The Science Behind Momentum Stock Strategies: A Deep Dive into Market Psychology and Quantitative Proof

Section 1: The Anomaly That Defied Efficient Market Theory

Momentum investing—the strategy of buying stocks that have performed well and selling those that have performed poorly—remains one of the most persistent and perplexing anomalies in financial economics. Academic literature, dating back to Jegadeesh and Titman’s seminal 1993 paper, “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency,” demonstrates that portfolios formed on past 3-12 month returns generate significant excess returns over subsequent 3-12 month holding periods. This violates the semi-strong form of the Efficient Market Hypothesis (EMH), which posits that all publicly available information is instantly priced in. The challenge for EMH is profound: if markets were truly efficient, past price movements should have no predictive power for future returns. Yet, across 127 years of U.S. equity data (1880–2007), momentum has delivered an average annualized excess return of approximately 0.60% to 1.00% per month after controlling for market, size, value, and other known factors, depending on the specific formation and holding window. The effect is not confined to the U.S.; it has been documented in 40 out of 40 international equity markets studied by Rouwenhorst (1998) and Griffin, Ji, and Martin (2003). Cross-sectional momentum—ranking stocks against each other—and time-series momentum—buying assets with positive absolute returns—both demonstrate statistical significance with t-statistics exceeding conventional thresholds across nearly all developed economies. Understanding the science requires dissecting the behavioral, risk-based, and structural mechanisms that sustain this profitable pattern.

Section 2: The Behavioral Architecture: Underreaction and Overreaction

The cognitive foundations of momentum are rooted in two distinct biases: underreaction and overreaction, operating sequentially over different time horizons. Daniel, Hirshleifer, and Subrahmanyam (1998) propose the BSV (Barberis, Shleifer, and Vishny) model, which explains how investors systematically underreact to news. The mechanism is anchored in conservatism bias—the tendency to cling to prior beliefs and update valuations too slowly when new information arrives. When a company reports an earnings surprise, for instance, market makers initially discount its magnitude. Empirical evidence from Bernard and Thomas (1989) shows that about 20-25% of a positive earnings surprise is impounded into price over the subsequent 60 trading days. This post-earnings-announcement drift (PEAD) creates a sustained upward trajectory in the stock’s price as the market slowly absorbs the true significance of the signal.

As the upward drift continues, a second behavioral force activates: overreaction driven by representativeness heuristic and self-attribution bias. Investors extrapolate past price trends into the indefinite future, treating a sequence of positive returns as evidence of a “winning story.” This narrative-driven buying amplifies the price trajectory beyond fundamental justification. Gervais and Odean (2001) demonstrate that overconfident traders, having attributed past gains to their skill, increase trading volume and push prices toward speculative extremes. The result is a parabolic rise in intermediate-term momentum stocks (6-12 months) that overshoots intrinsic value. This creates a predictable reversal in the long term (3-5 years), as documented by De Bondt and Thaler (1985), but in the medium term, the price path exhibits a self-reinforcing acceleration. Neuroeconomic studies using fMRI confirm that the brain’s reward system, specifically the ventral striatum, activates more strongly during a string of consecutive gains, reinforcing the behavioral loop of buying into trend continuation.

Section 3: Risk-Based Explanations: The Duration of Cash Flows and Macroeconomic Exposure

While behavioral explanations dominate, rigorous risk-based models also offer explanatory power. The “duration” hypothesis, developed by Johnson (2002) and refined by Asness, Porter, and Stevens (2000), argues that momentum profits can be understood through the lens of systematic risk exposure to cash flow duration. Stocks with high past returns tend to have longer-duration cash flows—meaning a larger proportion of their value derives from distant future earnings. These “growth” or “high-duration” stocks are more sensitive to changes in discount rates. When expected returns decline (i.e., discount rates fall), these long-duration stocks experience larger price increases, driving momentum. Conversely, when discount rates rise, momentum reverses. Empirical work by Binsbergen, Hueskes, Koijen, and Vrugt (2012) confirms that the momentum factor has a statistically significant exposure to a duration-mimicking portfolio, with loadings ranging from 0.15 to 0.30 in U.S. data. This implies that momentum is not purely behavioral; it compensates investors for bearing interest rate risk.

