university of copenhagen department of economics

Econometrics II
Introduction to Economic Time Series
Morten Nyboe Tabor
university of copenhagen department of economics

Learning Goals

1 Give an account for the important differences between (independent)

cross-sectional data and time series data.

2 Give a precise definition and interpretation of the concept of stationarity

of time series data, and explain the consequences of stationarity and weak
dependence of a stochastic process.

3 Evaluate and justify if a time series is stationary based on a graphical

analysis.

Econometrics II — Introduction to Economic Time Series — Slide 2

university of copenhagen department of economics

Outline

1 The difference between cross-sectional and time series data

2 Stationarity

3 Measuring time dependence

4 Characteristics of economic time series data

5 Non-stationarity and transformations to stationarity

Econometrics II — Introduction to Economic Time Series — Slide 3

1. The difference between cross-sectional
and time series data
university of copenhagen department of economics

Random Sampling (Cross-Sectional Data)

• The observations x1 , ..., xN are randomly drawn from a fixed population.

Random
N observations Sampling
from (Cross-Sectional
the same distribution. Data)
No ordering.
• The observations 1   are randomly drawn from a fixed population.
• When N is large we
 Observations fromcan characterize
the same the
distribution. Nodistribution.
ordering.P
N
Law of large numbers (for I.I.D. variables): N −1 i=1
xi → E (xi ).
• When  is large we can characterize the distribution.
X
Law of large numbers:  −1  → ().
=1

Random sampling from population Sample distribution approximates the population

y y

y2 y1 y3 y8 y7 y4 y9 y6 y5

5 of 17

Econometrics II — Introduction to Economic Time Series — Slide 5

university of copenhagen department of economics

Time Series Data

• A time series is a set of observations

y1 , y2 , ..., yt , ..., yT ,

where t is the time index.

Natural temporal ordering: 1 < 2 < ... < t < ... < T .

• It holds that yt−1 is realized (and often observed) when yt is determined.

We often focus on conditional models yt | yt−1 , yt−2 , ....
E.g. a model for the conditional mean, or for the entire conditional
distribution.

• Most data in macro-economics and finance come in this form.

• Very different characteristics.

Crucial: The tools should match the characteristics of the data!

Econometrics II — Introduction to Economic Time Series — Slide 6

 Stochastic Processes (Time Series Data)
university of copenhagen department of economics

Stochastic Processes (Time Series Data)

servation  is a realization of a random variable .
• Observation yt is a realization of a random variable yt .
ly one observation per random variable!
Only one observation per random variable!

• The sequence of random variables y1 , ..., yT is denoted a stochastic

of random variables 1   is denoted a stochastic process.
e sequenceprocess.

t
1 2 3 4 5 6 7

Econometrics II — Introduction to Economic Time Series — Slide 7

university of copenhagen department of economics

• If we could rerun history M times:

Stochastic process: y1 , y2 , ..., yt , ..., yT

(1) (1) (1) (1)

Realization 1: y1 , y2 , ..., yt , ..., yT
.. .. .. ..
. . . .
(m) (m) (m) (m)
Realization m: y1 , y2 , ..., yt , ..., yT
.. .. .. ..
. . . .
(M) (M) (M) (M)
Realization M: y1 , y2 , ..., yt , ..., yT

• Define the ensemble mean, E (yt ). Estimated with the cross-sectional

average
M
1 X (m)
E
b (yt ) = yt .
M
m=1

• Fundamentally different from the time average of a realized sample path

T
1 X
yT = yt .
T
t=1

Econometrics II — Introduction to Economic Time Series — Slide 8

2. Stationarity
university of copenhagen department of economics

Stationarity

Strict stationarity
A time series, y1 , y2 , ..., yt , ..., yT , is called strictly stationary if the
distributions of

(yt , yt+1 , ..., yt+s ) and (yt+h , yt+1+h , ..., yt+s+h )

are the same for all h. The distribution of yt does not depend on t.

Weak stationarity
The time series is called weakly stationary if

E (yt ) = µ
V (yt ) = E ((yt − µ)2 ) = γ0
Cov (yt , yt−h ) = E ((yt − µ) (yt−h − µ)) = γh for h = 1, 2, ...

