6/11/2023
Cointegration and Error Correction Models
Macroeconometrics
Prof. Freddy Espino
Macroeconometrics Cointegration and Error Correction Models 1
Readings
• Enders, Ch. 6
• Engle and Granger (1987) “Co-Integration and Error Correction:
Representation, Estimation, and Testing”, Econometrica, Vol. 55, No. 2,
pp. 251 - 276.
• Johansen, Søren (1988) “The Statistical Analysis of Cointegration
Vectors”, Journal of Economic Dynamics and Control, 12(2-3):231-254.
Macroeconometrics Cointegration and Error Correction Models 2
1
2
6/11/2023
1. Introduction
Macroeconometrics Cointegration and Error Correction Models 3
1. Introduction
• Cointegration is the existence of long-run relationship between two or
more variables.
• In econometrics, cointegration analysis is used to estimate and test
stationary linear relations, or cointegration relations, between non-
stationary time series variables.
• Typically refers to a linear combination of nonstationary variables.
Macroeconometrics Cointegration and Error Correction Models 4
2
4
6/11/2023
1. Introduction
• Theoretically, it is quite possible that nonlinear long-run relationships
exist among a set of integrated variables.
• However, the current state of econometric practice is just beginning to
allow for tests of nonlinear cointegrating relationships.
Macroeconometrics Cointegration and Error Correction Models 5
1. Introduction
• An Error Correction Model (ECM) is most used for data where the
underlying variables have a long-run stochastic trend, also known as
cointegration.
• The term error-correction relates to the fact that last-period's deviation from a
long-run equilibrium, the error, influences its short-run dynamics.
• Thus, ECMs directly estimate the speed at which a dependent variable
returns to equilibrium after a change in other variables.
Macroeconometrics Cointegration and Error Correction Models 6
3
6
6/11/2023
2. Integrated Stochastic Process
Macroeconometrics Cointegration and Error Correction Models 7
2.1 Integrated Stochastic Process
• The random walk model is but a specific case of a more general class of
stochastic processes known as integrated processes.
• In general, if a (nonstationary) time series must be differenced d times
to make it stationary, that time series is said to be integrated of order
d.
• A time series 𝑌 integrated of order d is denoted as 𝑌 ~𝐼(𝑑).
Macroeconometrics Cointegration and Error Correction Models 8
4
8
6/11/2023
2.1 Integrated Stochastic Process
• If a time series 𝑌 to begin with (i.e., it does not require any
differencing), it is said to be integrated of order zero, denoted by
𝑌 ~𝐼(0).
• Most economic time series are generally 𝐼(1).
Macroeconometrics Cointegration and Error Correction Models 9
2.1 Integrated Stochastic Process
1. If 𝑋 ~𝐼(0) and 𝑌 ~𝐼(1) , then 𝑋 + 𝑌 = 𝑍 ~𝐼(1) ; that is, a linear
combination or sum of stationary and nonstationary time series is
nonstationary.
2. If 𝑋 ~𝐼(𝑑), then 𝑎 + 𝑏𝑋 = 𝑍 ~𝐼(𝑑), where a and b are constants.
3. If 𝑋 ~𝐼(𝑑 ) and 𝑌 ~𝐼(𝑑 ), then 𝑎𝑋 + 𝑏𝑌 = 𝑍 ~𝐼(𝑑 ), where 𝑑 < 𝑑 .
Macroeconometrics Cointegration and Error Correction Models 10
5
10
6/11/2023
2.1 Integrated Stochastic Process
4. If 𝑋 ~𝐼(𝑑) and 𝑌 ~𝐼(𝑑), then 𝑎𝑋 + 𝑏𝑌 = 𝑍 ~𝐼(𝑑 ∗ ); 𝑑 ∗ is generally
equal to d, but in some cases 𝑑 ∗ < 𝑑 (cointegration).
• As you can see from the preceding statements, one must pay careful
attention in combining two or more time series that are integrated of
different order.
Macroeconometrics Cointegration and Error Correction Models 11
11
2.2 Spurious Regression
• Let’s consider the following two random walk models 𝑦 and 𝑥 .
• Suppose we regress 𝑦 on 𝑥 .
• Since 𝑦 and 𝑥 are uncorrelated I(1) processes, the R2 from the
regression of 𝑦 on 𝑥 should tend to zero.
