Tuesday 28 July 2020

Market inefficiency, liquidity flywheels, asset class arbitrage, and Hong Kong Land

Conventional economic theory holds that the marketplace in financial assets ought to be 'efficient'. Large numbers of intelligent and diligent investors have access to largely the same pool of information, and are highly motivated to root out and exploit any underpricings that exist. It is believed this competitive process will inevitably drive assets to their fair value - i.e. those that accurately reflect their risk and reward characteristics, and also price assets correctly relative to one another.

If only the world were so simple. While this theory is seductive in its simplicity, it is also wrong in its simplicity. While much has been written about the role fear, greed, and various cognitive biases play in mispricings (also important), often overlooked or underestimated is the role structural institutional market factors play, which encapsulate traditional economic ideas of agency conflicts and information asymmetry. What is little understood - and I hope to argue in this article - is that market inefficiency is structural and behavioural, rather than informational, which is why I believe it will always exist, and provide ample opportunity for the well-heeled.

The efficient market hypothesis (EMH) misses one very important point - the majority of security/asset-level investment decisions are not made by the ultimate owners of capital, but instead via their appointed intermediaries - whether they be traditional active asset managers, or passive vehicles (which mechanically buy stocks in proportion to their index representation). The EMH implicitly assumes end investors are making direct purchases of individual securities; are fully informed about the idiosyncratic risk and reward characteristics of those individual securities; and are investing with time horizons that accord with the underlying duration of the cash flows of the assets they purchase.

However, in reality, most ultimate owners of capital are not making direct purchases of securities; are less than perfectly informed about the risk and reward characteristics of the underlying assets in the portfolio (only the realised returns and volatility outcomes from month-to-month); and are often investing with time horizons considerably shorter than the duration of the underlying assets - particularly for stocks (where durations average 50 years, compared to the typical time horizon of most investors, which is perhaps 3-5 years at most).

Ultimate owners of capital, or their direct appointed representatives (advisors or asset allocators), are at least one-step removed from grass roots individual security selection. Instead, they are more concerned with higher-level asset allocation decisions, which are influenced by perceptions of risk, reward, growth potential, and volatility associated with aggregated asset classes, which may be radically disconnected from the actual risk/reward characteristics of the individual securities comprising these larger clusters. They also often react procyclically to realised return and volatility outcomes, exiting assets classes that have exhibited poor returns and/or high volatility, and increasing allocations to assets that have performed well - particularly with limited downside volatility.

This has significant consequences for the way assets are priced, and at a system level, can result in truly massive levels of market inefficiency,* and this inefficiency can also be highly persistent, because the arbitrage forces that are supposed to correct mispricings are not only relatively ineffective in the short term, but frequently overwhelmed by structural flows-based feedback loops that have a tendency to amplify rather than moderate these inefficiencies.

More pointedly, markets have a structural tendency towards the emergence of what I describe as 'liquidity flywheels', which in my view is the largest and most under-appreciated contributor to market inefficiency ('liquidity' here is defined as 'liquid capital' that is available to be invested into various financial or real assets). Liquidity flywheels are discussed in the section immediately below. Later, I discuss how the distortionary impact of institutional asset allocation practices can lead to the same underlying assets being valued at very different levels with very different costs of capital, and use Hong Kong Land as an example.

Liquidity flywheels

What is a liquidity flywheel? A liquidity flywheel is a situation where inflows into an asset class lead to buying pressure that pushes up prices, leading to favourable apparent return and volatility characteristics in the said asset class. This favourable outcome then attracts yet more inflows, leading to yet more buying, etc. Conversely, poorly performing asset classes with significant downside volatility can lead to investor redemptions, leading to forced selling that contributes to yet further price declines, yielding even worse returns and even greater redemptions, and so on. This process can go on for years, and sometimes even for decades, and is a fundamental contributor - perhaps the most important contributor - to both major asset-class bubbles, as well as asset price busts and secular lows that lead to fire sales prices (which are 'anti-bubbles' driven by the same drivers of bubbles in reverse). The disconnect between the ultimate owner of funds and the at-the-coal-face investors actually engaged in individual security analysis is fundamental to this process, because end investors have little to go on other than realised investment returns and volatility, and it introduces both information asymmetries and agency conflicts that can drive radical market inefficiency.