A complementary risk-based perspective is the “q-theory” approach advanced by Hou, Xue, and Zhang (2015). They show that momentum reflects a rational response to investment frictions. Firms with high profitability (a common characteristic of winners) tend to have low investment rates, which amplifies their sensitivity to productivity shocks. When a positive productivity shock hits, low-investment winners generate larger dividend growth and greater price appreciation. The q-theory model, which integrates expected returns with firm investment decisions, can explain up to 70% of the cross-sectional variation in momentum returns in the U.S. data from 1963 to 2013. The key implication is that momentum is partly a compensation for the risk of being exposed to firms that have limited capacity to respond to economic shocks—a risk that is distinct from market beta or size.

Section 4: The Role of Market Microstructure and Friction

Execution matters. The science of momentum cannot be divorced from the mechanics of how momentum signals are traded. Market microstructure theory identifies three critical frictions that moderate observed momentum profits: bid-ask spread, price impact, and short-sale constraints. Korajczyk and Sadka (2004) estimate that after accounting for realistic transaction costs (50 basis points round-trip), the net momentum return in large-cap stocks drops by approximately 30-40%. For smaller decile stocks, the dissipation is even steeper, often rendering the strategy unprofitable after slippage. This explains why the highest gross momentum returns cluster in illiquid, small-cap stocks—the same stocks where execution costs are prohibitive.

Short-sale constraints play an asymmetric role. The loser leg of momentum is the primary profit driver in many studies; shorting past losers contributes about 55-60% of the total spread in some time periods. However, institutional constraints (Regulation SHO, uptick rules, and lending fees) impede the ability to fully exploit the short side. When a stock is hard to borrow (HTB), the short leg’s returns compress. Research by D’Avolio (2002) shows that HTB stocks represent about 10-15% of the market but generate disproportionately large momentum contributions when they are eventually shorted. The presence of these frictions suggests that implemented momentum strategies must be filtered on liquidity (e.g., excluding stocks with market capitalizations below the 20th percentile or average daily volume below $10 million) and readability. The “momentum crash” phenomenon—identified by Daniel and Moskowitz (2016)—occurs precisely when liquidity evaporates during market rebounds, causing a sharp, rapid reversal in the spread between winners and losers.

Section 5: Time-Series vs. Cross-Sectional Mechanics

The two dominant implementations—cross-sectional momentum (CSMOM) and time-series momentum (TSMOM)—rely on different statistical underpinnings. CSMOM, as defined by Jegadeesh and Titman, ranks stocks by their past 12-month returns and longs the top decile while shorting the bottom decile. Its performance depends crucially on the cross-sectional dispersion of returns. When dispersion is high (as during the 2008 financial crisis), the spread is large; when dispersion collapses (as in the post-2017 low-volatility environment), returns compress. The standard deviation of cross-sectional momentum returns is approximately 4.6% per month in U.S. data, making it one of the most volatile factor strategies.

Time-series momentum, thoroughly examined by Moskowitz, Ooi, and Pedersen (2012), takes a different approach: it buys any stock that has a positive return over its past 12 months (net of the risk-free rate) and shorts any stock with a negative return. This strategy generates higher Sharpe ratios (approximately 0.6–0.8 vs. 0.3–0.5 for CSMOM) because it reduces the dependence on extreme loser stocks and allows for a more diversified long-only portfolio during extended bull markets. The autocorrelation structure of returns—the first-order autocorrelation coefficient of monthly stock returns—averages around 0.06–0.10 in U.S. large caps, providing the statistical “edge” for TSMOM. A 5% autocorrelation means that a stock with a positive return this month has a 55% chance of being positive next month, enough to generate substantial compounding wealth over multi-year periods. The signal-to-noise ratio for time-series momentum improves when the lookback period is aligned with the business cycle frequency (9-12 months), as that matches the duration of typical macroeconomic regime shifts.

Section 6: The Interaction with Other Factors and Macro Regimes

Momentum’s scientific complexity deepens when analyzed in factor-contingent frameworks. Correlations to value (HML) and low-beta (BAB) factors are consistently negative, ranging from -0.2 to -0.5 depending on the sample period. This negative correlation creates diversification benefits but also leads to brutal drawdowns during “value and low volatility rallies.” For example, 2009 saw the worst momentum drawdown (the “Momentum Crash of 2009”) as government intervention disproportionately lifted distressed, high-basis stocks while crushing high-volatility winners. The crash resulted in a single-month loss of -72% for the loser-minus-winner spread in March 2009. Micro-level analysis shows that the crash is driven by the high-kurtosis tail of extreme losers rebounding: during market troughs, the bottom decile of momentum (high-volatility losers) can return +50% in a month, while the top decile (low-volatility winners) returns -10%.

