% |
0.99 |
9.12% |
14.73% |
0.60 |
|
HFI 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) |
1633 |
-0.45% |
-0.62% |
4.83% |
1.86 |
11.08% |
16.94% |
0.64 |
|
SciBeta Global CW |
2590 |
6.25% |
16.82% |
0.35 |
|||||
Based on daily total returns in USD as of 30/06/2018. The base date is 21/06/2002 for Scientific Beta Multi-Beta Multi-Strategy 4-Factor EW indices, Scientific Beta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy 6-Factor 4-Strategy EW indices and Scientific Beta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy (Sector Neutral) 6-Factor 4-Strategy EW indices, 18/06/2004 for Scientific Beta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) indices and 19/12/2003 for Scientific Beta Multi-Beta Multi-Strategy CW indices. All statistics are annualised, except for the past quarter relative returns and year-to-date relative returns which are non-annualised. Performance ratios that involve the average returns are based on the geometric average, which reliably reflects multiple holding period returns for investors. The risk-free rates used are defined according to the regional universe of the index. The number of index constituents are as of the last quarterly rebalancing, i.e. 15/06/2018.
During the second quarter of 2018, relative returns varied from -1.26 % for the SciBeta United States Multi-Beta Multi-Strategy Four-Factor EW index to 0.01% for the SciBeta United States High-Factor-Intensity Diversified Multi-Beta Multi-Strategy (Sector Neutral) Six-Factor Four-Strategy EW index. Since the beginning of the year, factor strategies have clearly not been performing well as the implicit sector bets proposed by the factors have a considerable impact on short-term performance. Moreover, it can be observed that the sector neutral versions of the multi smart factor indices are the ones that are resisting best to this under-performance of factors.
Over the past ten years, the SciBeta Developed Multi-Beta Multi-Strategy Four-Factor EW index, the SciBeta Developed High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy EW index, the SciBeta Developed High-Factor-Intensity Diversified Multi-Beta Multi-Strategy (Sector Neutral) Six-Factor Four-Strategy EW index and the SciBeta Developed High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy EW Market Beta Adjusted (Leverage) index posted strong annual relative returns of 1.58%, 3.17%, 2.78% and 4.72% respectively, compared to cap-weighted indices. For the other regions, the highest performance over the past ten years was obtained by the SciBeta Global High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy EW Market Beta Adjusted (Leverage) index with a relative return of 4.83%, compared to cap-weighted indices, with the lowest performance posted by the SciBeta United States Multi-Beta Multi-Strategy Four-Factor EW index at 0.59%.
Over the long-term, all Scientific Beta Multi-Beta Multi-Strategy indices post positive excess returns compared to cap-weighted indices. Using long-term US track records from December 31, 1976 to December 31, 2016 (40 years), the SciBeta Multi-Beta Multi-Strategy Four-Factor EW, the SciBeta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy EW, the SciBeta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Sector Neutral Six-Factor Four-Strategy EW and the SciBeta High-Factor-Intensity Diversified Multi-Beta Multi-Strategy Six-Factor Four-Strategy EW Market Beta Adjusted (Leverage) indices post respective relative returns compared to cap-weighted indices of 2.56%, 3.09%, 2.80% and 4.48%.
The relative and absolute performance data for Scientific Beta regional universes is available here.
The table below reports the live performance of the Scientific Beta Multi-Beta Multi-Strategy indices for the Developed ex USA and United States regions as of 30 June, 2018.
