Correlation matters

Traditional portfolio theory assumes static correlations between asset classes and much portfolio construction is based on this simple assumption. The reality is very different. To see that, one needs to look no further than how U.S. equities have correlated with U.S. bonds since 1990. Whereas the correlation was positive during most of the 1990s, it was very unstable and quite unpredictable between 1998 and 2008, only to turn decisively negative after 2008 (chart 1).

Not surprisingly, as a consequence, a static 60-40 portfolio (60% in equities), as many continue to recommend, has not at all been the optimal solution it is often portrayed as. In terms of risk-adjusted returns (Sharpe ratio), if only equities and bonds are at our disposal, a 35-65 portfolio has actually done considerably better since 1990 (chart 2).

For that reason alone (but it is not the only one), the period since the financial crisis in 2008 has presented investors with extraordinary challenges. A traditional approach to portfolio construction has not worked very well. Either risk has been ‘on’ or it has been ‘off’. Correlations across risk assets have been very high, and volatility has been plagued by at least a couple of extreme bursts (chart 3).

What does risk on, risk off really mean?

A risk on, risk off environment is characterised by much higher correlations than you would normally expect to see, at least as far as risk assets are concerned. Low volatility prevails when risk is ‘on’, but occasionally it is replaced by relatively short, but very sharp, bursts of much higher volatility when risk goes off. Risk off assets like U.S. Treasuries (which have been a risk off asset at least since 2008) are negatively correlated with risk on assets in such an environment.  

Global markets have been sucked into a risk on, risk off environment since the near meltdown caused by the financial crisis in 2008. This research paper will look at which behavioural factors have been impacted the most, and whether there are any signs of a return to some sort of normality.

In a (theoretical) world where the correlation between two assets is perfect (i.e. it equals one), the risk-adjusted return (the Sharpe ratio) of the two assets must be exactly the same. I shall not tire you with the mathematical proof. Suffice to say that when the correlation between two assets is perfect, the less risky asset can always be synthetically created by combining the riskier asset with a risk-free asset. Practically, even if correlations are not perfect, in a risk on, risk off environment risk-adjusted returns on most risk assets should be quite similar.

This implies that investor behaviour changes when such dynamics apply. Remember what I just said. Correlations between risk assets will be much higher than usual in a risk on, risk off environment. As a consequence, the trade-off between risk and returns is now perceived to be very similar across asset classes (and rightly so). Therefore investors don’t distinguish much. They buy when risk is on and sell when risk is off – and they do so across the board because the choice of risk assets matters little. Hence the expression risk on, risk off.

One further implication: In a world where all Sharpe ratios are virtually the same (i.e. when all risk assets are nearly perfectly correlated), the only way investors can outperform is by taking more or less risk. If investors all take the same amount of risk, they will all deliver exactly the same return.

This warrants a delicate question. If the only way to outperform is by deviating on risk (volatility), is it really worth paying 2+20 for that? I wouldn’t have thought so. Of course, in the real world, risk assets are never perfectly correlated, but a surprisingly high number have been extraordinarily highly correlated since 2008.

Please note that this logic does not consider other risk factors than volatility – an issue I shall come back to later.

A snapshot of behavioural dynamics that have changed.  

A number of behavioural factors changed markedly in the aftermath of the financial crisis, which peaked in 2008. If we assume that a return to some sort of normality will drive those dynamics back to how they used to be like pre-2008, then it is probably worth taking a closer look at them in order to assess how far down the road we are in terms of normalisation.

You hear the argument virtually every day that the crisis isn’t over yet but, when it eventually is, interest rates will rise dramatically. It is therefore only reasonable to expect investor behaviour to gradually change as more and more investors subscribe to the view that the financial crisis is finally over, even if it is still a minority who think so.

