358 Time Series Econometrics

say β1 Yt + β2 Xt which is integrated of order d − b. The vector {β1 , β2 }

is called the cointegrating vector.

A straightforward generalization of the above definition can be made for the case of n
variables, as follows:

Definition 2 If Zt denotes an n×1 vector of series Z1t , Z 2t , Z3t , . . . , Znt and (a) each
Zit is I(d); and (b) there exists an n × 1 vector β such that Zt β ∼
I(d − b), then Zt˜CI(d, b).

For empirical econometrics, the most interesting case is where the series transformed
with the use of the cointegrating vector become stationary; that is, when d = b, and the
cointegrating coefficients can be identified as parameters in the long-run relationship
between the variables. The next sections of this chapter will deal with these cases.

Cointegration and the error-correction

mechanism (ECM): a general approach
The problem
As noted earlier, when there are non-stationary variables in a regression model we may
get results that are spurious. So if Yt and Xt are both I(1), if we regress:

Yt = β1 + β2 Xt + ut (17.6)

we will not generally get satisfactory estimates of β̂1 and β̂2 .

One way of resolving this is to difference the data to ensure stationarity of our
variables. After doing this, Yt ∼ I(0) and Xt ∼ I(0), and the regression model
will be:

Yt = a1 + a2 Xt + ut (17.7)

In this case, the regression model may give us correct estimates of the â1 and â2 parame-
ters and the spurious equation problem has been resolved. However, what we have from
Equation (17.7) is only the short-run relationship between the two variables. Remember
that, in the long-run relationship:

Yt∗ = β1 + β2 Xt (17.8)

so Yt is bound to give us no information about the long-run behaviour of our model.
Knowing that economists are interested mainly in long-run relationships, this consti-
tutes a big problem, and the concept of cointegration and the ECM are very useful to
resolve this.
Cointegration and Error-Correction Models 359

Cointegration (again)
We noted earlier that Yt and Xt are both I(1). In the special case that there is a linear
combination of Yt and Xt (that is, I(0)), then Yt and Xt are cointegrated. Thus, if this
is the case, the regression of Equation (17.6) is no longer spurious, and it also provides
us with the linear combination:

ût = Yt − β̂1 − β̂2 Xt (17.9)

which connects Yt and Xt in the long run.

The error-correction model (ECM)

If, then, Yt and Xt are cointegrated, by definition ût ∼ I(0). Thus we can express the
relationship between Yt and Xt with an ECM specification as:

Yt = a0 + b1 Xt − π ût−1 + et (17.10)

which will now have the advantage of including both long-run and short-run infor-
mation. In this model, b1 is the impact multiplier (the short-run effect) that measures
the immediate impact a change in Xt will have on a change in Yt . On the other hand,
π is the feedback effect, or the adjustment effect, and shows how much of the dise-
quilibrium is being corrected – that is the extent to which any disequilibrium in the
previous period affects any adjustment in Yt . Of course ût−1 = Yt−1 − β̂1 − β̂2 Xt−1 , and
therefore from this equation β2 is also the long-run response (note that it is estimated
by Equation (17.7)).
Equation (17.10) now emphasizes the basic approach of the cointegration and
error-correction models. The spurious regression problem arises because we are using
non-stationary data, but in Equation (17.10) everything is stationary, the change in
X and Y is stationary because they are assumed to be I(1) variables, and the residual
from the levels regression (17.9) is also stationary, by the assumption of cointegration.
So Equation (17.10) fully conforms to our set of assumptions about the classic linear
regression model and OLS should perform well.

Advantages of the ECM

The ECM is important and popular for many reasons:

1 First, it is a convenient model measuring the correction from disequilibrium of the

previous period, which has a very good economic implication.
2 Second, if we have cointegration, ECMs are formulated in terms of first differences,
which typically eliminate trends from the variables involved, and they resolve the
problem of spurious regressions.
360 Time Series Econometrics

3 A third, very important, advantage of ECMs is the ease with which they can fit into
the general to specific approach to econometric modelling, which is in fact a search
for the most parsimonious ECM model that best fits the given data sets.
4 Finally, the fourth and most important feature of the ECM comes from the fact that
the disequilibrium error term is a stationary variable (by definition of cointegration).
Because of this, the ECM has important implications: the fact that the two variables
are cointegrated implies that there is some adjustment process preventing the errors
in the long-run relationship from becoming larger and larger.

