Lecture III

By:

Amsalu B. (MSc.)

Lecturer, Department of Economics

Unity University

Email: [email protected]

Unity, Ethiopia
 3.1 Nature of the Time Series data

 Time series data have become so frequently and intensively used

in empirical research and thus, econometricians have recently
begun to pay very careful attention to such data.

 Time series data are data collected for a single entity (person,
firm, and country) collected (observed) at multiple time periods.

 A time series is a sequence of numerical data in which each item is

associated with a particular instant in time.

 Example: Monthly unemployment, weekly measures of money

supply, M1 and M2, daily closing prices of stock indices, and so on
 3.1 Nature of the Time Series data

 Thus, the way we collect time series data can be characterised as

daily (stock prices, weather report), weekly (gasoline supplied in
thousands of barrels), monthly (unemployment rate, CPI),
quarterly (GDP), & annual (GDP, Budget).

 Quinquennially: Every 5 years (e.g: the census of manufactures)

 Decennially (e.g: Census of population)

 Exchange rates daily date for 2 years=730 observation

 Inflation rate for Ethiopia, quarterly data for 30 years =30*4=120

observation.
 3.1 Nature of the Time Series data
 Gross domestic investment in Ethiopia of annual data for 40 yrs; 40*1= 40
observations.

 Time Series data Vs Cross sectional data

Time Series Cross sectional

• It coming with temporal ordering over
period of time on a single entity. • At a given point of time
• the past can affect the future, but not • a different sample has drawn from the
vice versa population will generally yield different
• We have d/t data for d/t samples values of the independent and
• Also viewed as random varaibles; we dependent variables
do not know what the annual growth • We have d/t data for d/t year
in output will be in Ethiopia during the • The OLS estimates computed from d/t
coming year. i.e, the outcomes of these random samples will generally differ
random variables are not foreknown. and this why we consider OLS
estimators to be random variables.
 3.1 Nature of the Time Series data
 A sequence of random variables indexed by time is called a
stochastic process or a time series process. (“Stochastic” is a
synonym for random.)

 When we collect a time series data set, we obtain one possible

outcome, or realization, of the stochastic process.

 We can only see a single realization, because we cannot go back

in time and start the process over again.

 This is analogous to cross-sectional analysis where we can collect

only one random sample.)
 3.1 Nature of the Time Series data

Important terminology :
 Univariate analysis examines a single data
series.
 Bivariate analysis examines a pair of series.
 The term vector indicates that we are
considering a number of series: two, three, or
more.
 The term „„vector‟‟ is a generalization of the
univariate and bivariate cases.
 3.2 Stationary and non-stationary Stochastic Processes
 A random or stochastic process is a collection of random
variables ordered in time.

 If we let Y denote a random variable, and if it is continuous, we

denote it as Y(t), but if it is discrete, we denoted it as Yt.

 Example of Yt is GDP, CPI, PDI; since most economic data are

collected at discrete points in time, we use Yt notation.

 If we let Y represent GDS, for our data we have

Y1,Y2,Y3,...,Y21,Y22,Y23, where the subscript 1 denotes the first
observation (i.e., GDS of 1991/1992) and the subscript 23 denotes
the last observation (i.e., GDS of 2014/2015).
 3.2 Stationary and non-stationary Stochastic Processes

 Example of Y(t) is electro cardiogram, record of heart activity.

 N.B: Each of these Y‟s is a random variable.

A. Stationary Stochastic Processes

 A stochastic process is said to be stationary “if its mean and

variance are constant over time and the value of the covariance
between the two time periods depends only on the distance or
gap or lag between the two time periods and not the actual time
at which the covariance is computed.
 3.2 Stationary and non-stationary Stochastic Processes
 In the time series literature, such a stochastic
process is known as a weakly stationary, or
covariance stationary, or second-order
stationary, or wide sense, stochastic process.
 To explain weak stationarity, let Yt be a
stochastic time series with these properties:
Mean : E (Yt )  
Variance : Var(Yt )  E(Yt   ) 2   2

Co var iance :  k  E[(Yt   )(Yt k   )]

 3.2 Stationary and non-stationary Stochastic
Processes
Where γk, the covariance (or autocovariance) at
lag k, is the covariance between the values of Yt
and Yt+k, that is, between two Y values k periods
apart.
If k=0, we obtain γ0, which is simply the variance
of Y(=σ2); if k=1,γ1 is the covariance between two
adjacent values of Y.
 1  E[(Y1   )(Yt 1   )]