Region |
Multi-Beta
Multi-Strategy Index |
Nº Constituents |
Annualised Relative Return Compared to Cap-Weighted |
Annualised Absolute Return |
Volatility |
Sharpe Ratio |
From 20/12/2013 to 30/06/2018
|
||||||
Developed ex USA |
4-Factor EW |
1196 |
2.55% |
7.02% |
11.03% |
0.59 |
HFI 6-Factor 4-Strategy EW |
812 |
3.84% |
8.30% |
11.04% |
0.71 |
|
HFI 6-Factor 4-Strategy EW Sector Neutral |
822 |
3.86% |
8.32% |
10.96% |
0.72 |
|
HFI 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) |
812 |
4.77% |
9.23% |
12.33% |
0.71 |
|
United States |
4-Factor EW |
470 |
-0.15% |
11.28% |
11.72% |
0.92 |
HFI 6-Factor 4-Strategy EW |
317 |
0.62% |
12.06% |
11.46% |
1.01 |
|
HFI 6-Factor 4-Strategy EW Sector Neutral |
319 |
1.23% |
12.67% |
11.93% |
1.02 |
|
HFI 6-Factor 4-Strategy EW Market Beta Adjusted (Leverage) |
317 |
2.33% |
13.77% |
12.92% |
1.03 |
|
Based on daily total returns in USD. The live date of the first generation Multi-Beta Multi-Strategy 4-Factor EW indices, i.e. 20/12/2013, is used as the basis. Although the other indices were created more recently, the longest live period is used for comparison purposes. The statistics are annualised and performance ratios that involve the average returns are based on the geometric average, which reliably reflects multiple holding period returns for investors. The risk-free rates used are defined according to the regional universe of the index.
Over their live period, all but one of the above Scientific Beta Multi-Beta Multi-Strategy indices posted positive relative returns compared to the cap-weighted benchmark.
The live performance data for Scientific Beta regional universes is available here.
Quarterly
Smart Beta Performance Report, June 2018
Publications
Documents authored by ERI Scientific Beta, featuring research in the field of smart beta or on related topics.
White Papers
In a concern for transparency, and as part of its aim to help investors to understand and to invest in smart beta equity strategies, ERI Scientific Beta has published a large number of white papers that are freely available on the Scientific Beta platform.
Highlighted White Papers
Tackling the Market Beta Gap: Taking Market Beta Risk into Account in Long-Only Multi-Factor Strategies
March 2018
The authors argue that more attention ought to be paid to market exposure when conducting analyses of smart beta strategies. They point out that most research proposing new multi-factor investment methodologies essentially ignores exposure to the market factor, which is the most consensual among all factors and often the most influential factor for a strategy. Overlooking the dominant factor is in stark contrast to the objective of factor investing, which aims to identify and manage the main drivers of risk and return. The authors document that different levels of market beta indeed have a strong impact on the performance and risk of smart beta strategies. The impact is visible in terms of long-term returns, volatility, and the dependence of performance on market conditions. Such effects need to be properly documented to allow investors to make explicit choices on their risk exposures. In case of a mismatch with investor preferences, it is possible to adjust multi-factor strategies to respect target levels of market beta.
Misconceptions and Mis-selling in Smart Beta: Improving the Risk Conversation in the Smart Beta Space
February 2018
Although gaining explicit exposure to priced risk factors is expected to provide good long-term risk-adjusted performance, investing in these very factors also exposes investors to a number of hidden or implicit risks that could be important drivers of short-term performance. To reconcile such risk exposures with investors’ preferences, it is crucial that they be well documented. With cap-weighted indices, which represent the default option in terms of a passive investment reference, being increasingly called into question, smart beta’s main fiduciary message is that there is no best solution in general, but rather a best solution that allows the investor’s fiduciary choices to be executed in the most efficient way. Ultimately, the choice on managing these risks is a key fiduciary decision that cannot be left to the discretion of an index provider that has no status to do so. Asset owners should also start improving governance practices by starting a risk conversation on smart beta investments with stakeholders.
