Trade the spread between two correlated stocks. When the spread widens beyond historical norms, short the outperformer and long the underperformer. Profit when the spread reverts.
History
Pairs trading was pioneered by Gerry Bamberger at Morgan Stanley in the mid-1980s, then refined by Nunzio Tartaglia's quantitative group. The strategy became one of the first systematic, market-neutral approaches on Wall Street. David Shaw, who worked in Tartaglia's group, later founded D.E. Shaw & Co. largely on the basis of pairs and statistical arbitrage strategies. The approach gained academic legitimacy through Gatev, Goetzmann, and Rouwenhorst's 2006 paper demonstrating consistent profitability from 1962 to 2002.
How It Works
Identify pairs of stocks with high historical correlation or cointegration (e.g., Coca-Cola and Pepsi, Visa and Mastercard)
Calculate the spread (price ratio or difference) between the pair and its historical mean and standard deviation
When the spread exceeds 2 standard deviations from the mean, enter a trade: long the underperformer, short the outperformer
Close the position when the spread reverts to the mean, capturing the convergence
Use cointegration tests (Engle-Granger, Johansen) rather than simple correlation to identify stable pairs
Position size each leg to be dollar-neutral, so the trade is market-neutral
Example Trades
KO/PEP spread widens to +2.5 sigma after Pepsi earnings miss
entry Long PEP at $165, short KO at $62 (dollar-neutral)
exit Spread reverts over 8 trading days
result +1.8% on the pair, ~3.6% annualized Sharpe contribution
V/MA spread compresses to -2 sigma during Visa-specific regulatory news
entry Long V at $275, short MA at $390 (dollar-neutral)
exit Spread normalizes in 12 days
result +2.1% on the pair
Related Charts
Who Runs This
When It Works vs. Fails
works
Stable, range-bound markets with normal volatility. Sectors with clear competitive pairs. Low correlation between macro and micro drivers.
fails
Regime changes, structural industry shifts, and crowded-trade unwinds. The 2007 quant crisis wiped out years of gains in days.
Risks
01 Pairs can decouple permanently due to fundamental changes (merger, bankruptcy, business model shift)
02 Crowding risk: when too many quant funds run the same pairs, convergence slows and divergences deepen
03 The August 2007 Quant Crisis saw massive losses in pairs/stat-arb as crowded positions unwound simultaneously
04 Requires significant leverage to generate meaningful returns from small spread movements
Research
Gatev, Goetzmann, Rouwenhorst, 2006
Engle, Granger, 1987
Rotondi, Russo, 2024