Regional Science and Urban Economics 45 (2014) 1–21

Contents lists available at ScienceDirect

Regional Science and Urban Economics

journal homepage: www.elsevier.com/locate/regec

Spatial autoregressive models with unknown heteroskedasticity:

A comparison of Bayesian and robust GMM approach☆
Osman Doğan ⁎, Süleyman Taşpınar
Program in Economics, The Graduate School and University Center, The City University of New York, New York, United States

a r t i c l e i n f o a b s t r a c t

Article history: Most of the estimators suggested for the estimation of spatial autoregressive models are generally inconsistent in
Received 29 August 2013 the presence of an unknown form of heteroskedasticity in the disturbance term. The estimators formulated from
Received in revised form 9 December 2013 the generalized method of moments (GMM) and the Bayesian Markov Chain Monte Carlo (MCMC) frameworks
Accepted 20 December 2013
can be robust to unknown forms of heteroskedasticity. In this study, the finite sample properties of the robust
Available online 4 January 2014
GMM estimator are compared with the estimators based on the Bayesian MCMC approach for the spatial
JEL classification:
autoregressive models with heteroskedasticity of an unknown form. A Monte Carlo simulation study provides
C13 evaluation of the performance of the heteroskedasticity robust estimators. Our results indicate that the MLE
C21 and the Bayesian estimators impose relatively greater bias on the spatial autoregressive parameter when there
C31 is negative spatial dependence in the model. In terms of finite sample efficiency, the Bayesian estimators perform
better than the robust GMM estimator. In addition, two empirical applications are provided to evaluate relative
Keywords: performance of heteroskedasticity robust estimators.
Spatial autoregressive models © 2013 Elsevier B.V. All rights reserved.
Unknown heteroskedasticity
Robustness
GMM
Asymptotics
MLE
Markov Chain Monte Carlo (MCMC)

1. Introduction numbers of observations when the data is getting aggregated (Griffiths,

Most of the estimation methods suggested in the literature are valid
under the assumption that the disturbance terms of spatial models
are independently and identically distributed (i.i.d.) or i.i.d. normal.
However, in many regression applications, heteroskedasticity may well
be present. For example, cross-sectional units usually differ in size and
some other characteristics, which in turn implies that the disturbance
terms in the regression analyses across these cross-sectional units may
be heteroskedastic. It may also be present in a random coefficient
model, where the parameters of the model are random around fixed
values. In this case, heteroskedasticity depends on the exogenous vari-
ables of the spatial models. Moreover, in regression analysis, many de-
pendent variables are constructed by data aggregation. In such a case,
heteroskedasticity arises from the process of averaging with different
☆ We would like to thank Wim Vijverberg for the insightful and instructive comments
on this study. We are also grateful to James P. LeSage for his helpful comments for the im-
provement of this study. This research was supported, in part, by a grant of computer time
from the City University of New York High Performance Computing Center under NSF
Grants CNS-0855217 and CNS-0958379.
⁎ Corresponding author at: The Graduate Center, Economics 365 Fifth Avenue NY, NY
10016-4309, United States. Tel.: +1 212 817 8255; fax: +1 212 817 1514.
E-mail addresses: odogan@gc.cuny.edu (O. Doğan), STaspinar@gc.cuny.edu (S. Taşpınar).
2007; Lee, 2010).1
In the present study, we evaluate the performance of various
heteroskedasticity robust estimators suggested in the literature for
spatial autoregressive models. To this end, we conduct a Monte Carlo
study and provide two empirical illustrations to show how these esti-
mators perform in applied research.
In the presence of heteroskedastic disturbances, the ML and GMM
estimators are generally inconsistent. The ML estimator is inconsistent
if heteroskedasticity is not incorporated into estimation, because the
likelihood function is not maximized at the true parameter values.2
The GMM estimators are also inconsistent since the moment functions
are often designed under the assumption that disturbances are i.i.d.
(Kelejian and Prucha, 1998, 1999; Liu et al., 2010). Hence, the orthogo-
nality conditions for the moment functions may not be satisfied.
To handle unknown forms of heteroskedasticity, Kelejian and Prucha
(2010) extend their two-step GMM estimation approach by modifying
the moment functions for a spatial model that has a spatial lag in the
1
For some recent empirical studies, see Lin and Lee (2010) and Doğan and Taşpınar
(2013b).
2
For many spatial model specifications, the ML estimation has been the most widely
used technique and has often been the only technique that is implemented. The ML ap-
proach is well treated in Anselin (1988), Lee (2004) and LeSage and Pace (2009).

0166-0462/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.regsciurbeco.2013.12.003
2 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21
dependent variable and the disturbance term (for short SARAR(1,1)).
Badinger and Egger (2011) extend the robust estimation approach in
Kelejian and Prucha (2010) to the case of SARAR(p,q) specification.
Lin and Lee (2010) suggest a one-step robust GMM estimator for the
model with only spatial dependence in the dependent variable (for
short SARAR(1,0)).3 In this approach, the parameters of a spatial
model are simultaneously estimated by a GMM estimator formulated
from the combination of a set of linear and quadratic moment functions
(Lee, 2007a, 2007b; Lee and Liu, 2010; Liu et al., 2010).
The moment functions considered for both two-step and one-step
GMM estimators are motivated by the reduced form of the spatial
models. For example, the reduced form for the case of SARAR(1,0)
indicates that the endogenous spatial lag of the dependent variable is
a function of a stochastic and a non-stochastic variable. The linear mo-
ment functions are formulated from the non-stochastic part and the
quadratic moment functions are formulated for the stochastic part.
The two-step GMM approach of Kelejian and Prucha is motivated by
computational simplicity as the ML estimation involves significant com-
putational burden for the large samples. Despite the computational
advantage, the two step GMM estimator is inefficient relative to the
one-step GMM estimators suggested in Lee (2007a, 2007b), Liu et al.
(2010) and Lee and Liu (2010).
An alternative to ML and GMM/IV estimation methods is the Bayesian
estimation method, which has been receiving attention in recent
years. Bayesian data analysis is distinctly different from the classical
(or frequentist) analysis in its treatment of parameters of the model.
In Bayesian econometrics the parameter vector is a random variable,
and Bayesian analysts formulate probabilistic statements about the
parameters before observing any data. These ex-ante probabilities
are called priors and usually take the form of a probability distribu-
tion with known moments. This notion of the prior probabilities
(or subjective probabilities) is totally absent in classical economet-
rics where all estimation and inference are based on observed data.
In both approaches, the likelihood function has the same functional
form and reflects the relation between data and the parameters of inter-
est. In Bayesian approach, the likelihood function is combined with prior
functions via Bayes' rule to construct the posterior probability distribution
of parameters. The posterior distribution function of the parameter vector
contains the information necessary for the estimation and inference.
The Bayesian approach requires the evaluation of higher dimensional
integrals to obtain posterior expectations, marginal likelihoods, and
predictive densities. The application of the Bayesian methods to the
estimation of spatial models follows the path of the progress that has
been made in the context of Bayesian computation techniques. Hepple
(1995b) identifies four phases for the development of Bayesian computa-
tion techniques. In the first phase, Bayesian studies involved problems
that can be characterized with well known probability distributions
such that the characteristics of posterior distributions such as means
and covariances can be analytically derived. In the second phase, Bayesian
studies focused on techniques through which problems involving
multidimensional probability distributions can be reduced to univariate
or bivariate integrations. In this phase, numerical techniques for the uni-
variate and bivariate integration were used. In the third phase, Bayesian
analysts worked on efficient procedures for higher order integrations.
Gauss–Hermite, importance sampling and Monte Carlo integration tech-
niques were used to tackle complicated and high-dimensional problems.
In the fourth phase, Markov Chain Monte Carlo (MCMC) simulation tech-
niques were introduced that make the computation of higher dimension
problems feasible. The advent of the MCMC approach represents a shift in
thinking, where the focus on the question of analytical moment calcula-
tion is replaced with a more general question of sampling issues from
the posterior distributions (Albert and Chib, 1993; Casella and George,
1992; Chib, 2001). In the Gibbs sampling version of the MCMC approach,
3
For a robust 2SLS estimator of the SARAR(1,0) specification, see Anselin (2006).
the joint posterior distribution is decomposed into conditional posterior
distributions through which random draws (or a simulated sample) can
be obtained. Inferences such as posterior mean and posterior covariance
matrix can be estimated from the simulated sample obtained from the
conditional posterior densities.
The development in Bayesian computation techniques provides a
wide range of tools that can be applied to the estimation of spatial
models. The early literature on the Bayesian perspective on spatial
models uses combinations of tools developed during the period from
the first to the third phase. For example, Hepple (1979) analytically
derives the joint posterior and marginal posterior distributions of
parameters for a spatial model containing a spatial lag in the distur-
bance term. The posterior moments are calculated through numerical
univariate and bivariate integration techniques.
Anselin (1982, 1988) considers the Bayesian approach for pure
spatial autoregressive and spatial error models. Diffuse priors for
parameters of models are suggested and marginal posterior distribu-
tions of parameters are analytically derived. The posterior mean of
the autoregressive parameter for a pure spatial autoregressive model in
Anselin (1982) is estimated with univariate numerical integration. A
small Monte Carlo simulation study in Anselin (1982) demonstrates
that the Bayesian estimator performs as well as the ML estimator for
larger values of the autoregressive parameter and larger samples.4
Hepple (1995a, 1995b) develops Bayesian analyses for major
spatial specifications including the SARAR(1,0) model, SARAR(0,1)
(or the SEM) model and spatial moving average models (for short
SARMA(0,1)). In each case, the joint posterior distributions of the param-
eters are stated from which the marginal posterior of the spatial
autoregressive parameters is analytically derived. The analytical deriva-
tion for the marginal posterior distribution of the parameters of the
exogenous variables is not available as the spatial autoregressive param-
eters can not be analytically integrated out from the joint posterior distri-
bution. However, the dimension of joint posterior can be reduced to two
dimensions so that bivariate numerical integration techniques can be
used for the estimation of the marginal posterior moments. As a result,
the estimate of the spatial autoregressive parameters can be obtained
through univariate numerical integration, and the estimates of the pa-
rameters of the exogenous variables can be obtained through bivariate
numerical integration over a grid of pairs of values.5
Hepple (2002, 2003) provides further analytical simplifications such
that the moments of the marginal posterior distributions of the
exogenous variables in spatial models with only one autoregressive pa-
rameter can be obtained through univariate numerical integration.
However, for the case of SARAR(1,1) and SARMA(1,1) where there are
two spatial autoregressive parameters the calculation of these moments
again requires bivariate integration over the parameter space of
autoregressive parameters.
The recent studies use the MCMC approach to estimate spatial
models. This approach is more appropriate for cases where marginal
posterior distributions are difficult to simplify analytically and to inte-
grate numerically. The MCMC approach is introduced for most types
of spatial models in LeSage (1997) and LeSage and Pace (2009).
Kakamu and Wago (2008) compare finite sample properties of the
Bayesian estimators based on the MCMC approach with that of the ML
estimator for the static panel spatial autoregressive model. The Monte
Carlo simulation results in Kakamu and Wago (2008) show that the
Bayesian estimator is virtually as efficient as the ML estimator.
In spatial models, the boundaries of the parameter space for
spatial autoregressive parameters are known, which facilitate the
4
To the best of our knowledge, Anselin (1982) is the first study, where small sample
properties of the ML estimator are compared with the frequentist properties of the Bayesian
estimator in the context of spatial models.
5
Note that as the dimension of spatial parameter increases, the dimension of numerical
integration rises. For example, for the case of SARAR(1,1) the marginal posterior of
autoregressive parameters is two dimensional.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21
selection of the proper uninformative priors for these parameters.
Thus, in most studies, uniform priors over the parameter space are
assigned to spatial autoregressive parameters (as in LeSage and Pace,
2009; Parent and LeSage, 2008; LeSage, 1997; Kakamu and Wago,
2008). Oliveira and Song (2008) consider two versions of Jeffreys' prior,
called independence Jeffreys and Jeffreys-rule priors for the spatial
autoregressive parameter for the case of SARAR(0,1). Jeffreys priors are
uninformative improper priors that are constructed from the information
matrix. The appealing aspect of these priors is that they are robust to re-
parametrization of the model. For the case of SARAR(0,1), Oliveira and
Song (2008) state the marginal posterior density of autoregressive
parameter (up to a constant of proportionality) and use the adaptive
rejection Metropolis sampling algorithm to simulate random draws. In a
Monte Carlo study, they show that the Bayesian inferences based on the
independence Jeffreys prior are slightly better than the inferences
obtained based on uniform and Jeffreys-rule priors.
The Bayesian estimators mentioned above are designed for the cases
where spatial models have homoskedastic disturbances. The treatment
of the unknown form of heteroskedasticity in the Bayesian approach
starts by assuming a structure for the unknown covariance matrix of
disturbances. First, it is assumed that the covariance matrix can be
decomposed into a constant component and a component that varies
over observations. Second, a prior density function is assigned to each
component so that the joint posterior density of the parameters can
be constructed. The most important assumption is that the component
that varies over observations is assumed to be i.i.d. with a common
prior density function. In this approach, a prior density function for
the parameter of the common density function is also specified. In the
literature, the priors determined in this way are known as hierarchical
priors. This feature of the Bayesian approach allows flexible models
through which any distribution for the disturbance term can be approx-
imated with a high degree of accuracy (Koop, 2003). Once the prior
density functions are determined, the joint posterior density function
can be easily constructed, from which a full set of conditional posterior
density functions can be obtained to form an MCMC sampling scheme.
The random draws (or the simulated sample) are then used to calculate
the relevant moments for the estimation.
LeSage (1997) is the first study in which the above outlined hierar-
chical Bayesian approach of unknown heteroskedasticity is introduced
for the spatial autoregressive models. Following Geweke (1993);
LeSage (1997) assigns an i.i.d. Chi-square distribution as the prior
for the elements of the component of the covariance matrix that
is assumed to vary over observations. The sampling for the spatial
autoregressive parameter is carried out through the ratio of uniform
sampling algorithm. LeSage and Pace (2009) use a Metropolis Hasting
algorithm based on a tuned random walk procedure to sample from
the conditional posterior density function of autoregressive parameters
to form the Metropolis within Gibbs sampling procedure for the whole
model. Parent and LeSage (2008) consider the hierarchical Bayesian
approach for the estimation of a conditional autoregressive (CAR)
spatial model, in which the spatial structured random component of
the model is assumed to have an autoregressive process.
It is of interest to investigate the performance of the aforementioned
heteroskedasticity-robust GMM and Bayesian estimators through a
Monte Carlo study and applied applications. In this study, we compare
the small sample properties of these estimators through a Monte Carlo
study.6 In this respect, we make two contributions to the literature.
First, this study is the first to compare the mentioned approaches of
the unknown heteroskedasticity.7 For the case of SARAR(1,0), the robust
6
To the best of our knowledge, there is no such study in the literature.
7 dent variable. By this assumption, the reduced form of the model
Anselin (1988, p.254) states promising directions of research and he writes “An as-
sessment of the relative merits of alternative paradigms for the analysis of spatial data,
such as parametric analysis vs. nonparametrics, classical probability vs. Bayesian ap-
proaches, and exploratory vs. explanatory analysis” as the first direction. The present
study is in this first direction as it compares classical probability and Bayesian approaches
of the estimation.
3
estimation methods from both perspectives are outlined. Second,
we provide two empirical examples to shed light on the effect of
heteroskedasticity on the parameter estimates.
In the following, Section 2 provides model assumptions that are
required for the formal study of the autoregressive spatial models.
Sections 3 and 4 build the robust GMM approach and the Bayesian hier-
archical approach for the case of SARAR(1,0). Section 5 contains the
Monte Carlo experiments we use to provide finite sample properties
of the estimators. In Section 6, we provide two empirical illustrations.
Section 7 concludes.
2. Model specification and assumptions
In this study, the following first order SARAR(1,0) specification is
considered:
Y n ¼ λ0 W n Y n þ X n β0 þ εn ; ð2:1Þ
where Yn is the n × 1 vector of dependent variable, Xn is the n × k ma-
trix of non-stochastic exogenous variables, Wn is the n × n spatial
weight matrix of known constants with zero diagonal elements, and
εn is the n × 1 vector of disturbances (or innovations). The variable
WnYn is known as the spatial lag of the dependent variable. The spatial
effect parameter λ0 is known as the spatial autoregressive parameter.
As spatial data are characterized with triangular arrays, the variables
in Eq. (2.1) have subscript n.8 Let Θ be the parameter space of the
model. In order to distinguish the true parameter vector from other pos-
sible values in Θ, the model is stated with the true parameter vector
θ0 = (λ0,β0′)′.
For the notational simplicity, we denote Sn(λ) = (In–λWn) and
Gn(λ) = WnS− 1
n (λ). Also, at the true parameter value λ 0 , we de-
note S n (λ0 ) = S n and Gn (λ0 ) = Gn . Let Diag(∙) be the operator
that creates a diagonal matrix from the diagonal elements of the
input matrix. Finally, let ∑n = Diag(σ2n1,…,σ2nn) be n × n diagonal
covariance matrix, where σ2ni = E(ε2ni), for i = 1,…,n.
Next, assumptions that are required for the asymptotic properties of
estimators are elaborated and then their interpretations are considered
for the case of SARAR(1,0).
Assumption 1. The elements εni of the disturbance term εn are distrib-
uted independently with mean zero and variance σ2ni, and E|εin|v b ∞ for
some v N 4 for all n and i.
This assumption allows independent and heteroskedastic distur-
bances. The elements of the disturbance term have moments higher
than the fourth moment. The existence of moment condition is required
for the application of the central limit theorem for the quadratic form
given in Kelejian and Prucha (2010) for the GMM estimator. In addition,
the variance of quadratic form in εn exists and is finite when the first
four moments are finite.9 Finally, Lyapunov's inequality guarantees that
the moment less than ν is also uniformly bounded for all n and i.
Assumption 2. The spatial weight matrix Wn is uniformly bounded in
absolute value in row and column sums. Moreover, S− n
1
and S− 1
n (λ)
exist and are uniformly bounded in absolute value in row and column
sums for all values of λ in a compact parameter space.
The uniform boundedness of the terms in Assumptions 1 and 2 is
motivated to control spatial autocorrelations in the model at a tractable
level (Kelejian and Prucha, 1998).10 Assumption 2 also implies that the
model in Eq. (2.1) represents an equilibrium relation for the depen-
8
See Kelejian and Prucha (2010).
9
For the variance of a quadratic form in εn, see Lemma 2(3).
10
For a definition and some properties of uniform boundedness see Kelejian and Prucha
(2010).
4 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

