Spurious Regression and Cointegration

Olusanya E. Olubusoye, Isaac Essi & OlaOluwa S. Yaya

2022-12-02

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 1 / 23
 Bibliography

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators

for Autoregressive Time Series with a Unit Root.” Journal of the
American Statistical Association 74 (366). American Statistical
Association, Taylor & Francis, Ltd.:427–31.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 2 / 23
 Bibliography

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators

for Autoregressive Time Series with a Unit Root.” Journal of the
American Statistical Association 74 (366). American Statistical
Association, Taylor & Francis, Ltd.:427–31.
Granger, C. W. J., and P. Newbold. 1974. “Spurious Regressions in
Econometrics.” Journal of Econometrics 2 (2):111–20.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 2 / 23
 Bibliography

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators

for Autoregressive Time Series with a Unit Root.” Journal of the
American Statistical Association 74 (366). American Statistical
Association, Taylor & Francis, Ltd.:427–31.
Granger, C. W. J., and P. Newbold. 1974. “Spurious Regressions in
Econometrics.” Journal of Econometrics 2 (2):111–20.
Racine, J. S. 2019. Reproducible Econometrics using R. Oxford
University Press, New York.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 2 / 23
 Bibliography

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators

for Autoregressive Time Series with a Unit Root.” Journal of the
American Statistical Association 74 (366). American Statistical
Association, Taylor & Francis, Ltd.:427–31.
Granger, C. W. J., and P. Newbold. 1974. “Spurious Regressions in
Econometrics.” Journal of Econometrics 2 (2):111–20.
Racine, J. S. 2019. Reproducible Econometrics using R. Oxford
University Press, New York.
Said, S. E., and D. A. Dickey. 1984. “Testing for Unit Roots in
Autoregressive-Moving Average Models of Unknown Order.”
Biometrika 71 (3):599.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 2 / 23
 Bibliography

Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Estimators

for Autoregressive Time Series with a Unit Root.” Journal of the
American Statistical Association 74 (366). American Statistical
Association, Taylor & Francis, Ltd.:427–31.
Granger, C. W. J., and P. Newbold. 1974. “Spurious Regressions in
Econometrics.” Journal of Econometrics 2 (2):111–20.
Racine, J. S. 2019. Reproducible Econometrics using R. Oxford
University Press, New York.
Said, S. E., and D. A. Dickey. 1984. “Testing for Unit Roots in
Autoregressive-Moving Average Models of Unknown Order.”
Biometrika 71 (3):599.
Yaya, O. S. 2022. Compendium of Time Series Econometrics with
Applications. Ibadan University Press.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 2 / 23
 Random Walks and Spurious Regression
@GRANGER_NEWBOLD:1974 point out that it is quite common to
encounter applied time series analysis with an apparently high degree
of fit as measured by, say, R 2

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 3 / 23
 Random Walks and Spurious Regression
@GRANGER_NEWBOLD:1974 point out that it is quite common to
encounter applied time series analysis with an apparently high degree
of fit as measured by, say, R 2
They pointed out how two random walk time series which are
statistically independent may nonetheless show a significant correlation,
and this phenomenon is called spurious correlation

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 3 / 23
 Random Walks and Spurious Regression
@GRANGER_NEWBOLD:1974 point out that it is quite common to
encounter applied time series analysis with an apparently high degree
of fit as measured by, say, R 2
They pointed out how two random walk time series which are
statistically independent may nonetheless show a significant correlation,
and this phenomenon is called spurious correlation
In their paper, @GRANGER_NEWBOLD:1974 examine the potential
for “discovering” spurious (nonexistent) relationships in time series
settings

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 3 / 23
 Random Walks and Spurious Regression
@GRANGER_NEWBOLD:1974 point out that it is quite common to
encounter applied time series analysis with an apparently high degree
of fit as measured by, say, R 2
They pointed out how two random walk time series which are
statistically independent may nonetheless show a significant correlation,
and this phenomenon is called spurious correlation
In their paper, @GRANGER_NEWBOLD:1974 examine the potential
for “discovering” spurious (nonexistent) relationships in time series
settings
They indicate exactly how “nonsense regression” relating two or more
time series can arise

