Global Investment Outlook 2018 — The Volatility Unwind

2018 opened by demonstrating the mechanism the 2017 report described.

In early February, volatility rose from historically low levels. The equity decline that accompanied it was significant but not extraordinary by historical standards. The losses in volatility-linked products were extraordinary — several instruments designed to profit from calm markets lost the overwhelming majority of their value within days, and some were terminated outright.

The disproportion is the point. The size of the loss was determined by positioning, not by the size of the market move. Short-volatility exposure had accumulated over a long calm period precisely because it had been profitable. When volatility rose, that exposure had to be covered, and covering it required buying volatility — which pushed volatility higher, which forced more covering.

The rest of 2018 was defined by two further developments. Trade policy moved from background condition to active variable, forcing a reassessment of businesses whose economics depended on cross-border supply chains. And central banks tightened — raising rates while beginning to reduce balance sheets — which produced a year in which both equities and bonds struggled at once.

That last observation is the most under-remembered feature of 2018 and the most relevant in hindsight. The assumption that bonds cushion equity declines held for most of the post-crisis period. In a year where the driver was tighter policy rather than weaker growth, it did not. That was a preview, at small scale, of 2022.

Anatomy of a positioning event

Distinguishing a positioning event from a news event is one of the more useful skills in reading market dislocations, and February 2018 is the textbook case.

A news event involves new information that changes the fundamental value of assets. The price moves to reflect the new information and stays there. The magnitude relates to the significance of the news.

A positioning event involves no meaningful new information. The move is caused by participants being forced to transact — by margin calls, risk limits, or mechanical rebalancing. The magnitude relates to the size and concentration of the positions, not to any fundamental development.

The sequence in February ran:

1. Volatility rose modestly from very low levels, for ordinary reasons.
2. Products designed to profit from low volatility began losing money.
3. Their hedging rules required buying volatility as it rose.
4. That buying pushed volatility higher.
5. Higher volatility caused larger losses, requiring more buying.
6. Meanwhile, volatility-targeting strategies elsewhere saw measured risk rise and mechanically reduced equity exposure — selling into the decline.

Steps 3–5 form a feedback loop with no external input. Nothing about the economy changed during the episode.

The market move was ordinary. The loss was not. The difference was entirely in what people owned before it started.

The practical lesson is a question worth asking of any crowded position: what would this strategy be forced to do if it started losing money, and how many others would be doing the same thing at the same moment? A strategy whose loss-response is to transact in the direction that worsens the loss is a strategy whose risk is not captured by its historical volatility.

Trade policy becomes a variable

2018 saw trade policy move from background condition to active market variable through tariff measures and the responses to them.

The reason this mattered beyond the directly affected goods is that decades of trade liberalisation had shaped how businesses were built. Supply chains were optimised for cost, which meant locating each stage wherever it was cheapest, on the assumption that goods could cross borders freely and predictably.

When that assumption weakens, several things change at once:

• Cost structures change for businesses with cross-border production, directly and immediately.
• Optimisation has to be redone with a different objective — resilience alongside cost — which is expensive and slow.
• Capital investment decisions become harder, because a plant built to serve a global market may be stranded if trade conditions change again. Uncertainty about future policy suppresses investment independently of current policy.
• The valuation of supply-chain exposure changed. A business with concentrated single-country production acquired a risk that had not previously been priced.

That third effect — the suppression of investment by uncertainty rather than by actual measures — is the one most often missed. A company facing an unresolved question about future trade rules may delay a decision regardless of how the question resolves. The cost is incurred during the uncertainty.

This proved durable. The reassessment of supply-chain concentration that began in 2018 was reinforced by the pandemic in 2020 and has continued since. What started as a trade dispute became a structural change in how manufacturing footprints are designed.

When both sides of the portfolio fall

2018 produced a genuinely unusual outcome: both equities and high-quality bonds delivered poor returns.

The standard balanced portfolio rests on an assumption about correlation. Equities carry growth risk; bonds are expected to rise when growth disappoints, because weaker growth prompts lower rates. The two are expected to offset.

That relationship holds when growth is the dominant driver. It fails when policy or inflation is the driver, because tighter policy hurts both simultaneously — equities through the discount rate applied to future earnings, bonds through the direct effect on yields.

2018 was a policy-driven year. Central banks were raising rates and reducing balance sheets against a backdrop of adequate growth. The result was pressure on both asset classes.

The episode was mild compared to 2022, but the mechanism was identical, and it should have prompted a broader reassessment of the diversification assumption than it did. It did not, largely because the losses were modest and the following year was strong — which is a recurring pattern. A mechanism that produces a small loss teaches less than one that produces a large one, even when the lesson is the same.

Quantitative tightening begins

The reduction of central bank balance sheets began in earnest, and the asymmetry deserves noting.

Quantitative easing worked partly through a portfolio-balance channel: a central bank buying government bonds removed them from private hands, and those holders had to buy something else, pushing capital into riskier assets and supporting their prices.

The reverse was expected to work symmetrically. It did not, for a reason worth understanding: QE was announced and predictable, while its withdrawal interacted with market conditions that had themselves been shaped by QE. Assets held by investors who had been pushed into them by the shortage of alternatives were not the same as assets held by natural owners. The order of unwinding mattered, and it could not be predicted from the order of accumulation.

The broader lesson — that the exit from an unconventional policy is not the entry run backwards — has recurred each time balance sheet reduction has been attempted since.