Macroeconometrics Cointegration and Error Correction Models 12
6
12
6/11/2023
2.2 Spurious Regression
• However, it usually does not happen → You may be tempted to conclude
that there is a significant statistical relationship between 𝑦 and 𝑥 ,
whereas a priori there should be none.
• This is in a nutshell the phenomenon of spurious or nonsense
regression.
Macroeconometrics Cointegration and Error Correction Models 13
13
2.2 Spurious Regression
• Yule (1926) showed that (spurious) correlation could persist in
nonstationary time series even if the sample is very large.
• According to Granger and Newbold (1974), an 𝑅 > 𝑑 (Durbin Watson
statistic) is a good rule of thumb to suspect that the estimated
regression is spurious.
Macroeconometrics Cointegration and Error Correction Models 14
7
14
6/11/2023
2.2 Spurious Regression
• One should be extremely wary of conducting regression analysis based
on time series that exhibit stochastic trends.
• And one should therefore be extremely cautious in reading too much in
the regression results based on I(1) variables.
Macroeconometrics Cointegration and Error Correction Models 15
15
3. Cointegration
Macroeconometrics Cointegration and Error Correction Models 16
8
16
6/11/2023
3.1 Cointegration
• We have warned that the regression of a nonstationary time series on
another nonstationary time series may produce a spurious
regression.
• Subjecting these time series individually to unit root analysis, you will
find that they both are I(1); that is, they contain a unit root.
Macroeconometrics Cointegration and Error Correction Models 17
17
3.1 Cointegration
• Suppose, then, that we regress 𝑦 ~𝐼(1) on 𝑥 ~𝐼(1) as follows:
• 𝑦 = 𝛽 +𝛽 𝑥 +𝑢
• 𝑢 =𝑦 −𝛽 𝑥 −𝛽
• We now subject 𝑢 to unit root analysis and find that it is stationary; that is,
it is I(0).
• Although 𝑦 and 𝑥 are individually I(1), their linear combination is I(0).
Macroeconometrics Cointegration and Error Correction Models 18
9
18
6/11/2023
3.1 Cointegration
• As a result, a regression of 𝑦 on 𝑥 as in would be meaningful (i.e., not
spurious).
• In this case we say that the two variables are cointegrated, this is,
they have a long-term, or equilibrium, relationship between them.
• The valuable contribution of the concepts of unit root, cointegration,
etc. is to force us to find out if the regression residuals are stationary.
Macroeconometrics Cointegration and Error Correction Models 19
19
3.1 Cointegration
• As Granger notes, “A test for cointegration can be thought of as a pre-
test to avoid ‘spurious regression’ situations.”
• In the language of cointegration theory, that regression is known as a
cointegrating regression and the slope parameter 𝛽 is known as the
cointegrating parameter.
Macroeconometrics Cointegration and Error Correction Models 20
10
20
6/11/2023
3.1 Cointegration
• Notice that 𝑢 = 𝑦 − 𝛽 𝑥 − 𝛽 is equal to:
𝑦
• 𝑢 = 1−𝛽 𝑥 −𝛽
Macroeconometrics Cointegration and Error Correction Models 21
21
3.2 The Engle-Granger Method
• Engle and Granger (1987) developed this crucial technique.
• According to them, the steps for determining whether two integrated
variables cointegrate of the same order are the following:
1. Determine its order of integration
2. Test whether 𝑢 is stationary or not.
Macroeconometrics Cointegration and Error Correction Models 22
11
22
6/11/2023
3.2 The Engle-Granger Method
• If the variables cointegrate, an OLS regression equation yields a
super-consistent" estimator.
• This means that there is a strong linear relationship between the
variables under study.
• The residual sequence, denoted by 𝑢 is a series of estimated values of
the deviation from the long-run relationship.
Macroeconometrics Cointegration and Error Correction Models 23
23
3.2 The Engle-Granger Method
• Testing for unit roots on residuals aims at determining whether these
deviations are stationary or not.
• If they are stationary, then the series cointegrate.
• If the residuals are not stationary, there is no cointegration.
• For example, the ADF test is performed on the following model:
• ∆𝑢 = 𝛿𝑢 +𝜀
Macroeconometrics Cointegration and Error Correction Models 24
12
24
6/11/2023
4. Error Correction Mechanism
Macroeconometrics Cointegration and Error Correction Models 25
25
4.1 Error Correction Mechanism
• We just showed that 𝑦 and 𝑥 are cointegrated; that is, there is a long-term,
or equilibrium, relationship between the two.
