Liquidity-Adjusted Momentum: How the Amihud Ratio Transforms Position Sizing

Liquidity-Adjusted Momentum: How the Amihud Ratio Transforms Position Sizing

Between March and May 2009, momentum strategies suffered one of their worst drawdowns in history. The standard 12-1 month momentum portfolio lost over 52% peak-to-trough as beaten-down financials and cyclicals violently snapped back while momentum winners collapsed. Yet a closer look at the anatomy of this crash reveals something striking: the losses were overwhelmingly concentrated in illiquid stocks. The most illiquid decile of momentum winners fell three times as much as the most liquid decile, and the most illiquid losers squeezed three times as hard. This pattern is not unique to 2009. Across every major momentum crash since 1990, illiquid positions have been the primary source of tail risk.

This observation motivates a simple but powerful modification to the standard momentum strategy: adjust position sizes by liquidity. Specifically, scale each position inversely by its Amihud (2002) illiquidity ratio, so that liquid momentum stocks receive larger weights and illiquid ones receive smaller weights, or are excluded entirely. Quant Decoded's original backtest shows this adjustment improves the Sharpe ratio from 0.55 to 0.72 with sizing alone, and to 0.82 when the most illiquid stocks are filtered out entirely, while reducing the maximum drawdown from -52% to -29%.

The result is a momentum strategy that retains most of the upside but dramatically reduces the tail risk that makes traditional momentum dangerous in practice. The improvement comes not from a new alpha signal but from better risk management of an existing one, consistent with the academic finding that much of momentum's apparent alpha is illusory because it accrues in stocks where trading costs make it uncapturable (Lesmond, Schill, and Zhou, 2004).

The Liquidity Problem in Momentum Strategies

The standard momentum portfolio, as defined by Jegadeesh and Titman (1993), ranks stocks by their past 12-1 month returns and goes long the top decile while shorting the bottom decile. This construction is agnostic to liquidity. A stock with $500 million in daily turnover and a stock with $2 million receive the same weight if they fall in the same decile.

This creates two related problems. First, illiquid momentum winners are the hardest to exit when the trade reverses. During momentum crashes, these positions gap down violently because there are no natural buyers. Second, illiquid momentum losers are the hardest to cover during short squeezes. When beaten-down stocks rally, the most illiquid shorts produce the largest losses because covering drives prices even higher.

Avramov, Cheng, and Hameed (2016) documented this pattern formally, showing that momentum profits are significantly time-varying and that liquidity conditions are a primary driver. In low-liquidity environments, momentum crashes are both more frequent and more severe. Pastor and Stambaugh (2003) demonstrated more broadly that stocks with high liquidity risk command a return premium, but that premium comes with extreme left-tail risk that momentum strategies inadvertently concentrate.

The core insight is that momentum and liquidity risk interact in a particularly dangerous way. Momentum selects for recently extreme performers. Extreme performance often coincides with deteriorating liquidity (winners become crowded, losers become distressed). The strategy therefore systematically overweights the most fragile positions at the worst possible time.

The Amihud Illiquidity Ratio

The Amihud (2002) illiquidity ratio provides a simple, robust measure of price impact per unit of trading volume. For stock i on day d, the ratio is defined as:

ILLIQ = |Return| / Dollar Volume

The daily ratios are averaged over a trailing window (we use 21 trading days) to produce a monthly illiquidity estimate. Higher values indicate that small amounts of trading volume move prices more, signaling lower liquidity. The measure has become the standard academic proxy for liquidity due to its simplicity, data availability, and strong correlation with more sophisticated measures such as the Kyle (1985) lambda and the bid-ask spread.

For our backtest, we compute the Amihud ratio monthly for each stock in the CRSP universe, then sort stocks into liquidity quintiles within each momentum decile. This dual-sort framework allows us to analyze how momentum performance varies across the liquidity spectrum and to construct liquidity-adjusted strategies.

Backtest Design and Data

Universe and Sample Period

The backtest covers US equities from January 1990 through December 2025 (432 months). The universe consists of all common stocks (share codes 10, 11) on NYSE, AMEX, and NASDAQ with sufficient data to compute 12-month returns and 21-day Amihud ratios. Micro-caps below the 5th percentile of NYSE market cap are excluded to avoid penny stock contamination.

