CFA CFA Level 2 Unit Root Test of Nonstationarity

Unit Root Test of Nonstationarity

  • This topic has 1 reply, 2 voices, and was last updated Jan-19 by ec_test.
  • Author
    • Beezy849

      Can anyone help clarify on this? I’ve been having trouble wrapping my head around the concept, specifically with how it applies to random walk and first difference. 

    • ec_test

      Hello there! 

      An AR(1) model equation should be Xt=b0 + b1 X(t-1) + Et

      For an AR(1) to have an adequate forecasting utility, b1 < 1

      A unit root means that the coefficient (b1) of the AR(1) model is equal to 1

      In other words, the slope of the coefficient is one (b1=1). Picture a perfectly straight line in a 45 degree angle in an X and Y graph. 

      When there is a “unit root” b1=1 and a random walk occurs.

      There are two types of random walks: Without a drift and with a drift

      No drift – b0=0 and b1=1 (no intercept) 

      Xt = b0 + b1 X(t-1) + Et
      Xt = 0 + (1) X(t-1) +Et
      Xt = X(t-1) + Et

      With a drift b0>0 and b1=1(there is an intercept)

      Xt = b0 + b1 X(t-1) + Et
      Xt = b0 + (1) X(t-1) + Et
      Xt = b0 + X(t-1) + Et

      Why b1=1 – unit roots a problem? Because the mean reverting level is undefined

      Mean Reverting Level = b0/(1-b1), if b1=1, then the Mean Reverting Level is b0/0 which is undefined. 
      When the Mean Reverting Level is undefined, the time series is non-stationary. 

      I hope this helps! 

Viewing 1 reply thread
  • You must be logged in to reply to this topic.