becomes feasible as Yn = S−1 −1

n Xnβ0 + Sn εn. Finally, the statement of on Pjn is quadratic in εn. For the selection of the quadratic moment
Assumption 2 is assumed to hold at the true and arbitrary autoregressive matrices Pjn s, the asymptotic efficiency of the GMM estimator is consid-
parameter vector. The uniform boundedness of S−1 n (λ) is required for ered. Lee (2007a) shows that when disturbance terms are i.i.d., the best
the ML estimator, not for the GMM estimator (Liu et al., 2010). selection of Pjn ∈ P2n is P∗n = (Gn − Diag(Gn)). Lin and Lee (2010)
In the literature, the parameter space for spatial autoregressive suggest that this selection can also be used for the case where distur-
parameter λ0 is restricted to the interval (−1,1), when spatial weight bance terms satisfy Assumption 1. Consider the following set of moment
matrices are row normalized.11 functions
The following assumptions are the usual regularity conditions
required for the GMM estimator. Throughout this study, the vector of mo- !
′
ment functions considered for the GMM estimator is in the form of εn ðθÞðGn −Diag ðGn ÞÞεn ðθÞ
g n ðθÞ ¼ ′ ; ð3:1Þ
g(θ0) = (εn′P1nεn,…,εn′Pmnεn, εn′Qn)′. The moment functions involving ðGn X n β0 ; X n Þ εn ðθÞ
n × n constant matrices Pjn for j = 1,…,m are known as quadratic mo-
ment functions. The last moment function Qn′εn is the linear moment func-
where εn(θ) = Sn(λ)Yn − Xnβ. Let Ωn = E[g n(θ0)g n′(θ0)] and Γ n ¼
tion, where Qn is an n × k⁎ with full column rank and with k⁎ ≥ k + 1.  
∂gn ðθ0 Þ
The matrices Pjns and Qn are chosen in such way that orthogonality con- −E ′
. Straightforward calculations show that13
ditions of population moment functions are not violated. Let P 2n be the ∂θ
class of n × n constant matrices with zero diagonal elements. The qua- 0
X
′  1
  
dratic moment functions involving matrices from this class satisfy the @ tr P P n Σn þ Σn P n 01ðkþ1Þ A
Ωn ¼ v n ; ð3:2Þ
orthogonality conditions when disturbance terms satisfy Assumption 1. 0ðkþ1Þ1 Q ′n Σn Q n
Assumption 4 states regularity conditions for these matrices and the
last assumption characterizes the parameter space. 0

  1
 ′
Γn ¼ @ tr Σn P n þ P n Gn 01k A
; ð3:3Þ
Assumption 3. The regressors matrix Xn is an n × k matrix consisting of ′ ′
Q n Gn X n β0 Q nXn
uniformly bounded constant elements. It has full column rank of k.
Moreover, limn→∞ 1n X ′n X n exists and is nonsingular.
where Σn = Diag(σ2n1,…,σ2nn) is again the n × n covariance matrix of dis-
Assumption 4. Elements of IV matrix Q n are uniformly bounded. turbances. Then, the robust GMM estimator is defined by
Pjn ∈ P 2n for j = 1,…,m is uniformly bounded in absolute value in  
^
λ
row and column sums. ^θ ¼ n ′ ^ −1 g ðθÞ;
¼ argminθ∈Θ g n ðθÞΩ ð3:4Þ
n ^
β
n n
n
k+1
Assumption 5. The parameter space Θ is a compact subset of ℝ
and θ0 ∈ Int(Θ). where Ω ^ is a consistent estimate of Ω n . An estimate of Ω n
n


′ 
   ^ ¼
requires estimates of tr Σn P n P n Σn þ Σn P n and Qn′ΣnQn. Let Σ
3. Robust GMM estimation of spatial autoregressive models n


^2
^2
^
Diag εn1 …; εnn , where εni s are residuals based on a consistent es-
In this section, the robust GMM estimator for the case of SARAR(1,0)
timator of θ0. Among others, a consistent estimator can be a 2SLS
is reviewed.12 For the case of SARAR(1,0), the reduced form of the
estimator based on the instrument matrix (W 2n X n , W n X n , X n ).
model is Yn = S−1 −1
n Xnβ0 + Sn εn. Then, the spatial lag of the dependent
Under our stated assumptions, it can be shown that the components of
variable can be expressed as W n Yn = W n S − 1 −1
n Xnβ0 + WnSn εn =
Ω can be consistently estimated by simply replacing Σ with Σ ^ .
G n X n β 0 + G n ε n . This result indicates that the endogenous variable n n n

WnYn in a SARAR(1,0) model is a function of a non-stochastic term The fact that quadratic moment matrices are uniformly bounded in
GnXnβ0 and a stochastic term Gnεn. This decomposition motivates the absolute value in row and column sums and the fact that distur-
1
formation of the moment functions for the GMM estimation. The non- bances terms have uniformly bounded fourth moments ensure that
n
stochastic part is instrumented by Qn = (GnXnβ0,Xn), which forms the




tr Σ ^ þΣ
^ P  P ′ Σ ^ P  −1 tr Σ P  P ′ Σ þ Σ P  ¼ o ð1Þ. The as-
population moment function Qn′εn (Lee, 2003, 2007a). The stochastic n n n n n n n n n n
n n n p

part is instrumented by Pjnεn, for j = 1,…,m, where Pjn ∈ P 2n. In this 1 ′^ 1 ′

ymptotic argument for the remaining component Q n Σ n Q n − Q n Σn
case, the moment function is in the form of εn′Pjnεn (Lee, 2007b; Lee n n
and Liu, 2010; Lin and Lee, 2010; Liu et al., 2010). Q n ¼ op ð1Þ is in line with the argument given by White (1980).14
Under Assumption 1, the orthogonality condition of moment Under Assumptions 1–5, Lin and Lee (2010) show that the robust
functions hold as shown below: GMM estimator defined in Eq. (3.4) is consistent and asymptotically
normally distributed, namely15


′ ′
E Q n εn ¼ Q n Eðεn Þ ¼ 0ðkþ1Þ1 ;





    −1 
′ ′ ′
E εn P jn εn ¼ E tr εn P jn εn ¼ E tr P jn εn εn pffiffiffi 1 ′ −1
n ^θn −θ0 Þ → N 0; lim Γ n Ωn Γ n
d



 : ð3:5Þ
′ n→∞ n
¼ tr P jn E ε n εn ¼ tr P jn Σn ¼ 0;

where Σn = Diag(σ2n1,…,σ2nn) is the n × n covariance matrix of distur- pffiffiffi
^ 

bances. The moment function based on Qn is linear and the one based
11
There are some other formulations for the parameter space of autoregressive param-
eter, for details see Kelejian and Prucha (2007, 2010), LeSage and Pace (2009) and Elhorst
et al. (2012).
12
Lin and Lee (2010) suggest a one-step robust GMM estimator for the case of
SARAR(1,0). Doğan and Taşpınar (2013a) extend the one-step robust GMM approach of
Lin and Lee (2010) for the case of SARAR(1,1). For a two-step robust GMM estimator for
the case of SARAR(1,1), see Kelejian and Prucha (2010). Badinger and Egger (2011) extend
the two-step robust estimation of Kelejian and Prucha (2010) to higher order spatial
autoregressive models, i.e., SARAR(p,q).
An estimate of the asymptotic covariance matrix of n θn −θ0 is
needed to make asymptotically valid inferences and to construct
asymptotically correct confidence regions. Lin and Lee (2010) show
13
Lemma 2 in Appendix A can be used to derive Ωn matrices in this section.
14
The same argument also holds for Γn. That is a consistent estimate of Γn can be obtained
by simply replacing Σn with Σ ^ .
n
15
The identification of the parameters in a GMM framework requires limn→∞ 1n Eðg n ðθ0 ÞÞ ¼
0 (Newey and McFadden, 1994). Lin and Lee (2010) investigate this condition and state the
identification conditions for the parameters. Here, we simply assume that the parameters are
identified.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 5

   
1 ^′ ^ −1 ^ −1 1 1
r ~ G(m,k), (iv) λ∼U − ; , where U(∙) denotes uniform distribu-
that a consistent estimator is Γ n Ωn Γ n , which is constructed from τ τ
n
a consistent estimator of θ0. tion function, and τ is the spectral radius of Wn.18
Note that Pn∗ and Q n also involve unknown parameters. In practice, an All prior distribution functions are assumed to be independent. The
initial consistent estimator can be used to construct these matrices. For r
prior distribution of is considered by Geweke (1993). A non-spatial
example, an initial estimator can be a GMM estimator based on the vi
moment functions involving P+ ′Wn − Diag(Wn′Wn) and Q + linear regression model with covariance matrix σ20Vn, where vis are
n = Wn n =
(W2nXn, WnXn, Xn).16 independently and identically distributed, corresponds to a linear
regression model with disturbances having a scale mixture of normal
4. Bayesian MCMC estimation of spatial autoregressive models distribution (Geweke, 1993; Chib, 2001). In particular, Geweke (1993)
shows that the posterior distribution function of the parameter of
The model in Eq. (2.1) under Assumption 1 has an unknown form of a non-spatial linear regression model with Y ni ∼ N(X ni′β 0 , σ 20 v i ),
r
heteroskedasticity. For Bayesian estimation, we assume that where ∼ i.i.d. χ2(r) for i = 1,…,n is proportional to the posterior
vi
Σn = σ02Vn, where Vn = Diag(v1,…,vn) is a diagonal n × n matrix with
distribution function of a non-spatial linear regression model with
σ2 Yni ∼ t(Xni′β0, σ02, r), where t(∙) denotes the univariate Student-t distri-
the diagonal elements vi ¼ ni2 for i = 1,…,n. This assumption
σ0 bution with mean Xni′β0, variance parameter σ20, and degrees-of-
indicates that the heteroskedasticity specification has two components: freedom parameter r.19
(i) a constant component σ02, and (ii) a component vi that varies over The prior density functions are given in the following equations.
observations. The constant component is an arbitrary term that facili-
n o
tates the Bayesian analysis.17 pðβÞ ¼ ð2πÞ
−k=2
jT j
−1=2 ′ −1
exp ðβ−cÞ T ðβ−cÞ ; ð4:4Þ
In this section, we review the Bayesian MCMC approach for the
spatial autoregressive model. Let v = (v1,…,vn)′ be an n × 1 vector.
 b
2 −ðaþ1Þ
a
2 b
Then, the likelihood function of this model can be written as p σ ¼ σ exp − 2 ; ð4:5Þ
Γ ðaÞ σ


− n
2 2 −n
n
−1=2
r nr=2 h
r i−n n
L Y n jθ; σ ; ν ¼ ð2πÞ σ ∏ vi jSn ðλÞj r
2 2
−ðrþ2Þ=2
pðν Þ ¼ Γ ∏ vi exp − ; ð4:6Þ
( i¼1
) ð4:1Þ 2 2 i¼1 2vi
1
−1
′ 2
exp − ðSn ðλÞY n −X n βÞ σ V n ðSn ðλÞY n −X n βÞ : n ro
2 1 1 m−1
pðr Þ ¼ r exp − ; for r N 0; and m; k N 0; ð4:7Þ
km Γ ðmÞ k
It will be convenient to write the exponent of the above likelihood 8  
<τ 1 1
function in terms of generalized least squares (GLS) quantities. To ; if λ∈ − ;
pðλÞ ¼ 2 τ τ ð4:8Þ
this end, for a given value of λ, the GLS estimator from the regression :

−1 0 otherwise:
Sn(λ)Yn = Xnβ + εn is β e ¼ X ′ V −1 X n X ′n V −1
n n n n Sn ðλÞY n . Let SSE ¼

′ −1

 Note that p(v) is the joint prior distribution function of vi for i = 1,…,
e
Sn ðλÞY n −X n β σ 2V n e be the weighted residual
Sn ðλÞY n −X n β
n n n.20 In the formulation of prior density functions, Γ(∙) denotes Gamma
sum of squares. Then, the exponent of the likelihood function can be function defined by Γ(a) = ∫ ∞ 0 μ
a − 1 −μ
e dμ. The uniform prior distribu-
written as tion for the autoregressive parameter in Eq. (4.8) indicates that all values


1 1
in the interval − ; are equally probable, where τ is the spectral radius
1
−1 τ τ
′ 2
exp − ðSn ðλÞY n −X n βÞ σ V n ðSn ðλÞY n −X n βÞ of Wn. Parent and LeSage (2008) introduce an alternative prior distribu-
2
( !)

