Upload
others
View
9
Download
0
Embed Size (px)
Citation preview
Exponent of Cross-sectional Dependence for Residuals∗
Natalia BaileyMonash University
George KapetaniosKing’s College London
M. Hashem PesaranUniversity of Southern California, and Trinity College, Cambridge
April 4, 2019
Abstract
In this paper, we focus on estimating the degree of cross-sectional dependence in theerror terms of a classical panel data regression model. For this purpose we propose anestimator of the exponent of cross-sectional dependence denoted by α, which is based onthe number of non-zero pair-wise cross correlations of these errors. We prove that ourestimator, α, is consistent and derive the rate at which α approaches its true value. Wealso propose a resampling procedure for the construction of confidence bounds around theestimator of α. We evaluate the finite sample properties of the proposed estimator by useof a Monte Carlo simulation study. The numerical results are encouraging and supportiveof the theoretical findings. Finally, we undertake an empirical investigation of α for theerrors of the CAPM model and its Fama-French extensions using 10-year rolling samplesfrom S&P 500 securities over the period Sept 1989 - May 2018.
Keywords: Pair-wise correlations, Cross-sectional dependence, Cross-sectional aver-ages, Weak and strong factor models, CAPM and Fama-French Factors.
JEL Codes: C21, C32
∗This paper is written in honour of the 125th anniversary of PC Mahalanobis, for his important contributionsto statistics and econometrics. We would like to acknowledge helpful comments from Alex Chudik. NataliaBailey and Hashem Pesaran acknowledge financial support under ESRC Grant ES/I031626/1.
1 Introduction
Interest in the analysis of cross-sectional dependence applied to households, firms, markets,
regional and national economies has become prominent over the past decade, especially so in
the aftermath of the latest financial crisis given its effects on the global economy. Researchers
in many fields have turned to network theory, spatial and factor models to obtain a better
understanding of the extent and nature of such cross dependencies. There are many issues
to be considered: how to test for the presence of cross-sectional dependence, how to measure
the degree of cross-sectional dependence, how to model cross-sectional dependence, and how
to carry out counterfactual exercises under alternative network formations or market inter-
connections. Many of these topics are the subject of ongoing research. In this paper we focus
on measuring cross-sectional dependence.
Bailey, Kapetanios and Pesaran (2016, BKP hereafter) give a thorough account of the ratio-
nale and motivation behind the need for determining the extent of cross-sectional dependence,
be it in finance, micro or macroeconomics. They focus on the asymptotic behaviour of the
variance of the cross section average of the observations on a double array of random variables,
say xit, indexed by i = 1, 2, . . . , N and t = 1, 2, . . . , T, over space and time. In particular, they
analyse the rate at which this variance tends to zero and show that it depends on the degree or
exponent of cross-sectional dependence which they denote by α. They explore a factor model
setting as a vehicle for characterising strong and semi-strong covariance structures as defined in
Chudik et al. (2011). They relate these to the degree of pervasiveness of factors in unobserved
factor models often used in the literature to model cross-sectional dependence.
In this paper we build on BKP and extend the analysis in two respects. First, we consider
a more generic setting which does not require a common factor representation and holds more
generally for both moderate to sizable cross-sectional dependence. We achieve this by directly
considering the significance of individual pair-wise correlations, and do not concern ourselves
with the factors that might underlie these pair-wise correlations. Second, we consider estimating
the exponent of cross-sectional dependence, α, of the residuals obtained from a panel data
regression model.
We propose a new estimator of α based on the number of statistically significant pair-wise
correlations of the residuals from the panel regression under consideration. To establish the
statistical significance of the correlation coefficients we adopt the thresholding multiple testing
(MT) estimator proposed by Bailey et al. (2018), BPS. Other thresholding estimators can also
be used. See, for example, Bickel and Levina (2008) or Karoui (2008) and Cai and Liu (2011) or
Fan et al. (2013). The MT testing procedure advanced by BPS has the advantage that it directly
considers the statistical significance of the correlation coefficient which is invariant to scales.
Other thresholding procedures focus on the sample covariances and resort to cross validation to
identify the threshold. Bickel and Levina (2008) use universal thresholding, namely comparing
all the sample covariances to the same threshold value, whilst Cai and Liu (2011) propose an
‘adaptive’ thresholding procedure that allows for differing thresholds across the different pairs
of sample covariances. Other contributions to this literature include the work of Huang et al.
(2006), Rothman et al. (2009), Cai and Zhou (2011) and Cai and Zhou (2012), Wang and
1
Zhou (2010), and Fan et al. (2011).1 All these contributions apply the thresholding procedure
to sample covariances and do not apply to the residuals from a panel regression model that
concerns us in this paper. It is also important to bear in mind that when estimating α we do
not assume that the underlying error covariance matrix is sparse, as is assumed in the literature
on regularization of the sample covariance. Our objective is to estimate the degree of sparsity
of the covariance matrix rather than assume sparsity for the purpose of consistent estimation
of the covariance matrix or its inverse. What matters for estimation of α is to ensure that all
non-zero entries of the correlation matrix are correctly identified.
We establish consistency of our estimator under the assumptions of exogeneity of regressors
and symmetry of the error distribution. We also explain how the derivations can be extended
to the case when weakly exogenous variables are present, as for example in a dynamic panel
data setting. The proposed estimator is simple to compute and is shown to perform well in
small samples, for a variety of correlation matrices, irrespective of whether the cross correlations
are generated from a multi-factor structure or specified by a given correlation matrix with a
specified degree of sparsity. This is especially the case as compared to basing the estimation of
α on the largest eigenvalue of the correlation matrix, which performs particularly poorly. The
rate of convergence of our preferred estimator is complex and depends on an interplay of the
cross-sectional and time dimensions, N and T . The Monte Carlo results also show that the error
in estimating α is smaller for values of α close to unity, which is likely to be of greater interest
in practice. The problem of making inference about the value of α raises additional technical
difficulties and will not be addressed in this paper. In practice, bootstrap techniques can be
used to obtain confidence bounds around our proposed estimator. We provide some Monte
Carlo results in support of estimating the empirical distribution of the proposed estimator of
α, using cross-sectional resampling as suggested in Kapetanios (2008). Finally, we provide an
empirical application investigating the degree of inter-linkages between financial variables using
the Standard & Poor’s 500 index. We present 10-year rolling estimates of α applied to excess
returns on securities included in the S&P 500 data set as well as α estimates applied to the
residuals obtained from the CAPM and its Fama-French extensions used extensively in the
finance literature.
The rest of the paper is organised as follows: Section 2 discusses alternative characterisa-
tions of α, the exponent of cross-sectional dependence, and the conditions under which these
measures are equivalent as N → ∞. Section 3 sets up the panel data model and discusses its
underlying assumptions. Section 4 proposes the estimator of α in terms of the number of statis-
tically significant non-zero pair-wise correlations of the residuals. Section 5 presents the main
theoretical results of the paper for a static panel data model with strictly exogenous regressors.
Extensions to dynamic panels or panels with weakly exogenous regressors are discussed in the
sub-section 5.2 while inference of α by use of bootstrap procedures is discussed in sub-section
5.3. Section 6 presents a detailed Monte Carlo simulation study. The empirical application
is discussed in Section 7. Finally, Section 8 concludes. Proofs of all theoretical results are
provided in the Appendix.
1Shrinkage procedures have also been proposed in the literature for regularization of covariance matrices.See, for example, Stein (1956), Ledoit and Wolf (2003) and Ledoit and Wolf (2004). However, the shrinkageprocedure does not set any elements of the covariance matrix to zero, and is not suitable for estimation of αwhich builds on the support recovery properties of the estimated covariance matrix.
2
2 Degrees of cross-sectional dependence: alternative mea-
sures
Our analysis focuses on the covariance matrix of εt = (ε1t, ε2t, . . . , εNt)′, where εt is the N × 1
vector of errors from a panel data regression model. Let ΣN = E (εtε′t) = (σij), and denote
its largest eigenvalue by λmax (ΣN) > 0. The errors εit are said to be strongly cross-sectionally
correlated, if λmax (ΣN) = (N), where denotes exact order of magnitude, and they are said
to be weakly cross-sectionally correlated, if λmax (ΣN) is bounded in N . All intermediate cases
can be parameterized in terms of the exponent αλ, such that
λmax (ΣN) = (Nαλ). (1)
The weak and strong cross dependence cases then relate to αλ = 0 and αλ = 1, respectively. It
is important to emphasise that the exponent, αλ, is an asymptotic concept, in the sense that
αλ can be identified only as N →∞, as the definition in (1) makes clear.
Suppose now that the cross dependence of εit is characterized by the following approximate
multiple-factor error process
εit = β′ift + uit, (2)
where ft is the m × 1 vector of unobserved common factors with zero means, and βi =
(βi1, βi2, . . . , βim)′ is the associated m × 1 vector of factor loadings, and uit is the idiosyn-
cratic component assumed to have mean zero and the covariance matrix V =E(utu′t), where
ut = (u1t, u2t, . . . , uNt)′. Then
ΣN = E (εtε′t) = BB′ + V,
where B = (β1,β2, . . . ,βN)′, and without loss of generality we have set E (ftf′t) = Im. To
identify the factor component from the idiosyncratic component we assume that λmax (V) =
O(1), but allow the factor loadings to satisfy the condition
B′B =N∑i=1
βiβ′i = (Nαβ), αβ > 0, (3)
where αβ measures the degree to which the factors are pervasive, in the sense that they have
non-zero effects on the individual errors, εit. In what follows we refer to αβ as the exponent of
factor loadings. In the standard approximate factor models it is assumed that αβ = 1, whilst
in practice, where the possibility of weak factors can not be ruled out, αβ could be a parameter
of interest to be estimated.
To see how αβ and αλ are related note that
λmax (ΣN) ≤ ‖BB′ + V‖1 ≤ ‖B‖1 ‖B‖∞ + ‖V‖1 ,
where ‖B‖1 and ‖B‖∞ are column and row norms of B, respectively. To ensure that V ar(εit)
is bounded we must have ‖B‖∞ < K. Also to ensure that λmax (V) = O(1), we must have
‖V‖1 < K. Therefore, the rate at which λmax (ΣN) can rise with N is controlled by
‖B‖1 = max1≤j≤N
N∑i=1
|βij| . (4)
3
Setting∑N
i=1 |βij| = (Nαβj ), for j = 1, 2, . . . ,m, then ‖B‖1 = (Nαβ), where αβ = maxj(αβj).
Then
λmax (ΣN) = (Nαβ) +O(1).
To distinguish the effects of the factor component from those of the idiosyncractic component
we must have αβ > 0. Comparing this result with (1) establishes that αλ = αβ > 0, as N →∞.
The above analysis suggests two alternative ways of estimating αλ. A direct procedure
would be to base the estimate of αλ on λmax (ΣN) and set
λmax (ΣN) = O (Nαλ) = κNαλ ,
where κ is a constant independent of N . Then
αλ =ln [λmax (ΣN)]
ln (N)− ln (κ)
ln (N). (5)
In order to identify αλ, as N →∞, we set κ = 1, so that (5) becomes
αλ =ln [λmax (ΣN)]
ln (N). (6)
In this form, the value of αλ is susceptible to the scaling of elements in εt. For this reason
we focus our attention rather on the corresponding correlation matrix RN = (ρij) given by
RN = D−1/2N ΣND
−1/2N ,
where
DN = diag(σii, i = 1, 2, . . . , N). (7)
Hence, (6) finally becomes
αλ =ln [λmax (RN)]
ln (N), (8)
and αλ has fixed bounds at zero and unity, as N →∞.
Developing a theory based on the maximum eigenvalue of the correlation matrix RN can
be challenging. To avoid some of the technical problems involved in estimating λmax (RN),
and noting that V ar(εt) ≤ N−1λmax (ΣN) , BKP propose basing the estimation of αλ on
V ar(εt), where εt = N−1∑N
i=1 εit. In the case where εit has a factor representation, BKP
show that V ar(εt) = O[max
(N2(αβ−1), N−1
)], which reduces to Std(εt) = O
(N(αβ−1)
), if
2 (αβ − 1) > −1, or if αβ > 1/2. This means that at least N1/2 of the factor loadings must
have non-zero values for ΣN to differ sufficiently from a diagonal ΣN .
In this paper we consider an alternative estimation strategy that does not require εit to have
a factor representation. Since
λmax (RN) ≤ ‖RN‖1 = max1≤j≤N
N∑i=1
|ρij| ,
we focus directly on estimation of ρij and distinguish between values of ρij that are close to
zero and those that are significantly different from zero, and measure the exponent of cross-
sectional dependence in terms of the number of significant (non-zero) cross-correlation coef-
ficients. Specifically, we define α such that MN = N2α where M is the number of non-zero
4
elements of RN which can be written equivalently as MN = τ ′N∆NτN , where τN is an N × 1
vector of ones and ∆N = (δij) is an N × N matrix of population correlation indicators with
typical elements given by
δij = I (ρij 6= 0) , i, j = 1, 2, . . . , N,
in which I(A) is equal to unity if A is true and zero otherwise. Note that by construction
δii = 1. Hence,
α =ln (MN)
ln(N2)=
ln (τ ′N∆NτN)
lnN2. (9)
Cross-sectional independence refers to the case when RN = IN , and α = 1/2, while the case
of cross-sectional strong dependence corresponds to all pair-wise correlation coefficients being
non-zero such that α = 1. Note that by construction 1/2 ≤ α ≤ 1, with α = 1/2 arising when
∆N = IN , and α = 1 if ρij 6= 0 for all i and j.
Other exponents of cross-sectional dependence can be defined by focussing only on the off-
diagonal elements of RN and consider the following exponent of cross-sectional dependence:2
α =ln [τ ′N (∆N − IN) τN ]
lnN (N − 1),
assuming that ∆N 6= IN . Unlike α the above measure is not defined if RN = IN . The two
measures coincide, namely α = α = 1, if ρij 6= 0 for all i and j, as N →∞. In cases where εithave a multi-factor error representation given by (2), the largest exponent of the factor loadings
is given by αβ > 0. Assuming, for simplicity that V is diagonal, it then readily follows that
τ ′N (∆N − IN) τN = N2αβ − Nαβ , where (N2αβ −Nαβ) = Nαβ (Nαβ − 1) is the total number
of off-diagonal non-zero pair-wise cross correlations of the errors. In such a multi-factor error
set up we have
α =ln (τ ′N∆NτN)
lnN2=
ln (N2αβ −Nαβ +N)
lnN2,
and
α =ln [τ ′N (∆N − IN) τN ]
lnN (N − 1)=
ln (N2αβ −Nαβ)
lnN (N − 1).
Recall that we must have αβ > 1/2 for factors to be distinguishable from the idiosyncratic
components. It is then easily seen that limN→∞ α = limN→∞ α = αβ. However, the two
measures could differ if N is not sufficiently large. In finite samples α can be written in terms
of α by first solving the quadratic equation
N2αβ + (N −Nαβ) = N2α, (10)
2One can also consider only the distinct off-diagonal elements of RN and define α as
α =ln[
12τ′N (∆N − IN ) τN
]ln[
12N (N − 1)
] =ln [τ ′N (∆N − IN ) τN ]− ln(2)
ln [N (N − 1)]− ln(2)
=α− ln(2)
ln[N(N−1)]
1− ln(2)ln[N(N−1)]
→ α, as N →∞.
5
for αβ, namely
αβ =ln(
1 +√
1− 4 (N −N2α))− ln 2
lnN. (11)
Since [Nαβ ] can only take positive or zero values the second root of (10) is clearly redundant.
In what follows, we focus on α since it is defined even if RN = IN , and α is suitably scaled to
lie in the range (1/2, 1] .
3 Panel data model
Consider the panel data regression model
yit = γ′ixit + εit, for i = 1, 2, . . . , N ; t = 1, 2, . . . , T, (12)
where xit is a k× 1 vector of observed regressors, γi is the associated vector of coefficients, and
εit are the model’s errors. We are interested in estimating the exponent of the cross-sectional
dependence of the errors, εit, defined by (9). First, we obtain residuals eit computed as
eit = yit − x′itγi = yit − x′it(X′iXi)
−1X′iyi, (13)
where Xi is the T × k matrix of observations on the regressors for the ith unit, and yi is the
T × 1 vector of observations on the dependent variable of the ith unit. We assume that the
regressors are strictly exogenous.
We define the standardized errors, ξit, and the associated standardized residuals, zit, as
ξit =εit
(T−1ε′iMiεi)1/2, (14)
zit =eit
(T−1e′iMiei)1/2
=eit
(T−1e′iei)1/2, (15)
where εi = (εi1, εi2, . . . , εiT )′, ei = (ei1, ei2, . . . , eiT )′, and Mi = IT− Xi(X′iXi)
−1X′i. Note that
(since ei = Miεi)
zit =εit + x′it(γi − γi)(T−1ε′iMiεi)
1/2= ξit − a′itξi, (16)
where
ait = Xi (X′iXi)
−1xit,
and ξi = (ξi1, ξi2, . . . , ξiT )′.
Further, in what follows we assume that the error terms are symmetrically distributed.
Assumption 1 Conditional on Xi, the errors of the panel data model, (12), (a) εit are sym-
metrically distributed with zero means and variances 0 < c < σ2i < K <∞, (b) εit are serially
independent, (c) εit and εjt are distributed independently if E (εitεjt) = 0, for all i 6= j.
Under the above assumption, and using (14) it readily follows that
E (ξi |Xi ) = 0, (17)
6
and
E(ξiξ′j |Xi,Xj ) = ρijIT . (18)
Our main analysis will condition on the observed regressors. Remark 2 will discuss an
unconditional version of our results. For the observed regressors, we make the following as-
sumption:
Assumption 2 The k×1 vector of regressors xit in (12) is bounded: supi,t ‖xit‖ <∞. Further,
for some T0, we have
infT>T0,i
λmin
(X′iXi
T
)> 0.
Under Assumption 1 it readily follows that
E (a′itξi |Xi ) = x′it(X′iXi)
−1X′iE (ξi |Xi ) = 0,
where x′it is the tth row of Xi = (xi1,xi2, . . . ,xiT )′. Hence,
E (zit |Xi ) = 0, (19)
which in turn implies that zit is a martingale difference process with respect to the non-
decreasing information set, Ωit = (xi1,xi2, . . . ,xit). By construction ΩiT ≡ Xi. Indeed, since
Ωit is a subset of ΩiT , then by the chain rule of conditional expectations we have
E (zit |Ωit ) = E [E (zit |ΩiT ) |Ωit ] .
But in view of (19), E (zit |ΩiT ) = 0, and it also follows that E (zit |Ωit ) = 0, for all i and t.
We also require the following assumption that sets a lower bound condition on the non-zero
values of the pair-wise correlations.
Assumption 3 Let ρmin = infi,j (|ρij| |ρij 6= 0). Then,
limT→∞
ln(T )√Tρmin
= 0. (20)
Remark 1 This assumption is needed for successful recovery of non-zero pair-wise correlations,
and is weaker than requiring ρmin > 0, as it allows ρmin to tend to zero with N or T or both, so
long as its rate of decline is slower than ln(T )/√T .
4 Consistent estimation of α
Consider the sample estimate of the pair-wise correlation coefficients of the residuals from units
i and j,
ρij =T−1e′iej
(T−1e′iei)1/2 (T−1e′jej)1/2 , (21)
7
where ei = (ei1, ei2, . . . , eiT )′, eit is defined by (13), and by construction the sample mean of eitis exactly zero. We can re-write (21) equivalently as
ρij = T−1T∑t=1
zitzjt, (22)
where zit is defined by (15). In order to identify whether the pair-wise correlation coefficients
ρij are significantly different from zero we follow Bailey et al. (2018) and apply the multiple
testing estimator associated with ρij. This is defined by
ρij = ρijI
(|ρij| >
cp(n, δ)√T
), (23)
where
cp(n, δ) = Φ−1(
1− p/2
nδ
), n =
1
2N(N − 1), (24)
n is the number of tests carried out, p is the nominal size of the individual test, which can
be set to 1%, 5% or 10%, Φ−1(.) is the inverse of the standard normal distribution function,
and δ is a tuning parameter to be set a priori. This thresholding method is based on the
notion that for each unit (i, j) pairs we carry out a total of 12N(N − 1) individual tests of the
null hypothesis that ρij = 0 where j 6= i, i, j = 1, 2, . . . , N . Such tests can result in spurious
outcomes especially when N is larger than T . The critical value function, cp(n, δ), is therefore
adjusted using parameter δ to take account of the effects of the multiple testing procedure
for the estimation of α. It is important to bear in mind that the multiple testing problem
encountered here differs from the standard one studied in the literature by Bonferroni (1935),
Holm (1979) and others. Our focus here is on identifying the range of values for δ such that
α can be consistently estimated, rather than controlling the overall size of the multiple tests
being carried out.
Accordingly, we propose to estimate α by
α =ln(τ ′∆τ
)2 lnN
, (25)
where ∆ = (δij), with
δij =
ρij, if |ρij| > cp(n,δ)√
T, for i 6= j
1, for i = j0, otherwise
.
5 Theoretical Derivations
5.1 Main Results
To establish that α converges to α, in addition to Assumptions 1, 2 and 3, we also require the
following additional technical sub-exponential assumption:
8
Assumption 4 There exist sufficiently large positive constants C0, C1, s > 0, such that
supi,t Pr (|εit| > α) ≤ C0 exp (−C1αs) , for all α > 0. (26)
This assumption is used to allow a relatively simple bounding of an infinite sum of proba-
bilities, needed for the proof of Lemmas 2 and 3 in the Appendix. It can be relaxed to allow
for fatter tails, at the expense of smaller allowable values for N .
The rate of convergence of α to α is given in Theorem 1 below:
Theorem 1 Consider the panel data regression model (12) and suppose that Assumptions 1-4
hold. Let α and δ be defined by (25), (23), and (24). Then, conditional on the observed xit, as
N, T →∞, and if, for some d > 0, N = O(T d) = o (exp(T )) ,
2 (lnN) (α− α) = O(N2(1−α−κδ))+O
(N2(1−α) exp
(−C0T
C1))
+O(N−α
)+O
(N1−2α) =
(27)
O(T 2d(1−α−κδ))+O
(exp
[2d(1− α) ln(T )− C0T
C1])
+O(T−dα
)+O
(T d(1−2α)
)for any 0 < κ < 1, and some C0, C1 > 0.
As long as δ is set large enough (δ > 1−α), the first term on the RHS of (27) can be made
sufficiently small.
Remark 2 If we do not wish to condition on the observed xit, one could obtain the result
of Theorem 1, unconditionally, if it is assumed that the regressors satisfy the following sub-
exponential condition for some s > 0,
supi,t Pr (‖xit‖ > α) ≤ C0 exp (−C1αs) , for all α > 0, (28)
and if for some T0,(
X′iXi
T
)−1exists for all T > T0. Under these conditions on xit (which
replace Assumption 2), we can then use Lemma A6 of Chudik et al. (2018) to establish suitable
probability bounds on(
ξ′iXi
T
)(X′iXi
T
)−1 (X′iξiT
), and to show that, for some C0, C1 > 0, and
0 < π < 1,
supi,j
Pr(|∑T
t=1zitzjt| >√Tcp (n, δ)
)≤ sup
i,jPr(|∑T
t=1ξitξjt| > (1− π)√Tcp (n, δ)
)+exp
(−C0T
C1).
5.2 Extension to panels with weakly exogenous regressors
In the case of panels with lagged dependent variables, the use of OLS residuals for estimation of
α could still be justifiable so long as T is sufficiently large, such that the time series bias in the
estimated residuals is not too large. This is supported by the Monte Carlo evidence provided
for dynamic panels below.
However, the mathematical proofs provided above will not be applicable to the OLS residuals
if the panel regression model, (12), contains weakly exogenous regressors, such as lagged values
of yit. An alternative approach which avoids some of the technical issues associated with the use
9
of OLS residuals would be to base the estimation of α on the recursive residuals. Specifically,
one could consider the recursive residuals defined by
eit = yit − γ′i,t−1xit,
where
γi,t−1 =
(t−1∑τ=1
xiτx′iτ
)−1( t−1∑τ=1
xiτyiτ
).
Then the pair-wise correlations based on these recursive residuals are given by
ρij = (T − h)−1T∑t=h
zitzjt,
where
zit =eitσit,
and
σ2it =
1
t
t∑τ=1
e2iτ .
Here h is the size of the training period, which needs to be set by the researcher. It is then
easily seen that under cross-sectional independence, zitzjt is a martingale process with respect
to Ωιi,t−1, where Ωιi,t−1 = (yiτ , xiτ ; for τ = t−1, t−2, . . . , 1). This and other related results then
allow us to apply the mathematical analysis of the previous sections to the recursive residuals,
after suitable adjustments. The main open question is what critical value to use when checking
the significance of ρij, and hence the threshold value in the determination of the indicators δijdefined above. This issue will not be pursued in this paper.
5.3 Confidence intervals for α
Quantifying the uncertainty surrounding the proposed estimator of α is clearly of interest.
Given the complexity of developing asymptotic inferential theory, a fruitful avenue is to use
bootstrap procedures. In the case of panel datasets a number of important data features
matter. One possibility is to consider a parametric bootstrap where residuals are resampled.
Again there are a number of ways to construct such a bootstrap method. The first is to
resample, with replacement, from the rows of the residual matrix with the rows referring to
time periods. This procedure applies if the residuals are serially uncorrelated, although block
resampling methods can be considered to deal with the serial correlation. It is important to
resample whole rows as otherwise the nature of the cross sectional dependence of the original
sample, on which α depends, is not inherited by the bootstrap sample. Initial experimentation
suggests that coverage rates for such bootstrap methods are very low. An alternative is to
use some estimation method for large dimensional covariance matrices, such as thresholding,
to estimate the covariance matrix of the residuals and then implement a wild bootstrap. This
approach has been considered, for example, by Goncalves and Perron (2018). However, its
desirable properties depend on the true covariance being sparse and certainly sparser that the
10
structures we consider in this paper. Another alternative is to resample from columns (cross
section units) of the residual matrix, namely to resample across units keeping all the residuals
(observations) on a given together. This procedure is robust to the serial correlation problem.
This resampling procedure, initially proposed by Kapetanios (2008), will be used to obtain
bounds on the estimator of α in the Monte Carlo section below.
6 Monte Carlo Simulations
We investigate the small sample properties of our proposed estimator of α, defined by (25),
using a number of different simulation designs, allowing for dynamics as well as non-Gaussian
errors. We consider the following relatively general dynamic panel data model
yit = ai + ϑiyi,t−1 + γixit + εit, for i = 1, 2, . . . , N ; t = 2, 3, . . . , T, (29)
with exogenous, but serially correlated, regressors:
xit = ρixxi,t−1 +(1− ρ2ix
)1/2νit, for i = 1, 2, . . . , N ; t = −50,−49, . . . , 0, . . . , T,
where ρix ∼ IIDU (0, 0.95), and νit ∼ IIDN (0, 1) , with xi,−50 = νi,−50 for i = 1, 2, . . . , N. The
first 50 observations are disregarded for all units to minimize the effects of the initial values on
the observations used in the estimation.
We consider the following cases: (i) a static panel data model, where ϑi = 0, ai ∼IIDN (1, 1), and γi ∼ IIDN (1, 1) , for i = 1, 2, . . . , N ; and (ii) a dynamic panel data model
with exogenous regressors, where ϑi ∼ IIDU (0, 0.95), γi ∼ IIDN (1, 1) , and ai ∼ IIDN (1, 1),
for i = 1, 2, . . . , N .3 Our estimator is robust to possible correlations between the fixed effects,
αi, and the regressors, xit.
We consider two different designs for generating the errors, εit, both with the same exponent
of cross-sectional dependence, α :
Design 1 We draw N × 1 vector bN = (b1, b2, . . . , bN)′ as Uniform (0.7, 0.9) for the first
Nb (≤ N) elements, where Nb = [Nαβ ] and set the remaining elements to zero. Then, we
construct the correlation matrix RN given by
RN = IN + bNb′N −B2N , (30)
where BN = Diag (bN) is an N ×N diagonal matrix with its ith diagonal element given
by bi. The degree of sparseness of RN is determined by the choice of αβ. If αβ = 0 then
RN = IN and α = 1/2, while if αβ = 1 then all elements of RN will be non-zero and we
have α = 1. For all intermediate values of αβ, RN will have a total of [Nαβ (Nαβ − 1) +N ]
non-zero elements. The exact relationship between α and αβ is given by (11). Further,
we generate the variances of εit as σii ∼ IID 0.5 [1 + 0.5χ2(2)] , for i = 1, 2, . . . , N, and
set DN = Diag(σii, i = 1, 2, . . . , N). We now generate εit so that its correlation matrix is
3We also consider (iii) a dynamic panel data model with no exogenous regressors, where ϑi ∼ IIDU (0, 0.95)and γi = 0, for i = 1, 2, . . . , N . Simulation results for case (iii) are similar to those for cases (i) and (ii) and areavailable in the online supplement, Tables S1a-S1d.
11
equal to RN . To this end we first obtain matrix PN as the Cholesky factor of RN , and
then set WN = D1/2N PN = (wij), where wij is the (i, j) element of WN . We then generate
εit as
εit =N∑j=1
wijujt, i = 1, 2, . . . , N ; t = 1, 2, . . . , T, (31)
where ujt are IID draws from Gaussian or non-Gaussian distributions, to be specified
below.4
Design 2 The second design closely follows the set up in BKP and employs the two-factor
specification given by
εit = βi1f1t + βi2f2t + uit, for i = 1, 2, . . . , N ; t = 1, 2, . . . , T, (32)
where fjt ∼ IIDN(0, 1), j = 1, 2 and uit ∼ IIDN(0, 1) for i = 1, 2, . . . , N . With regard
to the factor loadings, we generate them as follows:
βi1 = vi1, for i = 1, 2, . . . , [Nαβ1 ] (33)
βi1 = 0, for i = [Nαβ1 ] + 1, [Nαβ1 ] + 2, . . . , N
βi2 = vi2, for i = 1, 2, . . . , [Nαβ2 ] ,
βi2 = 0, for i = [Nαβ2 ] + 1, [Nαβ2 ] + 1, . . . , N,
where βi2 are then randomised across i to achieve independence from βi1. The loadings
are generated as vij ∼ IIDU(µvj − 0.2, µvj + 0.2), for j = 1, 2. We examine the case
where αβ2 < αβ1 = αβ and consider values of αβ and αβ2 such that αβ2 =2αβ3
. We set
µv2 = 0.71 and µv1 =
√µ2v −N
2(αβ2−αβ)µ2v2
such that µ2v1
+ µ2v2
= µ2v = 0.75 - see BKP
for further details. Both µv1 and µv2 are chosen such that µvj 6= 0, j = 1, 2, without µ′vjs
being too distant from zero either. As before, the exact relationship between α and αβ is
given by (11).
In both designs, we examine two cases for the innovations ujt: (i) Gaussian, where ujt ∼IIDN(0, 1) for j = 1, 2, . . . , N ; (ii) non-Gaussian, where ujt follows a multivariate t-distribution
with v degrees of freedom. This is achieved by generating ujt as
ujt =
(v− 2
χ2v,t
)1/2
νjt, for j = 1, 2, . . . , N,
where νjt ∼ IIDN(0, 1) and χ2v,t is a chi-squared random variate with v = 8 degrees of freedom.
For the estimation of α, in each replication r = 1, 2, . . . , R, we first compute the OLS
residuals
e(r)it = yit − a(r)i − ϑ
(r)i yi,t−1 − γ(r)i xit, i = 1, 2, . . . , N ; t = 2, 3, . . . , T,
where a(r)i , ϑ
(r)i , and γ
(r)i are the OLS estimators of the regressions of yit on an intercept, yi,t−1
and xit, computed using the observations t = 2, 3, . . . , T, for each i. In the case of the static
4Note that εt = WNut, and V ar(εt) = WNW′N = D
1/2N PNP′ND
1/2N = D
1/2N RND
1/2N , as required.
12
regressions where ϑi = 0, the residuals are computed from regressions of yit on an intercept and
xit using the observations t = 1, 2, . . . , T . The sample covariance matrix, Σ(r)
N = (σ(r)ij ), is then
computed as
σ(r)ij = T−1
T∑t=2
e(r)it e
(r)jt , for i, j = 1, 2, . . . , N,
and diagonal elements σ(r)ii collected in D
(r)
N = Diag(σ(r)ii , i = 1, 2, . . . , N). The corresponding
sample correlation matrix is then given by
R(r)
N = D(r)−1/2N Σ
(r)
N D(r)−1/2N .
We evaluate the bias and root mean squared error (RMSE) of the exponent of cross-sectional
dependence α computed as in (25) with the critical value, cp(n, δ), given by (24). For p and
δ we consider the values p = 0.05, 0.10 and δ = 1/2, 1/3.5 Further, we compute the bias-
corrected version of the exponent of cross-sectional dependence estimator developed in BKP,
α, and compare its performance with that of α.6 However, it is important to bear in mind
that BKP provide theoretical justification for their estimator only in the case of demeaned
observations, namely xit − xi, and do not consider residuals from panel regressions as we do in
this paper. As a by-product, this paper also provides Monte Carlo evidence on the properties
of the estimator, α, when applied to residuals from panel regressions.
Finally, for the construction of confidence intervals for α, we propose the following bootstrap
procedure:
Bootstrap In each replication, r = 1, 2, . . . , R, we collect residuals in matrix E(r) =(e(r)it
)T×N
and proceed to resample (with replacement) from the columns a total number of B times.
More precisely, in each bootstrap b = 1, 2, . . . , B, the N cross section units are reshuf-
fled with replacement to generate a new matrix of residuals E(r),(b) =(e(r),(b)(i)t
)T×N
,
where (i) denotes the resampled index of the N units. For each b, we then compute
the correlation matrix R(r),(b)N corresponding to residuals E(r),(b) and estimate α, which
we denote by α(r),(b). The bootstrap estimates of α are collected in the vector α(r),B =(α(r),(1), α(r),(2), . . . , α(r),(B)
)′from where we obtain the estimates of α that correspond to
the 0.05 and 0.95 percentiles, and which we denote by α(r),B0.05 and α
(r),B0.95 , respectively. Fi-
nally, we evaluate the frequency at which α(r),B0.05 < α(r) < α
(r),B0.95 across the R replications
of each MC experiment.
5The value of δ = 1/4 was also considered. Results are in line with those for δ = 1/3 and are available inthe online supplement, Tables S2a-S2d.
6Recall that α corresponds to the most robust bias-adjusted estimator of the exponent of cross-sectionaldependence considered in Bailey et al. (2016) and allows for both serial correlation in the factors and weakcross-sectional dependence in the error terms. It is given by
α = α(µ2v
)= 1 +
1
2
ln(σ2x)
ln(N)−
ln(µ2v
)2 ln (N)
− cN2 [N ln(N)] σ2
x
, (34)
where σ2x, µ
2v and cN are consistent estimators of σ2
x, µ2v, and cN - see BKP for further details. We use four
principal components when estimating cN .
13
For all experiments we consider the values of α = 0.55, 0.60, . . . , 0.90, 0.95, 1.00, and the
sample sizes N, T = 100, 200, 500, and carry out all experiments with R = 2, 000 replications
and B = 500 bootstraps. The values of ai, ϑi and γi are drawn randomly in each replication.
6.1 Small sample results
First, we consider the small sample performance of our proposed estimator, α, and investigate
its robustness to different choices of p and δ, that govern the critical value, cp(n, δ), used in
estimating it. The results for Gaussian errors are provided in Tables A1a to A2b, and the
results for non-Gaussian errors are reported in Tables A3a to A4b. Each table reports bias and
RMSE of α computed using the residuals from either static or dynamic panel data regressions.