Cointegration and the error-correction

mechanism: a more mathematical approach
A simple model for only one lagged term of X and Y
The concepts of cointegration and the error-correction mechanism (ECM) are very
closely related. To understand the ECM it is better to think of it first as a convenient
reparametrization of the general linear autoregressive distributed lag (ARDL) model.
Consider the very simple dynamic ARDL model describing the behaviour of Y in
terms of X, as follows:

Yt = a0 + a1 Yt−1 + γ0 Xt + γ1 Xt−1 + ut (17.11)

where the residual ut ∼ iid(0, σ 2 ).

In this model the parameter γ0 denotes the short-run reaction of Yt after a change in
Xt . The long-run effect is given when the model is in equilibrium, where:

Yt∗ = β0 + β1 Xt∗ (17.12)

and for simplicity assume that

Xt∗ = Xt = Xt−1 = · · · = Xt−p (17.13)

Thus, it is given by:

Yt∗ = a0 + a1 Yt∗ + γ0 Xt∗ + γ1 Xt∗ + ut

Yt∗ (1 − a1 ) = a0 + (γ0 + γ1 )Xt∗ + ut
a0 γ + γ1 ∗
Yt∗ = + 0 X + ut
1 − a1 1 − a1 t
Yt∗ = β0 + β1 Xt∗ + ut (17.14)

So the long-run elasticity between Y and X is captured by β1 = (γ0 + γ1 )/(1 − a1 ). Here,

we need to make the assumption that a1 < 1 so that the short-run model in Equation
(17.11) converges to a long-run solution.
Cointegration and Error-Correction Models 361

We can then derive the ECM, which is a reparametrization of the original Equation
(17.11) model:

Yt = γ0 Xt − (1 − a)[Yt−1 − β0 − β1 Xt−1 ] + ut (17.15)

Yt = γ0 Xt − π [Yt−1 − β0 − β1 Xt−1 ] + ut (17.16)

Proof that the ECM is a reparametrization

of the ARDL
To show that this is the same as the original model, substitute the long-run solutions
for β0 = a0 /(1 − a1 ) and β1 = (γ0 + γ1 )/(1 − a1 ) to give:


a0 γ + γ1
Yt = γ0 Xt − (1 − a) Yt −1 − − 0 Xt −1 + ut (17.17)
1 − a1 1 − a1
Yt = γ0 Xt − (1 − a)Yt −1 − a0 + (γ0 + γ1 )Xt −1 + ut (17.18)
Yt − Yt −1 = γ0 Xt − γ0 Xt −1 − Yt −1 + aYt −1 − a0 − γ0 Xt −1 − γ1 Xt −1 + ut
(17.19)

and by rearranging and cancelling out terms that are added and subtracted at the
same time we get:

Yt = a0 + a1 Yt −1 + γ0 Xt + γ1 Xt −1 + ut (17.20)

which is the same as for the original model.

What is of importance here is that when the two variables Y and X are cointegrated,
the ECM incorporates not only short-run but also long-run effects. This is because the
long-run equilibrium Yt−1 − β0 − β1 Xt−1 is included in the model together with the
short-run dynamics captured by the differenced term. Another important advantage is
that all the terms in the ECM model are stationary, and standard OLS is therefore valid.
This is because if Y and X are I(1), then Y and X are I(0), and by definition if Y and
X are cointegrated then their linear combination (Yt−1 − β0 − β1 Xt−1 ) ∼ I(0).
A final, very important, point is that the coefficient π = (1 − a1 ) provides us with
information about the speed of adjustment in cases of disequilibrium. To understand
this better, consider the long-run condition. When equilibrium holds, then (Yt−1 −
β0 − β1 Xt−1 ) = 0. However, during periods of disequilibrium, this term will no longer
be zero and measures the distance the system is away from equilibrium. For example,
suppose that because of a series of negative shocks in the economy (captured by the
error term ut ) Yt increases less rapidly than is consistent with Equation (17.14). This
causes (Yt−1 − β0 − β1 Xt−1 ) to be negative, because Yt−1 has moved below its long-run
steady-state growth path. However, since π = (1 − a1 ) is positive (and because of the
minus sign in front of π) the overall effect is to boost Yt back towards its long-run path
as determined by Xt in Equation (17.14). The speed of this adjustment to equilibrium
is dependent on the magnitude of (1 − a1 ). The magnitude of π will be discussed in the
next section.
362 Time Series Econometrics