Several behavioural dynamics have changed as a result of the risk on, risk off environment which has ruled the world for nearly 7 years now. With the caveat that this list is by no means all encompassing, here are my candidates (in no particular order):

• Risk off assets: Investors haven’t embraced many asset classes as risk off assets but have actually been quite imaginative when doing so. For example, the Japanese Yen appears to have been widely accepted as a risk off asset more recently.
• Passive vs. active: Passive investments have grown in size as most investors find it hard to generate alpha when correlations are high.
• Value vs. growth: Value stocks have underperformed growth stocks – not every year but for much of the time in recent years.
• Correlations between equities and their currencies: Perhaps surprisingly, the correlation between many equity markets and their underlying currencies has changed significantly since 2008, but not always in the same direction.
• Correlations between different investment strategies: High correlations have not been limited to asset classes but have also had an effect on a broad number of investment strategies within the same asset class, making life difficult for active investment managers.
• Focus on other risk factors: Some investors have taken the consequence of all of this, and have as a result increasingly focused on other risk factors than volatility to provide the ‘extra kick’.

Let’s take a closer look at each one of them but, before I do so, one disclaimer: A thorough and proper analysis of each and every one of them would turn this into a 50 page document which few people have the appetite for, myself included. I shall therefore refer readers to future research papers from ARP which shall pick up on a few of the subjects only discussed rather superficially in this paper.

Risk off assets

Risk off assets tend to originate from the larger and more established economies around the world. As most investors would know, U. S. Treasury bonds have been a favourite risk off asset during the post crisis years.

A more surprising risk off asset in recent times has been Japanese Yen (JPY) which rose in value when risk assets fell out of bed in 2008 and again in 2011 when Japan ran into trouble as a result of the earthquake and tsunami. Since late 2011 JPY has been very weak, in particular vis-à-vis USD, as risk assets have continued to rise.

The JPY story is multi-dimensional, but one explanation could be that, going into the financial crisis, JPY had become a favourite funding currency for investors (speculators) involved in the carry trade. Hence, when the underlying risk on assets were offloaded as a result of the sudden crisis environment, JPY had to be bought in considerable amounts to unwind all the leverage. More recently, the weakness of JPY is probably a function of Japanese monetary policy under Abe.

The simple lesson from all of this is that one cannot necessarily assume that risk off assets are always the same. What has worked in one crisis may not work in the next. Admittedly, I need to do considerably more work on identifying the asset classes most likely to be used for risk off purposes in the next crisis. Suffice to say, CHF looks to me like a favourite risk off asset going forward. After untying it from EUR, the Swiss have created what looks to me like a future safe haven asset in crisis times.  

Passive vs. active

When most risk assets are highly correlated, investment managers struggle to identify alpha, and nowhere has it been more painful than in the equity space. The result? An uproar amongst clients who can’t understand why they should pay the higher fees for active management when results are no better, and in many cases worse, than when investing passively.

Consequently, large sums of money have moved from active to passive equity mandates. Other asset classes have suffered as well but not to the same degree (chart 4).

The significant flow of funds from active to passive mandates implies a risk that is not widely understood. Passive equity funds pay no attention whatsoever to valuations. They simply buy the constituents that make up the index. It is therefore conceivable that, over time, the largest companies will become disproportionately expensive as a result. I have looked at projected PEs on the ten largest U.S. companies (chart 5) and conclude that, whilst valuation is not a major issue yet, it certainly has the potential to become one.

One final note on passive investing: It is not just in mature markets like the U.S. that passive investing promises to cause problems down the road. Foreign investments, many of which are passive, are flowing into China in big numbers. That illustrates how passive investing can potentially reward bad behaviour. China, after all, has the world’s biggest crisis of corporate governance. Recent scandals have highlighted:

• failure to enforce securities laws;
• abuse of conflicts of interest in state-owned enterprises;
• seriously deficient property rights.

In other words, any investor in a passive fund investing in China will be exposed to serious corporate governance risk. When companies in the index raise capital, as they do regularly, that exposure will increase. To make matters worse, when China is eventually added to the MSCI EM index, the affect will be accentuated.