An 'asset class' here can relate to a broad category such as 'venture capital' or 'equities', or (more commonly for equities), various classes of equities divided by geography, market cap range, style, sector, or factor. Industries or countries enjoying and/or expected to enjoy rapid growth, which are perceived as offering significant opportunities, are the frequent locus of inward liquidity flywheels, as people crowd into a theme to get 'exposure' to growth opportunities, and the price increases such flows engender have a tendency to be seen as validating the narrative, which acts to trigger yet more inflows.

A perfect example of a liquidity flywheel is what happened with WeWork, which saw its valuation get bid up to an astonishing height of US$47bn, even though the company was ultimately worthless, with little to show for years of operating losses other than the accumulation of billions of dollars of long term lease liabilities. It occurred because a seductive growth and new-economy disruption narrative, spearheaded by Masayoshi Son, resulted in the Vision Fund and other Venture Capital funds attracting large inflows, and the rush to put those funds to work drove the valuations of companies like WeWork sharply higher. That resulted in higher mark-to-market returns in the said VC funds, which lead to greater investor demand to pour money into VC funds to cash in on the boom. This lead to yet more money pushing valuations even higher.

The process also works in reverse for asset classes that get cheaper and cheaper over time. In theoretical EMH academic models, investors have access to unlimited capital, and if they spot a mispricing, they will buy undervalued assets, borrowing capital when needed frictionlessly at the 'risk free rate' to facilitate this process. However, in the real world, investors' buying and selling behaviour is constrained by the amount of capital they have at their disposal - both with respect to having too little cash, and too much - and their capacity to borrow is also constrained. Indeed, investors that borrow in size to buy cheap stocks are apt to be margined out and forced to liquidate their holdings - particularly because margin requirements are often increased by brokers as prices fall.

A fund manager might have a huge number of very cheap stocks they would love to buy, but if they do not have any available cash, they do not get to 'vote' on the market price by buying in the open market, as they lack the liquidity to do so - in the short term at least (longer term, you can reinvest dividends). Furthermore, if the said manager is suffering investor redemptions due to recent returns being poor, then regardless of the underlying managers' views on the long term attractiveness of individual securities, they will be forced to sell. It is therefore not uncommon for those most informed about the opportunities in undervalued securities to be actually selling them rather than buying, in direct contradiction to the EMH.

The opposite is also true for fund managers receiving large inflows - they must buy regardless of their personal views on the valuation appeal of stocks within their purview. It is perfectly possible they believe the stocks to be overvalued and yet still buy them in size, because they have to. Many fund managers are explicitly constrained in how much cash they can hold by their fund charter, but even for those managers that are not so explicitly constrained, if the said manager elects to hold a large amount of cash hoping for a better opportunity to buy, and markets continue to rise, they risk potentially catastrophic levels of underperformance, and so is a luxury they can ill-afford.

If the said manager runs a technology fund, their clients allocated them money because from an asset allocation standpoint, they decided they wanted exposure to the technology sector. If six months later the tech sector is up 50% and this individual fund is only up 10% because they are holding cash waiting for a better opportunity, the end clients are not going to be very happy, and in all likelihood will ask for their money back, so they can give it to a better-performing tech fund manager who is up 60%. As a result, institutions will also tend to buy regardless of price, and at the very least hold large index weight stocks in proportion to their index representation, to avoid the risk of underperforming.

