2 
Unit Root 
Test
Objectives 
: 
 To explain the concept of Unit root test 
 To highlight the different names of unit root test 
 Explaining the, 
Dickey Fuller unit root test 
Augmented Dickey Fuller test 
Phillips -Perron test
Overview: 
A unit root is an attribute of a statistical model of a time series 
whose autoregressive parameter is one. 
Yt = ρ Yt + ut -1 ≤ ρ ≤ 1 
Thus, 
Non stationary, random walk and unit root are 
synonymous.
Testing for Unit 
root: 
Testing for the order of integrating 
A test for the order of integration is a test for the number of unit roots, and it 
includes the following, 
oStep 1: Test Yt to see if it is stationary. 
oStep 2: Take first difference of Yt as Δ Yt = Yt - Yt-1 
oStep 3: Take second difference of Yt as Δ2 Yt =Δ Yt -Δ Yt-1
Graphically
Hypothesis Testing for Unit 
Root 
 The basic objective of this test is to test null hypothesis that ρ = 1 in 
the following model. 
Yt = ρ Yt + ut
Dickey Fuller Test: 
• Done by Dickey (1979) and Fuller (1976) 
It is valid when time series is characterized by AR (1) with white noise 
errors. 
Testing Strategy: 
Time series plot-Graphing variables against time should always 
be the first step in any time series analysis.
Augmented Dickey Fuller test: 
• Most widely used technique. 
• It takes into account high order of correlation by adding ( p ) 
lags in the test. 
• It is used when time series is AR (p) and error term is not 
white noise. 
Δ Yt = β + β1 t + δ Yt-1 + Σ γ Δ Yt-I + u
Phillips – Perron Test: 
• Phillips and Perron (1988) developed a generalized for of ADF test procedure. 
• Developed a number of unit root tests. 
• PP test differs from the ADF in how they deal with serial correlation and 
heteroscedsticity in the errors. 
• When there is no autocorrelation, there PP test is identical to the DF test. 
Test Regression: 
For the PP test is the AR (1) process, 
Δ Yt = α+ γ Yt-I + et 
While ADF corrects high order serial correlation. 
Advantages: 
PP test are robust to general forms of heteroscedasticity. 
User does not have to specify lag length for the test regression.
References: 
www.timeseies.org 
www.wisegeek/dickey-fuller.org 
http://en.wikipedia.org/wiki/Phillips 
%E2%80%93Perron_test

Unit Root Test

  • 2.
  • 3.
    Objectives : To explain the concept of Unit root test  To highlight the different names of unit root test  Explaining the, Dickey Fuller unit root test Augmented Dickey Fuller test Phillips -Perron test
  • 4.
    Overview: A unitroot is an attribute of a statistical model of a time series whose autoregressive parameter is one. Yt = ρ Yt + ut -1 ≤ ρ ≤ 1 Thus, Non stationary, random walk and unit root are synonymous.
  • 5.
    Testing for Unit root: Testing for the order of integrating A test for the order of integration is a test for the number of unit roots, and it includes the following, oStep 1: Test Yt to see if it is stationary. oStep 2: Take first difference of Yt as Δ Yt = Yt - Yt-1 oStep 3: Take second difference of Yt as Δ2 Yt =Δ Yt -Δ Yt-1
  • 6.
  • 7.
    Hypothesis Testing forUnit Root  The basic objective of this test is to test null hypothesis that ρ = 1 in the following model. Yt = ρ Yt + ut
  • 8.
    Dickey Fuller Test: • Done by Dickey (1979) and Fuller (1976) It is valid when time series is characterized by AR (1) with white noise errors. Testing Strategy: Time series plot-Graphing variables against time should always be the first step in any time series analysis.
  • 9.
    Augmented Dickey Fullertest: • Most widely used technique. • It takes into account high order of correlation by adding ( p ) lags in the test. • It is used when time series is AR (p) and error term is not white noise. Δ Yt = β + β1 t + δ Yt-1 + Σ γ Δ Yt-I + u
  • 10.
    Phillips – PerronTest: • Phillips and Perron (1988) developed a generalized for of ADF test procedure. • Developed a number of unit root tests. • PP test differs from the ADF in how they deal with serial correlation and heteroscedsticity in the errors. • When there is no autocorrelation, there PP test is identical to the DF test. Test Regression: For the PP test is the AR (1) process, Δ Yt = α+ γ Yt-I + et While ADF corrects high order serial correlation. Advantages: PP test are robust to general forms of heteroscedasticity. User does not have to specify lag length for the test regression.
  • 11.
    References: www.timeseies.org www.wisegeek/dickey-fuller.org http://en.wikipedia.org/wiki/Phillips %E2%80%93Perron_test