Suppose we shift the origin of Y from Yt to Yt+m

(say, from 1997 to 2002 for our GDS data). Now if
Yt is to be stationary, the mean, variance, and
autocovariances of Yt+m must be the same as
those of Yt.
 3.2 Stationary and non-stationary Stochastic Processes
In short, if a time series is stationary, its mean, variance,
and autocovariance (at various lags) remain the same no
matter at what point we measure them; that is, they are
time invariant.
 Such a time series will tend to return to its mean
(called mean reversion) and fluctuations around this
mean (measured by its variance) will have a broadly
constant amplitude.
If a time series is not stationary in the sense just defined,
it is called a nonstationary time series (keep in mind we
are talking only about weak stationarity).
In other words, a nonstationary time series will have a
time-varying mean or a time-varying variance or both.
 3.2 Stationary and non-stationary Stochastic Processes
Why Stationary time series are important?
 Because if a time series is nonstationary, we can study its
behavior only for the time period under consideration.
 Each set of time series data will therefore be for a particular
episode.
 As a consequence, it is not possible to generalize it to other time
periods. Therefore, for the purpose of forecasting, such
(nonstationary) time series may be of little practical value.
A. Non stationary Stochastic Processes
 A special type of stochastic process (or time series), is called, a
purely random, or white noise, process.
 3.2 Stationary and non-stationary Stochastic Processes

 We call a stochastic process purely random if it has zero mean,

constant variance σ2, and is serially uncorrelated.
 One of the classical example of non stationary time series is the
random walk model (RWM).
 It is often said that asset prices, such as stock prices or exchange
rates, follow a random walk; that is, they are nonstationary.
 We have two types of random walks
(1) random walk without drift (i.e., no constant or intercept
term) and
(2) random walk with drift (i.e., a constant term is present).
 3.2 Stationary and non-stationary Stochastic Processes

1. Random walk without drift

 Suppose ut is a white noise error term with
mean 0 and variance 2
.
 Then the series Yt is said to be a random walk if
Yt  Yt 1  u t  shows, the value of Y at time t is
equal to its value at time (t−1) plus a random
shock; thus it is an AR (1) model.
 We can think of Yt  Yt 1  u t as a regression of
Y at time t on its value lagged one period.
Y1  Y0  u t
Y2  Y1  u 2  Yo  u1  u 2
Y3  Y2  u 3  Y0  u1  u 2  u 3
 3.2 Stationary and non-stationary Stochastic Processes
 If the process started at some time 0 with a value of Y0, we have

Y1  Y0   ut
E (Y1 )  E (Y0   ut )  Y0 Var(Yt )  t 2

 The mean of Y is equal to its initial, or starting, value, which is

constant, but as t increases, its variance increases indefinitely,
thus violating a condition of stationarity.

 In short, the RWM without drift is a nonstationary stochastic

process. In practice Y0 is often set at zero, in which case E (Yt)
=0.
 3.2 Stationary and non-stationary Stochastic Processes

 An interesting feature of RWM is the persistence of random

shocks (i.e., random errors), which is clear from
Y1  Y0   u t

 Yt is the sum of initial Y0 plus the sum of random shocks.

 As a result, the impact of a particular shock does not die away.
 For example, if U2=2 rather than U2=0, then all Yt‟s from Y2
onward will be 2 units higher and the effect of this shock never
dies out.
 That is why random walk is said to have an infinite memory. The
implication is that, random walk remembers the shock forever;
that is, it has infinite memory.
 3.2 Stationary and non-stationary Stochastic Processes
☞We can rewrite the above equation as: Y 1  Y0   u t
(Yt  Yt 1 )  Yt  u t

☞Where  is the first difference operator. It is

easy to show that, while Yt is nonstationary, its
first difference is stationary. In other words, the
first differences of a random walk time series are
stationary.
2. Random Walk with Drift
Let us modify Y  Y   u as follows:
1 0 i
 3.2 Stationary and non-stationary Stochastic Processes
Where δ is known as the drift parameter. The
name drift comes from the fact that if we write
the preceding equation as .
it shows that Yt drifts upward or downward,
depending on δ being positive or negative. It is
also an AR(1) model.
Following the procedure discussed for random
walk without drift, it can be shown that for the
random walk with drift model :
 3.2 Stationary and non-stationary Stochastic Processes
As you can see for RWM with drift, the mean as well as the
variance increases over time.
Thus, it violating the conditions of (weak) stationary. In short,
RWM, with or without drift, is a nonstationary stochastic
process.
The random walk model is an example of what is known in the
literature as a unit root process.
Unit Root Stochastic Process
Let us write the RWM as:
This model resembles the Markov first-order autoregressive
model that we discussed on autocorrelation.
 3.2 Stationary and non-stationary Stochastic Processes
 If ρ=1, becomes a RWM (without drift). If ρ is in fact 1, we face
what is known as the unit root problem, that is, a situation of
nonstationary; we already know that in this case the variance of
Yt is not stationary.

