Chapter 11: Panel Data Models

Financial Econometric Modeling

Stan Hurn, Vance Martin, Peter C.B. Phillips and Jun Yu

Oxford University Press, 2020

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 1 / 31
Introduction

Panel data has combined measurements amounting to N × T observations,

describing N individual units observed longitudinally over T time periods.
Panel data are often classified into long panels where the time series sample size
T is large and wide panels where cross section sample size N is large. Panels
with T small and N large and are called short wide panels. Similarly, long narrow
panels have large T and small N.
1 Empirical research on asset pricing and investments makes use of an abundance of
time series data and these problems are usually associated with large T and small N.
2 Empirical corporate finance research typically have short wide panels with small T and
large N.
Panel data can be also classified into balanced panels and unbalanced panels.
1 A balanced panel is one with no missing observations where every cell of the T × N
matrix contains an observed data point.
2 The presence of cells with missing observations yields an unbalanced panel as the
sample sizes for each cross section member are no longer the same.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 2 / 31
Types of Models

When using panel data a complete set of equations for the system comprises an
equation specified for each of the N cross sections. Such a complete system may
be estimated jointly.
Even though the combination of time series and cross section data produces a
large number, N × T , of observations, panel models also involve a large number of
unknown parameters. These may relate to
1 specific characteristics of each unit or individual in the cross section; or
2 common forms of financial agent responses.
Panel data models lead to large numbers of potentially nuisance parameters –
called incidental parameters – and a much smaller number of homogeneous
parameters that result from commonality restrictions that apply across the
parameters of the N equations for each cross section unit.
Panel models are often classified according to restrictions that are imposed on the
parameters.
1 No common effects model where no restrictions at all are imposed on the coefficients of
the equations.
2 Common effects where all equations have common coefficients imposed.
3 Intermediate cases are where individual specific incidental parameters – usually the
intercept coefficients in the equations – are allowed to differ but parameters on
explanatory variables are constrained.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 3 / 31
Reasons for Using Panel Data
Methods of estimating panel models are designed to take advantage of the
statistical power involved in cross section averaging whenever commonality is
present. This typically leads to greater efficiency and higher rates of consistency
in estimation.
A panel may improve reliability in estimation and inference by mitigating the effects
of some poor data that may be present in the panel by virtue of the averaging
process that includes both high and low quality data.
Panel data also enable analysts to monitor performance in financial assets over
time and compare time profiles of different assets. In panels this is facilitated by
pooling the cross section observations in a way that facilitates estimation of time
series features, such as dynamic responses, that might reasonably be assumed to
be common across the panel.
Combining time series data and cross section data into a panel offers the prospect
of improved accuracy in parameter estimation due to the augmented sample size
(of magnitude N × T ). The improvement in statistical efficiency from having a
larger sample size is reflected in smaller variances, sharper (more powerful) tests,
and shorter confidence intervals.
Rich data sets that come with panels also lead to an expanded range of
hypotheses of interest. Certain hypotheses may be tested using panel data that
are not possible with time series or cross section data alone.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 4 / 31
 Two Introductory Panel Models

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 5 / 31
No Common Effects

The most unrestricted panel model is the no common effects model

yjt = αj + βj xjt + ujt , ujt ∼ (0, σj2 ),
where yjt and xjt are respectively the dependent and explanatory variables
associated with the j th cross section at time t.
The no common effects model is the most unrestricted form within this class of
panel model as the intercepts αj , slopes βj , and variances σj2 are all allowed to
vary across section, thereby providing a framework that accommodates full
heterogeneity over individuals in the cross section.
An example is the no common effects CAPM
Stock 1: r1t − rft = α1 + β1 (rmt − rft ) + u1t , u1t ∼ (0, σ12 ),
Stock 2: r2t − rft = α2 + β2 (rmt − rft ) + u2t , u2t ∼ (0, σ22 ),
.. .. .. .. ..
. . . . .
Stock N: rNt − rft = αN + βN (rmt − rft ) + uNt , uNt ∼ (0, σN2 ).
This model can also be extended to allow for multiple explanatory variables.
This decomposition of the no common effects CAPM means that the parameters
of the model can be estimated by applying ordinary least squares to each
equation separately.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 6 / 31
Common Effects

The common effects model involves extreme pooling of information. Each equation
of this model is assumed to have the same intercept and same slope coefficients

yjt = α + βxjt + ujt , ujt ∼ (0, σj2 ),

where the regression coefficients are the same for each equation j = 1, · · · , N but
equation error variances σj2 are heterogeneous.
Tthe common effects CAPM has the explicit form

rjt − rft = α + β(rmt − rft ) + ujt ,

with ujt representing the idiosyncratic risk of the j th stock at time t.