A more general model for large numbers of lagged terms

Consider the following two-variable Yt and Xt ARDL:


n 
m
Yt = µ + ai Yt−i + γi Xt−i + ut (17.21)
i=1 i=0

Yt = µ + a1 Yt−1 + · · · + an Yt−n + γ0 Xt + γ1 Xt−1 + · · · + γm Xt−m + ut (17.22)

We want to obtain a long-run solution of the model, which would be defined as the
point where Yt and Xt settle down to constant steady-state levels Y ∗ and X ∗ , or more
simply when:

Y ∗ = β0 + β1 X ∗ (17.23)

and again assume X ∗ is constant

X ∗ = Xt = Xt−1 = · · · = Xt−m

So, putting this condition into Equation (17.21), we get the long-run solution, as:

µ γ
Y∗ =  + i X ∗
1− ai 1 − ai
µ (γ1 + γ2 + · · · + γm ) ∗
Y∗ = + X (17.24)
1 − a1 − a2 − · · · − an 1 − a1 − a2 − · · · − an

or:

Y ∗ = B0 + B1 X ∗ (17.25)

which means we can define Y ∗ conditional on a constant value of X at time t as:

Y ∗ = B0 + B1 Xt (17.26)

Here there is an obvious link to the discussion of cointegration in the previous section.
Defining et as the equilibrium error as in Equation (17.4), we get:

et ≡ Yt − Y ∗ = Yt − B0 − B1 Xt (17.27)

Therefore, what we need is to be able to estimate the parameters B0 and B1 . Clearly,

B0 and B1 can be derived by estimating Equation (17.21) by OLS and then calculating
  
A = µ/(1− ai ) and B = γi /(1− ai ). However, the results obtained by this method
are not transparent, and calculating the standard errors will be very difficult. However,
the ECM specification cuts through all these difficulties.
Take the following model, which (although it looks quite different) is a reparametriza-
tion of Equation (17.21):


n−1 
m−1
Yt = µ + ai Yt−i + γi Xt−i + θ1 Yt−1 + θ2 Xt−1 + ut (17.28)
i=1 i=0
Cointegration and Error-Correction Models 363

Note: for n = 1 the second term on the left-hand side of Equation (17.28) disappears.
From this equation we can see, with a bit of mathematics, that:


m
θ2 = γi (17.29)
i=1

which is the numerator of the long-run parameter, B1 , and that:

 

n
θ1 = − 1 − ai  (17.30)
i=1

So the long-run parameter B0 is given by B0 = 1/θ1 and the long-run parameter B1 =

−θ2 /θ1 . Therefore the level terms of Yt and Xt in the ECM tell us exclusively about
the long-run parameters. Given this, the most informative way to write the ECM is
as follows:


n−1 
m−1
1 θ
Yt = µ + ai Yt−i + γi Xt−i + θ1 Yt−1 − − 2 Xt−1 + ut (17.31)
θ1 θ1
i=1 i=0


n−1 
m−1
Yt = µ + ai Yt−i + γi Xt−i − θ1 (Yt−1 − β̂0 − β̂1 xt−1 ) + ut (17.32)
i=1 i=0

where θ1 = 0. Furthermore, knowing that Yt−1 − β̂0 − β̂1 xt−1 = et , our equilibrium
error, we can rewrite Equation (17.31) as:


n−1 
m−1
Yt = µ + ai Yt−i + γi Xt−i − π êt−1 + εt (17.33)
i=1 i=0

What is of major importance here is the interpretation of π . π is the error-correction

coefficient and is also called the adjustment coefficient. In fact, π tells us how much
of the adjustment to equilibrium takes place in each period, or how much of the
equilibrium error is corrected. Consider the following cases:

(a) If π = 1 then 100% of the adjustment takes place within a given period, or the
adjustment is instantaneous and full.
(b) If π = 0.5 then 50% of the adjustment takes place in each period.
(c) If π = 0 then there is no adjustment, and to claim that Yt∗ is the long-run part of
Yt no longer makes sense.

We need to connect this with the concept of cointegration. Because of co-

integration, êt ∼ I(0) and therefore also êt−1 ∼ I(0). Thus, in Equation (17.33), which
is the ECM representation, we have a regression that contains only I(0) variables and
allows us to use both long-run information and short-run disequilibrium dynamics,
which is the most important feature of the ECM.