Minimum variance

We doubt that the growth of investments in minimum variance over recent years has been due solely to the superior efficiency of this approach compared to the cap-weighted market index, especially since this evidence was provided by empirical financial market research decades ago. Rather, the exceptional market situation of negative or low interest rate policy worldwide, combined with a number of political, geopolitical and economic risks, will have led to this increase.

Because there is a lack of alternative investment opportunities, many investors are increasing their equity quota to reach their return target. For this reason, they must pay more attention to portfolio risk when planning asset allocation. In this way, they consciously or unconsciously take into account Markowitz’s core contribution to modern portfolio theory: optimising portfolio risk and portfolio diversification. Risk-based approaches are helpful in this context because they create degrees of freedom for decision-makers to optimise the use of the risk budget.

However, despite increasing acceptance, risk-based strategies continue to have a tough time on the market. Many investors are unaware that they are not only superior to the ‘Holy Grail’ of the market-capitalised index in terms of risk, but in the long term, they are also superior regarding return. Even scientific evidence cannot eliminate the widespread scepticism.

In this research note, we address the arguments of critics who are calling into question the added value of minimum variance, and who see confirmation of their reservations in the 2016 underperformance. In the following, we explain why these arguments do not apply and why it is worthwhile for investors to focus in the long term on optimised diversification and risk management, even when outperformance is not expected during a longer phase of unusually low market volatility, as in 2016.

1. Is there a risk of bubble formation with minimum variance?

The volume of risk-based strategies is small compared to index funds and ETF. Moreover, the approaches and their portfolios are different.We have identified 73 ETFs and 75 investment funds in Bloomberg which invest in minimum variance, low volatility or similar risk-based approaches. For this group, we have analysed the assets and cash flows (net subscriptions) at the end of September 2016. This period was chosen because the risk-based strategies achieved a strong underperformance in the following (fourth) quarter.

At the end of September, the selected products reported managed assets of USD 53 billion (ETF) or USD 38 billion (investment funds) under management. The total sum of USD 91 billion corresponds to less than half of the spider, the largest ETF on the S&P 500 with a volume of USD 195 billion. Although risk-based ETFs from September 2015 to September 2016 claimed almost 12 % of all equity ETF inflow and the assets under management nearly doubled, their share of the total investment volume in equity ETFs of USD 2.7 trillion was still marginal at around 2 %.

Risk-based strategies do not form a monolith – in total contrast to cap-weighted index funds / ETFs where the portfolio structures of the different providers are exactly the same. In the case of risk-based approaches, there are various strategies such as low volatility (selection of securities with the lowest volatility), minimum variance or minimum volatility (optimisation of securities weighting of the portfolio with lowest volatility). The implementations of the different providers differ even within the same strategy.

Figure 2 illustrates how different risk-based strategies differ:

The first comparison shows the overlap between two different strategies: minimum volatility (iShares MSCI USA Minimum Volatility ETF) vs. low volatility (Powershares S&P 500 Low Volatility ETF). About 40 % of the securities occur in both strategies. The cap-weighted overlap, the lowest common denominator, corresponds to approximately 50 %.

The second comparison shows two similar strategies: OLZ Efficient World® Equity vs. iShares Edge MSCI World Minimum Volatility. It makes clear that there are also major differences within the same strategy. The overlap of the two portfolios is just 34 %.

The differences are also observed in country and sector allocation, as shown in Figures 3 and 4.

Each provider has their own recipe. The strategies differ in terms of the restrictions on liquidity, the methodologies for parameter estimation, the lowest or highest weights, and the optimisation of rebalancing and turnover frequencies. Whilst the overlap for cap-weighted index funds and ETFs is 100 %, this is significantly less for risk-based approaches.

In summary, two points can be highlighted:

– Although the percentage growth of inflows was very high in risk-based strategies, the absolute investment sum relative to the assets under management in market cap-based approaches is still marginal.

