Chapter 9

Regression with Time Series

Data:
Stationary Variables

Walter R. Paczkowski
Rutgers University
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 1
Edition Stationary Variables
 Chapter Contents

 9.1 Introduction
 9.2 Stationary and Weak Dependence
 9.3 Serial Correlation
 9.4 Other Tests for Serially Correlated Errors
 9.5 Estimation with Serially Correlated Errors
 9.6 Autoregressive Distributed Lag Models
 9.7 Forecasting
 9.8 Multiplier Analysis

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 2
Edition Stationary Variables
 9.1
Introduction

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 3
Edition Stationary Variables
 When modelling relationships between variables, the
nature of the data that have been collected has an
important bearing on the appropriate choice of an
econometric model.
Two features of time-series data to consider:
Time-series observations on a given economic unit,
observed over a number of time periods, are likely to
be correlated.
Time-series data have a natural ordering according to
time.
There is also the possible existence of dynamic
relationships between variables.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 4
Edition Stationary Variables
 A dynamic relationship is one in which the change in a
variable now has an impact on that same variable, or
other variables, in one or more future time periods.
These effects do not occur instantaneously but are spread,
or distributed, over future time periods.
In particular, it is important to distinguish between cross-
section data (data on a number of economic units at a
particular point in time) and time-series data (data
collected over time on one particular economic unit).
Individuals, households, firms, geographical regions,
countries, or some other entity, etc.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 5
Edition Stationary Variables
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 6
Edition Stationary Variables
 Modelling Dynamic Relationships:
Specify that a dependent variable y is a function of
current and past values of an explanatory variable x.

(9.1) 𝑦 𝑡 =𝛼+ 𝛽0 𝑥𝑡 + 𝛽 1 𝑥 𝑡 −1 + 𝛽2 𝑥𝑡 −2 +… + 𝛽 𝑞 𝑥 𝑡 − 𝑞 +𝑒

Because of the existence of these lagged effects, (9.1) is

called a distributed lag model.

The model is called a finite distributed lag model

because the effect of x on y cuts off after a finite number
of periods q.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 7
Edition Stationary Variables
 Modelling Dynamic Relationships:
An autoregressive model, or an autoregressive process, is
one where a variable y depends on past values of itself.

(9.2) 𝑦 𝑡 =𝛿+𝜃1 𝑦 𝑡 −1 +𝜃2 𝑦 𝑡 − 2 +…+𝜃𝑝 𝑦 𝑡 −𝑝 +𝑒𝑡

A more general model that includes both finite distributed

lag models and autoregressive models as special cases is
the autoregressive distributed lag model.

𝑦 𝑡=𝛿+𝜃1 𝑦𝑡−1+…+𝜃𝑝 𝑦𝑡−𝑝+𝛿0 𝑥𝑡 +𝛿1 𝑥𝑡−1+…+𝛿𝑞 𝑥𝑡−𝑞 +𝑒𝑡

(9.3)

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 8
Edition Stationary Variables
 Modelling Dynamic Relationships:
If we take equation (9.1) and assume that the impact of
past, lagged x’s does not cut off after q periods but goes
back into the infinite past, then we have the infinite
distributed lag (IDL) model

(9.4) 𝑦 𝑡 =∝+ 𝛽 0 𝑥 𝑡 + 𝛽1 𝑥 𝑡 −1 + 𝛽 2 𝑥 𝑡 − 2+ 𝛽 3 𝑥 𝑡 −3 + …+𝑒 𝑡

Assume that the coefficients βs eventually decline in

magnitude with their effect becoming negligible at long
lags.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 9
Edition Stationary Variables
 Modelling Dynamic Relationships:
Another way in which lags can enter a model is through
the error term. For example, if the error et satisfies the
assumptions of an AR(1) model, it can be written as

(9.10) 𝑒 𝑡 =𝑝 𝑒 𝑡 −1 +𝑣 𝑡

with the being uncorrelated.