Tables A1a and A1b give the results for static and dynamic panels, respectively, when the
cross-sectional dependence in the errors are generated according to Design 1, whilst the same
results for Design 2 are summarized in Tables A2a and A2b. Similarly, the results in Tables
A3a and A3b give bias and RMSE of α for static and dynamic panels when the errors are
non-Gaussian and the cross correlations are generated according to Design 1, whilst the same
results for Design 2 are provided in Tables A4a and A4b. In each of these tables, the left panels
give bias and RMSE for p = 0.05, and the right panels for p = 0.10, whilst the top panels give
the results for δ = 1/2, and the bottom panels for δ = 1/3. Specifically, each Table gives four
sets of results for the combinations (p, δ), with p = 0.05, 0.10 and δ = 1/3, 1/2.
Comparing the left and right panels of the tables, it is clear that α is robust to the choice
of p, irrespective of the value of δ, and for all N and T combinations. Observing that n =
N (N − 1) /2 is quite large even for moderate values of N , the effective p-value of the underlying
individual tests is given by 2p/nδ, which is likely to be dominated by the choice of δ as compared
to p. Therefore, the test outcomes are more likely to be robust to the choice of p as compared
to δ.
Turning to the choice of δ, comparing the results reported in top and bottom panels of the
tables, we note that for all N and T combinations the choice of δ = 1/2 produces smaller bias
and RMSE as compared to δ = 1/3 for values of α close to 1/2 (α ≤ 0.75). The reverse is true
when considering values of α close to unity (α > 0.80). Again, this is consistent with the result
of Theorem 1 which requires δ to be larger than 1− α. Hence, for α → 1/2 setting δ = 1/2 is
more appropriate, while as α→ 1 values of δ below 1/2 are more appropriate. In cases where
there is no priori information regarding the range in which the true value of α might fall, the
simulation results suggest setting δ to its upper bound value of δ = 1/2.
Overall, irrespective of whether we consider static or dynamic panel regressions, with Gaus-
sian or non-Gaussian errors, the tabulated results show that the small sample performance of
α improves as the true exponent of cross-sectional dependence, α, rises from 0.55 towards 1.0,
uniformly over N and T combinations. This finding holds for both Designs, although α gen-
erally performs better when Design 1 is used to generate the error cross-sectional dependence.
Further, both bias and RMSE of α diminish as N rises for all values of T considered. These re-
sults are in line with our main theoretical findings as set out in Theorem 1. It is also interesting
to note that under Design 1, the bias and RMSE of α are particularly small for values of α in
the range of 0.9−1, even if we consider dynamic panels with non-Gaussian errors. For example,
14
for T = 100 and N = 500, p = 0.05, δ = 1/3, the bias and RMSE of estimating α = 0.95 by α,
in the case of dynamic panels with non-Gaussian errors are −0.00008 and 0.00067, respectively.
(see Table A3b).
Tables A2a and A2b summarize the results for static and dynamic panel data models,
respectively, when the error cross-sectional dependence is generated by Design 2 (the two-factor
structure). Compared with Tables A1a and A1b, both bias and RMSE are more sizeable across
the range of α when T = 100. However, as T increases, the performance of α improves for all
values of α, especially when α approaches unity, as to be expected. Perhaps, the signal-to-noise
ratio implied by (32) becomes somewhat distorted when the T dimension is short and adversely
affects the accuracy of the multiple testing procedure used to identify the non-zero elements of
the error correlation matrix, RN . To verify this conjecture, we repeated the same experiments
attaching a scaling parameter of ς =√
1/2 to uit in (32), in line with the simulation setup in
BKP. Performance of α is much improved in this case and comparable to those shown in Tables
A1a and A1b, even when T is small.7 Our conclusions regarding the robustness of α to the
choice of p and δ arrived at under Design 1 continue to hold for Design 2.
We now consider the small sample performance of α relative to that of α, the estimator of
α proposed in BKP. α is a biased-corrected estimator of α based on the standard deviation
of the cross-sectional average of the residuals. As noted earlier, the asymptotic properties of
α are established only for demeaned observations, but it is conjectured that these asymptotic
properties are likely to hold even if α is computed using residuals from panel regressions. For
comparison we consider bias and RMSE of α computed using p = 0.05 and δ = 1/2, and
note that similar results are obtained for other choices of p and δ. Table B1 compares the
resulting bias and RMSE of the two α estimators when applied to residuals obtained from a
static (top panel) and dynamic panel data models (bottom panel). These results refer to Design
1 with Gaussian errors. Both estimators perform well for all values of α, and irrespective of
whether the panel regressions are static or dynamic. This is particularly so for values of α >
0.8. In comparative terms, α outperforms α on average, for all values of α, and all N and
T combinations. The superior performance α is more pronounced when α ≤ 0.75 uniformly
over N and T . The results for Design 2 (where the cross-sectional dependence of the errors
are generated using a two-factor specification) are summarized in Table B2 which has the same
format as Table B1. In the case of these experiments, α (the estimator proposed by BKP)
performs better than α when T is small (T = 100), but the bias and RMSE of α becomes more
comparable to α as both N and T rise. As noted above, scaling uit by ς in (32) eliminates this
relative outperformance of α.8
Further, Table C displays bias and RMSE results for estimates of the exponent of cross-
sectional dependence, given by (8), and computed using the maximum eigenvalue of the corre-
lation matrices RN derived from the residuals from a static or dynamic panel data model with
Gaussian errors generated under Designs 1 or 2. It is clear that all eigenvalue based estimates
of α perform rather poorly even for large values of N and T , and even for values of α close to
1.
7These results are available in the online supplement - see Table S3.8Comparison of α and α when errors are non-Gaussian are provided in Tables S4a-S4b.of the online supple-
ment.
15
Turning to the problem of sampling uncertainty, we compute bootstrapped confidence inter-
vals for the estimate α in the case of Design 1 with Gaussian errors. The results are summarised
in Table D and give the average coverage rates (in percent) over the R simulations of α being
between the 5th and 95th percentiles of its empirical distribution. The related confidence in-
tervals are constructed by applying the resampling procedure with replacement to the residuals
obtained for the static and dynamic panel data models, with the results summarized in the
top and bottom panels of Table D, respectively. We set δ = 1/2 in the critical value function
cp(n, δ), and p = 0.05 (left panel) p = 0.10 (right panel), in each replication of the MC exper-
iments and in each bootstrap. For α = 0.60 coverage is low for small values of N , but rises
steadily towards unity as N increases and for all values of T considered. For 0.65 ≤ α ≤ 0.95
coverage stands universally at 100% but drops significantly when α = 1.00. Indeed, in this case
for small T coverage is relatively high and improves as N rises but drops toward zero when the
T dimension is increased for all values of N considered. This is to be expected since, as noted
earlier, the error in estimating α becomes negligible when α = 1.00.
Overall, we can conclude that using multiple testing for identifying non-zero elements of
RN when computing α is computationally attractive, has sound theoretical properties, with
comparable performance to the estimator of the exponent of cross-sectional dependence, α,
developed in BKP.
7 An Empirical Application: identifying the weak factor
component of CAPM
In their paper BKP investigate the extent to which excess returns on the Standard & Poor’s
500 (S&P 500) securities are interconnected through the market factor by computing rolling
estimates of α, the degree of cross-sectional dependence of S&P 500 securities. According to
asset pricing theories such as the capital asset pricing model (CAPM) of Sharpe (1964) and
Lintner (1965), and arbitrage pricing (APT) of Ross (1976), such estimates of α should be close
to unity at all times. This is because both CAPM and APT assume security returns have a
common factor representation with at least one strong common factor, with the idiosyncratic
component being weakly correlated - see also the approximate factor model due to Chamberlain
(1983). Such a factor structure implies that all individual stock returns are significantly affected
by the common factor(s) and in consequence they are all pair-wise correlated with varying
degrees.
The subsequent analysis in BKP reveals that a disconnect between some asset returns and
the market factor does occur particularly at times of stock market bubbles and crashes where
these asset returns could be driven by non-fundamentals. In this paper, we focus on the
exponent of cross-sectional dependence of the residuals obtained from different versions of the
CAPM model, and provide rolling estimates of the exponent of the cross-sectional dependence
of the errors from CAPM and related APT models. This is important since under CAPM, after
allowing for the market factor, the errors can not be cross-sectionally strongly correlated. It
is therefore of interest to see if this is in fact true at all times, or if there are episodes where
market factors are not sufficient to capture all the significant interdependencies that might exist
16
across the security returns.
We update the BKP analysis and consider monthly excess returns of the securities included
in the S&P 500 index over the period from September 1989 to May 2018. We obtain estimates
of α using rolling samples of 120 months (10 years) to capture possible time variations in the
degree of cross-sectional dependence.9 Since the composition of the S&P 500 index changes
over time, we compiled returns on all 500 securities at the end of each month and included
in our analysis only those securities that had at least 10 years of data in the month under
consideration. On average, we ended up with 442 securities at the end of each month for the
10-year rolling samples. The one-month US treasury bill rate was chosen as the risk-free rate
(rft), and excess returns were computed as rit = rit − rft, where rit is the monthly return on
the ith security in the sample inclusive of dividend payments (if any).10
First, following BKP, we estimated α for excess security returns to see the extent to which
securities in S&P 500 index are fully interconnected at all times. As noted above, under CAPM
we would expect estimates of α to be close to unity. To this end we used our proposed estimator,
α, defined by (25), and computed 10-year rolling estimates αt, for t = September 1989 to May
2018 - a total of 345 estimates - with p = 0.05. To check the robustness of the estimates to the
choice of δ we computed the rolling estimates for δ = 1/2 and 1/3.11 The resultant estimates
are shown in Figure 1.12 As expected, estimates of αt, for δ = 1/2, lie below those of αt, using
δ = 1/3, but the series track each other very closely. Also the quantitative differences between
the two estimates are not that large. Specifically, all the 345 rolling estimates αt (δ = 1/2) fall
in the interval 0.82 to 0.96, whilst the corresponding estimates αt (δ = 1/3) all lie in the range
0.86 − 0.97. These estimates show a high degree of inter-linkages across individual securities,
and are very close to unity at the start and at the end of the sample, with important departures
from unity in between. Considering αt (δ = 1/2), it recorded lows around 0.86 in 1998 before
recovering temporarily and falling further to 0.84. This episode coincides with the Russian and
the Long-Term Capital crises of 1998 which originated in bond markets but rapidly transmitted
through international equity markets. The α estimates remained low, falling to 0.82 − 0.83,
around the turn of the century which saw the burst of the Dotcom bubble and 9/11 terrorist
attacks in the US. During the less volatile period of 2003 − 2007, α rose slightly to about
0.85 before a new low of 0.82 was recorded in August 2008, around the time of the sub-prime
mortgage crisis in the US, and the ensuing global financial meltdown and economic recession.13
The estimates of α gradually increase to pre-1997 levels of 0.92 − 0.93 in 2011 and have since
remained in this range to the end of our sample, May 2018.
We now turn our attention to the exponent of cross-sectional correlations of the error terms
of the CAPM model, and two well known extensions using additional Fama and French factors.
9We also consider rolling samples of size 60 months (5 years). Results for this setting are shown in the onlinesupplement - Figures 3 and 4.
10For further details of data sources and definitions see Pesaran and Yamagata (2012).11For the remaining parameters in (25) we set p = 0.05 and n = Nt(Nt − 1)/2, where Nt is the number of
securities in a given 10-year rolling window (t = 1, 2, . . . , 345).12The same estimates including their 90% confidence intervals are shown in Figures 5 and 6 of the online
supplement, where in critical value cp(n, δ) we set δ = 1/2 and 1/3, respectively.13The measured increase in α estimates during 2003 − 2007 is partly attributed to the length of the rolling
windows being set to 10 years (120 months). Opting for 5-year rolling windows (60 months) produces morepronounced increments in α estimates which is expected - see Figure 3 in the online supplement.
17
Figure 1: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of S&P500 securities’ excess returns
0.75
0.80
0.85
0.90
0.95
1.00
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/2) based on excess returns10-year rolling estimates of α ( with δ=1/3) based on excess returns
αα
Specifically, the first regression is the usual CAPM one-factor representation given by
rit − rft = ai + βi (rmt − rft) + u1i,t, for i = 1, 2, . . . , N, (35)
where rmt is the market return computed as the value-weighed returns on all NYSE, AMEX,
and NASDAQ stocks. The second and third regressions assume the following extensions to (35)
proposed by Fama and French (2004):
rit − rft = ai + β1i (rmt − rft) + β2ismbt + u2i,t, for i = 1, 2, . . . , N (36)
and
rit − rft = ai + βi (rmt − rft) + β2ismbt + β3ihmlt + u3i,t, for i = 1, 2, . . . , N, (37)
where smbt stands for average return on the three small portfolios minus the average return
on the three big portfolios formed by size, while hmlt refers to the average return on securities
with high book value to market value ratio minus the average return of securities with low book
value to market value ratio.14
As noted previously, under CAPM we would expect the errors, u1i,t, to be cross-sectionally
weakly correlated, with αu1 to be close to 1/2. But this need not be the case in reality. In
fact the introduction of FF factors, smbt and hmlt, could be viewed as an attempt to ensure
cross-sectionally weakly correlated errors for the augmented CAPM model. It is therefore of
interest to consider the estimates of α for the errors, u1i,t, u2i,t, and u3i,t, and see if they are close
to 1/2 as required by the theory. To this end, we compute 10-year rolling estimates of α based
on the pair-wise correlations of the OLS residuals u1i,t, u2i,t and u3i,t in the panel regressions
14For further details of data sources and definitions see Pesaran and Yamagata (2012).
18
(35), (36) and (37), respectively. These estimates denoted by αujt, j = 1, 2, 3 for t = September
1989 to May 2018, are shown in Figure 2.15,16
Figure 2: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of residualsfrom CAPM and its two Fama-French extensions
0.60
0.62
0.64
0.66
0.68
0.70
0.72
0.74
0.76
0.78Se
p-89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model
10-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by the SMB factor
10-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by SMB and HML factors
α
α
α
Notes: CAPM model includes excess market returns, CAPM model augmented by SMBincludes excess market returns and small minus big (SMB) firm returns, and CAPM modelaugmented by SMB and HML includes excess market returns, small minus big (SMB) firmreturns and high minus low (HML) firm returns as regressors in (35), (36) and (37), respec-tively.
As expected, estimates of α based on the residuals are smaller compared to the estimates
obtained for the securities themselves (as depicted in Figure 1). It is also interesting that all
the three estimates αu1t, αu2t and αu3t are closely clustered over the two sub-periods September
1989 to September 1997, and February 2011 to May 2018, suggesting that the standard CAPM
model provides an adequate characterisation of the cross-sectional correlations of securities,
and the additional FF factors are not required in these sub-periods. It is also worth nothing
that, over these two sub-periods, estimates of α fall in the narrow range of 0.63-0.71 which are
sufficiently small and support CAPM as an adequate model for characterising cross-correlations
of S&P 500 security returns. In contrast, the estimates αu1t, αu2t and αu3t tend to diverge over
the period from October 1997 to January 2011, and more importantly they all start to rise
sharply, suggesting important departures from the basic CAPM model. Using only the market
15We set p = 0.05 and δ = 1/2 when estimating α for the residuals, since a priori we would expect the truevalue of α for the errors of CAPM models to be close to 1/2. See the discussion in Section 6.
16The same estimates including their 90% confidence intervals are shown in Figures 7, 8 and 9 of the onlinesupplement, respectively.
19
factor, as in (35), results in αu1t jumping to levels around 0.74 − 0.76. Adding smbt to (35)
reduces the α estimates of the resulting residuals to 0.69−0.73, suggesting that the size portfolio
does have some influence on individual security returns during this period. Adding the second
FF factor (as in (37)), further reduces the estimates of α to the range 0.66−0.68. These results
are also in line with the sharp drop in the estimates of α we reported for the excess returns
during the period 1998 − 2010, and provide further evidence in favour of the argument that
the presence of factors other than the market factor, namely smbt and hmlt, tend to become
relevant during periods of financial crises and turmoils.
8 Conclusions
Cross-sectional dependence and the extent to which it occurs in large multivariate data sets is
of great interest for a variety of economic, econometric and financial analyses. Such analyses
vary widely. Examples include the effects of idiosyncratic shocks on aggregate macroeconomic
variables, the extent to which financial risk can be diversified by investing in disparate assets
or asset classes and the performance of standard estimators such as principal components when
applied to data sets with unknown collinearity structures. A common characteristic of such
analyses is the need to quantify the degree of cross-sectional dependence, especially when it is
prevalent enough to materially affect the outcome of the analysis.
In this paper we generalize the work of Bailey et al. (2016) by proposing a method of
measuring the extent of inter-connections in the residuals of large panel data sets in terms of
a single parameter, α. We refer to this as the exponent of cross-sectional dependence of the
residuals. We show that this exponent can be used to characterize the degree of sparsity of
correlation matrices, or the prevalence of factors in multi-factor representations routinely used
in economic and financial analysis. We propose a simple consistent estimator of the cross-
sectional exponent and derive the rate at which it approaches its true value. We also propose
a resampling procedure for the construction of confidence bounds around the estimator of α.
A detailed Monte Carlo study suggests that the proposed estimator has desirable small
sample properties especially when α > 3/4. We apply our measure to the widely analysed
Standard & Poor’s 500 data set. We find that for individual securities in S&P 500 index, the
10-year rolling estimates of cross-sectional exponents are sufficiently close to unity over the two
sub-periods 1989− 1997 and 2011− 2018, but not during the intervening period 1998− 2010,
when markets have been subject to a number of financial turmoils, starting with the LTCM crisis
and the Dotcom bubble, and ending with the credit crunch of 2007− 2008. These results carry
over when we consider the cross-sectional dependence of errors from the CAPM model and its
multi-factor extensions using Fama-French factors. Estimates of α based on the residuals from
the CAPM model lend support to CAPM during the sub-periods 1989− 1997 and 2011− 2018,
but not when we consider the period 1998− 2010.
20
References
Bailey, N., G. Kapetanios, and M. H. Pesaran (2016). Exponent of cross-sectional dependence: Estimation andinference. Journal of Applied Econometrics 31 (6), 929–960.
Bailey, N., M. H. Pesaran, and L. V. Smith (2018). A multiple testing approach to the regularisation of samplecorrelation matrices. Journal of Econometrics: https://doi.org/10.1016/j.jeconom.2018.10.006 .
Bickel, P. J. and E. Levina (2008). Covariance regularization by thresholding. The Annals of Statistics 36,2577–2604.
Bonferroni, C. (1935). Il calcolo delle assicurazioni su gruppi di teste. Studi in Onore del Professore SalvatoreOrtu Carboni, Rome, Italy , 13–60.
Cai, T. and W. Liu (2011). Adaptive thresholding for sparse covariance matrix estimation. Journal of AmericanStatistical Association 106, 672–684.
Cai, T. and H. Zhou (2011). Minimax estimation of large covariance matrices under `1 norm (with discussion).Statistica Sinica 22, 1319–1378.
Cai, T. and H. Zhou (2012). Optimal rates of convergence for sparse covariance matrix estimation. The Annalsof Statistics 40, 2389–2420.
Chamberlain, G. (1983). Funds, factors and diversification in arbitrage pricing theory. Econometrica 51 (5),1305–1323.
Chudik, A., G. Kapetanios, and M. H. Pesaran (2018). A one covariate at a time, multiple testing approach tovariable selection in high-dimensional linear regression models. Econometrica 86 (4), 1479–1512.
Chudik, A., M. H. Pesaran, and E. Tosetti (2011). Weak and strong cross section dependence and estimationof large panels. The Econometrics Journal 14 (1), C45–C90.
Fama, E. F. and K. R. French (2004). The capital asset pricing model: Theory and evidence. Journal ofEconomic Perspectives 18, 25–46.
Fan, J., Y. Liao, and M. Mincheva (2011). High dimensional covariance matrix estimation in approximate factormodels. The Annals of Statistics 39, 3320–3356.
Fan, J., Y. Liao, and M. Mincheva (2013). Large covariance estimation by thresholding principal orthogonalcomplements. Journal of Royal Statistical Society, Series B 75, 1–44.
Goncalves, S. and P. Perron (2018). Bootstrapping factor models with cross-sectional dependence. Available athttps://papyrus.bib.umontreal.ca/xmlui/handle/1866/20818.
Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6,65–70.
Huang, J., N. Liu, M. Pourahmadi, and L. Liu (2006). Covariance matrix selection and estimation via penalisednormal likelihood. Biometrika 93, 85–98.
Kapetanios, G. (2008). A bootstrap procedure for panel datasets with many cross-sectional units. The Econo-metrics Journal 11 (2), 377–395.
Karoui, N. E. (2008). Operator norm consistent estimation of large dimensional sparse covariance matrices. TheAnnals of Statistics 36, 2717–2756.
Ledoit, O. and M. Wolf (2003). Improved estimation of the covariance matrix of stock returns with an applicationto portfolio selection. Journal of Empirical Finance 10, 603–621.
Ledoit, O. and M. Wolf (2004). A well-conditioned estimator for large-dimensional covariance matrices. Journalof Multivariate Analysis 88, 365–411.
21
Lintner, J. (1965). The valuation of risk assets and the selection of risky investments in stock portfolios andcapital budgets. Review of Economics and Statistics 47 (1), 13–37.
Pesaran, M. H. and T. Yamagata (2012). Testing capm with a large number of assets. Working Paper, Universityof Cambridge: revised Jan 2018 .
Ross, S. A. (1976). The arbitrage theory of capital asset pricing. Journal of Economic Theory 13, 341–360.
Rothman, A. J., E. Levina, and J. Zhu (2009). Generalised thresholding of large covariance matrices. Journalof American Statistical Association 104, 177–186.
Sharpe, W. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal ofFinance 19 (3), 425–442.
Stein, C. (1956). Inadmissibility of the usual estimator for the mean of a multivariate normal distribution.Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Univ. CaliforniaPress, Berkeley , 197–206.
Wang, Y. and H. Zhou (2010). Vast volatility matrix estimation for high-frequency financial data. The Annalsof Statistics 38, 943–978.
22
Appendices
Appendix I: Statement of Lemmas
Lemma 1 Consider the panel data regression model (12) and suppose that Assumptions 1-4 hold. Then,
supi,j
Pr(|∑Tt=1zitzjt − E (zitzjt|Ωi,j,t) | >
√Tcp (n, δ)
)≤ sup
i,jPr(|∑Tt=1ξitξjt − E (ξisξjs|Ωi,j,t) | > (1− π)
√Tcp (n, δ)
)+ exp
(−C0T
C1),
for some C0, C1 > 0, where zit = eit
(T−1e′iMiei)1/2 .
Lemma 2 Let
W 0NT = N−2α
N∑i 6=j
I
(∣∣∣∣∣T−1T∑t=1
zitzjt
∣∣∣∣∣ > cp(n, δ)√T|ρij = 0
)where zit = eit
(T−1e′iMiei)1/2 . Under Assumptions 1-4,
E(W 0NT
)= O
(N2(1−α−κδ)
)+O
(N2(1−α) exp
(−C0T
C1)),
for any 0 < κ < 1, and some C0, C1 > 0, where δ is defined below (23).
Lemma 3 Let
W 1NT = N−2α
N∑i 6=j
I
(∣∣∣∣∣T−1T∑t=1
zitzjt
∣∣∣∣∣ ≤ cp(n, δ)√T|ρij 6= 0
)Under Assumptions 1-4, and as long as N = o (exp(T )), infi,j |E(ρij)| > 0,
E(W 1NT
)= O
(exp
(−C0T
C1)),
for some C0, C1 > 0, where δ is defined below (23) and zit = eit
(T−1e′iMiei)1/2 .
Appendix II: Proofs of Lemmas
Proof of Lemma 1
Recall ai,t = (ai,t1, . . . , ai,tT )′
= Xi(X′iXi)
−1xit. Then, using (15),
T∑t=1
zitzjt =
T∑t=1
(ξit +
T∑s=1
ai,tsξis
)(ξjt +
T∑s=1
aj,tsξjs
)=
T∑t=1
ξitξjt +
T∑t=1
ξit
(T∑s=1
aj,tsξjs
)+
T∑t=1
ξjt
(T∑s=1
ai,tsξis
)+
T∑t=1
(T∑s=1
ai,tsξis
)(T∑
s′=1
aj,ts′ξjs′
)
=
4∑i=1
Ai.
23
We focus on A4. A2 and A3 can be treated similarly. We have
T∑t=1
(T∑s=1
ai,tsξis
)(T∑
s′=1
aj,ts′ξjs′
)=
T∑s=1
T∑s′=1
ξisξjs′T∑t=1
ai,tsaj,ts′
T∑s=1
T∑s′=1
ξisξjs′ =
T∑s=1
ξisξjs +
T∑s=1
ξis
T∑s6=s′,s′=1
ξjs′
Further,
T∑s=1
T∑s′=1
ξisξjs′T∑t=1
ai,tsaj,ts′ =
T∑s=1
ξisξjs
(T∑t=1
ai,tsaj,ts′
)+
T∑s=1
ξis
T∑s 6=s′,s′=1
ξjs′
( T∑t=1
ai,tsaj,ts′
).
Define
aT,ij,ss′ = TT∑t=1
ai,tsaj,ts′ .
By Assumption 2, supT,i,j,s,s′ aT,ij,ss′ < K <∞. Further, define
ξ1,ijs = ξisξjsaT,ij,ss′
and
ξ2,ijs = ξis
1√T
T∑s6=s′,s′=1
ξjs′
aT,ij,ss′ .
Then,T∑s=1
ξisξjs
(T∑t=1
ai,tsaj,ts′
)+
T∑s=1
ξis
T∑s6=s′,s′=1
ξjs′
( T∑t=1
ai,tsaj,ts′
)=
1
T
T∑s=1
ξ1,ijs +1√T
T∑s=1
ξ2,ijs.
It can be easily seen that ξ1,ijs − E(ξ1,ijs|Ωi,j,t
)and ξ2,ijs − E
(ξ2,ijs|Ωi,j,t
)are martingale difference
series with finite variances, and that if ρij = 0 then E(ξ1,ijs|Ωi,j,t
)= E
(ξ2,ijs|Ωi,j,t
)= 0. Define cij,t =
E (zitzjt|Ωi,j,t). Then, using Lemma A3 of Chudik et al. (2018), it easily follows that
supi,j
Pr(|∑Tt=1zitzjt − cij,t| >
√Tcp (n, δ)
)≤ sup
i,jPr(|∑Tt=1ξitξjt − E (ξisξjs|Ωi,j,t) | > (1− π)
√Tcp (n, δ)
)+ exp
(−C0T
C1),
for some C0, C1 > 0.
Proof of Lemma 2
Note that W 0NT can be written as
W 0NT = N−2α
N∑i 6=j
I
(∣∣∣∣∣T−1T∑t=1
zitzjt
∣∣∣∣∣ > cp (n, δ)√T|ρij = 0
)
Since by assumption ξit and ξjt are distributed independently when ρij = 0, it also follows that
E (zitzjt |Ωi,j,t ) = E (ξitξjt |Ωi,j,t ) + E (ξitqjt |Ωi,j,t ) + E (ξjtqit |Ωi,j,t ) + E (qitqjt |Ωi,j,t )= 0,
24
and hence zitzjt,Ωi,j,t , is a zero mean martingale difference sequence, where Ωi,j,t = Ωi,t ∪ Ωj,t. Then, wenote that zitzjt is a normalised process since zit = eit
(T−1e′iei)1/2 . Then, noting Lemma 1, using Chudik et al.
(2018), and, in particular, their Lemmas A3 (which provides a martingale difference exponential probabilityinequality under (26)), A4 (which handles exponential probability tails for products of random variables) and
A9 (which handles the normalisation by(T−1e′iei
)1/2), we have for any 0 < π < 1, any bounded sequence,
dT > 0, and some C0, C1 > 0,
supi,j
Pr(|∑Tt=1zitzjt| >
√Tcp (n, δ) |ρij = 0
)≤ exp
[− (1− π)2cp (n, δ)
2
2(1 + dT )
]+ exp
(−C0T
C1). (38)
Note that (1−π)2
1+dT< 1, but can be made arbitrarily close to 1, and that, by Lemma A2 of Chudik et al. (2018),
exp[−bc2p (n, δ)
]= O
(n−2bδ
). Then, for any 0 < κ < 1, and some C0, C1 > 0,
supi.j
Pr(|∑Tt=1zitzjt| >
√Tcp (n, δ) |ρij = 0
)= O(n−κδ)+O
(exp
(−C0T
C1))
= O(N−2κδ)+O(exp
(−C0T
C1)).
Then,for some some C0, C1 > 0, any 0 < κ < 1, and if N = O(T d)
E(W 0NT ) = N−2α
N∑i 6=j
Pr
(∣∣∣∣∣T−1T∑t=1
zitzjt
∣∣∣∣∣ > cp (n, δ)√T|ρij = 0
)= O
(N2(1−α−κδ)
)+O
(N2(1−α) exp
(−C0T
C1))
=
O(T 2d(1−α−κδ)
)+O
(exp
[2d(1− α) ln(T )− C0T
C1]).
Proof of Lemma 3
We need to derive hN,T in
Pr
[∣∣∣∣∣T∑t=1
zitzjt
∣∣∣∣∣ ≤ √Tcp (n, δ) |ρij 6= 0
]= O (hN,T ) .
Let cij,t = E (zitzjt|Ωi,j,t), and note the inequality
Pr (|X +B| ≤ C) ≤ Pr (|X| > |B| − C) ,
where X is a random variable, B and C are constants, and |B| ≥ C > 0. Then, for some 0 < π < 1
Pr
[∣∣∣∣∣T∑t=1
(zitzjt − cij,t) + cij,t − E (cij,t) + E (cij,t)
∣∣∣∣∣ ≤ √Tcp (n, δ) |ρij 6= 0
]
≤ Pr
[∣∣∣∣∣T∑t=1
(zitzjt − cij,t) + cij,t − E (cij,t)
∣∣∣∣∣ >∣∣∣∣∣T∑t=1
E(cij,t)
∣∣∣∣∣−√Tcp (n, δ) |ρij 6= 0
]≤
≤ Pr
[∣∣∣∣∣T∑t=1
(zitzjt − cij,t)
∣∣∣∣∣ > (1− π)
[∣∣∣∣∣T∑t=1
E(cij,t)
∣∣∣∣∣−√Tcp (n, δ)
]|ρij 6= 0
]+ (39)
Pr
[∣∣∣∣∣T∑t=1
cij,t − E (cij,t)
∣∣∣∣∣ > π
[∣∣∣∣∣T∑t=1
E(cij,t)
∣∣∣∣∣−√Tcp (n, δ)
]|ρij 6= 0
](40)
But, by (28), (40) is bounded by exp(−C0T
C1)
for some C0, C1 > 0. We consider
Pr
[∣∣∣∣∣T∑t=1
(zitzjt − cij,t)
∣∣∣∣∣ > (1− π)
[∣∣∣∣∣T∑t=1
E(cij,t)
∣∣∣∣∣−√Tcp (n, δ)
]|ρij 6= 0
]
But, by (20) of Assumption 1, limN,T→∞
√Tcp(n,δ)∑Tt=1 E(cij,t)
= 0. Therefore, using again Lemma 1 and (38) of Lemma
2, we have
supij,
Pr
[∣∣∣∣∣T∑t=1
(zitzjt − cij,t)
∣∣∣∣∣ > (1− π)
[∣∣∣∣∣T∑t=1
E(cij,t)
∣∣∣∣∣−√Tcp (n, δ)
]|ρij 6= 0
]≤ exp (−CT ) ,
for some C > 0, proving the result.
25
Appendix III: Proof of Theorem 1
We prove that α converges to α under our assumptions of exogenous regressors and symmetrically distributederrors. We first note that
α =1
2
ln(
τ ′∆τN2α N
2α)
lnN=
1
2
ln(
τ ′∆τN2α
)lnN
+ α,
(lnN) (α− α) =1
2ln
(τ ′∆τ
N2α− 1 + 1
), (41)
and
2 (lnN) (α− α) =τ ′∆τ
N2α− 1 +O
[τ ′∆τN2α
− 1
]2 .
Further
N−2ατ ′∆τ = N−2αN∑i 6=j
I
(|ρij | >
cp(n, δ)√T
),
and
N−2ατ ′∆τ − 1 = N−2αN∑i6=j
I
(|ρij | >
cp(n, δ)√T|ρij = 0
)(42)
+
N−2αN∑i 6=j
I
(|ρij | >
cp(n, δ)√T|ρij 6= 0
)− 1
= W 0
NT +W 1NT +O
(N1−2α
).
We now have
N−2αN∑i 6=j
I
(|ρij | >
cp(n, δ)√T|ρij 6= 0
)− 1
= N−2α(N2α −Nα)− 1−N−2αN∑i 6=j
I
(|ρij | ≤
cp(N, δ)√T|ρij 6= 0
)
= −N−2αN∑i 6=j
I
(|ρij | ≤
cp(n, δ)√T|ρij 6= 0
)−N−α.
Hence,
N−2ατ ′∆τ − 1 = N−2αN∑i6=j
I
(|ρij | >
cp(n, δ)√T|ρij = 0
)−
N−2αN∑i 6=j
I
(|ρij | ≤
cp(n, δ)√T|ρij 6= 0
)−N−α,
which we write more compactly as
N−2ατ ′∆τ − 1 = W 0NT − W 1
NT +Op(N−α
),
where
W 0NT = N−2α
N∑i 6=j
I
(|ρij | >
cp(n, δ)√T|ρij = 0
),
W 1NT = N−2α
N∑i 6=j
I
(|ρij | ≤
cp(n, δ)√T|ρij 6= 0
).
26
It is worth noting that E∣∣W 0
NT
∣∣ = E(W 0NT
), and E
∣∣∣W 1NT
∣∣∣ = E(W 1NT
), and hence,
E∣∣∣N−2ατ ′∆τ − 1
∣∣∣ ≤ E (W 0NT
)+ E
(W 1NT
)+O
(N−α
)+O
(N1−2α
).
Lemmas 2-3 provide bounds for E(W 0NT
)and E
(W 1NT
)proving the result.
27
Tab
leA
1a:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
cp
anel
dat
am
od
elC
ross
corr
elati
on
sare
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
00.
204
-0.6
09-0
.584
-0.3
55-0
.289
-0.0
01-0
.259
-0.2
79-0
.094
-0.0
1610
010
00.
634
-0.3
24-0
.405
-0.2
45-0
.218
0.04
2-0
.230
-0.2
61-0
.083
-0.0
0920
00.
066
-0.4
26-0
.156
-0.3
02-0
.254
-0.1
01-0
.165
-0.0
80-0
.058
-0.0
2520
00.
395
-0.2
19-0
.034
-0.2
25-0
.205
-0.0
70-0
.143
-0.0
65-0
.046
-0.0
1550
00.
090
-0.1
20-0
.162
-0.1
44-0
.149
-0.0
35-0
.042
-0.0
46-0
.050
-0.0
4150
00.
332
0.02
1-0
.079
-0.0
93-0
.115
-0.0
09-0
.021
-0.0
29-0
.034
-0.0
2520
010
00.
243
-0.5
84-0
.557
-0.3
28-0
.267
0.01
7-0
.240
-0.2
62-0
.078
0.00
020
010
00.
686
-0.2
91-0
.372
-0.2
11-0
.201
0.05
6-0
.218
-0.2
51-0
.074
0.00
020
00.
118
-0.3
80-0
.120
-0.2
71-0
.226
-0.0
74-0
.139
-0.0
57-0
.035
0.00
020
00.
470
-0.1
610.
003
-0.1
98-0
.184
-0.0
52-0
.127
-0.0
51-0
.033
0.00
050
00.
158
-0.0
64-0
.114
-0.1
01-0
.106
0.00
7-0
.001
-0.0
06-0
.011
0.00
050
00.
418
0.08
1-0
.035
-0.0
59-0
.084
0.01
90.
005
-0.0
04-0
.010
0.00
050
010
00.
277
-0.5
56-0
.549
-0.3
25-0
.261
0.01
9-0
.239
-0.2
62-0
.078
0.00
050
010
00.
718
-0.2
51-0
.356
-0.2
10-0
.191
0.05
9-0
.216
-0.2
51-0
.073
0.00
020
00.
157
-0.3
67-0
.110
-0.2
65-0
.223
-0.0
73-0
.138
-0.0
56-0
.035
0.00
020
00.
519
-0.1
390.