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 3 / 23
 Random Walks and Spurious Regression
@GRANGER_NEWBOLD:1974 point out that it is quite common to
encounter applied time series analysis with an apparently high degree
of fit as measured by, say, R 2
They pointed out how two random walk time series which are
statistically independent may nonetheless show a significant correlation,
and this phenomenon is called spurious correlation
In their paper, @GRANGER_NEWBOLD:1974 examine the potential
for “discovering” spurious (nonexistent) relationships in time series
settings
They indicate exactly how “nonsense regression” relating two or more
time series can arise
This phenomenon is not restricted to random walks and can occur in
many time series settings, particularly when the errors are
autocorrelated
Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious
S. Yaya Regression and Cointegration 2022-12-02 3 / 23
 Random Walks and Spurious Regression

Below we consider two code chunks, one for cross-section i.i.d. data,
and one for time series data that contains a unit root.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 4 / 23
 Random Walks and Spurious Regression

Below we consider two code chunks, one for cross-section i.i.d. data,
and one for time series data that contains a unit root.
In both cases X is unrelated to Y, yet when X and Y are time series
that contain a unit root, they appear to be significantly related.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 4 / 23
 Random Walks and Spurious Regression

Below we consider two code chunks, one for cross-section i.i.d. data,
and one for time series data that contains a unit root.
In both cases X is unrelated to Y, yet when X and Y are time series
that contain a unit root, they appear to be significantly related.
This illustrates the pitfalls of ignoring the time dimension of your data
and why the study of time series methods is of fundamental
importance for applied economists.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 4 / 23
 Random Walks and Spurious Regression
##
## Call:
## lm(formula = y ~ x)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.6994 -0.6110 0.1832 0.6013 1.5516
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.09385 0.12840 0.731 0.4684
## x -0.19202 0.11259 -1.706 0.0946 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 '
##
## Residual standard error: 0.9075 on 48 degrees of freedom
## E.Multiple
Olusanya Olubusoye, IsaacR-squared: S. Yaya0.05714,
Essi & OlaOluwaSpurious Adjusted R-squared:
Regression and Cointegration 2022-12-02 0.037
5 / 23
 Random Walks and Spurious Regression

Note that for the above code chunk, there is no relationship between X
and Y which are both i.i.d. random variables, and the standard
procedures for statistical inference appears to reflect this fact and to be
sound.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 6 / 23
 Random Walks and Spurious Regression

In the next code chunk there is again no relationship between X and Y

but they are both random walks; yet the standard inferential
procedures fail and indicate the presence of a strong relationship when
in fact there is none.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 7 / 23
 Random Walks and Spurious Regression
##
## Call:
## lm(formula = dyn(y ~ x))
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.0402 -0.8390 0.3903 1.1701 3.1612
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.93945 0.31171 15.846 < 2e-16 ***
## x -0.44794 0.07835 -5.717 6.76e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 '
##
## Residual standard error: 1.754 on 48 degrees of freedom
## E.Multiple
Olusanya Olubusoye, IsaacR-squared: S. Yaya0.4051,
Essi & OlaOluwaSpurious Regression andAdjusted
Cointegration R-squared:
2022-12-02 0.3927
8 / 23
 Cointegration

A time series contains d unit roots if it became stationary after being

differenced d times. Thus, it is integrated of order d, that is I(d)

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 9 / 23
 Cointegration

A time series contains d unit roots if it became stationary after being

differenced d times. Thus, it is integrated of order d, that is I(d)
Two time series yt and xt that are both I(d) form a linear combination
(of these two series) and this linear combination is also an I(d), that is,
the residuals obtained from regressing yt on xt are I(d)

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 9 / 23
 Cointegration

A time series contains d unit roots if it became stationary after being

differenced d times. Thus, it is integrated of order d, that is I(d)
Two time series yt and xt that are both I(d) form a linear combination
(of these two series) and this linear combination is also an I(d), that is,
the residuals obtained from regressing yt on xt are I(d)
If however there exists a vector β such that the disturbance term from
the regression ϵt = yt − β × xt is of a lower order of integration that is,
I(d − b), where b > 0, then yt and xt are cointegrated of order I(d, b)

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 9 / 23
 Cointegration

Thus, if yt and xt are both I(1) and ut I(0), then the two series will be
cointegrated of order CI(1, 1) - The general definition of cointegration
as CI(d − b) by Engle and Granger (1987) does not necessarily mean
that d and b are integers but these can be fractional values

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 10 / 23
 Cointegration

Thus, if yt and xt are both I(1) and ut I(0), then the two series will be
cointegrated of order CI(1, 1) - The general definition of cointegration
as CI(d − b) by Engle and Granger (1987) does not necessarily mean
that d and b are integers but these can be fractional values
Thus, in the case of fractional orders of integration, d and b such that
b > 0, and d > b, we experience fractional cointegration between yt
and xt

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 10 / 23
 Cointegration

The economic interpretation of cointegration is that if two (or more)

series are linked to form an equilibrium relationship spanning the long
run, then even though the series themselves may contain stochastic
trend (i.e. nonstationary), they will nevertheless be more closely
together over time and the difference between them is constant
(i.e. stationary).