This model means that the random error at time t is related

to the random error in the previous time period plus a
random component.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 10
Edition Stationary Variables
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 11
Edition Stationary Variables
 Autocorrelations:
If there is no linear association between the variables, then
both the covariance and the correlation are zero.

In any ARDL model where there is a linear relationship

between yt and its lags, yt must be correlated with lagged
values of itself.

Correlations of this kind are called autocorrelations.

When a variable exhibits correlation over time, we say it

is autocorrelated or serially correlated.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 12
Edition Stationary Variables
 Autocorrelations:
Sample autocorrelations are obtained using a sample of
observations for a finite time period, 𝑥_1, 𝑥_2,…, 𝑥_𝑇, to
estimate the population autocorrelations.

To estimate 𝑝1:
𝑇
1
c^
ov(𝑥 𝑡 , 𝑥𝑡 −1 )= ∑
𝑇 −1 𝑡 =2
( 𝑥 𝑡 − 𝑥 )( 𝑥𝑡 −1 − 𝑥 )
𝑇
1
var(𝑥𝑡 )=
^ ∑
𝑇 −1 𝑡 =1
( 𝑥𝑡 − 𝑥 )
2

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 13
Edition Stationary Variables
 Autocorrelations:
Making the substitutions, we get:
𝑇

∑ ( 𝑥𝑡 − 𝑥) ( 𝑥𝑡− 1− 𝑥 )
(9.19) 𝑟1= 𝑡 =2
𝑇

∑ ( 𝑥𝑡 − 𝑥 ) 2
𝑡 =1

More generally, the s-order sample autocorrelation for a

series x that gives the correlation between observations
that are s periods apart is:
𝑇

∑ ( 𝑥 𝑡 − 𝑥 ) ( 𝑥𝑡 − 𝑠 − 𝑥 )
(9.20) 𝑟 𝑠 = 𝑡 =𝑠 +1
𝑇

∑ ( 𝑥𝑡 − 𝑥 ) 2
𝑡 =1

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 14
Edition Stationary Variables
 Autocorrelations:
It is often useful to test whether a sample autocorrelation
is significantly different from zero.
When the null hypothesis 𝐻0 𝑝𝑝s = 0 is true, 𝑟s has an
approximate normal distribution with mean zero and
variance 1 ∕ T.
Thus, a suitable test statistic is:

𝑟𝑠 − 0
𝑍= = √𝑇 𝑟 𝑠 𝑎´ 𝑁 ( 0 ,1 )
√ 1/𝑇

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 15
Edition Stationary Variables
 Autocorrelations:
A useful device for assessing the significance of
autocorrelations is a diagrammatic representation called
the correlogram.

The correlogram, also called the sample autocorrelation

function, is the sequence of autocorrelations 𝑟1,𝑟2,𝑟3….

A typical diagram for a correlogram will have bars or

spikes to represent the magnitudes of the autocorrelations
and approximate significant bounds drawn at ±2/√𝑇.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 16
Edition Stationary Variables
 9.2
Stationary and Weak Dependence

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 17
Edition Stationary Variables
 A critical assumption that is maintained throughout this
chapter is that the variables in our equations are
stationary.
Stationary variables have means and variances that do not
change over time and autocorrelations that depend only
on how far apart the observations are in time.
In addition to assuming that the variables are stationary,
in this chapter we also assume they are weakly
dependent.
Weak dependence implies that, as s → ∞ (observations
get further and further apart in time), they become almost
independent.