All White Papers
- Misconceptions and Mis-selling in Smart Beta: Improving the Risk Conversation in the Smart Beta Space
- Scientific Beta Multi-Beta Multi-Strategy Solution Benchmarks
- Tackling the Market Beta Gap: Taking Market Beta Risk into Account in Long-Only Multi-Factor Strategies
- How to Harvest Factor Premia without Suffering from Market Volatility: The Case for a Long/Short Multi-Factor Strategy
- Accounting for Cross-Factor Interactions in Multi-Factor Portfolios: The Case for Multi-Beta Multi-Strategy High Factor Exposure Indices
- Why we do not Believe that Maximising Factor Intensity at Stock Level is a Robust Approach to Multi-Factor Investing
- Academic Lessons on Factor Investing
- The Link between Factor Investing and Carbon Emissions
- Robustness of Smart Beta Strategies
- Investability of Scientific Beta Indices
- Overview of Diversification Strategy Indices
- Scientific Beta Maximum Deconcentration Indices
- Scientific Beta Maximum Decorrelation Indices
- Scientific Beta Efficient Minimum Volatility Indices
- Scientific Beta Efficient Maximum Sharpe Ratio Indices
- Scientific Beta Diversified Risk Weighted Indices
- Scientific Beta Diversified Multi-Strategy Index
- Scientific Beta Global Universe
- Scientific Beta Analytics: Examining the Performance and Risks of Smart Beta Strategies
- ERI Scientific Beta Comments on the Mercer Report "Factor Investing: From Theory to Practice"
Practitioner Publications
The following section presents selected reference papers that have been published recently by ERI Scientific Beta in academic journals.
Multifactor Index Construction: A Skeptical Appraisal of Bottom-Up Approaches
Journal of Index Investing, Summer 2018
In this article, the authors contrast the claims of promoters of "bottom-up" approaches for constructing multi-factor equity portfolios with relevant findings in the academic literature. In particular, the authors review findings in the academic literature that raise questions on the reliability of the link between factor scores and returns, on possibilities of overstating the backtest performance of bottom-up portfolios, and on the cost of concentration resulting from the chase of factor champions. The article shows that, while bottom-up approaches are driven by a naïve belief into a fine-grain deterministic link between stock-level multi-factor exposures and returns, the empirical evidence in the asset pricing literature only supports the existence of a broad-stroke relationship. Moreover, it is emphasized that bottom-up approaches are prone to over-fitting and multiple testing biases. Without any adjustments for such biases, the backtest results of bottom-up approaches may be overstated. Finally, in the process of chasing factor champions, bottom-up portfolios tend to become highly concentrated while the academic literature stresses that diversification is crucial for the successful harvesting of factor premia. The authors conclude that findings in the academic literature on these three questions give rise to a healthy dose of skepticism concerning the superiority claims of bottom-up proponents.
Developments to the Scientific Beta Universe and Multi-Beta Multi-Strategy Indices in June 2018
On the occasion of the June 2018 rebalancing, a number of enhancements were made to the Scientific Beta universe and Multi-Strategy indices. These developments correspond to improvements that allow Scientific Beta's clients to benefit from considerable research efforts made over the last two years to improve the conditions for implementing Scientific Beta indices while respecting the objectives and methodological principles that drive the construction of the indices:
- A menu of consensual factors and a consistent and parsimonious definition of these factors.
- The diversification of the specific risk of each factor selection enabling an efficient capture of the risk premia associated with each factor.
- The taking into account of the interactions between factors enabling the transparency and granularity of a "top-down" approach to be reconciled with the search for strong factor intensity.
1. Scientific Beta Universe
The developments to the investment universe of the Scientific Beta indices are part of the annual adjustments to the universe as planned in the universe's construction guidelines.
They aim to reinforce the application of two fundamental principles/objectives that guide the construction of a universe that is appropriate for investment in non-cap-weighted universes:
i) Respecting geographic basic blocks. Both the factor selections and the application of the weighting schemes are carried out within these blocks. This regionalisation of the index construction avoids microeconomic factor choices resulting in macroeconomic imbalances. Ultimately, the index construction respects geo-neutrality with respect to cap-weighted indices.
ii) A strong liquidity requirement. Given the distance from cap-weighting that results from alternative weighting schemes and factor selection, it is important to ensure that the universe of investable stocks is highly liquid and guarantees sufficient investment capacity.