– In addition, risk-based solutions differ, which simply means that the money does not flow into the same securities.

The growth of risk-based funds and ETFs must therefore be assessed in a relativised and differentiated way. In absolute terms, most funding in recent years has continued to flow into market cap-weighted index funds / ETFs. The trend towards the market average (market index) and therefore towards herd behaviour is unbroken. Within risk-based strategies, each provider pursues a different approach with the result that the portfolios are not identical, which greatly reduces the risk of bubble formation.

2. Is the tracking error for minimum variance too high?

Tracking error is the wrong measure for assessing active strategies; the Sharpe ratio, which measures portfolio efficiency, is more meaningful.

Another criticism is the high tracking error of risk-based strategies versus the cap-weighted market index. Tracking error measures the deviation of an investment strategy from the benchmark. It is used as a risk measure by many investors, also as a result of widespread indexing. Tracking error is a suitable criterion for passive index replications, but it is not suitable for the assessment of active strategies. It would only be meaningful if the benchmark – the cap-weighted index – were efficient, as is often presumed. However, empirical financial market research has repeatedly shown that this is precisely not an efficient portfolio. There are portfolios with less risk and more return than the market index. Tracking error can thus be misleading for active strategies because a high value suggests high risks.

Figure 5 illustrates the problem with tracking error: it shows the value development of two strategies with different fluctuations, both of which are invested in the same universe. Both strategies achieve the same return over the time span, but one (Simulation 1) fluctuates much more strongly than the other (Simulation 2). If the risk of Simulation 2 was measured as deviation from Simulation 1 – which corresponds to tracking error – then Simulation 2 would look very bad, even though after risk-adjustment, it is much more attractive than Simulation 1.

Active strategies such as minimum variance deliberately accept deviations from the market index in order to improve the efficiency of the portfolios. By definition, they have quite a high tracking error relative to the index. The Sharpe ratio, which measures the excess return of an investment per unit of risk, is a more adequate indicator for the assessment of active strategies. The higher the Sharpe ratio, the better the return / risk ratio of an investment. Minimum variance strategies exhibit lower fluctuations and higher Sharpe ratios than the cap-weighted index, but this is not reflected by tracking error.

3. Is the added value of minimum variance just a temporary anomaly?

Long-term empirical evidence supports the minimum variance approach.

Is the long-term higher value of minimum variance merely a transient, soon to disappear anomaly, as voiced by critics? The positive low-risk effects are duly documented in empirical financial market research. One of the first and most important studies was published in 1972 by Rob Haugen and James Heins. On the basis of data dating back to 1926, they show that portfolios compiled according to the lower volatility criterion exhibit a higher return at the same or even lower risk than the benchmark index. This observation was confirmed by further studies, in particular by Haugen and Baker (1991, 2012).

4. Can the additional returns be explained by classical risk premiums?

Low-risk effects cannot be explained by other factors such as size, value or illiquidity.

The on average better performance of minimum variance cannot be explained by other risk factors such as illiquidity, size or value. In collaboration with Prof. Heinz Zimmermann (University of Basel) [1], OLZ was able to demonstrate that neither a value nor a small-cap or illiquidity factor is statistically significant for the surplus return of minimum variance. The latter is all the more logical since OLZ does not invest in small caps and requires strict liquidity requirements in the selection of securities.

5. Are low-volatility securities overbought (crowding) and the surplus return therefore a result of the rising valuation?

Surplus return relative to the market index cannot be explained by an increasing valuation (multiple expansion). Valuation ratios or multiples must also be interpreted with caution.

A further point of criticism is that risk-based strategies are overbought because of the large inflows (crowding). If this were the case, investors would have to fear two consequences: firstly, because of herd behaviour, many investors would be equally invested. If they back out of the strategy, there is the danger of a correction in corresponding securities. Secondly, it is argued, the large inflows will have pushed the price artificially upward. As a result, the profitability of low volatility strategies will suffer in the coming years. The second point regarding the over-inflation of low-risk securities in particular, has been actively debated in the financial media and among investors. It is based on the analysis of various financial ratios (multiples).