For this model both the α-risks and the β-risks are assumed to be the same across
all N stocks, although the model still allows for different idiosyncratic risks σj2 .
The common effects CAPM has only two regression parameters α and β. By
comparison, the number of unknown parameters in the no common effects model
is 2N. The difference of R = 2N − 2 parameters represents the number of
restrictions that the common effects model imposes on the no common effects
model.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 7 / 31
Stacking

To estimate the common effects model with the restrictions imposed the approach
adopted is to stack the data. This is equivalent to expressing the data in long form.
For the CAPM example, stacking requires that the dependent variable rjt − rft of
each stock are stacked on top of each other to produce a new variable that is an
(NT × 1) vector of excess returns on all N stocks over all time periods.
Stacking the explanatory variable is achieved in exactly the same way except that
the excess returns on the market are the same for all stocks, so the stacked
explanatory variable simply stacks the same variable on itself N times to produce
an (NT × 1) vector of excess returns of the market.
Estimation then involves a linear regression of the stacked dependent variable on
a constant vector – represented by an (NT × 1) vector of ones – and the stacked
explanatory variable.
This process of stacking the data and applying least squares to the stacked
system imposes the cross-equation restrictions that the α-risks and the β -risks
are the same across all N stocks.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 8 / 31
Testing
The common effects model is a restricted version of the no common effects model
and these restrictions may be tested.
The null and alternative hypotheses are
H0 : α1 = α2 = · · · = αN = α and β1 = β2 = · · · = βN = β
H1 : at least one restriction fails.
Under the null hypothesis there are two unknown parameters given by α and β
compared to the 2N unknown parameters in the no common effects model.
To test these restrictions, the F statistic is calculated according to the formula
RSSR − RSSU N(T − 2)
F = × ,
RSSU R
where RSSU is the unrestricted (no common effects) residual sum of squares,
RSSR is the restricted (common effects) residual sum of squares and R is the
number of restrictions. This statistic is distributed as FR,N(T −2) .
An asymptotic version of the F test is the chi-square test which is obtained by
multiplying both sides of the F statistic by R
RSSR − RSSU
R×F = × N(T − 2).
RSSU
This statistic is distributed asymptotically under the null hypothesis as χ2R .
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 9 / 31
 Fixed and Random Effects Models

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 10 / 31
The Fixed Effects Model

The simple bivariate panel regression then has the following form

yjt = αj + βxjt + ujt , ujt ∼ (0, σ 2 ),

where the simplifying assumption of constant equation error variance over the
cross sections is made, so that σ12 = σ22 = · · · = σN2 = σ 2 .
In this specification the αj are considered fixed parameters by assumption, thereby
giving the model its name. The fixed effects model is characterised by N
intercepts (αj : j = 1, · · · , N) and 1 slope parameter (β). As discussed earlier, the
intercepts are known as incidental parameters.
The fixed effects model imposes N − 1 restrictions on the no common effects
model by requiring the slope coefficient β to be common in each equation.
Alternatively, a comparison of the fixed effects model and the common effects
model suggests that the common effects model imposes N − 1 restrictions on the
fixed effects model by requiring the constant term to α to be common across all
equations.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 11 / 31
Dummy Variable Representation

An alternative way to specify the fixed effects model is to define a set of dummy
variables as 
1 cross section j
Djt =
0 otherwise,
The fixed effects model is now rewritten as

yjt = α1 D1t + α2 D2t + · · · + αN DNt + βxjt + ujt ,

or more compactly
N
X
yjt = αj Djt + βxjt + ujt .
j=1

The fixed effects model is estimated by an ordinary least squares regression of the
(NT × 1) stacked dependent variable yjt on the (NT × 1) stacked explanatory
variable xjt and the N stacked dummy variables, D1t , D2t , · · · , DNt . Alternatively, if a
constant is included in the regression equation then one of the dummy variables is
excluded to avoid the dummy variable trap of collinear regressors.
The dummy variable approach is only feasible in long panels as the number of
dummy variables required by a wide panel would make the approach inefficient.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 12 / 31
Estimating the Fixed Effects Model

A general method for estimating fixed effects models is as follows.