Geographic Basic Blocks
The ERI Scientific Beta universe, which forms the basis of all Scientific Beta equity indices, aims at constantly reflecting investment opportunities and providing broad coverage of the international equity markets. The universe is divided into Geographic Basic Blocks, each having a fixed number of securities. ERI Scientific Beta has reviewed its block structure as follows:
- Creation of dedicated blocks for China and India. The Emerging Asia-Pacific block thus includes three constituent blocks – Emerging Asia-Pacific ex-China ex-India and the China and India block indices.
- Discontinuation of the Israeli block since this block does not reach a minimal number of constituents. Israel will now be part of Developed Europe & Middle-East ex-Eurozone ex-UK.
- Resizing of Emerging basic blocks, Developed Eurozone and Canada in order to have more consistent coverage by capitalisation.
The new developments have led to changes in the sizes of the blocks compared to the existing universe.
Liquidity
Two developments were implemented in June 2018. They both aim to strengthen the liquidity of the Scientific Beta universe.
The liquidity screen used to remove half of the least liquid securities from each Geographic Basic Block, that is based on a composite score combining Trading Ratio and Absolute Volume rankings, has been reinforced by implementing three distinct components:
- A minimum Trading Ratio requirement.
- A screening and minimum requirement based on the Absolute Volume.
- A minimum requirement of Relative Volume measured in the short-term (3 months) and long-term (1 year).
To further account for the foreign room available to investors when a security is subject to a foreign ownership limit, an adjustment factor has been created. This adjustment factor is applied to mitigate the security's Relative Volume measure (while processing the liquidity screening) and the free-float capitalisation.
These developments have a significant impact on the investability of the Scientific Beta universe.
Altogether, the changes that impact the construction of the Scientific Beta universe lead to a serious improvement in the investability of our universe. We can mention the following for illustrative purposes:
- Average ADDTV of the emerging universe constituents increases by 94% and by 16% for the developed universe.
- Days to Trade (95th percentile value) of cap-weighted is reduced by 32% for the emerging universe and 8% for the developed universe.
- The reduction in the worst ADDTV for its part is improved by 31% for the developed universe and 89% for the emerging universe.
2. Scientific Beta Multi-Beta Multi-Strategy Indices
Developments have been implemented while respecting the objectives and methodological principles of the Scientific Beta Multi-Beta Multi-Strategy indices. These developments are improvements to all the construction stages of the multi-strategy smart factor indices that make up all of the Scientific Beta Multi-Beta Multi-Strategy indices.
These stages are represented below:
The developments are as follows:
2.1 Stock Selection
The univariate approach with consensual variables is maintained but regroups stocks to avoid certain discrepancies.
- Selection within mega-sectors for value, investment and profitability to avoid accounting discrepancies.
- Selection based on local currency return data for low volatility and momentum in multi-currency geographic blocks.
These corrections add consistency to the factor selections, which are always carried out using the same factor proxies in the same proportions (50% of the universe for the broad indices and 30% for the narrow indices) and also guarantee greater diversity in the stocks selected.
2.2 Factor Intensity Filter
The factor intensity filter maintains the current arithmetic average score across factors but reweights factors to reflect distributional properties of scores since the rankings used do not provide information on the differences and variations in the different factors' intensity distributions. This innovation is in response to a consistent criticism of the predominant approaches in index construction, whether "top-down" or "bottom-up", namely that rankings or scores are a simplification providing robustness in factor index construction but which unfortunately obscure an important amount of information, notably the variations in factor intensity over time and between factors. The approach proposed by Scientific Beta reconciles the robustness of rankings with the relevance of information on factor intensity distribution. Hence, each time the factors are reconstituted, this reweighting will give different relative weights to each of the five factor scores used in the residual multi-factor filter that is intended to improve the factor intensity of the indices according to the relative intensity of the universe for each of the factors.