The price-to-book ratio (P/B) of minimum volatility strategies has risen somewhat more in recent years relative to the market index and also has a higher value in absolute terms. On the basis of this figure, it is argued that minimum variance portfolios are expensive and this increase will be at the expense of profitability in the coming years [2]. It is further argued that the surplus return which has been achieved so far can only be explained by P/B development (multiple expansion). This is not due to a low-risk effect, but simply to a price bubble in low-volatility securities.

Regarding this, we would like to record the fol-lowing facts:

– Minimum variance portfolios have systematically beaten the market since the 1930s, as empirical studies prove. This surplus return relative to the market index cannot be explained by increasing valuation (multiple expansion).

– Multiples such as P/B are not particularly relevant to the intrinsic value of a company or portfolio. Poor companies with structural problems and low profitability have a lower P/B than good companies. This does not mean per se that these securities are “cheap” and the others are “expensive”.

– As at 30 September 2016, the MSCI Minimum Volatility Index had a P/B of 2.7 versus 2.1 (+ 30 %) of the cap-weighted index, while the OLZ Minimum Variance Portfolio is 20 % higher at 2.5 % (Source: Bloomberg). In a historical context, these values by no means match the valuations during price bubbles [3].

The varied price/book value development of the cap-weighted index and the minimum variance portfolio can be partially explained by the different sector weightings. The cap-weighted index has a very static structure over time, whilst minimum variance processes new information from the market. For example, market indices have higher weights in financial and energy markets, while minimum variance strategies currently underweight these.

Figure 6 compares the development of sector weights. Although the P/B of financial securities [4] has halved in the aftermath of the financial crisis (Figure 7), the sector weight in the market index remained relatively high at about 20 %, whereas it is around 5 % – 10 % at OLZ. The higher weighting of financial securities in the market index contributes to the lower P/B. It is only due to this effect that the OLZ portfolio P/B ratio is around 15 % higher than before the financial crisis. On the other hand, the minimum variance portfolio is less exposed compared to the financial securities that are burdened with problems and have become more risky. The same pattern is also found in the energy sector.

The example shows that the assessment of a portfolio with valuation figures should be interpreted with caution. Instead, the focus should also include quality and profitability.

We compared profitability (EBIT/Assets) and P/B ratio of OLZ Minimum Variance Equities World portfolio with the MSCI World Index (Figure 8). The higher P/B of the OLZ minimum variance portfolio can be partially explained by the systematically higher profitability, measured as EBIT (operating profit before tax and interest) in relation to the total assets of a company.

A differentiated analysis of the ratios is important in order to put the underperformance of risk-based strategies – compared to the market index of recent months – in the right context and to avoid false conclusions. According to our analysis, this underperformance has nothing to do with high multiples of low-risk equities or unfavourable valuation comparisons. Rather, the trigger is the very optimistic assessment of economic development and the reflationary dynamics. Procyclical and risky securities have become attractive, not because of their low valuation, but because of their sensitivity to economic growth.

6. Is factor timing the solution?

A structured and disciplined investment approach with focus on efficiency and diversification is more promising in the long term than all timing strategies.

Factor timing looks very tempting on paper, but usually leads to suboptimal decisions in reality. In a study, index provider MSCI examined the relationship between valuation and future performance for both the value and minimum volatility indexes. The ratio between the initial value of the P/B (relative to the cap-weighted market index) and the return over the following five years was compared. Whilst for value, an investment at low P/B tended to yield better returns (green cloud in Figure 9), this relationship is less pronounced in the minimum variance portfolio (red cloud in Figure 9).