020
-0.1
88-0
.179
-0.0
49-0
.125
-0.0
50-0
.033
0.00
050
00.
189
-0.0
47-0
.105
-0.0
97-0
.104
0.00
90.
000
-0.0
06-0
.011
0.00
050
00.
470
0.10
8-0
.021
-0.0
53-0
.080
0.02
10.
006
-0.0
04-0
.010
0.00
0T
NR
MSE
TN
RM
SE
100
100
0.31
10.
633
0.59
60.
364
0.29
60.
051
0.26
40.
282
0.10
20.
041
100
100
0.71
40.
407
0.43
80.
269
0.23
30.
072
0.23
50.
263
0.08
70.
026
200
0.14
50.
438
0.17
40.
308
0.26
10.
124
0.17
10.
093
0.07
00.
053
200
0.43
60.
255
0.09
40.
233
0.21
10.
089
0.14
70.
072
0.05
30.
033
500
0.11
80.
137
0.17
70.
157
0.16
50.
073
0.07
80.
076
0.07
70.
079
500
0.34
50.
068
0.09
80.
104
0.12
40.
045
0.04
90.
050
0.05
20.
052
200
100
0.34
70.
611
0.57
10.
338
0.27
20.
042
0.24
20.
263
0.07
90.
000
200
100
0.76
90.
381
0.41
00.
243
0.21
70.
081
0.22
20.
252
0.07
50.
000
200
0.18
30.
391
0.13
30.
274
0.22
80.
076
0.14
00.
057
0.03
50.
000
200
0.50
90.
210
0.08
20.
207
0.18
80.
058
0.12
80.
052
0.03
30.
000
500
0.17
30.
075
0.11
60.
102
0.10
70.
010
0.00
40.
007
0.01
10.
000
500
0.42
90.
100
0.05
00.
063
0.08
50.
021
0.00
80.
006
0.01
10.
000
500
100
0.37
40.
584
0.56
30.
337
0.26
80.
043
0.24
00.
263
0.07
80.
000
500
100
0.79
80.
356
0.39
70.
245
0.21
10.
083
0.22
00.
252
0.07
50.
000
200
0.21
80.
379
0.12
60.
268
0.22
50.
075
0.13
80.
057
0.03
50.
000
200
0.56
00.
193
0.08
90.
199
0.18
40.
056
0.12
60.
051
0.03
30.
000
500
0.20
30.
063
0.10
80.
098
0.10
40.
011
0.00
40.
007
0.01
10.
000
500
0.48
00.
124
0.04
20.
057
0.08
20.
023
0.00
90.
005
0.01
00.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.50
70.
288
-0.0
13-0
.003
-0.0
720.
127
-0.1
77-0
.232
-0.0
70-0
.005
100
100
3.07
11.
418
0.71
20.
452
0.21
20.
289
-0.0
78-0
.181
-0.0
50-0
.002
200
1.43
30.
445
0.35
40.
011
-0.0
660.
010
-0.0
97-0
.038
-0.0
32-0
.006
200
2.97
01.
475
0.97
40.
391
0.15
70.
136
-0.0
28-0
.004
-0.0
17-0
.003
500
1.53
30.
715
0.31
40.
123
0.00
80.
061
0.02
1-0
.001
-0.0
15-0
.009
500
3.03
71.
641
0.84
90.
421
0.17
00.
146
0.06
60.
021
-0.0
04-0
.005
200
100
1.58
50.
330
0.01
80.
030
-0.0
580.
141
-0.1
69-0
.226
-0.0
650.
000
200
100
3.15
91.
461
0.75
10.
491
0.21
80.
307
-0.0
75-0
.176
-0.0
460.
000
200
1.53
10.
529
0.40
30.
040
-0.0
460.
024
-0.0
87-0
.031
-0.0
260.
000
200
3.08
91.
572
1.03
00.
423
0.17
60.
148
-0.0
200.
001
-0.0
140.
000
500
1.68
40.
813
0.37
00.
161
0.03
20.
078
0.03
40.
009
-0.0
060.
000
500
3.21
71.
764
0.91
80.
464
0.19
60.
163
0.07
60.
028
0.00
10.
000
500
100
1.62
70.
386
0.04
10.
035
-0.0
440.
144
-0.1
67-0
.225
-0.0
640.
000
500
100
3.19
21.
517
0.77
80.
495
0.24
10.
310
-0.0
71-0
.176
-0.0
450.
000
200
1.61
30.
564
0.43
20.
054
-0.0
390.
028
-0.0
83-0
.030
-0.0
250.
000
200
3.17
51.
616
1.06
70.
441
0.18
50.
154
-0.0
160.
003
-0.0
130.
000
500
1.78
30.
867
0.39
90.
173
0.04
00.
082
0.03
70.
010
-0.0
050.
000
500
3.32
71.
827
0.95
70.
483
0.20
70.
168
0.08
00.
030
0.00
10.
000
TN
RM
SE
TN
RM
SE
100
100
1.57
10.
448
0.24
80.
170
0.14
30.
153
0.18
70.
236
0.07
50.
015
100
100
3.12
41.
492
0.79
00.
514
0.28
00.
316
0.12
30.
192
0.06
10.
008
200
1.46
10.
490
0.37
80.
098
0.09
30.
053
0.10
20.
046
0.03
60.
016
200
2.99
11.
498
0.99
30.
416
0.18
40.
153
0.05
50.
032
0.02
60.
009
500
1.54
10.
722
0.32
10.
133
0.03
70.
067
0.03
20.
019
0.02
30.
022
500
3.04
41.
646
0.85
40.
426
0.17
60.
150
0.07
00.
027
0.01
30.
013
200
100
1.65
20.
475
0.25
50.
180
0.13
70.
167
0.18
10.
230
0.06
80.
000
200
100
3.21
11.
531
0.82
60.
555
0.28
60.
335
0.12
20.
188
0.05
80.
000
200
1.55
90.
570
0.42
60.
109
0.08
20.
052
0.09
20.
037
0.02
80.
000
200
3.10
91.
598
1.04
80.
447
0.20
30.
163
0.05
30.
032
0.02
20.
000
500
1.69
20.
820
0.37
70.
167
0.04
40.
081
0.03
60.
013
0.00
70.
000
500
3.22
31.
769
0.92
30.
469
0.20
10.
166
0.07
90.
031
0.00
80.
000
500
100
1.69
30.
524
0.25
40.
191
0.13
80.
170
0.17
90.
229
0.06
80.
000
500
100
3.24
51.
589
0.85
40.
561
0.30
70.
338
0.12
10.
187
0.05
90.
000
200
1.64
10.
604
0.45
60.
116
0.08
00.
057
0.09
00.
037
0.02
80.
000
200
3.19
61.
640
1.08
50.
465
0.21
10.
171
0.05
30.
034
0.02
30.
000
500
1.79
00.
874
0.40
50.
180
0.05
10.
085
0.04
00.
014
0.00
70.
000
500
3.33
21.
832
0.96
30.
487
0.21
20.
171
0.08
30.
033
0.00
80.
000
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efirs
tNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
mat
rix
ofth
eer
rors
,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
al
valu
eu
sed
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
28
Tab
leA
1b
:B
ias
an
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
00.
198
-0.6
21-0
.594
-0.3
63-0
.295
-0.0
07-0
.261
-0.2
84-0
.099
-0.0
2110
010
00.
625
-0.3
30-0
.412
-0.2
45-0
.221
0.03
9-0
.230
-0.2
64-0
.085
-0.0
1220
00.
064
-0.4
32-0
.161
-0.3
09-0
.262
-0.1
08-0
.173
-0.0
88-0
.070
-0.0
3320
00.
398
-0.2
18-0
.035
-0.2
29-0
.211
-0.0
74-0
.148
-0.0
69-0
.053
-0.0
1950
00.
084
-0.1
23-0
.169
-0.1
58-0
.158
-0.0
46-0
.054
-0.0
59-0
.066
-0.0
4950
00.
326
0.01
8-0
.084
-0.1
02-0
.120
-0.0
16-0
.029
-0.0
37-0
.045
-0.0
3020
010
00.
255
-0.5
73-0
.561
-0.3
30-0
.266
0.01
8-0
.240
-0.2
63-0
.078
0.00
020
010
00.
687
-0.2
80-0
.376
-0.2
16-0
.200
0.05
7-0
.217
-0.2
52-0
.074
0.00
020
00.
123
-0.3
81-0
.118
-0.2
72-0
.225
-0.0
74-0
.139
-0.0
56-0
.035
0.00
020
00.
469
-0.1
600.
008
-0.1
98-0
.183
-0.0
51-0
.126
-0.0
50-0
.033
0.00
050
00.
156
-0.0
63-0
.113
-0.1
01-0
.106
0.00
7-0
.001
-0.0
07-0
.011
0.00
050
00.
417
0.08
2-0
.034
-0.0
60-0
.084
0.01
90.
005
-0.0
04-0
.010
0.00
050
010
00.
281
-0.5
50-0
.550
-0.3
22-0
.261
0.02
0-0
.240
-0.2
61-0
.078
0.00
050
010
00.
737
-0.2
46-0
.361
-0.2
05-0
.190
0.06
0-0
.216
-0.2
49-0
.073
0.00
020
00.
153
-0.3
66-0
.108
-0.2
66-0
.222
-0.0
72-0
.137
-0.0
56-0
.035
0.00
020
00.
514
-0.1
390.
022
-0.1
90-0
.179
-0.0
49-0
.125
-0.0
50-0
.033
0.00
050
00.
185
-0.0
49-0
.104
-0.0
97-0
.104
0.00
90.
000
-0.0
06-0
.011
0.00
050
00.
467
0.10
7-0
.020
-0.0
53-0
.080
0.02
10.
006
-0.0
04-0
.010
0.00
0T
NR
MSE
TN
RM
SE
100
100
0.30
20.
644
0.60
70.
375
0.30
50.
056
0.26
60.
289
0.10
80.
045
100
100
0.70
20.
411
0.44
40.
270
0.23
80.
070
0.23
50.
267
0.09
00.
027
200
0.15
30.
444
0.18
00.
316
0.27
20.
126
0.18
30.
106
0.09
40.
064
200
0.43
90.
255
0.09
60.
237
0.21
90.
088
0.15
40.
080
0.06
80.
041
500
0.11
70.
144
0.18
50.
179
0.17
40.
102
0.10
20.
102
0.10
70.
090
500
0.34
00.
073
0.10
20.
119
0.13
00.
066
0.06
70.
069
0.07
30.
060
200
100
0.35
40.
600
0.57
30.
339
0.27
20.
043
0.24
10.
264
0.07
90.
000
200
100
0.76
50.
374
0.41
30.
246
0.21
60.
082
0.22
10.
253
0.07
50.
000
200
0.19
00.
392
0.13
10.
274
0.22
60.
076
0.13
90.
057
0.03
50.
000
200
0.51
10.
208
0.08
50.
207
0.18
70.
057
0.12
70.
051
0.03
30.
000
500
0.16
90.
074
0.11
60.
102
0.10
70.
010
0.00
40.
007
0.01
10.
000
500
0.42
70.
100
0.04
90.
063
0.08
50.
021
0.00
80.
005
0.01
10.
000
500
100
0.37
50.
580
0.56
30.
332
0.26
70.
045
0.24
10.
262
0.07
90.
000
500
100
0.81
50.
360
0.40
20.
236
0.20
80.
084
0.22
00.
251
0.07
50.
000
200
0.21
30.
378
0.12
30.
270
0.22
40.
074
0.13
80.
057
0.03
50.
000
200
0.55
30.
195
0.08
90.
199
0.18
30.
056
0.12
70.
051
0.03
30.
000
500
0.19
80.
065
0.10
70.
098
0.10
40.
011
0.00
40.
007
0.01
10.
000
500
0.47
80.
123
0.04
20.
057
0.08
20.
023
0.00
90.
006
0.01
00.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.50
90.
279
-0.0
26-0
.005
-0.0
740.
125
-0.1
77-0
.233
-0.0
71-0
.006
100
100
3.07
01.
397
0.70
80.
455
0.20
80.
290
-0.0
77-0
.181
-0.0
49-0
.003
200
1.43
10.
450
0.35
90.
009
-0.0
690.
009
-0.0
99-0
.040
-0.0
35-0
.008
200
2.97
41.
486
0.98
30.
387
0.15
60.
135
-0.0
29-0
.004
-0.0
19-0
.004
500
1.53
20.
716
0.31
10.
119
0.00
70.
059
0.01
8-0
.004
-0.0
19-0
.011
500
3.03
81.
642
0.84
80.
418
0.16
90.
145
0.06
50.
020
-0.0
07-0
.006
200
100
1.56
90.
341
0.01
70.
027
-0.0
540.
142
-0.1
68-0
.227
-0.0
650.
000
200
100
3.12
21.
471
0.75
00.
487
0.23
00.
307
-0.0
73-0
.178
-0.0
470.
000
200
1.55
20.
529
0.40
70.
042
-0.0
450.
026
-0.0
85-0
.030
-0.0
260.
000
200
3.10
61.
571
1.03
50.
427
0.17
70.
150
-0.0
170.
003
-0.0
140.
000
500
1.68
60.
819
0.37
10.
158
0.03
20.
078
0.03
40.
010
-0.0
060.
000
500
3.22
01.
765
0.91
90.
462
0.19
60.
163
0.07
60.
029
0.00
10.
000
500
100
1.63
40.
384
0.04
00.
040
-0.0
400.
146
-0.1
67-0
.223
-0.0
640.
000
500
100
3.19
61.
512
0.77
80.
502
0.24
10.
310
-0.0
72-0
.174
-0.0
450.
000
200
1.60
90.
563
0.43
20.
057
-0.0
380.
029
-0.0
83-0
.030
-0.0
250.
000
200
3.17
21.
611
1.06
80.
443
0.18
90.
152
-0.0
150.
003
-0.0
130.
000
500
1.78
00.
864
0.40
10.
174
0.04
10.
082
0.03
60.
010
-0.0
050.
000
500
3.32
11.
824
0.96
10.
483
0.20
80.
168
0.07
90.
030
0.00
20.
000
TN
RM
SE
TN
RM
SE
100
100
1.57
60.
441
0.24
20.
165
0.14
40.
150
0.18
70.
238
0.07
60.
015
100
100
3.12
31.
466
0.78
70.
514
0.27
40.
315
0.12
20.
192
0.06
20.
008
200
1.45
90.
493
0.38
30.
095
0.09
80.
050
0.10
50.
049
0.04
30.
020
200
2.99
61.
509
1.00
10.
411
0.18
50.
152
0.05
60.
034
0.02
90.
012
500
1.54
10.
724
0.31
90.
130
0.03
60.
069
0.03
70.
029
0.03
30.
026
500
3.04
41.
648
0.85
30.
423
0.17
50.
149
0.07
00.
029
0.01
80.
015
200
100
1.63
20.
486
0.24
70.
175
0.13
50.
169
0.18
00.
230
0.06
90.
000
200
100
3.17
41.
540
0.82
30.
548
0.29
40.
335
0.12
00.
188
0.05
90.
000
200
1.57
90.
568
0.43
00.
109
0.08
40.
053
0.09
10.
037
0.02
80.
000
200
3.12
71.
595
1.05
50.
451
0.20
50.
166
0.05
30.
034
0.02
30.
000
500
1.69
40.
826
0.37
70.
165
0.04
50.
081
0.03
70.
013
0.00
70.
000
500
3.22
61.
771
0.92
50.
467
0.20
10.
166
0.07
90.
032
0.00
70.
000
500
100
1.70
00.
529
0.26
20.
182
0.13
40.
171
0.17
90.
228
0.06
80.
000
500
100
3.24
61.
581
0.85
60.
564
0.30
40.
337
0.12
10.
187
0.05
80.
000
200
1.63
60.
602
0.45
40.
116
0.08
10.
056
0.09
00.
037
0.02
80.
000
200
3.19
31.
635
1.08
60.
466
0.21
60.
168
0.05
40.
034
0.02
30.
000
500
1.78
70.
871
0.40
70.
180
0.05
20.
085
0.03
90.
014
0.00
70.
000
500
3.32
71.
830
0.96
50.
487
0.21
40.
171
0.08
20.
033
0.00
80.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vec
tor
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wher
eBN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
alva
lue
use
din
the
mu
ltip
lete
stin
gpro
ced
ure
show
nin
(23).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
29
Tab
leA
2a:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
cp
anel
dat
am
od
elC
ross
corr
elati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
0-0
.880
-2.3
78-2
.922
-3.0
26-3
.232
-3.1
31-3
.509
-3.6
48-3
.576
-3.5
9810
010
0-0
.092
-1.5
33-2
.041
-2.1
19-2
.328
-2.2
20-2
.609
-2.7
63-2
.699
-2.7
3820
0-1
.221
-2.4
80-2
.800
-3.2
67-3
.464
-3.4
34-3
.595
-3.5
91-3
.658
-3.7
1420
0-0
.542
-1.7
21-1
.988
-2.4
35-2
.620
-2.5
90-2
.750
-2.7
54-2
.824
-2.8
8350
0-1
.501
-2.6
58-3
.266
-3.5
81-3
.780
-3.7
79-3
.859
-3.9
34-4
.011
-4.0
5750
0-0
.910
-1.9
52-2
.492
-2.7
75-2
.953
-2.9
48-3
.026
-3.0
98-3
.175
-3.2
2220
010
00.
705
-0.0
90-0
.160
0.12
00.
009
0.20
5-0
.145
-0.2
99-0
.242
-0.3
0420
010
01.
207
0.29
40.
129
0.34
60.
192
0.36
2-0
.007
-0.1
77-0
.131
-0.2
0720
00.
375
-0.0
680.
134
-0.0
86-0
.146
-0.0
63-0
.191
-0.2
06-0
.269
-0.3
2620
00.
784
0.23
60.
360
0.09
40.
001
0.06
6-0
.071
-0.0
97-0
.166
-0.2
2850
00.
219
-0.0
24-0
.086
-0.1
58-0
.206
-0.1
53-0
.212
-0.2
67-0
.325
-0.3
6850
00.
541
0.20
90.
094
-0.0
12-0
.079
-0.0
38-0
.104
-0.1
63-0
.222
-0.2
6650
010
01.
043
0.29
50.
247
0.56
30.
433
0.62
50.
264
0.08
40.
114
-0.0
0150
010
01.
463
0.58
10.
426
0.67
20.
502
0.66
60.
289
0.09
80.
121
0.00
020
00.
683
0.31
30.
550
0.32
60.
254
0.32
90.
203
0.16
10.
087
-0.0
0120
01.
031
0.52
90.
674
0.40
10.
298
0.35
50.
218
0.16
90.
091
0.00
050
00.
530
0.35
50.
331
0.24
90.
205
0.24
60.
176
0.11
70.
056
-0.0
0150
00.
799
0.50
40.
412
0.29
40.
229
0.25
90.
184
0.12
10.
058
0.00
0T
NR
MSE
TN
RM
SE
100
100
1.25
22.
688
3.30
13.
469
3.69
93.
623
3.97
44.
094
4.02
94.
051
100
100
0.82
61.
877
2.41
22.
542
2.75
22.
671
3.02
33.
155
3.09
73.
137
200
1.45
52.
748
3.16
73.
655
3.86
83.
859
4.00
24.
009
4.07
04.
130
200
0.89
42.
002
2.35
12.
795
2.98
72.
976
3.11
23.
125
3.18
93.
251
500
1.66
22.
899
3.58
03.
938
4.15
34.
176
4.26
04.
336
4.41
14.
456
500
1.11
52.
198
2.79
13.
105
3.29
23.
306
3.38
63.
459
3.53
33.
580
200
100
0.80
30.
377
0.36
90.
346
0.29
80.
354
0.31
20.
405
0.35
90.
399
200
100
1.27
20.
451
0.31
40.
438
0.30
50.
426
0.21
70.
275
0.23
90.
281
200
0.45
70.
291
0.30
70.
274
0.29
60.
251
0.31
50.
318
0.36
70.
415
200
0.82
40.
336
0.42
10.
223
0.19
90.
198
0.20
60.
208
0.25
10.
299
500
0.30
30.
241
0.26
20.
299
0.32
80.
295
0.32
60.
366
0.41
80.
456
500
0.57
10.
283
0.21
40.
196
0.21
10.
197
0.21
60.
252
0.30
00.
337
500
100
1.06
80.
337
0.27
20.
571
0.44
20.
631
0.27
90.
116
0.13
60.
002
500
100
1.49
50.
621
0.45
10.
682
0.51
00.
672
0.30
30.
127
0.14
10.
001
200
0.69
70.
325
0.55
30.
329
0.25
70.
332
0.20
80.
166
0.09
30.
001
200
1.04
80.
542
0.67
80.
405
0.30
10.
357
0.22
30.
174
0.09
70.
001
500
0.53
40.
357
0.33
20.
250
0.20
60.
246
0.17
70.
118
0.05
70.
001
500
0.80
40.
507
0.41
40.
294
0.23
00.
260
0.18
40.
122
0.05
90.
001
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.12
1-0
.410
-1.0
23-1
.161
-1.4
37-1
.372
-1.7
89-1
.971
-1.9
26-1
.985
100
100
2.94
71.
111
0.20
8-0
.136
-0.5
73-0
.609
-1.0
93-1
.323
-1.3
12-1
.401
200
0.92
6-0
.377
-0.7
70-1
.305
-1.5
54-1
.571
-1.7
58-1
.790
-1.8
71-1
.940
200
2.75
01.
070
0.35
2-0
.393
-0.7
87-0
.898
-1.1
40-1
.210
-1.3
13-1
.396
500
0.90
5-0
.329
-1.0
08-1
.398
-1.6
34-1
.672
-1.7
75-1
.859
-1.9
39-1
.991
500
2.68
81.
017
0.01
4-0
.590
-0.9
50-1
.065
-1.2
10-1
.316
-1.4
07-1
.466
200
100
2.12
00.
962
0.59
00.
671
0.42
50.
538
0.13
4-0
.064
-0.0
38-0
.131
200
100
3.64
92.
070
1.33
81.
167
0.75
80.
763
0.29
30.
047
0.03
9-0
.080
200
1.90
00.
990
0.85
50.
438
0.24
60.
250
0.08
00.
027
-0.0
56-0
.130
200
3.43
62.
034
1.51
20.
863
0.52
00.
428
0.20
30.
113
0.00
9-0
.082
500
1.87
01.
040
0.62
20.
337
0.17
30.
156
0.05
9-0
.017
-0.0
85-0
.135
500
3.39
42.
001
1.20
00.
682
0.38
40.
290
0.15
00.
052
-0.0
29-0
.088
500
100
2.30
51.
160
0.79
70.
898
0.64
10.
748
0.33
80.
123
0.13
10.
000
500
100
3.78
52.
210
1.48
71.
322
0.90
50.
903
0.42
90.
171
0.15
00.
000
200
2.07
31.
186
1.05
70.
633
0.43
40.
431
0.26
10.
191
0.09
90.
000
200
3.57
12.
177
1.65
30.
999
0.65
00.
553
0.32
80.
224
0.11
20.
000
500
2.06
01.
231
0.81
50.
514
0.34
80.
322
0.21
60.
137
0.06
40.
000
500
3.55
62.
152
1.34
60.
811
0.51
00.
406
0.25
90.
157
0.07
10.
000
TN
RM
SE
TN
RM
SE
100
100
1.34
90.
991
1.45
81.
608
1.84
11.
798
2.15
32.
306
2.26
72.
325
100
100
3.02
91.
332
0.83
20.
857
1.04
91.
073
1.42
91.
611
1.60
31.
685
200
1.10
20.
867
1.21
01.
653
1.88
31.
906
2.06
12.
097
2.17
02.
241
200
2.79
81.
230
0.79
20.
864
1.13
21.
222
1.41
01.
477
1.56
61.
648
500
1.01
70.
766
1.32
51.
693
1.91
61.
966
2.06
32.
144
2.22
02.
272
500
2.71
11.
136
0.64
30.
938
1.22
41.
335
1.46
51.
562
1.64
71.
703
200
100
2.16
91.
026
0.64
90.
711
0.46
90.
568
0.21
60.
175
0.15
30.
187
200
100
3.68
82.
111
1.37
11.
190
0.78
00.
780
0.32
60.
140
0.12
00.
121
200
1.92
21.
014
0.87
30.
463
0.28
40.
281
0.15
50.
125
0.13
50.
181
200
3.45
32.
049
1.52
30.
874
0.53
40.
440
0.22
80.
145
0.08
90.
119
500
1.87
81.
049
0.63
30.
357
0.20
80.
192
0.12
40.
111
0.14
40.
180
500
3.39
92.
005
1.20
50.
688
0.39
40.
301
0.16
90.
093
0.08
80.
122
500
100
2.34
51.
200
0.82
40.
910
0.65
10.
755
0.35
00.
149
0.15
00.
001
500
100
3.82
32.
246
1.51
31.
336
0.91
70.
911
0.44
10.
192
0.16
80.
000
200
2.09
11.
200
1.06
30.
638
0.43
70.
434
0.26
60.
195
0.10
50.
000
200
3.58
62.
190
1.66
11.
005
0.65
50.
556
0.33
20.
228
0.11
70.
000
500
2.06
51.
234
0.81
70.
515
0.34
90.
322
0.21
70.
138
0.06
50.
000
500
3.56
02.
155
1.34
80.
813
0.51
10.
407
0.26
00.
158
0.07
30.
000
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
and
[Nαβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
,f jt
anduit∼IIDN
(0,1
),v ij∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
an
d
(33)
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=20
00.
30
Tab
leA
2b
:B
ias
an
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
0-1
.016
-2.5
71-3
.183
-3.3
19-3
.568
-3.4
79-3
.838
-3.9
98-3
.922
-3.9
5210
010
0-0
.216
-1.6
91-2
.254
-2.3
54-2
.602
-2.5
09-2
.879
-3.0
47-2
.978
-3.0
2420
0-1
.320
-2.6
42-3
.032
-3.5
58-3
.745
-3.7
63-3
.918
-3.9
25-4
.003
-4.0
5720
0-0
.627
-1.8
58-2
.175
-2.6
71-2
.846
-2.8
56-3
.011
-3.0
25-3
.104
-3.1
6350
0-1
.610
-2.8
49-3
.517
-3.8
90-4
.107
-4.1
24-4
.222
-4.3
04-4
.388
-4.4
3950
0-1
.007
-2.1
19-2
.705
-3.0
35-3
.227
-3.2
35-3
.330
-3.4
09-3
.490
-3.5
4320
010
00.
689
-0.1
09-0
.179
0.08
8-0
.030
0.16
6-0
.181
-0.3
29-0
.272
-0.3
3620
010
01.
194
0.28
00.
115
0.32
90.
162
0.33
2-0
.035
-0.1
99-0
.153
-0.2
3020
00.
357
-0.0
890.
115
-0.1
20-0
.177
-0.0
95-0
.227
-0.2
40-0
.301
-0.3
6020
00.
774
0.22
00.
345
0.06
8-0
.022
0.04
3-0
.099
-0.1
22-0
.189
-0.2
5350
00.
209
-0.0
33-0
.095
-0.1
70-0
.215
-0.1
65-0
.225
-0.2
81-0
.337
-0.3
8050
00.
538
0.20
40.
087
-0.0
22-0
.085
-0.0
47-0
.114
-0.1
73-0
.230
-0.2
7550
010
01.
040
0.29
40.
248
0.56
50.
436
0.62
70.
261
0.08
10.
115
-0.0
0150
010
01.
466
0.58
30.
430
0.67
60.
504
0.67
00.
286
0.09
50.
121
0.00
020
00.
685
0.31
30.
548
0.32
40.
253
0.33
00.
202
0.16
20.
086
-0.0
0120
01.
031
0.52
80.
672
0.39
90.
298
0.35
50.
217
0.17
00.
089
0.00
050
00.
527
0.35
30.
330
0.24
90.
204
0.24
60.
176
0.11
70.
055
-0.0
0150
00.
797
0.50
20.
412
0.29
40.
229
0.25
90.
183
0.12
20.
057
0.00
0T
NR
MSE
TN
RM
SE
100
100
1.36
02.
876
3.54
93.
762
4.03
13.
981
4.30
14.
455
4.37
84.
411
100
100
0.87
12.
032
2.62
12.
783
3.03
82.
976
3.29
63.
453
3.38
33.
430
200
1.53
62.
902
3.39
83.
954
4.17
04.
208
4.35
84.
367
4.44
54.
497
200
0.95
52.
133
2.53
63.
040
3.23
33.
258
3.40
43.
418
3.49
73.
553
500
1.76
93.
101
3.84
94.
273
4.51
24.
554
4.65
94.
740
4.82
24.
870
500
1.20
72.
375
3.02
43.
391
3.59
73.
625
3.72
53.
803
3.88
13.
932
200
100
0.79
40.
393
0.39
30.
369
0.32
60.
359
0.35
50.
440
0.39
80.
446
200
100
1.25
70.
440
0.31
80.
438
0.30
60.
415
0.24
20.
301
0.26
90.
317
200
0.44
80.
292
0.30
10.
306
0.33
00.
284
0.35
40.
355
0.40
10.
453
200
0.81
60.
323
0.41
20.
229
0.21
30.
207
0.23
00.
233
0.27
50.
327
500
0.29
50.
245
0.27
40.
306
0.32
90.
296
0.33
50.
376
0.42
20.
463
500
0.56
70.
281
0.21
70.
197
0.20
90.
194
0.22
10.
258
0.30
10.
342
500
100
1.06
50.
332
0.27
20.
573
0.44
30.
634
0.27
60.
118
0.13
50.
002
500
100
1.50
10.
622
0.45
70.
686
0.51
30.
676
0.30
00.
129
0.14
10.
001
200
0.69
80.
326
0.55
10.
328
0.25
70.
332
0.20
70.
166
0.09
20.
001
200
1.04
80.
540
0.67
60.
403
0.30
10.
357
0.22
10.
174
0.09
50.
001
500
0.53
10.
355
0.33
10.
250
0.20
50.
246
0.17
60.
118
0.05
70.
001
500
0.80
20.
505
0.41
30.
295
0.23
00.
260
0.18
40.
123
0.05
90.
001
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.04
5-0
.523
-1.1
86-1
.344
-1.6
52-1
.597
-1.9
97-2
.190
-2.1
41-2
.207
100
100
2.88
51.
029
0.09
1-0
.271
-0.7
32-0
.774
-1.2
47-1
.487
-1.4
74-1
.568
200
0.89
1-0
.465
-0.9
00-1
.471
-1.7
12-1
.762
-1.9
47-1
.984
-2.0
75-2
.143
200
2.71
71.
001
0.25
4-0
.516
-0.9
00-1
.042
-1.2
81-1
.359
-1.4
69-1
.550
500
0.84
0-0
.446
-1.1
57-1
.577
-1.8
22-1
.868
-1.9
84-2
.074
-2.1
57-2
.213
500
2.64
10.
932
-0.0
96-0
.725
-1.0
94-1
.216
-1.3
73-1
.485
-1.5
78-1
.640
200
100
2.11
20.
951
0.58
10.
663
0.40
60.
518
0.11
2-0
.078
-0.0
52-0
.148
200
100
3.66
22.
072
1.34
51.
171
0.74
90.
749
0.27
60.
039
0.03
0-0
.091
200
1.88
20.
979
0.84
10.
418
0.23
10.
234
0.06
10.
011
-0.0
70-0
.146
200
3.41
52.
020
1.49
60.
845
0.50
80.
416
0.18
90.
102
-0.0
01-0
.093
500
1.86
61.
037
0.61
70.
330
0.17
00.
151
0.05
3-0
.022
-0.0
88-0
.140
500
3.39
21.
999
1.19
80.
678
0.38
20.
286
0.14
60.
048
-0.0
31-0
.091
500
100
2.31
11.
162
0.79
90.
900
0.64
60.
752
0.33
50.
122
0.13
20.
000
500
100
3.79
82.
219
1.48
91.
327
0.91
30.
909
0.42
80.
170
0.15
20.
000
200
2.07
71.
187
1.05
70.
632
0.43
30.
431
0.26
00.
191
0.09
80.
000
200
3.57
32.
179
1.65
70.
999
0.64
90.
553
0.32
70.
224
0.11
10.
000
500
2.05
31.
226
0.81
10.
513
0.34
80.
322
0.21
60.
137
0.06
30.
000
500
3.54
92.
147
1.34
10.
810
0.50
90.
406
0.25
80.
157
0.07
10.
000
TN
RM
SE
TN
RM
SE
100
100
1.28
91.
085
1.61
31.
794
2.06
82.
041
2.36
62.
538
2.49
12.
557
100
100
2.97
31.
295
0.86
30.
961
1.20
61.
240
1.58
41.
785
1.77
31.
861
200
1.07
40.
929
1.33
51.
827
2.05
62.
113
2.27
62.
310
2.39
82.
463
200
2.76
61.
179
0.79
30.
977
1.25
51.
378
1.57
51.
640
1.74
31.
819
500
0.96
90.
873
1.48
81.
895
2.13
02.
189
2.30
32.
387
2.46
72.
521
500
2.66
71.
086
0.71
51.
086
1.39
01.
509
1.65
41.
756
1.84
31.
904
200
100
2.16
11.
015
0.64
50.
708
0.45
70.
555
0.21
60.
188
0.17
10.
213
200
100
3.70
02.
111
1.38
01.
196
0.77
40.
768
0.31
60.
140
0.12
70.
139
200
1.90
61.
006
0.86
10.
448
0.27
30.
272
0.15
30.
131
0.14
80.
198
200
3.43
22.
036
1.50
80.
858
0.52
20.
430
0.21
70.
140
0.09
30.
130
500
1.87
41.
046
0.62
90.
351
0.20
40.
186
0.12
10.
112
0.14
20.
183
500
3.39
72.
003
1.20
20.
684
0.39
10.
296
0.16
50.
092
0.08
50.
123
500
100
2.35
11.
200
0.82
50.
912
0.65
60.
759
0.34
80.
151
0.15
10.
001
500
100
3.83
52.
253
1.51
51.
342
0.92
40.
916
0.44
00.
194
0.16
90.
000
200
2.09
51.
201
1.06
40.
636
0.43
70.
434
0.26
40.
196
0.10
40.
000
200
3.58
92.
193
1.66
41.
004
0.65
40.
556
0.33
00.
228
0.11
60.
000
500
2.05
91.
229
0.81
30.
514
0.34
90.
322
0.21
60.
138
0.06
50.
000
500
3.55
32.
150
1.34
30.
811
0.51
00.
406
0.25
90.
158
0.07
20.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.es
ign
2as
sum
esa
two-
fact
orm
od
elw
ith
[Nαβ]
an
d[N
αβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3,
wh
ereαβ
rela
tes
toα
un
der
(11)
,f jt
anduit∼IIDN
(0,1
),v ij∼IIDU
(µvj−
0.2,µvj
+0.
2),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in
(32)
and
(33)
.T
he
num
ber
ofre
pli
cati
ons
isse
ttoR
=2000.
31
Tab
leA
3a:
Bia
san
dR
MS
E×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
om
ast
atic
pan
eld
ata
mod
elC
ross
corr
elat
ion
sare
gen
erate
du
sin
gD
esig
n1
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
02.
272
0.83
70.
313
0.22
10.
042
0.17
6-0
.159
-0.2
43-0
.103
-0.0
4710
010
03.
653
1.84
70.
991
0.67
00.
318
0.35
0-0
.048
-0.1
79-0
.068
-0.0
3020
02.
184
0.89
90.
639
0.17
90.
027
0.02
0-0
.114
-0.0
76-0
.078
-0.0
6220
03.
487
1.79
91.
201
0.53
90.
252
0.15
6-0
.032
-0.0
26-0
.048
-0.0
4150
02.
138
1.15
30.
485
0.21
30.
030
0.03
5-0
.036
-0.0
67-0
.090
-0.0
8650
03.
317
1.92
90.
948
0.49
30.
199
0.13
70.
029
-0.0
23-0
.057
-0.0
5920
010
02.
680
1.15
00.
594
0.35
90.
156
0.28
6-0
.090
-0.1
88-0
.050
0.00
020
010
04.