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 11 / 23
 Cointegration - Granger Causality test

Causality (in Granger’s sense), either in one direction or both directions

is not explained by cointegration since the essence of cointegration is
to obtain a long-run regression, while causality tells us if one variable
causes the other

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 12 / 23
 Cointegration - Granger Causality test

Causality (in Granger’s sense), either in one direction or both directions

is not explained by cointegration since the essence of cointegration is
to obtain a long-run regression, while causality tells us if one variable
causes the other
Following Granger’s (1969) causality test, for example, we can
investigate whether a variable yt causes xt , that is to see how much of
the current yt can be explained by past values of yt and then to see
whether adding lagged values of xt (that is xt−1 , xt−2 , ..., xt−k ) can
help to explain its behaviour. Thus, yt is said to be Granger-caused by
xt if xt helps in predicting yt .

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 12 / 23
 Cointegration - Granger Causality test
This can be two-way causation, given in Vector AutoRegressive (VAR)
form:

yt = α0 + α1 yt−1 + ... + αk yt−k + β1 xt−1 + ... + βk x−k + ut

xt = α0 + α1 xt−1 + ... + αk xt−k + β1 yt−1 + ... + βk y−k + ut

for all possible pairs of (yt , xt ). The reported F statistics in the

computation are the Wald statistics for the joint hypothesis:

H0 : β1 = β2 = ... = βk = 0

for each equation. The null hypothesis is that xt does not Granger-cause yt
in the first equation, and that yt does not Granger-cause xt in the second
equation
Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious
S. Yaya Regression and Cointegration 2022-12-02 13 / 23
 Cointegration - Granger Causality test

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 14 / 23
 Cointegration - Granger Causality test

There are three models here

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 15 / 23
 Cointegration - Granger Causality test

There are three models here

Model 1: It predicts rainfall using rainfall in previous three months and
temperature in previous three months

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 15 / 23
 Cointegration - Granger Causality test

There are three models here

Model 1: It predicts rainfall using rainfall in previous three months and
temperature in previous three months
Model 2: It predicts rainfall using only rainfall in previous three months

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 15 / 23
 Cointegration - Granger Causality test

There are three models here

Model 1: It predicts rainfall using rainfall in previous three months and
temperature in previous three months
Model 2: It predicts rainfall using only rainfall in previous three months
Another one with lag 1 below

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 15 / 23
 Cointegration - Granger Causality test
## Granger causality test
##
## Model 1: Rainfall ~ Lags(Rainfall, 1:1) + Lags(Temperature,
## Model 2: Rainfall ~ Lags(Rainfall, 1:1)
## Res.Df Df F Pr(>F)
## 1 1340
## 2 1341 -1 255.93 < 2.2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 '
## Granger causality test
##
## Model 1: Temperature ~ Lags(Temperature, 1:1) + Lags(Rainfa
## Model 2: Temperature ~ Lags(Temperature, 1:1)
## Res.Df Df F Pr(>F)
## 1 1340
## 2 1341 -1 105.5 S.< Yaya 2.2e-16
Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious
***
Regression and Cointegration 2022-12-02 16 / 23
 Cointegration - Johansen test

Johansen cointegration test allows us to determine if three or more

time series are cointegrated by allowing a stationary series of a linear
combination of the underlying series to be formed

xt = µ + A1 xt−1 + A2 xt−2 + ... + Ap xt−p + wt

where µ is the vector-valued mean of the series, Ai are the coefficient
matrices for each lag and wt is a multivariate Gaussian noise term with
mean zero.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 17 / 23
 Cointegration - Johansen test