Econometrics II — Introduction to Economic Time Series — Slide 10

university of copenhagen department of economics

Stationarity
y

t
1 2 3 4 5 6 7

Strict stationarity implies that yt (t = 1, 2, ..., T ) are random draws from the
same unconditional distribution.
Weak stationary is only a statement about the mean and the variance, i.e. the
first two moments of the distribution.
And yt fluctuates around a constnat level: equilibrium correcting / mean
reverting.
Econometrics II — Introduction to Economic Time Series — Slide 11
university of copenhagen department of economics

Simulation Illustration
We consider three simulated time series:

We plot the simulated series along with the sample average.

We increase the sample size by adding new observations and recursively update
the sample average. What do you see?
Econometrics II — Introduction to Economic Time Series — Slide 12
university of copenhagen department of economics

Simulation Illustration: What Do We See?

(A) IID observations: y 1t = t (B) Recursive sample average of y 1t

2.5
y 1t 0.5 Recursive sample average, ȳ 1(T )

0.0
0.0
-2.5

0 150 300 450 600 750 900 0 150 300 450 600 750 900
(C) Stationary AR(1): y 2t =0.85 y 2t 1  t (D) Recursive sample average of y 2t
5 y 2t
1
Recursive sample average, ȳ 2(T )

0
0

-5
0 150 300 450 600 750 900 0 150 300 450 600 750 900
(E) Non-stat. AR(1): y 3t = t (F) Recursive sample average of y 3t
15
y 3t Recursive sample average, ȳ 3(T )
30
10

5
10
0
0 150 300 450 600 750 900 0 150 300 450 600 750 900

Page: 1 of 1
Econometrics II — Introduction to Economic Time Series — Slide 13
university of copenhagen department of economics

Why do we care if a time series process

is stationary and weakly dependent?

Econometrics II — Introduction to Economic Time Series — Slide 14

university of copenhagen department of economics

Stationarity: Main Result

Strict stationarity implies that yt (t = 1, 2, ..., T ) contains information about
the same distribution. And, in terms of weak stationarity, yt fluctuates around
a constant level: equilibrium correcting.
To estimate parameters, we need a law of a large numbers (LLN) to hold.
• We make an additional technical assumption called weak dependence:
Observations yt and yt−h becomes approximately independent for h → ∞.
Then y is a consistent estimator of E (yt ).
• Given stationarity and weak dependence of yt and xt , most properties of
OLS in the IID case carry over to the time series regression

yt = xt0 β + t .

We return to this issue later.

• Compared to the I.I.D. case, weak stationarity replaces the condition that
observations are Identically Distributed , whereas weak dependence
replaces the Independence condition.

The most important distinction in time series econometrics is whether the

time series of interest are stationary or not!
→ Determines which methods we should use.
Econometrics II — Introduction to Economic Time Series — Slide 15
3. Measuring time dependence
university of copenhagen department of economics

Measuring Time Dependence

A characteristic feature of economic time series is a clear dependence over time.
We can measure the dependence by the correlation
Cov (yt , yt−h )
Corr (yt , yt−h ) = p , h = ..., −2, −1, 0, 1, 2, ...
V (yt ) · V (yt−h )

As a function of h this is called a correlogram.

Under stationarity the formula simplifies to
Cov (yt , yt−h )
ρh = , h = ..., −2, −1, 0, 1, 2, ...
V (yt )

called the autocorrelation function (ACF).

You may think of weak dependence as ρh → 0 as h → ∞.
Both can be estimated by replacing population moments with sample moments:
T
1 X
C
d ov (yt , yt−h ) = (yt − y ) (yt−h − y ) .
T −h
t=h+1

Under stationarity the associated estimators are asymptotically equivalent.

Econometrics II — Introduction to Economic Time Series — Slide 17
university of copenhagen department of economics

Empirical Example
Look at the US unemployment rate.Empirical Example
Look at
• Ittheis US
notunemployment
clear whether urate.
t equilibrium corrects.
• It is not clear whether  equilibrium corrects.
• It fluctuates within bounds, but deviations are very persistent and
• It fluctuates within bounds, but deviations are very persistent and equilibrium cor-
equilibrium correction is very slow. ut and ut−h are highly correlated for
rection is very slow.  and − are highly correlated for large values of .
large values of h.
• We return to formal testing later.
• We return to formal testing later.

(A) US unemployment rate (B) ACF for (A)

1.0
10.0

7.5
0.5

5.0

1950 1960 1970 1980 1990 2000 0 5 10 15

Econometrics II — Introduction to Economic Time Series —13 of 17

Slide 18
4. Characteristics of economic time series
data
university of copenhagen department of economics

What areCharacteristics
the characteristics ofof
economic
Time time seriesData
Series data?