140
2.23
51.
324
0.82
30.
446
0.46
10.
013
-0.1
35-0
.030
0.00
020
02.
604
1.30
90.
890
0.32
30.
134
0.11
1-0
.022
-0.0
04-0
.017
0.00
020
03.
993
2.28
81.
499
0.70
30.
364
0.23
60.
051
0.03
1-0
.005
0.00
050
02.
642
1.44
00.
792
0.41
80.
178
0.14
00.
069
0.02
70.
000
0.00
050
03.
934
2.28
61.
311
0.71
70.
345
0.22
40.
112
0.04
70.
007
0.00
050
010
02.
988
1.36
60.
677
0.43
10.
202
0.30
4-0
.079
-0.1
77-0
.047
0.00
050
010
04.
527
2.50
71.
450
0.92
50.
513
0.48
90.
028
-0.1
19-0
.025
0.00
020
03.
035
1.52
21.
012
0.39
90.
164
0.15
7-0
.019
0.00
1-0
.014
0.00
020
04.
505
2.56
51.
667
0.80
70.
408
0.29
70.
056
0.03
80.
000
0.00
050
03.
170
1.71
90.
910
0.44
90.
203
0.16
20.
079
0.02
90.
003
0.00
050
04.
577
2.64
01.
470
0.77
00.
384
0.25
50.
127
0.05
00.
011
0.00
0T
NR
MSE
TN
RM
SE
100
100
2.87
01.
712
1.03
20.
720
0.45
40.
330
0.28
70.
282
0.14
40.
098
100
100
4.18
32.
530
1.54
31.
081
0.64
10.
493
0.28
80.
243
0.11
20.
068
200
3.01
81.
642
1.16
10.
731
0.58
60.
219
0.21
40.
139
0.12
00.
112
200
4.19
32.
418
1.67
51.
017
0.72
60.
312
0.20
30.
114
0.08
60.
078
500
3.06
61.
984
1.03
10.
793
0.39
70.
250
0.15
30.
149
0.15
90.
145
500
4.12
72.
686
1.44
61.
000
0.51
90.
323
0.15
10.
120
0.11
40.
104
200
100
3.19
11.
765
1.19
00.
747
0.45
60.
459
0.21
10.
217
0.07
50.
003
200
100
4.57
22.
725
1.80
11.
136
0.70
00.
633
0.24
50.
195
0.07
60.
002
200
3.15
51.
943
1.34
90.
662
0.46
20.
240
0.20
00.
076
0.03
40.
002
200
4.46
52.
817
1.90
91.
000
0.64
90.
358
0.24
60.
103
0.04
00.
001
500
3.24
01.
965
1.26
50.
867
0.43
90.
230
0.13
90.
108
0.02
70.
001
500
4.45
72.
753
1.74
61.
129
0.59
60.
320
0.18
80.
128
0.03
40.
001
500
100
3.35
11.
900
1.01
50.
680
0.41
20.
445
0.17
60.
202
0.06
40.
000
500
100
4.83
22.
903
1.71
51.
134
0.68
20.
624
0.20
40.
174
0.06
40.
000
200
3.54
11.
946
1.27
80.
616
0.34
70.
294
0.13
80.
066
0.03
20.
000
200
4.91
72.
913
1.90
90.
996
0.56
00.
424
0.17
70.
093
0.03
70.
000
500
3.70
42.
155
1.29
80.
660
0.40
80.
239
0.12
80.
057
0.04
00.
000
500
5.01
83.
016
1.81
00.
964
0.56
70.
334
0.17
90.
079
0.04
70.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
5.75
03.
447
2.10
51.
411
0.78
50.
634
0.12
7-0
.084
-0.0
24-0
.018
100
100
8.49
65.
631
3.68
12.
501
1.49
21.
069
0.39
40.
058
0.03
5-0
.010
200
6.20
23.
781
2.49
61.
376
0.77
70.
461
0.14
30.
069
-0.0
02-0
.021
200
8.93
65.
912
3.98
52.
381
1.41
40.
837
0.35
50.
177
0.04
3-0
.013
500
6.77
54.
358
2.48
81.
433
0.75
30.
444
0.20
30.
073
-0.0
03-0
.027
500
9.46
96.
422
3.91
32.
350
1.30
30.
749
0.36
70.
154
0.03
2-0
.017
200
100
6.30
43.
897
2.49
31.
587
0.92
90.
757
0.18
9-0
.046
0.00
40.
000
200
100
9.06
06.
128
4.12
42.
705
1.66
11.
210
0.46
20.
094
0.05
90.
000
200
6.79
54.
372
2.87
31.
591
0.90
90.
543
0.22
80.
116
0.02
70.
000
200
9.53
26.
541
4.40
62.
637
1.57
10.
929
0.45
00.
225
0.06
70.
000
500
7.52
44.
846
2.98
31.
731
0.93
10.
527
0.26
90.
119
0.03
30.
000
500
10.2
066.
939
4.46
82.
694
1.51
20.
841
0.43
40.
196
0.06
10.
000
500
100
6.74
74.
228
2.66
41.
727
1.02
90.
796
0.21
0-0
.022
0.01
20.
000
500
100
9.53
26.
485
4.33
52.
877
1.78
71.
264
0.49
20.
128
0.06
90.
000
200
7.39
04.
731
3.11
41.
747
0.98
00.
633
0.23
90.
128
0.03
30.
000
200
10.1
576.
940
4.69
52.
827
1.66
91.
043
0.46
70.
241
0.07
50.
000
500
8.30
45.
326
3.23
41.
844
1.00
70.
585
0.29
70.
127
0.03
90.
000
500
10.9
897.
458
4.76
42.
843
1.61
70.
920
0.47
20.
209
0.06
80.
000
TN
RM
SE
TN
RM
SE
100
100
6.19
23.
978
2.53
61.
756
1.05
40.
778
0.37
50.
220
0.09
90.
045
100
100
8.83
66.
029
4.01
82.
794
1.72
11.
212
0.58
80.
264
0.12
20.
029
200
6.72
24.
248
2.89
21.
766
1.15
60.
603
0.30
10.
152
0.06
60.
045
200
9.32
06.
270
4.31
32.
708
1.73
20.
974
0.48
80.
248
0.09
00.
030
500
7.32
14.
922
2.89
01.
828
1.02
80.
621
0.29
10.
154
0.06
50.
053
500
9.86
76.
859
4.25
02.
686
1.55
50.
922
0.45
50.
225
0.07
00.
035
200
100
6.65
24.
277
2.87
51.
848
1.14
90.
926
0.36
60.
189
0.08
80.
002
200
100
9.32
16.
410
4.42
52.
920
1.85
21.
369
0.60
70.
254
0.12
80.
001
200
7.14
34.
765
3.20
71.
839
1.14
60.
657
0.38
90.
181
0.06
40.
001
200
9.79
06.
839
4.67
72.
848
1.77
61.
038
0.59
20.
286
0.10
10.
001
500
7.87
75.
184
3.32
52.
062
1.15
60.
631
0.35
10.
196
0.06
00.
001
500
10.4
647.
200
4.74
92.
975
1.71
70.
946
0.51
60.
268
0.08
80.
001
500
100
6.98
84.
518
2.87
71.
903
1.17
60.
924
0.33
40.
163
0.08
00.
000
500
100
9.70
86.
698
4.50
33.
028
1.91
61.
383
0.59
30.
243
0.12
50.
000
200
7.67
64.
985
3.31
61.
905
1.10
50.
746
0.32
80.
175
0.06
30.
000
200
10.3
627.
130
4.86
12.
963
1.78
01.
146
0.54
30.
285
0.10
30.
000
500
8.58
15.
585
3.48
32.
007
1.15
50.
663
0.35
20.
158
0.07
20.
000
500
11.1
897.
655
4.96
22.
985
1.74
80.
996
0.52
70.
239
0.09
90.
000
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.
Non
-Gau
ssia
ner
rors
are
gen
erat
edasuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31)
.D
esig
n1
assu
mesb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elati
on
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
give
nby
(30)
,w
her
eBN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
al
valu
eu
sed
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23)
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
32
Tab
leA
3b
:B
ias
an
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n1
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
02.
224
0.77
00.
293
0.19
00.
027
0.17
1-0
.178
-0.2
56-0
.110
-0.0
5410
010
03.
581
1.76
50.
966
0.62
00.
305
0.34
7-0
.067
-0.1
88-0
.073
-0.0
3520
02.
032
0.82
10.
589
0.14
0-0
.032
0.00
5-0
.128
-0.0
90-0
.096
-0.0
7720
03.
308
1.69
91.
143
0.49
50.
187
0.14
1-0
.042
-0.0
37-0
.059
-0.0
5150
02.
043
0.96
60.
483
0.17
10.
015
0.01
7-0
.051
-0.0
83-0
.105
-0.1
0550
03.
212
1.71
30.
954
0.44
60.
187
0.11
90.
015
-0.0
35-0
.067
-0.0
7220
010
02.
719
1.15
90.
531
0.33
80.
149
0.26
9-0
.085
-0.1
83-0
.051
0.00
020
010
04.
192
2.23
91.
248
0.79
60.
437
0.44
00.
017
-0.1
29-0
.032
0.00
020
02.
632
1.28
00.
894
0.33
40.
116
0.12
2-0
.025
-0.0
05-0
.016
0.00
020
04.
013
2.24
71.
499
0.71
30.
340
0.24
90.
046
0.02
9-0
.003
0.00
050
02.
622
1.42
00.
791
0.37
00.
164
0.13
60.
069
0.02
60.
001
0.00
050
03.
915
2.25
41.
300
0.65
90.
326
0.21
80.
112
0.04
50.
008
0.00
050
010
02.
974
1.35
00.
694
0.40
50.
203
0.30
3-0
.076
-0.1
79-0
.048
0.00
050
010
04.
489
2.48
71.
462
0.89
10.
511
0.48
60.
033
-0.1
23-0
.027
0.00
020
02.
913
1.47
41.
005
0.40
50.
170
0.14
6-0
.013
-0.0
02-0
.014
0.00
020
04.
383
2.50
21.
652
0.81
30.
415
0.28
40.
062
0.03
40.
000
0.00
050
03.
138
1.67
40.
918
0.47
80.
198
0.17
60.
078
0.02
90.
001
0.00
050
04.
541
2.59
21.
482
0.80
40.
377
0.27
30.
125
0.05
00.
008
0.00
0T
NR
MSE
TN
RM
SE
100
100
2.81
81.
527
0.98
10.
768
0.44
40.
457
0.25
70.
301
0.15
60.
110
100
100
4.11
62.
368
1.49
61.
097
0.63
20.
593
0.23
10.
253
0.12
20.
076
200
2.73
01.
507
1.09
60.
686
0.41
60.
262
0.21
30.
159
0.15
60.
136
200
3.91
72.
278
1.59
10.
959
0.53
20.
341
0.19
40.
133
0.11
40.
097
500
2.93
91.
547
1.05
50.
623
0.54
90.
314
0.16
50.
161
0.17
00.
175
500
3.99
72.
246
1.48
30.
848
0.65
20.
370
0.15
30.
121
0.12
10.
127
200
100
3.25
61.
787
1.04
20.
722
0.41
40.
404
0.33
10.
244
0.07
10.
002
200
100
4.64
52.
741
1.65
41.
111
0.65
60.
580
0.36
10.
231
0.07
10.
001
200
3.29
91.
992
1.49
00.
790
0.42
30.
291
0.20
50.
099
0.05
40.
006
200
4.57
22.
834
2.00
41.
101
0.59
90.
412
0.24
70.
123
0.06
20.
005
500
3.19
51.
991
1.38
60.
626
0.40
30.
229
0.15
50.
075
0.04
00.
002
500
4.41
42.
754
1.82
80.
908
0.55
90.
316
0.20
10.
097
0.04
70.
001
500
100
3.36
31.
821
1.09
50.
623
0.43
60.
403
0.18
90.
210
0.06
50.
000
500
100
4.81
32.
851
1.77
11.
073
0.69
60.
588
0.22
00.
185
0.06
40.
000
200
3.25
71.
957
1.35
60.
704
0.37
10.
286
0.23
70.
060
0.03
50.
000
200
4.67
22.
885
1.95
21.
060
0.58
00.
410
0.27
40.
085
0.04
10.
000
500
3.65
11.
969
1.28
70.
764
0.37
00.
331
0.14
30.
053
0.01
50.
000
500
4.97
12.
860
1.81
01.
064
0.53
60.
420
0.19
00.
075
0.02
10.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
5.65
93.
338
2.05
81.
326
0.77
00.
632
0.10
9-0
.090
-0.0
28-0
.021
100
100
8.37
75.
505
3.63
32.
386
1.47
21.
067
0.37
30.
053
0.03
1-0
.012
200
6.00
83.
638
2.42
11.
324
0.69
00.
444
0.13
80.
058
-0.0
08-0
.028
200
8.74
35.
747
3.89
02.
309
1.30
80.
816
0.35
20.
165
0.04
0-0
.017
500
6.64
54.
087
2.49
91.
366
0.74
00.
422
0.18
50.
065
-0.0
08-0
.034
500
9.33
26.
129
3.92
22.
265
1.28
50.
719
0.34
40.
147
0.02
9-0
.021
200
100
6.34
93.
883
2.40
51.
544
0.92
10.
730
0.19
1-0
.037
0.00
20.
000
200
100
9.10
26.
090
4.03
22.
643
1.65
31.
178
0.46
00.
106
0.05
60.
000
200
6.81
54.
303
2.87
41.
600
0.87
90.
558
0.22
00.
112
0.02
80.
000
200
9.55
96.
462
4.40
72.
641
1.53
40.
944
0.43
80.
218
0.06
80.
000
500
7.51
04.
793
2.96
11.
652
0.90
00.
517
0.26
90.
117
0.03
40.
000
500
10.1
976.
880
4.44
22.
604
1.47
10.
828
0.43
20.
194
0.06
20.
000
500
100
6.70
44.
202
2.67
31.
683
1.02
10.
796
0.21
6-0
.027
0.00
90.
000
500
100
9.48
36.
464
4.33
02.
827
1.77
51.
265
0.49
70.
119
0.06
50.
000
200
7.27
14.
659
3.08
91.
752
0.99
30.
612
0.24
50.
122
0.03
30.
000
200
10.0
356.
857
4.66
12.
829
1.68
31.
016
0.47
20.
233
0.07
50.
000
500
8.25
95.
286
3.25
31.
885
0.99
30.
609
0.29
20.
128
0.03
50.
000
500
10.9
427.
418
4.78
42.
888
1.59
80.
948
0.46
60.
209
0.06
40.
000
TN
RM
SE
TN
RM
SE
100
100
6.10
43.
821
2.48
11.
718
1.04
10.
848
0.30
50.
223
0.10
60.
049
100
100
8.72
45.
874
3.96
82.
707
1.70
41.
259
0.52
40.
261
0.12
50.
032
200
6.46
64.
085
2.79
41.
699
0.95
40.
613
0.29
20.
165
0.08
40.
058
200
9.08
66.
094
4.20
12.
624
1.53
50.
974
0.48
20.
253
0.10
00.
039
500
7.17
74.
504
2.93
41.
700
1.10
50.
629
0.27
30.
141
0.06
70.
067
500
9.72
26.
462
4.28
72.
559
1.60
00.
909
0.42
80.
212
0.07
10.
046
200
100
6.71
24.
274
2.73
21.
811
1.11
10.
871
0.45
80.
236
0.08
50.
001
200
100
9.37
56.
387
4.28
82.
868
1.82
31.
315
0.66
90.
301
0.12
40.
001
200
7.21
24.
731
3.25
91.
907
1.09
00.
709
0.38
20.
192
0.08
50.
004
200
9.84
86.
783
4.70
72.
894
1.71
71.
083
0.57
80.
291
0.11
90.
004
500
7.84
75.
147
3.34
81.
876
1.11
20.
620
0.36
00.
171
0.07
20.
001
500
10.4
447.
151
4.75
02.
805
1.66
60.
930
0.52
30.
248
0.09
80.
001
500
100
6.95
64.
478
2.91
31.
839
1.17
40.
897
0.35
10.
177
0.07
80.
000
500
100
9.66
66.
666
4.51
52.
959
1.90
81.
363
0.60
50.
250
0.12
10.
000
200
7.48
34.
928
3.32
01.
944
1.12
80.
722
0.40
20.
165
0.06
70.
000
200
10.1
927.
055
4.84
52.
987
1.80
21.
116
0.59
80.
274
0.10
60.
000
500
8.53
15.
484
3.49
82.
093
1.13
00.
737
0.35
60.
155
0.04
70.
000
500
11.1
377.
573
4.98
03.
065
1.72
21.
064
0.52
70.
237
0.07
60.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
r
i=
1,2,...,N
.N
on-G
auss
ian
erro
rsar
ege
ner
ated
asuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
h
v=
8d
egre
esof
free
dom
,in
(31)
.D
esig
n1
assu
mesb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,2,...,N
,in
the
const
ruct
ion
of
the
corr
elat
ion
mat
rix
ofth
eer
rors
,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
al
valu
eu
sed
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23)
.c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
al
valu
eu
sed
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=20
00.
33
Tab
leA
4a:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
cp
anel
dat
am
od
elC
ross
corr
elati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
01.
116
-0.9
49-2
.057
-2.5
27-3
.024
-3.0
83-3
.551
-3.7
45-3
.713
-3.7
5410
010
02.
735
0.49
7-0
.797
-1.4
07-2
.004
-2.1
34-2
.645
-2.8
79-2
.869
-2.9
3120
00.
933
-1.0
04-1
.930
-2.7
75-3
.210
-3.3
57-3
.631
-3.6
93-3
.777
-3.8
4420
02.
503
0.35
6-0
.758
-1.7
38-2
.272
-2.4
85-2
.792
-2.8
79-2
.979
-3.0
5550
00.
898
-1.0
48-2
.261
-3.0
08-3
.456
-3.6
15-3
.782
-3.9
01-3
.998
-4.0
6350
02.
404
0.23
8-1
.163
-2.0
48-2
.582
-2.7
95-2
.993
-3.1
26-3
.229
-3.2
9620
010
02.
764
1.26
50.
642
0.53
50.
221
0.27
0-0
.159
-0.3
46-0
.318
-0.3
9420
010
04.
186
2.32
21.
376
1.03
90.
575
0.53
00.
042
-0.1
87-0
.184
-0.2
8120
02.
655
1.33
20.
886
0.30
70.
020
-0.0
36-0
.226
-0.2
83-0
.361
-0.4
3020
04.
035
2.31
11.
535
0.75
10.
324
0.18
5-0
.053
-0.1
41-0
.235
-0.3
1550
02.
623
1.35
90.
641
0.19
0-0
.076
-0.1
38-0
.258
-0.3
42-0
.412
-0.4
6050
03.
925
2.23
61.
209
0.55
80.
175
0.04
3-0
.112
-0.2
15-0
.294
-0.3
4650
010
03.
337
1.82
41.
180
1.09
60.
741
0.79
40.
349
0.12
10.
125
-0.0
0250
010
04.
715
2.80
61.
821
1.48
10.
974
0.92
60.
424
0.15
90.
140
-0.0
0120
03.
163
1.85
11.
417
0.82
20.
525
0.46
70.
270
0.18
70.
094
-0.0
0220
04.
521
2.76
41.
968
1.15
50.
717
0.56
90.
324
0.21
30.
104
-0.0
0150
03.
186
1.89
61.
174
0.69
70.
431
0.35
30.
223
0.13
50.
060
-0.0
0250
04.
510
2.73
91.
667
0.97
10.
576
0.42
60.
258
0.15
10.
065
-0.0
01T
NR
MSE
TN
RM
SE
100
100
2.16
51.
928
2.63
13.
035
3.49
93.
576
3.99
94.
181
4.16
24.
206
100
100
3.41
31.
803
1.73
32.
023
2.48
32.
605
3.05
33.
266
3.27
13.
336
200
2.17
91.
962
2.49
93.
204
3.62
43.
796
4.06
64.
140
4.22
74.
292
200
3.32
51.
811
1.67
52.
227
2.67
52.
893
3.18
43.
281
3.38
43.
460
500
2.48
92.
230
2.88
93.
493
3.90
54.
068
4.23
74.
357
4.45
24.
516
500
3.48
22.
058
2.07
82.
589
3.02
03.
217
3.40
93.
540
3.64
23.
709
200
100
3.21
31.
794
1.13
10.
851
0.54
10.
478
0.38
30.
473
0.45
10.
508
200
100
4.56
02.
717
1.71
61.
255
0.75
90.
637
0.28
90.
313
0.30
70.
373
200
3.21
61.
863
1.23
10.
650
0.38
80.
317
0.37
20.
413
0.47
20.
532
200
4.51
42.
751
1.82
80.
996
0.52
10.
334
0.24
00.
272
0.33
40.
399
500
3.27
91.
932
1.10
20.
598
0.39
40.
343
0.40
00.
467
0.52
90.
571
500
4.49
22.
735
1.59
70.
857
0.43
40.
274
0.26
50.
330
0.39
40.
440
500
100
3.56
92.
057
1.36
01.
190
0.80
60.
821
0.37
70.
156
0.14
40.
010
500
100
4.92
63.
012
1.98
01.
573
1.03
90.
955
0.45
20.
190
0.15
80.
009
200
3.46
62.
107
1.55
60.
915
0.57
80.
486
0.28
00.
193
0.10
00.
004
200
4.79
23.
000
2.11
41.
256
0.77
70.
594
0.33
70.
219
0.10
90.
003
500
3.60
52.
250
1.43
20.
871
0.53
00.
390
0.23
80.
140
0.06
20.
004
500
4.86
73.
055
1.91
11.
144
0.68
00.
469
0.27
70.
157
0.06
80.
003
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
5.02
92.
438
0.80
5-0
.099
-0.8
95-1
.169
-1.7
67-2
.070
-2.1
01-2
.197
100
100
7.90
24.
856
2.73
61.
406
0.29
7-0
.215
-0.9
57-1
.367
-1.4
67-1
.611
200
5.51
72.
858
1.22
4-0
.128
-0.9
37-1
.333
-1.7
41-1
.900
-2.0
40-2
.139
200
8.40
25.
243
3.06
81.
291
0.15
0-0
.479
-1.0
25-1
.275
-1.4
66-1
.596
500
6.30
73.
401
1.32
0-0
.089
-0.9
74-1
.409
-1.7
29-1
.926
-2.0
60-2
.141
500
9.14
75.
730
3.11
21.
241
0.01
7-0
.644
-1.0
95-1
.363
-1.5
34-1
.633
200
100
6.27
03.
901
2.47
91.
776
1.07
00.
858
0.27
6-0
.022
-0.0
62-0
.190
200
100
8.91
15.
987
3.99
12.
793
1.74
61.
286
0.55
80.
157
0.04
9-0
.125
200
6.77
44.
318
2.87
21.
642
0.90
20.
562
0.20
60.
044
-0.0
90-0
.194
200
9.45
96.
401
4.32
62.
628
1.53
40.
950
0.44
60.
192
0.00
5-0
.131
500
7.48
64.
767
2.86
91.
599
0.81
90.
447
0.16
1-0
.009
-0.1
22-0
.192
500
10.1
316.
809
4.29
82.
520
1.37
80.
770
0.35
10.
108
-0.0
43-0
.133
500
100
6.77
24.
331
2.85
52.
129
1.37
81.
155
0.55
50.
225
0.16
4-0
.001
500
100
9.41
26.
412
4.34
53.
101
2.00
21.
517
0.76
50.
332
0.20
40.
000
200
7.25
24.
723
3.22
21.
942
1.17
90.
822
0.45
70.
274
0.12
60.
000
200
9.92
66.
788
4.64
22.
883
1.75
31.
145
0.62
80.
355
0.15
50.
000
500
8.09
85.
256
3.26
01.
907
1.08
90.
687
0.38
50.
206
0.08
40.
000
500
10.7
307.
297
4.68
02.
806
1.60
70.
960
0.51
90.
265
0.10
40.
000
TN
RM
SE
TN
RM
SE
100
100
5.48
93.
043
1.69
01.
275
1.49
31.
667
2.14
02.
409
2.45
42.
551
100
100
8.22
45.
216
3.11
91.
827
1.03
40.
949
1.33
51.
668
1.77
51.
915
200
6.02
53.
453
1.91
91.
205
1.43
91.
730
2.08
52.
242
2.38
12.
480
200
8.75
85.
623
3.43
31.
738
0.95
20.
995
1.35
91.
581
1.76
31.
890
500
6.88
64.
072
2.17
81.
368
1.53
01.
809
2.08
22.
266
2.39
62.
478
500
9.55
16.
175
3.59
41.
836
1.05
41.
131
1.43
31.
665
1.82
71.
926
200
100
6.57
14.
206
2.73
61.
945
1.19
80.
928
0.37
50.
198
0.19
20.
263
200
100
9.14
16.
221
4.19
32.
934
1.85
11.
344
0.61
40.
229
0.14
50.
182
200
7.13
44.
656
3.12
11.
838
1.03
50.
635
0.27
90.
164
0.18
30.
257
200
9.72
86.
666
4.53
92.
801
1.65
31.
015
0.48
90.
229
0.11
50.
180
500
7.86
75.
131
3.17
31.
828
0.97
40.
530
0.23
40.
151
0.20
30.
258
500
10.4
127.
093
4.55
22.
724
1.51
80.
844
0.39
20.
157
0.12
50.
186
500
100
6.95
24.
506
2.99
62.
219
1.44
21.
186
0.58
30.
252
0.18
10.
008
500
100
9.55
06.
551
4.46
23.
182
2.06
21.
551
0.79
30.
357
0.22
00.
007
200
7.45
84.
915
3.36
12.
045
1.24
50.
853
0.47
30.
282
0.13
20.
002
200
10.0
816.
940
4.76
42.
980
1.82
01.
181
0.64
80.
365
0.16
00.
001
500
8.33
35.
484
3.45
72.
060
1.19
20.
739
0.40
90.
215
0.08
70.
001
500
10.9
027.
473
4.84
32.
941
1.70
41.
014
0.54
70.
276
0.10
70.
001
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]an
d[N
αβ2]n
on
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vel
y.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
andf jt∼IIDN
(0,1
).N
on-G
auss
ian
erro
rsare
gen
erate
das:uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,
2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
d
ran
dom
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31)
.v ij∼IIDU
(µvj−
0.2,µvj
+0.
2),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
and
(33)
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=20
00.
34
Tab
leA
4b
:B
ias
an
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n2
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
01.
016
-1.1
03-2
.241
-2.7
71-3
.270
-3.3
57-3
.838
-4.0
45-4
.008
-4.0
3610
010
02.
662
0.36
5-0
.939
-1.5
97-2
.202
-2.3
58-2
.883
-3.1
24-3
.111
-3.1
6220
00.
813
-1.1
92-2
.165
-3.0
69-3
.534
-3.7
08-3
.969
-4.0
29-4
.118
-4.1
8420
02.
380
0.19
1-0
.959
-1.9
89-2
.541
-2.7
74-3
.072
-3.1
59-3
.261
-3.3
3750
00.
669
-1.2
97-2
.534
-3.2
94-3
.752
-3.9
26-4
.100
-4.2
16-4
.312
-4.3
8050
02.
163
-0.0
04-1
.413
-2.2
97-2
.832
-3.0
55-3
.256
-3.3
87-3
.488
-3.5
6020
010
02.
749
1.24
90.
619
0.50
60.
180
0.23
5-0
.194
-0.3
87-0
.359
-0.4
3120
010
04.
150
2.29
41.
352
1.01
30.
542
0.50
10.
014
-0.2
17-0
.216
-0.3
0720
02.
578
1.26
70.
834
0.25
6-0
.018
-0.0
70-0
.268
-0.3
16-0
.397
-0.4
6420
03.
950
2.24
11.
482
0.70
10.
292
0.15
7-0
.086
-0.1
66-0
.263
-0.3
4050
02.
568
1.31
10.
610
0.16
1-0
.098
-0.1
61-0
.281
-0.3
66-0
.436
-0.4
8550
03.
869
2.18
81.
178
0.53
10.
155
0.02
5-0
.130
-0.2
33-0
.312
-0.3
6550
010
03.
419
1.88
81.
222
1.11
30.
749
0.79
20.
349
0.12
50.
124
-0.0
0250
010
04.
812
2.88
61.
868
1.50
50.
989
0.92
70.
426
0.16
40.
139
-0.0
0120
03.
202
1.88
21.
439
0.83
90.
535
0.47
10.
272
0.19
10.
092
-0.0
0220
04.
557
2.79
61.
992
1.17
30.
727
0.57
50.
326
0.21
60.
102
-0.0
0150
03.
166
1.87
91.
158
0.68
50.
422
0.34
80.
221
0.13
30.
059
-0.0
0250
04.
489
2.72
21.
651
0.95
90.
567
0.42
00.
256
0.14
90.
065
-0.0
01T
NR
MSE
TN
RM
SE
100
100
2.11
72.
014
2.78
23.
266
3.74
93.
863
4.30
64.
507
4.48
84.
515
100
100
3.35
81.
765
1.79
72.
176
2.67
62.
840
3.30
73.
537
3.54
03.
591
200
2.21
92.
155
2.74
43.
521
3.96
94.
155
4.40
64.
474
4.55
94.
626
200
3.28
51.
877
1.86
12.
502
2.96
73.
189
3.46
93.
562
3.66
13.
737
500
2.42
42.
367
3.10
43.
732
4.15
14.
331
4.50
64.
622
4.71
74.
785
500
3.32
12.
061
2.23
22.
785
3.22
33.
432
3.62
83.
757
3.85
83.
930
200
100
3.19
11.
751
1.07
00.
792
0.49
20.
453
0.40
70.
521
0.50
20.
555
200
100
4.53
32.
687
1.67
51.
210
0.70
70.
603
0.28
70.
347
0.34
70.
408
200
3.13
51.
816
1.21
00.
658
0.43
10.
357
0.42
40.
456
0.52
40.
579
200
4.42
12.
684
1.78
60.
970
0.52
60.
343
0.27
50.
304
0.37
40.
436
500
3.27
01.
965
1.18
10.
708
0.48
60.
394
0.42
50.
483
0.54
50.
591
500
4.45
82.
730
1.63
20.
920
0.51
20.
328
0.29
10.
341
0.40
60.
455
500
100
3.67
42.
127
1.38
61.
183
0.79
10.
808
0.36
60.
154
0.14
30.
006
500
100
5.05
13.
105
2.02
71.
583
1.03
70.
947
0.44
40.
189
0.15
70.
004
200
3.61
82.
254
1.66
11.
001
0.63
10.
509
0.29
10.
198
0.09
90.
006
200
4.91
43.
122
2.20
71.
335
0.82
80.
619
0.34
90.
225
0.10
90.
005
500
3.50
22.
134
1.31
90.
778
0.46
80.
362
0.22
60.
135
0.06
10.
003
500
4.79
22.
967
1.81
91.
063
0.62
10.
439
0.26
30.
151
0.06
70.
002
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
4.96
72.
337
0.69
1-0
.247
-1.0
47-1
.341
-1.9
60-2
.264
-2.2
93-2
.378
100
100
7.86
84.
785
2.65
51.
301
0.18
6-0
.344
-1.1
12-1
.516
-1.6
17-1
.751
200
5.38
62.
703
1.05
9-0
.320
-1.1
44-1
.548
-1.9
52-2
.111
-2.2
51-2
.349
200
8.27
25.
102
2.92
51.
133
-0.0
17-0
.647
-1.1
90-1
.440
-1.6
32-1
.760
500
6.06
03.
176
1.11
0-0
.281
-1.1
55-1
.589
-1.9
10-2
.103
-2.2
34-2
.319
500
8.90
55.
513
2.91
71.
072
-0.1
36-0
.790
-1.2
39-1
.502
-1.6
69-1
.771
200
100
6.21
63.
864
2.45
61.
755
1.04
60.
840
0.25
6-0
.043
-0.0
86-0
.208
200
100
8.87
95.
963
3.97
92.
784
1.73
21.
277
0.54
50.
142
0.03
1-0
.136
200
6.67
44.
238
2.81
51.
590
0.87
20.
538
0.18
20.
028
-0.1
08-0
.211
200
9.35
06.
310
4.26
22.
570
1.49
80.
925
0.42
50.
179
-0.0
09-0
.142
500
7.41
14.
705
2.82
81.
567
0.79
90.
433
0.15
0-0
.020
-0.1
32-0
.203
500
10.0
626.
749
4.25
52.
487
1.35
60.
756
0.34
20.
100
-0.0
50-0
.141
500
100
6.88
24.
427
2.91
92.
162
1.39
91.
163
0.55
90.
232
0.16
5-0
.001
500
100
9.53
26.
514
4.41
63.
144
2.02
91.
530
0.77
20.
340
0.20
50.
000
200
7.28
44.
756
3.25
01.
961
1.19
30.
830
0.46
00.
279
0.12
5-0
.001
200
9.94
16.
805
4.66
32.
897
1.76
51.
152
0.63
20.
360
0.15
40.
000
500
8.07
35.
235
3.24
31.
895
1.07
90.
680
0.38
10.
203
0.08
40.
000
500
10.7
057.
276
4.66
12.
793
1.59
80.
953
0.51
50.
262
0.10
30.
000
TN
RM
SE
TN
RM
SE
100
100
5.44
02.
965
1.63
81.
304
1.61
81.
840
2.34
92.
627
2.66
82.
753
100
100
8.20
05.
156
3.05
81.
759
1.04
11.
045
1.50
01.
839
1.94
42.
073
200
5.92
33.
366
1.88
51.
347
1.66
01.
950
2.30
12.
456
2.59
02.
688
200
8.64
35.
511
3.33
61.
690
1.03
61.
148
1.52
71.
750
1.92
72.
055
500
6.65
53.
889
2.06
51.
402
1.64
81.
947
2.22
52.
407
2.53
62.
622
500
9.31
65.
974
3.42
91.
730
1.06
21.
215
1.54
01.
773
1.93
32.
036
200
100
6.53
44.
180
2.71
41.
916
1.16
30.
902
0.35
20.
211
0.21
90.
287
200
100
9.12
16.
208
4.18
62.
924
1.83
41.
330
0.59
40.
222
0.15
30.
198
200
7.02
54.
571
3.06
31.
791
1.01
60.
622
0.27
60.
175
0.21
00.
281
200
9.61
36.
569
4.47
22.
744
1.62
20.
994
0.47
50.
225
0.13
20.
198
500
7.79
95.
083
3.15
51.
831
0.99
60.
554
0.25
50.
156
0.20
70.
266
500
10.3
467.
040
4.52
22.
710
1.52
10.
855
0.40
40.
160
0.12
60.
191
500
100
7.08
04.
614
3.06
52.
246
1.45
41.
188
0.57
90.
255
0.18
10.
003
500
100
9.68
56.
665
4.54
23.
226
2.08
51.
559
0.79
50.
361
0.22
00.
002
200
7.54
45.
006
3.43
92.
111
1.29
40.
881
0.48
80.
290
0.13
20.
003
200
10.1
316.
996
4.82
13.
029
1.86
01.
206
0.66
20.
373
0.16
10.
003
500
8.28
55.
432
3.40
02.
006
1.14
60.
710
0.39
40.
208
0.08
60.
001
500
10.8
637.
433
4.79
92.
899
1.66
70.
988
0.53
10.
268
0.10
50.
001
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
model
,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
an
d[N
αβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:
αβ
2=
2αβ/3,
wh
ereαβ
rela
tes
toα
un
der
(11)
andf jt∼IIDN
(0,1
).N
on
-Gau
ssia
ner
rors
are
gen
erate
das:
uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,
ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
ared
ran
dom
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).
v ij∼IIDU
(µvj−
0.2,µvj
+0.
2),j
=1,
2,µv
=0.
87,
µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
35
Tab
leB
1:C
omp
aris
onof
Bia
san
dR
MS
E(×
100)
for
theα
andα
esti
mat
esof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stat
ican
dd
yn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Bia
sR
MSE
Sta
tic
model
:ϑi
=0
fori
=1,
2,...,N
TN
αT
Nα
100
100
0.20
4-0
.609
-0.5
84-0
.355
-0.2
89-0
.001
-0.2
59-0
.279
-0.0
94-0
.016
100
100
0.31
10.