What is notable about this process is how little all this buying and selling, and all the pronounced volatility in asset prices in drives (including major secular bull and bear markets), has anything at all to do with rational 'price discovery' in the traditional EMH sense. It explains why markets can be populated by highly intelligent, informed, and hardworking people, and also be grossly inefficient and absurdly volatile. Liquidity flywheels can go on for many many years, and are a frequent cause of long periods of underperformance for value managers, because it is the effect of liquidity flywheels that cause stocks to become undervalued in the first place, and there is no reason to expect a liquidity flywheel to suddenly reverse just because certain stocks have already become cheap. 

Much has been said about the dreadful performance of value over the past decade, and in particular the last few years, and while there are many causes for this (a topic for another blog post), what is most missing from the dialogue is a recognition of the fact that this is not unusual, and long stretches of value underperformance have happened many times before, and for largely the same reasons. It happened in the 1990s tech bubble, and it also happened in the 1960-early 1970s 'Nifty 50' bubble as well. Including the most recent tech/growth bubble, that is about once every 20 years. The major cause in all these cases was a liquidity flywheel that lead to certain parts of the market becoming extremely overpriced, due to a decade of procyclical inflows that drove multiples to the stratosphere.

The fundamental issue is that the ultimate owners of capital, or their immediate representatives (advisors and asset allocators), often make their asset allocation decisions on the basis of backward-looking return and volatility realisations, as without direct knowledge of the underlying securities, that is all they have to go on. And the important thing is that this process is inherently self-sustaining, not self-correcting, such that inefficiencies tend to get larger over time, not smaller (until an inflection point is reached, following which very dramatic reversals can happen), and the ability of investors to arbitrage these inefficiencies is largely absent.

Liquidity flywheels are one of the most important forces in markets, and one of the most under-appreciated. They are also the fundamental reason why momentum strategies work (until they don't) - something that the EMH also declared to be impossible, and yet which quantitative analysis of past market action clearly refutes. So long as there are liquidity flywheels in markets, momentum will be a strategy that works, provided one has a reliable means by which to determine when momentum has turned, because when momentum reversals happen, they happen big and fast.

Different prices and costs of capital for the same asset

Another important consequence of these institutional forces is that radically different costs of capital (and hence asset valuations) can emerge for the very same assets, depending on how they are packaged, and the differences are often not trivial. If the EMH held, the same assets should be priced in the same way, regardless of how they are packaged (they are, after all, the same assets), but in the real world, we often see very considerable difference emerge, and those differences can often go years, or even decades, without being arbitraged away.

Consider for instance a stock like Hong Kong Land (HKL SP). HKL has a market capitalisation of about US$9bn, and yet holds about US$37bn of real estate at its most recently appraised market value (end of calendar 2019). A meaningful proportion of HKL's real estate is super prime office buildings in Singapore and HK, such as the Marina Bay Financial Center in Singapore, and Exchange Square in Hong Kong. These are amongst the most desirable prime locations in HK and Singapore, and as a result almost always enjoy close to 100% occupancy and favourable ongoing rent revisions. The company has only very modest levels of debt (US$3bn), so high leverage is not the cause of the disconnect, and this modest level of debt is offset by the value of various development projects they hold, which are of a slightly larger magnitude to the company's debt burden. Assuming the latter two net, the stock is therefore trading at about 25c in the dollar of its unleveraged real estate holdings.

Now to be sure, it is possible that recent events - not just covid-19, but recent political developments in HK - have increased the risk of long term value erosion, should HK for e.g. lose its station as a desirable offshore financial center. However, this very large gap between HKL's stock market valuation, and the private market valuation of its assets, has persisted for a long time, and well before these recent developments. The stock is down some 50% from its 2017 highs, but even at its recent peaks, the stock was still trading at 50c in the dollar of the open market value of its prime real estate assets.

