It’s All about Unit Root

We noted in Chapter 1 that one of the important types of data used in empirical analysis is
time series data. In this and the following chapter we take a closer look at such data not only
because of the frequency with which they are used in practice but also because they pose
several challenges to econometricians and practitioners.
First, empirical work based on time series data assumes that the underlying time series is
stationary. Althoughwe have discussed the concept of stationarity intuitively in Chapter 1,
we discuss it more fully in this chapter. More specifically, we will try to find out what
stationarity means and why one should worry about it.

Second, in Chapter 12, on autocorrelation, we discussed several causes of autocorrelation.

Sometimes autocorrelation results because the underlying time series is nonstationary.

Third, in regressing a time series variable on another time series variable(s), one often obtains
a very high R2 (in excess of 0.9) even though there is no meaningful relationship between the
two variables. Sometimes we expect no relationship between two variables, yet a regression
of one on the other variable often shows a significant relationship. This situation exemplifies
the problem of spurious, or nonsense, regression, whose nature will be explored shortly. It
is therefore very important to find out if the relationship between economic variables is
spurious or nonsensical. We will see in this chapter how spurious regressions can arise if time
series are not stationary.

Fourth, some financial time series, such as stock prices, exhibit what is known as the
random walk phenomenon. This means the best prediction of the price of a stock, say IBM,
tomorrow is equal to its price today plus a purely random shock (or error term). If this were
in fact the case, forecasting asset prices would be a futile exercise.

Fifth, regression models involving time series data are often used for forecasting. In view of
the preceding discussion, we would like to know if such forecasting is valid if the underlying
time series are not stationary.

Finally, causality tests (recall the Granger and Sims causality tests discussed in Chapter 17)
assume that the time series involved in analysis are stationary. Therefore, tests of stationarity
should precede tests of causality.

There are several tests to detect the unit root in the series

1. Tradition Test

a) Graphical Analysis.

b) The correlogram Test.

1
 OLS AFTRT AVERAGING ITEM

LATANT
SEM-CB VARIABLES SEM-PLUS

Simply OLS, Logit. Probit

Unit-Root

Co-integration /ARDL
CROSS ECONOMETRIC MODELING
2SLS/IVREG TIME SERIES
SECTIONAL VARIABLES

Non Paramatric
SEM

Panel Unit Root, Cintegration

FE, RE, BE, FGLS PANAL DATA

2SLS/GMM

2
 UNIT ROOT TEST
(TO MEASURE THE STATIONARITY)

INTEGRATED AT
SAME ORDER

IF NO Johansen Test Of IF YES

Cointegration
(To measure The Long-run
Relationship)

Cointegration Does Cointegration does

Not exists exists

GRANGER CAUSALITY TEST VECTOR ERROR CORRECTION MODEL

(TO MEASURE THE CAUSAL (TO MEASURES THE SHORT RUN AND
AND FEEFCT RELATIONSHIP) LONGRUN RELATIONSHIP)

3
4
2. Modern (Statistical) Method

a) The Augmented Dickey-Fuller (ADF) Test.

b) The Pillips-Perron (PP) Test.

c) Dickey-Fuller GLS (ERS) Test.

d) Kwiatkowski-Philips-Schmidt-Shin (KPSS) Test.

e) Elliott-Rothenberg-Stock Point-Optimal Test.

f) Ng-Perron Test.

3. Panel Unit Root

a) First Generation Test [Cross Section Independence]

If Homogeneity

(Non Stationary Test, where null hypothesis of a unit root.)

i) Lm Levin Choi (LLC)

ii) Breitung (2000)

(Stationary Test, null hypothesis of no unit root)

iii) Hadri (2000)

If Heterogeneity

i) Im, Pesaran, Shin Test.

ii) Maddala and Wu

b) Second Generation Tests [Cross Section Dependence]

If Factor structure

i) Pesaran (2007)

ii) Choi (2002)

If other approach

iii) O’Connell (1998)

iv) Chang (2002, 2004)

5
6
 Unit Root Test for Time Series Data

Traditional Method of Unit Root

1. Graphical Analysis

Testing for Structural Changes

The macroeconomic data introduced in Section 21.1 (see the book’s website for the actual
data) are for the period 1947–2007, a period of 61 years. In this period the U.S. economy
experienced several business cycles of varying durations. Business cycles are marked by
periods of recessions and periods of expansions. It is quite likely that one business cycle is
different from another, which may reflect structural breaks or structural changes in the
economy.