• Of course, in the short run there may be disequilibrium.
• Therefore, one can treat the error term as the “equilibrium error.”
• The error correction mechanism (ECM) first used by Sargan (1984) and later
popularized by Engle and Granger (1987) corrects for disequilibrium.
Macroeconometrics Cointegration and Error Correction Models 26
13
26
6/11/2023
4.1 Error Correction Mechanism
• An important theorem, known as the Granger representation theorem:
• If two variables 𝑦 and 𝑥 are cointegrated, then the relationship
between the two can be expressed as ECM.
Macroeconometrics Cointegration and Error Correction Models 27
27
4.1 Error Correction Mechanism
• Consider the following model
• 𝑦 =𝛽 +𝛽 𝑥 +𝑢
• Δ𝑦 = 𝛼 + 𝛼 Δ𝑥 + 𝛼 𝑢 + 𝜀 ……………… (ECM)
𝑦
• Δ𝑦 = 𝛼 + 𝛼 Δ𝑥 + 𝛼 1−𝛽 𝑥 −𝛽 +𝜀
Macroeconometrics Cointegration and Error Correction Models 28
14
28
6/11/2023
4.1 Error Correction Mechanism
• ECM equation states that Δ𝑦 depends on Δ𝑥 and on the equilibrium
error term 𝑢 .
• The absolute value of 𝛼 decides how quickly the equilibrium is
restored.
• The significance of 𝛼 will tell us whether if the equilibrium error term
is zero or not.
Macroeconometrics Cointegration and Error Correction Models 29
29
4.2 Advantages of the Engle and Granger approach
• Relatively simple.
• Useful as a first indication of the existence of a long run equilibrium
relationship.
• Where there is a consistent Cointegration vector it allows us to use the
super consistency property of OLS to obtain consistent estimates of the
cointegrating vector.
Macroeconometrics Cointegration and Error Correction Models 30
15
30
6/11/2023
4.2 Advantages of the Engle and Granger approach
• Provides long run equilibrium information and the short-term
dynamics.
• Provides speed of adjustment to equilibrium.
Macroeconometrics Cointegration and Error Correction Models 31
31
4.3 Limitations of the Engle and Granger approach
• For more than two variables it is no longer possible to demonstrate the
uniqueness of the Cointegration vector:
• If we have a vector of 𝐾 variables each integrated of the same order,
we can have up to 𝐾 − 1 Cointegration vectors.
• Has no systematic procedure to estimate multiple Cointegration
vectors.
Macroeconometrics Cointegration and Error Correction Models 32
16
32
6/11/2023
4.3 Limitations of the Engle and Granger approach
• Distribution of test statistics is only a rough guide and will be slightly
different in any application.
• Results are based on asymptotic theory, but we do not have infinitely
large samples in practice.
• Carry over error bias.
Macroeconometrics Cointegration and Error Correction Models 33
33
5. Vector Error Correction Model (VECM)
Macroeconometrics Cointegration and Error Correction Models 34
17
34
6/11/2023
5.1 Vector Error Correction Model (VECM)
• The issue of whether the variables in a VAR need to be stationary
exists:
• Sims (1980) and others recommend against differencing even if the
variable contain a unit root.
• They argue that the goal of VAR analysis is to determine the
interrelationships among the variables, not the parameter
estimates.
Macroeconometrics Cointegration and Error Correction Models 35
35
5.1 Vector Error Correction Model (VECM)
• The main argument against differencing is that it “throws away”
information concerning the comovements in the data (such as the
possibility of cointegrating relationships).
• When there are unit roots in the model, it is convenient to
reformulate the VAR into a Vector Error Correction Model (VECM).
Macroeconometrics Cointegration and Error Correction Models 36
18
36
6/11/2023
5.1 Vector Error Correction Model (VECM)
• If two series are both integrated (of order one, or I(1)) we could model
their interrelationship by taking first differences of each series and
including the differences in a VAR or a structural model.
• However, this approach would be suboptimal if it was determined that
these series are indeed cointegrated.
Macroeconometrics Cointegration and Error Correction Models 37
37
5.1 Vector Error Correction Model (VECM)
• This implies that the simple regression in first differences is
misspecified.