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 18
Edition Stationary Variables
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 19
Edition Stationary Variables
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 20
Edition Stationary Variables
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 21
Edition Stationary Variables
 9.3
Serial Correlation

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 22
Edition Stationary Variables
 9.3
Serial FIGURE 9.5 Scatter diagram for DUt and DUt-1
Correlation

9.3.1
Serial
Correlation in
DUt is the change 1.25
Scatter for change in U
Output Growth
in unemployment 1.00
rate in quarter t.
0.75

DU t is described 0.50

DU_t-1
as autocorrelated 0.25
or serially
0.00
correlated
-0.25

It is likely that -0.50

zero- -0.50 -0.25 0.00 0.25 0.50 0.75 1.00 1.25

autocorrelation DU_t

assumption is
violated
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 23
Edition Stationary Variables
 9.3
Serial FIGURE 9.5 Scatter diagram for Gt and Gt-1
Correlation

9.3.1
Serial
Correlation in
Output Growth

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 24
Edition Stationary Variables
 9.3
Serial
Correlation

9.3.1a
Computing To test if cov(DUt, DUs) = 0, we can use the correlation
Autocorrelation
between DU’s that are k periods apart, which is known as
the k-th order sample autocorrelation (ρk):


T
( DU t  DU )( DU t  k  DU )
Eq. 9.14 rk  t k 1

 t 1 t
T 2
( DU  DU )

The null hypothesis is H0: ρk = 0

Eq. 9.17 – A suitable test statistics is
rk  0
Z  T rk  N 0,1
1T
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 25
Edition Stationary Variables
 9.2
Finite
Distributed Lags

Consider the following distributed lag model

DU t   β 0Gt  β1Gt  1    β q Gt  q  et
– which says the change in the unemployment rate (DU)
depends on the contemporary and past GDP growth
rates (G’s).

Assumption: cov(et, es) = 0 for t ≠ s

– To test the validity of the assumption, can also calculate
the k-th order sample autocorrelation rk of eˆt where
eˆt DU t  (  b 0Gt  b1Gt  1    b q Gt  q )

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 26
Edition Stationary Variables
 9.3
Serial
Correlation

9.3.1a
Computing
Autocorrelation
For our problem, we have:

Z1  98 0.494 4.89, Z 2  98 0.414 4.10

Z 3  98 0.154 1.52, Z 4  98 0.200 1.98
– We reject the hypotheses H0: ρ1 = 0 and H0: ρ2
=0
– We have insufficient evidence to reject H0: ρ3
=0
– ρ4 is on the borderline of being significant.
– We conclude that G, the quarterly growth rate in
U.S. GDP, exhibits significant serial correlation at
lags oneChapter
and 9:two
Regression with Time Series Data:
Principles of Econometrics, 4t
h
Page 27
Edition Stationary Variables
 9.3
Serial
Correlation

9.3.1b
The Correlagram

The correlogram, also called the sample

autocorrelation function, is the sequence of
autocorrelations r1, r2, r3, …
– It shows the correlation between observations
that are one period apart, two periods apart,
three periods apart, and so on

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 28
Edition Stationary Variables
 9.3
Serial
Correlation

9.3.2a
A Phillips Curve The k-th order autocorrelation for the residuals can
be written as: T

 eˆ eˆ t t k
Eq. 9.21 rk  t k T1
t
ˆ
e 2

t 1
– The least squares equation is:

 0.7776  0.5279 DU
INF
Eq. 9.22
se  0.0658 0.2294 

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 29
Edition Stationary Variables
 9.3 FIGURE 9.8 Correlogram for residuals from least-squares estimated
Serial
Correlation Phillips curve

9.3.2a
A Phillips Curve
The correlogram, a
sequence of Correlogram of Residuals
autocorrelations r1, r2,
r3,…, can be used to
check whether the
assumption cov(et, es)
= 0 for t ≠ s is violated

The correlogram is
also called the sample
autocorrelation
function

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 30
Edition Stationary Variables
 9.4
Other Tests for Serially Correlated
Errors

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 31
Edition Stationary Variables
 9.4
Other Tests for
Serially A Lagrange Multiplier Test
Correlated
Errors

9.4.1
A Lagrange
Multiplier Test
Consider
yt β1  β 2 xt  et

If et and et-1 are correlated, then one way to model

the relationship between them is to write:
Eq. 9.23 et ρet  1  vt

– We can substitute this into a simple regression

equation:
Eq. 9.24 yt β1  β 2 xt  ρet  1  vt
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 32
Edition Stationary Variables
 9.4
Other Tests for
Serially A Lagrange Multiplier Test
Correlated
Errors