 
 
 ð4:2Þ tion function for λ based on Beta function. In this formulation
1 e ′ X ′ σ 2 V −1 X e
¼ exp − SSE þ β−β n n n n β−β n :
2 ⋆ 1 ð1 þ λÞ ð1−λÞ
d−1 d−1
p ðλÞ ¼
Betaðd; dÞ 22d−1
d−1 d−1
Thus, the likelihood function in Eq. (4.1) can be written in terms of Γ ðdÞΓ ðdÞ ð1 þ λÞ ð1−λÞ
¼ ; ð4:9Þ
GLS quantities in the following way: Γ ð2dÞ 2 2d−1

−n

−n n where Beta(d,d) = ∫ t0 μ d − 1 (1 − μ) d − 1 dμ is the standard
2 2 −1=2
LðY n jθ; σ ; νÞ ¼ ð2π Þ 2 σ 2 ∏ vi jSn ðλÞj Beta function. Parent and LeSage (2008) show that when d is close to
( i¼1
!)
1

 
 unity and λ ∈ (− 1,1), p⋆(λ) produces uninformative prior over the
exp − SSE þ β−β e ′ X ′ σ 2 V −1 X e
β−β :
2 n n n n n interval (−1,1) such that the end-points get zero prior weight, which
is consistent with the parameter space of λ.21
ð4:3Þ
18
Note that the parameter space of autoregressive parameter is ð−1τ; 1τÞ. For a proof, see
Following Geweke (1993) and LeSage and Pace (2009), we Kelejian and Prucha (2007).
19
assume the following prior distributions for the parameters of Thus, t(Xni′β0, σ20, r) can be treated as a mixture of normal and chi-square distributions.
For the multivariate version, see Liu and Rubin (1995).
SARAR(1,0): (i) β ~ N(c,T), (ii) σ 2 ~ IG(a,b), where IG(a,b) is the 20
The density of vi a can be obtained from that of r/vi. To this end, for any r, let X ¼ vri
inverse gamma density function with the shape parameter a and with density function PX(∙). Define Y = r/X = v i, then the derived density function
r of Y is given by pY ðyÞ ¼ pX ðr=yÞ  dX . The straightforward calculation shows that pðvi Þ ¼
the scale parameter b, (iii) ∼ i.i.d. χ2(r), for i = 1,…,n, where the n o dY
vi ðrþ2Þ
−
½2r r=2 ½Γ ðr=2Þ−1 vi
2
exp −2vr . Let G(∙) denote the gamma distribution function. Note
degrees of freedom parameter r has gamma distribution, namely
i

that 1/v i ~ χ (r)/r. Since χ 2 (r)/r = G(r/2,2/r), we have 1/v i ~ G(r/2,2/r) which
2

implies that v i ~ IG(r/2,r/2).

16
Note that P+
n along with Wn is considered by Kelejian and Prucha (2010) for a two-step
21
Note that, the parameter space of autoregressive parameters is (–1,1), when the
GMM estimation method for a SARAR(1,1) model. weight matrices are row normalized. If weight matrices are not row normalized,
17
For the case of non-spatial models, see Koop (2003) and Griffiths (2007). then (In–λ0Wn) can be singular for certain values from the interval (–1,1).
6 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Let p(β,σ 2 ,v,r,λ|Y n ) be the joint posterior density function of Now, we return to joint posterior density in Eq. (4.10) to determine
parameters. Bayes' Theorem indicates that the joint posterior densi- the conditional posterior distribution of σ2. By collecting all terms in
ty is proportional to the product of the likelihood function and Eq. (4.10) that are not multiplicatively separable from the components
prior densities. Under the assumption that prior distributions of that include σ2, the conditional posterior density of σ2 can be obtained
parameters are independent, the joint posterior density can be as follows:
written as ( )


− 1



e Þ′ X ′ σ 2 V −1 X Þ β−β
n



2
p σ β; ν; r; λ; Y n Þ∝ σ
2
 exp − SSE þ β−β
2
e
n n n n n
2 2 2
p β; σ ; ν; r; λjY n ∝LðY n jθ; σ ; v1 ; …; vn ÞpðβÞp σ pðνÞpðr ÞpðλÞ 2

− n
−ðaþ1Þ b
2
 σ exp − 2
n
−n
2 −1=2
∝ð2πÞ σ ∏ vi jSn ðλÞj
2 2

σ
 i¼1

 
 ( )
1 e ′ X ′ σ 2 V −1 X
− −ðaþ1Þ
exp − SSE þ β−β β−βe 2
n
1 ðS ðλÞY n −X n β Þ′ V −1
n ðSn ðλÞY n −X n β Þ þ 2b
n n n n n ∝ σ exp − 2 n :
2
2
n o
−ðaþ1Þ b σ 2
−1=2 ′ −1 2
jTj exp ðβ−cÞ T ðβ−cÞ  σ exp − 2

r nr=2 h
r i−n n
σ
n
ð4:15Þ
−ðrþ2Þ=2 r m−1 ro
 Γ ∏ vi exp − r exp −  pðλÞ:
2 2 i¼1 2v i k
The comparison of the above kernel with that of an inverse gamma
ð4:10Þ
density function indicates that the conditional posterior distribution of σ2
is an inverse gamma distribution with the shape parameter a⋆ ¼ n2 þ a
⋆ ðSn ðλÞY n −X n β Þ′ V −1
n ðSn ðλÞY n −X n β Þ þ 2b
The MCMC approach requires a full set of conditional posterior and with the scale parameter b ¼ .
2
density functions of parameters to form an algorithm through which Thus,
random draws (or simulated sample) can be obtained. The simulated
sample obtained from the conditional posterior densities is a Markov

2 ⋆ ⋆
chain with the property that its limiting distribution is the joint posteri- σ jβ; ν; r; λ; Y n ∼IG a ; b : ð4:16Þ
or distribution (Albert and Chib, 1993; Casella and George, 1992; Chib,
2001; Tierney, 1994).
Conditional on the other parameters of the model, the conditional In the same way, the conditional posterior distribution of v can be
posterior density of β can be obtained by collecting all terms in obtained by ignoring all terms in the joint posterior density function
Eq. (4.10) that are not multiplicatively separable from components that are not related to v. Thus,
that include β. Let p(β|σ2,v,r,λ,Yn) be the conditional posterior density
function of β. Then,
 n 1
−1
2 −1=2 ′ 2
p νjβ; σ ; r; λ; Y n ∝∏ vi exp − ðSn ðλÞY n −X n βÞ σ V n ðSn ðλÞY n −X n βÞ
i¼1 2
1


 
 ð4:17Þ
2
pðβ jσ ; ν; r; λ; Y n Þ∝exp − e ′ X ′ σ 2 V −1 X
SSE þ β−β e
β−β n r
n n n n n −ðrþ2Þ=2
2 ∏ vi exp − :
 i¼1 2vi
′ −1
þðβ−cÞ T ðβ−cÞ :

ð4:11Þ
For notational simplification, let en = Sn(λ)Yn–Xnβ. Then ðSn ðλÞY n −
−1 −1 e2nj
X n βÞ′ σ 2 V n
n
ðSn ðλÞY n −X n βÞ ¼ e′n σ 2 V n en ¼ ∑ j¼1 , where enj
By completion of square (see Lemma 1 in Appendix), the last two σ 2v j
terms of the exponent in Eq. (4.11) can be written as is the jth element of en. Then, Eq. (4.17) can be written as

8 9


 
 <X
e ′ X ′ σ 2 V −1 X
 n enj =
n 2
β−β β−β e þ ðβ−cÞ′ T −1 ðβ−cÞ 2 −1=2
n
−ðrþ2Þ=2 r
n n n n n p νjβ; σ ; r; λ; Y n ∝∏ vi exp − 2  ∏ vi exp −


−1 
 i¼1 : j¼1
2σ v j
; i¼1 2vi
⋆ ′ ′ 2 −1 ⋆
¼ β−β Xn σ V n Xn þ T β−β ð4:12Þ ( )
n e2 r

 
−1 −1 −1
 ∝∏ vi
−ðrþ3Þ=2
exp − ni2 − ð4:18Þ
þ β e −c −1 X
′
σ
2
V X þ T βe −c ; i¼1 2σ vi 2vi
n n n n n
8
9
n <− σ −2 e2ni þ r =
−ðrþ3Þ=2
∝∏ vi exp :
i¼1 : 2vi ;
where β⋆ = (Xn′V−1 n Xn + σ T
2 −1 −1
) (Xn′V−1
n Sn(λ)Yn + σ T
2 −1
c). Using
Eq. (4.12), the conditional posterior density of β can be written as

1

−1 
 The above result shows that the conditional posterior density of each
2 ⋆ ′ ′ 2 −1 ⋆
pðβjσ ; ν; r; λ; Y n Þ∝exp − β−β Xn σ V n Xn þ T β−β ; vi can be written as
2
ð4:13Þ
8
9
<− σ −2 e2ni þ r =
2 −ðrþ3Þ=2
pðvi jβ; σ ; r; λ; Y n Þ∝vi exp ; for i ¼ 1; …; n:
where the terms that are multiplicatively separable from β are : 2vi ;
absorbed into the constant of proportionality. The kernel of p(β|σ2, ν, r, ð4:19Þ
λ, Yn) in Eq. (4.14) is proportional to the kernel of a multivariate
normal density with mean β⋆ and covariance (Xn′(σ2Vn )− 1Xn +
T− 1)− 1. Thus,
The functional form in the right hand side of Eq. (4.19) does not
correspond to any known form of density function. Geweke (1993,
 
−1 −1 
2 ⋆ ′ 2 −1 p.S26) suggests a method through which the conditional posterior
βjσ ; ν; r; λ; Y n ∼N β ; X n σ V n Xn þ T : ð4:14Þ σ −2 e2ni þ r
distribution of v i can be determined. Let Ψ ¼ . Then, the
vi
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 7

result in Eq. (4.20) implies that the density function of Ψ is propor- density functions in Eq. (4.10) can not be obtained. In the MCMC
tional to Ψ(r − 1)/2exp{− Ψ/2}, which implies22 approach, a Markov chain is constructed from the conditional pos-
terior distributions so that a large number of a simulated sample of
parameters can be obtained. The mean of the simulated sample is
σ −2 e2ni þ r 2 2 then used as an estimate of E(ζ|Y n ) (Casella and George, 1992;
β; σ ; r; λ; Y n ∼χ ðr þ 1Þ; for i ¼ 1; :::; n: ð4:20Þ
2vi Chib, 2001).
The previous section shows that the joint posterior densities of
Now, we return to conditional posterior distribution of the hyper-
spatial autoregressive models can be decomposed into conditional
parameter r. By collecting all terms that are not multiplicatively
posterior density functions of parameters through which random
separable from the components that include r from joint posterior
draws can be obtained. When conditional posterior distributions are
density yields,
known, the Gibbs sampling method can be used to draw from condi-

r nr=2 h
r i−n n r ðk þ 2vi Þ tional distributions. As the conditional posterior distributions of β, σ2
2 m−1 −r=2
pðr jβ; σ ; ν; λ; Y n Þ∝ r Γ ∏ vi exp − : and vi are in the form of known distributions, random draws can be
2 2 i¼1 2kvi
easily obtained through the Gibbs sampling method for these parame-
ð4:21Þ
ters. Unfortunately, the conditional posterior distributions of the
autoregressive parameter and hyper-parameter r do not correspond to
The above density does not correspond to a standard distribution,
known distributions. The sampling for these parameters can be accom-
therefore a Metropolis Hasting type algorithm can be employed to
plished through a Metropolis–Hasting algorithm. As a result, the
obtain random draws.
sampling scheme for spatial autoregressive models consists of the
Finally, the conditional posterior distribution of λ takes the form:
combination of the Gibbs sampling and the Metropolis–Hasting
 1
−1 algorithm. This whole procedure of sampling is called Metropolis within
2 ′ 2
pðλjβ; σ ; ν; r; Y n ∝jSn ðλÞjexp − ðSn ðλÞY n −X n βÞ σ V n ðSn ðλÞY n −X n βÞ : Gibbs sampling (Geweke and Keane, 2001).
2
The Metropolis–Hasting sampling method requires a proposal distri-
ð4:22Þ
bution (or candidate generating distribution) from which the candidate
values for the autoregressive parameters can be generated. LeSage and
The uniform prior density for λ, which is absorbed in the proportion-
Pace (2009) and Kakamu and Wago (2008) use a normal distribution
ality constant is used for the result in Eq. (4.22). For the case where prior
as the proposal distribution with a tuned random-walk procedure.24
takes the form suggested by Parent and LeSage (2007), the conditional
According to this method, the candidate values for λnew are generated
posterior distribution of λ takes the following form:
by

1
−1
2 ′ 2
pðλjβ; σ ; ν; r; Y n Þ∝jSn ðλÞjexp − ðSn ðλÞY n −X n β Þ σ V n ðSn ðλÞY n −X n β Þ new old
2 λ ¼λ þ c  ϕ; where ϕ∼Nð0; 1Þ: ð4:25Þ
d−1 d−1
1 ð1 þ λÞ ð1−λÞ
 :
Betaðd; dÞ 22d−1
ð4:23Þ The constant parameter c in Eq. (4.25) is called tuning parameter,
which ensures that the sampler moves over the entire conditional pos-
terior distributions of the autoregressive parameters. The increment
Both conditional posterior densities in Eqs. (4.22) and (4.23) random variable denoted by ϕ in Eq. (4.25) is symmetric so that the
are not in the form of a known density. Like hyper-parameter r, acceptance probability value for the candidate value λnew takes the
sampling for λ can be accomplished through the Metropolis–Hasting following form:
approach.23
( )
4.1. Bayesian MCMC computation
 p λnew β; σ 2 ; ν; r; Y n Þ
new old
p λ ;λ ¼ min 1; ;
p λold β; σ 2 ; ν; r; Y n Þ
In Bayesian analysis, the joint posterior distribution of parame-
ters embodies all information about the parameters. A cost (loss)
function is specified as a criterion for determining an optimal point The new candidate λnew is accepted with a probability of p(λnew,λold).
estimate. Let ζ = (λ,β′,σ 2 ,ν,r)′ be the parameter vector. One of The tuning parameter or the spread of the candidate generating density
the most popular

 loss

function
′
is
 the weighted squared error affects the behavior of the chain in at least two ways: (i) it affects the ac-
function, C ζ^ ; ζ ¼ ζ^ −ζ Q ζ^ −ζ , where Q is a positive definite ceptance rate of new candidate values through acceptance probability,
matrix. Then, the Bayesian point estimate is defined by (ii) it also affects the region where the new candidate values are sam-
pled. To illustrate the role of the tuning parameter, consider a chain
h
i that has converged so that the new candidates are sampled around
ζ^ ¼ argminζ^ Eζ jY n C ζ^ ; ζ
Z
′
 the mode of the marginal posterior densities of autoregressive parame-
¼ argminζ^ ζ^ −ζ Q ζ^ −ζ pðζ jY n Þdζ ; ð4:24Þ ters. If the tuning parameter is chosen to be large to allow the new gen-
Θ erated candidates to be far from the current value (or mode), then the
new candidates will have low probability of being accepted. In that
where p(ζ|Y n ) is the posterior density of ζ. The solution of above case, the chain may get stuck at the current value (the candidates will
minimization problem yields ζ^ ¼ Eðζ jY n Þ ¼ ∫Θ ζpðζ jY n Þdζ , which is never be accepted). Reducing the tuning parameter will correct this
the mean of the posterior distribution of ζ (Koop et al., 2007, problem. On the other hand, a small tuning parameter value will reduce
p.34). The closed form solution for ζ^ from the joint posterior the chance of getting candidates that are far away from the current
value. In this case, it will take longer time for the chain to traverse the
¼ −ð Ψ Þ, support of the density, as a result, low probability regions of the density
22 σ −2 e2ni þr
This result can be seen through the derived density function of Ψ. Since dvi

n o dΨ 2

then pðΨÞ∝ð
σ e þr Þ
−2 2
ni
−ðrþ3Þ=2

exp − 1
σ −2 2
e þ r Ψ
 ðσ e þrÞ
−2 2
ni
∝Ψ ðr−1Þ=2
expf −Ψ=2g. The support will be under-sampled. These two issues raise the question of
ðΨÞ−ðrþ3Þ=2 2 ni ðσ e þrÞ
−2 2
ni
2
Ψ

kernel of p(Ψ) implies Ψ ~ χ2(r + 1).