• If the series are cointegrated, they move together in the long run.
• A VAR in first differences, although properly specified in terms of
covariance-stationary series, will not capture those long-run
tendencies.
Macroeconometrics Cointegration and Error Correction Models 38
19
38
6/11/2023
5.1 Vector Error Correction Model (VECM)
• Accordingly, the VAR concept may be extended to the vector error-
correction model, or VECM, where there is evidence of cointegration
among two or more series.
• The model is fit to the first differences of the nonstationary variables,
but a lagged error-correction term is added to the relationship.
Macroeconometrics Cointegration and Error Correction Models 39
39
5.1 Vector Error Correction Model (VECM)
• In the case of two variables, this term is the lagged residual from
the cointegrating regression, of one of the series on the other in levels.
• In the case of multiple variables, there is a vector of error-
correction terms, of length equal to the number of cointegrating
relationships, or cointegrating vectors, among the series.
Macroeconometrics Cointegration and Error Correction Models 40
20
40
6/11/2023
5.1 Vector Error Correction Model (VECM)
• Consider a first difference stationary, I(1), vector time series 𝑦 .
• The elements of 𝑦 are cointegrated if there is at least one vector 𝛽 such
that 𝛽′𝑦 is stationary in levels.
• 𝜷 is known as the cointegrating vector.
• In practice, most empirical applications analyze multivariate systems,
so the rest of our discussion focuses on that case.
Macroeconometrics Cointegration and Error Correction Models 41
41
5.1 Vector Error Correction Model (VECM)
• Consider a VAR with p lags:
𝑦 =𝑣+𝐴 𝑦 +⋯+ 𝐴 𝑦 +𝜀
• Can be rewritten as:
∆𝑦 = 𝑣 + Π𝑦 + Γ Δ𝑦 +𝜀
Macroeconometrics Cointegration and Error Correction Models 42
21
42
6/11/2023
5.1 Vector Error Correction Model (VECM)
• Where:
Π= A −𝐼
Γ =− A
Macroeconometrics Cointegration and Error Correction Models 43
43
5.1 Vector Error Correction Model (VECM)
Engle and Granger (1987) show that if the variables 𝑦 are I(1) the
matrix Π has rank 0 ≤ r < K, where r is the number of linearly
independent cointegrating vectors
Macroeconometrics Cointegration and Error Correction Models 44
22
44
6/11/2023
5.1 Vector Error Correction Model (VECM)
1. If the variables in 𝑦 are I(1) but not cointegrated, Π is a matrix of zeros and
thus Rank(Π) = 0, then there is no co-integrating vector, implying that the
system is not co-integrated, thus, a VAR in their first differences is
consistent.
2. If all the variables are I(0), Rank(Π) = K; full rank, thus, a VAR in their
levels is consistent.
3. If the variables cointegrate, 0 < Rank(Π) = r < K, thus, a VAR in first
differences is misspecified because it omits the lagged level term Π𝑦 .
Macroeconometrics Cointegration and Error Correction Models 45
45
5.1 Vector Error Correction Model (VECM)
• If we express Π = 𝛼𝛽′, the linear combination given by 𝛽′𝑦 are stationary.
• In the terminology of Engle-Granger, this means that the vector process 𝑦 is
cointegrated with cointegration vector 𝜷 and 𝜶 is the adjusting factor:
∆𝑦 = 𝑣 + 𝛼𝛽′𝑦 + Γ Δ𝑦 +𝜀
Macroeconometrics Cointegration and Error Correction Models 46
23
46
6/11/2023
5.1 Vector Error Correction Model (VECM)
• Bivariate case:
𝑦
Δ𝑦 = 𝛼 + 𝛼 1−𝛽 𝑥 −𝛽 + 𝛼 Δ𝑥 + +𝜀
Π = 𝛼𝛽′
Macroeconometrics Cointegration and Error Correction Models 47
47
5.1 Vector Error Correction Model (VECM)
• The general representation is:
∆𝑦 = 𝛼 𝛽 𝑦 + 𝜇 + 𝜌𝑡 + Γ Δ𝑦 + 𝛾 + 𝜏𝑡 + 𝜀
Macroeconometrics Cointegration and Error Correction Models 48
24
48
6/11/2023
6. The Johansen Test for Cointegration
Macroeconometrics Cointegration and Error Correction Models 49
49
6.1 The Johansen Test for Cointegration
The Johansen (1988) test is a test for cointegration that allows for more
than one cointegrating relationship, unlike the Engle-Granger method.
