Monday 2 September 2019

A duration-bubble or a low-volatility bubble?

I'm a fan of Antipodes Partners, and always find their commentaries insightful. In the company's latest investor update (see below), founder and CIO Jacob Mitchell provides a number of useful insights, but one of the arguments he makes is that investors have been 'chasing duration' to an unprecedented extent in the post-GFC era. Slide 7 is quite revealing. Mitchell's full presentation runs from 5.50 to 22.00 (all of which I recommend), with the relevant section for this article on duration running from 10.05-14.45.

It's an interesting argument. However, I'm not convinced Jacob's characterization of today's bifurcated markets as being primarily a function of investors chasing 'duration' is entirely accurate. I think what has really been happening is that in the post GFC era, investors have been chasing asset classes/factors/styles which have (or are perceived to have) low volatility/drawdown risk, and concomitantly eschewing those with higher volatility/drawdown risk. It has little to do with duration per se, in my view. After all, per 11.15-14.45/slide 8 of Mitchell's presentation, the duration of the cash flows associated with ownership of European bank equities is longer than many of the low-yielding debt instruments used as a comparison (in an otherwise very good argument).

The origin of this more than decade-long megatrend lies in the wreckage of the Global Financial Crisis (GFC), which was a formative event for an entire generation of investors and asset allocators. It was a 'return shock' of tremendous proportions, where pretty much anything exposed to equities was down 50% or more on a marked-to-market basis in as little as 6-12 months. Furthermore, unlike the bust, there was nowhere to hide. In 2000-02, lots of old-world value sectors/stocks surged as tech stumbled, and emerging market equities also boomed during 2000-2007. The GFC was different - 'correlations went to 1', as they say, and equities everywhere tanked. Many value guys were also heavily exposed to financials and suffered badly. This GFC experience has cast a long shadow, and has had significant implications on how markets have functioned ever since.

It has always been the case that meaningful portions of the investment world cannot afford to sustain large mark-to-market (M2M) drawdowns, but the GFC and a number of other developments I discuss below have meaningfully amplified this behaviour. Insurers, for instance, are required to mark equity investments to market, and large drawdowns can therefore pressure regulatory capital and imperil rating-agency product ratings, which can in turn impair their ability to write new business (or even continue as a going concern).

As a result, insurers remain focused on high-grade fixed income securities, almost irrespective of yield and fundamental risk/reward, due to the relatively low associated M2M risks, and the extent of the GFC drawdown has caused them to allocate even less to equities, and particularly high-risk equity exposures. However, as sufficient yield has become increasingly had to come by in the QE-addled post-GFC period, they have sought to supplement their fixed income portfolios with increasing allocations to 'alternatives'. These initially included hedge funds, but in more recent times, the preference has migrated towards private equity and venture capital.

Pension funds have behaved in a similar manner post-GFC - albeit not to the same extent, and they retain much larger equity exposures. In theory, pension funds ought to have the best ability of all market participants to withstand short term M2M volatility and invest in long-duration assets such as stocks, to match off against their long-duration liabilities. However, the duration of pension funds' liabilities usually exceeds the duration of the employment tenures of most of the people running them, and no one wants to preside over a significant portion of the assets backing their pension liabilities declining by 50% on their watch. Because many organisations saw exactly that happen to their equities exposure in the GFC, a major aversion to equities in general, and more volatile/economically-sensitive sectors/regions/style factors in particular, has developed.

In the post GFC period, hedge funds were initially seen as an attractive alternative asset class, as they promised to provide equities exposure with downside volatility protection - particularly because several hedge funds succeeded in not only protecting capital during the GFC, but also profiting mightily from 'the big short'. However, due precisely to many of these funds' focus being on hedging downside volatility/drawdown risk, they have also relinquished most of their returns over the past decade as well. GMO's research has shown that the long term return on writing cash-covered put options on market indices has been about the same as the market's total return. The implication of this is that you have historically needed to give up 100% of your expected return being long equity indices in order to hedge the downside by buying index put options. With equity indices - particularly in the US - having risen incessantly over the past decade, hedge funds have accordingly delivered poor returns for this reason, exacerbated by limited alpha generation and high fees.

The GFC also had the effect of causing many active managers to move from under-estimating the likelihood of crises before the GFC, to dramatically over-estimating the likelihood in the GFC's wake - particularly because the GFC set in motion various regulatory/policy/behavioural responses designed specifically to prevent a recurrence (including much tighter banking regulation, and much more conservative lending). The GFC catapulted to fame and fortune many hedge fund managers with relatively short track records that had successfully bet on disaster. Funds flooded into these managers. But their poor subsequent records - as they have continue to bet on myriad other macro blow-ups which have not occurred - indicate that they were mostly just lucky pre GFC, rather than prescient. Many copy-cat macro-oriented hedge funds also emerged that tried to be the next Paulson or Burry, who bet heavily on various crises occurring which did not. This 'generals fighting the last war' approach cost their investors a lot of money, and discredited hedge funds. Investors fled.