For instance, take the first oil embargo in 1973. It quadrupled oil prices. Prices again
increased substantially after the second oil embargo in 1979. Naturally, these shocks will
affect economic behavior. Therefore, if we were to regress personal consumption expenditure
(PCE) on disposable personal income (DPI), the intercept, the slope, or both are likely to
change from one business cycle to another (recall the Chow test of structural breaks). This is
what is meant by structural changes.

Perron, for instance, has argued that the standard tests of the unit root hypothesis may not be
reliable in the presence of structural changes. There are ways to test for structural changes
and to account for them, the simplest involving the use of dummy variables. But a discussion
of the various tests of structural breaks will take us far afield and is best left for the
references.

7
For the References:

1. James H. Stock and Mark W. Watson, Introduction to Econometrics, 2d ed.,

Pearson/Addison-Wesley, Boston, 2007, pp. 565–571.

2. G. S. Maddala and In-Moo Kim, Unit Roots, Cointegration, and Structural Change,
Cambridge University Press, New York, 1998.

2) The correlogram Test

Date: 02/01/24 Time: 07:31

Sample: 1970Q1 1991Q4
Included observations: 88
Autocorrelation Partial Correlation AC PAC Q-Stat Prob

1 0.969 0.969 85.462 0.000

2 0.935 -0.058 166.02 0.000
3 0.901 -0.020 241.72 0.000
4 0.866 -0.045 312.39 0.000
5 0.830 -0.024 378.10 0.000
6 0.791 -0.062 438.57 0.000
7 0.752 -0.029 493.85 0.000
8 0.713 -0.024 544.11 0.000
9 0.675 0.009 589.77 0.000
10 0.638 -0.010 631.12 0.000
11 0.601 -0.020 668.33 0.000
12 0.565 -0.012 701.65 0.000
13 0.532 0.020 731.56 0.000
14 0.500 -0.012 758.29 0.000
15 0.468 -0.021 782.02 0.000
16 0.437 -0.001 803.03 0.000
17 0.405 -0.041 821.35 0.000
18 0.375 -0.005 837.24 0.000
19 0.344 -0.038 850.79 0.000
20 0.313 -0.017 862.17 0.000
21 0.279 -0.066 871.39 0.000
22 0.246 -0.019 878.65 0.000
23 0.214 -0.008 884.22 0.000
24 0.182 -0.018 888.31 0.000
25 0.153 0.017 891.25 0.000
26 0.123 -0.024 893.19 0.000
27 0.095 -0.007 894.38 0.000
28 0.068 -0.012 894.99 0.000
29 0.043 -0.007 895.24 0.000
30 0.019 -0.005 895.29 0.000

8
Taking The First Difference
Date: 02/01/24 Time: 07:34
Sample (adjusted): 1970Q2 1991Q4
Included observations: 87 after adjustments
Autocorrelation Partial Correlation AC PAC Q-Stat Prob

1 0.316 0.316 9.0136 0.003

2 0.186 0.095 12.165 0.002
3 0.049 -0.038 12.389 0.006
4 0.051 0.033 12.631 0.013
5 -0.007 -0.032 12.636 0.027
6 -0.019 -0.020 12.672 0.049
7 -0.073 -0.062 13.188 0.068
8 -0.289 -0.280 21.380 0.006
9 -0.067 0.128 21.820 0.009
10 0.019 0.100 21.855 0.016
11 0.037 -0.008 21.991 0.024
12 -0.239 -0.311 27.892 0.006
13 -0.117 0.011 29.314 0.006
14 -0.204 -0.114 33.712 0.002
15 -0.128 -0.051 35.474 0.002
16 -0.035 -0.021 35.610 0.003
17 -0.056 -0.019 35.956 0.005
18 0.009 0.122 35.965 0.007
19 -0.045 -0.071 36.195 0.010
20 0.066 -0.126 36.694 0.013
21 0.084 0.089 37.519 0.015
22 0.039 -0.060 37.696 0.020
23 -0.068 -0.121 38.259 0.024
24 -0.032 -0.041 38.384 0.032
25 0.013 0.092 38.406 0.042
26 -0.064 -0.143 38.932 0.049
27 -0.017 -0.081 38.970 0.064
28 -0.038 -0.051 39.156 0.078
29 0.005 0.056 39.160 0.099
30 -0.100 -0.141 40.516 0.095