- The High Factor Intensity filter (40% removal) benefits from the evolution of weighting in the composite.
- The standard indices benefit from a standard filter (20% removal) based on the same methodology.
These developments strengthen the factor intensity of the indices, especially as part of the standard non-High-Factor-Intensity Multi-Beta Multi-Strategy indices, and above all reduce the number of cases where these indices have one or more negative factor betas.
Impact of the Changes on the High-Factor-Intensity Multi-Beta Indices
The universe is US LTTR. The time period is from June 1970 to December 2016. The regression analysis is performed using weekly total returns in USD. A 7-Factor model is used that has cap-weighted market factor and 6 L/S EW factors. The yield on Secondary US Treasury Bills (3M) is used as a proxy for the risk-free rate.
Impact of the Changes on the Standard Multi-Beta Indices
The universe is US LTTR. The time period is from June 1970 to December 2016. The regression analysis is performed using weekly total returns in USD. A 7-Factor model is used that has cap-weighted market factor and 6 L/S EW factors. The yield on Secondary US Treasury Bills (3M) is used as a proxy for the risk-free rate.
2.3 Multi-Strategy Weighting Scheme
The Efficient Minimum Volatility Weighting component has been excluded from the Multi-Strategy Weighting Scheme as it leads to pronounced exposure biases and increases cross-factor correlation.
This evolution of the multi-strategy weighting scheme has no significant impact on the level of diversification of the specific risks of the Scientific Beta multi-strategy indices and moreover, the robustness of the diversification obtained through the diversification of the inherent model risk in a multi-strategy scheme is retained.
2.4 Liquidity Capping
The enhancement implemented aims to strengthen the investability of the indices in reference to capacity/liquidity expressed in relation to the cap-weighting of the stocks.
The current capping at a multiple of market cap is maintained but the level is tightened to reduce the exposure to stocks with low capitalisations:
- The multiple is set at 3 instead of 5. It also allows weight constraints to be loosened to lambda = 4 from lambda = 3.
These new liquidity rules have a very positive impact, increase the capacity of the Multi-Beta Multi-Strategy indices by 20% on average, and avoid strong exposure to stocks with low levels of liquidity.
For further information, please contact our Client Services department on +33 493 187 851 from 9.00am to 6.00pm CET or at clientservices@scientificbeta.com.
Events
As smart beta is a new frontier for the investment industry, research and education on this subject are a core part of ERI Scientific Beta's strategy. That is why we regularly invite interested parties to participate in events on the subject of smart beta investing.
EDHEC Scientific Beta Days North America 2018 Conference
25-26 October, 2018 – Boston, United States
This two-day conference, organised by ERI Scientific Beta, will present the asset owner and financial advisory communities with the latest conceptual advances and research results in smart beta investing, enabling their implications and applications to be discussed with researchers who combine expertise of advanced financial techniques with a sound awareness of their industry relevance. The event will include multiple plenary sessions, workshops and practical sessions allowing professionals to review major industry challenges, explore state-of-the-art investment techniques and benchmark practices to advances in research.
The conference will focus on the following themes:
Day One: Smart Beta Risk Management Solutions
- Distinguishing Factor Timing and Dynamic Risk Allocation
- How to Select a Multi-Factor Portfolio Construction Method
- Regime Premia Diversification in Multi-Factor Investing
- Measuring and Controlling the Transaction Costs of Factor Investing
Day Two: Case Studies and Practical Applications
- Smart Factor Selection Process
- Long/Short Strategies
- Implementing Factor Risk Allocation without Compromising its Existing Allocation – The Case of Factor Overlay Portfolio
- Integrating Multi-Factor and Smart Beta Investing in the Portfolio Allocation and Construction Process of a Pension Fund
The detailed program is available here.
The conference is reserved for asset owners (including pension schemes, charities, endowments, foundations, insurance companies, single family offices and financial executives from non-financial companies) and institutional consultants.