1 Express the equation for the j th cross section in time-averaged form as
T T T
1 X 1 X 1 X
yjt = αj + β xjt + ujt
T T T
t=1 t=1 t=1

yj = αj + βx j + u j .

2 Express the data in deviation-from-the-mean form, given by

yjt − y j = β(xjt − x j ) + ujt − u j ,

in which the fixed effects αj have been eliminated.

3 Estimate the resultant equation by ordinary least squares to obtain β.
b
4 The estimates of the fixed effects for each of the j assets are recovered as

bj = y i − βx
α b j,

where βb is the least squares estimator from Step 3.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 13 / 31
The Random Effects Model
Assume that the differences in the cross section elements involve random variable
intercepts so that
αj = α + vj , vj ∼ (0, σv2 ).
where the vj are random effects that lead to the differences in the intercepts αj .
The vj are assumed to be distributed with zero mean and constant variance, so
that the random intercepts in each equation all have the same mean α. More
importantly, the random errors vj in these intercepts are typically assumed to be
independent of the equation errors ujt .
The random effects model is then
yjt = α + vj + βxjt + ujt , ujt ∼ N(0, σ 2 ),
where, once again, the assumption of constant variance over the cross sections is
used. The random effects, vj , are assumed to be independent of both the
disturbances ujt , so that E(vj ujt ) = 0, and the explanatory variable xjt , so that
E(vj xjt ) = 0, for all j and t.
The random effects model may be written in terms of a single disturbance as
yjt = α + βxjt + wjt ,
where the composite disturbance wjt is
wjt = ujt + vj .
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 14 / 31
Covariance of the Composite Error

Estimation of the random effects model is somewhat more involved than the fixed
effects model. The complexity arises from the composite form of the equation
disturbance term wjt .
The covariance structure of wjt is

E(wjt ) = E(ujt + vj ) = E(ujt ) + E(vj ) = 0

var(wjt2 ) = E[(ujt + vj )2 ] = E(ujt2 ) + E(vj2 ) + 2E(ujt vj ) = σ 2 + σv2
cov(wjt wjs ) = E[(ujt + vj )(ujs + vj )] = E(ujt ujs ) + E(ujt vj ) + E(ujs vj ) + E(vj2 )
= σv2 .

There is therefore a specific correlation structure which is constant through time

cov(wjt wjs ) σ2
ρts = p = 2 v 2.
var(wjt )var(wjs ) σ + σv

Estimation of the model must take account of this correlation structure.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 15 / 31
Estimating the Random Effects Model

The steps involved to estimate the parameters of the random effects model in this more
general case are as follows.
1 Choose starting values for σ 2 and σv2 .
2 Define the quasi-difference parameter
σ
λ=1− p ,
σ 2 + T σv2
and compute the quasi-differenced data

yejt = yjt − λy j , xejt = xjt − λx jt ,

where y j and x j are time averages.

3 The generalised least squares estimator based on the starting values for (σ 2 , σv2 )
is obtained by estimating the transformed equation

yejt = α(1 − λ) + β xejt + w

ejt ,

in which w
ejt is a disturbance term, by ordinary least squares.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 16 / 31
Hausman Test

A formal test of fixed effects versus random effects in modelling panel data can be
based on the fundamental assumption that the random effects, vj , are considered
to be independent of the explanatory variables, xjt , that is E(vj xjt ) = 0.
Formally the hypotheses are

H0 : random effects
H1 : fixed effects.