So why are the same assets attracting 2-3x fold differences in prices? It has a lot to do simply with the way the assets are packaged from an asset class standpoint, and the different costs of capital that apply to different investor constituencies and asset classes. HKL's super-prime real estate assets are valued on the books (and in the real world) at cap rates as low as 3% (and often closer to 2% net of costs and tax). In a world with zero interest rates, this is not an unrealistically low yield for super-prime assets with favourable trends in rent revisions, A-grade tenants, long lease terms, and an income stream that is inflation protected. Even a modest pace of 2% annual rental growth (below historical averages) would generate all-in after-tax returns of some 4%, which with inflation protection, is very attractive relative to bonds and other high-grade debt.

However, the issue is that active equity managers - who are the natural buyers of HKL stock - are not benchmarked against cash and high-grade bond returns, or even high-grade real estate. HKL, as a listed company, is part of the "listed equities" bucket (and more specifically, the "Asia ex-Japan listed equities" bucket), and the performance of the institutions that purchase HKL are therefore benchmarked against equity market indices, rather than cash or bonds. The average stock in Singapore and HK is currently priced (in my estimation) with an expected return of perhaps 10%, which is another way of saying the cost of (listed) equity in these regions is currently about 10%. If equity managers were to buy HKL and realise only a 4% return (which they likely would if the stock was priced at 1x book), but the index was to generate 10% pa, it would be of little benefit to the fund manager to argue to their clients that the returns are low risk and quite attractive relative to fixed income. The clients would say, we allocated you money to get "equities exposure", and you're only up 4% and the market is up 10%, and that is not satisfactory performance. Because HKL is part of the "listed equities" bucket, it is expected to deliver "listed equities" returns of 10%.

Consequently, listed on public markets, the assets that underlie HKL are priced with a cost of capital reflective of Singaporean and HK equities in general, which is a cost of capital that bears no relation to the cost of capital private buyers of prime A-grade real estate are subject to. Because HKL's underlying assets generate only perhaps 4%, this requires the stock trade at about one third of book value (ignoring recent declines which are covid/HK related). Another way to think about this would be to imagine a company that owned only 30-yr treasuries yielding (say) 3%. If the stock traded at 0.3x book, it would offer a return of 10% instead of 3%. The underlying assets and cash flow stream are the same, but the costs of capital are different (here, 3% for long bonds, 10% for equities).

The cost of capital for listed equities is higher because equities are volatile, and in general, the cash flows of businesses (averaged across all types) are much less dependable than those of real estate. The problem is that from a fundamental/operational perspective, stocks are not "more risky" by definition, because it very much depends on what type of assets the company holds. Individual companies can range from the most speculative, risky biotech startups or resource exploration company, to holders of some of the lowest risk, most dependable cash-generating assets in the world (A-grade real estate, franchised incumbent businesses with huge moats and steady cash flow streams). In this respect, the very idea that "stocks" in general are an "asset class" is deeply flawed - they are an aggregation of individual businesses with hugely variant risk profiles. However, in the world of industrial scale asset allocation, stock-level differences are typically lost in translation when large scale asset allocation decisions are made between 'listed equities', 'fixed income', 'alternatives', 'developed market equity vs. emerging market equity', etc, and then results are assessed against an overall equity benchmark.

The ultimate effect of this is that the cost of capital that is applied to assets can come to reflect less the idiosyncratic nature of the underlying assets, and more simply the asset class packaging. Put long bonds or high-grade real estate into a company and then list that company on the stock market, and suddenly the same underlying bonds or real estate are transformed into a "listed equity", and because a higher "listed equity" cost of capital is then applied to those assets, the entity will trade at a steep discount. Change the packaging, and suddenly you see the same assets trade at radically different prices.

Repackaging listed real estate into a REIT vehicle (Real Estate Investment Trust) can sometimes bridge this gap. There are funds that can invest in REITs and benchmark their returns to the overall performance of REITs, and/or other yield-based strategies/assets ('yield alternatives'). This can lower the cost of capital by changing the benchmark for comparison from stocks in general, to fixed income or a specific class of yield stocks. It is for this reason that companies like HKL will sometimes spin off real estate holdings into REITs to 'unlock value'. In an efficient market, there ought to be no such opportunity to 'unlock value', but in the real world where institutional realities can create large differences in the cost of capital for the same assets, the opportunity is very real.

