Time-Series Econometrics

1. Stochastic Process

 Analysis of time-series is based on the modelling of stochastic process.

 A stochastic process is a collection of random variables ordered in time.
 The stochastic process evolves in time according to probabilistic laws.
 An observed time-series is considered to be one realization of a stochastic process.

Type of Stochastic Definition Functional Specification

Process
Purely Random (White Each element has independent and Yt  ut with
Noise) identical distribution with constant
E (u t )    t; var(u t )   2  t ;
(or zero) mean and constant
variance cov(u t , u t s )  0  s  0

For simplicity, one may also assume that:

E (u t )  0  t ; var(ut )   2  t ;
cov(ut , ut s )  0  s  0

Example: The random disturbance terms

in CLRM, ut ~ IIN(0, )
2

Random Walk In each period, the variable takes a Yt  Yt  1  u t or Yt  Yt1  ut

random deviation from its previous If ut is white noise with
value, and the deviations are
E (u t )    t ; var(ut )   2  t ;
independently and identically
distributed in size cov(ut , ut s )  0  s  0
=> E(Yt )  t; var(Yt )  t.
2

Moving Average (MA) Value of Y at time point t is moving Moving Average process of order m:
Process average of the current and past Yt   0 u t   1u t 1   2 u t  2  ...   m u t  m
values of the random disturbance Here, ut is white noise with
term. E (u t )  0  t ; var(ut )   2  t ;
cov(ut , u t s )  0  s  0
 m 
=> E( Yt )  0; var( Yt )   2    j2 
 j 0 
Autoregressive (AR) Value of Y at time point t depends Autoregressive process of order r:
Process on its previous values and random Yt   1Yt 1   2Yt  2  ...   r Yt  r  u t
disturbance at that time point.
ARMA Process The variable Y has characteristics of ARMA process of order r and m, i.e.
both AR and MA ARMA (r, m):
Yt  1Yt 1   2Yt 2  ...   r Yt r  ut 
1ut 1   2ut 2  ...   m ut m
Here,
E (u t )  0  t ; var(ut )   2  t ;
cov(ut , u t s )  0  s  0
ARIMA Process Differencing non-stationary time- ARIMA (r, d, m) means the time-series
series to make it integrated and has to be differenced d times to make it
modelling the differenced series as stationary and the stationary time series
ARMA process can be modelled as ARMA(r, m)

ARIMA (r, 0, 0) indicates purely AR(r)

stationary process

ARIMA (0, 0, m) indicates purely MA(m)

stationary process

ARIMA (0, d, 0) indicates that the time-

series is integrated of order d, i.e., I(d)

2. Random Walk Models

Basic Model: Yt   0   1t   2Yt 1   t

Random Walk Econometric Specification

Random walk without drift Yt  Yt 1  u t This is an AR(1) model
and trend If ut is white noise with
0  0;1  0;2  1 E (u t )  0  t ; var(ut )   2  t ;
cov(ut , u t s )  0  s  0
E(Yt )  0, var(Yt )  t 2
(Assuming Y0  0 )
Violates the condition of stationarity
Random walk with drift, but Yt   0  Yt 1  u t Drift parameter: θ0
no trend Drift will be upward or downward depending
0  0;1  0;2  1 on θ0>0 or θ0<0
If ut is white noise with
E (u t )  0  t ; var(ut )   2  t ;
cov(ut , u t s )  0  s  0
E(Yt )  t 0 , var(Yt )  t 2
 (Assuming Y0  0 )
Violates the condition of stationarity
Deterministic Trend Yt   0   1 t  u t If ut is white noise with
0  0;1  0;2  0 E (u t )  0  t ; var(ut )   2  t ;
cov(ut , u t s )  0  s  0
E (Yt )   0  1t , var(Yt )   2
Violates the condition of stationarity
Random walk with drift and Yt   0   1t  Yt 1  u t Yt is non-stationary
deterministic trend
0  0;1  0;2  1
Stochastic process with drift and Yt   0   1t   2Yt 1  u t Yt is stationary around a deterministic trend
deterministic trend
0  0;1  0;2  1

3. An Example
Consider the model: Yt  Yt 1  ut

Three possibilities
(1) When   1 , the series is non-stationary and explosive. Past shocks have a greater
impact than current ones.
(2) When   1, the series is non-stationary shocks persist at full force, and the series is not
mean-reverting. This is the random walk model and where the variance increases with t
and we have the infinite variance problem.
(3) When   1 , series is stationary and the effects of shocks die out exponentially according
to. The series reverts to its mean.
Typically, we are interested in the last two scenarios, i.e.,   1 and   1 . The question is
whether we have a unit root or not (also known as a random walk) i.e.,   1?