23 24
The Metropolis-Hasting algorithm is described in Section 4.1. Chib (2001) calls this method as random walk M–H chain.
8 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Disturbance terms
10

−5

−10

−15
0 20 40 60 80 100

Group variances
15

10

0
0 20 40 60 80 100

a) Group variance sand disturbance terms of true data generating process

Estimates of variance components: ν Estimates of variance components: ν
15 i 15 i
r=5 r=5
r=6 r=10
r=7 r=50

10 10

5 5

0 0
0 20 40 60 80 100 0 20 40 60 80 100

b) Posterior mean of variance components ( ) c) Posterior mean of variance components ( )

Fig. 1. The effect of hyper-parameter r on the estimates of variance components vis. (a) Group variances and disturbance terms of true data generating process. (b) Posterior mean of
variance components (vi). (c) Posterior mean of variance components (vi).

how to choose the optimal tuning parameter. For our case, the literature
has suggested a tuning parameter that can result in approximately 0.5
acceptance rate (Chib, 2001; Chib and Greenberg, 1995). In practice,
the tuning parameter c is adjusted in such way that the acceptance
rate falls between 40% and 60%.25
LeSage and Pace (2009) consider a more efficient sampling proce-
dure known as the Griddy Gibbs sampler for the autoregressive param-
eter. This procedure involves a numerical integration technique to
calculate the normalizing constant and the related cumulative distribu-
tion function (CDF) of the autoregressive parameter. Then, the random
draws are taken from the CDF by an inversion approach in each pass
through the sampler. In the Monte Carlo study, we compare the perfor-
mance of this approach with the M–H random walk procedure.
Finally, the elicitation of the parameters of the prior distributions
for β and σ2 is required. The prior distributions for β can be made
relatively diffuse by assigning a large prior variance. This can be
accomplished by setting c = 0 and T = Ik × 10 5 . The conditional
posterior distribution of σ2 is an inverse Gamma distribution. The
analytical results provided for the mean of this distribution indicate
that a relatively diffuse prior for σ 2 can be obtained by assigning
values of a = 0 and b = 0.26

25
In practice, c can be set 0.5 at the initial step. LeSage and Pace (2009) suggest that 26
For the case of SARAR(1,0), we show that σ2|β, ν, r, λ, Yn ∼ IG(a⋆ b⋆), which is given in
⋆
c′ = c/1.1 for the case where the acceptance rate falls below 40%, and c′ = 1.1 for the Eq. (4.17). Then, the conditional posterior mean E σ 2 jβ; ν; r; λ; Y n ¼ a b−1. The choice of
⋆

case where the acceptance rate rises above 60%. This can be accomplished through a a = 0 and b = 0 eliminates the effect of the prior distribution on the conditional posterior
while loop in the algorithm. mean.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 9

For illustration, a single pass through the Metropolis within Gibbs plots in Fig. 1 picture the effects of various r values on the estimates
sampling scheme consists of the following steps: of variance components v i . Fig. 1(b) indicates that the estimates
are similar for r = 5,6,7. On the other hand, Fig. 1(c) shows that
1. Let ζ(0) = (λ(0),β′(0),σ2(0),ν(0),r(0))′ be the initial parameter values.
the r-values bigger than 10 give estimates close to unity, suggesting
2. Update β: Draw β(1) from p(β|λ(0),σ2(0),v(0),r(0),Yn).
a homoskedastic assumption. This example indicates that a hyper-
3. Update σ2: Draw σ2(1) from p(σ2|λ(0),β(1),r(0),Yn).
  parameter value near 5 is an optimal choice for the estimation of
σ −2 e2ni þ r ð0Þ ð1Þ
4. Update v: Draw v(1) i from p jλ ; β ; σ 2ð1Þ ; r ð0Þ ; Y n for heteroskedastic models.
vi
; σ 2ð1Þ ; r ð0Þ ; Y n Þ for i ¼ 1; …; n.
new (0)
5. Update λ: Calculate λ = λ +(c ×
ϕ, ) 5. Monte Carlo experiments

 p λnew jβ ð1Þ ; σ 2ð1Þ ; νð1Þ ; r ð0Þ ; Y n
(a) Calculate p λnew ; λð0Þ ¼ min 1;
0 ð1Þ  ,
p λ jβ ; σ 2ð1Þ ; νð1Þ ; r ð0Þ ; Y n In the previous sections, we provided analytical results for the
(b) Draw U from Uniform(0,1),
(1) new estimators. The conditional posterior distribution of the parameters
(c) Set λ = λ if p(λ ,λ ) N U, otherwise λ(1) = λ(0).
new (0)
indicates that the Bayesian estimators resemble to the generalized
least square (GLS) estimators such that the effect of outliers is down-
The sequence {ζ (0) ,ζ (1) ,…,ζ (M) } obtained from the Metropolis weighted in the estimation. In the robust GMM approach, the sample
within Gibbs sampling form a Markov chain whose probability moment functions are weighted with the inverse of their covariances
density function converges to the joint posterior density p(ζ|Yn) as so that they are made as close as possible to zero. Despite these
M goes to infinity (Chib, 2001; Tierney, 1994).27 To discard the ef- analytical results, it is difficult to make any general statement about
fect of the initial value ζ(0) on the chain, iterates from some initial the performance of estimators in finite samples. Hence, to evaluate
periods are usually excluded from the chain (burn-in repetitions). the finite sample properties of estimators, we resort to a Monte Carlo
Let h be a function defined on the parameter space of the model. simulation.
Under a suitable law of large numbers for Markov chains, it can be
shown that M− 1 ∑M (j)
j = 1 h(ζ ) → ∫ Θ h(ζ)p(ζ|Y n )dζ as M → ∞. Note 5.1. Design
that the convergence is established in terms of the simulated sam-
ple size M and not in terms of the data sample size n (Chib, 2001; In this section, the small sample properties of the ML estimator,
Geweke and Keane, 2001; Hoogerheide et al., 2009). Thus, posterior the robust GMM estimator and the Bayesian estimator are compared
moments such as mean and second moment can be easily estimated through Monte Carlo simulation experiments. The model is specified
by taking h(ζ) = ζ and h(ζ) = ζζ′, respectively. as
The hyper-parameter r is not updated in the above illustrated
Metropolis within Gibbs sampling scheme.28 The sampling from the
conditional posterior density of the hyper-parameter r can be carried Y n ¼ λ0 W n Y n þ X n β0 þ εn : ð5:1Þ
out with an argument similar to the one described above for λ. The
hyper-parameter r plays a central role in generating heteroskedasticity There are two regressors and no intercept term such that Xn = [xn,1,
robust estimates. In the preceding section, the prior of vi is assumed to xn,2] and β0 = (β10,β20)′ = (1,1)′, where xn,1 and xn,2 are n × 1 vectors
be IG(r/2,r/2). Then, the prior mean and variance of vi are respectively of variables. We use the date set described in Kelley Pace and Barry
r 2r 2 (1997) on the votes cast in the 1980 presidential election across 3107
given as and . The prior variance approaches to zero
r−2 r 3 −8r 2 þ 20r−16
U.S. counties. We use the normalized version of the income per-capita
and the prior mean approaches to 1 as r goes to infinity. This observation
and the homeownership from this data set for xn,1 and xn,2, respectively.
implies that vi converges in probability to its mean 1 as r → ∞. Thus,
These variables are normalized by subtracting the corresponding
higher values of r are associated with a homoskedastic assumption
sample averages, and then dividing by the sample standard deviations.
and lower values imply a heteroskedastic assumption.29 In our Monte
The first n observations of these variables are used in the Monte Carlo
Carlo simulation we set r = 5 as the data generating process is assumed
experiment. The spatial autoregressive parameter λ0 takes values from
to be heteroskedastic.
the set K = (−0.9,−0.6,−0.3,0,0.3,0.6,0.9) in order to allow for weak
The hyper-parameter affects estimates of variance components
and strong spatial interactions.
v i s through the conditional posterior distribution given in Eq. (4.20).
The row normalized spatial weight matrices are based on the inter-
Fig. 1 illustrates the effect of this parameter on the posterior mean
action scenario described in Arraiz et al. (2010) and Drukker et al.
of v for a sample size of n = 100 for the case of a heteroskedastic
(2013). In this interaction scenario, units are distributed across four
SARAR(1,0) specification. For this illustration, the spatial weight ma-
quadrants of a space in a such way that the proportions of number of
trix Wn is based on a small group interaction scenario, where group
units in the northeast quadrant to the total number of units are allowed
sizes are random draws from the interval (3,20).30 Fig. 1(a) shows
to vary by design. It is interesting to see the results for high density and
true group variances and the resulting disturbance terms. The other
low density cases. In a high density case, the units in the northeast por-
tion are closer to each other. As stated in Arraiz et al. (2010), the location
specification of units in this interaction scenario reflects the distribution
27
Theorem 1 of Tierney (1994) and Theorem 2 of Chib (2001) give conditions under of the U.S. states over space, where the northeastern portion contains
which the convergence holds. Theorems 4–5 in Geweke (1993, p.S28–S29) show that the-
relatively more states. The location of each unit across the space is
se conditions hold for a non-spatial linear regression model. Therefore, it can be shown
that these conditions also hold for linear spatial models. determined by the xy-coordinates over a square grid. Let m and m be
28
Following LeSage and Pace (2009), we assume that this parameter is known. In the two integers. Then, the units in the northeast quadrant of the space
Monte Carlo design, we set r = 5. have discrete coordinates satisfying m þ 1 ≤x ≤m and m þ 1≤y≤m,
29
Geweke (1993) shows that in the case where uninformative priors are assumed, the with an increment value of 0.5. For the other quadrants, the location
second moment of joint posterior density exists only if r N 4. For our case, the uninforma-
tive priors (uniform distributions) are assumed for the autoregressive parameters, there-
coordinates are integers satisfying 1 ≤ x ≤ m, 1≤y≤m, and 1≤x ≤m,
fore the lowest value that r can take is bigger than 4. LeSage and Pace (2009) suggest a 1 ≤ y ≤ m.
value of r between 2 and 7 as an optimal choice for the estimation of heteroskedastic The distance dij between any two units i and j, located at (x1,y1)
models. and (x2,y2) is measured by the Euclidean distance given by
30
The block diagonal weight matrix is row normalized. Members of a group have the
same variance. For each group, if the group size is greater than 10, we set the variance h i
equal to the group size; otherwise we set the variance to the square of the inverse of the 2 2 1=2
dij ¼ ðx1 −x2 Þ þ ðy1 −y2 Þ :
group size. This setup is similar to the one in Lin and Lee (2010).
10 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Then, the (i,j) th element of the weight matrix is defined by is generally inconsistent. The log-likelihood function under the
assumption that the disturbance terms are i.i.d. normal(0, σ 20 ) is
given by
1 if 0 ≤dij ≤1;
wij ¼
0 otherwise:

 n n
2 1 ′
2
ln Ln θ; σ ¼ − lnð2πÞ− ln σ þ lnjSn ðλÞj− 2 εn ðθÞεn ðθÞ; ð5:4Þ
2 2 2σ
We generate heteroskedasticity by using the number of neigh-
bors a unit has. Let hi be the number of neighbors for unit i. Then,
the ith element εin of the innovation vector is generated in the where εn(θ) = Sn(λ)Yn–Xnβ. The solution of the first order conditions


following way for β and σ2 yields the ML estimators β ^ ðλÞ ¼ X ′ X −1 X ′ S ðλÞY
n n n n n n




and σ
2 1 ′ ^ ^ ^
^ n ðλÞ ¼ ε n λ; βn ðλÞ εn λ; βn ðλÞ , where εn λ; βn ðλÞ ¼ Sn ðλÞ
n
ε ni ¼ σ ni ξni ; ^ ðλÞ. Concentrating out β and σ2 from the log-likelihood yields
Y n −X n β n
hi ð5:2Þ
σ ni ¼ c Xn
h j =n;
n
2 
j¼1
n
^ n ðλÞ þ lnjSn ðλÞj:
ln Ln ðλÞ ¼ − lnð2π Þ− ln σ ð5:5Þ
2 2
where σni is the standard error for the i-th observation and ξni's are
i.i.d. Normal(0,1). Here, the noise variance in a location is a function
of a ratio of neighbors of that location to the average number of The MLE of λ0 is the extremum estimator obtained from the maximi-
neighbors in the model. The idea is that having relatively larger zation of the concentrated log-likelihood function. The necessary condi-
number of neighbors in a neighborhood implies a larger noise vari- ^ n requires plimn→∞ ∂lnLn ðλÞ ¼ 0. Lin and Lee
tion for the consistency of λ
ance. The constant c adjusts the signal–noise ratio of the model. For ∂λ
example, c = 10 yields an approximate signal–noise ratio of 0.55. (2010) show that this condition is not satisfied in the presence of
∂lnLn ðλÞ
We calculate the signal–noise ratio by the R2 statistic suggested in heteroskedastic disturbances. More specifically, the condition is ¼
∂λ
Pace et al. (2012): cov Gn;ii ; σ 2ni
þ op ð1Þ , where G n,ii s are diagonal elements of G n and

′
 σ2
2
E ε′n S−1
n S−1
n εn
2
σ ¼ 1 n
∑i¼1 σ 2ni .
Hence, the inconsistency of the MLE λ ^ n depends
R ¼ 1−
 n
′
β′0 X ′n S−1 S−1 ′ −1′ −1 on the covariance between the diagonal elements of Gn and the
n X n β 0 þ E ε n Sn Sn ε n

 variance terms σ2nis. For certain spatial weight matrices, the diagonal
−1′ −1
tr Sn Sn Σn elements of G n are the same, implying a zero covariance with the
¼ 1−
: ð5:3Þ
′
β′0 X ′n S−1 S−1 −1′ −1 individual variances. For a circular world weight matrix with equal
n X n β 0 þ tr Sn Sn Σn
diagonal elements that relate each unit to the units in front and in
back, there is no variation in the diagonal element of G n . In this
The Monte Carlo experiments are based on the combinations case, the necessary condition is not violated, even if the disturbance
ðc; m; mÞ ¼ fð10; 5; 15Þ; ð10; 7; 21Þ; ð10; 14; 20Þ; ð10; 20; 28Þg, where the terms are heteroskedastic.
first two combinations yield high density cases whereas the last two The inconsistency of λ ^ n contaminates the MLE of β 0 ,
combinations yield the low density cases in the northeastern portion.31 since it is a function of λ ^ n . The MLE of β0 has an asymptotic bias
These combinations produce the corresponding sample sizes of n =