To perform the test, let βbFE and βbRE represent, respectively, the fixed and random
effects estimates of the slope parameters.
The general form of the test is based on the Wald statistic
h i0 h i−1 h i
WD = βbFE − βbRE cov(βbFE ) − cov(βbRE ) βbFE − βbRE ,

where cov(βbFE ) and cov(βbRE ) are the estimated covariances of βbFE and βbRE .
The test statistic WD is distributed asymptotically under the null hypothesis as χ2R
where R represents the number of explanatory variables in the model excluding
the constant.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 17 / 31
 Dynamic Panel Data Models

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 18 / 31
Introducing Dynamics

The panel data models discussed so far specify a relationship between the
dependent variable, yjt , and the explanatory variable, xjt , both at time t.
Dynamics may be introduced into the model by including lags of both yjt and xjt .
The respective fixed effects and random effects versions of the so-called panel
autoregressive model are as follows:

Fixed effects : yjt = αj + βxjt + ρyjt−1 + ujt

Random effects : yjt = α + vj + βxjt + ρyjt−1 + ujt .

where ujt is the disturbance term which is assumed to be independent over time
E(ujt ujs ) = 0, for all t 6= s.
The autoregressive parameter ρ controls the strength of the dynamics and, in this
formulation of the model, is assumed to take a common value across section.
1 If ρ = 0 the model reduces to the non-dynamic or static panel data model.
2 If −1 < ρ < 1 the autoregressive dynamics are stationary with the successive effects of
shocks on the dependent variable dissipating over time.
3 If ρ = 1 the shocks have persistent effects and do not dissipate over time, resulting in a
nonstationary panel model.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 19 / 31
Nickell Bias

Elimination of the fixed effect intercepts from each equation by removing

time-averages, produces a form of endogeneity in the regressor that influences
estimation by introducing a new form of bias.
The bias arises from the correlation of the lagged regressor with the equation
error. The transformed lagged dependent variable yjt−1 − y jt−1 is correlated with
the transformed disturbance term ujt − u j because

σ2
cov(y jt−1 , ujt ) = 6= 0, for all t.
T
This correlation violates one of the key conditions needed for the validity of least
squares. In particular, this condition is needed to ensure consistent estimation of
the autoregressive coefficient ρ as N → ∞ with finite time series sample size T .
This non-zero correlation is present even when the disturbance term, ujt , is itself
not autocorrelated.
The resultant bias in estimating ρ is commonly referred to as Nickell bias, following
its original analysis by Nickell (1981). It has the asymptotic form
1+ρ
ρ − ρ) ' −
plim (b .
N→∞ T

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 20 / 31
Illustrating Nickell Bias

Consider a simple simulation experiment with a dynamic panel model to illustrate

the impact of the Nickell bias. The data generating process is
yjt = αj + βxjt + ρyjt−1 + ujt , ujt ∼ N(0, σ 2 ),
xjt = γxj t−1 + ejt , ejt ∼ N(0, σe2 ),

with population parameters ρ = {0.2, 0.8}, β = 0.8, γ = 0.5 and σ 2 = 1.0.

σe2 is determined by specifying a value for the signal-to-noise ratio defined as
−1
(γ + ρ)2 ρ2

 1 
σs2 = var yjt − αj = β 2 σe2 1 + (ργ − 1) − (ργ)2 + σ2 .
1−ρ 1 + ργ 1 − γρ2

The higher the value of σs2 the more useful xjt is in explaining yjt .
The settings T = {10, 20, 30} reveal the performance of the fixed effects estimator
in short, medium and long panels; and the settings N = {20, 100} illustrate
estimator performance in narrow and wide panels. Each parameter combination is
assessed using 1000 replications.
Results should demonstrate that the fixed effects estimator of ρ suffers from
significant bias and that the bias of the fixed effects estimator increases with ρ and
decreases as T increases.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 21 / 31
Results

N T ρ = 0.2 ρ = 0.8
Mean Std. Dev. Mean Std. Dev.
20 10 0.138 0.051 0.560 0.070
20 20 0.175 0.034 0.693 0.040
20 30 0.181 0.027 0.731 0.029
100 10 0.140 0.023 0.567 0.030
100 20 0.172 0.015 0.694 0.018
100 30 0.182 0.012 0.733 0.013

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 22 / 31
Estimating Dynamic Panel Models

To avoid Nickell bias is to adopt a two step approach.