But let's go further. Let's suppose HKL's yield assets were instead packaged into an unlisted/illiquid "alternatives" asset class manufactured by a private equity sponsor, which marketed the vehicle as a low-risk yield product that aimed to generate a superior yield to those available on bonds/fixed income. The assets could also be leveraged up to juice returns and increase the manufactured yield from 2-4% unlevered to perhaps 4-6% levered. The 4-6% yield could then be marketed to institutional investors and asset allocators looking for replacements for their miserly bond yields. When the basis for comparison becomes long bonds at 1%, rather than equity benchmarks, suddenly 4% looks good. The result? A 4% cost of capital is used instead of 10%, and the same assets are valued at radically different prices, depending on the packaging.

The reason this alchemy can work is that there is an institutionalised aversion to volatility. Volatility is not just uncomfortable, but it is a career/reputation risk to whoever decided to make such an asset allocation decision - i.e. a financial advisor, or asset allocator with fiduciary responsibilities. If somebody puts capital into HKL stock and the stock falls 20%, that looks bad. It looks like a 20% loss, even if the underlying assets are still generating the same cash flows as before, and the dividend yield of the stock has merely risen. This is where the duration mismatch issue rears its head - if one buys and holds HKL for 50 years, temporary share price volatility won't make a shred of long term difference - in fact share price declines will be positive, as dividends can be reinvested at a higher yield. However, in reality, the duration of the career interests of advisors and asset allocators is much shorter than that - they are concerned with how returns fare over the the next few quarters and years, how that looks to clients, and how that is likely to influence their compensation.

Consequently, the minimization of volatility/drawdown risk is highly prized, and by keeping the assets 'private', these "alternative" vehicles can provide investors with the comfortable and expedient illusion of a lack of volatility. HKL's stock price, for instance, has fallen some 50% in the past 3 years, as the market has repriced its assets from an expected return of perhaps 6% to 10%. The company's dividend has remained stable, with the stock's yield increasing from 3% to 6%. By keeping the assets private, however, you can report the underlying cash flow returns of 2-4% (leveraged up to 4-6%), and claim that the value of your assets has not changed. After all, the private market value of the assets has not fallen (it has not for HKL either). The end results is an attractive yield for clients of 4-5%, without having to bear the risk of explicit volatility/drawdowns.

The desire to avoid looking bad by suffering such a drawdown is what is driving a huge wave of money into private equity and other "alternative" vehicles, which are often just a repackaging of assets you can buy on stock markets, with higher leverage and much higher valuations, but with the ability to spare clients of the appearance of volatility/drawdowns. The cost of capital with this newfangled packaging is structurally lower because returns are not benchmarked against listed equities, but instead cash and bond yields, and other low risk yield alternatives. Because interest rates and bond yields are so low, 4% returns appear quite attractive (vs. say 1%). Investors are happy. They allocate more capital, because 4% beats their 1% cost of capital. Again, it is important to emphasise that these are exactly the same assets, just with different packaging, and the packaging can change the cost of capital dramatically. This is a very clear refutation of the EMH.

There are some arbitrage forces in markets that can help correct these differences, but they are highly imperfect. If HKL were smart, for instance, they would take advantage of the large arbitrage opportunity that exists between the private and public market costs of capital by selling some of their real estate assets and using the proceeds to buy back stock. Activist investors and corporate take-outs can play a role, but in HKL's case, it already has a controlling shareholder (Jardine Matheson), which forecloses the opportunity for activism.