9
2. Modern (Statistical) Method

a) The Augmented Dickey-Fuller (ADF) Test

In conducting the DF test as in Eqs. (21.9.2), (21.9.4), and (21.9.5), it was assumed that the
error term ut was uncorrelated. But in case the ut are correlated, Dickey and Fuller have
developed another test, known as the augmented Dickey–Fuller (ADF) test.

Null Hypothesis: GDP has a unit root

Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=11)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -0.547205 0.8756

Test critical values: 1% level -3.508326
5% level -2.895512
10% level -2.584952

*MacKinnon (1996) one-sided p-values.

Null Hypothesis: D(GDP) has a unit root

Exogenous: Constant
Lag Length: 0 (Automatic - based on SIC, maxlag=11)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -6.630339 0.0000

Test critical values: 1% level -3.508326
5% level -2.895512
10% level -2.584952

b) The Pillips-Perron (PP) Test.

An important assumption of the DF test is that the error terms ut are independently and identically distributed.
The ADF test adjusts the DF test to take care of possible serial correlation in the error terms by adding the
lagged difference terms of the regressand. Phillips and Perron use nonparametric statistical methods to take care
of the serial correlation in the error terms without adding lagged difference terms. Since the asymptotic
distribution of the PP test is the same as the ADF test statistic, we will not pursue this topic here.

Null Hypothesis: GDP has a unit root

Exogenous: Constant
Bandwidth: 4 (Newey-West automatic) using Bartlett kernel

Adj. t-Stat Prob.*

Phillips-Perron test statistic -0.322401 0.9162

Test critical values: 1% level -3.507394
5% level -2.895109
10% level -2.584738

*MacKinnon (1996) one-sided p-values.

10
Residual variance (no correction) 1275.723
HAC corrected variance (Bartlett kernel) 2284.997

Null Hypothesis: D(GDP) has a unit root

Exogenous: Constant
Bandwidth: 1 (Newey-West automatic) using Bartlett kernel

Adj. t-Stat Prob.*

Phillips-Perron test statistic -6.607376 0.0000

Test critical values: 1% level -3.508326
5% level -2.895512
10% level -2.584952

c) Dickey-Fuller GLS (ERS) Test.

In statistics and econometrics, the ADF-GLS test (or DF-GLS test) is a test for a unit root in
an economic time series sample. It was developed by Elliott, Rothenberg and Stock (ERS) in
1992 as a modification of the augmented Dickey–Fuller test (ADF).[1]

A unit root test determines whether a time series variable is non-stationary using an
autoregressive model. For series featuring deterministic components in the form of a constant
or a linear trend then ERS developed an asymptotically point optimal test to detect a unit
root. This testing procedure dominates other existing unit root tests in terms of power. It
locally de-trends (de-means) data series to efficiently estimate the deterministic parameters of
the series, and use the transformed data to perform a usual ADF unit root test. This procedure
helps to remove the means and linear trends for series that are not far from the non-stationary
region.[2]

Null Hypothesis: GDP has a unit root

Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=11)

t-Statistic

Elliott-Rothenberg-Stock DF-GLS test statistic 1.580128

Test critical values: 1% level -2.592129
5% level -1.944619
10% level -1.614288

*MacKinnon (1996)

Null Hypothesis: D(GDP) has a unit root

Exogenous: Constant
Lag Length: 0 (Automatic - based on SIC, maxlag=11)

t-Statistic

Elliott-Rothenberg-Stock DF-GLS test statistic -5.284282

Test critical values: 1% level -2.592129

11
 5% level -1.944619
10% level -1.614288

d) Kwiatkowski-Philips-Schmidt-Shin (KPSS) Test.

e) Elliott-Rothenberg-Stock Point-Optimal Test.

f) Ng-Perron Test.

12
13
14