−1
of λ0 −λ^n X′ Xn X ′ Gn X n β0 . 33 For our Monte Carlo setup, there
{486,974,485,945}. n
In all experiments, the following estimators are considered: (1) the is variation in the diagonal elements of G n , so the MLE of λ 0 and
Gaussian maximum likelihood estimator (MLE), (2) the robust general- β0 is expected to produce inconsistent estimates.
ized method of moment estimator (RGMME) of (3.7),32 (3) the Bayes- Table 3 shows the simulation results for ðc; m; m; nÞ ¼ ð10; 5; 15; 486Þ.
ian MCMC estimator (BE1), where the Griddy Gibbs sampling method The MLE imposes substantial amount of bias on λ0 for the case where
is used for λ, and (4) the Bayesian MCMC estimator (BE2), where the there is negative strong spatial dependence relative to the case of strong
random-walk Metropolis Hasting algorithm is used for λ. For each positive spatial dependence. The same pattern shows itself for the BE
specification, the Monte Carlo experiment is based on 1000 repetitions. estimates. The RGMME imposes relatively smaller bias on λ0 in all
For the BEs, we run the MCMC algorithm 6000 times and discard the cases. In cases where there is a positive spatial dependence, the MLE
first 1000 iterations. and BEs underestimate λ0, while the RGMME overestimates. In terms
of finite sample efficiency measured by RMSE, the BEs of λ0 are more
5.2. Simulation results efficient. Table 3 also reports the estimation results for β10 and β20.
Theoretically, the bias reported for λ0 also contaminates the estimates
The simulation results are presented in Appendix B. In each of β10 and β20 for the case of the MLE (Lin and Lee, 2010). The results
table, the empirical mean (mean), the bias (Bias), the empirical for β10 and β20 are consistent with this theoretical prediction. The
standard error (SD) and the root mean square error (RMSE) of MLE of β10 and β20 has more than 5% bias in many cases. In terms of
parameter estimates are presented near to each other for easy finite sample efficiency and bias, the BEs are performing better than
comparisons. the RGMME.
Before presenting the simulation results, a couple of points Table 4 reports the estimation results for ðc; m; m; nÞ ¼ ð10; 7; 21; 974Þ.
for the MLE should be considered. Theoretically, the MLE of λ 0 The MLE results repeat the same pattern for λ0, β10 and β20. Again, the
BEs impose substantial amount of bias on λ0 when there is a strong
31
For details, see Arraiz et al. (2010). negative spatial dependence. In other cases, both BE1 and BE2 impose
32
For the RGMME, we use two initial estimators: (i) the2SLS estimator formulated less than 5% bias on λ0. When there is positive spatial dependence, the
from the instrument matrix (W 2n X n , W n Xn , X n ), and (ii) the initial GMM estimator
MLE and the BEs underestimate λ0. The RGMME imposes almost no
based on Qn = (W2nXn, WnXn, Xn) and Pn = (Wn′Wn − Diag(Wn′Wn)). Both initial esti-
mators produce similar results, therefore we only report the results based on the ini-
33
tial GMM estimator. For an asymptotic argument, see Doğan and Taşpınar (2013a).
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21
bias on λ0 in all cases and is slightly more efficient. The finite sample
efficiency is higher for all estimators relative to Table 3. For the case of
β10 and β20, the estimators report similar results and the BEs are
relatively more efficient.
Table 5 reports the estimation results for ðc; m; m; nÞ ¼ ð10 ; 14; 20;
485Þ. Again, the MLE results repeat the same pattern for λ0, β10 and
β20. That is, the MLE imposes relatively greater bias on λ0 when there
is a strong negative spatial dependence. The BEs of λ0 impose less
than 5% bias in many cases and are relatively more efficient. For the
case of β10 and β20, the BEs are performing better than the RGMME in
terms of bias and efficiency in most cases.
Finally, Table 6 reports the estimation results for ðc; m; m; nÞ ¼
ð10; 20; 28; 945Þ. Again, the MLE results repeat the same pattern for λ0,
β10 and β20. The RGMME imposes almost no bias on λ0, and it is
as efficient as the BEs. Again, both BE1 and BE2 report almost identical
results and they impose less than 5% bias on λ0 in many cases. For the
case of β10 and β20, although the RGMME imposes relatively smaller
bias, it is generally inefficient relative to the BEs.
Overall, the MLE imposes substantial amount of bias on λ0 when
there is strong negative spatial dependence in the model. The
amount of bias does not decrease as the sample size increases. This
pattern suggests inconsistency for the MLE. In the case of BEs, the
amount of bias is generally greater relative to the RGMME. In terms
of efficiency, the BEs are generally more efficient and the BE 1
performs better than the BE2.
In spatial models, the marginal effect calculations for the exogenous
variables require a different analysis. For the model in Eq. (5.1), the
marginal effects are given by
∂Y n −1
¼ ðI n −λ0 W n Þ In βk0 ; for k ¼ 1; 2: ð5:6Þ
∂x′n;k
The marginal effects take the form of n × n matrix (In–λ0Wn)−1Inβk0.
LeSage and Pace (2009) suggest scalar summary measures for the
marginal effects. They label the average of the diagonal elements
as the direct effects and the average of either the row sums or the
column sums of the non-diagonal elements as the indirect effects.
The total effect is the sum of the direct and indirect effects.
Tables 7 and 8 of Appendix B report the simulation results for
the total effects.34 The general pattern from the estimates reported in
these tables is that all estimators impose relatively larger bias on the
total effect estimates when there is strong positive spatial dependence.
Especially, the amount of bias is substantial when λ0 = 0.9. As the
sample size increases, the amount of bias decreases more in the case of
RGMME. In terms of bias, the RGMME and the BE1 perform better
than the MLE, and the RGMME is slightly better than the BE 1. In
terms of the finite sample RMSE, the BEs and the MLE perform better
than the RGMME and the BE1 is relatively more efficient.
In general, the Monte Carlo results show that the RGMME is
relatively inefficient when the sample size is 486 and when there
exists positive spatial dependence. The efficiency of the RGMME
depends on the selection of the moment functions for the estima-
tion. As stated in Section 3, the best selection of the quadratic
moment matrices is not available for the case where disturbance
terms satisfy Assumption 1. Hence, the best selection of the linear
and quadratic moment functions in the presence of heteroskedastic
disturbances would provide further efficiency improvement for
the RGMME. 35 Another related issue is about the calculation of
the measure of dispersion of the total effects. In practice, measures
of dispersion of the total effects are needed to draw inferences
34
The results based on BE1 are reported. The total effects results of BE2 are similar and
are available upon request.
35
The issue of selection of the moment functions in the heteroskedastic case is a poten-
tial topic to be explored in the future studies.
11
regarding the point estimates. LeSage and Pace (2009) suggest a
simulation method for the calculation of the dispersions of total
effects, where the parameter estimates and asymptotic covariances
of estimators are used. Hence, in this simulation method, the trade-
offs between bias and precisions of estimators may determine the
quality of the inference regarding the marginal effect of explanato-
ry variables.36
6. Empirical illustrations
6.1. Housing price model
In the first illustrative application, the Harrison and Rubinfeld
(1978) data set described in Gilley and Kelley Pace (1996) is used
to estimate a hedonic housing-price model for a sample of 506
observations of census tracts in the Boston Standard Metropolitan
Area in 1970. This data set has been used by many authors to
illustrate various econometric techniques, see for example LeSage
(1999) and Baltagi (2008).
In this application, the median house price is explained by
thirteen variables that are described more fully in Table 9 of
Appendix C: per-capita crime rate by town, proportion of residential
land zoned for lots over 25,000 ft2, proportion of non-retail business
acres per town, Charles River dummy variable, nitric oxides concen-
tration parts per 10 million, average number of rooms per dwelling,
proportion of owner-occupied units built prior to 1940, weighted
distance to five Boston employment centers, index of accessibility
to radial highways, full-value property tax rate, pupil–teacher ratio
by town, proportion of blacks per town, and proportion of lower
class per town.37
We construct a contiguity weight matrix following LeSage (1999,
p.115).38 The estimation results are given in Table 1. First of all, the
null hypothesis of no heteroskedasticity is rejected as indicated by the
Breusch–Pagan LM tests and its Koenker version.39 The presence of
heteroskedasticity can also be tested from the Bayesian perspective by
investigating the posterior mean of variance components. Recall that
the Bayesian estimators are based on the assumption that the covariance
matrix Σn can be decomposed as σ2Vn, where Vn = Diag(v1,…,vn). Thus,
the estimates of vis that deviate substantially from unity indicate
the presence of heteroskedasticity. The posterior mean of variance
components vis are presented in Fig. 2. The estimates of the variance
components indicate that the constant variance assumption may not
hold for this application.
The parameter estimates in Table 1 are mostly in agreement
except for the pollution (denoted by nox) and house age variables.
The Bayesian estimator reports a negative insignificant estimate
for the pollution variable.40 Both MLE and RGMME report negative
significant estimates for the pollution variable, which represents
important policy implications for this variable. In the case of
house age variable, the Bayesian estimate is negative and signifi-
cant, while both MLE and RGMME report insignificant positive
estimates.
36
This issue is pointed out by a referee, and we left it for future studies.
37
The dependent variable is in log form. Following LeSage (1999), we scale each variable
by subtracting the mean and dividing with the standard deviation.
38
We use the ‘xy2cont’ function of Spatial Econometric Toolbox to create the weight ma-
trix, which returns three weight matrices: W1, W2, and W3. Following LeSage (1999,
p.115), we use W2 for this application.
39
The LM test is specified based on the square of the crime, rooms, tax rate, and pu-
pil/teacher variables. We examined the relationship between the residuals and the
explanatory variables: The scatter plots indicate that the most likely candidates are
these variables.
40
LeSage (1999, p.117) also reports an insignificant estimate for this variable. In the case
of Charles River dummy variable, spatial model specifications have insignificant estimates.
LeSage (1999) states that this locational dummy variable is unnecessary, since the spatial
models account explicitly the spatial nature of the data.
12 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Posterior mean of variances

25
SARAR(1,0) Heteroscedastic variance

20

15

10

0
0 50 100 150 200 250 300 350 400 450 500
Observations
Fig. 2. Estimates of vi: Housing Price Model.

Table 1
Hedonic Housing-Price.

Parameter estimates Total effects

Log(hprice) ML RGMM BE ML RGMM BE

Crime −0.1648 −0.1294 −0.1154 −0.2993 −0.2312 −0.1972

[−6.8730] [−3.1908] [−4.6229] [−6.8828] [−3.4600] [−4.4388]
Zoning 0.0808 0.0805 0.0507 0.1471 0.1438 0.0866
[3.0149] [3.3951] [2.4450] [2.9246] [3.1578] [2.4175]
Industry 0.0443 0.0468 0.0360 0.0862 0.0849 0.0614
[1.2565] [1.6834] [1.3428] [1.3406] [1.7181] [1.3385]
Charlesr 0.0167 −0.0384 0.0011 0.0304 −0.0707 0.0019
[0.8969] [−1.4151] [0.0743] [0.8799] [−1.3033] [0.0753]
Nox −0.1288 −0.1292 −0.0542 −0.2361 −0.2289 −0.0929
[−3.4138] [−2.9714] [−1.7469] [−3.5177] [−3.3085] [−1.7354]
Rooms 0.1609 0.1329 0.2974 0.2958 0.2394 0.5079
[6.5518] [3.2520] [11.7856] [5.9919] [2.9171] [9.8774]
House age 0.0186 0.0272 −0.0598 0.0318 0.0507 −0.1024
[0.5971] [0.8053] [−2.2384] [0.5471] [0.8341] [−2.2083]
Distance −0.2149 −0.2041 −0.1695 −0.3938 −0.3613 −0.2897
[−6.0965] [−5.3051] [−5.9347] [−5.7026] [−5.2180] [−5.5538]
Access 0.2719 0.2208 0.1647 0.4983 0.3927 0.2815
[5.6216] [4.0229] [3.8799] [5.4681] [4.0784] [3.7533]
Tax rate −0.2209 −0.2255 −0.2045 −0.4088 −0.4033 −0.3494
[−4.1614] [−4.5818] [−4.7507] [−4.1635] [−4.0740] [−4.5295]
Pupil/teacher −0.1015 −0.1201 −0.0811 −0.1871 −0.2142 −0.1385
[−4.0559] [−4.8498] [−4.2052] [−4.4142] [−5.4084] [−4.0704]
Black pop 0.0774 0.0699 0.1073 0.1437 0.1251 0.1832
[3.7669] [2.6762] [5.8033] [3.7259] [2.7106] [5.5015]
Lower class −0.3361 −0.3393 −0.2188 −0.6179 −0.6050 −0.3736
[−10.1178] [−5.3235] [−6.4348] [−10.9477] [−6.8317] [−6.0629]
λ 0.4549 0.4373 0.4126 – – –
[12.5379] [6.4776] [12.3915]
BP LM 142.5982 – – – – –
[0.0000]
BP Koenker 45.1843 – – – – –
[0.0000]
Jarque-Bera 475.8595 – – – – –
[0.0010]
R ̅
2
0.7709 0.8313 0.7475 – – –
σ 0.1575 – 0.2525 – – –

Note: The brackets contain t-statistics for the parameter estimates and p-values for the test statistics.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21
The estimate of the spatial autoregressive parameter λ is 0.4549
for the MLE, 0.4373 for the RGMME and 0.4126 for the BE. For
similar sample sizes, our Monte Carlo results in Tables 3 and 5
show that both the MLE and BE underestimate this parameter,
while the RGMME imposes almost no bias. The estimation results
for λ from this empirical application are not consistent with the
Monte Carlo results.
The last three columns of Table 1 report the total effect esti-
mates of the exogenous variables. Again the important difference
is observed for the total effect of the pollution and house age
variables. These differences imply important policy implications
toward the effect of these variables on house prices. The t-statistics
reported for the RGMM total effect estimates are relatively smaller in
most cases, which are consistent with the Monte Carlo results of
Tables 7 and 8 of Appendix B, where RMSE of the total effect estimates
from the RGMME are relatively greater.
The question that how to choose between estimators for this
application should be considered along the specification issues
for the housing price model.41 LeSage (1999) provides evidences
that the SARAR(1,0) specification may not be the correct specifica-
tion for this application. In the presence of mis-specifications,
the performance of estimators can change as shown in the next
application.
6.2. Dual gravity model
This second empirical illustration checks for the robustness of
some of the results presented in Behrens et al. (2012), which es-
timates the so-called gravity equation system using techniques
borrowed from spatial econometrics. For the analysis bilateral
trade flows, the gravity equation system yields convenient re-
gression specifications. Although theoretical ground is simple
and powerful, estimation is not straightforward owing to the
problem of how to account for the interdependence among
trade flows. If one were to implicitly assume that trade flows be-
tween any two trading partners are independent from the rest of
the world, then including ad hoc regional remoteness indices or
origin and destination specific fixed effects into the specification
could provide a way to control for interdependence. However,
the independence assumption between trade flows lacks theo-
retical foundation.
Behrens et al. (2012) make the observation that structural esti-
mation of gravity equation systems depends on the correct treat-
ment of the unobservable price indices. Their idea is to utilize,
instead, some other observed factor that has one-to-one correspon-
dence with unobserved price indices. To do so, they derive a gravity
equation from the quantity-based version of the constant elasticity
of substitution (CES) model by making the observation that price in-
dices are implicit functions of trade flows. Conveniently, the solution
yields an econometric specification that properly accounts for inter-
dependence between trade flows. Estimation can be carried out with
spatial econometrics methods which explicitly impose certain struc-
tures on spatial dependence over cross sectional units via spatial
41
The specification for the spatial weight matrix should also be considered. We also inves-
tigated the performance of estimators for this application through a Monte Carlo study in
which we use the data set of this application. We consider two data generating processes:
(i) the true data generating process is characterized by the parameter estimates reported
by the RGMME, (ii) the true data generating process is based on the parameter estimates
reported by the BE. The heteroskedastic disturbances are generated by using posterior
mean of vi. The simulation results for both data generating processes are inconsistent with
the results reported by the RGMME and BE in Table 1. This result may indicate the presence
of a model mis-specification for this application. The results are available upon request.
13
weight matrices. Moreover, the structural model yields the spatial
dependence structure, which allows skipping extensive sensitivity
analysis.
The model consists of n regions, each of which is populated with
workers who have inelastic labor supply. Region i is endowed with
Li unit of labor, which is the only production factor. Let Z ij ≡
F ij
be
Y iY j
the GDP-standardized trade flow from region i to region j, where
F ij is the value of trade flows from i to j, and Y i and Y j stand for
GDP level in i and j, respectively. The CES monopolistic competition
model with free entry in Behrens et al. (2012) yields the following
expression for the trade flows from region i to j


log Z ij ¼ −σlog ðLÞ−ðσ −1Þ
!