Anyone who practises factor timing faces the same challenge as with any timing strategy: the lack of market predictability. Both inflation and interest rate developments as well as improved economic development have decisively determined the performance of recent months. But these have repeatedly been wrongly predicted in recent years. Portfolio reallocation was at the expense of diversification and incurred high transaction and opportunity costs. Figure 9 shows that, particularly for minimum volatility, there is no statistically strong correlation between the current valuation and the performance of the following years.

Minimum variance, and factor strategies in general, give investors much more freedom in designing efficient asset allocation. OLZ is of the definite opinion that much more time and energy should be put into factor analysis rather than an attempt to time the market. A structured and disciplined approach with focus on efficiency and diversification is more promising in the long term than all timing strategies.

7. Do low-risk strategies exhibit a high level of debt and thus correspondingly high interest rate sensitivity?

The level of debt is equal or in fact lower than that of the market index. It is not interest rate sensitivity but rather economic growth which is an important factor.

Risk-based strategies are often associated with investments in securities with high levels of debt, which is accompanied by high interest rate sensitivity. This would indeed be extremely problematic in the current low interest rate environment, because it would compromise diversification to bonds at the level of the overall portfolio.

Figure 10 shows the degree of debt at the sector level of the OLZ Minimum Variance Equities World strategy compared to the MSCI World (Debt-to- Equity ratio as at 30 September 2016).

From the figure it can be seen that within the individual sectors OLZ minimum variance predominantly favours securities with a fairly low level of debt. Only in the energy sector is the level of debt considerably higher than MSCI World. The energy sector has the smallest weight in our portfolio and is only represented by a few securities, some of which have an increased debt level.

In the sector weighting (Figure 11) it can be seen that the OLZ model favours sectors with higher levels of debt (eg telecommunications, utilities or consumer staples). Financials (not including real estate) have the highest weights in the index. As expected, this sector has the highest debt; however, it cannot be compared with all other sectors. With and without taking the financial sector into account, the degree of debt in the OLZ portfolio is lower than in the benchmark.

The result is consistent for all our funds for different regions. Risk-based strategies therefore do not per se show a high level of debt and thus high interest rate sensitivity. This finding is important in order to better understand the cause of the underperformance with rising interest rates in the second half of 2016. This is not due to debt level but rather to the underweighting of cyclical sectors and/or securities. The expectations of higher economic growth and rising inflation rates have fuelled cyclical securities, which are typically underweighted in a risk-based strategy.

But we need to be aware that interest rates can also rise with declining or shrinking economic growth (stagflation). According to many analysts, such a market development is no longer just a textbook scenario. We expect that minimum variance can deliver significant added value in such a market phase.

Conclusion

The discussion about the appeal of risk-based strategies revolves almost exclusively around (short-term) return. The risk characteristics of investments and their contribution to the formation of efficiently diversified portfolios are only marginally discussed. Yet in today’s market environment, it is crucial for long-term investment success to optimise the risk-return profile of the investments and to make the most of the existing risk budget. OLZ is a Markovitz advocate and focuses on portfolio efficiency: we think in terms of risks, and not simply short-term return. We are convinced that efficiently diversified portfolios, such as the minimum variance portfolio, make an important contribution to the stability of a portfolio – and thus achieve a better return over the long term.

[1] The results of the study were presented by Prof. Heinz Zimmermann at the 3rd OLZ Smart Investing Day (21.01.2015 ).

[2] Arnott, Beck, Kalesnik, West, “How can Smart Beta Go Horribly Wrong”, Research Affiliates, 2016; Arnott, Beck, Kalesnik, “To Win with Smart Beta ask if the Price is Right”, Research Affiliates, 2016

[3] Clifford Asness in My Factor Philippic, AQR 2016 summarizes an interesting and incisive comment on Research Affiliates’ articles listed in Footer 2.

[4] At MSCI, the financial sector also included Real Estate until 30 September 2016. In historical comparison, we show the sector weight of financials (in the index and in our fund) according to the old definition.