Admission is complimentary and by invitation only. To request an invitation to the conference, please visit the dedicated registration website.
For further information about this event, please contact Joanne Finlay at scientificbetadays@scientificbeta.com.
Press Review
A selection of articles published in the international business or specialised press featuring ERI Scientific Beta and its research.
Smart beta is getting even smarter
Moneyweb (25/06/2018)
"(...) The solution being proposed by many product providers is therefore to blend smart beta factors together in 'multi-factor' offerings. These use a number of different strategies such as value, momentum and quality in a single fund. "Factors priced by the market tend to be decoupled from each other," explained Shirbini. "When one factor is performing poorly, another will compensate. In the long term they all produce good risk-adjusted return, but a multi factor index smooths performance over the short term. For retail investors a multi-factor index is therefore a more suitable investment product." (...)"
Copyright Moneyweb Holdings Ltd
ERI Scientific Beta highlights the importance of market risk in smart beta strategies
ETF Strategy (04/06/2018)
"(...) The study found that, over the long term, adjusting the market beta of a multi-factor strategy with that of a cap-weighted index corresponds to an annual gain in performance of more than 1%. More generally, the authors demonstrate that different levels of market beta have a strong impact on the performance and risk of smart beta strategies in terms of long-term returns, volatility, and the dependence of performance on market conditions. (...)"
Copyright ETF Strategy Ltd
Single-Factor Products Gather Most Assets
Nordic Investor (31/05/2018)
"(...) The ERI Scientific Beta indices (a venture of the EDHEC Risk Institute) are based on four smart factor indices: mid-cap, value, momentum and low volatility, and have now gathered over $10 billion of assets in a range of products run by Global X, Amundi, ETFS and Morgan Stanley: Global X Scientific Beta US, Europe, Japan and Asia ex-Japan (tickers SCIU, SCID, SCIJ an SCIX); Amundi ETF Global and Europe Equity Multi Smart Allocation Scientific Beta UCITS ETF (tickers SMRT and SMRE); ETFS Diversified-Factor US Large Cap, and Developed Europe, Index Funds (SBUS, and SBEU); and Morgan Stanley Scientific Beta Global, and US, Equity Factors UCITS ETF (GEF and USEF). The EDHEC approach is relatively sophisticated in that the factors are proprietary rather than generic definitions. (...)"
Copyright Nordic Investor
Lyxor launches multi-factor long/short equity ETF
ETF Strategy (22/05/2018)
"(...) Lyxor has launched the Lyxor Scientific Beta Developed Long/Short UCITS ETF, providing exposure to an equity market neutral index that takes a long position in a number of portfolios designed to maximise exposure to different factors while shorting the broad cap-weighted market. The underlying reference, created by smart beta index provider ERI Scientific Beta, is the Scientific Beta Multi-Beta Multi-Strategy Managed Volatility L/S Equity Market Neutral (x 3.5) Index which includes large- and mid-cap equities from global developed markets. The index implements its long/short strategy through a short position in the cap-weighted reference index and a quarterly allocation to sub-portfolios in the long leg. (...)"
Copyright ETF Strategy Ltd
Market Exposure Needs Attention
Benefits and Pensions Monitor (16/05/2018)
"(...) More attention should be paid to market exposure when conducting analyses of smart beta strategies, says Scientific Beta. Its research paper ‒ ‘Mind the Gap: On the Importance of Understanding and Controlling Market Risk in Smart Beta Strategies’ ‒ argues most research proposing new multi-factor investment methodologies essentially ignores exposure to the market factor, which is the most consensual among all factors and often the most influential factor for a strategy. It demonstrates that different levels of market beta have a strong impact on the performance and risk of smart beta strategies in terms of long-term returns, volatility, and the dependence of performance on market conditions. (...)"