With hedge funds disappointing, and with yields on fixed income instruments - the ultimate asset class providing a refuge from 'volatility risk' - declining to a mere pittance (if that), money has instead rushed into private equity and venture capital. This is intriguing, because private equity is essentially little more than leveraged equities exposure - albeit with greater ability to influence management behaviour and capital allocation - while venture capital is also involved in the riskiest equity bets of all - funding start-ups, most of whom are destined to go to zero. It appears to be a strange place for investors to seek out low-risk returns.

However, the risk investors are seeking to avoid is not fundamental/valuation risk, but mark-to-market drawdown risk - particularly over shorter time horizons (by the time things blow up, the management teams in charge of making making asset allocation decisions may have already retired with their bonuses safely banked). The advantage PE/VC 'asset classes' have over traditional equities is that they don't need to mark their positions to market every day/month/quarter, and are therefore able to manufacture the illusion of returns without volatility - something that has considerable practical value for many of their volatility-phobic investors. Tech VC also has the further advantage of being able to mark the carrying value of their investments to the latest capital raising valuation - even if they are the only fund participating in such a raising.

However, the impact of this volatility-phobia has not just been a reallocation of funds from equities into 'alternatives', but also into equities exposures that promise returns with relatively limited volatility/risk. This has included a rotation into 'bond proxy' equities such as consumer staples, utilities, etc; dependable structural growth stories such as large cap tech; and safe jurisdictions without Fx risk such as the US; while avoiding sectors subject to macro sensitivity or other uncertainties, and jurisdiction with greater political, economic, or Fx risk.

This aversion to volatility/risk has also often been effectuated through the use of ETFs coupled with 'smart-beta', 'risk-parity' and 'factor-based' investment strategies, while investors have also favoured active managers delivering performance with limited drawdowns. This volatility-phobic trend has been accentuated by robo-advisers and other algorithmic asset allocation strategies, which all seek to optimise the same equation - relatively high returns with relatively low 'risk', with risk defined as volatility/drawdown risk (i.e. the classic 'sharp' ratio). And the algobots assess these characteristics using historical data series.

What this has meant in practice is that equity market 'factors' that have been delivering good returns with relatively limited volatility/drawdowns have been attracting more and more money. These inflows have then in turn contributed to further strong returns with comparatively limited drawdowns (as when markets swoon, incessant inflows moderate the size of the drawdown vs. factors experiencing outflows, which fall even further in market sell-offs). This flywheel effect has accentuated and prolonging pre-existing market momentum (kickstarted by a flight to quality/low-volatility in the risk-averse climate of the GFC's aftermath), and tricked algobots into believing certain 'factors' such as 'growth' and 'quality' have better enduring risk/reward characteristics than other factors such as 'value', when in reality, it merely reflects a bubble in flows.

An prime example of the folly is 'low volatility' ETFs - ETFs that invest in 'low volatility' stocks, which have grown significantly in popularity. Imagine if a significant fraction of the stocks comprising low-vol ETFs came to be held by 'low-volatility' ETFs. What would happen if the stocks suddenly spiked down and became highly volatile? Two things: (1) the ETF would have to sell them, because the stocks are no longer 'low volatility'; and (2) the investors in the low-volatility ETF would experience highly-volatile outcomes. They would say, 'this is not what we signed up for', and pull their money out. In combination, this could drive a crash in 'low volatility' stocks. It would be hard to design an ETF more foolish and prone to a total blow up. But until that inflection point is reached, so long as more money keeps flowing into low volatility ETFs, the stocks will perform well and won't go down, as the inflows continue to support their 'low volatility' performance.

Incidentally, this is one of the fundamental problems/dangers with 'big data' in the investment world. Big data in markets is different to big data in, say, baseball (I highly recommend the book 'Big Data Baseball', which is Moneyball on steroids). Big data in baseball makes it possible to more and more accurately appraise the true value of individual players (which as discussed in the book, has hitherto sometimes been hidden), but although it may affect player selection and on-field strategies, it does not ultimately affect those underlying contributions/attributes. Furthermore, virtually 100% of what happens during a baseball play has happened before - often thousands and thousands of times. Algos are great at analysing this sort of voluminous data and sifting out hidden relationships/patterns invisible to mere mortals. In these sorts of contexts, big data can have a huge impact, and make the market in baseball players (and on-field strategies) far more 'efficient'.

Financial markets don't work that way, because the return and volatility data sets are backwards looking, and the algos cannot anticipate discontinuities, or something happening that has not happened before in the datasets. History, and our data collection to date, has simply been too short to account for the myriad of all possible outcomes. If there had already been 500,000 housing bubbles and busts, maybe the data sets could predict future bubbles and all the relevant factors, but there haven't been. There had never been a national housing bust in the US before the GFC, so it wasn't in the data, and was therefore invisible to the bots. If you asked the algos in 2005, they would have told you a nationwide crash was a 10-sigma event almost impossible to occur. But they were wrong, as the past data sets reflected ex post outcomes, not ex ante probabilities in a novel situation. It's like setting baseball 'big data' bots to work when they only have three pitches worth of data. If all three pitches have been strikes, then the bots will think a strike happens every ball.