1 Remove the fixed effects (αj ) or random effects (vj ) by taking first differences
∆yjt = β∆xjt + ρ∆yjt−1 + ∆ujt .
While this equation removes intercepts, the differencing operation induces correlation
between the lagged differenced regressor ∆yjt−1 and the equation error ∆ujt . In
particular E(∆yjt−1 ∆ujt ) = −E(yjt−1 ujt−1 ) = −σ 2 . Thus, least squares estimation
leads to biased and inconsistent estimates of the remaining parameters {β, ρ}.
2 To correct the bias and inconsistency of least squares estimation, instrumental variable
or GMM methods are employed. Suitable instruments may be constructed using
carefully chosen lagged values of the dependent variable and exogenous regressors.
In choosing the set of instruments to employ at each time period t, there are three
commonly-used approaches which involve different choices of instruments. In
general all the approaches use combinations of lagged levels and differences of
the dependent variable as suitable instruments.

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 23 / 31
 Nonstionary Panel Data Models

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 24 / 31
Preamble

If the autoregressive coefficient ρ in a dynamic panel model does not satisfy the
stability restriction |ρ| < 1, the variables for each cross section are nonstationary
and may be cointegrated with common long run parameters.
An example is the present value model applied to a cross section of countries.
The question is whether these countries have a common cointegrating vector that
embodies the same present value relationship.
A general specification of the present value model for a set of N, countries is
written as
pjt = αj + βj djt + ujt , j = 1, 2, · · · , N,
where pjt is the log equity price of country j, djt is the log dividend of country j and
ujt is a disturbance term.
Assuming that pjt and djt are nonstationary time series and ujt is stationary, the
parameters αj and βj represent the cointegrating parameters which in this
formulation may vary for each country. T
The parameter αj is a function of the relevant discount factor that relates to
country j.
This model is comparable to the no common effects panel model with the
parameters for each cross section being estimated separately using appropriate
methods.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 25 / 31
Panel Scatter Plots of the Present Value Model for 12 Countries

Australia Brazil Canada

9.0 11.0 9.5
Log Price

Log Price

Log Price
8.5 10.0 9.0
8.0 9.0 8.5
7.5 8.0 8.0
4 4.5 5 5.5 3 4 5 6 7 4 4.5 5 5.5 6
Log Dividends Log Dividends Log Dividends
China France Germany
7.0 9.0 9.5
6.5 8.8
Log Price

Log Price

Log Price
8.6 9.0
6.0
8.4 8.5
5.5 8.2
5.0 8.0 8.0

0 .5 1 1.5 2 2.5 4 4.5 5 5.5 4 4.5 5 5.5 6

Log Dividends Log Dividends Log Dividends
Hong Kong Japan Mexico
8.5
5.0 8.0
Log Price

Log Price

Log Price
8.0 4.8 7.5
4.6 7.0
7.5
4.4 6.5
7.0 4.2 6.0
3.5 4 4.5 5 -.5 0 .5 1 2.5 3 3.5 4 4.5
Log Dividends Log Dividends Log Dividends
South Korea U.K. U.S.
1.0 9.6 7.6
0.5 9.4 7.4
Log Price

Log Price

9.2 Log Price 7.2

0.0
9.0 7.0
-0.5 8.8 6.8
-1.0 8.6 6.6
-5 -4.5 -4 -3.5 5.4 5.6 5.8 6 6.2 2.8 3 3.2 3.4 3.6 3.8
Log Dividends c Log Dividends Log Dividends
Financial Econometric Modeling Hurn, Martin, Phillips & Yu Oxford University Press, 2020 26 / 31
Restrictions

Restricting the slope parameters to be the same across the N countries

β1 = β2 = · · · = βN = β,

imposes a common long run relationship between equity prices and dividends.
The result is a present value fixed effects cointegrating model with homogeneous
slopes of the form
pjt = αj + βdjt + ujt .
The fixed effects are given by the parameters {α1 , α2 , · · · , αN }, which allow the
discount factor for each country to differ.
A further set of restrictions is to assume that the discount factors are also the
same across all N countries by imposing the restrictions

α1 = α2 = · · · = αN = α.