Copyright Benefits and Pensions Monitor
ERI Scientific Beta names best performing indexes
Money Management (27/04/2018)
"(...) The best performing index in the March quarter in the developed universe among those smart factor indices was the SciBeta Developed Narrow High Factor Intensity High Momentum Diversified Multi-Strategy index with a relative return of 2.03 per cent compared to the broad cap-weighted index, according to the smart beta indices platform provider ERI Scientific Beta. (...) As far as the multi smart factor indices were concerned, the SciBeta Developed Multi-Beta Multi-Strategy 4-Factor EW index, the SciBeta Developed High Factor Intensity Multi-Beta Multi-Strategy 6-Factor EW index and the SciBeta Developed High Factor Intensity Multi-Beta Multi-Strategy 6-Factor EW Market Beta Adjusted (Leverage) index all posted positive relative returns of 0.80 per cent, 0.25 per cent and 0.10 per cent, respectively compared to cap-weighted indices. (...)"
Copyright Financial Express
Warnings About Smart Beta From Smart-Beta Index Provider
Validea (04/04/2018)
"(...) A white paper on smart-beta investing released this week by "veteran quant" Eric Shirbini of ERI Scientific Beta says that "second-order risks and exposures aren't being sufficiently accounted for by investors or disclosed by providers." This according to a recent article in Institutional Investor. "Just like monitoring the style drift of active managers," writes Shirbini, "investors need to monitor the risk dynamics of factor strategies." Specifically, Shirbini cites three categories of risk to be monitored: market beta, macroeconomic and sector/geographical. If not managed, he argues that a smart-beta strategy could lose out on the long-term equity market risk premium. (...)"
Copyright Validea
As part of its international development programme and in order to strengthen its index development activity, the EDHEC group, one of Europe's leading research and teaching institutions, is recruiting for positions within Scientific Analytics in London and ERI Scientific Beta in Nice. To apply, please send your CV and a cover letter to Laurence Kriloff: laurence.kriloff@edhec-risk.com.
- Salaries are determined according to the EDHEC pay scale, based on qualifications and prior experience.
- Written and spoken English is essential, and basic spoken French is a plus.
- An EU work permit is mandatory for the positions in ERI Scientific Beta's Nice and London offices.
For more information about ERI Scientific Beta, please visit our website and our corporate YouTube channel.
Director of Research, Scientific Analytics (London)
As part of its policy of transferring know-how to the industry, EDHEC is setting up Scientific Analytics. Scientific Analytics is an original Fintech initiative that aims to favour the adoption of the latest advances in factor allocation and its implementation by institutional investors, asset managers and wealth managers. Its academic origin provides the foundation for its strategy: offer, in the best economic conditions possible, tools to conduct risk analysis and design innovative solutions based on factor allocation risks. The objective is to make Scientific Analytics a provider of disruptive technology for the whole of the investment industry.
Scientific Analytics is part of a series of initiatives conducted by EDHEC to develop a connection between research and the investment industry, among which we can cite:
- Scientific Beta, which, with more than USD 25 billion under replication, is one of the worldwide leaders in the design and calculation of smart beta indices (more on www.scientificbeta.com)
- EDHEC Infra, which has built the largest database of unlisted infrastructure investments in Singapore and produces indices on all the equity and debt dimensions of unlisted infrastructure projects
As part of its R&D effort, and to coordinate the activity of all of its research teams, Scientific Analytics is recruiting a Director of Research. There are ambitious plans for this position. The Director of Research will be one of the key members of the management team and as such will participate in the development of a firm that has a worldwide business plan and appropriate funding for this ambition.
The ideal candidate has experience of at least seven years in an equivalent position within an institutional investor or asset manager firm. The candidate's academic record must certify genuine competence not only in advanced quantitative techniques, but also in portfolio construction. A PhD would be a highly appreciated bonus.