Johansen cointegration test allows us to determine if three or more

time series are cointegrated by allowing a stationary series of a linear
combination of the underlying series to be formed
A general vector autoregressive model is similar to the AR(p) model
except that each quantity is vector valued and matrices are used as the
coefficients. The general form of the VAR(p) model, with drift, is given
by:

xt = µ + A1 xt−1 + A2 xt−2 + ... + Ap xt−p + wt

where µ is the vector-valued mean of the series, Ai are the coefficient
matrices for each lag and wt is a multivariate Gaussian noise term with
mean zero.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 17 / 23
 Cointegration - Johansen test
##
## ######################
## # Johansen-Procedure #
## ######################
##
## Test type: trace statistic , with linear trend
##
## Eigenvalues (lambda):
## [1] 0.3377689 0.3302054 0.0023202
##
## Values of teststatistic and critical values of test:
##
## test 10pct 5pct 1pct
## r <= 2 | 23.22 6.50 8.18 11.65
## r <= 1 | 4030.26 15.66 17.95 23.52
## r = 0 | 8150.85 28.71 31.52 37.22
## E. Olubusoye, Isaac Essi & OlaOluwaSpurious
Olusanya S. Yaya Regression and Cointegration 2022-12-02 18 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

Investigating cointegrating vector using the Narrow-Band Frequency

Domain Least Squares (NBFDLS) estimator for obtaining the residuals
from the cointegrating variables is a new method

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 19 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

Investigating cointegrating vector using the Narrow-Band Frequency

Domain Least Squares (NBFDLS) estimator for obtaining the residuals
from the cointegrating variables is a new method
The Ordinary Least Squares (OLS) estimator is standard but estimates
of residuals from cointegrating variables are usually biased, while the
NBFDLS estimator of Christensen and Nielsen (2006) is preferred to
the least square estimator and least square in the frequency domain
with periodograms as in Robinson (1994b).

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 19 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

Investigating cointegrating vector using the Narrow-Band Frequency

Domain Least Squares (NBFDLS) estimator for obtaining the residuals
from the cointegrating variables is a new method
The Ordinary Least Squares (OLS) estimator is standard but estimates
of residuals from cointegrating variables are usually biased, while the
NBFDLS estimator of Christensen and Nielsen (2006) is preferred to
the least square estimator and least square in the frequency domain
with periodograms as in Robinson (1994b).
The narrow band frequency domain least square (NBFDLS) estimator
would be used to obtain the cointegrating vector of the variables
provided each variable possesses long memory (0 < d < 0.5) and its
errors are weakly dependent.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 19 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.
Two, are differecing orders the same in the Unit differecing or fractional
unit root (fractional integration test.). Carry out homogeneity (equality
test) of fractional orders.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.
Two, are differecing orders the same in the Unit differecing or fractional
unit root (fractional integration test.). Carry out homogeneity (equality
test) of fractional orders.
Three, in the case of unit differencing, where series are either 1,1
orders or 2,2„ obtain the residual series using yt − β ∗ xt = ut where ut
is the residuals in this sense.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.
Two, are differecing orders the same in the Unit differecing or fractional
unit root (fractional integration test.). Carry out homogeneity (equality
test) of fractional orders.
Three, in the case of unit differencing, where series are either 1,1
orders or 2,2„ obtain the residual series using yt − β ∗ xt = ut where ut
is the residuals in this sense.
Four, test for differencing order of the residuals using ADF test. Sure,
this should be I(0) series.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.
Two, are differecing orders the same in the Unit differecing or fractional
unit root (fractional integration test.). Carry out homogeneity (equality
test) of fractional orders.
Three, in the case of unit differencing, where series are either 1,1
orders or 2,2„ obtain the residual series using yt − β ∗ xt = ut where ut
is the residuals in this sense.
Four, test for differencing order of the residuals using ADF test. Sure,
this should be I(0) series.
Then, since 0 is less that 1,1, cointegration is established in unit sense.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

One, Difference the series to be cointegrated to check for orders of

differencing. Use ADF test or fractional integration test.
Two, are differecing orders the same in the Unit differecing or fractional
unit root (fractional integration test.). Carry out homogeneity (equality
test) of fractional orders.
Three, in the case of unit differencing, where series are either 1,1
orders or 2,2„ obtain the residual series using yt − β ∗ xt = ut where ut
is the residuals in this sense.
Four, test for differencing order of the residuals using ADF test. Sure,
this should be I(0) series.
Then, since 0 is less that 1,1, cointegration is established in unit sense.
Note, you can only use OLS here but not Narrow-Band FDLS since
series are nonstationary and not in long memory range

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 20 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

In the case of fractional cointegration, it is expected that

du < min(dx , dy ) for cointegration to be established

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 21 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

In the case of fractional cointegration, it is expected that

du < min(dx , dy ) for cointegration to be established
If the cointegrating variables are in stationary long memory range, i.e
0 < d < 0.5, OLS estimates for residuals ut are not reliable but
Narrow-Band FDLS.