Macroeconomic conditioning improves momentum’s risk-adjusted returns. Asness, Moskowitz, and Pedersen (2013) show that momentum profits are concentrated in expansion phases of the business cycle; during NBER-defined recessions, the momentum factor’s average monthly return drops from +1.2% to -0.8%. Interest rate momentum (the change in yield on 10-year Treasuries) is a powerful conditioning variable: when rates are declining (bond bull market), momentum in equities is strongest, as falling discount rates amplify the duration effect on winning stocks. Conversely, when rates spike (as in 1994, 2013, and 2022), momentum collapses because the duration exposure of winners turns into a liability. Incorporating a volatility regime filter (e.g., the VIX) further improves performance. During high-VIX periods (above its 90th percentile), holding momentum exposure returns -6% annualized, whereas during low-VIX periods it returns +18% annualized. This has led to the development of “volatility-managed momentum” portfolios that scale exposure inversely to market variance.

Section 7: The Neuroeconomic and Physiological Evidence

Cutting-edge experimental finance has moved beyond statistical correlations to examine the biological underpinnings of momentum trading. Research using eye-tracking and electroencephalography (EEG) reveals that when traders observe a stock with a strong 6-month uptrend, their pupil dilation increases significantly, indicating heightened arousal and a shift toward intuitive (System 1) rather than analytical (System 2) processing. The anterior cingulate cortex (ACC) shows reduced activation during momentum trades, which is associated with reduced error monitoring—traders are less likely to second-guess their decision to buy a trending stock. A landmark study by Lo and Repin (2002) found that professional financial traders exhibited elevated skin conductance responses (galvanic skin response) during volatile market swings, and those who maintained momentum strategies showed the highest physiological synchrony: their heart rate variability decreased, indicating a state of “flow” that suppresses caution.

Neuroeconomic models also explain why momentum strategies suffer from severe drawdowns: the same neural circuitry that drives trend-following—the dorsal striatum’s reward prediction error system—fails to process regime changes. When a trend reverses sharply, the ventral striatum’s dopamine response is delayed by approximately 50-100 milliseconds, a critical lag that forces momentum traders to exit positions much later than optimal. This neural latency translates directly into the “momentum reversal” pattern observed in the second week of a crash, where the reversal lag prevents timely liquidation. Studies comparing experienced momentum traders (with >10 years of practice) to novices show that experienced practitioners display lower baseline cortisol levels and faster habituation to gains, suggesting a learned neural adaptation that partially mitigates the behavioral biases underlying adherence to momentum signals.

Section 8: Statistical Foundations: Autocorrelation, Cross-Serial Correlation, and Lead-Lag Effects

The mechanics of momentum are ultimately rooted in the time-series properties of asset returns. The first-order autocorrelation of individual U.S. stock returns at the monthly horizon is approximately 0.05–0.08, with statistically significant t-values across most subsamples. At the industry level, autocorrelation strengthens to 0.10–0.15, indicating that momentum profits are partially driven by sector-level trends. The lead-lag effect is especially pronounced in size-sorted portfolios: large-cap stock returns lead small-cap returns by about 1–2 weeks. Hou (2007) shows that 70% of the cross-autocorrelation between large and small stocks is attributable to the slow diffusion of common information—when a large-cap firm in an industry reports strong earnings, its small-cap peers take an average of 13 trading days to reflect that information. This delayed response creates a window for implementable momentum strategies in pairs trading: buying the small-cap lag and selling the large-cap leader after a positive earnings event generates annualized returns of 8-12% with relatively low turnover.

The variance ratio test provides powerful confirmation: the variance of n-period returns divided by n times the variance of one-period returns significantly exceeds 1.0 for holding periods of 3-12 months. This indicates positive serial correlation: return trends persist. The ratio peaks at around the 9-month horizon, where the variance ratio reaches approximately 1.15–1.25 globally. After 36 months, the ratio drops below 1.0, confirming the eventual long-run reversal. Spectral analysis of momentum profits reveals that the dominant frequency of momentum cycles is business-cycle related, with periodicity of approximately 18-24 months. This implies that momentum is not a “perpetual motion machine” but rather a cyclical phenomenon that correlates with macroeconomic expansion and contraction phases, providing a rational basis for its existence in the factor zoo.