Furthermore, the very behaviours these bots drive can actually change the future probabilities in a reflexive and (to the bots) unpredictable way. The very belief that a nationwide housing crash was a 10-sigma event was a key contributor to the creation of the housing bust (as it caused ratings on MBSs to be far too high, enabling an excessive flow of credit into the sector). It was an active contributor, not just a failure of passive prediction. In much the same way, the algobot search for high returns with low volatility/risk is creating a situation where the stocks with the best best returns potential and with the lowest risk actually reside in the exact same stocks the bot algos eschew, and vice versa.

It hasn't just been algobots driving a bubble in low volatility, however. The behaviour of both end investors and active equity managers has also reflected these megatrends. Aging demographics means that there is a large cohort of investors approaching retirement age in development markets, and the GFC was a major shock to the retirement accounts of many of this cohort. They can ill-afford to see 50% of their retirement savings wiped out again on the eve of their retirement. This large class of capital is therefore now very sensitive to volatility and drawdown risk, and this risk aversion is impacting not just their asset allocation, but also their selection of asset managers and the types of equity exposures they are prepared to run: they have shifted funds into managers delivering steady gains with low volatility, and away from those with variable returns, and/or have allocated funds to ETFs or 'smart beta' ETFs with similar dynamics. But it is this very behaviour, writ large, that has been driving the high returns and low volatility in the fist place.

End-investor behaviour has also affected the behaviour of active fund managers, because a long-standing issue for equity managers has always been that they are only able to be as patient and volatility-tolerant as their end investors. If their investor base cannot 'wear' downside volatility, these organisations will actively avoid asset classes, regions and sectors that expose their funds to meaningful short term drawdown risk. They will do this because even if buying cheap but volatile stocks will lead to a better long term outcome for their investors, it represents an unacceptably high business risk. If they suffer massive redemptions, they may go bust or (from the perspective of the individual PM) be laid off long before the point of reversal arrives.

Exacerbating these trends further, the world of institutional funds management has been consolidating over the past decade, as more and more money has flowed into passive strategies, and fee compression has emerged. Many smaller shops have folded, and many large managers have been merging in order to reduce costs (and therefore fees). However, this has resulted in larger and larger funds, and more and more money flooding into the same large-cap structural growth or quality/low-volatility names, pushing up multiples further.

All of these forces in combination have resulted in crowding into the same, now highly-expensive segments of the market, and the wholesale abandonment of other sectors, and in my submission is the real reason we see the wide disparity in multiples and return profiles Antipodes has ably referenced. Its not that investors aren't prepared to look for structural growth in 'less obvious places' - it's that they aren't prepared to/can't afford to weather short term drawdowns.

Sure, dividend yields on European banks in sound financial positions are now 7% (on 50% dividend payouts), vs. close to zero for junior debt instruments owed by those same institutions. But if you buy the stock, you might be down 5-10% in a day on a marked-to-market basis (as occurred in August), whereas that will not happen for a held-to-maturity debt instrument (debt instruments can be accounted for as 'held to maturity' by financial institutions, and hence not marked to market (unless 'impaired'), but there is no such thing as 'held to maturity' for equities). It's not about duration - it's about mark-to-market volatility risk. GE and Siemens offer structural growth in less obvious places at a much lower price? Sure, but you might be down 5-10% in a day if people worry about the outlook for global growth, or the size of GE's LTC insurance reserves.

This is how you end up in a world where many markets are expensive, and nominal long bond yields are approximately zero, but you can still buy some stocks on extraordinarily low prices - often market leading companies with long term growth potential with dividend yields of 10%, but which happen to be cyclical or subject to some other form of uncertainty that is inducing share price volatility. It makes no sense, until you recognise that most investors simply can't weather the volatility, and algobots will dump stocks exhibiting poor short term Sharp ratio optimisations in a self-reinforcing manner. It's also how you end up in a world where the vast majority of value investors with great long term records are underperforming the indices - sometimes by a sizable margin.

The result is that there are now extraordinary opportunities emerging in global markets for the volatility-tolerant long term investor. The world has a vastly excessive abundance of low-volatility capital, but a growing and acute shortage of high-volatility capital. Like all opportunities in markets, however, the magnitude of the opportunity is proportional to the challenges of successfully exploiting it at scale - it is in this respect that markets are both highly inefficient, and offer no free lunch (something the academics missed). The opportunity is large because it's hard to attract a lot of money at the moment to implement such a strategy - most of the funds doing so are currently suffering redemptions. If it were easy exploit, the opportunity would disappear. That's how markets work.

Personally speaking, I have the good fortune of being a small-time manager with a low cost base and my own permanent equity capital invested, and only need a handful of loyal, long term investors to make a decent living. As far as I'm concerned, the current environment is a wonderful opportunity if you're fortunate enough to be in a position to exploit it.