9.4.1
A Lagrange
Multiplier Test
For the autocorrelation LM test, we write:
yt β1  β 2 xt  ρeˆt  1  vt

– But since yt b1  b2 xt  eˆt , we get:

b1  b2 xt  eˆt β1  β 2 xt  ρeˆt  1  vt

Eq. 9.26 Rearranging, we get:
eˆt β1  b1   β 2  b2  xt  ρeˆt  1  vt
γ1  γ 2 xt  ρeˆt  1  v

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 33
Edition Stationary Variables
 9.4
Other Tests for
Serially A Lagrange Multiplier Test
Correlated
Errors

9.4.1
A Lagrange
Multiplier Test
If H0: ρ = 0 is true, then LM = T x R2 has an
approximate χ2(1) distribution
– T and R2 are the sample size and goodness-of-
fit statistic, respectively, from least squares
estimation of Eq. 9.26

Consider a model due to the Phillips Curve:

INFt β1  β 2 DU t  et

– Where INF is the inflation rate and DU is the

change in unemployment rate.
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 34
Edition Stationary Variables
 9.4
Other Tests for
Serially A Lagrange Multiplier Test
Correlated
Errors

9.4.1
A Lagrange Applying the autocorrelation LM test:
Multiplier Test

LM T R 2 90 0.3066 27.59

– The null hypothesis H0: ρ = 0 is rejected at all

conventional significance levels
– We conclude that the errors are serially correlated

For a four-period lag (i.e.

2
up to et-4), we obtain:
LM T R 90 0.4075 36.7

– Because the 5% critical value from a χ2(4)-distribution is

9.49, these LM values lead us to conclude that the errors
are serially
Principles of Econometrics, 4t
h
correlated
Chapter 9: Regression with Time Series Data:
Page 35
Edition Stationary Variables
 9.5
Estimation with Serially Correlated
Errors

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 36
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

We will encounter models with a lagged dependent

variable, such as:
yt δ  θ1 yt  1  δ 0 xt  δ1 xt  1  vt

– Assumption: vt is uncorrelated with yt-1, xt and xt-1

Three estimation procedures are considered:

1. Least squares estimation
2. An estimation procedure
et ρet  1that
 vtis relevant when the
errors follow

3. A general estimation strategy for estimating models

Principles of Econometrics, 4t
h
with serially correlated errors
Chapter 9: Regression with Time Series Data:
Page 37
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.1
Least Squares Consequences of using least squares estimation without
Estimation
recognizing the existence of serially correlated errors:
1. Still a linear unbiased estimator, but no longer the
best
2. The standard errors computed on the basis of
uncorrelated errors are no longer correct

The correct standard errors for the least squares estimator:

– HAC (Heteroskedasticity and Autocorrelation
Consistent) standard errors, or Newey-West
standard errors
• These are analogous to the heteroskedasticity
consistent standard errors
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 38
Edition Stationary Variables
 9.5
Estimation with
Serially Least Squares Estimation
Correlated
Errors

9.5.1
Least Squares
Estimation
Let’s reconsider the Phillips Curve model:
Eq. 9.29  0.7776  0.5279 DU
INF
0.0658  0.2294  incorrect se 
0.1030  0.3127  HAC se 

The t and p-values for testing H0: β2 = 0 are:

t  0.5279 0.2294  2.301 p 0.0238 from LS standard errors 
t  0.5279 0.3127  1.688 p 0.0950 from HAC standard errors 

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 39
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2
Estimating an
AR(1) Error
Model
Return to the Lagrange multiplier test for serially
correlated errors where we used the equation:
Eq. 9.30 et ρet  1  vt

– Assume the vt are uncorrelated random errors

with zero mean and constant variances:

Eq. 9.31 E vt  0 var vt   v2 cov vt , vs  0 for t s

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 40
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2
Estimating an
AR(1) Error
Eq. 9.30 describes a first-order autoregressive
Model
model or a first-order autoregressive process for
et
– The term AR(1) model is used as an
abbreviation for first-order autoregressive
model
– It is called an autoregressive model because it
can be viewed as a regression model
– It is called first-order because the right-hand-
side variable is et lagged one period

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 41
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2a
Properties of an
AR(1) Error
We assume that:

Eq. 9.32  1  ρ 1
The mean and variance of et are:
 2
Eq. 9.33 E et  0
var et    2
e
v
1  ρ2
The covariance term is:
ρ k 2
Eq. 9.34 cov et , et  k   , k 0 v
2
1 ρ

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 42
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2a
Properties of an The correlation implied by the covariance is:
ρ k corr et , et  k 
AR(1) Error

cov et , et  k 

var et  var et  k 
cov et , et  k 
Eq. 9.35 
var et 


ρ k v2 1  ρ 2 
 v2 1  ρ 2 
ρ k
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 43
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2a
Properties of an
AR(1) Error
Setting k = 1:
Eq. 9.36 ρ1 corr et , et  1  ρ

– ρ represents the correlation between two errors that

are one period apart
• It is the first-order autocorrelation for e,
sometimes simply called the autocorrelation
coefficient
• It is the population autocorrelation at lag one for
a time series that can be described by an AR(1)
model
• r1 is an estimate for ρ when we assume a series is
Principles of Econometrics, 4t AR(1)
h Chapter 9: Regression with Time Series Data:
Page 44
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2a
Properties of an
AR(1) Error
Each et depends on all past values of the errors vt:
2 3
Eq. 9.37 et vt  ρvt  1  ρ vt  2  ρ vt  3 

Obtained by substituting the value of et-1, et-2, … in

et ρet  1  vt

– For an AR(1) model, we have:

ρˆ 1 ρˆ r1 0.549

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 45
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.2a
Properties of an
AR(1) Error

For longer lags, we have:

ρˆ 2 ρˆ 0.549  0.301

2 2

ρˆ 3 ρˆ 3 0.549  0.165

3

ρˆ 4 ρˆ 0.549  0.091

4 4

ρˆ 5 ρˆ 5 0.549  0.050

5

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 46
Edition Stationary Variables
 9.5
Estimation with
Serially Nonlinear Least Squares
Correlated Estimation
Errors

9.5.2b
Nonlinear Least Write the Phillips curve model with an AR(1) error as:
Squares
Estimation
yt β1  β 2 xt  et with et ρet  1  vt ,  1  ρ  1
Eq. 9.38

– vt is uncorrelated and E(vt) = 0, var(vt) =  v

2

Using the relationship et = ρet-1 + vt

yt β1  β 2 xt  ρet  1  vt

Also, ρet  1 ρyt  1  ρβ1  ρβ 2 xt  1, substituting, we get:

Eq. 9.43 yt β1 1  ρ   β 2 xt  ρyt  1  ρβ 2 xt  1  vt

Above is an autoregressive distributed lag (ARDL) model

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 47
Edition Stationary Variables
 9.5
Estimation with
Serially Nonlinear Least Squares
Correlated Estimation
Errors

9.5.2b
Nonlinear Least Although Eq. 9.43 is a linear function of the variables xt ,
Squares
Estimation
yt-1 and xt-1, it is not a linear function of the parameters (β1,
β2, ρ)
– The coefficient of xt-1 equals -ρβ2
– These are nonlinear least squares estimates (NLS)
Eviews: INF = C(1)*(1-C(3)) + C(2)*DU +C(3)*INF(-1) -C(3)*C(2)*DU(-1)