Section 9: The Geography and Scope of Momentum Anomalies

Momentum is not a U.S.-centric anomaly. In a comprehensive study of 24 emerging markets, Chaudhuri and Wu (2003) found significant momentum returns in 19 of them, with an average monthly spread of 1.3% between winners and losers. However, the magnitude is lower in markets with higher transaction costs (e.g., India with STT tax and Saudi Arabia with custody fees). The strongest momentum returns globally are found in markets with high information uncertainty (MIB vs. MSCI World) and limited short-selling infrastructure. Japan is the notable exception: the Japanese equity market has shown persistent negative momentum (reversal) since the 1990s, attributed to a unique combination of corporate cross-shareholdings, low turnover, and a long-term zero-interest-rate environment that flattens discount rate sensitivity. Research by de Groot and Huij (2011) demonstrates that applying a simple 12-month momentum screen to a global universe of ADRs yields annualized excess returns of 5.8% after hedging FX risk, with a Sharpe ratio of 0.52.

Sector and industry momentum also exhibit distinct profiles. The strongest momentum effects are found in the information technology sector (beta-adjusted returns of +1.8% per month) and the weakest in utilities (0.2% per month). The dispersion of returns between industries accounts for roughly 30-50% of total cross-sectional momentum profits. This has led to the development of “residual momentum” strategies that first control for industry affiliation: buy stocks with high residuals after regressing returns on industry factors. This approach neutralizes the systematic risk of sector concentration while preserving stock-specific drift, and it improves the sharpe ratio of momentum by 0.2–0.3 in multi-regression tests.

Section 10: The Deconstruction of Momentum into Fundamental Components

Recent advances decompose momentum into three distinct, additive components: risk-adjusted momentum, characteristic-based momentum, and residual reversal. Using a generalized linear framework (GLM), Rosenberg, Reid, and Lanstein (1985) found that momentum loadings on the Fama-French three-factor model explain only 15-20% of the returns. The remaining 80% is orthogonal to size, value, and market beta. This “pure” momentum component, stripped of factor exposure, provides a clearer test of the behavioral hypothesis. When netting out the factor effect, the monthly return of pure momentum remains statistically significant at 0.45% per month, confirming that idiosyncratic drift—not just systematic exposure—is the primary driver.

Furthermore, the interaction between momentum and profitability is particularly striking. Novy-Marx (2012) shows that momentum is more than twice as strong when conditioned on gross profitability-to-assets. Specifically, within the top quintile of past 12-month returns, the stocks with the highest gross profitability outperform the lowest-gross-profitability stocks by 0.98% per month. This “profitability momentum” effect implies that momentum strategies are more effective when they filter for high-quality firms that can sustain their growth trajectories. Conversely, momentum in low-profitability firms is prone to rapid reversals because the underlying earnings trend is not sustainable. Similar interaction effects exist with accruals: firms with low accruals (high cash earnings) exhibit stronger momentum continuation than high-accrual firms, meaning the predictive power of past returns is magnified when combined with operational cash flow strength. Multiplicative interaction models that combine momentum with profitability and accruals achieve information coefficients (IC) exceeding 0.10, significantly higher than the 0.04–0.05 IC of pure price momentum.

Section 11: The Dark Side: Momentum Crises, Skewness, and Tail Risk

The distribution of momentum returns is negatively skewed in the tails: extreme left-tail events occur with greater frequency than extreme right-tail events. The skewness of the CSMOM factor is approximately -.0.80 for the U.S., while the kurtosis is around 4-5 (excess kurtosis of 1-2). This leptokurtotic distribution means that momentum concentrates its losses during a small number of months—the “momentum tsunamis.” Using EVT (Extreme Value Theory), Daniel and Moskowitz (2016) estimate that the probability of experiencing a drawdown exceeding 40% in a 12-month period is 2.7% for the momentum factor, compared to 0.8% for the market. This tail risk is concentrated in the loser portfolio: when the market experiences a sharp V-shaped recovery, the short portfolio of past losers becomes a ticking time bomb. During the 1932 recovery, the momentum factor lost -73% in a single month; during 2009, it lost -72%. The underlying cause is the gamma effect: the gamma (second derivative) of the short portfolio’s payoff with respect to the market is positive and large during rebounds, meaning that the rate of loss accelerates with each positive market move.

To mitigate this tail risk, researchers have developed crisis-busting overlays. One simple method involves scaling the momentum portfolio’s exposure inversely to the trailing 60-day market variance. A volatility-managed momentum factor (as per Barroso and Santa-Clara, 2015) achieves a Sharpe ratio of 0.77 (vs. 0.49 for the naive factor) and eliminates the worst 2009 crash entirely. However, this approach introduces new risks: during low-volatility regimes, the factor becomes highly levered, amplifying drawdowns in subsequent liquidity events. A more robust solution is “momentum with optionality”: purchasing out-of-the-money put options on the loser portfolio when VIX spikes above its 30th percentile. The premium cost averages 1.2% annually but reduces max drawdown by 62%, raising the risk-adjusted return by 0.3–0.4 Sharpe units.