(A) US unemployment rate (B) Danish productivity (logs)

12.5 1.00

10.0
0.75
7.5

5.0 0.50

2.5
0.25
0.0
1950 1960 1970 1980 1990 2000 1970 1980 1990 2000

(C) Danish income and consumption (logs) (D) Daily change in the NASDAQ index (%)
6.5
Income
6.4 Consumption 10
6.3 5
6.2
0
6.1
6.0 -5

5.9 -10 Period: 3/1-2000 to 26/2-2004

1970 1980 1990 2000 0 150 300 450 600 750 900 1050

4 of 17

Econometrics II — Introduction to Economic Time Series — Slide 20

5. Non-stationarity and transformations to
stationarity
university of copenhagen department of economics

Transformations to Stationarity

Many economic time series are not stationary.

But sometimes a non-stationary time series can be transformed to stationarity.

Three important cases:

(A) Remove a deterministic trend.
Trend stationary.

(B) Take first differences.

Difference stationary or integrated or first order.

(C) Combine several variables.

Cointegration.

Econometrics II — Introduction to Economic Time Series — Slide 22

university of copenhagen department of economics
(A) If the non-stationarity of  is due to a deterministic trend, then the de-trended
variable
(A) If the non-stationarity of yt is∗ due to a deterministic trend, then the
de-trended variable  =  − 0 − 1
yt∗=isytcalled
might be stationary. In this case, − µ0 trend-stationary.
− µ1 t,

Themight be stationary.
de-trended In be
variable can thisfound
case,asytthe
is estimated
called trend-stationary.
residual in the linear regression
The de-trended variable can be found as the estimated residual in the
∗
linear regression  = 0 + 1 +  
As an example look at the Danish yt = µ0 + µ1 t + yt∗ .
productivity.
As an example look at the Danish productivity:
(C) Danish productivity (log) minus trend (D) ACF for (C)
1.0

0.05

0.5
0.00

-0.05
1970 1980 1990 2000 0 5 10 15

15 of 17

Econometrics II — Introduction to Economic Time Series — Slide 23

university of copenhagen department of economics

(B) Alternatively it might turn out that yt is non-stationary while the first
difference,it might turn out that  is non-stationary while the first difference,
(B) Alternatively 
∆yt = yt − yt−1 ,
is stationary. In this case, yt ∆  =  − stationary
is difference −1 or integrated of first
order, I(1).
is stationary. In this case,  is difference stationary or integrated of first order, I(1).

As an example look at Danish private consumption.

As an example look at Danish private consumption:

(E) change in Danish consumption (log) (F) ACF for (E)

1

0.05

0.00 0

-0.05

1970 1980 1990 2000 0 5 10 15

Econometrics II — Introduction to Economic Time Series — Slide 24

university of copenhagen department of economics

(C) Finally, it might turn out that two variables, yt and xt , are non-stationary,
(C) Finally,
but it mightso
related turn outathat
that two
linear variables,  and , are non-stationary, but related
combination
so that a linear combination
zt = yt − β · xt ,
 =  −  · 
is stationary. Here yt and xt are I(1) but so-called co-integrated.
is stationary. Here  and  are I(1) but so-called co-integrated.
As an example consider consumption, ct , and income, yt . Both are I(1)
As an
andexample
have noconsider
equilibrium. They arerelated,
consumption, , and income, . Both
however, are savings
and the I(1) andrate,
have no
equilibrium. They are related, however, and the savings rate,
st = yt − ct ,
 =  − 
seems to be stationary and equilibrium corrects.
seems to be stationary and equilibrium corrects.
(G) Danish savings rate (log) (H) ACF for (G)
1

0.15

0.10
0

0.05

0.00
1970 1980 1990 2000 0 5 10 15
17 of 17

Econometrics II — Introduction to Economic Time Series — Slide 25

university of copenhagen department of economics

Learning Goals, Again

1 Give an account of the important differences between (independent)

cross-sectional data and time series data.

2 Give a precise definition and interpretation of the concept of stationarity

and weak dependence, and explain the important consequences of
stationarity and weak dependence of a time series process.

3 Evaluate and justify if a time series is stationary based on a graphical

analysis.

Next: The linear regression model for stationary time series.

Econometrics II — Introduction to Economic Time Series — Slide 26