Strategy Specifications

Three strategy variants are tested:

1. Standard Momentum: Long top return decile, short bottom decile, equal weight within deciles. Monthly rebalancing with a one-month skip (12-1 formation).

2. Liquidity-Adjusted Sizing: Same momentum signal and decile breakpoints, but position sizes within each decile are inversely proportional to the stock's Amihud ratio. Specifically, the weight of stock i is proportional to 1/ILLIQ_i, normalized to sum to 1 within each leg. This gives liquid stocks larger positions and illiquid stocks smaller positions.

3. Liquidity-Filtered: Same momentum signal, but stocks in the bottom 20% of liquidity (highest Amihud quintile) are excluded entirely before forming decile portfolios. Equal weight within the remaining positions.

All returns are gross of transaction costs unless otherwise noted. The transaction cost analysis section addresses implementation drag separately.

Results: Performance Comparison

The table below presents the headline performance statistics for each strategy variant.

The results reveal a clear pattern. Liquidity adjustment sacrifices a modest amount of return (8.2% to 7.5%) but dramatically reduces volatility (14.9% to 9.1%) and tail risk (maximum drawdown from -52.1% to -29.3%). The Sharpe ratio improves from 0.55 to 0.82, a 49% increase. The Sortino ratio improvement is even more pronounced, from 0.71 to 1.21, reflecting the disproportionate reduction in downside volatility. Perhaps most importantly, the return distribution shifts from significantly negatively skewed (-1.82) to only mildly negative (-0.51), eliminating the crash-prone character that has historically made momentum one of the most dangerous factor strategies.

Drawdown Analysis by Liquidity Quintile

To understand why liquidity adjustment works, we decompose the standard momentum strategy's returns by the liquidity quintile of its constituent positions.

The data confirms the central thesis. The most illiquid quintile (Q5) generates the highest raw return (12.3%) but with enormous volatility (24.8%) and catastrophic drawdowns (-68.4%). During the 2009 momentum crash, Q5 positions contributed 37% of total losses despite representing only 20% of positions. Conversely, the most liquid quintile (Q1) delivers a more modest 6.4% return but with a maximum drawdown of just -18.2% and only 8% contribution to crash losses.

This pattern creates a highly favorable tradeoff for liquidity adjustment. By reducing or eliminating exposure to Q4 and Q5 stocks, the strategy surrenders 1-2 percentage points of annual return but eliminates the positions responsible for over 60% of crash losses. The risk-adjusted improvement is substantial because the marginal return from illiquid momentum stocks does not compensate for their marginal risk contribution.

Momentum Crash Episodes: With and Without the Liquidity Filter

The table below examines every momentum drawdown exceeding -15% during the sample period and compares the standard strategy against the liquidity-filtered version.

The liquidity filter consistently reduces crash severity by 34% to 45% across all major episodes. The largest absolute improvement occurs during the 2009 Global Financial Crisis, where the drawdown is cut from -52.1% to -29.3%, a reduction of 22.8 percentage points. The Quant Quake of August 2007 shows the largest proportional improvement (45% reduction), which is intuitive because that event was specifically driven by crowding and forced liquidation in quantitatively selected positions, conditions that disproportionately affect illiquid stocks.

The consistency of improvement across different market regimes and crash triggers is notable. Whether the crash is driven by macro reversal (2009), sector rotation (2001), systematic deleveraging (2007), or pandemic-driven dislocation (2020), illiquid positions are invariably the primary source of tail risk.

Turnover and Capacity Analysis

A practical concern with any strategy modification is whether it introduces excessive turnover or reduces investable capacity. The table below addresses these questions.

The results are counterintuitive in one important respect. Although the liquidity-adjusted sizing strategy has slightly higher turnover (24.8% vs. 21.4%), its estimated transaction costs are actually lower (1.5% vs. 1.8%) because it concentrates weight in liquid stocks where trading costs are minimal. The liquidity-filtered strategy reduces both turnover and costs further because excluding illiquid stocks naturally reduces the churn associated with volatile, hard-to-trade names.

Capacity improves dramatically. The standard momentum strategy has an estimated capacity of roughly $3.2 billion before market impact becomes significant. The liquidity-filtered version more than doubles this to $8.1 billion because it trades only in stocks with sufficient depth. For institutional investors managing large pools of capital, this capacity advantage may be as important as the Sharpe improvement.