 The name unit root is due to the fact that ρ=1. Thus the terms
nonstationary, random walk, and unit root can be treated as
synonymous.

 If, however, |ρ|<1, that is if the absolute value of ρ is less than

one, then it can be shown that the time series Yt is stationary in
the sense we have defined it.
 3.3 Trend Stationary and Difference Stationary Stochastic Processes
 If the trend in a time series is completely predictable and not variable, we call
it a deterministic trend, whereas if it is not predictable, we call it a stochastic
trend. To make the definition more formal, consider the following model of the
time series Yt.

 ----------------(a)

 Where ut is a white noise error term and where t is time measured

chronologically. Now we have the following possibilities:

 RWM without drift:

--------(b)=non stationary

--------(c)=stationary, Hence, a RWM without

drift is a difference stationary process (DSP).
Pure random walk : If in (a),  0  0, 1  0,  2  1, we get
3.3 Trend Stationary and Difference Stationary Stochastic Processes

 RWM with drift:Pure random walk with drift : If in (a),   0,   0,  0 1 2  1, we get

----------(d)-non stationary
-----------(e)—stationary, this means Yt will exhibit a
positive (β1>0) or negative (β1<0) trend. Such a trend is called a stochastic trend.
Equation (e) is a DSP process because the nonstationarity in Yt can be eliminated
by taking first differences of the time series.

 Deterministic trend:
Pure random walk with drift : If in (a),  0  0, 1  0,  2  1, we get

Yt   0  1t  ut
which is called a trend stationary process (TSP).
 3.3 Trend Stationary and Difference Stationary Stochastic Processes

Although the mean of Yt is β0+β1t, which is not

constant, its variance (=σ2) is constant.

Once the values of β0 and β1 are known, the mean can

be forecasted perfectly. Therefore, if we subtract the
mean of Yt from Yt, the resulting series will be
stationary, hence the name trend stationary.

This procedure of removing the (deterministic) trend is

called detrending.
3.3 Trend Stationary and Difference Stationary
Stochastic Processes
Random walk with drift & deterministic
trend: If in (a)  0  0, 1  0,  2  1, we get
-----(f) non stationary
Deterministic trend with stationary AR (1)
component:
If in (a)  0  0, 1  0,  2  1, we get
Yt   0  1t   2Yt 1  U t , which is
stationary around the deterministic trend.
 3.4 Integrated Stochastic Process

The random walk model is a specific case of a

more general class of stochastic processes known
as integrated processes.
the RWM without drift is nonstationary, but its
first difference is stationary.
the RWM without drift integrated of order 1,
denoted as I(1).
Similarly, if a time series has to be differenced
twice (i.e., take the first difference of the first
differences) to make it stationary, we call such a
time series integrated of order 2.
 3.4 Integrated Stochastic Process
 In general, if a (nonstationary) time series has to be differenced d
times to make it stationary, that time series is said to be integrated
of order d. A time series Yt integrated of order d is denoted as Yt
∼I(d).

 If a time series Yt is stationary to begin with (i.e., it does not

require any differencing), it is said to be integrated of order zero,
denoted by Yt ∼I(0).

 Thus, we will use the terms “stationary time series” and “time
series integrated of order zero” to mean the same thing.
 3.4 Integrated Stochastic Process
 Most economic time series are generally I(1); that is, they
generally become stationary only after taking their first
differences.

 Properties of Integrated Series

Let Xt, Yt and Zt be three time series

 i. If Xt ∼I(0) and Yt ∼I(1),then Zt =(Xt +Yt)=I(1); that is, a linear

combination or sum of stationary and nonstationary time series is
nonstationary.

 ii. If Xt ∼I(d), then Zt =(a+bXt)=I(d), where a and b are constants.

That is, a linear combination of an I(d) series is also I(d). Thus, if
Xt ∼I(0), then Zt =(a+bXt)∼I(0).
 3.4 Integrated Stochastic Process
 iii. If Xt ∼I(d1) and Yt ∼I(d2), then Zt =(aXt +bYt)∼I(d2), where
d1<d2.

 iv. If Xt ∼I(d) and Yt ∼I(d), then Zt =(aXt +bYt)∼I(d*); d* is

generally equal to d, but in some cases d*<d.
 3.5 Tests of Stationarity: The Unit Root Test

A test of stationarity (or nonstationarity)

that has become widely popular over the
past several years is the unit root test.
 ------------------(i)
where ut is a white noise error term
We know that if ρ =1, that is, in the case of
the unit root, (i) becomes a random walk
model without drift, which we know is a
nonstationary stochastic process.
 3.5 Tests of Stationarity: The Unit Root Test

 Therefore, why not simply regress Yt on its (one period) lagged

value Yt −1 and find out if the estimated ρ is statistically equal to
1?