 X n
Lk
 X n
Lk

log τ ij − log τkj −σ log ðwi Þ−ðσ −1Þ log Z kj ;
k¼1
L k¼1
L
ð6:1Þ
where L i is the population of region i, L ≡ ∑ nk = 1 L k is the total
world population, τ ij is the iceberg trade costs of shipping one
unit of any variety between region i and j, and σ N 1 denotes the
constant elasticity of substitution between any two product varie-
ties. The last term in Eq. (6.1) shows that trade flow between region
i and j depends on all the trade flows from other regions k to region
j. Since varieties are substitutes, higher flows from other regions k
to region j reduce flows from region i to region j. Lower relative
trade barriers measured by the deviation of bilateral trade barriers
from the population weighted average, which are denoted by




n L
log τ ij −∑k¼1 k log τkj Þ, increase trade flows from i to j. Also,
L
lower wages in the origin region i improve the competitiveness of
region i's firms in market j and thus increase the trade flows from
i to j.
In matrix form Eq. (6.1) can be written as
ℤ ¼ −σ lnðLÞln2 −ðσ −1Þ In2 −W τ−σ w−ðσ−1ÞWℤ; ð6:2Þ
where ℤ denotes the n2 × 1 vector of log of GDP-standardized trade
flows stacked, l is the n2 × 1 vector of ones, τ is the n2 × 1 vector of
the log trade costs, w is the n2 × 1 vector of log origin wages, In2 is
the n2 × n2 identity matrix, and W = (S × D(L)) ⊗ In is the n2 × n2
spatial weight matrix. S is the n × n matrix whose elements are
equal to 1 and D(L) denotes the n × n diagonal matrix with kth
diagonal term of LL .42 k
Behrens et al. (2012) assume that the iceberg trade costs of shipping
one unit of any variety between region i and j relate to distance and
γ
border effects by τij = dijexp(ξbij), where the distance between regions
i and j is denoted by dij, and bij is a dummy variable taking the value 1 if
the trade flows cross an international border and zero otherwise. Taking
the logarithm and rewriting it in matrix form yields τ = γd + ξb,
where d is the n2 × 1 vector with elements log(dij) and b is the n2 × 1
vector of dummy variables indicating cross-border flows. Then, e τ¼
ðIn2 −WÞτ ¼ ðIn2 −WÞðγd þ ξbÞ ¼ γd e þ ξb
e . Substituting back into
Eq. (6.2) yields
e þ β w þ θb
ℤ ¼ β 0 ln 2 þ β 1 d e þ λWℤ; ð6:3Þ
2
where β0 ≡ −σln(L) b 0 is the constant term, β1 ≡ −(σ–1)γ b 0
is the distance coefficient of deviations from population weighted average
42
Note that the diagonal elements of W are not zero. The spatial weight matrices are
characterized with zero diagonal elements. A zero diagonal weight matrix can be ob-
tained by removing diagonal elements from W. Specifically, define Wdiag = D(L) ⊗ In.
Then, subtracting (σ–1)diagℤ from both sides of (6.2) yields the zero diagonal weight
matrix Wd = (W–Wdiag). However, this operation produces a heterogeneous coeffi-
cient model. In our application, we only consider the homogeneous coefficient model
so that W is assumed to have zerodiagonal elements.
14 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Table 2
SARAR(1,0) estimation results.

log(Zij) log(Zij)(US − US flows × 1.3)

ML RGMM BE 1 ML RGMM BE1

Intercept −13.8985 −14.2546 −13.2967 −14.0349 −15.6591 −13.5307

[−32.0439] [−17.2922] [−23.2082] [−19.9543] [−18.8065] [−23.7556]
log(wi) 1.1617 1.0889 1.1122 0.9393 0.8603 −0.8627
[−6.4775] [−5.2064] [−7.0701] [−5.2942] [−4.1377] [−5.5056]
e
dij −1.2550 −1.2445 −1.2296 −1.2519 −1.2336 −1.2243
[−36.6686] [−34.7267] [−39.7895] [−35.9740] −35.2560 [−40.0829]
e
bij −1.0458 −1.0631 −1.1767 −1.1493 −1.1606 −1.2944
[−15.9843] [−13.3942] [−19.2912] [−17.3638] [−15.0120] [−21.7863]
λ 0.0299 0.0203 0.0676 0.0439 0.0317 0.0794
[1.0677] [0.5915] [3.2101] [1.5774] [0.9245] [3.8114]
BP LM test 315.8548 – – 286.0448 – –
[0.0000] [0.0000]
Border Effects

CA (total) 6.3691 6.5670 8.0300 7.6498 7.8032 9.8882

[5.8951] [4.9702] [5.8121] [5.4576] [4.6372] [5.4914]
CA–CA (intra) 2.5237 2.5626 2.8337 2.7658 2.7934 3.1446
[43.4416] [36.4753] [52.4833] [47.2068] [40.8218] [60.5362]
CA–US (inter) 0.3962 0.3902 0.3529 0.3616 0.3580 0.3180
[6.8206] [5.5544] [6.5359] [6.1710] [5.2314] [6.1220]
US (total) 1.2717 1.2767 1.3105 1.3023 1.3056 1.3464
[65.8322] [54.1974] [70.6978] [65.1785] [55.7965] [72.4465]
US–US (intra) 1.1277 1.1299 1.1448 1.1412 1.1426 1.1603
[149.5478] [122.5164] [163.3460] [150.0577] [128.6453] [169.9878]
US–CA (inter) 0.8868 0.8850 0.8735 0.8763 0.8752 0.8618
[117.6011] [97.1922] [124.6458] [115.2290] [98.5322] [126.2545]
Average 2.8459 2.8955 3.2439 3.1563 3.1919 3.6487
[12.9498] [10.9439] [12.6654] [11.9126] [10.1149] [11.8836]

Note: The brackets contain t-statistics for the parameter estimates and p-values for the test statistics. The BP LM test stands for the Breusch–Pagan Lagrange multiplier test of
heteroskedasticity.

distances, β2 ≡ −σ b 0 is the coefficient of wage at the origin region,
θ ≡ −(σ–1)ξ b 0 denotes the coefficient of border effects, and the spatial
autoregressive parameter λ ≡ −(σ–1) b 0 captures the strength of the
interdependence across trade flows, i.e., a measure of spatial competition.
Adding a disturbance term to Eq. (6.3), we obtain the following
SARAR(1,0) regression specification
ℤ ¼ λWℤ þ Xγ þ ϵ; ð6:4Þ


e w; b
where X ¼ ln2 ; d; e , γ = (β ,β ,β ,θ)′, and ϵ is the n 2 × 1
0 1 2
column vector of independently distributed disturbance term. 43
6.2.1. Data and the results
The data set contains information on bilateral trade flows
between 30 U.S. states and 10 Canadian provinces.44 The aggregate
manufacturing exports F ij , internal absorption F ii , and regional
GDPs Yi are measured in millions of U.S. dollars for the year 1993.
The bilateral distances between the states and provinces using
the great circle formula applied to the state and province capitals'
geographic coordinates: between any two regions i and j





dij ¼ ℝ0  arccos cos ϱi −ϱ j cosðψi Þcos ψ j þ sinðψi Þsin ψ j ;
ð6:5Þ
43
Behrens et al. (2012) makes the observation that the multilateral resistance variable and
the border effects in any region may depend on a spatially weighted average of resistance
variables and border effects in all the other regions. Hence, they conjecture that error terms
may exhibit some form spatial dependence as well and consider two ways of introducing the
γ
error terms: (i) as noise in the measurement of trade costs, i.e., τij = dij exp(ξbij + ϵij), where
ϵij is an i.i.d. normal error term; and (ii) as noise in the measurement of the trade
flows, i.e., Zijreal ≡ Zijobsexp(ϵij). But, the authors assume a first-order spatial moving av-
erage process for the error process in an ad hoc manner and specify: u = ϵ–ρWdϵ,
where W d ≡ (W–D(L)⊗In ) and ρ is the spatial moving average parameter. Then,
Eq. (6.3) and the specification for u correspond to the SARMA(1,1) model.
44
Available at Robert C. Feenstra's data page: http://cid.econ.ucdavis.edu/.
where R0 is the Earth's radius, and ϱi and ψi denote region i's longi-


1 surf i 1=2
tude and latitude, respectively. We use dii ¼ to measure
3 π
internal trade costs, where surfi refers to region i's surface. 45 The
regional wages w i are average hourly manufacturing wages for
the year 1988 at the state and province level. The regional and pro-
vincial population shares are computed from 1988 population
figures. 46 The reason for choosing 5-year lagged values of regional
population and wage values is to control for potential endogeneity.
Behrens et al. (2012) successfully replicate the results in Anderson
and van Wincoop (2003) and Feenstra (2004) and show that there re-
mains a significant amount of cross-sectional correlation in the OLS resid-
uals. Therefore, the OLS estimator is inefficient and may be biased. Their
OLS estimates confirm that treating trade flows as independent yields
puzzling large values for the border coefficient.47
Table 2 presents the ML, the RGMM, and the Bayesian (BE1) esti-
mates for the SARAR(1,0) model. The columns 1–4 indicate that the
ML results are similar with the results presented in Behrens et al.
(2012). The Breusch–Pagan LM statistic provides strong evidence for
heteroskedasticity in the ML residuals, suggesting inconsistency for
the MLE. Fig. 3 shows the plots of the posterior mean of vis for this
application. The substantial deviation of the posterior mean of vi from
the unity also indicates the presence of heteroskedasticity in this
application.
The estimated parameter values for the spatial autoregressive pa-
rameter λ from all estimators are positive, which is opposite of the
45
Behrens et al. (2012) also employ two other measures of internal trade costs:
1=2
(i) 14min j dij , and (ii) 23ðsurf
i
π Þ .
46
See Behrens et al., 2012, Appendix B) for further details.
47
McCallum (1995) finds that the Canadian provinces seem to trade 21.5 to 27 times
more with themselves than with US states of equal size and distance.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 15

Posterior mean of variance components Posterior mean of variance components

16 18
SARAR(1,0) SARAR(1,0)

14 16

14
12

12
10
10
8
8
6
6
4
4

2 2

0 0
0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Observations Observations

a) Posterior mean of b) Posterior mean of : US-US flows×1.3

Fig. 3. Posterior mean of variance components vis: dual gravity model. (a) Posterior mean of vi. (b) Posterior mean of vi: US–US flows × 1:3.

predicted sign. The RGMM and ML estimates of the spatial The reduced form of the above equation can be written in vector
autoregressive parameter are insignificant, whereas the BE estimates form as follows
are significant. The results are incompatible with the qualitative predic-
tion of the model. Behrens et al. 2012 state that this anomaly stems from B ¼ θ In2 −λW
−1
b; ð6:7Þ
the specification adopted for the disturbance terms. There is no as-
sumed process for the disturbance terms in a SARAR(1,0) specification.  !
Behrens et al. (2012) show that when a SARAR(1,1) or a SARMA(1,1) Z
where B is the n2 × 1 vector stacking log ij , and b is the n 2 × 1
specification is adopted, the estimate of λ turns to be negative and sig- Z ij


n L
nificant as predicted by the underlying theory. The disturbance process vector with ijth element bij −∑k¼1 k bkj Þ . Under the assumption
L
is specified as ϵ = ρWϵ + ζ for a SARAR(1,1) model and ϵ = ρWζ + ζ −1
that I n2 −λW exists, the above equation can be written as B ¼
for a SARMA(1,1) model. Here, ζ is n2 × 1 vector of innovations and ρ is ∞ l l
θ∑l¼0 λ W b. The special structure of W implies that Wl b ¼ 0, for
another dependence parameter. In both specifications, the diagonal ele-
l = 1,2,…. (Behrens et al., 2012). Then, the log of border effects
ments of the covariance matrix of ϵ are not the same, suggesting
simplifies as

f
heteroskedasticity. Hence, when the true data generating process is
characterized by a SARAR(1,1) or a SARMA(1,1) model, the regression
X L
disturbance terms, i.e., the elements of ϵ, are heteroskedastic even if −θ k∈US k if i ∈CA; j∈CA ðCA−CAÞ;
the elements of innovation vector ζ are homoskedastic. Our results in L
Table 2 show that estimating the misspecified model SARAR(1,0) in- X L

 X n θ k∈US k ifi∈CA; j∈US ðCA−USÞ;
stead of SARAR(1,1) or SARMA(1,1), the heteroskedasticity robust esti- Lk L
log Bij ¼ θðbij − b Þ¼
mators do not yield the predicted sign for λ. Hence, the correct k¼1
L kj X Lk
specification of the disturbance process plays a vital role for the estima- θ k∈CA ifi∈US; j∈CA ðUS−CAÞ;
L
tion of the autoregressive parameter for this application. X L
Turning to the explanatory variables, we see that all estimators −θ k∈CA k ifi∈US; j∈US ðUS−USÞ:
L
capture the predicted signs. There does not seem to be a significant
difference between the estimated coefficients. In addition, the estimated ð6:8Þ
coefficients are almost identical to the ones reported for a SARAR(1,1)
and a SARMA(1,1) model in Behrens et al. (2012). This result indicates Thus, the border effects are decomposed into two components: the
that the coefficients of explanatory variables remain fairly stable across trade-boosting intra-national effect denoted by CA–CA and US–US,
specifications. and the trade reducing international effect denoted by CA–US and
The bottom lines of Table 2 report the estimated border effects, US–CA. The trade boosting intra-national effect is defined as the
which are defined as the ratio of trade flows in a world with border to increase in the intra-national trade due to the presence of the borders.
that in a borderless world. Let Zij be the GDP standardized trade flows The rationale is that the international borders protect domestic firms
in a world with border, and Z ij be the GDP standardized trade flows in against the competition of foreign firms and thus create an advantage
a borderless world. From Eq. (6.1), the logarithm of the border effects for the domestic firm. The trade reducing international effect is defined
 
