The previous analysis demonstrated the nonlinear mathematics of drawdown recovery — how a 50% loss requires a 100% gain to restore, and why avoiding deep drawdowns contributes more to terminal wealth than maximizing returns. This piece examines the mechanism that addresses it: volatility targeting.

The Simple Idea Behind a Robust Technique

Volatility targeting can be stated in one sentence: scale your position size inversely to realized volatility, so that your actual risk exposure remains approximately constant through time.

When volatility is low, hold more. When volatility is high, hold less. The target is not a return level — it is a risk level. Instead of saying “I want to hold 60% in equities,” the framework says “I want to hold an amount of equities such that my expected daily loss in a bad scenario does not exceed X percent of my portfolio.”

The practical implementation is straightforward. Measure realized volatility over a recent lookback window — often 21 to 63 trading days, though the optimal window varies by asset class. Compute the ratio of your target volatility to realized volatility. Scale your position by that ratio.

If your target annual volatility is 10% and current realized volatility is 20%, hold half your normal position. If realized volatility falls to 5%, hold double. The mechanism is continuous and mechanical: no forecasting, no market views, no discretion required.

When volatility doubles, hold half as much. This is not a prediction about where markets will go. It is an acknowledgment that risk has changed — and that position size should reflect the actual risk being borne, not the risk that prevailed on the day the original allocation was set.

The Evidence

Moreira and Muir (2017), in “Volatility-Managed Portfolios” published in the Journal of Finance, conducted the most systematic empirical test of volatility targeting across asset classes. Their findings were unambiguous.

They applied a simple volatility-managed strategy to the equity premium, the value factor, the momentum factor, bond returns, and currency returns. In each case, the strategy held more of the asset when recent volatility was low and less when it was high, scaled to a constant target volatility. The results: volatility management significantly improved the Sharpe ratio for every one of the five asset classes tested. For momentum specifically, volatility targeting increased the annualized Sharpe ratio from 0.53 to 0.97 — nearly doubling it.

The mechanism behind the improvement was not alpha generation in the conventional sense. It was loss reduction. Volatility tends to spike precisely during market dislocations — the periods when drawdowns are largest and most damaging to compounding. By automatically reducing position size when volatility rises, the strategy mechanically reduces exposure during the exact environments where large losses are most likely to occur. The result is a portfolio that participates in normal-volatility return environments but carries less exposure during crisis-volatility environments.

The outperformance was not confined to specific time periods or market conditions. Moreira and Muir (2017) tested the strategy over multiple decades of data and found the results persistent across market cycles. The mechanism is robust because it does not rely on any prediction being correct — it relies only on the empirical regularity that volatility tends to cluster: high-volatility periods are followed by high volatility, low-volatility periods by low volatility. This autocorrelation in volatility, documented extensively in the GARCH literature, is the statistical foundation that makes volatility targeting work.

Harvey, Liu, Zhu, and colleagues, in their 2018 analysis of managed volatility strategies across risk premia, extended the Moreira-Muir finding to a broader set of factors and international markets. The result held: volatility-scaled exposures consistently produced better risk-adjusted outcomes than fixed-weight exposures. The improvement was most pronounced in assets that exhibit strong volatility clustering and the most severe left-tail events — the exact environments where the nonlinear drawdown mathematics documented in the prior analysis impose the greatest compounding cost.

The Momentum Crash Problem

The most striking demonstration of what volatility targeting can accomplish comes from the literature on momentum crashes.

Momentum — buying recent winners and selling recent losers — is one of the most documented return premia in finance, present across decades and geographies. It is also subject to spectacular crashes: sudden, severe drawdowns that can erase years of accumulated profits in weeks. The crashes occur predominantly at market turning points, when the most beaten-down stocks — the short side of a momentum portfolio — suddenly reverse sharply.

Barroso and Santa-Clara (2015), in “Momentum Has Its Moments” published in the Journal of Financial Economics, examined the momentum factor through the lens of volatility targeting. Their findings were striking: simple volatility scaling, applied to the standard momentum factor, reduced maximum drawdown from 91.6% to 40.8% — a reduction of more than 50 percentage points.

The maximum drawdown of 91.6% is not a typographical error. The raw momentum factor, held at a constant notional exposure, experienced a 91.6% drawdown during a single six-month episode in 2009. This is the compounding trap described in the prior analysis: a 91.6% loss requires an 1,086% gain to recover. No realistic future return stream recovers from that. The strategy is mathematically over.

With volatility targeting applied, the same factor, over the same period, experienced a maximum drawdown of 40.8% — severe, but survivable. The required recovery is 69%, not 1,086%. The strategy can realistically reach prior highs and continue compounding.

The mechanism is direct. Momentum crashes follow periods of elevated market volatility. By cutting position size when volatility rises — before the crash has fully materialized — the volatility-scaled strategy automatically enters the crash period with reduced exposure. The drawdown is not avoided, but it is substantially shallowed.