Telling the two apart in real time

The distinction between a positioning event and a news event is analytically clean and operationally difficult, because both look the same on a price chart. It is worth setting out how they can be distinguished while they are happening, since that is when the distinction is worth money.

Signals that a move is a positioning event:

• The move is disproportionate to the news. If the price move is large and the identifiable new information is small or absent, positioning is doing the work.
• The affected assets are related by ownership, not by fundamentals. A positioning event propagates through who owns what. If assets with no economic connection are moving together, the connection is a common holder.
• Volume is concentrated and one-directional. Forced selling produces sustained one-way flow rather than the two-way trading that follows genuine disagreement about value.
• The move accelerates rather than decays. A news event's price impact typically decays as the information is absorbed. A positioning event accelerates, because each increment of price forces the next increment of transacting.
• Liquidity deteriorates in a specific pattern. Market makers widen spreads when they cannot identify the informed side. In a positioning event there is no informed side, which paradoxically makes liquidity worse rather than better.

Signals that a move is a news event:

• Identifiable new information of a magnitude commensurate with the move.
• The affected assets share an economic exposure to that information.
• The price stabilises at a new level rather than reverting.
• Two-way trading as participants disagree about the implication.

Why the distinction is worth money: a positioning event creates opportunity, because prices move away from value for reasons unrelated to value and revert when the forced transacting completes. A news event does not — the price has moved to a new correct level and will stay there.

The reflexive response to a sharp decline is to ask what went wrong. The better first question is what was owned, because in a substantial fraction of sharp declines the answer to the first question is "nothing".

What an allocator could act on

2018's mechanisms translate into a small number of checkable practices.

Map the loss-response of every crowded position. The 2017 report frames the question; 2018 demonstrated the cost of not asking it. A position whose response to losses is to transact in the loss-worsening direction should be sized smaller than its measured volatility suggests, and the adjustment should be structural rather than tactical.

Do not assume bonds will cushion equities. 2018 provided the first clear demonstration that the negative stock-bond correlation depends on growth being the dominant driver. In a policy- or inflation-driven year it fails, and it fails at the moment the cushion is most needed. A portfolio whose risk model assumes a stable negative correlation is under-hedged in exactly the scenario that breaks the assumption, and the correction is not to abandon bonds but to hold the correlation as conditional rather than fixed.

Price supply-chain concentration explicitly. 2018 established that a business with concentrated single-country production carries a risk that had not previously been priced. That risk is assessable from disclosure — where production sits, what proportion crosses which borders — and it entered standard diligence after 2018 for good reason.

Distinguish policy uncertainty from policy content. The investment-suppressing effect of unresolved trade rules operated regardless of how the rules eventually settled. An analysis focused on which outcome would occur missed the cost being incurred while the question remained open. Uncertainty is itself a variable with an economic cost, and it is measurable — the trade policy uncertainty indices track it directly.

Treat unconventional policy exits as non-symmetric. Quantitative tightening was not quantitative easing reversed, and no subsequent balance sheet reduction has been either. The general principle — that the exit from an unconventional intervention interacts with conditions the intervention itself created — applies to any large-scale policy that has been in place long enough to shape who owns what.

What 2018 established

• Positioning events were demonstrated to be distinguishable from news events, and the distinction determines whether a dislocation is an opportunity or a warning.
• The 2017 fragility resolved as its mechanism predicted, confirming that endogenous risk is identifiable in advance.
• Trade policy re-entered the risk framework as a first-order variable, with supply-chain concentration becoming a priced exposure.
• The equity-bond diversification assumption showed its first crack in a policy-driven year — a preview of 2022.
• Quantitative tightening was shown to be asymmetric with easing rather than its mirror image.

Why the 2018 correlation lesson was ignored

2018 delivered a clear demonstration that equities and bonds can fall together, and the lesson did not take. Understanding why is useful, because the same failure recurred at far greater cost in 2022.

The demonstration was real but mild. Both asset classes delivered poor returns, and the mechanism — a policy-driven rather than growth-driven year — was identifiable. But the losses were modest by historical standards.

2019 was strong, which meant a portfolio that had absorbed the 2018 loss recovered quickly. An investor who had drawn a conclusion from 2018 and changed their allocation would have underperformed in 2019 relative to one who had not.

This is the same competitive selection problem the 2016 US venture report describes for short corrections: the environment punishes adaptation when the adverse condition does not persist. A lesson that costs money to act on and produces no visible benefit within a year does not spread.

Three further reasons the lesson did not take:

• The magnitude was small enough to attribute to noise. A modest simultaneous decline is consistent with correlation being unstable rather than conditional, and the second explanation requires changing a framework while the first does not.
• The theoretical basis was already known and already dismissed. That stock-bond correlation depends on whether growth or inflation is the dominant driver was not a discovery in 2018 — it was in the literature. It had simply not been relevant for long enough that it had stopped being operationally live.
• The alternative was unattractive. An investor who accepted the conditional nature of the hedge faced the question of what to hold instead, and in 2018 the honest answer was that there were no good alternatives at acceptable cost.

That last point deserves emphasis, because it is the reason the failure was not simply carelessness. Recognising that a hedge is conditional does not produce a better hedge. It produces the knowledge that the portfolio is more exposed than it appears, with no obvious remedy. That is uncomfortable and it is the correct state of knowledge.

The useful response to a conditional hedge is not to replace it. It is to know the condition, watch it, and size the rest of the portfolio on the assumption that the hedge may not be there. 2018 made the condition visible. 2022 made it expensive.