Value vs. growth

For the very same reason that capital has parted company with active investment managers in recent years, growth stocks have outperformed value stocks. Only in 2012 did value outperform growth. In all other years since 2008 value underperformed (chart 6).

The reason is straightforward. As ever larger amounts of capital are managed passively, more and more capital flows towards the largest companies, and the majority of these are growth stocks – not value stocks.

This is yet another topic I plan to come back to in a future research paper. Suffice to say that the underperformance of value vs. growth has created significant opportunities in the value space. Having said that, in the short term, the drive towards passive investing may cause the underperformance to continue for a while yet.

Correlations between equities and their currencies

Before you read any further, take a quick look at chart 7. What does that chart tell you? If you are like me, not a lot at first sight. Then look at it again, but divide it into two parts - before and after 2008. An interesting picture emerges. It becomes evident that equity markets correlate very differently with the local currency today than they did prior to the financial crisis.

Take the U.S. equity market vs. USD. Prior to 2008 the correlation between the two was in a range of -0.25 to 0.25. Since 2008 the correlation has been very negative – almost -0.8 at one point, even if it has moderated a little bit more recently.

A similar kind of story has developed in Japan, although the trend changed earlier there, probably because the crisis unfolded earlier in Japan. In almost all other countries in the survey, the correlation has risen post 2008.

What does all of this mean?

As I prepared for this paper, I came across a research note prepared by Wai Lee, Ph.D. and CIO of Neuberger Berman called Risk On, Risk Off. If you are interested in the topic, I would strongly recommend you read Mr. Lee’s paper. He summarized the consequences for investors the following way:

“For Japan and the U.S., stocks typically move in the opposite direction of their respective currencies, while for all other regions the reverse is true, and stocks generally move in the same direction as currencies. As a result, performance […] depends entirely on whether the investor correctly identifies the set of common factors driving the two asset classes, as well as taking positions that are consistent with the investor’s view on how assets will behave as these factors change.

Furthermore, stock market selection and currency selection also collapse into the same strategy when the two assets in the same region are perfectly correlated and, therefore, have identical Sharpe ratios. In this environment, for those investors who continue to maintain their distinct processes of managing stocks and currencies, their performance may be confined to a relatively narrow range, as the chance of getting both stock and currency strategies right or wrong at the same time is lower given that the two strategies’ underlying investment processes are not built upon an identical set of factors.”

In essence, there have effectively been two clusters since 2008. One cluster around the risk off assets USD and JPY, where the correlation has been, and continues to be, strongly negative, and another cluster around the risk on assets, where the correlation is very high. This may continue until we leave the risk on, risk off environment.

Correlations between different investment strategies

As I have just documented, correlations between most equity markets and their currencies have changed quite dramatically post 2008, but it doesn’t stop there. Correlations between risk assets in general have been very high in the post-crisis environment, and global equities have also been highly correlated with most alternative investment strategies (and the two are definitely related).

One of the best proxies for hedge funds is probably the HFRI Composite Index and, as is pretty obvious when you look at chart 8a, I am not exaggerating when I suggest that global equities have been extraordinarily highly correlated with hedge funds in recent years. Now, we all know that a small number of strategies represent a high percentage of the hedge fund universe. Equities (long only, long biased and long/short) and macro strategies between them account for about 50% of all hedge fund AuM, and it is therefore not fair to conclude that all hedge funds are highly correlated with equities, but a surprisingly high number of them have been in recent years – and still are.

In our ongoing search for low correlation alternative investment strategies, we have only managed to identify two in the liquid space which have a correlation profile that looks profoundly different from the one depicted in chart 8a. One is Volatility (chart 8b) and the other one is Systematic Macro (chart 8c).

As it is a relatively new addition to HFR’s universe of hedge fund indices, Volatility’s history is much shorter than Systematic Macro’s, so one would have to be a little bit careful as far as conclusions are concerned, but both strategies appear to deliver that all important low correlation with equities in times of distress.