633
0.59
60.
364
0.29
60.
051
0.26
40.
282
0.10
20.
041
200
0.06
6-0
.426
-0.1
56-0
.302
-0.2
54-0
.101
-0.1
65-0
.080
-0.0
58-0
.025
200
0.14
50.
438
0.17
40.
308
0.26
10.
124
0.17
10.
093
0.07
00.
053
500
0.09
0-0
.120
-0.1
62-0
.144
-0.1
49-0
.035
-0.0
42-0
.046
-0.0
50-0
.041
500
0.11
80.
137
0.17
70.
157
0.16
50.
073
0.07
80.
076
0.07
70.
079
200
100
0.24
3-0
.584
-0.5
57-0
.328
-0.2
670.
017
-0.2
40-0
.262
-0.0
780.
000
200
100
0.34
70.
611
0.57
10.
338
0.27
20.
042
0.24
20.
263
0.07
90.
000
200
0.11
8-0
.380
-0.1
20-0
.271
-0.2
26-0
.074
-0.1
39-0
.057
-0.0
350.
000
200
0.18
30.
391
0.13
30.
274
0.22
80.
076
0.14
00.
057
0.03
50.
000
500
0.15
8-0
.064
-0.1
14-0
.101
-0.1
060.
007
-0.0
01-0
.006
-0.0
110.
000
500
0.17
30.
075
0.11
60.
102
0.10
70.
010
0.00
40.
007
0.01
10.
000
500
100
0.27
7-0
.556
-0.5
49-0
.325
-0.2
610.
019
-0.2
39-0
.262
-0.0
780.
000
500
100
0.37
40.
584
0.56
30.
337
0.26
80.
043
0.24
00.
263
0.07
80.
000
200
0.15
7-0
.367
-0.1
10-0
.265
-0.2
23-0
.073
-0.1
38-0
.056
-0.0
350.
000
200
0.21
80.
379
0.12
60.
268
0.22
50.
075
0.13
80.
057
0.03
50.
000
500
0.18
9-0
.047
-0.1
05-0
.097
-0.1
040.
009
0.00
0-0
.006
-0.0
110.
000
500
0.20
30.
063
0.10
80.
098
0.10
40.
011
0.00
40.
007
0.01
10.
000
TN
αT
Nα
100
100
1.22
0-0
.126
-0.2
87-0
.090
-0.0
300.
216
-0.0
96-0
.115
0.05
10.
105
100
100
2.20
81.
607
1.43
51.
046
0.78
60.
610
0.43
80.
324
0.18
80.
105
200
0.46
3-0
.305
-0.0
46-0
.191
-0.0
910.
040
-0.0
220.
056
0.07
00.
093
200
1.39
21.
292
1.00
60.
826
0.58
90.
416
0.29
10.
206
0.13
60.
093
500
0.16
2-0
.182
-0.1
16-0
.010
-0.0
160.
091
0.08
00.
080
0.07
00.
080
500
1.05
70.
996
0.76
20.
575
0.38
60.
295
0.20
30.
142
0.09
70.
080
200
100
1.91
60.
155
-0.2
08-0
.091
-0.0
930.
147
-0.1
34-0
.168
-0.0
010.
050
200
100
2.57
61.
473
1.15
20.
842
0.60
50.
452
0.34
60.
275
0.13
90.
050
200
0.53
0-0
.375
-0.1
43-0
.221
-0.1
760.
003
-0.0
75-0
.002
0.02
20.
045
200
1.12
31.
036
0.76
70.
615
0.45
40.
298
0.22
20.
139
0.08
60.
045
500
0.14
4-0
.244
-0.1
88-0
.086
-0.0
810.
052
0.04
40.
036
0.03
10.
040
500
0.74
20.
759
0.57
50.
425
0.29
10.
202
0.14
10.
094
0.05
60.
040
500
100
6.30
12.
825
0.92
80.
325
0.04
50.
186
-0.1
33-0
.190
-0.0
300.
018
500
100
6.67
53.
383
1.62
20.
932
0.56
50.
415
0.29
10.
256
0.11
10.
018
200
1.49
50.
021
0.03
3-0
.221
-0.1
77-0
.037
-0.1
04-0
.027
-0.0
120.
017
200
1.86
90.
881
0.60
70.
477
0.34
30.
212
0.17
80.
098
0.05
70.
017
500
0.14
8-0
.269
-0.2
16-0
.118
-0.0
990.
022
0.01
70.
011
0.00
70.
015
500
0.54
70.
544
0.42
70.
280
0.20
50.
123
0.08
90.
055
0.03
10.
015
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs:ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αT
Nα
100
100
0.19
8-0
.621
-0.5
94-0
.363
-0.2
95-0
.007
-0.2
61-0
.284
-0.0
99-0
.021
100
100
0.30
20.
644
0.60
70.
375
0.30
50.
056
0.26
60.
289
0.10
80.
045
200
0.06
4-0
.432
-0.1
61-0
.309
-0.2
62-0
.108
-0.1
73-0
.088
-0.0
70-0
.033
200
0.15
30.
444
0.18
00.
316
0.27
20.
126
0.18
30.
106
0.09
40.
064
500
0.08
4-0
.123
-0.1
69-0
.158
-0.1
58-0
.046
-0.0
54-0
.059
-0.0
66-0
.049
500
0.11
70.
144
0.18
50.
179
0.17
40.
102
0.10
20.
102
0.10
70.
090
200
100
0.25
5-0
.573
-0.5
61-0
.330
-0.2
660.
018
-0.2
40-0
.263
-0.0
780.
000
200
100
0.35
40.
600
0.57
30.
339
0.27
20.
043
0.24
10.
264
0.07
90.
000
200
0.12
3-0
.381
-0.1
18-0
.272
-0.2
25-0
.074
-0.1
39-0
.056
-0.0
350.
000
200
0.19
00.
392
0.13
10.
274
0.22
60.
076
0.13
90.
057
0.03
50.
000
500
0.15
6-0
.063
-0.1
13-0
.101
-0.1
060.
007
-0.0
01-0
.007
-0.0
110.
000
500
0.16
90.
074
0.11
60.
102
0.10
70.
010
0.00
40.
007
0.01
10.
000
500
100
0.28
1-0
.550
-0.5
50-0
.322
-0.2
610.
020
-0.2
40-0
.261
-0.0
780.
000
500
100
0.37
50.
580
0.56
30.
332
0.26
70.
045
0.24
10.
262
0.07
90.
000
200
0.15
3-0
.366
-0.1
08-0
.266
-0.2
22-0
.072
-0.1
37-0
.056
-0.0
350.
000
200
0.21
30.
378
0.12
30.
270
0.22
40.
074
0.13
80.
057
0.03
50.
000
500
0.18
5-0
.049
-0.1
04-0
.097
-0.1
040.
009
0.00
0-0
.006
-0.0
110.
000
500
0.19
80.
065
0.10
70.
098
0.10
40.
011
0.00
40.
007
0.01
10.
000
TN
αT
Nα
100
100
1.21
9-0
.011
-0.2
92-0
.137
-0.0
330.
188
-0.0
85-0
.123
0.05
60.
106
100
100
2.16
01.
722
1.41
01.
091
0.80
90.
619
0.44
60.
317
0.19
20.
106
200
0.47
2-0
.353
-0.0
80-0
.172
-0.1
020.
043
-0.0
310.
057
0.06
90.
094
200
1.48
01.
319
1.04
40.
812
0.60
80.
421
0.30
70.
205
0.13
60.
094
500
0.15
2-0
.178
-0.1
28-0
.053
-0.0
310.
092
0.09
10.
082
0.07
30.
081
500
1.05
90.
986
0.81
00.
564
0.40
70.
294
0.21
00.
146
0.09
90.
081
200
100
1.93
80.
167
-0.1
58-0
.096
-0.0
830.
155
-0.1
19-0
.170
-0.0
030.
051
200
100
2.59
81.
444
1.15
20.
863
0.60
10.
476
0.34
60.
276
0.13
50.
051
200
0.54
7-0
.363
-0.1
01-0
.227
-0.1
49-0
.011
-0.0
810.
001
0.01
80.
046
200
1.16
61.
023
0.76
50.
612
0.43
90.
294
0.22
10.
139
0.08
50.
046
500
0.13
4-0
.238
-0.2
00-0
.100
-0.0
690.
048
0.04
20.
034
0.03
20.
040
500
0.74
10.
746
0.58
90.
403
0.29
60.
205
0.13
80.
093
0.05
80.
040
500
100
6.14
52.
677
0.95
20.
313
0.04
30.
184
-0.1
42-0
.187
-0.0
310.
018
500
100
6.49
43.
260
1.65
80.
900
0.55
40.
410
0.29
20.
256
0.11
10.
018
200
1.48
90.
038
0.02
4-0
.205
-0.1
91-0
.031
-0.1
04-0
.028
-0.0
100.
017
200
1.89
60.
881
0.60
60.
464
0.35
40.
204
0.17
80.
097
0.05
90.
017
500
0.16
5-0
.264
-0.2
12-0
.124
-0.1
030.
020
0.01
70.
009
0.00
50.
015
500
0.57
10.
561
0.40
40.
284
0.20
60.
130
0.08
60.
056
0.03
10.
015
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit
∼IIDN
(0,1
)in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vec
tor
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
mat
rix
ofth
eer
rors
,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).α
isco
mp
ute
du
sin
gc p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
2an
dp
=0.
05in
the
mu
ltip
lete
stin
gp
roce
du
resh
own
in(2
3).α
corr
esp
on
ds
toth
em
ost
rob
ust
esti
mato
rof
the
exp
on
ent
of
cros
s-se
ctio
nal
dep
end
ence
con
sid
ered
inB
aile
yet
al.
(2016)
an
dall
ows
for
both
seri
al
corr
elati
on
inth
efa
ctors
an
dw
eak
cross
-sec
tion
al
dep
end
ence
inth
eer
ror
term
s.W
eu
sefo
ur
pri
nci
pal
com
pon
ents
wh
enes
tim
ati
ngc N
inth
eex
pre
ssio
nfo
rα
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
36
Tab
leB
2:C
om
pari
son
ofB
ias
an
dR
MS
E(×
100
)fo
rth
eα
andα
ofth
ecr
oss-
sect
ion
alex
pon
ent
ofth
eer
rors
ofa
stat
ican
dd
yn
am
icp
anel
dat
am
od
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Bia
sR
MSE
Sta
tic
model
:ϑi
=0,
fori
=1,
2,...,N
TN
αT
Nα
100
100
-0.8
80-2
.378
-2.9
22-3
.026
-3.2
32-3
.131
-3.5
09-3
.648
-3.5
76-3
.598
100
100
1.25
22.
688
3.30
13.
469
3.69
93.
623
3.97
44.
094
4.02
94.
051
200
-1.2
21-2
.480
-2.8
00-3
.267
-3.4
64-3
.434
-3.5
95-3
.591
-3.6
58-3
.714
200
1.45
52.
748
3.16
73.
655
3.86
83.
859
4.00
24.
009
4.07
04.
130
500
-1.5
01-2
.658
-3.2
66-3
.581
-3.7
80-3
.779
-3.8
59-3
.934
-4.0
11-4
.057
500
1.66
22.
899
3.58
03.
938
4.15
34.
176
4.26
04.
336
4.41
14.
456
200
100
0.70
5-0
.090
-0.1
600.
120
0.00
90.
205
-0.1
45-0
.299
-0.2
42-0
.304
200
100
0.80
30.
377
0.36
90.
346
0.29
80.
354
0.31
20.
405
0.35
90.
399
200
0.37
5-0
.068
0.13
4-0
.086
-0.1
46-0
.063
-0.1
91-0
.206
-0.2
69-0
.326
200
0.45
70.
291
0.30
70.
274
0.29
60.
251
0.31
50.
318
0.36
70.
415
500
0.21
9-0
.024
-0.0
86-0
.158
-0.2
06-0
.153
-0.2
12-0
.267
-0.3
25-0
.368
500
0.30
30.
241
0.26
20.
299
0.32
80.
295
0.32
60.
366
0.41
80.
456
500
100
1.04
30.
295
0.24
70.
563
0.43
30.
625
0.26
40.
084
0.11
4-0
.001
500
100
1.06
80.
337
0.27
20.
571
0.44
20.
631
0.27
90.
116
0.13
60.
002
200
0.68
30.
313
0.55
00.
326
0.25
40.
329
0.20
30.
161
0.08
7-0
.001
200
0.69
70.
325
0.55
30.
329
0.25
70.
332
0.20
80.
166
0.09
30.
001
500
0.53
00.
355
0.33
10.
249
0.20
50.
246
0.17
60.
117
0.05
6-0
.001
500
0.53
40.
357
0.33
20.
250
0.20
60.
246
0.17
70.
118
0.05
70.
001
TN
αT
Nα
100
100
1.17
20.
453
0.38
30.
727
0.53
10.
700
0.31
00.
144
0.19
80.
104
100
100
2.34
91.
859
1.54
31.
430
1.07
31.
002
0.61
40.
393
0.30
10.
104
200
0.25
00.
045
0.46
70.
320
0.27
20.
358
0.22
40.
207
0.15
20.
093
200
1.45
31.
314
1.14
40.
878
0.67
20.
590
0.41
20.
312
0.20
50.
093
500
-0.1
48-0
.003
0.18
30.
210
0.20
40.
269
0.21
30.
164
0.12
00.
080
500
1.06
60.
940
0.75
50.
583
0.45
70.
400
0.30
10.
218
0.14
40.
080
200
100
2.80
61.
587
1.10
81.
329
0.84
40.
894
0.42
80.
181
0.18
10.
051
200
100
3.43
82.
401
1.86
21.
879
1.27
71.
157
0.69
90.
414
0.28
80.
051
200
0.90
60.
459
0.68
70.
431
0.30
10.
357
0.21
80.
178
0.11
70.
046
200
1.47
81.
251
1.17
90.
869
0.61
20.
540
0.37
30.
262
0.16
40.
046
500
0.24
60.
109
0.21
90.
197
0.17
20.
229
0.17
40.
127
0.08
10.
040
500
0.78
10.
714
0.61
40.
468
0.36
30.
325
0.23
90.
166
0.10
00.
040
500
100
7.42
54.
897
3.57
23.
447
2.35
51.
976
1.18
40.
587
0.37
40.
019
500
100
7.72
35.
218
3.88
13.
685
2.63
32.
223
1.45
10.
825
0.52
20.
019
200
2.63
72.
154
2.10
21.
495
0.94
00.
763
0.49
00.
275
0.15
60.
018
200
2.95
62.
505
2.39
71.
811
1.23
10.
984
0.67
00.
383
0.22
30.
018
500
0.62
30.
529
0.65
20.
399
0.30
50.
281
0.18
50.
121
0.06
40.
016
500
0.94
80.
917
0.95
30.
625
0.47
90.
370
0.24
20.
156
0.08
50.
016
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs:ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αT
Nα
100
100
-1.0
16-2
.571
-3.1
83-3
.319
-3.5
68-3
.479
-3.8
38-3
.998
-3.9
22-3
.952
100
100
1.36
02.
876
3.54
93.
762
4.03
13.
981
4.30
14.
455
4.37
84.
411
200
-1.3
20-2
.642
-3.0
32-3
.558
-3.7
45-3
.763
-3.9
18-3
.925
-4.0
03-4
.057
200
1.53
62.
902
3.39
83.
954
4.17
04.
208
4.35
84.
367
4.44
54.
497
500
-1.6
10-2
.849
-3.5
17-3
.890
-4.1
07-4
.124
-4.2
22-4
.304
-4.3
88-4
.439
500
1.76
93.
101
3.84
94.
273
4.51
24.
554
4.65
94.
740
4.82
24.
870
200
100
0.68
9-0
.109
-0.1
790.
088
-0.0
300.
166
-0.1
81-0
.329
-0.2
72-0
.336
200
100
0.79
40.
393
0.39
30.
369
0.32
60.
359
0.35
50.
440
0.39
80.
446
200
0.35
7-0
.089
0.11
5-0
.120
-0.1
77-0
.095
-0.2
27-0
.240
-0.3
01-0
.360
200
0.44
80.
292
0.30
10.
306
0.33
00.
284
0.35
40.
355
0.40
10.
453
500
0.20
9-0
.033
-0.0
95-0
.170
-0.2
15-0
.165
-0.2
25-0
.281
-0.3
37-0
.380
500
0.29
50.
245
0.27
40.
306
0.32
90.
296
0.33
50.
376
0.42
20.
463
500
100
1.04
00.
294
0.24
80.
565
0.43
60.
627
0.26
10.
081
0.11
5-0
.001
500
100
1.06
50.
332
0.27
20.
573
0.44
30.
634
0.27
60.
118
0.13
50.
002
200
0.68
50.
313
0.54
80.
324
0.25
30.
330
0.20
20.
162
0.08
6-0
.001
200
0.69
80.
326
0.55
10.
328
0.25
70.
332
0.20
70.
166
0.09
20.
001
500
0.52
70.
353
0.33
00.
249
0.20
40.
246
0.17
60.
117
0.05
5-0
.001
500
0.53
10.
355
0.33
10.
250
0.20
50.
246
0.17
60.
118
0.05
70.
001
TN
αT
Nα
100
100
1.09
20.
325
0.31
20.
674
0.49
60.
664
0.29
10.
130
0.18
70.
105
100
100
2.27
41.
834
1.49
21.
393
1.04
70.
958
0.59
70.
386
0.28
90.
105
200
0.15
7-0
.008
0.43
20.
286
0.25
10.
341
0.22
20.
201
0.14
80.
094
200
1.47
01.
363
1.16
30.
888
0.68
70.
586
0.41
80.
308
0.20
10.
094
500
-0.1
60-0
.033
0.16
20.
193
0.18
80.
257
0.20
40.
157
0.11
60.
081
500
1.11
00.
978
0.77
20.
587
0.45
70.
399
0.29
80.
215
0.14
20.
081
200
100
2.79
11.
573
1.13
21.
355
0.83
20.
878
0.41
40.
168
0.18
70.
051
200
100
3.39
52.
377
1.89
01.
911
1.28
01.
161
0.70
80.
414
0.30
50.
051
200
0.88
10.
432
0.67
90.
416
0.29
00.
345
0.21
10.
175
0.11
40.
046
200
1.46
31.
254
1.19
70.
864
0.62
30.
542
0.37
20.
261
0.16
20.
046
500
0.26
10.
128
0.24
00.
206
0.18
10.
237
0.18
00.
133
0.08
10.
040
500
0.80
10.
733
0.63
00.
473
0.36
90.
332
0.24
40.
174
0.10
10.
040
500
100
7.38
84.
904
3.52
93.
455
2.33
41.
980
1.18
80.
579
0.37
70.
019
500
100
7.71
45.
231
3.84
93.
696
2.62
02.
222
1.44
80.
829
0.52
70.
019
200
2.60
32.
107
2.04
61.
437
0.88
30.
723
0.45
90.
266
0.14
30.
018
200
2.93
22.
460
2.34
31.
753
1.17
60.
939
0.64
20.
370
0.20
90.
018
500
0.63
60.
531
0.64
40.
397
0.29
20.
274
0.18
30.
121
0.06
30.
016
500
0.95
50.
904
0.93
50.
620
0.47
20.
363
0.24
00.
156
0.08
40.
016
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
and
[Nαβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
,f jt
anduit∼IIDN
(0,1
),v ij∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
an
d
(33)
.α
isco
mp
ute
du
sin
gc p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/2
an
dp
=0.0
5in
the
mu
ltip
lete
stin
gp
roce
du
resh
own
in(2
3).α
corr
esp
on
ds
toth
em
ost
rob
ust
esti
mat
orof
the
exp
onen
tof
cros
s-se
ctio
nal
dep
end
ence
con
sid
ered
inB
ailey
etal.
(2016)
an
dall
ows
for
both
seri
al
corr
elati
on
inth
efa
ctors
an
dw
eak
cros
s-se
ctio
nal
dep
end
ence
inth
eer
ror
term
s.W
eu
sefo
ur
pri
nci
pal
com
pon
ents
wh
enes
tim
ati
ngc N
inth
eex
pre
ssio
nfo
rα
.T
he
num
ber
of
rep
lica
tion
sis
set
toR
=20
00.
37
Tab
leC
:B
ias
an
dR
MS
E(×
100)
ofα
esti
mate
sof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsu
sin
gth
em
axim
um
eige
nva
lue
ofth
eir
sam
ple
corr
elati
on
mat
rix
ina
stat
ican
dd
yn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gre
ssor
sC
ross
corr
elat
ion
sare
gen
erate
du
sin
gD
esig
ns
1an
d2
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Bia
sR
MSE
Des
ign
1:Sta
tic
model
-ϑi
=0
,fo
ri
=1,
2,...,N
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
100
100
-14.
080
-12.
332
-10.
881
-10.
088
-9.8
17-9
.406
-9.7
05-9
.782
-9.7
00-9
.706
100
100
14.1
3112
.391
10.9
4210
.152
9.88
99.
471
9.77
19.
845
9.76
39.
772
200
-11.
253
-9.6
74-8
.509
-8.3
38-8
.267
-8.1
89-8
.364
-8.3
52-8
.443
-8.5
3420
011
.291
9.72
38.
564
8.39
38.
323
8.24
58.
418
8.40
68.
496
8.58
850
0-7
.719
-6.6
71-6
.427
-6.4
69-6
.682
-6.7
50-6
.931
-7.0
69-7
.197
-7.2
8350
07.
751
6.71
06.
471
6.51
56.
728
6.79
66.
977
7.11
47.
239
7.32
820
010
0-1
5.98
7-1
3.47
9-1
1.63
4-1
0.52
0-1
0.07
6-9
.572
-9.8
06-9
.839
-9.6
50-9
.641
200
100
16.0
2413
.519
11.6
7410
.563
10.1
179.
614
9.84
39.
874
9.68
59.
676
200
-13.
357
-10.
916
-9.2
94-8
.883
-8.6
09-8
.368
-8.4
50-8
.371
-8.3
95-8
.375
200
13.3
8410
.946
9.32
78.
915
8.64
08.
400
8.47
98.
400
8.42
38.
402
500
-10.
018
-8.1
02-7
.384
-7.1
10-7
.032
-6.9
77-7
.013
-7.1
23-7
.159
-7.1
7750
010
.037
8.12
57.
408
7.13
57.
055
7.00
07.
038
7.14
87.
182
7.19
950
010
0-1
7.22
1-1
4.20
8-1
2.09
5-1
0.83
0-1
0.23
0-9
.670
-9.8
26-9
.780
-9.5
77-9
.513
500
100
17.2
4814
.236
12.1
2210
.858
10.2
569.
694
9.84
69.
800
9.59
69.
529
200
-14.
655
-11.
707
-9.7
62-9
.195
-8.7
99-8
.474
-8.5
01-8
.360
-8.3
37-8
.320
200
14.6
7511
.728
9.78
29.
213
8.81
68.
490
8.51
58.
373
8.35
08.
332
500
-11.
613
-9.0
02-7
.948
-7.4
76-7
.302
-7.0
85-7
.074
-7.0
95-7
.106
-7.1
1450
011
.626
9.01
57.
961
7.48
97.
314
7.09
67.
085
7.10
57.
115
7.12
4D
esig
n1:
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs-ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
100
100
-14.
273
-12.
513
-11.
143
-10.
300
-10.
022
-9.6
08-9
.905
-10.
049
-9.9
07-9
.935
100
100
14.3
2312
.572
11.2
1010
.368
10.0
949.
678
9.96
710
.112
9.97
19.
997
200
-11.
387
-9.7
44-8
.615
-8.5
08-8
.457
-8.3
29-8
.544
-8.5
36-8
.669
-8.7
2320
011
.425
9.79
28.
669
8.56
18.
512
8.38
48.
596
8.59
08.
725
8.77
650
0-7
.782
-6.7
13-6
.555
-6.6
33-6
.792
-6.8
89-7
.059
-7.2
31-7
.380
-7.4
0350
07.
813
6.75
36.
599
6.68
06.
836
6.93
47.
108
7.27
77.
426
7.44
620
010
0-1
6.10
9-1
3.52
4-1
1.71
7-1
0.65
7-1
0.16
9-9
.691
-9.8
70-9
.912
-9.7
64-9
.682
200
100
16.1
4413
.564
11.7
6010
.701
10.2
109.
733
9.90
79.
949
9.79
89.
718
200
-13.
395
-10.
991
-9.3
55-8
.997
-8.6
83-8
.451
-8.4
96-8
.453
-8.4
70-8
.438
200
13.4
2011
.022
9.38
89.
028
8.71
28.
482
8.52
78.
481
8.50
08.
464
500
-10.
067
-8.1
54-7
.470
-7.1
69-7
.125
-7.0
15-7
.111
-7.1
89-7
.205
-7.2
6950
010
.087
8.17
67.
493
7.19
37.
149
7.03
97.
134
7.21
27.
227
7.29
250
010
0-1
7.23
7-1
4.21
7-1
2.11
6-1
0.86
2-1
0.27
6-9
.736
-9.8
76-9
.808
-9.6
32-9
.554
500
100
17.2
6614
.245
12.1
4310
.889
10.3
029.
759
9.89
69.
826
9.65
09.
570
200
-14.
701
-11.
724
-9.7
74-9
.242
-8.8
28-8
.512
-8.5
08-8
.399
-8.3
83-8
.344
200
14.7
2111
.744
9.79
49.
260
8.84
48.
528
8.52
28.
412
8.39
58.
357
500
-11.
622
-9.0
37-7
.975
-7.5
25-7
.308
-7.1
10-7
.118
-7.1
31-7
.143
-7.1
4550
011
.635
9.05
17.
989
7.53
77.
320
7.12
17.
129
7.14
07.
152
7.15
4D
esig
n2:
Sta
tic
model
-ϑi
=0
,fo
ri
=1,
2,...,N
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
100
100
-20.
441
-19.
904
-19.
231
-18.
859
-18.
828
-18.
720
-19.
151
-19.
367
-19.
377
-19.
458
100
100
20.4
9419
.971
19.3
0618
.939
18.9
1118
.804
19.2
3519
.451
19.4
6119
.542
200
-16.
689
-16.
128
-15.
570
-15.
791
-15.
961
-16.
051
-16.
357
-16.
489
-16.
672
-16.
813
200
16.7
3616
.185
15.6
3615
.859
16.0
3016
.122
16.4
2716
.560
16.7
4316
.885
500
-12.
052
-11.
948
-12.
233
-12.
612
-13.
038
-13.
298
-13.
613
-13.
879
-14.
103
-14.
260
500
12.0
8311
.990
12.2
8212
.664
13.0
9213
.355
13.6
7013
.938
14.1
6214
.320
200
100
-23.
448
-21.
789
-20.
444
-19.
630
-19.
337
-19.
024
-19.
310
-19.
402
-19.
314
-19.
324
200
100
23.4
8821
.833
20.4
9119
.678
19.3
8419
.072
19.3
5619
.447
19.3
5919
.369
200
-19.
823
-18.
132
-16.
909
-16.
718
-16.
586
-16.
469
-16.
617
-16.
628
-16.
714
-16.
780
200
19.8
5318
.167
16.9
4716
.756
16.6
2416
.507
16.6
5416
.665
16.7
5116
.817
500
-15.
427
-14.
141
-13.
740
-13.
655
-13.
743
-13.
750
-13.
888
-14.
025
-14.
149
-14.
240
500
15.4
5014
.168
13.7
7013
.686
13.7
7413
.781
13.9
1914
.056
14.1
8014
.271
500
100
-25.
468
-23.
006
-21.
249
-20.
175
-19.
708
-19.
258
-19.
452
-19.
476
-19.
322
-19.
289
500
100
25.4
9623
.034
21.2
7720
.201
19.7
3219
.281
19.4
7519
.498
19.3
4219
.309
200
-21.
987
-19.
443
-17.
732
-17.
272
-16.
946
-16.
678
-16.
734
-16.
667
-16.
696
-16.
715
200
22.0
0819
.465
17.7
5317
.292
16.9
6416
.695
16.7
5116
.683
16.7
1216
.731
500
-17.
787
-15.
539
-14.
644
-14.
250
-14.
124
-13.
980
-14.
009
-14.
065
-14.
127
-14.
177
500
17.8
0115
.553
14.6
5714
.263
14.1
3713
.993
14.0
2214
.078
14.1
4014
.189
Des
ign
2:D
ynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs-ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
TN
αes
tim
ate
usi
ng
eige
nva
lue
appro
ach
100
100
-20.
516
-20.
025
-19.
395
-19.
051
-19.
063
-18.
960
-19.
387
-19.
614
-19.
633
-19.
726
100
100
20.5
6820
.092
19.4
6919
.130
19.1
4419
.043
19.4
6919
.696
19.7
1519
.809
200
-16.
722
-16.
214
-15.
701
-15.
947
-16.
112
-16.
236
-16.
543
-16.
686
-16.
880
-17.
026
200
16.7
6516
.269
15.7
6516
.015
16.1
8216
.307
16.6
1616
.759
16.9
5517
.100
500
-12.
067
-12.
003
-12.
324
-12.
736
-13.
174
-13.
452
-13.
782
-14.
057
-14.
287
-14.
454
500
12.0
9912
.047
12.3
7512
.791
13.2
3113
.511
13.8
4214
.119
14.3
4914
.516
200
100
-23.
502
-21.
871
-20.
547
-19.
757
-19.
473
-19.
155
-19.
429
-19.
532
-19.
454
-19.
459
200
100
23.5
4221
.917
20.5
9619
.807
19.5
2219
.204
19.4
7719
.578
19.5
0019
.505
200
-19.
885
-18.
203
-16.
982
-16.
815
-16.
671
-16.
562
-16.
722
-16.
742
-16.
825
-16.
896
200
19.9
1518
.238
17.0
1916
.853
16.7
1016
.600
16.7
6116
.780
16.8
6316
.934
500
-15.
445
-14.
156
-13.
756
-13.
679
-13.
765
-13.
786
-13.
930
-14.
068
-14.
191
-14.
277
500
15.4
6614
.182
13.7
8513
.708
13.7
9513
.816
13.9
6014
.098
14.2
2114
.307
500
100
-25.
472
-23.
054
-21.
306
-20.
220
-19.
759
-19.
304
-19.
518
-19.
538
-19.
386
-19.
356
500
100
25.5
0123
.085
21.3
3620
.249
19.7
8619
.329
19.5
4119
.559
19.4
0719
.376
200
-21.
991
-19.
433
-17.
726
-17.
280
-16.
950
-16.
703
-16.
762
-16.
700
-16.
729
-16.
755
200
22.0
1219
.455
17.7
4817
.300
16.9
6816
.721
16.7
7916
.716
16.7
4516
.770
500
-17.
821
-15.
585
-14.
692
-14.
280
-14.
152
-14.
010
-14.
044
-14.
095
-14.
160
-14.
209
500
17.8
3515
.600
14.7
0714
.295
14.1
6614
.024
14.0
5714
.108
14.1
7314
.222
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
For
det
ail
son
Des
ign
s1
an
d2
refe
rto
footn
ote
sof
Tab
les
A1a
an
dA
2a.
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=20
00.
38
Tab
leD
:C
over
age
(×100
)fo
rth
eα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
can
dd
yn
am
icp
anel
dat
am
od
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
2an
dp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
2an
dp
=0.
10Sta
tic
model
:ϑi
=0
fori
=1,
2,...,N
TN
TN
100
100
0.00
20.0
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
65.9
510
010
00.
0027
.35
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0053
.70
200
0.00
89.3
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
92.3
020
00.
0090
.15
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0085
.30
500
0.00
99.7
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
99.6
050
00.
0099
.70
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0098
.90
200
100
0.00
19.8
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.15
200
100
0.00
25.0
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
200
0.00
88.4
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.65
200
0.00
88.8
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.25
500
0.00
99.9
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
6.35
500
0.00
99.9
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
2.85
500
100
0.00
20.9
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
100
0.00
28.2
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
200
0.00
88.9
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
200
0.00
90.2
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
0.00
99.7
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
0.00
99.9
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs:ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
TN
100
100
0.00
18.3
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
69.9
510
010
00.
0024
.35
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0058
.20
200
0.00
86.7
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
94.6
020
00.
0088
.20
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0089
.60
500
0.00
99.6
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
99.8
550
00.
0099
.70
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0099
.65
200
100
0.00
20.7
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.10
200
100
0.00
26.9
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.10
200
0.00
88.9
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.70
200
0.00
90.0
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.30
500
0.00
99.7
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
7.60
500
0.00
99.8
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
4.25
500
100
0.00
21.5
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
100
0.00
28.0
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
200
0.00
88.2
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
200
0.00
89.4
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
0.00
99.7
010
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
500
0.00
99.5
510
0.00
100.
0010
0.00
100.
0010
0.00
100.
0010
0.00
0.00
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efirs
tNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
mat
rix
ofth
eer
rors
,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).α
an
dit
sem
pir
ical
dis
trib
uti
on
are
com
pu
ted
usi
ngc p
(n,δ
)w
ithn
=N
(N−
1)/2,δ
=1/
2an
dp
=0.
05,
0.10
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23).
Cov
erage
refe
rsto
the
freq
uen
cyat
wh
ichα
(r),B
0.0
5<α
(r)<α
(r),B
0.9
5,
wh
erer
=1,
2,...,R
andb
=1,
2,...,B
.T
he
nu
mb
erof
rep
lica
tions
isse
ttoR
=2000
an
dth
enu
mb
erof
boots
trap
sis
set
toB
=50
0.
39
An Online Supplement for
Exponent of Cross-sectional Dependence for Residuals
Natalia BaileyDepartment of Econometrics and Business Statistics,
Monash University
George KapetaniosKing’s Business School,King’s College London
M. Hashem PesaranDepartment of Economics,
University of Southern California and Trinity College, Cambridge
April 4, 2019
This online supplement provides additional Monte Carlo and empirical results.
Appendix A
Additional Monte Carlo results
The Monte Carlo results provided in the tables below are based on the designs set out in Section6 of the paper.
1
Tab
leS
1a:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
no
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
00.
211
-0.6
18-0
.595
-0.3
57-0
.287
-0.0
02-0
.259
-0.2
81-0
.094
-0.0
1710
010
00.
639
-0.3
27-0
.415
-0.2
46-0
.216
0.04
3-0
.230
-0.2
62-0
.083
-0.0
0920
00.
074
-0.4
24-0
.155
-0.3
02-0
.257
-0.1
02-0
.166
-0.0
81-0
.064
-0.0
2920
00.
406
-0.2
14-0
.030
-0.2
25-0
.207
-0.0
70-0
.144
-0.0
66-0
.050
-0.0
1750
00.
087
-0.1
21-0
.164
-0.1
49-0
.152
-0.0
36-0
.043
-0.0
53-0
.051
-0.0
4250
00.
331
0.02
0-0
.081
-0.0
96-0
.117
-0.0
10-0
.022
-0.0
34-0
.035
-0.0
2520
010
00.
267
-0.5
79-0
.553
-0.3
29-0
.270
0.01
7-0
.241
-0.2
63-0
.078
0.00
020
010
00.
709
-0.2
85-0
.366
-0.2
20-0
.204
0.05
7-0
.218
-0.2
51-0
.074
0.00
020
00.
124
-0.3
85-0
.117
-0.2
71-0
.226
-0.0
74-0
.139
-0.0
56-0
.035
0.00
020
00.
471
-0.1
630.
006
-0.1
97-0
.183
-0.0
51-0
.127
-0.0
50-0
.033
0.00
050
00.
153
-0.0
64-0
.112
-0.1
01-0
.106
0.00
7-0
.001
-0.0
07-0
.011
0.00
050
00.
416
0.08
1-0
.033
-0.0
59-0
.084
0.01
80.
005
-0.0
04-0
.010
0.00
050
010
00.
282
-0.5
55-0
.549
-0.3
23-0
.264
0.02
0-0
.239
-0.2
62-0
.078
0.00
050
010
00.