In addition, the candidate will have published research papers on quantitative portfolio construction or factor allocation implementation. In addition to theoretical expertise, the successful candidate will also have a strong interest in the development of concrete investment solutions and factor strategies both in equities and in a multi-asset context as part of a business project.
The candidate will have to assemble a diverse group of experts. He/she will be passionate about team spirit and will connect closely with each of the team members.
Quantitative Equity Analyst (Research), ERI Scientific Beta (Nice)
The successful candidate will be a quantitative equity analyst with significant skills in the construction of quantitative equity portfolios, the implementation of equity factor investing, and the publication of research reports.
The position requires a Master’s degree in Finance or Financial Engineering from a leading institution and experience in constructing quantitative equity portfolios within an equity index provider, investment bank, or asset management firm.
Experience in drafting and publishing research reports for a broad audience in English is a must.
The candidate also needs to have a sound command of Matlab for financial computations, in particular in the area of portfolio construction and performance analysis.
Experience in working with large equity datasets (such as CRSP, Compustat) would be an advantage.
The successful candidate, who will report to Scientific Beta’s director of research, will participate in Scientific Beta's research work in factor investing and will contribute to the design or validation of equity investment strategies and performance and risk analytics.
ERI Scientific Beta
ERI Scientific Beta aims to be the first provider of a smart beta indices platform to help investors understand and invest in advanced beta equity strategies.
Established by EDHEC-Risk Institute, one of the very top academic institutions in the field of fundamental and applied research for the investment industry, ERI Scientific Beta shares the same concern for scientific rigour and veracity, which it applies to all the services that it offers investors and asset managers.
The ERI Scientific Beta offering covers three major services:
- Scientific Beta Indices are smart beta indices that aim to be the reference for the investment and analysis of alternative beta strategies. Scientific Beta Indices reflect the state-of-the-art in the construction of different alternative beta strategies and allow for a flexible choice among a wide range of options at each stage of their construction process. This choice enables users of the platform to construct their own benchmark, thus controlling the risks of investing in this new type of beta (Smart Beta 2.0).
Within the framework of Smart Beta 2.0 offerings, ERI Scientific Beta provides access to smart factor indices, which give exposure to risk factors that are well rewarded over the long term while at the same time diversifying away unrewarded specific risks. By combining these smart factor indices, one can design very high performance passive investment solutions.
- Scientific Beta Analytics are detailed analytics and exhaustive information on its smart beta indices to allow investors to evaluate the advanced beta strategies in terms of risk and performance. The analytics capabilities include risk and performance assessments, factor and sector attribution, and relative risk assessment. Scientific Beta Analytics also allow the liquidity, turnover and diversification quality of the indices offered to be analysed. In the same way, analytics provide an evaluation of the probability of out-of-sample outperformance of the various strategies present on the platform.
- Scientific Beta Fully-Customised Benchmarks and Smart Beta Solutions is a service proposed by ERI Scientific Beta, and its partners, in the context of an advisory relationship for the construction and implementation of benchmarks specially designed to meet the specific objectives and constraints of investors and asset managers. This service notably offers the possibility of determining specific combinations of factors, considering optimal combinations of smart beta strategies, defining a stock universe specific to the investor, and taking account of specific risk constraints during the benchmark construction process.
With a concern to provide worldwide client servicing, ERI Scientific Beta is present in Boston, London, Nice, Singapore and Tokyo. As of December 31, 2017, there was USD 25bn in assets replicating Scientific Beta indices. 35% of these assets under replication are ESG-compliant. ERI Scientific Beta has a dedicated team of 45 people who cover not only client support from Nice, Singapore and Boston, but also the development, production and promotion of its index offering. ERI Scientific Beta signed the United Nations-supported Principles for Responsible Investment (PRI) on September 27, 2016.
Contact
ERI Scientific Beta
1 George Street, #15-02, Singapore 049145
Tel. +33 493 187 851 (from 9.00am to 6.00pm CET)
E-mail: clientservices@scientificbeta.com | Website: www.scientificbeta.com