As we mentioned, 𝒓 is the number of cointegrating relationships, the
elements of 𝜶 are known as the adjustment parameters in the vector
error correction model and each column of 𝜷 is a cointegrating vector.
Macroeconometrics Cointegration and Error Correction Models 50
25
50
6/11/2023
6.1 The Johansen Test for Cointegration
Let 𝜆 , … , 𝜆 be the K eigenvalues used in computing the log likelihood
at the optimum.
Furthermore, assume that these eigenvalues are sorted from the
largest 𝜆 to the smallest 𝜆 .
Macroeconometrics Cointegration and Error Correction Models 51
51
6.1 The Johansen Test for Cointegration
If there are 𝑟 < 𝐾 cointegrating equations, 𝛼 and 𝛽 have rank 𝑟 and the
eigenvalues 𝜆 , … , 𝜆 are zero.
The number of non-zero eigenvalues of a matrix Π is at most Rank(Π)
Macroeconometrics Cointegration and Error Correction Models 52
26
52
6/11/2023
6.1 The Johansen Test for Cointegration
1. The Trace Test:
𝐽 = 𝐿𝑅 𝑟, 𝑛 = −𝑇 ln(1 − 𝜆 )
• H0: cointegrating vectors ≤ 𝒓.
• H1: cointegrating vectors > 𝒓.
Macroeconometrics Cointegration and Error Correction Models 53
53
6.1 The Johansen Test for Cointegration
2. Maximum Eigenvalue Test:
𝐽 = 𝐿𝑅 𝑟, 𝑟 + 1 = −𝑇ln(1 − 𝜆 )
• H0: cointegrating vectors = 𝒓.
• H1: cointegrating vectors = 𝒓 + 1.
Macroeconometrics Cointegration and Error Correction Models 54
27
54
6/11/2023
6.1 The Johansen Test for Cointegration
• At the core of the Johansen method is the relationship between the
rank of the impact matrix 𝚷 = 𝜶𝜷′ and the size of its eigenvalues.
• The method infers the cointegration rank by testing the number of
eigenvalues that are statistically different from 0, then conducts
model estimation under the rank constraints.
Macroeconometrics Cointegration and Error Correction Models 55
55
6.1 The Johansen Test for Cointegration
• Sometimes the two tests may give conflicting results
• Harris(1995)
• The sequence of the trace tests leads to a consistent procedure.
• The maximum eigenvalue has a sharper alternative
hypothesis and is preferred to pin down the number of
cointegrating vectors.
Macroeconometrics Cointegration and Error Correction Models 56
28
56
6/11/2023
6.1 The Johansen Test for Cointegration
• Cheung and Lai (1993) propose choosing cointegration rank based on
the trace statistic.
• They state that the trace statistic is more robust to skewness and
excess kurtosis in residuals than the maximum Eigenvalue statistic.
Macroeconometrics Cointegration and Error Correction Models 57
57
6.1 The Johansen Test for Cointegration
• Enders (2010) concurs with these findings and states that when the
two tests for cointegration rank are in conflict the trace statistic is
likely to give more reliable results.
• Current practice is to only consider the trace test.
Macroeconometrics Cointegration and Error Correction Models 58
29
58
6/11/2023
7. Deterministic Trend Specification
Macroeconometrics Cointegration and Error Correction Models 59
59
7.1 The Johansen Test for Cointegration
• Johansen (1988) develop max likelihood procedure to test for
Cointegration.
• Could estimate and test the number of cointegration equations and to
test restricted versions of the cointegrating vectors and speeds of
adjustment.
• Allows verification of theories through coefficient restrictions etc.
• The test based on the stationary VAR.
Macroeconometrics Cointegration and Error Correction Models 60
30
60
6/11/2023
7.1 The Johansen Test for Cointegration
• Although the method appears to be very different from the Engle-
Granger method, it is essentially a multivariate generalization of the
augmented Dickey-Fuller test for unit roots.
• The eigenvalues depend on the form of the VEC model, and on the
composition of its deterministic terms.