INFt β1 1  ρ   β 2 DU t  ρINFt  1  ρβ 2 DU t  1  vt

Eq. 9.44
Our Phillips Curve model assuming AR(1) errors is:
 0.7609  0.6944 DU
INF et 0.557et  1  vt
se  0.1245 0.2479  0.090 
Eq. 9.45 Applying NLS and the original untransformed model is
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 48
Edition Stationary Variables
 9.5
Estimation with
Serially Estimating a More General Model
Correlated
Errors

9.5.3
Estimating a Eq. 9.43 is
yt β1 1  ρ   β 2 xt  ρyt  1  ρβ 2 xt  1  vt
More General
Model

Now consider the autoregressive distributed lag (ARDL)

model
Eq. 9.47 yt δ  θ1 yt  1  δ 0 xt  δ1 xt  1  vt

Note that
δ β1 1  ρ  δ0 β 2 δ1  ρβ2 θ1 ρ
Eq. 9.48

Eq. 9.43 is a restricted version of Eq. 9.47 with the

restriction δ1 = -θ1δ0 imposed

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 49
Edition Stationary Variables
 9.5
Estimation with
Serially Estimating a More General Model
Correlated
Errors

9.5.3
Estimating a Applying the least squares estimator to Eq. 9.47 using the
More General
Model data for the Phillips curve example yields:
 t 0.3336  0.5593INF  0.6882 DU  0.3200 DU
INF t 1 t t1
Eq. 9.49
se  0.0899  0.0908  0.2575  0.2499 

The equivalent AR(1) estimates are similar:

δˆ βˆ 1 1  ρˆ  0.7609 1  0.5574  0.3368
θˆ 1 ρˆ 0.5574
δˆ βˆ  0.6944
0 2

δˆ 1  ρβ
ˆ ˆ 2  0.5574  0.6944  0.3871

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 50
Edition Stationary Variables
 9.5
Estimation with
Serially
Correlated
Errors

9.5.4
Summary of
Section 9.5 and
We have described three ways of overcoming the
Looking Ahead
effect of serially correlated errors:
1. Estimate the model using least squares with
HAC standard errors
2. Use nonlinear least squares to estimate the
model with a lagged x, a lagged y, and the
restriction implied by an AR(1) error
specification
3. Use least squares to estimate the model with a
lagged x and a lagged y, but without the
restriction implied by an AR(1) error
specification
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 51
Edition Stationary Variables
 9.6
Autoregressive Distributed Lag
Models

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 52
Edition Stationary Variables
 9.6
Autoregressive
Distributed Lag
Models

An autoregressive distributed lag (ARDL) model

is one that contains both lagged xt’s and lagged yt’s

Eq. 9.52 yt   0 xt  1 xt  1     q xt  q  1 yt  1     p yt  p  vt

– Two examples:
 0.3336  0.5593INF  0.6882 DU  0.3200 DU
ADRL 1,1 : INFt t 1 t t1

 0.3548  0.5282 INF  0.4909 DU

ADRL 1, 0  : INFt t 1 t

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 53
Edition Stationary Variables
 9.6
Autoregressive
Distributed Lag
Models

Four possible criteria for choosing p and q:

1. Has serial correlation in the errors been
eliminated?
2. Are the signs and magnitudes of the estimates
consistent with our expectations from
economic theory?
3. Are the estimates significantly different from
zero, particularly those at the longest lags?
4. What values for p and q minimize information
criteria such as the AIC and SC?

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 54
Edition Stationary Variables
 9.6
Autoregressive
Distributed Lag
Models

9.6.1
The Phillips Consider the previously estimated ARDL(1,0) model:
Curve
 t 0.3548  0.5282 INF  0.4909 DU , obs 90
INF t 1 t
Eq. 9.56
se  0.0876  0.0851 0.1921
For an ARDL(4,0) version of the model:
 t 0.1001  0.2354 INF  0.1213INF  0.1677 INF
INF t 1 t 2 t 3

se  0.0983 0.1016  0.1038  0.1050 

Eq. 9.57
 0.2819 INFt -4  0.7902 DU t
0.1014  0.1885 obs 87

Table 9.4
AIC and SC
Values for
Phillips
Curve
ARDL
Models
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 55
Edition Stationary Variables
 9.6
Autoregressive
Distributed Lag
Models