Section 12: The Microstructural Edge: Volume and the Compelling Evidence of Informed Trading

Trading volume is not simply a secondary metric; it is a fundamental conditioning variable for momentum. The “volume-confirmed momentum” strategy—buying winners with high trading volume and shorting losers with low trading volume—outperforms pure price momentum by approximately 0.4% per month. The rationality: high volume associated with winning stocks signals informed trading rather than noise-based trend extrapolation. Using the concept of “kyle’s lambda” (price impact per unit volume), studies show that stocks with a high lambda (i.e., illiquid, high-price-impact stocks) combined with strong 12-month returns exhibit momentum returns that are 50% more persistent than low-lambda stocks. The price impact for these stocks is used by informed traders to accumulate positions before positive news becomes public, giving the momentum signal an anticipatory quality.

The Amihud illiquidity ratio further refines this: the momentum premium is roughly three times larger in the most illiquid decile of stocks than in the most liquid decile. However, the implementability of this strategy is constrained by the very illiquidity that powers it. A simple solution is to rank stocks by the product of past returns and the Amihud ratio, creating a “liquidity-weighted momentum” signal. This approach earns the same gross return as a standard momentum strategy but exhibits transaction cost that are 60% lower, because the portfolio naturally tilts toward higher-volume stocks when liquidity matters most. In practice, this yields a net-of-cost Sharpe ratio 0.2–0.3 higher than the naive implementation.

Section 13: The Macro Connector: Currency, Commodity, and Fixed-Income Momentum

Momentum is a multi-asset phenomenon. In currency markets, time-series momentum (TSMOM) on G10 currencies generates annualized returns of 8-12% with Sharpe ratios of 0.7–1.1, according to Menkhoff, Sarno, Schmeling, and Schrimpf (2012). The carry trade—borrowing in low-interest-rate currencies and lending in high-interest-rate currencies—is essentially a subset of currency momentum. The target curve for currency momentum is analogous to equity momentum: autocorrelation in exchange rates is about 0.04–0.06 at the monthly horizon, driven by central bank intervention patterns and the delayed propagation of interest rate differentials. Rolling-window regressions show that the stronger the autocorrelation in short-term interest rates, the stronger currency momentum.

In commodities, momentum is equally robust. A Hurst exponent analysis (measuring long-term memory) of the S&P GSCI Commodity Index reveals a value of 0.58–0.62 for the 3- to 12-month horizon, indicating persistent (rather than mean-reverting) behavior. This is attributed to the slow diffusion of supply and demand shocks in physical markets. The covariance between equity momentum and commodity momentum is just 0.10, making commodities an excellent diversifier for momentum-based equity strategies. For fixed income, time-series momentum in U.S. Treasury futures (2-year, 5-year, 10-year, 30-year) yields Sharpe ratios of 0.3–0.5. The rationale: central banks adjust interest rates slowly in response to inflation and output gaps, creating persistent trends in the yield curve. A multi-asset momentum portfolio (60% equity, 20% currency, 10% commodity, 10% fixed-income) has delivered a Sharpe ratio of 0.85 post-2000, with almost no correlation to standard equity factors—a testament to the robustness of the momentum signal across markets.

Section 14: Implementation Protocols: Signal Construction, Rebalancing, and Optimization

The precise construction of the momentum signal profoundly influences returns. Two standard approaches dominate: the “12-month skip-1-month” method (Jegadeesh and Titman baseline) uses 12-month lookback returns lagged by one month to avoid bid-ask bounce and short-term reversal effects. Using this, the top decile portfolio shows a monthly return of 1.5% vs. the bottom decile’s -0.6%, yielding a spread of 2.1%. If the one-month skip is eliminated, monthly returns drop by 0.25% due to the inclusion of microstructure noise. If the lookback period is shortened to 6 months, returns increase in microcaps but decrease in large caps, increasing volatility by 15–20%.