Z
log Bij ¼ log ij can be written as as the effect of the international border on trade flows across countries.
Z ij The volume of trade flows will be higher in a borderless world than in a
world with borders. Finally, the logarithm of total border effects is
! ! ! obtained by combining the trade boosting and reducing effects, which

 Z ij X n
Lk X
n
L Z kj L
log Bij ¼ log ¼ θ bij − b þλ k
log : ð6:6Þ are given by −2ξλ∑k∈US k for the Canadian provinces and −2ξλ
L kj L L
Z ij Z kj Lk
k¼1 k¼1 ∑k∈CA for the US states.
L
16 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

The estimated border effects for Canadian provinces and US
states in Table 2 are similar to those reported by Behrens et al.
(2012) and they range from 6.3 to 9.8 for Canadian provinces and
cluster around 1.3 for US states. 48 The total border effect for the
Canadian provinces is estimated as 6.5670 by the RGMME in the
second column. The same column shows that the total border effect
of U.S. is 1.2767.
As a robustness check, columns 4–6 of Table 2 replicates the re-
sults when all US state–state trade flows are inflated by a factor of
1.3, because the construction of data set involves scaling of shipping
data, which may understate trade flows between US states while
making the Canadian intra border effects seem significantly larger
than that of US states.49 The estimated parameter coefficients are
similar to the ones in columns 1–3. The estimated values of elasticity of
substitution however are less than unity contrary to the model's
assumption.
Rubinfeld (1978) data set described in Gilley and Kelley Pace
(1996) is used to estimate a hedonic housing-price model for a
sample of 506 observations of census tracts in the Boston Standard
Metropolitan Area in 1970. The estimation results show that the pa-
rameter estimates for the pollution and housing age variables do not
agree, which in turn imply important policy implication toward
these variables. In the second application, we try to see the effect
of ignoring potential heteroskedasticity in a gravity model of bilat-
eral trade flows between Canadian provinces and US states. We
replicate some of the results in Behrens et al. (2012). Although
the structural model dictates negative cross-sectional correlation
among trade flows, the heteroskedasticity robust RGMM and
Bayesian estimators do not able to capture it. This result may indicate
that the correct specification of the process for the disturbance terms is
vital for the estimation of the autoregressive parameter in spatial
models.

7. Conclusion Appendix A. Some useful lemmas

This study examines the effect of unknown forms of heteroskedasticity

on spatial models that have a spatial autoregressive process in the Lemma 1. Completion of Square Let A and B be positive definite n × n
dependent variable. The GMM and Bayesian methodologies have matrices. Define C−1 = A−1 + B−1 and D = A + B. Let x, α and γ be
the advantage that the estimator formulated from these estimation n × 1 vectors. Then, there exists an n × 1 vector μ such that
frameworks can be robust to an unknown form of heteroskedasticity. ′ −1 ′ −1 ′ −1 ′ −1
ðx−α Þ A ðx−α Þ þ ðx−γÞ B ðx−γÞ ¼ ðx−μ Þ C ðx−μ Þ þ ðα−γÞ D ðα−γÞ;
In the GMM framework, the unknown heteroskedasticity is incorpo-
rated into estimation through the formulation of appropriate mo- ðA:1Þ
ment functions. The quadratic moment functions considered for
the estimation are designed in a such way that the orthogonality where μ = C(A−1α + B−1γ).
condition of the moment functions is not violated in the presence Proof. See Abadir and Magnus (2005, Exercise 8.15, p. 216). □
of heteroskedastic disturbance terms. In the Bayesian framework,
the structure of the heteroskedasticity is assumed to have two Lemma 2. Let An, Bn and Cn be n × n matrices with ijth elements respec-
components: (i) a constant component and (ii) a component that tively denoted by an,ij, bn,ij and cn,ij. Assume that An and Bn have zero diag-
varies across observations. The important assumption of the Bayes- onal elements, and Cn has uniformly bounded row and column sums in
ian approach is that the component that varies over observations is absolute value. Assume that εn satisfies Assumption 1 with covariance
assumed to be i.i.d. to facilitate the estimation. The mean of the matrix denoted by Σn = Diag{σ2n1,…,σ2nn}. Then,
resulting conditional posterior distribution of the parameters of
(1)
the exogenous variables indicates that the Bayesian estimator
 Xn X
n

′ ′ 2 2
mimics the generalized least squares estimator, where the E ε n An εn  εn Bn ε n ¼ an;ij bn;ij þ bn;ji σ ni σ nj
i¼1 j¼1
aberrant observations are automatically down-weighted in the


′
estimation. ¼ tr Σn An Bn Σn þ Σn Bn
To compare finite sample properties of the heteroskedastiticy

2 2
robust estimators, a Monte Carlo simulation is designed for the where Σn ¼ Diag σ n1 ; …; σ nn :
case of SARAR(1,0). The Monte Carlo simulation results show that
the MLE of the autoregressive parameter imposes significant bias (2)
and the bias amount does not decrease as the sample size increases. h
 i
2
X
n
2 4 4
The RGMME imposes trivial biases in all cases. The Bayesian estima- Eðε n C n εÞ ¼ cn;ii E ε ni −3σ ni
tor is efficient relative to the RGMME, but it imposes relatively larger i¼1
X
n 2 n X
X n

bias on the autoregressive parameter when there is strong negative 2 2 2
þ cn;ii σ ni Þ þ cn;ij cn;ij þ cn;ji σ ni σ nj
spatial dependence. The Monte Carlo results for the estimates of i¼1 i¼1 j¼1
the total effects of exogenous variables show that the RGMME is X
n h
 i
2 4 4 2
slightly better than the Bayesian estimator in terms of bias. In ¼ cn;ii E εni −3σ ni þ tr ðΣn C n Þ
i¼1
terms of finite sample efficiency, the Bayesian estimator is more

′
efficient. þ tr Σn C n C n Σn þ Σn C n Σn C n ;
In the final section, we present two empirical illustrations to
show how the heteroskedasticity robust estimators perform in ap- (3)
plied research. In the first illustrative application, the Harrison and
X
n h
 i Xn X
n

2 4 4 2 2
varðε n C n ε Þ ¼ cn;ii E ε ni −3σ ni þ cn;ij cn;ij þ cn;ji σ ni σ nj
i¼1 i¼1 j¼1

X
n h
 i

2 4 4 ′
¼ cn;ii E ε ni −3σ ni þ tr Σn C n C n Σn þ Σn C n Σn C n :
i¼1
48
Behrens et al. 2012 report that the border effects cluster respectively around 14.5
and 1.5 for Canadian provinces and US states when spatial dependence structure is
ignored.
49
When constructing the data set, Anderson and van Wincoop (2003) employs a
scaling scheme to US state–state trade flows taken from the Commodity Flow Survey
in order to make it more comparable to the Canada–US flows. Proof. See Lin and Lee (2010). □
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 17

Appendix B. Monte Carlo results

Table 3
c ¼ 10; m ¼ 5; m ¼ 15; N ¼ 486.

λ (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE]

λML λRGMM λBE1 λBE2

−0.9 (−0.7203)[0.1797](0.0551)[0.1879] (−0.9110)[−0.0110](0.1090)[0.1096] (−0.7589)[0.1411](0.0523)[0.1505] (−0.7557)[0.1443](0.0520)[0.1534]

−0.6 (−0.4613)[0.1387](0.0751)[0.1578] (−0.5912)[0.0088](0.1025)[0.1028] (−0.4804)[0.1196](0.0714)[0.1393] (−0.4784)[0.1216](0.0710)[0.1408]
−0.3 (−0.2395)[0.0605](0.0838)[0.1033] (−0.2982)[0.0018](0.1112)[0.1112] (−0.2477)[0.0523](0.0827)[0.0978] (−0.2463)[0.0537](0.0827)[0.0986]
0.0 (−0.0072)[−0.0072](0.0896)[0.0899] (−0.0031)[−0.0031](0.1086)[0.1086] (−0.0079)[−0.0079](0.0884)[0.0887] (−0.0064)[−0.0064](0.0884)[0.0886]
0.3 (0.2479)[−0.0521](0.0825)[0.0976] (0.2913)[−0.0087](0.1017)[0.1021] (0.2501)[−0.0499](0.0806)[0.0948] (0.2514)[−0.0486](0.0806)[0.0941]
0.6 (0.5372)[−0.0628](0.0704)[0.0944] (0.6005)[0.0005](0.1180)[0.1180] (0.5424)[−0.0576](0.0688)[0.0897] (0.5437)[−0.0563](0.0688)[0.0889]
0.9 (0.8674)[−0.0326](0.0358)[0.0484] (0.9125)[0.0125](0.1052)[0.1059] (0.8699)[−0.0301](0.0336)[0.0451] (0.8709)[−0.0291](0.0336)[0.0444]

β10,ML β10,RGMM β 10;BE1 β10;BE2

−0.9 (0.9261)[−0.0739](1.7065)[1.7081] (1.1290)[0.1290](1.7011)[1.7060] (0.9006)[−0.0994](1.5768)[1.5800] (0.8942)[−0.1058](1.5804)[1.5839]

−0.6 (0.9745)[−0.0255](1.6667)[1.6669] (1.0447)[0.0447](1.6786)[1.6792] (0.9757)[−0.0243](1.5551)[1.5553] (0.9740)[−0.0260](1.5545)[1.5547]
−0.3 (1.0712)[0.0712](1.6753)[1.6769] (1.0948)[0.0948](1.6830)[1.6856] (1.0632)[0.0632](1.5654)[1.5667] (1.0637)[0.0637](1.5659)[1.5672]
0.0 (1.0596)[0.0596](1.7130)[1.7141] (1.0677)[0.0677](1.7169)[1.7183] (1.0704)[0.0704](1.6005)[1.6021] (1.0723)[0.0723](1.6013)[1.6029]
0.3 (0.9477)[−0.0523](1.7354)[1.7362] (0.9509)[−0.0491](1.7225)[1.7232] (0.9514)[−0.0486](1.6123)[1.6130] (0.9497)[−0.0503](1.6113)[1.6121]
0.6 (0.9920)[−0.0080](1.7060)[1.7061] (1.0264)[0.0264](1.6966)[1.6968] (0.9923)[−0.0077](1.5789)[1.5790] (0.9935)[−0.0065](1.5772)[1.5772]
0.9 (1.0623)[0.0623](1.7871)[1.7882] (1.1268)[0.1268](1.7835)[1.7880] (1.0401)[0.0401](1.6837)[1.6842] (1.0431)[0.0431](1.6822)[1.6827]

β20,ML β20,RGMM β 20;BE1 β20;BE2

−0.9 (1.0875)[0.0875](1.7203)[1.7225] (0.8462)[−0.1538](1.7133)[1.7202] (1.1158)[0.1158](1.5641)[1.5684] (1.1232)[0.1232](1.5687)[1.5735]

−0.6 (1.0352)[0.0352](1.6866)[1.6869] (0.9537)[−0.0463](1.7018)[1.7024] (1.0343)[0.0343](1.5396)[1.5399] (1.0362)[0.0362](1.5386)[1.5390]
−0.3 (0.9308)[−0.0692](1.6607)[1.6621] (0.9034)[−0.0966](1.6671)[1.6699] (0.9421)[−0.0579](1.5188)[1.5199] (0.9412)[−0.0588](1.5185)[1.5197]
0.0 (0.9321)[−0.0679](1.7374)[1.7387] (0.9199)[−0.0801](1.7400)[1.7419] (0.9249)[−0.0751](1.5812)[1.5829] (0.9229)[−0.0771](1.5819)[1.5838]
0.3 (1.0569)[0.0569](1.7367)[1.7377] (1.0412)[0.0412](1.7237)[1.7242] (1.0572)[0.0572](1.5793)[1.5803] (1.0583)[0.0583](1.5784)[1.5795]
0.6 (1.0315)[0.0315](1.7241)[1.7244] (0.9730)[−0.0270](1.7139)[1.7141] (1.0295)[0.0295](1.5605)[1.5608] (1.0278)[0.0278](1.5584)[1.5586]
0.9 (0.9725)[−0.0275](1.8035)[1.8037] (0.8681)[−0.1319](1.8027)[1.8076] (0.9973)[−0.0027](1.6606)[1.6606] (0.9932)[−0.0068](1.6591)[1.6591]

Table 4
c ¼ 10; m ¼ 7; m ¼ 21; N ¼ 974.

λ (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE]

λML λRGMM λBE1 λBE2

−0.9 (−0.7196)[0.1804](0.0407)[0.1850] (−0.9032)[−0.0032](0.0675)[0.0676] (−0.7616)[0.1384](0.0377)[0.1434] (−0.7598)[0.1402](0.0377)[0.1452]

−0.6 (−0.4653)[0.1347](0.0553)[0.1456] (−0.5985)[0.0015](0.0743)[0.0743] (−0.4864)[0.1136](0.0523)[0.1251] (−0.4849)[0.1151](0.0522)[0.1264]
−0.3 (−0.2398)[0.0602](0.0614)[0.0859] (−0.3002)[−0.0002](0.0808)[0.0808] (−0.2506)[0.0494](0.0605)[0.0782] (−0.2492)[0.0508](0.0607)[0.0792]
0.0 (−0.0000)[−0.0000](0.0652)[0.0652] (0.0034)[0.0034](0.0778)[0.0779] (−0.0003)[−0.0003](0.0638)[0.0638] (0.0011)[0.0011](0.0639)[0.0639]
0.3 (0.2604)[−0.0396](0.0589)[0.0710] (0.3016)[0.0016](0.0661)[0.0661] (0.2630)[−0.0370](0.0587)[0.0694] (0.2644)[−0.0356](0.0588)[0.0687]
0.6 (0.5461)[−0.0539](0.0475)[0.0718] (0.5991)[−0.0009](0.0483)[0.0483] (0.5515)[−0.0485](0.0463)[0.0671] (0.5530)[−0.0470](0.0464)[0.0661]
0.9 (0.8721)[−0.0279](0.0248)[0.0373] (0.9050)[0.0050](0.0591)[0.0593] (0.8746)[−0.0254](0.0237)[0.0348] (0.8760)[−0.0240](0.0237)[0.0337]

β10,ML β10,RGMM β 10;BE1 β 10;BE2

−0.9 (1.0211)[0.0211](1.6046)[1.6047] (1.0610)[0.0610](1.6076)[1.6087] (0.9887)[−0.0113](1.5532)[1.5532] (0.9894)[−0.0106](1.5513)[1.5513]