 If it is, then Yt is nonstationary. This is the general idea behind

the unit root test of stationarit y.

 For theoretical reasons, we manipulate (i) as follows: Subtract Yt

−1 from both sides of (i) to obtain: Yt  Yt 1  Yt 1  Yt 1  ut

, which can be written alternatively -----------------(ii)

 3.5 Tests of Stationarity: The Unit Root Test

 Where δ = (ρ−1) and , as usual, is the first-difference operator.

 In practice, therefore, instead of estimating (i), we estimate (ii)

and test:

 the (null) hypothesis that δ = 0, then ρ = 1, that is we have a unit

root (nonstationary).

the t value of the estimated coefficient of Yt −1 does not follow

the t distribution even in large samples; that is, it does not have
an asymptotic normal distribution. Thus, we use τ(tau)
statistic.
 3.5 Tests of Stationarity: The Unit Root Test

1. Dickey–Fuller (DF) test

 Dickey and Fuller have shown that under the
null hypothesis that δ = 0, the estimated t value
of the coefficient of Yt −1 follows the τ(tau)
statistic.
 In the literature the tau statistic or test is known
as the Dickey–Fuller (DF) test
 In conducting the DF test, it was assumed that
the error term ut was uncorrelated.
 the DF test is estimated in three different forms,
that is, under three different null hypotheses.
 3.5 Tests of Stationarity: The Unit Root Test

 Where t is the time or trend variable.

 Null hypothesis, δ = 0; that is, there is a unit root—the time
series is nonstationary.
 Alternative hypothesis, δ is less than zero; that is, the time
series is stationary.
 Decision rule: If /tau statistics/> /tau critical value/; we rejected
the null hypothesis, it means that Yt is a stationary time series.
 3.5 Tests of Stationarity: The Unit Root Test

2. Augmented Dickey–Fuller (ADF) Test

 In this case the ut are correlated, Dickey and
Fuller have developed a test, known as the
Augmented Dickey–Fuller (ADF) test.
 This test is conducted by “augmenting” the
preceding three equations by adding the lagged
values of the dependent variable ΔYt -i.
The ADF test here consists of estimating the
following regression:
 3.5 Tests of Stationarity: The Unit Root Test

 In ADF we still test whether δ = 0 and the ADF test

follows the same asymptotic distribution as the DF
statistic, so the same critical values can be used.
Hypothesis we use under this test is:

Example: the GDP series using one lagged difference of

natural log of GDP of Ethiopia; the results were as
follows:
 3.5 Tests of Stationarity: The Unit Root Test

Decision rule: The t(=τ) value of the Yt −1 coefficient ( =δ) is

0.34, but this value in absolute terms is much l ess than even the
1% and 5% critical τ value of −4.242 and -3.540 respectively,
again suggesting that even after taking care of possible
autocorrelation in the error term, the Y series is not stationary.
 3.5 Tests of Stationarity: The Unit Root Test

3. The Phillips–Perron (PP) Unit Root Tests

 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.
 3.5 Tests of Stationarity: The Unit Root Test
Phillips and Perron use nonparametric statistical methods
to take care of the serial correlation in the error terms
without adding lagged difference terms.
The Phillips-Perron test involves fitting the following
regression: Yt   0   1t  Yt 1  u t
Under the null hypothesis that ρ = 0, the PP Z(t) and Z(
) statistics have the same asymptotic distributions as the
ADF t-statistic and normalized bias statistics.
 One advantage of the PP tests over the ADF tests is that
the PP tests are robust to general forms of
heteroscedasticity in the error term ut. Another
advantage is that the user does not have to specify a lag
length for the test regression.
 3.5 Tests of Stationarity: The Unit Root Test

In some situation, lack of power in both the ADF

and PPtests is widely acknowledged,
Usually ADF yields superior results than PP test,
if the data set has no missing observations and
structural breaks whilst PP test also yields
superior results than ADF test, if the dataset have
some missing observations and have structural
breaks
Decision rule: since tests statistic, Z(t) value is
greater that critical values we reject the null
hypotheses of non stationary.
THANKS!!!

END!!!

THANKS!!!