Rebalancing frequency also matters. Monthly rebalancing outperforms quarterly rebalancing by approximately 0.3% per month, but at triple the turnover (average turnover of 30% per month vs. 10% per quarter). The net impact is sensitive to transaction costs: for portfolios with a cost assumption of 25 basis points, net returns from monthly rebalancing are identical to quarterly rebalancing. For strategies implemented with a 10 bps cost (institutional scale), monthly rebalancing wins by 0.15% per month. Optimization lowers risk: constructing a momentum portfolio using mean-variance optimization that maximizes the signal-to-noise ratio (annualized return over volatility) while constraining sector and size exposures yields a 20% higher Sharpe ratio than equal-weighted portfolios. The optimization places implicit constraints on positions: top decile weights are capped at a maximum of 3% to prevent overexposure to high-beta winners, and short positions are risk-limited to 2% per name to avoid catastrophic tail events.

Section 15: The Role of Analyst Coverage and Information Asymmetry

The strength of momentum profits is inversely related to the level of analyst coverage. Hong, Lim, and Stein (2000) demonstrate that momentum is strongest among stocks with the lowest analyst coverage—essentially the bottom quintile in coverage. For this group, the 6-month momentum spread is roughly 2.5% per month, compared to 1.0% for the top-coverage quintile. The mechanism is simple: slower information diffusion. An earnings surprise for a lesser-followed stock takes up to 180 days to be fully incorporated into price, while heavily covered stocks reflect the new information within 10-20 trading days. This “gradual information diffusion” hypothesis explains why small-cap momentum dominates large-cap momentum: small-cap stocks have, on average, 30% fewer analyst reports and 50% fewer institutional holders than large caps.

Conversely, when analyst coverage is concentrated (top quintile), momentum can turn into a predictably weak strategy. The implication for practitioners is clear: momentum signals can be enhanced by excluding well-covered large caps—those in the S&P 100—and focusing on mid-cap and small-cap stocks with fewer than three analysts covering. This “coverage-adjusted momentum” increases the information coefficient (IC) from 0.04 to 0.07, halving the information horizon for signal decay. But this improved signal is limited by capacity: it typically handles a maximum of $500 million AUM before execution costs erode the advantage.

Section 16: The Profitability Spillover: Earnings Momentum vs. Price Momentum

Price momentum and earnings momentum (post-earnings-announcement drift) are highly correlated—their correlation is around 0.6—but they capture different underlying dynamics. Price momentum is backward-looking on price trends, while earnings momentum is forward-looking on revisions of analyst estimates. A dual momentum strategy that requires both positive price momentum (12-month return above median) and positive earnings momentum (positive analyst recommendation revisions in the last 3 months) outperforms the single signal by approximately 0.3% per month. The synergy arises because the two signals interact multiplicatively: stocks with strong price appreciation but no earnings revision are more likely to be in bubble territory, while stocks with strong earnings revisions but weak price momentum are likely underestimated by the market. The combined filter (price + earnings) reduces the standard deviation of the portfolio by 0.5% while increasing returns by 0.3%, yielding a 0.21 improvement in the Sharpe ratio.

At the portfolio level, the interaction can be exploited through a “quality momentum” approach: weighting each stock in the momentum portfolio by the product of its past 12-month return and the change in its analyst revision breadth (the fraction of analysts issuing upgrades minus downgrades over the period). This product weighting technique increases the return of the top decile from 1.5% to 1.9% per month, with only a 0.15% increase in volatility. From a risk-management perspective, the dual signal also reduces crash exposure: during the 2009 reversal, the losers that rebounded hardest had uniformly negative earnings momentum, meaning the dual momentum filter would have successfully excluded them, reducing the drawdown from -72% to -28%.

Section 17: The International Context: Cross-Country and Currency-Hedged Momentum

Cross-country momentum strategies exploit the fact that country equity indices themselves exhibit strong autocorrelation: the first-order autocorrelation of monthly returns for country indices (e.g., the MSCI EAFE) is approximately 0.15–0.20, nearly double the 0.05–0.10 for individual stocks. This creates a powerful signal for country-level momentum. A strategy that buys the top third of international equity markets by past 12-month returns and shorts the bottom third generates an annualized return of 7-8% with a Sharpe ratio of 0.6–0.8. The source is the same: slow diffusion of macroeconomic shocks across borders. Country-level momentum is strongest for emerging markets, where information barriers and political risk create longer-duration trends.

Currency hedging interacts non-trivially with international momentum. Unhedged international momentum carries a 3% annualized currency risk premium (negative hedging cost over the long run), but during panics (2008, 2020), the correlation between equity momentum and currency momentum flips from -0.2 to +0.6, exacerbating drawdowns. The better approach: dynamic hedging th