What Volatility Targeting Is Not

Clarity on what this framework does not claim is as important as what it does.

Volatility targeting is not a return-forecasting model. It makes no prediction about whether the market will go up or down. It observes only what risk has recently been — not what risk will be, and not what returns will be. An investor who implements volatility targeting and expects to avoid losses because they hold less during high-volatility periods will be disappointed. The strategy can and does lose money. What it limits is the depth of those losses.

Volatility targeting is not a guarantee against drawdowns. If volatility spikes suddenly — as it did in the immediate hours following the Lehman Brothers filing, before most systematic volatility estimates could adjust — the strategy enters the drawdown with the prior period’s allocation before the scale-down can occur. Jump volatility — sudden large moves not preceded by elevated measured volatility — is the mechanism’s primary limitation. The strategy manages gradual volatility regimes better than abrupt discontinuities.

Volatility targeting is not free. Scaling up when volatility is low and scaling down when it is high requires active rebalancing, which generates transaction costs. In markets with high bid-ask spreads or limited liquidity, the cost of implementing the strategy can offset part of the risk-adjusted benefit. The economic value of the strategy is highest in liquid markets with modest transaction costs.

Volatility targeting does not eliminate regime risk. As the first analysis in this series established, regime transitions are when the most severe and sudden portfolio damage occurs. Volatility targeting attenuates the damage because volatility typically rises during transitions — but the lag between when volatility rises and when the position is scaled down means some transition losses are absorbed before the adjustment occurs.

None of these limitations negate the core finding. They define the conditions under which the strategy works best and the conditions under which it is most stressed.

Every Investor Already Sizes Positions — The Question is How

The framing of volatility targeting as an exotic institutional technique misses the fundamental point. Every investor who makes a portfolio decision is implicitly implementing a position-sizing rule. The question is not whether to have one — it is whether the rule is systematic or emotional.

The default rule for most investors is fixed allocation: hold a constant percentage of the portfolio in each asset class, rebalancing periodically to restore that percentage. This rule is simple and has merit — it implements a mechanical “buy low, sell high” through rebalancing. But it contains a hidden assumption: that the risk of each position is constant over time. When the equity allocation is set at 60%, the implicit assumption is that 60% today poses the same risk as 60% in a month, a year, or a decade.

This assumption is demonstrably wrong. A 60% equity allocation in early 2007, when realized equity volatility was approximately 10% annualized, posed a fundamentally different risk than 60% in late October 2008, when realized equity volatility exceeded 70% annualized. The investor who maintained 60% equity throughout that period was not maintaining constant risk — they were accepting a sevenfold increase in actual risk while their portfolio reports said “60% equity, as planned.”

Volatility targeting makes the risk constant by allowing the weight to vary. Instead of asking “how much equity do I own?” it asks “how much equity volatility am I bearing?” These are different questions, and the second one is the right one to answer.

The alternative to systematic position sizing is emotional position sizing — which the behavioral finance literature documents reliably produces the opposite of optimal behavior. Investors hold too much risk during euphoric low-volatility periods because they are extrapolating good recent outcomes. They sell too much during crises because the losses feel unbearable. The net effect of emotional position sizing is the opposite of what volatility targeting implements: they hold more when risk is high and less when risk is low, systematically amplifying both the drawdowns and the behavioral errors that crystallize them into permanent losses.

The Broader Implication for Portfolio Construction

The evidence from Moreira and Muir (2017), Harvey et al. (2018), and Barroso and Santa-Clara (2015) converges on a single conclusion: risk exposure should be a variable, not a fixed input to portfolio construction.

The traditional framework treats the asset allocation as the primary decision and accepts whatever volatility the market happens to deliver. The volatility targeting framework inverts this: it treats the risk budget as the primary decision and varies the allocation to keep actual risk within that budget.

This inversion has practical consequences that extend beyond the immediate technique. It implies that the right question before adding any position is not “what return do I expect?” but “what risk am I currently bearing, and does this position fit within my risk budget given current market conditions?” It implies that portfolio rebalancing should be triggered by risk drift, not just weight drift. It implies that the sizing decision is ongoing and dynamic, not a one-time allocation choice.

Volatility targeting is one implementation of a broader principle: that risk management is a continuous input to portfolio construction, not a periodic audit. The next question in this line of reasoning is about the research process itself — specifically, whether the strategies that claim to offer systematic edges actually do.

The next analysis examines why most quantitative research is wrong — and why the overfitting epidemic makes backtests the least trustworthy evidence in finance.

References: Moreira and Muir (2017), “Volatility-Managed Portfolios,” Journal of Finance, Vol. 72, No. 4. Barroso and Santa-Clara (2015), “Momentum Has Its Moments,” Journal of Financial Economics, Vol. 116, No. 1. Harvey, Liu, Zhu, et al. (2018), analysis of managed volatility across risk premia. GARCH volatility clustering: Engle (1982), “Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation,” Econometrica.