734
-0.2
45-0
.358
-0.2
08-0
.196
0.06
0-0
.216
-0.2
49-0
.073
0.00
020
00.
154
-0.3
63-0
.107
-0.2
66-0
.224
-0.0
72-0
.138
-0.0
56-0
.035
0.00
020
00.
520
-0.1
350.
023
-0.1
91-0
.181
-0.0
48-0
.125
-0.0
50-0
.032
0.00
050
00.
188
-0.0
48-0
.104
-0.0
97-0
.103
0.00
90.
000
-0.0
06-0
.011
0.00
050
00.
468
0.10
9-0
.020
-0.0
52-0
.079
0.02
10.
006
-0.0
04-0
.010
0.00
0T
NR
MSE
TN
RM
SE
100
100
0.32
30.
640
0.60
80.
367
0.29
60.
055
0.26
30.
284
0.10
10.
041
100
100
0.72
00.
403
0.44
70.
271
0.23
20.
074
0.23
50.
264
0.08
70.
025
200
0.15
60.
436
0.17
40.
308
0.26
40.
117
0.17
50.
096
0.08
80.
061
200
0.44
90.
251
0.09
40.
234
0.21
40.
083
0.14
90.
074
0.06
50.
038
500
0.11
70.
140
0.18
00.
165
0.16
80.
085
0.08
50.
099
0.08
40.
078
500
0.34
40.
070
0.10
00.
109
0.12
70.
054
0.05
50.
067
0.05
70.
051
200
100
0.35
90.
602
0.56
60.
339
0.27
50.
043
0.24
20.
263
0.07
90.
000
200
100
0.78
60.
379
0.40
70.
252
0.21
80.
082
0.22
20.
252
0.07
50.
000
200
0.18
70.
396
0.13
10.
274
0.22
70.
076
0.14
00.
057
0.03
50.
000
200
0.51
00.
209
0.08
80.
206
0.18
80.
058
0.12
80.
052
0.03
30.
000
500
0.16
80.
075
0.11
50.
102
0.10
60.
009
0.00
40.
007
0.01
10.
000
500
0.42
70.
098
0.04
90.
063
0.08
50.
021
0.00
80.
006
0.01
10.
000
500
100
0.37
90.
586
0.56
20.
333
0.27
00.
044
0.24
10.
262
0.07
80.
000
500
100
0.81
30.
360
0.39
90.
240
0.21
30.
084
0.22
10.
251
0.07
50.
000
200
0.21
00.
376
0.12
40.
269
0.22
50.
075
0.13
80.
057
0.03
50.
000
200
0.55
60.
192
0.09
20.
200
0.18
50.
056
0.12
60.
051
0.03
30.
000
500
0.20
20.
063
0.10
80.
098
0.10
40.
011
0.00
40.
007
0.01
10.
000
500
0.47
90.
124
0.04
20.
056
0.08
10.
023
0.00
90.
006
0.01
00.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.51
50.
274
-0.0
30-0
.008
-0.0
680.
129
-0.1
77-0
.232
-0.0
69-0
.005
100
100
3.09
91.
393
0.69
90.
444
0.21
50.
296
-0.0
79-0
.180
-0.0
49-0
.002
200
1.45
30.
459
0.36
30.
014
-0.0
650.
010
-0.0
98-0
.039
-0.0
34-0
.007
200
3.00
51.
495
0.98
80.
398
0.15
70.
135
-0.0
30-0
.004
-0.0
19-0
.004
500
1.54
40.
723
0.31
50.
123
0.00
70.
061
0.02
1-0
.003
-0.0
15-0
.009
500
3.05
21.
650
0.85
30.
422
0.16
90.
147
0.06
60.
020
-0.0
04-0
.005
200
100
1.60
40.
335
0.02
40.
022
-0.0
610.
142
-0.1
69-0
.225
-0.0
650.
000
200
100
3.16
31.
461
0.75
20.
480
0.22
00.
306
-0.0
73-0
.176
-0.0
460.
000
200
1.55
20.
528
0.40
80.
042
-0.0
450.
024
-0.0
86-0
.031
-0.0
260.
000
200
3.10
31.
570
1.03
70.
425
0.18
00.
149
-0.0
190.
002
-0.0
140.
000
500
1.68
40.
816
0.37
30.
160
0.03
20.
077
0.03
50.
009
-0.0
060.
000
500
3.21
71.
765
0.92
30.
464
0.19
60.
162
0.07
70.
028
0.00
10.
000
500
100
1.63
30.
378
0.03
60.
033
-0.0
500.
146
-0.1
67-0
.224
-0.0
630.
000
500
100
3.21
11.
498
0.77
50.
491
0.23
40.
313
-0.0
71-0
.173
-0.0
440.
000
200
1.62
10.
574
0.43
30.
054
-0.0
410.
029
-0.0
84-0
.029
-0.0
250.
000
200
3.18
31.
620
1.06
80.
440
0.18
20.
155
-0.0
160.
004
-0.0
130.
000
500
1.77
50.
869
0.40
00.
175
0.04
20.
082
0.03
70.
010
-0.0
050.
000
500
3.32
21.
833
0.96
10.
485
0.21
00.
168
0.08
00.
030
0.00
20.
000
TN
RM
SE
TN
RM
SE
100
100
1.57
90.
437
0.23
90.
168
0.14
00.
156
0.18
80.
236
0.07
40.
014
100
100
3.14
81.
464
0.77
50.
509
0.28
20.
323
0.12
20.
191
0.06
10.
008
200
1.48
20.
503
0.38
70.
097
0.09
50.
049
0.10
40.
047
0.04
20.
019
200
3.02
71.
518
1.00
60.
422
0.18
60.
151
0.05
60.
033
0.02
80.
011
500
1.55
20.
730
0.32
20.
132
0.03
70.
068
0.03
40.
029
0.02
50.
022
500
3.05
81.
656
0.85
80.
427
0.17
50.
150
0.07
00.
029
0.01
40.
013
200
100
1.67
00.
487
0.25
50.
181
0.13
40.
168
0.18
10.
230
0.06
90.
000
200
100
3.21
41.
531
0.82
40.
543
0.28
30.
334
0.12
20.
188
0.05
90.
000
200
1.58
00.
568
0.43
10.
108
0.08
40.
053
0.09
20.
037
0.02
90.
000
200
3.12
31.
592
1.05
50.
449
0.20
80.
166
0.05
40.
032
0.02
40.
000
500
1.69
10.
822
0.37
90.
167
0.04
40.
080
0.03
80.
013
0.00
70.
000
500
3.22
21.
770
0.92
80.
470
0.20
10.
165
0.08
00.
031
0.00
80.
000
500
100
1.70
10.
520
0.25
30.
180
0.13
40.
171
0.18
00.
228
0.06
80.
000
500
100
3.26
11.
568
0.84
60.
553
0.29
80.
341
0.12
20.
186
0.05
90.
000
200
1.64
70.
612
0.45
80.
114
0.08
00.
057
0.09
00.
037
0.02
80.
000
200
3.20
31.
643
1.08
60.
464
0.20
80.
172
0.05
30.
036
0.02
30.
000
500
1.78
30.
876
0.40
70.
181
0.05
40.
085
0.04
00.
014
0.00
70.
000
500
3.32
81.
838
0.96
60.
490
0.21
60.
171
0.08
30.
033
0.00
80.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:
ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi
=0,
for
i=
1,2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vec
tor
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
alva
lue
use
din
the
mu
ltip
lete
stin
gpro
ced
ure
show
nin
(23).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
2
Tab
leS
1b:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
no
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
Bia
s10
010
0-0
.928
-2.4
37-3
.044
-3.1
37-3
.332
-3.2
47-3
.622
-3.7
52-3
.669
-3.6
9310
010
0-0
.125
-1.5
82-2
.137
-2.2
06-2
.403
-2.3
21-2
.710
-2.8
51-2
.777
-2.8
1620
0-1
.276
-2.5
35-2
.883
-3.3
50-3
.548
-3.5
38-3
.722
-3.7
28-3
.791
-3.8
3720
0-0
.595
-1.7
63-2
.057
-2.4
99-2
.683
-2.6
70-2
.854
-2.8
66-2
.932
-2.9
8350
0-1
.531
-2.7
37-3
.388
-3.7
23-3
.923
-3.9
37-4
.023
-4.0
94-4
.174
-4.2
3950
0-0
.941
-2.0
29-2
.601
-2.8
97-3
.078
-3.0
84-3
.167
-3.2
37-3
.316
-3.3
8020
010
00.
716
-0.0
78-0
.151
0.11
20.
008
0.20
1-0
.158
-0.3
07-0
.249
-0.3
1220
010
01.
211
0.29
80.
129
0.33
90.
188
0.35
6-0
.018
-0.1
83-0
.136
-0.2
1220
00.
368
-0.0
730.
136
-0.0
88-0
.145
-0.0
67-0
.198
-0.2
12-0
.273
-0.3
3120
00.
787
0.23
50.
362
0.09
30.
004
0.06
3-0
.077
-0.1
02-0
.168
-0.2
3150
00.
219
-0.0
26-0
.082
-0.1
54-0
.199
-0.1
48-0
.210
-0.2
65-0
.321
-0.3
6550
00.
546
0.20
90.
097
-0.0
09-0
.073
-0.0
33-0
.102
-0.1
61-0
.218
-0.2
6350
010
01.
044
0.29
80.
250
0.56
60.
432
0.62
90.
260
0.08
60.
113
0.00
050
010
01.
464
0.58
10.
431
0.67
60.
501
0.66
90.
284
0.10
00.
119
0.00
020
00.
674
0.31
00.
547
0.32
50.
254
0.32
90.
203
0.16
20.
085
-0.0
0120
01.
022
0.52
40.
672
0.39
90.
298
0.35
50.
218
0.17
10.
089
0.00
050
00.
529
0.35
50.
331
0.24
90.
205
0.24
70.
176
0.11
70.
056
-0.0
0150
00.
797
0.50
40.
412
0.29
40.
230
0.26
00.
184
0.12
10.
058
0.00
0T
NR
MSE
TN
RM
SE
100
100
1.30
22.
747
3.43
03.
585
3.79
23.
754
4.09
24.
217
4.14
04.
162
100
100
0.85
01.
930
2.51
82.
637
2.82
92.
789
3.13
03.
258
3.19
03.
230
200
1.49
42.
797
3.25
23.
742
3.96
23.
974
4.15
24.
164
4.22
14.
265
200
0.92
82.
042
2.42
32.
862
3.05
83.
064
3.23
63.
253
3.31
33.
363
500
1.70
43.
009
3.73
84.
119
4.34
04.
378
4.46
74.
532
4.61
34.
675
500
1.16
02.
306
2.93
63.
264
3.45
63.
484
3.56
73.
630
3.71
03.
772
200
100
0.81
00.
360
0.35
70.
337
0.30
80.
355
0.32
70.
409
0.36
50.
405
200
100
1.27
70.
449
0.30
50.
427
0.31
00.
425
0.22
20.
277
0.24
40.
286
200
0.45
90.
300
0.31
30.
276
0.29
50.
260
0.31
90.
325
0.37
20.
419
200
0.83
10.
341
0.42
70.
226
0.19
80.
203
0.20
70.
212
0.25
40.
301
500
0.30
10.
237
0.25
80.
288
0.31
00.
280
0.31
70.
357
0.40
40.
446
500
0.57
40.
280
0.21
40.
188
0.19
60.
184
0.20
80.
243
0.28
70.
329
500
100
1.06
90.
339
0.27
40.
575
0.44
00.
635
0.27
50.
120
0.13
40.
002
500
100
1.50
00.
623
0.45
70.
686
0.50
90.
675
0.29
80.
130
0.13
90.
001
200
0.68
70.
322
0.55
00.
328
0.25
70.
332
0.20
80.
167
0.09
20.
001
200
1.03
90.
538
0.67
60.
403
0.30
10.
357
0.22
20.
175
0.09
50.
001
500
0.53
40.
357
0.33
20.
250
0.20
60.
247
0.17
70.
118
0.05
70.
002
500
0.80
30.
507
0.41
40.
295
0.23
00.
261
0.18
40.
122
0.05
90.
001
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
1.11
5-0
.445
-1.0
89-1
.228
-1.4
93-1
.453
-1.8
71-2
.042
-1.9
89-2
.049
100
100
2.94
21.
087
0.16
1-0
.183
-0.6
08-0
.668
-1.1
53-1
.376
-1.3
63-1
.451
200
0.91
3-0
.396
-0.8
12-1
.348
-1.5
94-1
.626
-1.8
32-1
.870
-1.9
49-2
.012
200
2.74
71.
058
0.32
5-0
.425
-0.8
13-0
.936
-1.1
94-1
.272
-1.3
71-1
.451
500
0.89
1-0
.383
-1.0
88-1
.487
-1.7
25-1
.771
-1.8
77-1
.958
-2.0
42-2
.106
500
2.69
20.
986
-0.0
41-0
.655
-1.0
21-1
.144
-1.2
91-1
.395
-1.4
90-1
.559
200
100
2.12
30.
960
0.58
70.
663
0.42
50.
534
0.12
4-0
.068
-0.0
40-0
.134
200
100
3.67
12.
077
1.34
91.
164
0.76
20.
760
0.28
60.
045
0.03
8-0
.083
200
1.89
20.
989
0.85
50.
434
0.24
80.
248
0.07
60.
025
-0.0
57-0
.132
200
3.42
52.
029
1.50
80.
860
0.52
00.
426
0.20
00.
112
0.00
8-0
.083
500
1.87
01.
040
0.62
50.
340
0.17
70.
159
0.06
0-0
.015
-0.0
81-0
.133
500
3.39
62.
001
1.20
30.
685
0.38
70.
292
0.15
10.
052
-0.0
26-0
.086
500
100
2.30
51.
161
0.80
10.
900
0.64
10.
752
0.33
30.
126
0.12
90.
000
500
100
3.78
02.
216
1.48
71.
324
0.90
80.
908
0.42
50.
174
0.14
80.
000
200
2.06
51.
182
1.05
80.
633
0.43
40.
431
0.26
20.
192
0.09
80.
000
200
3.56
92.
180
1.65
71.
001
0.65
00.
553
0.32
90.
225
0.11
00.
000
500
2.06
21.
231
0.81
50.
514
0.34
90.
323
0.21
60.
137
0.06
40.
000
500
3.56
42.
157
1.34
90.
813
0.51
20.
408
0.25
90.
157
0.07
10.
000
TN
RM
SE
RM
SE
100
100
1.35
51.
026
1.52
71.
678
1.90
21.
893
2.23
82.
390
2.34
42.
402
100
100
3.02
71.
330
0.85
50.
895
1.08
61.
142
1.48
71.
674
1.66
61.
745
200
1.09
00.
883
1.25
31.
696
1.92
41.
967
2.15
12.
189
2.26
22.
323
200
2.79
41.
222
0.79
10.
889
1.15
81.
266
1.47
61.
548
1.63
61.
711
500
1.01
70.
848
1.43
91.
816
2.04
12.
099
2.19
62.
269
2.35
12.
414
500
2.71
81.
134
0.71
21.
035
1.32
91.
447
1.57
31.
663
1.75
41.
821
200
100
2.17
31.
024
0.64
50.
700
0.47
00.
566
0.21
40.
173
0.15
50.
189
200
100
3.71
12.
118
1.38
21.
187
0.78
50.
777
0.32
00.
135
0.12
20.
122
200
1.91
51.
016
0.87
50.
460
0.28
50.
281
0.15
10.
126
0.13
70.
181
200
3.44
22.
045
1.51
90.
872
0.53
30.
439
0.22
40.
144
0.09
00.
119
500
1.87
81.
048
0.63
70.
358
0.20
60.
190
0.11
90.
105
0.13
50.
175
500
3.40
12.
006
1.20
80.
690
0.39
50.
301
0.16
80.
090
0.08
00.
118
500
100
2.34
21.
199
0.82
60.
911
0.65
10.
758
0.34
60.
152
0.14
90.
001
500
100
3.81
52.
250
1.51
31.
339
0.91
90.
915
0.43
70.
196
0.16
60.
000
200
2.08
21.
196
1.06
40.
638
0.43
80.
434
0.26
50.
197
0.10
30.
000
200
3.58
42.
193
1.66
41.
007
0.65
50.
556
0.33
30.
230
0.11
60.
000
500
2.06
81.
234
0.81
70.
515
0.35
00.
323
0.21
70.
138
0.06
50.
000
500
3.56
82.
160
1.35
10.
815
0.51
30.
408
0.26
00.
158
0.07
20.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:
ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi
=0,
for
i=
1,2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
an
d[N
αβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3,
wh
ereαβ
rela
tes
toα
un
der
(11)
,f jt
anduit∼IIDN
(0,1
),v ij∼IIDU
(µvj−
0.2,µvj
+0.
2),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in
(32)
and
(33)
.T
he
num
ber
ofre
pli
cati
ons
isse
ttoR
=2000.
3
Tab
leS
1c:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
om
ad
yn
am
icp
anel
dat
am
od
elw
ith
no
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n1
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
02.
253
0.81
20.
371
0.20
10.
044
0.18
8-0
.164
-0.2
50-0
.103
-0.0
5010
010
03.
623
1.81
61.
052
0.64
20.
327
0.36
4-0
.053
-0.1
84-0
.067
-0.0
3220
02.
107
0.94
40.
664
0.16
2-0
.008
0.03
0-0
.108
-0.0
80-0
.085
-0.0
6620
03.
396
1.84
01.
242
0.52
30.
214
0.16
6-0
.021
-0.0
26-0
.052
-0.0
4450
02.
050
1.04
30.
488
0.20
40.
012
0.03
4-0
.040
-0.0
72-0
.096
-0.0
9250
03.
226
1.80
20.
952
0.48
30.
183
0.14
00.
026
-0.0
27-0
.061
-0.0
6420
010
02.
761
1.16
40.
575
0.38
50.
144
0.26
9-0
.096
-0.1
83-0
.049
0.00
020
010
04.
226
2.26
21.
302
0.85
40.
432
0.44
10.
005
-0.1
30-0
.028
0.00
020
02.
628
1.26
80.
853
0.33
70.
136
0.10
9-0
.028
-0.0
07-0
.016
0.00
020
04.
016
2.24
21.
454
0.71
80.
363
0.23
50.
042
0.02
7-0
.003
0.00
050
02.
659
1.46
80.
747
0.39
60.
170
0.15
90.
071
0.02
80.
000
0.00
050
03.
961
2.31
71.
259
0.69
20.
334
0.24
80.
114
0.04
80.
007
0.00
050
010
03.
035
1.32
10.
722
0.44
40.
218
0.29
4-0
.071
-0.1
74-0
.043
0.00
050
010
04.
588
2.46
21.
491
0.94
20.
531
0.47
50.
038
-0.1
17-0
.021
0.00
020
02.
930
1.52
91.
049
0.41
10.
180
0.14
6-0
.019
-0.0
03-0
.013
0.00
020
04.
403
2.56
71.
713
0.82
40.
427
0.28
60.
057
0.03
20.
001
0.00
050
03.
062
1.71
70.
888
0.49
20.
215
0.16
80.
076
0.02
80.
001
0.00
050
04.
461
2.64
81.
446
0.82
00.
396
0.26
10.
123
0.05
00.
008
0.00
0T
NR
MSE
TN
RM
SE
100
100
2.88
81.
621
1.23
60.
625
0.56
10.
382
0.27
40.
283
0.14
00.
100
100
100
4.18
32.
465
1.73
40.
993
0.73
50.
551
0.27
20.
235
0.10
70.
069
200
2.85
11.
818
1.22
10.
717
0.40
10.
401
0.28
30.
147
0.13
20.
117
200
4.04
12.
563
1.73
60.
981
0.54
10.
472
0.29
20.
125
0.09
30.
082
500
2.84
11.
783
1.05
00.
679
0.33
10.
282
0.19
20.
166
0.16
40.
163
500
3.92
22.
464
1.46
90.
917
0.44
90.
351
0.18
90.
131
0.11
80.
120
200
100
3.39
91.
838
1.15
60.
836
0.37
50.
398
0.20
60.
226
0.06
70.
001
200
100
4.76
22.
789
1.75
91.
229
0.62
30.
574
0.23
10.
207
0.06
80.
001
200
3.17
31.
903
1.24
90.
742
0.57
30.
215
0.24
50.
069
0.04
40.
001
200
4.48
72.
766
1.81
01.
066
0.73
30.
335
0.27
90.
094
0.05
10.
000
500
3.21
22.
004
1.10
10.
750
0.46
30.
339
0.17
70.
105
0.02
80.
003
500
4.45
32.
797
1.59
41.
028
0.60
90.
428
0.21
70.
127
0.03
50.
002
500
100
3.41
11.
715
1.18
60.
728
0.42
70.
380
0.21
80.
210
0.08
70.
000
500
100
4.90
42.
765
1.84
71.
171
0.70
20.
565
0.25
20.
187
0.09
20.
000
200
3.31
22.
098
1.43
20.
749
0.42
30.
264
0.10
80.
056
0.03
30.
000
200
4.71
73.
015
2.04
81.
102
0.62
90.
394
0.14
80.
079
0.04
00.
000
500
3.46
82.
134
1.18
20.
818
0.45
40.
249
0.11
50.
052
0.01
50.
000
500
4.80
73.
004
1.71
11.
117
0.61
60.
346
0.16
50.
075
0.02
20.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
5.72
23.
405
2.16
01.
368
0.79
30.
652
0.12
0-0
.086
-0.0
23-0
.019
100
100
8.46
75.
599
3.74
42.
451
1.50
21.
085
0.38
40.
059
0.03
7-0
.011
200
6.09
93.
823
2.55
41.
360
0.72
60.
473
0.16
00.
072
-0.0
03-0
.023
200
8.83
15.
954
4.05
12.
362
1.35
40.
849
0.37
80.
182
0.04
4-0
.014
500
6.68
34.
203
2.49
41.
419
0.73
80.
455
0.20
00.
070
-0.0
03-0
.030
500
9.37
36.
256
3.92
12.
329
1.28
90.
763
0.36
30.
151
0.03
3-0
.019
200
100
6.36
13.
929
2.46
01.
621
0.91
60.
731
0.17
7-0
.038
0.00
80.
000
200
100
9.12
96.
167
4.08
22.
744
1.64
21.
179
0.44
40.
106
0.06
40.
000
200
6.81
84.
315
2.81
61.
609
0.90
60.
542
0.21
30.
111
0.02
80.
000
200
9.56
46.
471
4.34
22.
652
1.56
50.
926
0.43
10.
219
0.06
80.
000
500
7.56
94.
881
2.92
21.
698
0.91
10.
561
0.27
20.
121
0.03
30.
000
500
10.2
536.
974
4.40
82.
656
1.48
80.
881
0.43
80.
198
0.06
10.
000
500
100
6.81
94.
181
2.70
01.
743
1.04
50.
780
0.22
3-0
.020
0.01
70.
000
500
100
9.61
16.
456
4.36
72.
898
1.81
21.
239
0.50
70.
129
0.07
60.
000
200
7.29
54.
739
3.16
51.
761
1.00
70.
616
0.24
30.
119
0.03
40.
000
200
10.0
766.
945
4.75
22.
843
1.70
11.
021
0.47
30.
229
0.07
60.
000
500
8.17
45.
343
3.20
71.
906
1.02
10.
593
0.29
10.
127
0.03
50.
000
500
10.8
617.
477
4.73
62.
913
1.63
00.
929
0.46
60.
208
0.06
40.
000
TN
RM
SE
TN
RM
SE
100
100
6.18
03.
909
2.68
91.
670
1.11
90.
835
0.35
80.
202
0.09
30.
045
100
100
8.81
65.
980
4.15
22.
713
1.76
61.
264
0.57
10.
247
0.11
80.
030
200
6.58
14.
355
2.96
21.
722
0.98
80.
713
0.39
50.
169
0.06
90.
048
200
9.19
06.
355
4.38
82.
661
1.58
31.
056
0.57
50.
266
0.09
10.
031
500
7.16
04.
694
2.91
51.
783
0.97
10.
647
0.31
40.
153
0.06
90.
065
500
9.72
66.
639
4.27
12.
646
1.50
80.
950
0.46
90.
222
0.07
60.
046
200
100
6.78
64.
339
2.81
91.
929
1.08
10.
866
0.34
70.
206
0.08
50.
001
200
100
9.43
96.
473
4.36
22.
994
1.78
91.
310
0.58
00.
273
0.12
80.
000
200
7.17
04.
700
3.10
71.
889
1.19
10.
640
0.39
90.
170
0.07
40.
000
200
9.82
46.
760
4.58
02.
885
1.80
01.
023
0.58
40.
275
0.11
00.
000
500
7.90
85.
236
3.20
01.
981
1.14
50.
732
0.36
80.
198
0.06
00.
001
500
10.5
047.
252
4.64
02.
901
1.69
61.
043
0.52
90.
271
0.08
80.
001
500
100
7.06
94.
413
2.97
41.
931
1.19
40.
873
0.38
00.
185
0.11
20.
000
500
100
9.79
76.
629
4.57
53.
057
1.94
11.
330
0.63
10.
263
0.15
50.
000
200
7.52
25.
049
3.42
71.
974
1.17
10.
713
0.31
20.
158
0.06
60.
000
200
10.2
417.
171
4.96
03.
014
1.84
11.
112
0.53
50.
266
0.10
60.
000
500
8.40
15.
589
3.41
32.
142
1.20
00.
681
0.33
70.
154
0.04
90.
000
500
11.0
267.
665
4.90
53.
110
1.78
71.
015
0.51
30.
236
0.07
80.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:
ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi
=0,
for
i=
1,2,...,N
.N
on-G
auss
ian
erro
rsar
ege
ner
ated
asuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
h
v=
8d
egre
esof
free
dom
,in
(31)
.D
esig
n1
assu
mesb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,2,...,N
,in
the
const
ruct
ion
of
the
corr
elat
ion
mat
rix
ofth
eer
rors
,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
al
valu
eu
sed
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23)
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
4
Tab
leS
1d:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
no
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
Bia
s10
010
01.
186
-0.8
98-2
.047
-2.5
70-3
.052
-3.1
25-3
.605
-3.8
14-3
.780
-3.8
2010
010
02.
840
0.56
4-0
.764
-1.4
30-2
.018
-2.1
63-2
.693
-2.9
35-2
.926
-2.9
8920
00.
953
-1.0
04-1
.983
-2.8
65-3
.320
-3.4
93-3
.754
-3.8
05-3
.909
-3.9
7020
02.
530
0.36
8-0
.797
-1.8
11-2
.361
-2.5
95-2
.894
-2.9
73-3
.087
-3.1
6050
00.
712
-1.1
98-2
.406
-3.1
40-3
.569
-3.7
22-3
.886
-4.0
08-4
.101
-4.1
5950
02.
201
0.07
3-1
.310
-2.1
73-2
.686
-2.8
89-3
.081
-3.2
15-3
.314
-3.3
7520
010
02.
816
1.27
90.
648
0.53
10.
198
0.24
1-0
.187
-0.3
70-0
.345
-0.4
2420
010
04.
235
2.34
91.
391
1.04
30.
563
0.50
80.
021
-0.2
05-0
.205
-0.3
0420
02.
495
1.21
50.
809
0.25
5-0
.007
-0.0
49-0
.242
-0.2
91-0
.368
-0.4
3420
03.
846
2.17
11.
441
0.68
60.
290
0.16
8-0
.070
-0.1
49-0
.241
-0.3
1750
02.
592
1.33
00.
625
0.17
4-0
.084
-0.1
47-0
.266
-0.3
50-0
.419
-0.4
6950
03.
906
2.21
21.
194
0.54
40.
168
0.03
6-0
.118
-0.2
20-0
.299
-0.3
5250
010
03.
405
1.87
31.
221
1.12
50.
752
0.79
80.
355
0.12
40.
127
-0.0
0250
010
04.
805
2.87
21.
872
1.52
00.
993
0.93
30.
432
0.16
30.
142
-0.0
0120
03.
125
1.82
41.
398
0.81
30.
519
0.46
30.
266
0.18
60.
094
-0.0
0220
04.
487
2.73
61.
948
1.14
30.
708
0.56
50.
320
0.21
20.
104
-0.0
0150
03.
153
1.87
61.
159
0.68
60.
425
0.35
00.
222
0.13
40.
059
-0.0
0250
04.
471
2.71
31.
649
0.95
80.
569
0.42
20.
256
0.14
90.
065
-0.0
01T
NR
MSE
TN
RM
SE
100
100
2.40
92.
073
2.71
83.
131
3.55
63.
628
4.05
94.
266
4.24
34.
287
100
100
3.64
42.
010
1.86
02.
119
2.53
92.
647
3.10
73.
338
3.34
13.
408
200
2.28
72.
057
2.59
93.
322
3.75
33.
940
4.19
74.
254
4.35
84.
418
200
3.40
41.
901
1.77
12.
336
2.78
63.
011
3.29
63.
377
3.49
23.
564
500
2.16
92.
100
2.89
93.
554
3.97
24.
144
4.31
94.
442
4.53
84.
595
500
3.16
01.
817
2.02
12.
622
3.07
13.
278
3.47
63.
611
3.71
23.
773
200
100
3.28
11.
792
1.11
00.
820
0.50
90.
451
0.40
10.
507
0.48
70.
548
200
100
4.63
72.
751
1.72
81.
249
0.73
70.
609
0.28
90.
339
0.33
70.
406
200
2.86
71.
563
1.02
60.
508
0.34
90.
321
0.40
40.
434
0.49
30.
549
200
4.17
92.
465
1.62
30.
842
0.44
10.
309
0.26
20.
291
0.35
20.
413
500
3.18
41.
857
1.05
70.
575
0.37
80.
334
0.39
20.
459
0.52
00.
566
500
4.41
22.
662
1.54
80.
824
0.41
50.
263
0.25
70.
320
0.38
50.
434
500
100
3.76
42.
240
1.49
81.
264
0.84
00.
829
0.38
50.
159
0.14
60.
009
500
100
5.11
93.
185
2.11
81.
660
1.08
40.
970
0.46
50.
196
0.16
10.
008
200
3.40
22.
053
1.51
70.
892
0.56
10.
477
0.27
50.
192
0.10
00.
004
200
4.73
62.
948
2.07
51.
231
0.75
70.
584
0.33
00.
217
0.11
00.
003
500
3.51
82.
167
1.35
70.
815
0.50
00.
378
0.23
30.
138
0.06
10.
005
500
4.79
22.
983
1.84
41.
089
0.64
80.
454
0.27
00.
154
0.06
70.
004
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
5.16
12.
533
0.85
2-0
.095
-0.9
01-1
.189
-1.8
10-2
.115
-2.1
49-2
.246
100
100
8.03
44.
962
2.79
91.
420
0.29
9-0
.228
-0.9
93-1
.402
-1.5
05-1
.651
200
5.53
82.
858
1.19
2-0
.184
-1.0
04-1
.417
-1.8
20-1
.972
-2.1
22-2
.218
200
8.40
85.
235
3.03
91.
248
0.09
9-0
.543
-1.0
85-1
.330
-1.5
31-1
.658
500
6.09
33.
222
1.16
9-0
.209
-1.0
67-1
.483
-1.7
94-1
.989
-2.1
17-2
.193
500
8.92
85.
545
2.95
81.
121
-0.0
73-0
.712
-1.1
51-1
.415
-1.5
78-1
.672
200
100
6.31
13.
936
2.50
51.
789
1.06
80.
846
0.26
3-0
.033
-0.0
78-0
.206
200
100
8.96
66.
036
4.03
02.
820
1.75
51.
287
0.55
00.
151
0.03
7-0
.136
200
6.56
94.
157
2.76
01.
563
0.85
60.
535
0.18
70.
036
-0.0
95-0
.195
200
9.24
86.
224
4.19
52.
534
1.47
40.
914
0.42
40.
182
0.00
0-0
.131
500
7.48
14.
754
2.85
91.
586
0.81
20.
442
0.15
9-0
.012
-0.1
24-0
.195
500
10.1
376.
805
4.29
52.
513
1.37
20.
766
0.35
00.
107
-0.0
43-0
.135
500
100
6.84
84.
402
2.90
82.
167
1.39
71.
164
0.56
30.
229
0.16
7-0
.001
500
100
9.48
56.
474
4.39
23.
140
2.01
91.
528
0.77
50.
337
0.20
70.
000
200
7.22
54.
701
3.20
61.
933
1.17
00.
818
0.45
30.
274
0.12
60.
000
200
9.88
56.
757
4.61
72.
867
1.74
11.
140
0.62
50.
355
0.15
60.
000
500
8.05
15.
221
3.23
51.
890
1.07
90.
681
0.38
10.
204
0.08
4-0
.001
500
10.6
857.
261
4.65
42.
787
1.59
60.
953
0.51
50.
262
0.10
30.
000
TN
RM
SE
TN
RM
SE
100
100
5.71
13.
248
1.86
81.
395
1.56
91.
708
2.19
22.
470
2.51
32.
614
100
100
8.41
75.
394
3.26
51.
932
1.14
51.
011
1.38
71.
721
1.82
41.
967
200
6.07
13.
497
1.95
91.
290
1.53
41.
825
2.17
42.
317
2.46
42.
558
200
8.78
55.
643
3.44
11.
759
1.01
01.
066
1.43
11.
640
1.82
81.
950
500
6.60
23.
810
1.94
41.
247
1.52
61.
840
2.12
52.
313
2.44
12.
517
500
9.28
55.
935
3.37
61.
651
0.95
61.
127
1.46
31.
702
1.86
11.
955
200
100
6.63
84.
256
2.77
11.
960
1.19
50.
911
0.36
30.
211
0.21
40.
286
200
100
9.21
76.
285
4.24
42.
967
1.86
31.
342
0.60
50.
231
0.15
50.
197
200
6.83
14.
394
2.92
31.
685
0.94
10.
585
0.26
40.
173
0.19
80.
266
200
9.44
96.
415
4.34
12.
647
1.55
10.
954
0.45
90.
223
0.12
70.
186
500
7.82
25.
079
3.13
11.
793
0.95
30.
521
0.22
90.
139
0.19
30.
252
500
10.3
897.
060
4.52
22.
695
1.49
90.
836
0.39
00.
150
0.11
50.
179
500
100
7.10
04.
652
3.11
52.
299
1.48
91.
206
0.59
80.
260
0.18
50.
007
500
100
9.67
56.
665
4.56
03.
259
2.10
71.
574
0.81
20.
366
0.22
40.
006
200
7.41
94.
878
3.33
12.
026
1.22
70.
844
0.46
70.
281
0.13
20.
001
200
10.0
326.
900
4.73
02.
956
1.80
01.
171
0.64
20.
363
0.16
10.
001
500
8.26
75.
427
3.40
52.
015
1.16
00.
721
0.40
00.
211
0.08
60.
002
500
10.8
457.
423
4.79
82.
902
1.67
50.
996
0.53
70.
271
0.10
70.
002
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
mod
el,
(29),
are
gen
erate
das:
ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi
=0,
for
i=
1,2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
an
d[N
αβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3,
wh
ereαβ
rela
tes
toα
un
der
(11)
andf jt∼IIDN
(0,1
).N
on
-Gau
ssia
ner
rors
are
gen
erate
das:
uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,ν it∼IIDN
(0,1
)
andχ
2 v,t
isa
chi-
squ
ared
ran
dom
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).
v ij∼
IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,2
,µv
=0.8
7,µv2
=0.7
1,
µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
5
Tab
leS
2a:
Bia
san
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
cp
anel
dat
am
od
elC
ross
corr
elati
on
sare
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
nan
dn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
10G
auss
ian
erro
rsG
auss
ian
erro
rsT
NB
ias
Bia
s10
010
03.
117
1.45
10.
736
0.46
60.
220
0.29
4-0
.075
-0.1
79-0
.049
-0.0
0210
010
05.
785
3.47
02.
112
1.35
60.
779
0.62
70.
124
-0.0
76-0
.010
-0.0
0120
03.
367
1.75
01.
145
0.49
60.
219
0.17
1-0
.009
0.00
6-0
.013
-0.0
0320
06.
148
3.75
72.