Macroeconometrics Cointegration and Error Correction Models 61
61
7.1 The Johansen Test for Cointegration
• The critical values for the tests are obtained using Monte Carlo
approach.
• The distribution of statistics depends on two components:
1. The number of nonstationary components under the null
hypothesis.
2. The form of the deterministic components, constant, trend or both.
Has similarity with the Dickey fuller test.
Macroeconometrics Cointegration and Error Correction Models 62
31
62
6/11/2023
7.2 Deterministic Trend Specification
• Deterministic trends in a cointegrating VEC can stem from two distinct
sources; the mean of the cointegrating relationship and the mean of the
differenced series:
∆𝑦 = 𝛼 𝛽 𝑦 + 𝜇 + 𝜌𝑡 + Γ Δ𝑦 + 𝛾 + 𝜏𝑡 + 𝜀
Macroeconometrics Cointegration and Error Correction Models 63
63
7.2 Deterministic Trend Specification
𝒑 𝟏
∆𝒚𝒕 = 𝜶 𝜷 𝒚𝒕 + 𝝁 + 𝝆𝒕 + 𝚪𝒊 𝚫𝒚𝒕 𝒊 + 𝜸 + 𝝉𝒕 + 𝜺𝒕
𝒊 𝟏
1. No intercept or trend in CE or test VAR τ=ρ=γ=μ=0
2. Intercept (no trend) in CE – no intercept and no trend in VAR τ=ρ=γ=0
3. Intercept (no trend) in CE and test VAR τ=ρ=0
4. Intercept and trend in CE – intercept and no trend in VAR τ=0
5. Intercept and trend in CE – intercept and trend in VAR none
Macroeconometrics Cointegration and Error Correction Models 64
32
64
6/11/2023
8. Example
Macroeconometrics Cointegration and Error Correction Models 65
65
8. Example
• 𝑐 : Personal Consumption Expenditure
• 𝑦 : Disposable Personal Income
• 𝑐 ~𝐼 1 ; 𝑦 ~𝐼 1
• Consumption-Income Relationship: 𝑐 = 𝛽 + 𝛽 𝑦 + 𝜖
• Cointegrated 𝑐 and 𝑦 if 𝜖 ~𝐼(0)
• Example: Johansen Test with constant model and 10 lags
Macroeconometrics Cointegration and Error Correction Models 66
33
66
6/11/2023
8. Example
• VAR
𝑐 𝑎 𝑎 () 𝑎 () 𝑐 𝜀
𝑦 = 𝑎 + 𝑎 () 𝑎 () 𝑦 + 𝜀
Macroeconometrics Cointegration and Error Correction Models 67
67
8. Example
• VEC:
Δ𝑐 𝑎 𝜋 𝜋 𝑐 𝑎 () 𝑎 () Δ𝑐 𝜀
= 𝑎 + 𝜋 𝜋 𝑦 + 𝑎 𝑎 + 𝜀
Δ𝑦 () () Δ𝑦
Macroeconometrics Cointegration and Error Correction Models 68
34
68
6/11/2023
8. Example
• Rank Π = 𝛼𝛽′
Δ𝑐 𝛾 𝜏 𝛼 𝑐 𝜇 𝜌 𝑎 () 𝑎 () Δ𝑐 𝜀
= 𝛾 + 𝜏 𝑡+ 𝛼 𝛽 𝛽 + 𝜇 + 𝜌 𝑡 + 𝑎 𝑎 + 𝜀
Δ𝑦 𝑦 () () Δ𝑦
• r = 1 (one cointegrating vector)
Macroeconometrics Cointegration and Error Correction Models 69
69
8. Example
• Rank Π = 𝛼𝛽′
Δ𝑐 𝛾 𝜏 𝛼 𝛼 𝛽 𝛽 𝑐 𝜇 𝜌 𝑎 () 𝑎 () Δ𝑐 𝜀
= 𝛾 + 𝜏 𝑡+ 𝛼 𝛼 𝑦 + 𝜇 + 𝜌 𝑡 + 𝑎 𝑎 + 𝜀
Δ𝑦 𝛽 𝛽 () () Δ𝑦
• r = 2 (full cointegration)
• In the case of full cointegration, since all variables are stationary, the above VECM model
reduces to a VAR model with level variables.
Macroeconometrics Cointegration and Error Correction Models 70
35
70