9.6.1
The Phillips
Curve

FIGURE 9.9 Correlogram for Table 9.3 p-values for LM Test

residuals from Phillips curve for Autocorrelation
ARDL(1,0) model
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 56
Edition Stationary Variables
 9.7
Forecasting

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 57
Edition Stationary Variables
 9.7
Forecasting

We consider forecasting using three different

models:
1. AR model
2. ARDL model

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 58
Edition Stationary Variables
 9.7
Forecasting

9.7.1
Forecasting with
an AR Model

Consider an AR(2) model for real GDP growth:

Eq. 9.62 Gt δ  θ1Gt  1  θ 2Gt  2  vt

The model to forecast GT+1 is:

GT 1 δ  θ1GT  θ 2GT  1  vT 1

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 59
Edition Stationary Variables
 9.7
Forecasting

9.7.1
Forecasting with
an AR Model

The growth values for the two most recent

quarters are:
GT = G2009Q3 = 0.8
GT-1 = G2009Q2 = -0.2
The forecast for G2009Q4 is:
GˆT 1 δˆ  θˆ 1GT  θˆ 2GT  1
Eq. 9.63 0.46573  0.37700 0.8  0.24624  0.2 
0.7181

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 60
Edition Stationary Variables
 9.7
Forecasting

9.7.1
Forecasting with
an AR Model
For two quarters ahead, the forecast for G2010Q1 is:

GˆT 2 δˆ  θˆ 1GT 1  θˆ 2GT

Eq. 9.64 0.46573  0.37700 0.71808  0.24624 0.8
0.9334

For three periods out, it is:

GˆT 3 δˆ  θˆ 1GT 2  θˆ 2GT 1

Eq. 9.65 0.46573  0.37700 0.93343  0.24624 0.71808
0.9945

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 61
Edition Stationary Variables
 9.1
Introduction Stationary Time Series
9.1.2a
Stationarity

A stationary variable is one that is not

explosive, nor trending, and nor wandering
aimlessly without returning to its mean

FIGURE 9.2 (a) Time series of a stationary variable

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 62
Edition Stationary Variables
 9.1
Introduction
Non-stationary Time Series

FIGURE 9.2 (b)

time series of a
nonstationary
variable that is
‘‘slow-turning’’ or
‘‘wandering’’

FIGURE 9.2 (c)

time series of a
nonstationary
variable that
‘‘trends”

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 63
Edition Stationary Variables
 9.1
Introduction

Two features of time-series data to consider:

1. Have a natural ordering according to time
2. Observed over a number of time periods, and are
likely to be correlated

There is the possible existence of dynamic

relationships between variables
– A dynamic relationship is one in which the change in a
variable now has an impact on that same variable, or
other variables, in one or more future time periods

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 64
Edition Stationary Variables
 9.1
Introduction FIGURE 9.1 The distributed lag effect

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 65
Edition Stationary Variables
 9.1
Introduction

9.1.1
Dynamic Nature Ways to model the dynamic relationship:
of Relationships
1. Specify that a dependent variable y is a function of current
and past values of an explanatory variable x
Eq. 9.1 yt  f ( xt , xt  1 , xt  2 ,...)

2. Capture the dynamic characteristics of time-series by

specifying a model with a lagged dependent variable as one
of the explanatory variables
Eq. 9.3
yt  f ( yt  1 , xt , xt  1 ,...)