441
1.32
10.
716
0.45
30.
145
0.08
20.
016
-0.0
0150
03.
913
2.21
41.
191
0.61
30.
276
0.20
20.
095
0.03
50.
001
-0.0
0450
06.
774
4.22
12.
470
1.36
50.
698
0.42
50.
208
0.08
80.
021
-0.0
0220
010
03.
205
1.49
50.
774
0.50
50.
227
0.31
2-0
.072
-0.1
75-0
.045
0.00
020
010
05.
846
3.48
62.
139
1.39
80.
788
0.64
40.
123
-0.0
75-0
.006
0.00
020
03.
491
1.85
11.
203
0.52
80.
239
0.18
3-0
.001
0.01
0-0
.010
0.00
020
06.
231
3.84
52.
499
1.35
20.
734
0.46
50.
153
0.08
50.
018
0.00
050
04.
100
2.33
91.
264
0.66
10.
304
0.21
80.
104
0.04
10.
005
0.00
050
06.
942
4.33
82.
539
1.41
60.
728
0.44
30.
217
0.09
30.
024
0.00
050
010
03.
238
1.55
10.
802
0.50
90.
250
0.31
5-0
.068
-0.1
74-0
.044
0.00
050
010
05.
863
3.54
62.
162
1.39
10.
806
0.64
40.
125
-0.0
75-0
.006
0.00
020
03.
566
1.89
31.
238
0.54
80.
247
0.18
90.
003
0.01
2-0
.009
0.00
020
06.
299
3.88
42.
533
1.36
80.
741
0.47
10.
157
0.08
60.
018
0.00
050
04.
208
2.40
61.
307
0.68
10.
315
0.22
50.
108
0.04
30.
006
0.00
050
07.
028
4.40
62.
587
1.43
90.
742
0.45
00.
222
0.09
50.
024
0.00
0T
NR
MSE
RM
SE
100
100
3.16
91.
525
0.81
20.
527
0.28
70.
322
0.12
10.
190
0.06
10.
008
100
100
5.82
23.
515
2.16
01.
398
0.82
30.
657
0.19
00.
125
0.05
60.
004
200
3.38
61.
771
1.16
30.
518
0.24
30.
187
0.05
10.
035
0.02
50.
008
200
6.16
03.
772
2.45
61.
338
0.73
30.
466
0.16
40.
097
0.03
60.
004
500
3.91
82.
220
1.19
60.
619
0.28
10.
205
0.09
80.
039
0.01
20.
011
500
6.77
84.
224
2.47
31.
369
0.70
30.
429
0.21
10.
091
0.02
60.
006
200
100
3.25
81.
564
0.84
80.
568
0.29
30.
340
0.12
10.
187
0.05
80.
000
200
100
5.88
23.
530
2.18
51.
442
0.83
10.
675
0.19
10.
124
0.05
60.
000
200
3.51
01.
875
1.22
10.
550
0.26
20.
198
0.05
30.
036
0.02
10.
000
200
6.24
43.
861
2.51
41.
368
0.75
10.
478
0.17
30.
100
0.03
50.
000
500
4.10
52.
344
1.26
90.
665
0.30
80.
222
0.10
70.
044
0.01
00.
000
500
6.94
54.
341
2.54
21.
420
0.73
20.
446
0.22
00.
096
0.02
70.
000
500
100
3.29
11.
622
0.87
60.
575
0.31
40.
344
0.12
00.
186
0.05
90.
000
500
100
5.90
23.
594
2.21
01.
438
0.85
10.
675
0.19
30.
125
0.05
90.
000
200
3.58
51.
916
1.25
50.
570
0.27
00.
206
0.05
50.
039
0.02
20.
000
200
6.31
23.
899
2.54
71.
385
0.75
70.
486
0.17
70.
103
0.03
60.
000
500
4.21
22.
411
1.31
20.
685
0.32
00.
228
0.11
20.
046
0.01
10.
000
500
7.03
14.
410
2.59
11.
443
0.74
60.
454
0.22
60.
098
0.02
80.
000
Non
-Gau
ssia
ner
rors
Non
-Gau
ssia
ner
rors
TN
Bia
sT
NB
ias
100
100
8.56
85.
689
3.72
42.
532
1.51
21.
081
0.40
20.
062
0.03
6-0
.010
100
100
12.0
708.
641
5.99
34.
153
2.60
81.
779
0.83
10.
291
0.12
7-0
.006
200
9.53
06.
389
4.33
32.
622
1.56
90.
929
0.40
80.
203
0.05
4-0
.012
200
12.9
989.
290
6.52
44.
199
2.62
11.
572
0.77
70.
390
0.12
7-0
.007
500
10.7
247.
428
4.64
92.
841
1.60
70.
920
0.45
90.
198
0.05
0-0
.014
500
14.0
8410
.250
6.83
64.
378
2.59
51.
492
0.77
00.
348
0.10
7-0
.008
200
100
9.13
06.
188
4.16
82.
735
1.68
01.
223
0.47
00.
098
0.06
10.
000
200
100
12.5
839.
125
6.43
74.
381
2.80
11.
935
0.90
90.
331
0.15
20.
000
200
10.1
207.
023
4.75
92.
884
1.73
11.
024
0.50
50.
252
0.07
70.
000
200
13.5
329.
905
6.96
84.
496
2.80
71.
679
0.89
00.
444
0.15
00.
000
500
11.4
397.
948
5.22
03.
202
1.82
91.
017
0.52
80.
239
0.07
70.
000
500
14.7
0210
.742
7.41
34.
766
2.84
81.
606
0.84
70.
390
0.13
20.
000
500
100
9.60
36.
544
4.37
92.
909
1.80
91.
277
0.50
00.
132
0.07
10.
000
500
100
13.0
549.
478
6.67
64.
564
2.95
92.
007
0.94
80.
375
0.16
60.
000
200
10.7
397.
425
5.05
33.
082
1.83
31.
144
0.52
30.
269
0.08
60.
000
200
14.1
0810
.304
7.27
94.
715
2.93
41.
827
0.91
40.
468
0.16
10.
000
500
12.2
068.
470
5.52
93.
365
1.94
71.
106
0.57
00.
255
0.08
50.
000
500
15.4
0411
.243
7.73
44.
956
2.99
31.
719
0.90
20.
412
0.14
20.
000
TN
RM
SE
RM
SE
100
100
8.90
56.
084
4.05
82.
823
1.74
01.
224
0.59
50.
266
0.12
30.
029
100
100
12.3
058.
914
6.23
84.
387
2.79
91.
917
0.99
10.
436
0.19
70.
018
200
9.88
96.
726
4.64
72.
937
1.87
51.
066
0.53
70.
273
0.09
80.
027
200
13.2
409.
528
6.76
44.
449
2.86
71.
701
0.89
10.
457
0.16
60.
018
500
11.0
677.
813
4.95
53.
152
1.84
91.
089
0.54
80.
268
0.08
20.
029
500
14.3
1310
.519
7.06
64.
624
2.80
51.
649
0.86
00.
414
0.13
40.
018
200
100
9.39
06.
467
4.46
62.
949
1.87
11.
381
0.61
40.
257
0.12
90.
001
200
100
12.7
609.
319
6.65
84.
552
2.96
12.
076
1.03
30.
447
0.21
00.
001
200
10.3
617.
303
5.01
73.
088
1.93
01.
132
0.64
40.
313
0.11
10.
001
200
13.6
9410
.100
7.16
24.
658
2.97
01.
778
1.01
00.
503
0.18
30.
000
500
11.6
628.
179
5.47
43.
460
2.02
51.
121
0.61
00.
309
0.10
30.
001
500
14.8
5210
.903
7.60
24.
968
3.01
51.
705
0.92
60.
454
0.15
80.
000
500
100
9.77
86.
756
4.54
73.
059
1.93
71.
396
0.60
00.
246
0.12
70.
000
500
100
13.1
749.
623
6.80
04.
685
3.06
82.
114
1.03
50.
458
0.21
50.
000
200
10.9
317.
603
5.21
23.
213
1.94
11.
245
0.59
80.
313
0.11
30.
000
200
14.2
3510
.427
7.39
94.
820
3.02
61.
917
0.98
00.
510
0.18
80.
000
500
12.3
788.
643
5.70
63.
498
2.07
01.
180
0.62
50.
286
0.11
40.
000
500
15.5
1911
.362
7.86
25.
062
3.09
61.
787
0.95
50.
443
0.16
90.
000
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.
Gau
ssia
ner
rors
are
gen
erat
edasuit∼IIDN
(0,1
)in
(31).
Non
-Gau
ssia
ner
rors
are
gen
erate
dasuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,
2,...,N
,ν it∼IIDN
(0,1
)an
d
χ2 v,t
isa
chi-
squ
ared
ran
dom
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elat
ion
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
alva
lue
use
din
the
mu
ltip
lete
stin
gpro
ced
ure
show
nin
(23).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
6
Tab
leS
2b
:B
ias
and
RM
SE
(×100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
om
ast
atic
pan
eld
ata
mod
elC
ross
corr
elati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
nan
dn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
10G
auss
ian
erro
rsG
auss
ian
erro
rsT
NB
ias
TN
Bia
s10
010
03.
000
1.15
40.
241
-0.1
09-0
.553
-0.5
91-1
.078
-1.3
09-1
.300
-1.3
8910
010
05.
848
3.44
71.
981
1.21
20.
465
0.21
1-0
.407
-0.7
35-0
.792
-0.9
3120
03.
189
1.40
80.
603
-0.2
00-0
.634
-0.7
71-1
.028
-1.1
09-1
.218
-1.3
0520
06.
118
3.68
92.
255
1.02
00.
278
-0.0
67-0
.450
-0.6
15-0
.773
-0.8
9050
03.
662
1.73
50.
529
-0.2
12-0
.655
-0.8
22-0
.996
-1.1
17-1
.215
-1.2
7950
06.
695
4.02
02.
156
0.92
10.
159
-0.2
06-0
.493
-0.6
77-0
.808
-0.8
9020
010
03.
693
2.10
21.
360
1.18
20.
768
0.76
90.
297
0.04
90.
041
-0.0
7920
010
06.
255
4.03
22.
697
2.05
61.
346
1.13
20.
530
0.19
10.
120
-0.0
4620
03.
826
2.30
51.
683
0.97
30.
589
0.47
10.
231
0.13
10.
020
-0.0
7520
06.
517
4.25
02.
952
1.80
11.
105
0.78
10.
420
0.24
10.
083
-0.0
4450
04.
262
2.57
11.
549
0.88
90.
506
0.36
10.
194
0.08
0-0
.009
-0.0
7350
07.
065
4.53
92.
811
1.65
40.
951
0.61
20.
337
0.16
30.
040
-0.0
4450
010
03.
829
2.24
11.
509
1.33
50.
914
0.90
80.
432
0.17
20.
150
0.00
050
010
06.
320
4.10
52.
786
2.15
21.
435
1.21
60.
615
0.26
90.
188
0.00
020
03.
954
2.44
01.
816
1.10
00.
711
0.58
70.
347
0.23
30.
116
0.00
020
06.
589
4.33
03.
033
1.88
01.
183
0.85
60.
495
0.30
60.
143
0.00
050
04.
410
2.70
71.
679
1.00
20.
614
0.46
10.
287
0.17
00.
076
0.00
050
07.
169
4.63
42.
902
1.73
11.
025
0.67
90.
398
0.22
10.
094
0.00
0T
NR
MSE
TN
RM
SE
100
100
3.08
01.
365
0.83
50.
847
1.03
31.
059
1.41
41.
596
1.59
01.
672
100
100
5.88
63.
501
2.07
61.
363
0.78
90.
675
0.78
60.
996
1.04
01.
158
200
3.22
71.
519
0.89
90.
752
0.99
71.
100
1.29
41.
370
1.46
31.
548
200
6.13
33.
716
2.30
51.
144
0.61
90.
572
0.73
00.
855
0.97
81.
086
500
3.67
51.
789
0.77
10.
676
0.94
81.
092
1.24
01.
348
1.43
91.
499
500
6.70
04.
031
2.18
81.
022
0.51
40.
564
0.73
60.
881
0.99
31.
068
200
100
3.73
22.
143
1.39
31.
205
0.79
00.
785
0.32
90.
140
0.12
00.
120
200
100
6.28
54.
061
2.72
12.
075
1.36
21.
144
0.54
80.
226
0.15
30.
074
200
3.84
12.
319
1.69
30.
983
0.60
10.
481
0.25
10.
157
0.08
60.
110
200
6.52
84.
260
2.96
11.
808
1.11
20.
787
0.42
90.
251
0.10
30.
069
500
4.26
62.
575
1.55
20.
893
0.51
10.
368
0.20
50.
105
0.07
20.
102
500
7.06
84.
541
2.81
31.
656
0.95
40.
615
0.34
00.
170
0.06
30.
065
500
100
3.86
72.
277
1.53
51.
349
0.92
50.
916
0.44
40.
194
0.16
80.
000
500
100
6.35
04.
134
2.80
92.
168
1.44
81.
225
0.62
70.
286
0.20
40.
000
200
3.96
82.
452
1.82
41.
106
0.71
60.
590
0.35
10.
237
0.12
10.
000
200
6.59
94.
339
3.04
11.
886
1.18
80.
860
0.50
00.
310
0.14
80.
000
500
4.41
42.
710
1.68
11.
003
0.61
50.
462
0.28
70.
171
0.07
70.
000
500
7.17
24.
636
2.90
41.
732
1.02
60.
680
0.39
90.
222
0.09
50.
000
Non
-Gau
ssia
ner
rors
Non
-Gau
ssia
ner
rors
TN
Bia
sT
NB
ias
100
100
7.97
34.
917
2.78
51.
443
0.32
6-0
.192
-0.9
39-1
.352
-1.4
54-1
.599
100
100
11.5
657.
997
5.25
93.
336
1.75
60.
878
-0.1
01-0
.684
-0.9
01-1
.123
200
9.01
55.
756
3.46
81.
595
0.37
6-0
.308
-0.8
89-1
.162
-1.3
66-1
.504
200
12.5
738.
804
5.87
93.
441
1.73
70.
678
-0.1
43-0
.580
-0.8
80-1
.073
500
10.4
406.
818
3.96
21.
870
0.47
1-0
.313
-0.8
37-1
.146
-1.3
38-1
.448
500
13.8
669.
790
6.35
73.
666
1.75
20.
582
-0.1
86-0
.638
-0.9
07-1
.056
200
100
8.97
96.
043
4.03
12.
821
1.76
51.
298
0.56
50.
161
0.05
2-0
.123
200
100
12.3
548.
843
6.16
54.
314
2.78
51.
939
0.97
00.
395
0.17
2-0
.077
200
10.0
366.
865
4.66
02.
858
1.68
31.
041
0.50
10.
224
0.02
3-0
.121
200
13.3
959.
656
6.76
14.
363
2.67
61.
648
0.86
00.
421
0.12
4-0
.077
500
11.3
527.
797
5.02
23.
004
1.67
60.
941
0.44
70.
162
-0.0
11-0
.113
500
14.5
9610
.544
7.14
64.
495
2.62
51.
491
0.74
90.
320
0.06
9-0
.074
500
100
9.48
36.
468
4.38
63.
128
2.01
91.
528
0.77
10.
335
0.20
50.
000
500
100
12.8
259.
246
6.49
44.
590
3.00
12.
122
1.12
40.
517
0.27
30.
000
200
10.4
967.
246
4.97
03.
107
1.89
41.
226
0.67
20.
376
0.16
20.
000
200
13.8
1010
.006
7.04
54.
582
2.85
51.
790
0.98
30.
525
0.21
60.
000
500
11.9
338.
278
5.40
03.
284
1.89
31.
115
0.59
70.
299
0.11
50.
000
500
15.1
1410
.992
7.50
94.
766
2.82
61.
640
0.86
70.
420
0.15
60.
000
TN
RM
SE
TN
RM
SE
100
100
8.29
25.
272
3.16
21.
856
1.03
90.
938
1.31
81.
652
1.76
11.
901
100
100
11.7
798.
225
5.48
13.
535
1.97
01.
143
0.69
90.
982
1.17
31.
375
200
9.34
76.
109
3.80
01.
975
0.99
50.
889
1.22
81.
463
1.65
51.
789
200
12.7
949.
037
6.09
73.
649
1.94
90.
962
0.65
30.
882
1.13
21.
310
500
10.7
867.
197
4.36
12.
315
1.12
80.
915
1.18
41.
437
1.61
51.
723
500
14.0
9510
.043
6.62
03.
929
2.02
30.
944
0.67
60.
918
1.14
61.
288
200
100
9.20
86.
276
4.23
22.
961
1.87
01.
356
0.62
10.
232
0.14
60.
181
200
100
12.5
159.
010
6.31
74.
430
2.87
51.
993
1.01
40.
430
0.20
50.
121
200
10.2
897.
114
4.86
53.
027
1.80
01.
105
0.54
10.
255
0.11
10.
168
200
13.5
699.
836
6.92
34.
504
2.78
11.
712
0.89
60.
439
0.14
70.
112
500
11.5
958.
048
5.25
53.
196
1.81
11.
014
0.48
50.
193
0.10
30.
161
500
14.7
6010
.721
7.32
34.
653
2.74
61.
563
0.78
70.
338
0.09
90.
111
500
100
9.62
16.
606
4.50
33.
210
2.08
01.
561
0.79
90.
360
0.22
10.
007
500
100
12.9
229.
346
6.58
64.
660
3.05
62.
155
1.15
10.
539
0.28
70.
006
200
10.6
427.
390
5.08
83.
202
1.96
11.
262
0.69
20.
385
0.16
80.
001
200
13.9
1010
.110
7.13
94.
664
2.91
61.
828
1.00
50.
536
0.22
20.
001
500
12.0
828.
434
5.54
83.
410
1.98
61.
169
0.62
50.
311
0.11
90.
001
500
15.2
1311
.101
7.61
94.
867
2.90
61.
691
0.89
60.
434
0.16
10.
001
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
and
[Nαβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
andf jt∼IIDN
(0,1
).G
auss
ian
erro
rsare
giv
enby:uit∼IIDN
(0,1
)an
dn
on
Gau
ssia
ner
rors
are
gen
erate
das:uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,
2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).v ij∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,
µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
7
Tab
leS
2c:
Bia
san
dR
MS
E(×
100
)fo
rth
eα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n1
wit
hG
au
ssia
nan
dn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
10G
auss
ian
erro
rsG
auss
ian
erro
rsT
NB
ias
TN
Bia
s10
010
03.
118
1.43
10.
731
0.46
90.
217
0.29
5-0
.074
-0.1
79-0
.048
-0.0
0310
010
05.
795
3.45
12.
105
1.35
70.
780
0.63
10.
123
-0.0
75-0
.008
-0.0
0120
03.
378
1.75
91.
154
0.49
40.
218
0.17
0-0
.010
0.00
5-0
.015
-0.0
0420
06.
150
3.77
02.
457
1.32
10.
717
0.45
30.
146
0.08
20.
014
-0.0
0250
03.
916
2.21
21.
190
0.61
10.
275
0.20
10.
093
0.03
4-0
.001
-0.0
0550
06.
783
4.21
82.
466
1.36
30.
697
0.42
40.
208
0.08
70.
020
-0.0
0220
010
03.
169
1.50
60.
773
0.50
20.
238
0.31
2-0
.069
-0.1
76-0
.046
0.00
020
010
05.
829
3.49
42.
147
1.38
80.
795
0.64
00.
125
-0.0
75-0
.008
0.00
020
03.
510
1.84
71.
206
0.53
30.
240
0.18
60.
002
0.01
2-0
.010
0.00
020
06.
257
3.84
52.
503
1.36
40.
736
0.46
80.
157
0.08
70.
018
0.00
050
04.
097
2.34
21.
265
0.65
80.
303
0.21
90.
104
0.04
20.
005
0.00
050
06.
937
4.34
22.
540
1.41
30.
727
0.44
40.
217
0.09
40.
024
0.00
050
010
03.
242
1.54
60.
801
0.51
60.
250
0.31
5-0
.069
-0.1
72-0
.045
0.00
050
010
05.
851
3.52
82.
170
1.39
80.
806
0.64
20.
125
-0.0
72-0
.007
0.00
020
03.
568
1.88
91.
240
0.54
90.
252
0.18
70.
005
0.01
2-0
.010
0.00
020
06.
308
3.88
52.
529
1.37
50.
747
0.46
80.
159
0.08
60.
017
0.00
050
04.
201
2.40
11.
309
0.68
10.
317
0.22
40.
107
0.04
30.
006
0.00
050
07.
024
4.39
92.
589
1.44
00.
745
0.44
90.
221
0.09
60.
025
0.00
0T
NR
MSE
TN
RM
SE
100
100
3.17
11.
500
0.80
90.
528
0.28
20.
321
0.12
10.
191
0.06
20.
008
100
100
5.83
33.
494
2.15
21.
399
0.82
50.
659
0.18
80.
124
0.05
80.
004
200
3.39
91.
781
1.17
10.
515
0.24
30.
186
0.05
20.
036
0.02
70.
011
200
6.16
33.
785
2.47
11.
337
0.73
30.
467
0.16
50.
098
0.03
30.
005
500
3.92
12.
217
1.19
50.
616
0.28
00.
205
0.09
80.
040
0.01
50.
012
500
6.78
74.
221
2.47
01.
367
0.70
10.
428
0.21
20.
091
0.02
50.
006
200
100
3.22
01.
574
0.84
50.
562
0.30
20.
340
0.11
90.
187
0.05
90.
000
200
100
5.86
53.
538
2.19
21.
429
0.83
90.
670
0.19
00.
122
0.05
60.
000
200
3.52
91.
869
1.22
40.
555
0.26
40.
201
0.05
50.
038
0.02
20.
000
200
6.27
03.
860
2.51
71.
380
0.75
30.
482
0.17
80.
103
0.03
50.
000
500
4.10
22.
347
1.27
00.
663
0.30
80.
222
0.10
70.
045
0.01
00.
000
500
6.94
04.
345
2.54
51.
417
0.73
10.
447
0.22
10.
097
0.02
70.
000
500
100
3.29
11.
615
0.87
70.
578
0.31
30.
343
0.12
00.
186
0.05
80.
000
500
100
5.88
43.
573
2.21
91.
444
0.85
00.
671
0.19
40.
126
0.05
70.
000
200
3.58
81.
911
1.25
70.
570
0.27
60.
202
0.05
60.
039
0.02
30.
000
200
6.32
13.
901
2.54
31.
391
0.76
40.
481
0.18
00.
103
0.03
50.
000
500
4.20
72.
406
1.31
30.
686
0.32
20.
228
0.11
10.
046
0.01
10.
000
500
7.02
74.
403
2.59
31.
444
0.74
90.
452
0.22
50.
099
0.02
80.
000
Non
-Gau
ssia
ner
rors
Non
-Gau
ssia
ner
rors
TN
Bia
sT
NB
ias
100
100
8.44
75.
562
3.67
72.
417
1.49
11.
079
0.38
10.
057
0.03
2-0
.012
100
100
11.9
738.
498
5.93
24.
013
2.58
51.
768
0.80
50.
286
0.12
4-0
.007
200
9.34
06.
220
4.23
12.
548
1.46
00.
907
0.40
50.
192
0.05
1-0
.015
200
12.8
099.
116
6.41
74.
116
2.49
61.
543
0.77
80.
375
0.12
6-0
.009
500
10.5
887.
135
4.65
72.
749
1.58
70.
886
0.43
30.
191
0.04
7-0
.017
500
13.9
509.
966
6.83
64.
270
2.56
91.
448
0.73
60.
340
0.10
5-0
.010
200
100
9.17
26.
148
4.07
62.
673
1.67
41.
191
0.46
80.
110
0.05
80.
000
200
100
12.6
369.
066
6.34
94.
291
2.79
51.
901
0.90
00.
343
0.14
60.
000
200
10.1
466.
941
4.76
12.
886
1.69
21.
040
0.49
30.
245
0.07
80.
000
200
13.5
549.
818
6.96
54.
486
2.76
11.
695
0.87
40.
435
0.15
10.
000
500
11.4
297.
889
5.19
23.
109
1.78
41.
002
0.52
60.
238
0.07
80.
000
500
14.6
9810
.680
7.38
14.
666
2.79
21.
586
0.84
30.
389
0.13
30.
000
500
100
9.55
16.
522
4.37
52.
858
1.79
71.
278
0.50
50.
124
0.06
70.
000
500
100
13.0
039.
468
6.68
14.
519
2.93
92.
009
0.95
00.
362
0.16
00.
000
200
10.6
267.
339
5.01
63.
082
1.84
91.
115
0.52
90.
261
0.08
60.
000
200
14.0
0210
.220
7.24
24.
716
2.95
21.
790
0.92
10.
457
0.16
10.
000
500
12.1
608.
434
5.54
93.
413
1.92
61.
135
0.56
40.
255
0.08
00.
000
500
15.3
5511
.209
7.75
75.
006
2.97
01.
751
0.89
40.
413
0.13
70.
000
TN
RM
SE
TN
RM
SE
100
100
8.79
35.
928
4.00
92.
737
1.72
31.
271
0.53
10.
263
0.12
60.
031
100
100
12.2
138.
756
6.18
24.
264
2.78
01.
936
0.93
20.
431
0.19
80.
019
200
9.66
16.
547
4.53
02.
851
1.68
01.
063
0.53
20.
277
0.10
70.
037
200
13.0
299.
348
6.64
64.
355
2.68
01.
687
0.89
10.
453
0.17
10.
024
500
10.9
257.
431
4.99
03.
023
1.88
11.
068
0.51
60.
255
0.08
30.
039
500
14.1
7610
.176
7.08
74.
492
2.81
11.
609
0.81
80.
402
0.13
40.
026
200
100
9.44
26.
443
4.33
02.
897
1.84
31.
328
0.67
60.
304
0.12
50.
001
200
100
12.8
229.
271
6.53
74.
469
2.94
12.
028
1.06
30.
486
0.20
60.
001
200
10.4
167.
244
5.04
73.
129
1.87
01.
177
0.62
90.
316
0.12
80.
003
200
13.7
3510
.029
7.17
34.
675
2.91
01.
817
0.99
10.
499
0.19
60.
003
500
11.6
438.
128
5.46
73.
298
1.97
01.
104
0.61
50.
292
0.11
40.
001
500
14.8
4210
.846
7.58
24.
820
2.95
21.
681
0.92
80.
442
0.16
70.
000
500
100
9.73
36.
723
4.55
92.
989
1.92
91.
377
0.61
20.
253
0.12
20.
000
500
100
13.1
259.
607
6.81
54.
624
3.04
92.
099
1.04
20.
455
0.20
90.
000
200
10.7
737.
524
5.19
13.
233
1.96
51.
212
0.65
00.
301
0.11
60.
000
200
14.1
0110
.346
7.37
14.
834
3.05
11.
875
1.01
80.
495
0.18
90.
000
500
12.3
288.
570
5.72
63.
576
2.04
41.
246
0.62
40.
284
0.09
30.
000
500
15.4
6811
.304
7.88
55.
133
3.07
01.
847
0.95
10.
442
0.15
00.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
model
,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),
fori
=1,2,...,N
.G
auss
ian
erro
rsar
ege
ner
ated
asuit∼IIDN
(0,1
)in
(31).
Non-G
au
ssia
ner
rors
are
gen
erate
dasuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,
ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
ared
ran
dom
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).
Des
ign
1ass
um
esb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
orbN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elati
on
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
giv
enby
(30),
wh
ere
BN
=D
iag
(bN
).c p
(n,δ
)co
rres
pon
ds
toth
ecr
itic
alva
lue
use
din
the
mu
ltip
lete
stin
gp
roce
du
resh
own
in(2
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
8
Tab
leS
2d
:B
ias
and
RM
SE
(×100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
dyn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
atio
ns
are
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
nan
dn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
c p(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
,δ
=1/
4an
dp
=0.
10G
auss
ian
erro
rsG
auss
ian
erro
rsT
NB
ias
TN
Bia
s10
010
02.
937
1.07
20.
123
-0.2
44-0
.710
-0.7
55-1
.231
-1.4
71-1
.460
-1.5
5510
010
05.
779
3.38
21.
900
1.11
30.
353
0.09
5-0
.515
-0.8
50-0
.907
-1.0
5020
03.
160
1.34
70.
512
-0.3
15-0
.740
-0.9
08-1
.162
-1.2
50-1
.366
-1.4
5220
06.
113
3.65
52.
200
0.94
70.
210
-0.1
63-0
.545
-0.7
16-0
.880
-0.9
9750
03.
620
1.66
20.
436
-0.3
31-0
.784
-0.9
57-1
.142
-1.2
68-1
.368
-1.4
3550
06.
666
3.97
22.
092
0.83
70.
066
-0.3
06-0
.603
-0.7
90-0
.922
-1.0
0820
010
03.
709
2.10
31.
368
1.18
50.
759
0.75
50.
280
0.04
20.
032
-0.0
9020
010
06.
276
4.04
32.
708
2.06
31.
341
1.12
40.
518
0.18
80.
115
-0.0
5220
03.
807
2.29
01.
669
0.95
60.
577
0.45
90.
217
0.12
00.
011
-0.0
8420
06.
494
4.22
92.
939
1.78
31.
094
0.77
10.
409
0.23
20.
077
-0.0
5050
04.
264
2.57
21.
549
0.88
60.
504
0.35
80.
190
0.07
7-0
.010
-0.0
7550
07.
057
4.53
32.
808
1.64
90.
948
0.60
90.
334
0.16
10.
039
-0.0
4650
010
03.
840
2.25
01.
510
1.34
00.
921
0.91
40.
431
0.17
20.
152
0.00
050
010
06.
358
4.13
52.
802
2.16
31.
447
1.22
60.
615
0.27
00.
191
0.00
020
03.
955
2.44
01.
818
1.09
90.
709
0.58
70.
345
0.23
30.
114
0.00
020
06.
603
4.33
33.
042
1.88
11.
183
0.85
60.
494
0.30
60.
142
0.00
050
04.
407
2.70
41.
675
1.00
10.
614
0.46
10.
287
0.17
00.
075
0.00
050
07.
166
4.63
02.
898
1.72
91.
024
0.67
90.
398
0.22
20.
094
0.00
0T
NR
MSE
TN
RM
SE
100
100
3.02
31.
327
0.86
30.
947
1.18
71.
223
1.56
71.
769
1.75
81.
847
100
100
5.82
03.
447
2.01
11.
305
0.77
70.
712
0.88
01.
119
1.16
21.
287
200
3.20
01.
471
0.87
40.
844
1.11
01.
247
1.45
01.
523
1.63
21.
711
200
6.12
93.
683
2.25
91.
100
0.62
90.
644
0.84
10.
964
1.10
31.
205
500
3.63
51.
730
0.75
70.
787
1.09
41.
248
1.41
21.
524
1.61
61.
681
500
6.67
13.
984
2.13
30.
974
0.55
20.
663
0.86
41.
015
1.13
01.
209
200
100
3.74
72.
142
1.40
31.
210
0.78
30.
774
0.32
00.
141
0.12
60.
138
200
100
6.30
54.
072
2.73
32.
082
1.35
81.
137
0.53
90.
222
0.15
20.
085
200
3.82
32.
305
1.67
90.
967
0.58
90.
471
0.24
00.
150
0.08
80.
120
200
6.50
54.
239
2.94
61.
791
1.10
10.
778
0.41
80.
243
0.10
00.
074
500
4.26
82.
575
1.55
20.
890
0.51
00.
364
0.20
20.
103
0.06
80.
103
500
7.05
94.
536
2.81
01.
651
0.95
10.
612
0.33
80.
168
0.06
00.
065
500
100
3.87
72.
283
1.53
51.
354
0.93
20.
921
0.44
30.
196
0.17
00.
000
500
100
6.38
64.
162
2.82
52.
178
1.46
01.
235
0.62
70.
289
0.20
60.
000
200
3.96
92.
453
1.82
61.
105
0.71
40.
590
0.34
90.
237
0.12
00.
000
200
6.61
34.
343
3.04
91.
887
1.18
80.
860
0.49
80.
310
0.14
70.
000
500
4.41
12.
707
1.67
71.
002
0.61
50.
462
0.28
70.
171
0.07
70.
000
500
7.16
84.
632
2.90
01.
731
1.02
50.
680
0.39
90.
223
0.09
50.
000
Non
-Gau
ssia
ner
rors
Non
-Gau
ssia
ner
rors
TN
Bia
sT
NB
ias
100
100
7.94
34.
850
2.70
71.
341
0.21
7-0
.321
-1.0
92-1
.499
-1.6
03-1
.738
100
100
11.5
427.
952
5.21
23.
270
1.68
50.
788
-0.2
11-0
.790
-1.0
15-1
.228
200
8.88
45.
619
3.33
01.
443
0.21
7-0
.469
-1.0
46-1
.320
-1.5
24-1
.660
200
12.4
498.
683
5.76
13.
318
1.60
90.
552
-0.2
65-0
.700
-0.9
98-1
.191
500
10.2
096.
610
3.77
71.
710
0.32
9-0
.446
-0.9
67-1
.272
-1.4
59-1
.571
500
13.6
519.
595
6.18
43.
520
1.62
80.
472
-0.2
88-0
.733
-0.9
95-1
.147
200
100
8.94
66.
018
4.02
02.
811
1.75
01.
289
0.55
30.
147
0.03
3-0
.135
200
100
12.3
288.
823
6.15
44.
304
2.76
91.
932
0.96
10.
386
0.15
7-0
.085
200
9.92
36.
769
4.59
22.
799
1.64
61.
015
0.47
90.
211
0.01
0-0
.131
200
13.2
829.
557
6.68
54.
302
2.63
51.
622
0.84
00.
410
0.11
4-0
.084
500
11.2
797.
734
4.97
52.
969
1.65
20.
926
0.43
70.
154
-0.0
18-0
.120
500
14.5
2810
.481
7.09
74.
456
2.59
91.
475
0.74
00.
314
0.06
5-0
.079
500
100
9.60
16.
570
4.45
73.
173
2.04
71.
540
0.77
80.
343
0.20
60.
000
500
100
12.9
469.
356
6.57
94.
646
3.03
42.
138
1.13
30.
527
0.27
50.
000
200
10.5
117.
264
4.99
13.
122
1.90
71.
233
0.67
60.
381
0.16
20.
000
200
13.8
2010
.018
7.06
04.
596
2.86
41.
799
0.98
60.
531
0.21
70.
000
500
11.9
108.
257
5.38
23.
272
1.88
41.
108
0.59
30.
296
0.11
50.
000
500
15.0
9410
.975
7.49
44.
755
2.81
61.
632
0.86
20.
417
0.15
60.
000
TN
RM
SE
TN
RM
SE
100
100
8.27
35.
217
3.10
41.
788
1.04
41.
032
1.48
21.
822
1.92
82.
059
100
100
11.7
638.
187
5.44
23.
474
1.91
71.
104
0.78
01.
106
1.30
11.
495
200
9.23
05.
995
3.70
11.
910
1.03
71.
023
1.38
71.
625
1.81
21.
946
200
12.6
768.
927
5.99
33.
555
1.86
90.
927
0.74
11.
002
1.24
91.
432
500
10.5
597.
000
4.19
42.
191
1.07
70.
965
1.27
31.
531
1.70
91.
821
500
13.8
819.
852
6.45
33.
794
1.91
70.
878
0.70
60.
981
1.21
21.
358
200
100
9.18
76.
261
4.22
62.
951
1.85
21.
342
0.60
20.
225
0.15
20.
196
200
100
12.4
958.
996
6.31
04.
422
2.85
81.
983
0.99
90.
421
0.19
60.
130
200
10.1
707.
014
4.79
42.
967
1.76
71.
083
0.52
60.
249
0.12
40.
184
200
13.4
509.
732
6.84
24.
440
2.74
11.
686
0.88
00.
432
0.14
60.
124
500
11.5
267.
991
5.21
73.
175
1.80
71.
020
0.49
30.
195
0.10
10.
166
500
14.6
9410
.662
7.27
84.
621
2.73
01.
559
0.78
90.
