[go: up one dir, main page]
Scribd
Upload
40 views7 pages

Unit Root Test in Time Series

The document discusses unit root tests for time series data. It defines stationary and non-stationary time series, and explains that a unit root occurs when a time series follows a random walk process and shocks to the series are permanent. The Dickey-Fuller test is introduced as a method to test for a unit root by examining whether the autoregressive parameter φ is equal to 1 or less than 1. The test regression is specified under different assumptions about the inclusion of a constant, trend, or both. The test statistic involves the estimated φ parameter and its standard error to test the null hypothesis of a unit root.

Uploaded by

Altaf Kondakamarala
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
40 views7 pages

Unit Root Test in Time Series

The document discusses unit root tests for time series data. It defines stationary and non-stationary time series, and explains that a unit root occurs when a time series follows a random walk process and shocks to the series are permanent. The Dickey-Fuller test is introduced as a method to test for a unit root by examining whether the autoregressive parameter φ is equal to 1 or less than 1. The test regression is specified under different assumptions about the inclusion of a constant, trend, or both. The test statistic involves the estimated φ parameter and its standard error to test the null hypothesis of a unit root.

Uploaded by

Altaf Kondakamarala
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

02-May-19

Unit Root Test

Outline
• Stationary and non-stationary time series.
• Deterministic and Stochastic trend
• Random Walk
• Unit root test

1
02-May-19

Outline
• Stationary and non-stationary time series.
• Deterministic and Stochastic trend
• Random Walk
• Unit root test

Non-stationary time series model

2
02-May-19

Non-stationarity in mean
• Deterministic trend
• Stochastic trend

Deterministic Trend
• A deterministic trend is when we say a series is trending because it is
an explicit function of time
Yt = α+ βt+ut

The parameter β measure the average change in Yt from one period to


the another
ΔYt = Yt - Yt-1 =α+ βt+ut – [α+ β(t-1)+ut-1 ]
=β+ ut - ut-1
Now, E(Δyt ) = β,
This model is called a trend stationary model

3
02-May-19

Stochastic Trend
AR(1) model : Yt = c+ ϕ Yt-1+ut

As long as |ϕ|< 1, it is stationary and everything is fine.

Now, consider the extreme case where ϕ =1,


Yt = c+ Yt-1+ut
= c+(c+ Yt-2+ut-1)+ ut
=tc+Y0+ 𝑡𝑖=1 𝑢𝑖
This is what we call a random walk with drift. If C=0, it is a random
walk.

Random Walk process

4
02-May-19

Random Walk process

Random Walk Process


• Behaviour of the stock market
• Brownian motion
• Movement of a drunken man
• It is a limiting process of AR(1)

Yt = ϕ Yt-1+ut
It can be shown that mean of random walk process is constant, but
variance is not. So, random walk process is non-stationary.

5
02-May-19

Unit root in time series


When we have a stationary system, effect of a shock will die out
gradually
When we have a non-stationary system, effect of a shock is permanent

Two types of non-stationarity:


- Unit root, i.e., |βi|=1: Homogeneous non-stationarity, ϕ=1
- |βi|> 1: explosive non-stationarity.

DF Test
Dickey-Fuller test:
The simplest approach to test for a unit root begins with AR(1) model
Yt = θ0+ ϕ Yt-1+ut
Where, ut ~ N (0, σu2 )

Ho : |ϕ| = 1 => Yt ~ I (1)


H1 : |ϕ| < 1 => Yt ~ I (0)
Now, ΔYt = Yt - Yt-1 = θ0+ ϕ Yt-1+ut -Yt-1 = θ0+ (ϕ -1)Yt-1+ut
= θ0 + δ Yt-1 +ut , δ = ϕ -1
If δ = 0, system has a unit root
Ho : δ = 0
H1 : δ < 0

6
02-May-19

DF Test
Δyt = δ Yt-1 +ut -> Pure random walk/ Without constant and trend
Δyt = θ0 +δ Yt-1 +ut -> Random walk with drift/ with constant
Δyt = θ0 + θ1 t+δ Yt-1 +ut -> Random walk with drift and linear trend/ with constant
and trend
Test Statistic:
After applying OLS method and finding the estimator of ϕ,
t =(ϕ – 1)/ se (ϕ)
At level yt is stationary means I(0)
At first difference yt is stationary means I(1)
At second difference yt is stationary means I(2)

You might also like