3. Model the continuing impact of change over several periods

Eq. 9.4 via the error term
yt  f ( yt  1, xt , xt  1 ,...)  et et  f (et  1 )

• et is correlated
Principles of Econometrics, 4th
with et - 1Series Data:
Chapter 9: Regression with Time
Page 66
Edition Stationary Variables
 9.1
Introduction

9.1.2
Least Squares
Assumptions
The primary assumption is Assumption MR4:
cov  yi , y j  cov ei , e j  0 for i  j

• For time series, this is written as:

cov  yt , ys  cov et , es  0 for t s

– The dynamic models in Eqs. 9.1, 9.3 and 9.4

imply correlation between yt and yt - 1 or et and et
- 1 or both, so they clearly violate assumption
MR4
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 67
Edition Stationary Variables
 9.2
Finite Distributed Lags

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 68
Edition Stationary Variables
 9.2
Finite
Distributed Lags

Consider a linear model in which, after q time periods,

changes in x no longer have an impact on y
Eq. 9.5 yt   0 xt  1 xt  1  2 xt  2     q xt  q  et
– Note the notation change: βs is used to denote the
coefficient of xt-s and α is introduced to denote the
intercept

Model 9.5 has two uses:

– Forecasting
yT 1   0 xT 1  1 xT  2 xT  1     q xT  q 1  eT 1
– Policy analysis
E ( yt ) E ( yt  s ) What is the effect of a
  s
xt  s xt change in x on y?
Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:
Page 69
Edition Stationary Variables
 9.2
Finite
Distributed Lags

Assume xt is increased by one unit and then

maintained at its new level in subsequent periods
– The immediate impact will be β0
– the total effect in period t + 1 will be β0 + β1, in
period t + 2 it will be β0 + β1 + β2, and so on
• These quantities are called interim
multipliers
– The total multiplier is the final effect on y of
the sustained increase
q after q or more periods
have elapsed s0
βs

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 70
Edition Stationary Variables
 9.2
Finite
Distributed Lags

The effect of a one-unit change in xt is distributed

over the current and next q periods, from which we
get the term ‘‘distributed lag model’’
– It is called a finite distributed lag model of
order q
• It is assumed that after a finite number of
periods q, changes in x no longer have an
impact on y
– The coefficient βs is called a distributed-lag
weight or an s-period delay multiplier
– The coefficient β0 (s = 0) is called the impact
multiplier
Principles of Econometrics, 4t
h Chapter 9: Regression with Time Series Data:
Page 71
Edition Stationary Variables
 9.2
Finite
Distributed Lags
ASSUMPTIONS OF THE DISTRIBUTED LAG MODEL

9.2.1
Assumptions

TSMR1. yt   β 0 xt  β1 xt  1  β 2 xt  2    β q xt  q  et , t q  1, , T
TSMR2. y and x are stationary random variables, and et is independent of
current, past and future values of x.
TSMR3. E(et) = 0
TSMR4. var(et) = σ2
TSMR5. cov(et, es) = 0 t ≠ s
TSMR6. et ~ N(0, σ2)

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 72
Edition Stationary Variables
 9.2
Finite
Distributed Lags

9.2.2
An Example: Okun’s Law implies that
Okun’s Law

DU t   β 0Gt  et

where DU = Ut - Ut-1 is change in unemployment

rate,
G is GDP growth rate

We can
DUexpand
 this
β Gtoinclude
β G lags:
βG    β qGt  q  et
t 0 t 1 t 1 2 t 2

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 73
Edition Stationary Variables
 9.2
Finite
Unemployment rate and GDP growth Time series for U.S:
Distributed Lags
1985Q2 to 2009Q3
9.2.2
An Example:
Okun’s Law
FIGURE 9.4 (a) Time series for FIGURE 9.4 (b) Time series for
the change in the U.S. U.S. GDP growth: 1985Q2 to
unemployment rate: 1985Q3 to 2009Q3
2009Q3

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 74
Edition Stationary Variables
 9.2
Finite
Distributed Lags
Table 9.2 Estimates for Okun’s Law Finite Distributed Lag Model

9.2.2
An Example:
Okun’s Law

Principles of Econometrics, 4th Chapter 9: Regression with Time Series Data:

Page 75
Edition Stationary Variables