337
0.09
50.
114
500
100
9.75
36.
720
4.58
33.
254
2.10
41.
569
0.80
10.
364
0.22
10.
002
500
100
13.0
539.
466
6.67
84.
719
3.08
82.
170
1.15
80.
547
0.29
00.
002
200
10.6
897.
444
5.14
33.
251
2.00
01.
287
0.70
60.
395
0.16
90.
003
200
13.9
4010
.145
7.17
74.
700
2.94
61.
851
1.01
80.
546
0.22
40.
002
500
12.0
478.
398
5.51
03.
374
1.95
41.
145
0.61
00.
303
0.11
70.
001
500
15.1
8711
.076
7.59
24.
842
2.88
11.
670
0.88
20.
426
0.15
90.
000
Not
es:
Par
amet
ers
ofth
ed
yn
amic
pan
eld
ata
model
,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
an
d[N
αβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3,
wh
ereαβ
rela
tes
toα
un
der
(11)
andf jt∼
IIDN
(0,1
).G
au
ssia
ner
rors
are
giv
enby:uit
∼IIDN
(0,1
)an
dn
on
Gau
ssia
ner
rors
are
gen
erat
edas
:uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,
2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).v ij
∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
9
Tab
leS
3:B
ias
an
dR
MS
E(×
100)
for
theα
esti
mat
eof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stati
cp
anel
dat
am
od
el
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hG
au
ssia
ner
rors
scale
dbyς
=√ 1/
2α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
05c p
(n,δ
)w
ithn
=N
(N−
1)/2
andp
=0.
10δ
=1/
2δ
=1/
2T
NB
ias
TN
Bia
s10
010
00.
812
0.02
7-0
.037
0.23
40.
107
0.29
5-0
.080
-0.2
44-0
.199
-0.2
7510
010
01.
253
0.35
90.
203
0.42
30.
256
0.42
40.
037
-0.1
37-0
.100
-0.1
8420
00.
446
0.04
60.
268
0.03
9-0
.039
0.03
2-0
.112
-0.1
38-0
.213
-0.2
7920
00.
802
0.29
80.
449
0.17
90.
076
0.13
3-0
.015
-0.0
49-0
.125
-0.1
9350
00.
303
0.11
40.
070
-0.0
03-0
.061
-0.0
23-0
.096
-0.1
64-0
.234
-0.2
8850
00.
575
0.29
70.
205
0.10
10.
029
0.06
0-0
.016
-0.0
84-0
.152
-0.2
0420
010
01.
013
0.27
80.
233
0.55
00.
416
0.61
60.
252
0.07
30.
105
-0.0
0320
010
01.
426
0.55
80.
406
0.66
00.
487
0.65
90.
280
0.09
10.
114
-0.0
0220
00.
649
0.28
90.
530
0.31
10.
241
0.31
90.
192
0.15
30.
077
-0.0
0420
00.
982
0.49
60.
650
0.38
40.
286
0.34
60.
210
0.16
30.
083
-0.0
0250
00.
497
0.33
50.
317
0.23
90.
195
0.23
60.
168
0.10
90.
048
-0.0
0450
00.
748
0.47
40.
393
0.28
10.
219
0.25
10.
177
0.11
50.
052
-0.0
0350
010
01.
046
0.30
10.
254
0.57
40.
444
0.63
60.
275
0.09
20.
120
0.00
050
010
01.
465
0.58
70.
432
0.68
20.
509
0.67
40.
296
0.10
30.
124
0.00
020
00.
688
0.32
10.
558
0.33
50.
262
0.33
80.
213
0.16
80.
092
0.00
020
01.
035
0.53
50.
680
0.40
80.
304
0.36
20.
225
0.17
50.
094
0.00
050
00.
534
0.36
10.
339
0.25
60.
213
0.25
30.
183
0.12
30.
060
0.00
050
00.
801
0.50
80.
418
0.29
90.
235
0.26
50.
189
0.12
60.
061
0.00
0T
NR
MSE
TN
RM
SE
100
100
0.87
50.
315
0.30
20.
388
0.31
90.
417
0.32
50.
390
0.37
20.
428
100
100
1.30
40.
470
0.33
10.
495
0.35
30.
486
0.25
40.
275
0.26
10.
309
200
0.49
60.
235
0.35
80.
238
0.24
30.
242
0.27
00.
292
0.34
30.
405
200
0.83
20.
362
0.48
90.
259
0.20
10.
228
0.19
00.
204
0.24
10.
295
500
0.34
10.
210
0.20
20.
197
0.20
90.
214
0.24
00.
286
0.34
60.
397
500
0.59
10.
329
0.25
10.
179
0.15
20.
171
0.16
70.
196
0.24
70.
293
200
100
1.03
80.
325
0.26
30.
561
0.42
70.
623
0.26
90.
115
0.12
70.
013
200
100
1.46
10.
604
0.43
70.
672
0.49
80.
666
0.29
60.
127
0.13
50.
009
200
0.66
30.
303
0.53
40.
316
0.24
60.
323
0.19
80.
158
0.08
50.
012
200
0.99
90.
510
0.65
50.
390
0.29
10.
350
0.21
50.
168
0.09
00.
008
500
0.50
10.
338
0.31
80.
240
0.19
60.
237
0.16
90.
110
0.05
10.
010
500
0.75
30.
477
0.39
50.
283
0.22
10.
252
0.17
80.
116
0.05
40.
007
500
100
1.07
10.
343
0.27
90.
582
0.45
10.
642
0.28
90.
123
0.14
10.
000
500
100
1.49
80.
630
0.46
00.
693
0.51
80.
680
0.31
00.
133
0.14
50.
000
200
0.70
20.
333
0.56
10.
339
0.26
60.
341
0.21
70.
173
0.09
80.
000
200
1.05
30.
550
0.68
50.
413
0.30
80.
365
0.23
00.
179
0.10
00.
000
500
0.53
80.
363
0.34
00.
257
0.21
30.
254
0.18
40.
124
0.06
10.
000
500
0.80
70.
511
0.41
90.
300
0.23
60.
266
0.19
00.
127
0.06
20.
000
δ=
1/3
δ=
1/3
TN
Bia
sT
NB
ias
100
100
2.10
40.
973
0.61
70.
712
0.46
10.
578
0.15
9-0
.038
-0.0
16-0
.116
100
100
3.60
32.
048
1.33
81.
180
0.77
60.
785
0.30
50.
063
0.05
4-0
.071
200
1.82
60.
977
0.88
00.
472
0.28
00.
284
0.11
00.
055
-0.0
32-0
.109
200
3.33
81.
987
1.50
20.
869
0.53
00.
445
0.21
90.
130
0.02
3-0
.069
500
1.78
41.
026
0.65
00.
381
0.22
40.
205
0.10
40.
025
-0.0
45-0
.101
500
3.25
51.
936
1.18
60.
692
0.40
80.
317
0.17
80.
079
-0.0
03-0
.065
200
100
2.26
11.
131
0.77
40.
886
0.62
80.
743
0.33
10.
119
0.12
6-0
.001
200
100
3.73
52.
174
1.45
71.
308
0.89
70.
902
0.42
50.
169
0.14
6-0
.001
200
2.01
01.
144
1.03
00.
617
0.42
30.
424
0.25
50.
187
0.09
4-0
.001
200
3.50
42.
131
1.62
70.
983
0.64
00.
547
0.32
30.
221
0.10
8-0
.001
500
1.96
91.
175
0.78
20.
495
0.33
60.
314
0.21
10.
133
0.06
1-0
.001
500
3.45
02.
084
1.30
40.
787
0.49
50.
398
0.25
40.
153
0.06
9-0
.001
500
100
2.30
41.
161
0.79
80.
903
0.64
40.
752
0.34
10.
126
0.13
30.
000
500
100
3.78
02.
208
1.48
91.
325
0.90
60.
905
0.43
10.
173
0.15
10.
000
200
2.07
31.
190
1.06
00.
637
0.43
70.
436
0.26
50.
194
0.10
10.
000
200
3.57
52.
182
1.65
71.
004
0.65
30.
556
0.33
10.
227
0.11
30.
000
500
2.06
21.
233
0.81
70.
516
0.35
10.
325
0.21
90.
139
0.06
60.
000
500
3.55
52.
152
1.34
80.
813
0.51
20.
408
0.26
10.
159
0.07
20.
000
TN
RM
SE
TN
RM
SE
100
100
2.14
81.
029
0.67
10.
750
0.50
50.
610
0.25
50.
186
0.18
00.
212
100
100
3.64
32.
090
1.37
31.
206
0.80
10.
805
0.34
90.
161
0.14
50.
143
200
1.84
81.
002
0.89
70.
496
0.31
20.
314
0.17
30.
145
0.14
20.
182
200
3.35
52.
004
1.51
40.
882
0.54
50.
459
0.24
40.
165
0.10
40.
123
500
1.79
11.
033
0.65
70.
391
0.24
00.
225
0.14
10.
105
0.12
10.
158
500
3.26
01.
940
1.19
00.
697
0.41
40.
325
0.19
10.
108
0.08
00.
108
200
100
2.30
01.
173
0.80
50.
901
0.64
10.
752
0.34
70.
152
0.14
60.
006
200
100
3.77
42.
214
1.48
81.
327
0.91
30.
913
0.44
10.
198
0.16
60.
004
200
2.02
91.
159
1.03
80.
624
0.42
90.
428
0.26
10.
192
0.10
10.
004
200
3.52
12.
147
1.63
70.
992
0.64
70.
551
0.32
90.
226
0.11
40.
003
500
1.97
51.
179
0.78
40.
497
0.33
70.
315
0.21
20.
134
0.06
20.
003
500
3.45
42.
088
1.30
70.
790
0.49
70.
399
0.25
50.
154
0.07
00.
002
500
100
2.34
61.
205
0.83
00.
919
0.65
60.
759
0.35
60.
154
0.15
30.
000
500
100
3.81
92.
250
1.52
11.
344
0.92
20.
916
0.44
70.
199
0.17
00.
000
200
2.09
31.
206
1.06
90.
644
0.44
30.
439
0.27
00.
199
0.10
70.
000
200
3.59
22.
197
1.66
71.
013
0.66
10.
561
0.33
70.
231
0.11
90.
000
500
2.06
71.
237
0.82
00.
518
0.35
30.
326
0.22
00.
140
0.06
70.
000
500
3.55
92.
156
1.35
10.
815
0.51
30.
409
0.26
20.
160
0.07
40.
000
Not
es:
Par
am
eter
sof
the
stat
icp
anel
dat
am
od
el,(2
9),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5),ϑi
=0
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]an
d[N
αβ2]n
on
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vel
y.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
,f jt
anduit∼IIDN
(0,1
)sc
aled
byς
=√ 1
/2,v ij∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
10
Tab
leS
4a:
Com
par
ison
of
Bia
san
dR
MS
E(×
100)
for
theα
andα
esti
mat
esof
the
cros
s-se
ctio
nal
exp
onen
tof
the
erro
rsfr
oma
stat
ican
dd
yn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n1
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Bia
sR
MSE
Sta
tic
model
:ϑi
=0,
fori
=1,
2,...,N
TN
αT
Nα
100
100
2.27
20.
837
0.31
30.
221
0.04
20.
176
-0.1
59-0
.243
-0.1
03-0
.047
100
100
2.87
01.
712
1.03
20.
720
0.45
40.
330
0.28
70.
282
0.14
40.
098
200
2.18
40.
899
0.63
90.
179
0.02
70.
020
-0.1
14-0
.076
-0.0
78-0
.062
200
3.01
81.
642
1.16
10.
731
0.58
60.
219
0.21
40.
139
0.12
00.
112
500
2.13
81.
153
0.48
50.
213
0.03
00.
035
-0.0
36-0
.067
-0.0
90-0
.086
500
3.06
61.
984
1.03
10.
793
0.39
70.
250
0.15
30.
149
0.15
90.
145
200
100
2.68
01.
150
0.59
40.
359
0.15
60.
286
-0.0
90-0
.188
-0.0
500.
000
200
100
3.19
11.
765
1.19
00.
747
0.45
60.
459
0.21
10.
217
0.07
50.
003
200
2.60
41.
309
0.89
00.
323
0.13
40.
111
-0.0
22-0
.004
-0.0
170.
000
200
3.15
51.
943
1.34
90.
662
0.46
20.
240
0.20
00.
076
0.03
40.
002
500
2.64
21.
440
0.79
20.
418
0.17
80.
140
0.06
90.
027
0.00
00.
000
500
3.24
01.
965
1.26
50.
867
0.43
90.
230
0.13
90.
108
0.02
70.
001
500
100
2.98
81.
366
0.67
70.
431
0.20
20.
304
-0.0
79-0
.177
-0.0
470.
000
500
100
3.35
11.
900
1.01
50.
680
0.41
20.
445
0.17
60.
202
0.06
40.
000
200
3.03
51.
522
1.01
20.
399
0.16
40.
157
-0.0
190.
001
-0.0
140.
000
200
3.54
11.
946
1.27
80.
616
0.34
70.
294
0.13
80.
066
0.03
20.
000
500
3.17
01.
719
0.91
00.
449
0.20
30.
162
0.07
90.
029
0.00
30.
000
500
3.70
42.
155
1.29
80.
660
0.40
80.
239
0.12
80.
057
0.04
00.
000
TN
αT
Nα
100
100
2.18
10.
661
0.21
70.
273
0.17
80.
399
0.05
8-0
.009
0.09
80.
104
100
100
3.49
72.
601
2.03
61.
541
1.16
60.
901
0.63
90.
451
0.27
20.
104
200
1.14
50.
159
0.30
60.
023
0.10
00.
174
0.06
80.
105
0.10
80.
093
200
2.49
91.
945
1.62
41.
164
0.94
60.
615
0.41
90.
286
0.19
00.
093
500
0.48
50.
174
0.03
80.
073
0.09
20.
156
0.14
30.
116
0.09
40.
080
500
1.86
21.
757
1.11
70.
943
0.64
80.
468
0.31
10.
223
0.13
40.
080
200
100
3.29
31.
318
0.61
50.
402
0.22
10.
372
0.01
6-0
.075
0.05
00.
050
200
100
4.21
72.
571
2.02
21.
392
0.98
80.
825
0.51
40.
342
0.21
60.
050
200
1.44
60.
283
0.27
40.
061
0.01
40.
117
0.01
00.
066
0.05
70.
045
200
2.31
81.
734
1.22
00.
906
0.65
50.
474
0.38
10.
233
0.13
90.
045
500
0.60
10.
104
0.07
10.
075
0.04
20.
121
0.09
70.
074
0.05
20.
040
500
1.52
81.
103
1.04
00.
757
0.46
40.
330
0.24
10.
210
0.09
40.
040
500
100
8.33
04.
834
2.30
01.
191
0.59
40.
484
0.05
1-0
.078
0.03
30.
018
500
100
8.77
55.
465
2.96
01.
806
1.12
70.
815
0.44
70.
300
0.19
40.
018
200
3.32
61.
257
0.73
30.
201
0.06
60.
137
-0.0
040.
040
0.02
70.
017
200
3.90
51.
988
1.37
20.
775
0.52
90.
408
0.25
40.
179
0.10
40.
017
500
0.97
90.
238
0.11
10.
061
0.02
60.
097
0.06
70.
043
0.02
90.
015
500
1.52
50.
917
0.81
00.
485
0.41
00.
250
0.15
80.
104
0.08
50.
015
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs:ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αT
Nα
100
100
2.22
40.
770
0.29
30.
190
0.02
70.
171
-0.1
78-0
.256
-0.1
10-0
.054
100
100
2.81
81.
527
0.98
10.
768
0.44
40.
457
0.25
70.
301
0.15
60.
110
200
2.03
20.
821
0.58
90.
140
-0.0
320.
005
-0.1
28-0
.090
-0.0
96-0
.077
200
2.73
01.
507
1.09
60.
686
0.41
60.
262
0.21
30.
159
0.15
60.
136
500
2.04
30.
966
0.48
30.
171
0.01
50.
017
-0.0
51-0
.083
-0.1
05-0
.105
500
2.93
91.
547
1.05
50.
623
0.54
90.
314
0.16
50.
161
0.17
00.
175
200
100
2.71
91.
159
0.53
10.
338
0.14
90.
269
-0.0
85-0
.183
-0.0
510.
000
200
100
3.25
61.
787
1.04
20.
722
0.41
40.
404
0.33
10.
244
0.07
10.
002
200
2.63
21.
280
0.89
40.
334
0.11
60.
122
-0.0
25-0
.005
-0.0
160.
000
200
3.29
91.
992
1.49
00.
790
0.42
30.
291
0.20
50.
099
0.05
40.
006
500
2.62
21.
420
0.79
10.
370
0.16
40.
136
0.06
90.
026
0.00
10.
000
500
3.19
51.
991
1.38
60.
626
0.40
30.
229
0.15
50.
075
0.04
00.
002
500
100
2.97
41.
350
0.69
40.
405
0.20
30.
303
-0.0
76-0
.179
-0.0
480.
000
500
100
3.36
31.
821
1.09
50.
623
0.43
60.
403
0.18
90.
210
0.06
50.
000
200
2.91
31.
474
1.00
50.
405
0.17
00.
146
-0.0
13-0
.002
-0.0
140.
000
200
3.25
71.
957
1.35
60.
704
0.37
10.
286
0.23
70.
060
0.03
50.
000
500
3.13
81.
674
0.91
80.
478
0.19
80.
176
0.07
80.
029
0.00
10.
000
500
3.65
11.
969
1.28
70.
764
0.37
00.
331
0.14
30.
053
0.01
50.
000
TN
αT
Nα
100
100
2.10
40.
586
0.25
50.
282
0.17
30.
365
0.06
3-0
.034
0.10
50.
105
100
100
3.36
82.
626
1.98
61.
530
1.17
60.
983
0.61
00.
434
0.27
30.
105
200
1.09
60.
120
0.25
50.
033
0.09
40.
165
0.06
30.
123
0.10
90.
093
200
2.41
51.
888
1.55
11.
145
0.84
70.
638
0.42
90.
309
0.20
40.
093
500
0.46
10.
050
0.07
00.
138
0.12
00.
163
0.13
70.
118
0.09
30.
081
500
1.87
81.
414
1.23
50.
914
0.68
80.
487
0.30
20.
218
0.13
40.
081
200
100
3.37
11.
310
0.53
40.
356
0.21
40.
347
0.01
6-0
.068
0.05
50.
051
200
100
4.32
22.
617
1.82
21.
382
0.95
50.
765
0.57
50.
388
0.22
60.
051
200
1.45
10.
234
0.33
60.
056
0.01
10.
121
0.02
50.
067
0.05
90.
046
200
2.45
91.
627
1.45
90.
955
0.69
40.
499
0.37
20.
252
0.16
20.
046
500
0.60
30.
097
0.04
90.
061
0.04
40.
125
0.09
60.
071
0.05
70.
040
500
1.72
41.
229
1.12
60.
660
0.46
60.
318
0.27
20.
155
0.10
70.
040
500
100
8.19
44.
656
2.33
51.
190
0.55
70.
499
0.05
1-0
.072
0.03
50.
018
500
100
8.62
45.
271
3.00
31.
796
1.08
80.
819
0.45
70.
335
0.18
30.
018
200
3.27
51.
278
0.70
60.
195
0.06
90.
117
0.00
60.
036
0.02
80.
017
200
3.85
22.
096
1.35
20.
771
0.55
20.
407
0.33
20.
173
0.11
60.
017
500
0.97
60.
245
0.12
20.
076
0.03
10.
113
0.06
90.
045
0.02
50.
015
500
1.63
00.
927
0.82
70.
572
0.38
00.
368
0.20
40.
104
0.05
80.
015
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.
Non
-Gau
ssia
ner
rors
are
gen
erat
edasuit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31)
.D
esig
n1
assu
mesb i
∼U
(0.7,0.9
)fo
rth
efi
rstNb
(≤N
)el
emen
tsof
vect
or
bN
,i
=1,
2,...,N
,in
the
con
stru
ctio
nof
the
corr
elati
on
matr
ixof
the
erro
rs,RN
=I N
+bN
b′ N−
B2 N
give
nby
(30)
,w
her
eBN
=D
iag
(bN
).α
isco
mp
ute
du
sin
gc p
(n,δ
)w
ithn
=N
(N−
1)/
2,δ
=1/
2an
dp
=0.
05
inth
em
ult
iple
test
ing
pro
ced
ure
show
nin
(23)
.α
corr
esp
ond
sto
the
most
rob
ust
esti
mato
rof
the
exp
on
ent
of
cros
s-se
ctio
nal
dep
end
ence
con
sid
ered
inB
ail
eyet
al.
(201
6)an
dal
low
sfo
rb
oth
seri
alco
rrel
atio
nin
the
fact
ors
an
dw
eak
cross
-sec
tion
al
dep
end
ence
inth
eer
ror
term
s.W
euse
fou
rp
rin
cip
al
com
pon
ents
wh
enes
tim
atin
gc N
inth
eex
pre
ssio
nfo
rα
.T
he
nu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
11
Tab
leS
4b
:C
omp
aris
onof
Bia
san
dR
MS
E(×
100)
for
theα
andα
ofth
ecr
oss-
sect
ion
alex
pon
ent
of
the
erro
rsof
ast
ati
can
dd
yn
amic
pan
eld
ata
mod
elw
ith
exog
enou
sre
gres
sors
Cro
ssco
rrel
ati
on
sare
gen
erate
du
sin
gD
esig
n2
wit
hn
on
-Gau
ssia
ner
rors
α0.
550.
600.
650.
700.
750.
800.
850.
900.
951.
00α
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Bia
sB
ias
Sta
tic
model
:ϑi
=0,
fori
=1,
2,...,N
TN
αα
100
100
1.11
6-0
.949
-2.0
57-2
.527
-3.0
24-3
.083
-3.5
51-3
.745
-3.7
13-3
.754
100
100
2.16
51.
928
2.63
13.
035
3.49
93.
576
3.99
94.
181
4.16
24.
206
200
0.93
3-1
.004
-1.9
30-2
.775
-3.2
10-3
.357
-3.6
31-3
.693
-3.7
77-3
.844
200
2.17
91.
962
2.49
93.
204
3.62
43.
796
4.06
64.
140
4.22
74.
292
500
0.89
8-1
.048
-2.2
61-3
.008
-3.4
56-3
.615
-3.7
82-3
.901
-3.9
98-4
.063
500
2.48
92.
230
2.88
93.
493
3.90
54.
068
4.23
74.
357
4.45
24.
516
200
100
2.76
41.
265
0.64
20.
535
0.22
10.
270
-0.1
59-0
.346
-0.3
18-0
.394
200
100
3.21
31.
794
1.13
10.
851
0.54
10.
478
0.38
30.
473
0.45
10.
508
200
2.65
51.
332
0.88
60.
307
0.02
0-0
.036
-0.2
26-0
.283
-0.3
61-0
.430
200
3.21
61.
863
1.23
10.
650
0.38
80.
317
0.37
20.
413
0.47
20.
532
500
2.62
31.
359
0.64
10.
190
-0.0
76-0
.138
-0.2
58-0
.342
-0.4
12-0
.460
500
3.27
91.
932
1.10
20.
598
0.39
40.
343
0.40
00.
467
0.52
90.
571
500
100
3.33
71.
824
1.18
01.
096
0.74
10.
794
0.34
90.
121
0.12
5-0
.002
500
100
3.56
92.
057
1.36
01.
190
0.80
60.
821
0.37
70.
156
0.14
40.
010
200
3.16
31.
851
1.41
70.
822
0.52
50.
467
0.27
00.
187
0.09
4-0
.002
200
3.46
62.
107
1.55
60.
915
0.57
80.
486
0.28
00.
193
0.10
00.
004
500
3.18
61.
896
1.17
40.
697
0.43
10.
353
0.22
30.
135
0.06
0-0
.002
500
3.60
52.
250
1.43
20.
871
0.53
00.
390
0.23
80.
140
0.06
20.
004
TN
αT
Nα
100
100
2.00
70.
792
0.56
40.
800
0.58
00.
729
0.32
70.
146
0.19
90.
104
100
100
3.30
32.
266
1.72
41.
516
1.11
41.
016
0.62
50.
395
0.30
10.
104
200
0.64
50.
087
0.43
70.
268
0.24
10.
330
0.21
40.
199
0.14
90.
092
200
2.03
41.
611
1.25
90.
928
0.70
00.
580
0.41
70.
308
0.20
20.
092
500
0.00
6-0
.107
0.05
70.
124
0.15
10.
237
0.19
50.
154
0.11
00.
080
500
1.64
91.
307
0.94
60.
667
0.48
00.
400
0.29
80.
216
0.13
70.
080
200
100
3.80
42.
191
1.41
61.
521
0.96
30.
939
0.45
90.
196
0.19
70.
051
200
100
4.58
93.
065
2.15
52.
028
1.40
01.
209
0.73
10.
422
0.30
50.
051
200
1.49
30.
719
0.81
60.
501
0.34
20.
382
0.23
20.
190
0.12
10.
046
200
2.26
01.
570
1.33
50.
929
0.65
10.
560
0.38
60.
271
0.16
90.
046
500
0.44
70.
118
0.21
40.
199
0.18
30.
238
0.17
80.
129
0.08
10.
040
500
1.25
80.
876
0.67
80.
493
0.37
50.
337
0.24
30.
170
0.10
10.
040
500
100
9.13
95.
921
3.98
13.
658
2.47
82.
109
1.24
50.
613
0.39
30.
019
500
100
9.49
66.
288
4.31
23.
896
2.76
22.
355
1.49
30.
836
0.53
50.
019
200
3.78
62.
622
2.23
01.
557
0.96
30.
765
0.48
70.
280
0.15
30.
018
200
4.22
43.
014
2.53
51.
877
1.24
80.
976
0.67
20.
394
0.22
10.
018
500
0.99
30.
664
0.72
00.
436
0.31
50.
290
0.18
50.
122
0.06
60.
016
500
1.41
21.
062
1.01
90.
667
0.49
20.
382
0.24
20.
158
0.08
60.
016
Dynam
icm
odel
wit
hex
ogen
ous
regr
esso
rs:ϑi∼U
(0,0.9
5)fo
ri
=1,
2,...,N
TN
αT
Nα
100
100
1.01
6-1
.103
-2.2
41-2
.771
-3.2
70-3
.357
-3.8
38-4
.045
-4.0
08-4
.036
100
100
2.11
72.
014
2.78
23.
266
3.74
93.
863
4.30
64.
507
4.48
84.
515
200
0.81
3-1
.192
-2.1
65-3
.069
-3.5
34-3
.708
-3.9
69-4
.029
-4.1
18-4
.184
200
2.21
92.
155
2.74
43.
521
3.96
94.
155
4.40
64.
474
4.55
94.
626
500
0.66
9-1
.297
-2.5
34-3
.294
-3.7
52-3
.926
-4.1
00-4
.216
-4.3
12-4
.380
500
2.42
42.
367
3.10
43.
732
4.15
14.
331
4.50
64.
622
4.71
74.
785
200
100
2.74
91.
249
0.61
90.
506
0.18
00.
235
-0.1
94-0
.387
-0.3
59-0
.431
200
100
3.19
11.
751
1.07
00.
792
0.49
20.
453
0.40
70.
521
0.50
20.
555
200
2.57
81.
267
0.83
40.
256
-0.0
18-0
.070
-0.2
68-0
.316
-0.3
97-0
.464
200
3.13
51.
816
1.21
00.
658
0.43
10.
357
0.42
40.
456
0.52
40.
579
500
2.56
81.
311
0.61
00.
161
-0.0
98-0
.161
-0.2
81-0
.366
-0.4
36-0
.485
500
3.27
01.
965
1.18
10.
708
0.48
60.
394
0.42
50.
483
0.54
50.
591
500
100
3.41
91.
888
1.22
21.
113
0.74
90.
792
0.34
90.
125
0.12
4-0
.002
500
100
3.67
42.
127
1.38
61.
183
0.79
10.
808
0.36
60.
154
0.14
30.
006
200
3.20
21.
882
1.43
90.
839
0.53
50.
471
0.27
20.
191
0.09
2-0
.002
200
3.61
82.
254
1.66
11.
001
0.63
10.
509
0.29
10.
198
0.09
90.
006
500
3.16
61.
879
1.15
80.
685
0.42
20.
348
0.22
10.
133
0.05
9-0
.002
500
3.50
22.
134
1.31
90.
778
0.46
80.
362
0.22
60.
135
0.06
10.
003
TN
αT
Nα
100
100
1.85
40.
704
0.48
10.
745
0.54
20.
707
0.30
50.
139
0.18
60.
105
100
100
3.17
72.
262
1.72
51.
472
1.08
91.
006
0.61
40.
388
0.28
70.
105
200
0.60
90.
031
0.38
50.
237
0.21
50.
316
0.20
20.
192
0.14
40.
093
200
2.04
21.
618
1.24
40.
914
0.69
00.
574
0.41
20.
303
0.19
90.
093
500
0.01
0-0
.106
0.06
00.
124
0.15
10.
242
0.19
70.
156
0.11
20.
080
500
1.52
21.
155
0.86
20.
636
0.47
60.
402
0.29
70.
215
0.13
90.
080
200
100
3.81
62.
133
1.37
11.
412
0.90
10.
891
0.42
40.
181
0.18
20.
051
200
100
4.63
63.
039
2.15
21.
945
1.34
41.
159
0.69
70.
422
0.29
40.
051
200
1.46
40.
701
0.81
10.
496
0.35
20.
379
0.22
80.
185
0.11
80.
046
200
2.27
31.
579
1.35
10.
952
0.68
20.
575
0.38
60.
274
0.16
90.
046
500
0.44
10.
101
0.20
40.
190
0.17
80.
237
0.17
80.
128
0.08
10.
040
500
1.40
41.
045
0.81
30.
565
0.41
00.
343
0.24
50.
169
0.10
20.
040
500
100
9.05
95.
866
3.98
03.
556
2.41
71.
994
1.20
20.
591
0.37
00.
019
500
100
9.45
06.
268
4.33
33.
831
2.71
92.
248
1.46
50.
843
0.51
50.
019
200
3.74
72.
521
2.19
31.
494
0.92
50.
732
0.47
80.
272
0.14
80.
018
200
4.20
12.
926
2.50
21.
816
1.23
30.
955
0.66
90.
387
0.21
80.
018
500
1.03
00.
646
0.70
50.
433
0.32
10.
286
0.19
10.
125
0.06
60.
016
500
1.45
61.
044
1.00
60.
661
0.49
60.
375
0.25
00.
161
0.08
60.
016
Not
es:
Rem
ain
ing
par
amet
ers
ofth
ep
anel
dat
am
od
el,
(29),
are
gen
erate
das:ai∼IIDN
(1,1
),ρix
∼U
(0,0.9
5)
an
dγi∼IIDN
(1,1
),fo
ri
=1,
2,...,N
.D
esig
n2
assu
mes
atw
o-fa
ctor
mod
elw
ith
[Nαβ]
and
[Nαβ2]
non
-zer
olo
ad
ings
for
the
firs
tan
dse
con
dfa
ctor,
resp
ecti
vely
.W
ese
t:αβ
2=
2αβ/3
,w
her
eαβ
rela
tes
toα
un
der
(11)
andf jt∼IIDN
(0,1
).G
auss
ian
erro
rsare
giv
enby:uit∼IIDN
(0,1
)an
dn
on
Gau
ssia
ner
rors
are
gen
erate
das:uit
=( v−
2χ2 v,t
) 1/2ν it,
fori
=1,
2,...,N
,ν it∼IIDN
(0,1
)an
dχ
2 v,t
isa
chi-
squ
are
dra
nd
om
vari
ate
wit
hv=
8d
egre
esof
free
dom
,in
(31).v ij∼IIDU
(µvj−
0.2,µvj
+0.2
),j
=1,
2,
µv
=0.
87,µv2
=0.
71,µv1
=√ µ
2 v−N
2(αβ2−αβ
)µ
2 v2),
in(3
2)
an
d(3
3).
Th
enu
mb
erof
rep
lica
tion
sis
set
toR
=2000.
12
Appendix B
Additional Empirical results
The figures below provide the estimates displayed in Figures 1 and 2 but using the shorter5-year rolling samples.
Figure 3: 5-year rolling estimates of the exponent of cross-sectional correlation (αt) of S&P 500securities’ excess returns
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
5-year rolling estimates of α ( with δ=1/2) based on excess returns5-year rolling estimates of α ( with δ=1/3) based on excess returns
αα
Figure 4: 5-year rolling estimates of the exponent of cross-sectional correlation (αt) of residualsfrom CAPM and its two Fama-French extensions
0.55
0.60
0.65
0.70
0.75
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
5-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model
5-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by the SMB factor
5-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by SMB and HML factors
α
α
α
Notes: CAPM model includes excess market returns, CAPM model augmented by SMB includesexcess market returns and small minus big (SMB) firm returns, and CAPM model augmented bySMB and HML includes excess market returns, small minus big (SMB) firm returns and high minuslow (HML) firm returns as regressors in (35), (36) and (37), respectively.
13
The figures below provide the estimates displayed in Figures 1 and 2 including their 95%confidence bands.
Figure 5: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of S&P500 securities’ excess returns, with 90% confidence intervals
0.80
0.82
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98Se
p-89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/2) based on excess returns
10-year rolling estimates of 0.05 percentile of α's empirical distribution
10-year rolling estimates of 0.95 percentile of α's empirical distribution
α
Notes: In the critical value, cp(n, δ) we set p = 0.05 and δ = 1/2.
Figure 6: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of S&P500 securities’ excess returns, with 90% confidence intervals
0.80
0.82
0.84
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/3) based on excess returns
10-year rolling estimates of 0.05 percentile of α's empirical distribution
10-year rolling estimates of 0.95 percentile of α's empirical distribution
α
Notes: In the critical value, cp(n, δ) we set p = 0.05 and δ = 1/3.
14
Figure 7: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of residualsfrom CAPM model, with 90% confidence intervals
0.60
0.62
0.64
0.66
0.68
0.70
0.72
0.74
0.76
0.78
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model
10-year rolling estimates of 0.05 percentile of α's empirical distribution
10-year rolling estimates of 0.95 percentile of α's empirical distribution
α
Notes: CAPM model includes excess market returns as regressor in (35). In the criticalvalue, cp(n, δ) we set p = 0.05 and δ = 1/2. Confidence intervals are computed using theboostrap procedure described in Section 6.
Figure 8: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of residualsfrom CAPM model augmented by SMB, with 90% confidence intervals
0.60
0.62
0.64
0.66
0.68
0.70
0.72
0.74
0.76
0.78
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
17
10-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by the SMB factor
10-year rolling estimates of 0.05 percentile of α's empirical distribution
10-year rolling estimates of 0.95 percentile of α's empirical distribution
α
Notes: CAPM model augmented by SMB includes excess market returns and small minusbig (SMB) firm returns as regressors in (36). In the critical value, cp(n, δ) we set p = 0.05and δ = 1/2. Confidence intervals are computed using the boostrap procedure described inSection 6.
15
Figure 9: 10-year rolling estimates of the exponent of cross-sectional correlation (αt) of residualsfrom CAPM model augmented by SMB and HML, with 90% confidence intervals
0.60
0.62
0.64
0.66
0.68
0.70
0.72
0.74
0.76
0.78
Sep-
89
Sep-
91
Sep-
93
Sep-
95
Sep-
97
Sep-
99
Sep-
01
Sep-
03
Sep-
05
Sep-
07
Sep-
09
Sep-
11
Sep-
13
Sep-
15
Sep-
1710-year rolling estimates of α ( with δ=1/2) for the residuals of the CAPM model augmented by SMB and HML factors10-year rolling estimates of 0.05 percentile of α's empirical distribution
10-year rolling estimates of 0.95 percentile of α's empirical distribution
α
Notes: CAPM model augmented by SMB and HML includes excess market returns, smallminus big (SMB) firm returns and high minus low (HML) firm returns as regressors in (37).In the critical value, cp(n, δ) we set p = 0.05 and δ = 1/2. Confidence intervals are computedusing the boostrap procedure described in Section 6.
16