In that context I should also mention that a few strategies, although their correlation profile in principle look very similar to chart 8a, have experienced a more significant recent drop in the correlation with global equities than most hedge fund strategies have. Emerging Markets, Multi-Strategy and Funds of Hedge Funds would all fall into this category.

I also need to point out again that my analysis has only been done at the strategy level. There are unquestionably funds that deviate significantly from the investment strategy they are part of. In other words, just because equity long/short looks as if it is a strategy where the equity beta is way too high to attract our attention, doesn’t mean that there aren’t managers who can deliver on our objective of low equity beta.

One final comment re correlated returns: If you take another look at chart 8a, you may note that the correlation between global equites and hedge funds has finally started to dip, even if the drop so far is minimal. When we looked at about 15 alternative investment strategies that were all highly correlated with global equities, we found exactly the same pattern developing in most cases.

As a result, the obvious question to ask is the following: Has the world finally begun to return to some sort of normality, or is it a statistical fluke? We shall return to this question in our conclusion.

Focus on other risk factors

Another valid question given the current environment: If correlations are so high, in theory narrowing the range of Sharpe ratios that investment managers can possibly deliver, why do Sharpe ratios still in practice vary considerably?

The answer is simple. Because volatility is only one of many risk factors that impact returns, and that range of factors has expanded quite significantly since 2008.

A few examples: Investors have assigned much importance to instant liquidity since the bad days of 2008. This has had the effect of establishing rather large illiquidity premia on less liquid risk assets. Where the equity risk premium is probably around 4-5% (as a proxy for the risk premium on an asset class offering instant liquidity), if you would be prepared to tie up your risk capital for 7-10 years, it is not unusual for the risk premium to be at least twice that level.  

Regulatory arbitrage would be another example. The financial crisis led regulatory authorities in many countries to tighten the rules that govern banks. The result of that has been that more and more lending takes place outside the banking sector, which has also altered the risk premium.

I could go on. Absolute Return Partners have identified a good handful of factors which allow investors to pick up higher returns than you would strictly speaking expect, assuming you look at the highly correlated investment environment in isolation.

Conclusion

Overall, I don’t see any clear and unequivocal signs that the risk on, risk off mentality, which has ruled since 2008, is finally coming to an end. Yes, correlations have begun to recede a little bit; however, if it is indeed a sign of bigger things to come, it is still very early days.

Part of the problem for investors is that events in the real economy are not doing them any favours. Economic growth continues to disappoint, and incidents like the Greek tragedy leave an impression – rightly or wrongly – that the crisis is still very much on.    

The fact that some alternative strategies, as already mentioned, have experienced a material drop in their correlation vis-à-vis global equities is probably not a coincidence. Many emerging markets have been in a bear market over the last couple of years, as the combination of a strong U.S. dollar and record high EM dollar borrowing has created significant concerns. This probably explains the falling correlation more than anything else.

Meanwhile, Funds of Hedge Funds and Multi-Strategy have both benefitted from the fact that they have multiple tools at their disposal, and being stuck in a high correlation environment for many years has undoubtedly taught them one or two things.

However, if the objective is very low correlation with equities, in particular in times of distress, the hedge fund industry offers comparatively few opportunities in the liquid space at present, and will probably continue to do so for as long as the risk on, risk off mentality continues to rule.  

I do realise that I haven’t touched at all on the subject of expected returns. Strictly speaking, our objective is not one but two. Not only are we searching for strategies that are likely to produce uncorrelated returns, but those returns also have to be attractive. In a risk on, risk off environment, where correlations are high, it is, as I have already pointed out, much harder than it normally is to produce alpha.

It would therefore only be reasonable to expect those alternative investment strategies that suffer from the highest correlation with equities to also suffer from the lowest expected returns.

Niels C. Jensen

24 June 2015