For the model Yt  Yt 1  ut with   1

 Y1  Y0  u1
 Y2  Y1  u2   ( Y0  u1 )  u2   Y0  u1  u2
2

 Y3  Y2  u3   (  Y0  u1  u2 )  u3   Y0   u1  u2  u3

2 3 2

This shows that the current error is the sum of all the previous shock weighted by the coefficient
() declining exponentially. How fast the effect of these previous errors will die out depends on
the value of 
When  < 1 the time-series is stationary. In this case, the time-series looks jagged and it never
wanders too far from the mean. The effect of the errors decay and disappear over time. Impact
of recent events are relatively more important than what happened a long time ago.

(1) Yt    Yt1  ut with   1

(2) Yt    Yt1  ut with   1

3.1 Model 1
t 1
Yt  Y0  t   ut 
 0

E Yt Y0   Y0  t and varYt Y0   t 2

Both conditional mean and variance depend on time

3.2 Model 2
t 1 
Yt  Y0     (  ut  )     (  ut  )
 0  0


 

E (Yt )       (Unconditional mean)
 0 1 

    2 2   2
 

var(Yt )  var   (  ut  )         (Unconditional variance)
 1 
2
  0    0
Both unconditional mean and variance are constant (independent of time)

3.3 Importance of Stationarity in Time-Series

 It is necessary to test if a time-series is stationary (i.e., if there is no change in property of

the series in respect of the mean, variance and autocorrelation structure over time)
 The null hypothesis is generally defined as the presence of a unit root and the alternative
hypothesis is either stationarity, trend stationarity or explosive root depending on the
test used
 Persistence of shocks for non-stationary series
 Problem of spurious regressions - two variables trending over time can have a high R 2
even if the two are totally unrelated
 Standard assumptions for asymptotic analysis will not be valid for non-stationary
variables - testing of hypothesis may not be valid

3.4 Trend Stationary and Difference Stationary

 Decision
General Specification Yt  0  1t 2Yt 1  ut Nature of the series depends on sign and
magnitude θ0, θ1 and θ2
Random walk without drift Yt  Yt 1  ut Difference stationary process (DSP)
and without trend
Or, Yt  Yt 1  ut
0  0;1  0;2  1
Or, Yt  ut
Random walk with drift, Yt  0  Yt 1  ut Y has positive stochastic trend if θ0 > 0
but no trend Y has negative stochastic trend if θ0 < 0
Yt  0  ut
0  0;1  0;2  1 Difference stationary process (DSP)
Deterministic Trend Yt   0   1 t  u t Trend stationary process (TSP)
0  0;1  0;2  0 Yt  E (Yt )  u t
Random walk with drift Yt   0   1t  Yt 1  u t Yt is non-stationary
and deterministic trend  Yt   0   1t  u t
0  0;1  0;2  1
Stochastic Process with Yt   0   1t   2Yt 1  u t Yt is stationary around a deterministic
drift and deterministic  Yt   0   1t  ( 2  1)Yt 1  u t trend
trend
0  0;1  0;2  1
 4. Testing for Presence of Unit Roots

4.1 Steps for Unit Root Test

1. Define the data as time-series
2. Make log transformation of variables, if necessary. Standard unit root tests assume linearity under both the null and the
alternative. Violation of this assumption may cause severe size and power distortions both in finite and large samples
3. Setting lags and differences of the variables
4. Setting the lag length (for ADF and PP tests)
5. Carrying out the test with different types of random walk
4.2 Specification of Different Unit Root Tests
No Drift; No Trend Drift; No Trend Drift and Trend Hypothesis
Dickey- Yt   2Yt 1  ut Yt   0   2Yt 1  ut Yt   0  1t   2Yt 1  ut H0 :  2  0
Fuller (DF)
Test Yt   2Yt 1  ut ; Yt  Yt 1   0   2Yt 1  Yt 1  ut Yt   0  1t   2Yt 1  ut ; H1 :  2  0
 2  2  1 Yt   0   2Yt 1  ut ; 2   2  1  2  2  1
Augmented p p p H0 :  2  0
Dickey- Yt   2Yt 1    j Yt  j  ut Yt   0   2Yt 1    j Yt  j  ut Yt   0  1t   2Yt 1    j Yt  j  ut
Fuller (ADF) j 1 j 1 j 1 H1 :  2  0
Test p p p
Yt   2Yt 1    j Yt  j  ut ; Yt  0   2Yt 1    j Yt  j  ut ; Yt  0  1t   2Yt 1    j Yt  j  ut ;
j 1 j 1 j 1