It is important to understand that market inefficiency is structural and behavioural, not informational. Many investors attempt to invest on the basis that market inefficiency is informational in nature, and dedicate tremendous amount of time and resource to trying to come up with better information than the next guy. However, in today's markets, the primary source of inefficiency is structural/agency driven, and the way to exploit that is not to acquire better information, but to have a structure that allows one to engage in long term value arbitrage that other investors cannot (often taking the form of buying underlying assets that are actually low risk, but are priced as if they were very high risk because they are part of an asset class that is generally perceived to be high risk). This requires a wide and unconstrained mandate (by geography, asset class, etc), long term capital, a rigorously long term approach, and an extreme tolerance for volatility and benchmark variation, which requires patience and emotional fortitude that is sorely lacking in today's instant gratification world.

Outperforming in the long term is actually not very difficult, but it requires highly lumpy results, often marked by long periods of lackluster returns, punctuated by short periods of spectacular results, which happen alongside liquidity flywheel/momentum reversals, which are inflection points that do not happen very often. Furthermore, usually, the worse value is performing, the closer one is to the end of a liquidity flywheel bubble cycle (value had a woeful time in 1999, for instance), because value is the 'anti-bubble' expression - a Newtonian equal and opposite reaction - of liquidity flywheels driving bubbles elsewhere in markets. It is redemption flywheels that drive value opportunities, and redemption flywheels are often the result of investors pulling money out of unpopular areas of the market in a rush to get exposure to hot areas of markets.

Taking advantage of long term opportunities is extremely difficult for investors with the wrong structure and wrong investors, however, as massive redemptions will likely happen right when opportunities are most ripe, and not all funds will be able to survive the inevitable lean periods. Value funds shut down en mass in 1999, for instance, and the same thing has been happening of late. Lean cost structures and - ideally - large principal FUM participation from the fund managers is a must.

Outperforming in the short term with consistency, by contrast, is extremely hard. The best way to do it is usually a momentum strategy, which works most of the time, but occasionally yields disastrous results on sudden momentum reversals. Momentum is the polar opposite of value - it generates good returns most of the time, and disastrous returns a minority of the time.

The latter strategy is a more remunerative strategy for fund managers, however, even if it often leaves long term investors worse off, which is why it is more popular/common. While the good times roll, large performance fees are banked, and it is investors that are left with the losses when it all turns to custard. This is why value investing remains relatively uncommon, despite its long track record of success, and in my view a combination of agency conflicts, information asymmetry, volatility-phobia, and the desire for quick results, will all but ensure market inefficiencies continue, and considerable opportunities for long term value investors will remain for many generations to come.


Disclosure: I do not own shares in HKL

*In my perception, a lot of investors confuse rising prices with increased market efficiency. Prices becoming more expensive, in general, has nothing to do with efficiency, even though it does have the practical effect of making value opportunities harder to come by. People also confuse the fact that value doesn't work in the short term - often for many years - with market efficiency. It is not the efficiency that makes it hard to outperform in the short term, but the inefficiency - prices are driven by flows, not fundamentals, often for many years, and so even very diligent and accurate valuation work can often not be rewarded by markets for years, leading to an erroneous perception of market efficiency.

Postscript: Just as this note was going to press, I noticed that HSKE-listed Dongfeng Motor (489 HK) - a company I blogged about in 2018 here - was trading up as much as 30% as the company announced it was planning an A-share listing on the frothy ChiNext exchange. A broker report noted (in an understated comment to say the least): "ChiNext-listed stocks are trading at high valuations (57x trailing 12-month PE average as of 27 July...), which should boost the valuation of Dongfeng's H-shares (trading at 2.8x FY20E PE)."

I cannot think of a better demonstration of many of the points raised in this article - different costs of capital for the same assets in different markets, and the effect of long term liquidity flywheel cycles. Positive liquidity flywheels have driven A-share valuations to extremes of 50-100x P/Es, and negative liquidity flywheel cycles have driven certain H-shares to as low as 2-3x. Dongfeng and H-shares generally (excepting the tech sector) are now 13 years into a liquidity flywheel downcycle, which started in 2007. Since my blog post, Dongfeng has continued to post strong operational results, earnings growth, and increased dividends. The stock has fallen from HK$9 to HK$5.