The result is the present value common effects cointegrating model

pjt = α + βdjt + ujt .

Financial Econometric Modeling c

Hurn, Martin, Phillips & Yu Oxford University Press, 2020 27 / 31
Panel Unit Root Tests - LLC

Rather than testing each variable separately for a unit root, a test statistic is
constructed to assess the presence of a unit root jointly among all N variables (or
cross sections) simultaneously. This strategy can be interpreted as an efficient
procedure of testing for the presence of a time series unit root in a variable that
varies not only over time but also over N cross sections.
The Levin, Lin and Chu (2002) test is based on the panel unit root regression
equation
XL
∆yjt = αj + βyj t−1 + γjk ∆yj t−k + ujt .
k=1
The null and alternative hypotheses of the panel unit root test are given by
H0 : β=0 [unit root]
H1 : β<0 [stationary].
Note that the null hypothesis requires that all time series in the panel have unit
roots (ρ := β + 1 = 1) and the alternative similarly requires that all time series
have stationary roots (|ρ| < 1).
This model is the panel analogue of the augmented Dickey-Fuller regression
equation with β = ρ − 1. Note however that the panel unit root t statistic is
asymptotically distributed as N(0, 1) under the null hypothesis.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 28 / 31
Panel Unit Root Tests - IPS
A less restrictive panel unit root test that allows for some stationary and
nonstationary elements in the cross section under the alternative hypothesis was
proposed by Im, Pesaran and Shin (2003).
This test involves the weaker formulation
XL
∆yjt = αj + βj yj t−1 + γjk ∆yj t−k + ujt ,
k=1

in which the slope coefficients βj are now not restricted.

The null and alternative hypotheses for this test are
H0 : βj = 0, j = 1, 2, · · · , N
βj < 0 for j = 1, · · · , N1
(
H1 :
βj = 0 for j = N1 + 1, · · · , N.
Under the null hypothesis all cross sections have unit roots as before, whereas
under the alternative hypothesis just N1 of the N cross sections are stationary.
An overall panel unit root test is computed by estimating separately for each cross
section and using a t statistic to test that βj = 0. An overall statistic is computed by
averaging across the N values of these t statistics. The averaged t statistic the
asymptotic distribution of the modified panel unit root test is also standard normal.
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 29 / 31
Panel Cointegration

As with cointegration testing in non-panel data models there exists a range of test
procedures for cointegration in dynamic panel models. Many of these tests rely on
Fisher’s (1932) idea of combining individual tests to produce a meta-analysis test.
If πj is the p value of the test for cross section j for a particular hypothesis, the joint
statistic based on the N cross sections is
N
X
−2 log (πj ) → χ22N ,
j=1

which is asymptotically distributed as chi square with 2N degrees of freedom

when T → ∞.
Having established cointegration in the panel, the cointegrating parameter vector
β together with the fixed effects αj , are estimated.
A dynamic least squares approach augments the panel equation to include leads
and lags of ∆djt as follows
L
X
pjt = αj + βdjt + γjk ∆djt−k + ujt ,
k =−L

which is the estimated by ordinary least squares.

Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 30 / 31
Panel Cointegration
As with cointegration testing in non-panel data models there exists a range of test
procedures for cointegration in dynamic panel models. Many of these tests rely on
Fisher’s (1932) idea of combining individual tests to produce a meta-analysis test.
In testing for panel cointegration, the procedure is to apply (for instance) the
Johansen cointegration test to each cross sectional unit separately. If πj is the p
value of the test for cross section j for a particular hypothesis, the joint statistic
based on the N cross section outcomes is constructed as
N
X
−2 log (πj ) → χ22N ,
j=1

which is asymptotically distributed as chi square with 2N degrees of freedom.

A dynamic least squares approach augments the panel equation to include leads
and lags of ∆djt as follows
L
X
pjt = αj + βdjt + γjk ∆djt−k + ujt ,
k =−L

which is the estimated by ordinary least squares.

The estimator of the cointegrating parameter vector β has t ratio statistics that are
asymptotically distributed as N(0, 1).
Financial Econometric Modeling c
Hurn, Martin, Phillips & Yu Oxford University Press, 2020 31 / 31