 2  2  1  2  2  1  2  2  1
Phillips- Yt  2Yt1  ut Yt   0   2Yt 1  ut Yt   0  1t   2Yt 1  ut H0 :  2  0
Perron (PP)
Test Yt  2Yt1  ut ; Yt  0   2Yt 1  ut ; Yt   0  1t   2Yt 1  ut ; H1 :  2  0
 2  2 1  2  2  1  2  2  1
 4.4 Choice of the Alternative Models

Situation Function Form Strategy

When the time series is flat (i.e. doesn’t have a No drift; No Non-rejection of the null hypothesis:
trend) and potentially slow-turning around Trend Series to be differenced for
zero stationarity
Rejection of the null hypothesis:
Stationary series- need not be
differenced
When the time series is flat and potentially Drift; No Trend Non-rejection of the null hypothesis:
slow-turning around a non-zero value Series to be differenced for
stationarity
Rejection of the null hypothesis:
Stationary series- need not be
differenced
When the time series has a trend in it (either Drift; Trend Non-rejection of the null hypothesis:
up or down) and is potentially slow-turning Series to be differenced for
around a trend line stationarity
Rejection of the null hypothesis:
Series is trend stationary; To be
analysed by using a time trend instead
of differencing the data
Note: If the series is exponentially trending, logarithmic transformation of the series is
necessary before differencing it.

Summary of the Steps Involved in the Tests

1. Test for unit roots in the process of the variable with the drift and the time trend.

If the null hypothesis (H0: 2 = 0) is not rejected, there are unit roots.

If the null hypothesis rejected, check for the presence of the time trend. If the corresponding
null hypothesis is rejected, the process is stationary around a time trend.

If the coefficient of time variable is significant and presence of unit roots is not rejected,
the variable has unit roots with the time trend.

2. If there is no time trend, test for the unit roots with drift.

Check the process of unit roots. If there are no unit roots, the variable is stationary.

If a constant term is significant, check the results for unit roots. If the null hypothesis is
not rejected, the variable has unit roots with the drift. If the null hypothesis is rejected,
the variable has no unit roots.

3. If there is no constant, test for unit roots with no drift and no time trend. If there are no
unit roots, the process is stationary.
Selection of Functional Form for Unit Root Test

p Visualization of the Time-series

Estimate the Model : Yt   0   1t   2Yt 1    j Yt  j  ut
j 1
with Trend and Intercept

Is H 0 :  2  0 Conclusion: Presence
If YES of No Unit Root
rejected?

If NOT: Testing for

If YES
Presence of Trend

Is H 0 :  1  0 rejected Is H 0 :  2  0 Conclusion: Presence

given non-rejection of If YES rejected using If NOT of Unit Root
H0 :  2  0 normal distribution?

If NOT:

p
Estimate the Model: Yt  0   2Yt1   Y
j1
j t j  ut

Conclusion: Presence
Is H 0 :  2  0
If YES No Unit Root
rejected?

If NOT: Testing for

Presence of Drift If YES

Is H 0 : 0  0 rejected Is H 0 :  2  0
Conclusion: Presence
given non-rejection of If YES rejected using If NOT
of Unit Root
H0 :  2  0 normal distribution?

If NOT:

p
Estimate the Model: Yt   2Yt1   Y
j1
j t j  ut

Is H 0 :  2  0 If YES Conclusion: Presence of No Unit Root

rejected?
If NOT Conclusion: Presence of Unit Root
4.6 Comparative Analysis among the DF, ADF and PP Tests of Unit Root

 Unlike the DF test, the ADF test allows for higher-order autoregressive processes by
p
including the term(s)  Yt  j
j 1

 Although the DF and the ADF tests are frequently used in testing for unit roots, there are
problems of size distortions and low power. In DF test, the problem of autocorrelation is
not corrected for. There is problem of selection of lag length in ADF test. The information
criteria such as AIC or BIC often select a low value of the lag length.

 The PP test is based on the similar equation as employed in the DF test (without the lagged
differenced terms included in the ADF test).

 The PP test incorporates automatic non-parametric correction procedure for

autocorrelated residuals, and usually gives the same conclusions as the ADF tests

 Monte Carlo studies suggest that the PP test has greater power than the ADF test