Using the \(\beta_1\) marginal posterior distribution to identify exit signals in pairs trading

Allan Quadros

Department of Statistics

Introduction

Problem



\[Y = \beta_0 + \beta_1X + \epsilon\]

  • Defining stop loss thresholds as \(\pm (2 + \delta)\) for the standardized residuals might cause too early or too late exits


Hypothesis


The proposed Bayesian method would …

  • better identify a moment of true rupture in the cointegration structure between stock Y and X.

  • overcome situations in which investors stop the operation too early but the residuals revert back to the mean 0 after some days.


Method - Step 1

Method - Step 2

Aplication


Perennial cointegration:

  • S&P500 ETFs SPY and IVV;
  • Gold (GLD) and Silver (IAU)

Non-perennial cointegration:

  • McKesson Corp. (MCK) vs Chevron Corp. (CVX);
  • Micron Technology Inc. (MU) vs ServiceNow Inc. (NOW);
  • United Airlines (UAL) vs American Airlines (AAL)

Results

Results (1)


Results (2)

Results (3)

Conclusions


New method:

  • worse results
  • more variability and drawdowns

But …

  • leave an open space for further study on the application of the Bayesian framework to pairs trading