After accounting for estimated transaction costs, the net-of-cost Sharpe ratio of the liquidity-filtered strategy (0.71) exceeds the gross Sharpe of the standard strategy (0.55). This is the key practical finding: the liquidity adjustment does not merely improve theoretical performance but improves implementable, after-cost performance by an even larger margin than gross returns suggest.

Why Liquidity Adjustment Works: The Mechanism

The effectiveness of liquidity-adjusted momentum rests on three complementary mechanisms.

First, illiquid momentum stocks exhibit asymmetric price dynamics. When momentum reverses, liquid stocks can be sold with minimal market impact, producing orderly drawdowns. Illiquid stocks cannot. Selling pressure in thin markets creates price cascades, where each sale pushes prices further down, triggering stop-losses and margin calls that generate additional selling. This is the mechanism behind the extreme negative skewness observed in illiquid momentum quintiles.

Second, illiquidity serves as a proxy for crowding risk. Stocks with declining liquidity often signal that a crowded trade is approaching its breaking point. Momentum strategies naturally accumulate positions in stocks that become increasingly crowded over the formation period. By underweighting illiquid stocks, the strategy implicitly reduces exposure to the most crowded positions.

Third, the Amihud ratio captures information about tradability that momentum signals miss. A stock with strong 12-month returns and declining volume is a very different proposition from one with strong returns and rising volume. The former suggests waning interest and potential reversal; the latter suggests sustainable demand. Liquidity adjustment incorporates this distinction without abandoning the momentum signal.

These mechanisms align with the broader academic literature on liquidity and asset pricing. Pastor and Stambaugh (2003) showed that liquidity risk is priced in the cross-section; stocks with high liquidity beta earn higher average returns but with substantial left-tail risk. Momentum strategies inadvertently load heavily on liquidity risk because extreme past performers tend to have experienced liquidity changes. Adjusting for this loading reduces return modestly but eliminates a disproportionate share of risk.

Robustness and Limitations

Several robustness checks support the main findings. The liquidity adjustment improves Sharpe ratios across all formation periods tested (3-1, 6-1, and 12-1 months), with the largest improvement for the standard 12-1 specification. Results are qualitatively similar when using market-cap-weighted momentum instead of equal-weighted, though the improvement is smaller because market-cap weighting already implicitly favors more liquid stocks. Using bid-ask spread as an alternative liquidity measure produces nearly identical results, confirming that the Amihud ratio is not driving the findings through measurement artifact.

However, several limitations apply. This is a single-country backtest; although the mechanisms should generalize internationally, out-of-sample testing in non-US markets is needed. The backtest uses end-of-day data and assumes execution at closing prices, which may be optimistic for the most illiquid positions (though this bias works against the standard strategy more than the liquidity-adjusted version). Finally, the specific liquidity threshold (bottom 20%) and the inverse-Amihud weighting scheme were chosen based on economic reasoning rather than in-sample optimization, but some data-mining risk remains.

Conclusion

The interaction between momentum and liquidity risk is one of the most important, and most underappreciated, dynamics in systematic investing. Standard momentum strategies load heavily on illiquid positions that generate attractive returns in normal markets but catastrophic losses during reversals. Adjusting position sizes by the Amihud illiquidity ratio, or simply excluding the most illiquid stocks, eliminates the majority of momentum's tail risk while preserving most of its return.

The practical implications are significant. A liquidity-filtered momentum strategy with a Sharpe of 0.82, maximum drawdown of -29%, and capacity exceeding $8 billion is a fundamentally different proposition from a standard momentum strategy with a Sharpe of 0.55 and -52% drawdowns. For allocators who have avoided momentum due to crash risk, the liquidity-adjusted version may represent a more palatable way to capture the momentum premium. For those already running momentum, a liquidity overlay provides meaningful risk reduction at modest return cost.

The finding also connects to a broader principle in portfolio construction: the best improvements often come not from discovering new signals but from more intelligently managing existing ones. Momentum remains one of the most robust and well-documented anomalies in finance. Its main weakness is not the signal itself but the way standard implementations handle liquidity risk. Fixing that weakness produces a strategy that is, by most measures, superior to the original.

References