Factor Cyclicality and Drawdowns: Sizing for the Droughts

Factors go through droughts. Not quarters, years. A value tilt can lag a plain index for the better part of a decade, a momentum tilt can hand back a large share of its accumulated gains in a few violent months, and a defensive tilt can look foolish through an entire recovery. Factor cyclicality analysis is the work of measuring how deep and how long those episodes have run historically, so that the exposure you take is one you can actually hold to the other side.

This is the least glamorous part of factor investing and the part that decides outcomes. Most factor programmes do not fail because the factor was wrong. They fail because the exposure was abandoned near the bottom of a drought, which converts a temporary underperformance into a permanent one.

Why the droughts happen

A factor is, by construction, a standing tilt toward companies that share a characteristic. That tilt earns whatever the market pays for that characteristic, and the market’s willingness to pay moves in long, slow cycles.

When capital is cheap and investors are willing to underwrite long-dated growth, valuation discipline is a handicap. Cheap companies are usually cheap because something is wrong or dull about them, and in an environment that rewards optimism they stay cheap while expensive companies get more expensive. That is not the value factor breaking. It is the value factor doing exactly what it does, into a headwind.

Momentum has a different shape of pain. It tends to work in a steady grind and then break sharply at turning points, because the trade is by definition crowded into whatever has already worked, and reversals hit that basket all at once. A momentum drawdown is usually short and violent rather than long and slow.

Quality and low volatility tend to lag hardest in sharp recoveries, when the most leveraged and least stable companies rebound the most. Their bad periods look like being left behind rather than losing money.

Size has its own cycle tied to liquidity conditions and risk appetite, and its drawdowns tend to be worse than the headline numbers suggest because the trading costs rise at the same time.

The point is not to memorise these patterns. It is to internalise that each factor has a characteristic failure mode, and that the failure mode arrives on its own schedule.

Measuring the drought properly

A drawdown on a factor is not the same as a drawdown on a portfolio. There are two different measurements and confusing them causes arguments.

Absolute drawdown measures the peak-to-trough fall in value of a portfolio built on the factor. It is what an investor actually experiences, and it mostly reflects the equity market rather than the factor. In a broad market fall, every long equity factor portfolio falls.

Relative drawdown measures the peak-to-trough fall in the factor portfolio’s value relative to its benchmark. It is the number that describes the factor itself, and it is the one people quietly stop tracking during the good years. A relative drawdown is what makes a client ask why they are paying for this.

Alongside those, three views earn their place in the analysis.

The underwater curve plots how far below the previous peak the strategy sits, at every point in time. Run on relative performance, it shows the shape of the drought clearly: how deep, how long to the bottom, and, crucially, how long from the bottom back to a new high. That last leg is usually the longest and the least discussed. The mechanics are covered in drawdown recovery analysis.

Rolling excess returns show what an investor who started at an arbitrary date would have experienced. A factor with a good full-period record can still have had many overlapping five-year windows of underperformance, and it is those windows, not the full-period figure, that determine whether real people stayed invested. Method notes in rolling returns explained.

Time under water, the count of months spent below the previous relative peak, is the single most useful summary statistic for governance, because it translates directly into review cycles. A drought that spans several annual reviews is a governance problem before it is an investment problem.

Sizing for a drought you cannot time

Once you know the historical shape of the droughts, the sizing question becomes concrete: at what exposure could this book have sat through the worst historical episode without the decision being taken out of your hands.

That framing matters because it inverts the usual approach. Most sizing is set by how attractive the factor looks. Better sizing is set by how much of the bad case the mandate, the client and the committee can absorb. The binding constraint is rarely capital. It is patience.

A few practices follow.

Write the expected bad case down in advance, in the investment note. State the deepest relative drawdown and the longest time under water observed historically for this factor, and state plainly that a repeat should be expected rather than treated as evidence of failure. A drawdown that was described in advance is a very different conversation from one that arrives as a surprise.

Define, before launch, what would actually falsify the thesis. There has to be a difference between a normal drought and a broken factor, and it has to be specified while you are calm. Candidate falsifiers are structural, not performance-based: the economic rationale no longer holds, the definition can no longer be computed consistently, the strategy’s capacity has been exceeded, or the crowding has reached a level that changes the risk. Performance alone is a poor falsifier because it is exactly what a normal drought looks like.

Diversify across factors whose droughts historically differ. This is the strongest argument for a multi-factor approach and it is a claim about survivability rather than returns. See multi-factor investing explained.

Separate the review of the process from the review of the results. In a drought the only honest thing to check is whether the process ran as specified: were the definitions applied unchanged, was the rebalance executed, was the data clean. If the answer is yes, then the poor result is the exposure behaving normally, and there is nothing to fix.

Avoid resizing mid-drought without a pre-agreed rule. Reducing exposure after a long period of underperformance is the single most reliable way to capture the pain and miss the recovery. If exposure is going to be variable, the rule for varying it must exist before it is needed.

What this analysis does not tell you

It cannot time the cycle. No part of this work tells you when a drought will end. Valuation spreads, the usual candidate indicator, are wide during droughts and can stay wide for years, so they are a poor timing tool even when the underlying logic is sound.

Historical drought depth is a floor, not a ceiling. The worst drawdown in the sample is simply the worst that has happened so far, in the periods you have data for. Treating it as a limit is a well-known way to be surprised. The next one can be deeper, longer, or both.

It cannot distinguish a drought from a death. This is the honest, unresolved problem at the centre of factor investing. A premium that has been arbitraged away and a premium that is temporarily out of favour look identical while it is happening. Any framework that claims to separate them in real time is claiming more than the evidence supports.

Factor return histories carry their own biases. They are usually constructed from data that has been through index changes, delistings and restatements, and the construction choices matter more than most published charts admit. Drought statistics inherit every one of those choices. The relevant traps are set out in survivorship bias in backtests and why restatements break models.

A long sample can hide a regime change. Statistics computed across many decades average across market structures that no longer exist. That does not make them useless, but it does mean the confidence they project is partly an artefact of the window.

It says nothing about any individual company. Cyclicality is a statement about a characteristic across a universe. It has no implication for any particular business, and a factor tailwind does not make a badly run company a good one.

The realistic conclusion is modest and useful. You cannot avoid factor droughts, and you cannot time them. What you can do is know their historical shape, size the exposure so the shape is survivable, and agree in advance what would count as evidence that something has genuinely changed. That is most of what separates a factor programme that lasts from one that gets abandoned at the worst possible moment.

Related reading

• Portfolio metrics explained: the hub for risk, return and factor measurement.
• Factor investing in India: the factors, the evidence, and the Indian context.
• Factor crowding explained: when a lot of money chases the same signal, and what it does to risk.
• What is maximum drawdown: the peak-to-trough measure this analysis leans on.
• Why backtest results do not repeat: regime change, decay, luck, and the honest expectation.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.