Pairs Trading

Pairs trading is a form of statistical arbitrage developed at Morgan Stanley in the mid-1980s. The strategy exploits the long-run equilibrium relationship between two highly correlated securities — most commonly two companies in the same industry with similar business models, economic sensitivities, and investor bases. When the price ratio or spread between the two securities deviates from its historical norm by a statistically significant amount, the strategy goes long the relatively cheap security and short the relatively expensive one, expecting the spread to revert.

The identification of valid pairs requires rigorous statistical testing. Common approaches include cointegration testing (using the Engle-Granger or Johansen procedure to determine whether the long-run price relationship between two series is stationary) and correlation analysis over various time horizons. Pairs with high correlation but no cointegration may experience temporary co-movement without any force driving reversion, making them unreliable for systematic mean reversion strategies.

Once a pair is identified and a position is entered, the manager defines entry and exit rules based on the number of standard deviations the spread has deviated from its historical mean. A common framework enters a trade at 2 standard deviations and exits at convergence to the mean, with a stop-loss at 3 standard deviations. The position sizing is calibrated to maintain dollar and beta neutrality between the long and short legs.

Pairs trading is theoretically market neutral because movements in the overall stock market should affect both securities in the pair roughly equally, leaving the spread unchanged. In practice, the two legs of a pair do not always respond symmetrically to market moves, particularly when one company has meaningfully different financial leverage, interest rate sensitivity, or sector exposure than the other.

The strategy tends to perform best in stable, mean-reverting market regimes and struggles during structural breaks — periods where the fundamental relationship between two companies changes due to a merger, regulatory development, competitive disruption, or macroeconomic shock that affects one company but not the other. When spread divergence reflects a genuine fundamental change rather than noise, the trade will not converge and losses accumulate. Risk management therefore requires distinguishing statistical outliers that represent noise from those that reflect a regime change in the underlying relationship.