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 21 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

In the case of fractional cointegration, it is expected that

du < min(dx , dy ) for cointegration to be established
If the cointegrating variables are in stationary long memory range, i.e
0 < d < 0.5, OLS estimates for residuals ut are not reliable but
Narrow-Band FDLS.
See in Christensen and Nielsen (2006) for the methodology

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 21 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS

In the case of fractional cointegration, it is expected that

du < min(dx , dy ) for cointegration to be established
If the cointegrating variables are in stationary long memory range, i.e
0 < d < 0.5, OLS estimates for residuals ut are not reliable but
Narrow-Band FDLS.
See in Christensen and Nielsen (2006) for the methodology
So, residuals from the cointegrating pairs are obtained using the
Narrow-Band betas β as yt − β ∗ xt = ut

Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious

S. Yaya Regression and Cointegration 2022-12-02 21 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS
## [1] 0.24 0.48 -0.24 0.24 0.24 0.47 2.12 2.58 3.30
## [13] 2.97 2.49 1.79 2.23 1.55 0.66 1.32 1.31 1.31
## [25] 1.51 1.29 1.29 0.00 0.00 0.85 1.49 1.91 1.91
## [37] 1.88 2.50 1.64 1.64 1.84 2.24 2.42 2.82 3.61
## [49] 3.10 3.25 3.98 3.97 4.51 3.52 3.83 4.36 3.78
## [61] 5.03 3.40 3.02 2.00 1.32 2.13 3.09 4.26 5.05
## [73] 5.74 7.07 7.97 9.10 9.68 11.05 11.00 12.18 11.63
## [85] 9.46 8.42 6.57 5.88 6.56 7.77 8.40 9.12 9.03
## [97] 8.95 9.38 8.80 9.44 9.51 9.50 10.48 10.97 12.31
## [109] 11.38 11.54 10.60 9.64 7.70 5.88 5.33 4.66 5.16
## [121] 3.62 3.98 4.11 4.13 4.30 3.93 4.16 4.39 4.02
## [133] 4.11 3.96 3.96 4.08 4.47 4.93 5.30 5.21 5.42
## [145] 6.44 6.21 5.81 4.09 1.58 1.37 1.20 1.81 2.12
## [157] 0.55 0.00 0.16 -0.04 1.52 2.70 2.33 2.05 1.46
## [169] 2.12 1.61 1.72 1.04 1.04 1.00 0.85 1.10 0.77
Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious
S. Yaya Regression and Cointegration 2022-12-02 22 / 23
 Fractional Cointegration - Using OLS and
Narrow-Band FDLS
## [1] -0.46234398 -0.16820497 -0.83552917 -0.32041197 -0.30
## [7] 1.67664536 2.12640285 2.72495587 3.39569153 2.50
## [13] 1.99403451 1.42624151 0.71160935 1.17209438 0.63
## [19] 0.50059869 0.46572401 0.38378387 0.96038053 -0.21
## [25] 0.39356571 0.18819788 0.27013801 -0.90719431 -0.88
## [31] 0.64133436 1.05109184 1.05109184 0.22620904 1.10
## [37] 1.16741350 1.80204567 0.97131000 0.99179503 1.23
## [43] 1.86836733 2.32250635 3.11689600 3.52640285 3.43
## [49] 2.55860985 2.70422020 3.44299950 3.35983867 3.86
## [55] 3.16716287 3.70594217 3.15520650 4.73057433 4.10
## [61] 4.31741350 2.55572401 2.11719534 1.10743786 0.40
## [67] 2.19743786 3.36304821 4.17645967 3.77182751 4.12
## [73] 4.88109184 6.27986302 7.17986302 8.29084120 8.91
## [79] 10.22888484 11.35035617 10.64964486 9.47062304 9.77
## [85] 8.44599088 7.41477018 5.52087368 4.79283005 5.42
Olusanya E. Olubusoye, Isaac Essi & OlaOluwaSpurious
S. Yaya Regression and Cointegration 2022-12-02 23 / 23