−0.6 (1.0065)[0.0065](1.6063)[1.6063] (1.0030)[0.0030](1.6123)[1.6123] (0.9683)[−0.0317](1.5478)[1.5481] (0.9680)[−0.0320](1.5495)[1.5499]
−0.3 (1.0701)[0.0701](1.6015)[1.6030] (1.0658)[0.0658](1.6139)[1.6153] (1.0779)[0.0779](1.5654)[1.5673] (1.0786)[0.0786](1.5674)[1.5694]
0.0 (1.0649)[0.0649](1.6261)[1.6274] (1.0731)[0.0731](1.6267)[1.6284] (1.0809)[0.0809](1.5482)[1.5503] (1.0809)[0.0809](1.5501)[1.5522]
0.3 (0.9337)[−0.0663](1.6115)[1.6128] (0.9535)[−0.0465](1.5974)[1.5981] (0.9377)[−0.0623](1.5471)[1.5483] (0.9398)[−0.0602](1.5444)[1.5456]
0.6 (0.9652)[−0.0348](1.6350)[1.6353] (1.0018)[0.0018](1.6136)[1.6136] (0.9817)[−0.0183](1.5528)[1.5529] (0.9814)[−0.0186](1.5545)[1.5546]
0.9 (0.9027)[−0.0973](1.6747)[1.6775] (0.9567)[−0.0433](1.6404)[1.6409] (0.8988)[−0.1012](1.6249)[1.6281] (0.9012)[−0.0988](1.6229)[1.6259]

β20,ML β20,RGMM β 20;BE1 β 20;BE2

−0.9 (1.0287)[0.0287](1.6837)[1.6839] (0.9457)[−0.0543](1.6895)[1.6903] (1.0572)[0.0572](1.6210)[1.6220] (1.0568)[0.0568](1.6193)[1.6203]

−0.6 (1.0047)[0.0047](1.6904)[1.6904] (0.9909)[−0.0091](1.6999)[1.6999] (1.0455)[0.0455](1.6202)[1.6209] (1.0459)[0.0459](1.6219)[1.6226]
−0.3 (0.9319)[−0.0681](1.6776)[1.6789] (0.9311)[−0.0689](1.6916)[1.6930] (0.9224)[−0.0776](1.6304)[1.6323] (0.9216)[−0.0784](1.6325)[1.6344]
0.0 (0.9348)[−0.0652](1.7068)[1.7081] (0.9217)[−0.0783](1.7087)[1.7105] (0.9188)[−0.0812](1.6124)[1.6144] (0.9192)[−0.0808](1.6146)[1.6166]
0.3 (1.0642)[0.0642](1.6934)[1.6946] (1.0330)[0.0330](1.6789)[1.6793] (1.0577)[0.0577](1.6188)[1.6199] (1.0554)[0.0554](1.6158)[1.6168]
0.6 (1.0585)[0.0585](1.7133)[1.7143] (0.9997)[−0.0003](1.6908)[1.6908] (1.0392)[0.0392](1.6154)[1.6159] (1.0389)[0.0389](1.6169)[1.6174]
0.9 (1.1256)[0.1256](1.7475)[1.7520] (1.0394)[0.0394](1.7094)[1.7099] (1.1259)[0.1259](1.6853)[1.6900] (1.1217)[0.1217](1.6828)[1.6872]
18 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Table 5
c ¼ 10; m ¼ 14; m ¼ 20; N ¼ 485.

λ (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE]

λML λRGMM λBE1 λBE2

−0.9 (−0.7979)[0.1021](0.0358)[0.1082] (−0.9074)[−0.0074](0.0646)[0.0650] (−0.8521)[0.0479](0.0274)[0.0552] (−0.8503)[0.0497](0.0274)[0.0567]

−0.6 (−0.4761)[0.1239](0.0540)[0.1352] (−0.5963)[0.0037](0.0653)[0.0654] (−0.5360)[0.0640](0.0480)[0.0799] (−0.5346)[0.0654](0.0481)[0.0812]
−0.3 (−0.2326)[0.0674](0.0697)[0.0969] (−0.2982)[0.0018](0.0906)[0.0906] (−0.2642)[0.0358](0.0625)[0.0720] (−0.2628)[0.0372](0.0625)[0.0728]
0.0 (−0.0031)[−0.0031](0.0741)[0.0742] (−0.0020)[−0.0020](0.0921)[0.0921] (−0.0051)[−0.0051](0.0668)[0.0670] (−0.0036)[−0.0036](0.0670)[0.0671]
0.3 (0.2378)[−0.0622](0.0780)[0.0998] (0.2921)[−0.0079](0.0891)[0.0894] (0.2623)[−0.0377](0.0663)[0.0762] (0.2639)[−0.0361](0.0664)[0.0755]
0.6 (0.5254)[−0.0746](0.0663)[0.0998] (0.5954)[−0.0046](0.0647)[0.0649] (0.5564)[−0.0436](0.0526)[0.0683] (0.5580)[−0.0420](0.0526)[0.0673]
0.9 (0.8604)[−0.0396](0.0361)[0.0536] (0.8961)[−0.0039](0.0345)[0.0348] (0.8746)[−0.0254](0.0267)[0.0368] (0.8759)[−0.0241](0.0267)[0.0359]

β10,ML β10,RGMM β 10;BE1 β10;BE2

−0.9 (1.2551)[0.2551](1.9547)[1.9713] (1.0779)[0.0779](1.9049)[1.9065] (1.1025)[0.1025](1.6337)[1.6369] (1.1071)[0.1071](1.6336)[1.6371]

−0.6 (1.0615)[0.0615](1.8573)[1.8583] (0.9797)[−0.0203](1.8598)[1.8599] (1.0142)[0.0142](1.5712)[1.5713] (1.0140)[0.0140](1.5726)[1.5727]
−0.3 (0.9810)[−0.0190](1.8731)[1.8731] (0.9801)[−0.0199](1.8646)[1.8647] (0.9874)[−0.0126](1.5379)[1.5380] (0.9858)[−0.0142](1.5400)[1.5401]
0.0 (1.0196)[0.0196](1.7893)[1.7894] (1.0574)[0.0574](1.7860)[1.7869] (1.0121)[0.0121](1.4868)[1.4869] (1.0119)[0.0119](1.4876)[1.4876]
0.3 (1.0065)[0.0065](1.9023)[1.9023] (1.0605)[0.0605](1.8507)[1.8517] (1.0109)[0.0109](1.5385)[1.5385] (1.0106)[0.0106](1.5382)[1.5382]
0.6 (1.0144)[0.0144](1.9185)[1.9185] (1.0679)[0.0679](1.8391)[1.8404] (1.0228)[0.0228](1.5285)[1.5286] (1.0220)[0.0220](1.5265)[1.5267]
0.9 (0.9293)[−0.0707](1.9317)[1.9330] (0.9907)[−0.0093](1.8532)[1.8532] (0.9569)[−0.0431](1.5931)[1.5936] (0.9586)[−0.0414](1.5892)[1.5898]

β20,ML β20,RGMM β 20;BE1 β20;BE2

−0.9 (0.8054)[−0.1946](2.0958)[2.1048] (0.9142)[−0.0858](2.0485)[2.0503] (0.9299)[−0.0701](1.7327)[1.7341] (0.9262)[−0.0738](1.7330)[1.7346]

−0.6 (0.9638)[−0.0362](1.9981)[1.9984] (1.0154)[0.0154](2.0047)[2.0047] (0.9977)[−0.0023](1.6740)[1.6740] (0.9982)[−0.0018](1.6750)[1.6750]
−0.3 (1.0238)[0.0238](2.0178)[2.0180] (1.0169)[0.0169](2.0113)[2.0113] (1.0123)[0.0123](1.6365)[1.6366] (1.0140)[0.0140](1.6384)[1.6385]
0.0 (0.9696)[−0.0304](1.9381)[1.9383] (0.9249)[−0.0751](1.9356)[1.9371] (0.9782)[−0.0218](1.5848)[1.5849] (0.9786)[−0.0214](1.5856)[1.5857]
0.3 (1.0087)[0.0087](2.0577)[2.0577] (0.9291)[−0.0709](2.0037)[2.0049] (0.9952)[−0.0048](1.6455)[1.6455] (0.9948)[−0.0052](1.6449)[1.6449]
0.6 (1.0320)[0.0320](2.0703)[2.0706] (0.9267)[−0.0733](1.9850)[1.9864] (1.0049)[0.0049](1.6295)[1.6295] (1.0049)[0.0049](1.6274)[1.6274]
0.9 (1.1401)[0.1401](2.0854)[2.0901] (1.0059)[0.0059](2.0041)[2.0041] (1.0850)[0.0850](1.6963)[1.6984] (1.0811)[0.0811](1.6921)[1.6941]

Table 6
c ¼ 10; m ¼ 20; m ¼ 28; N ¼ 945.

λ (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE] (Mean)[Bias](Std.)[RMSE]

λML λRGMM λBE1 λBE2

−0.9 (−0.7963)[0.1037](0.0275)[0.1073] (−0.9006)[−0.0006](0.0371)[0.0371] (−0.8542)[0.0458](0.0202)[0.0500] (−0.8528)[0.0472](0.0201)[0.0513]

−0.6 (−0.4706)[0.1294](0.0424)[0.1362] (−0.5961)[0.0039](0.0497)[0.0498] (−0.5348)[0.0652](0.0364)[0.0747] (−0.5334)[0.0666](0.0365)[0.0760]
−0.3 (−0.2329)[0.0671](0.0489)[0.0830] (−0.3030)[−0.0030](0.0646)[0.0647] (−0.2671)[0.0329](0.0435)[0.0545] (−0.2656)[0.0344](0.0434)[0.0554]
0.0 (0.0015)[0.0015](0.0587)[0.0587] (0.0022)[0.0022](0.0722)[0.0722] (0.0010)[0.0010](0.0511)[0.0511] (0.0025)[0.0025](0.0512)[0.0512]
0.3 (0.2421)[−0.0579](0.0591)[0.0827] (0.2964)[−0.0036](0.0667)[0.0668] (0.2675)[−0.0325](0.0489)[0.0587] (0.2689)[−0.0311](0.0489)[0.0580]
0.6 (0.5269)[−0.0731](0.0538)[0.0907] (0.6013)[0.0013](0.0527)[0.0527] (0.5602)[−0.0398](0.0420)[0.0579] (0.5616)[−0.0384](0.0420)[0.0569]
0.9 (0.8602)[−0.0398](0.0276)[0.0484] (0.8982)[−0.0018](0.0250)[0.0250] (0.8772)[−0.0228](0.0195)[0.0300] (0.8787)[−0.0213](0.0195)[0.0289]

β10,ML β10,RGMM β 10;BE1 β 10;BE2

−0.9 (1.1639)[0.1639](1.3388)[1.3488] (1.0210)[0.0210](1.2981)[1.2983] (1.0792)[0.0792](1.2288)[1.2313] (1.0812)[0.0812](1.2293)[1.2320]

−0.6 (1.0914)[0.0914](1.3231)[1.3263] (1.0228)[0.0228](1.3082)[1.3084] (1.0501)[0.0501](1.2298)[1.2308] (1.0501)[0.0501](1.2283)[1.2293]
−0.3 (1.1016)[0.1016](1.3052)[1.3091] (1.0753)[0.0753](1.3066)[1.3088] (1.0902)[0.0902](1.2222)[1.2255] (1.0919)[0.0919](1.2214)[1.2249]
0.0 (0.9676)[−0.0324](1.2531)[1.2535] (0.9726)[−0.0274](1.2533)[1.2536] (0.9700)[−0.0300](1.2016)[1.2020] (0.9708)[−0.0292](1.2024)[1.2028]
0.3 (1.0165)[0.0165](1.2885)[1.2886] (1.0348)[0.0348](1.2742)[1.2747] (1.0103)[0.0103](1.1888)[1.1888] (1.0089)[0.0089](1.1881)[1.1881]
0.6 (0.9443)[−0.0557](1.2591)[1.2603] (0.9747)[−0.0253](1.2198)[1.2200] (0.9769)[−0.0231](1.1702)[1.1704] (0.9769)[−0.0231](1.1702)[1.1704]
0.9 (0.9342)[−0.0658](1.3342)[1.3358] (1.0016)[0.0016](1.2815)[1.2815] (0.9722)[−0.0278](1.2024)[1.2027] (0.9718)[−0.0282](1.2024)[1.2028]

β20,ML β20,RGMM β 20;BE1 β 20;BE2

−0.9 (0.9104)[−0.0896](1.4123)[1.4152] (0.9605)[−0.0395](1.3741)[1.3747] (0.9466)[−0.0534](1.2838)[1.2849] (0.9457)[−0.0543](1.2840)[1.2852]

−0.6 (0.9490)[−0.0510](1.4096)[1.4106] (0.9742)[−0.0258](1.3957)[1.3960] (0.9702)[−0.0298](1.2991)[1.2994] (0.9706)[−0.0294](1.2974)[1.2977]
−0.3 (0.9017)[−0.0983](1.3883)[1.3918] (0.9169)[−0.0831](1.3910)[1.3935] (0.9087)[−0.0913](1.2894)[1.2926] (0.9070)[−0.0930](1.2887)[1.2921]
0.0 (1.0299)[0.0299](1.3247)[1.3251] (1.0210)[0.0210](1.3250)[1.3252] (1.0258)[0.0258](1.2601)[1.2604] (1.0248)[0.0248](1.2612)[1.2614]
0.3 (0.9876)[−0.0124](1.3666)[1.3667] (0.9528)[−0.0472](1.3510)[1.3518] (0.9887)[−0.0113](1.2478)[1.2478] (0.9899)[−0.0101](1.2474)[1.2474]
0.6 (1.0860)[0.0860](1.3324)[1.3351] (1.0149)[0.0149](1.2895)[1.2896] (1.0348)[0.0348](1.2278)[1.2283] (1.0340)[0.0340](1.2278)[1.2282]
0.9 (1.1057)[0.1057](1.4140)[1.4180] (0.9816)[−0.0184](1.3558)[1.3559] (1.0441)[0.0441](1.2568)[1.2576] (1.0427)[0.0427](1.2569)[1.2576]

Table 7
Total effects: xn,1.

MLE RGMME BE1

m ¼ 5; m ¼ 15; N ¼ 486
−0.9 (0.4487)[−0.0776](1.0363)[1.0392] (0.5128)[−0.0135](0.9256)[0.9257] (0.4484)[−0.0779](0.9593)[0.9624]
−0.6 (0.7363)[0.1113](1.1686)[1.1738] (0.7153)[0.0903](1.0786)[1.0824] (0.7276)[0.1026](1.0744)[1.0792]
−0.3 (0.8202)[0.0510](1.4148)[1.4157] (0.8014)[0.0322](1.3711)[1.3715] (0.8060)[0.0368](1.2885)[1.2890]
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21 19

Table 7 (continued)
MLE RGMME BE1

0.0 (0.8703)[−0.1297](1.7656)[1.7703] (0.8642)[−0.1358](1.8425)[1.8475] (0.8827)[−0.1173](1.6362)[1.6404]

0.3 (1.3902)[−0.0384](2.3117)[2.3120] (1.4769)[0.0483](2.5512)[2.5516] (1.4106)[−0.0180](2.1430)[2.1431]
0.6 (2.2905)[−0.2095](3.8475)[3.8532] (2.6775)[0.1775](4.4527)[4.4563] (2.2950)[−0.2050](3.6186)[3.6245]
0.9 (7.3677)[−2.6323](14.645)[14.880] (9.0038)[−0.9962](24.232)[24.252] (7.9447)[−2.0553](13.801)[13.953]

m ¼ 7; m ¼ 21; N ¼ 974
−0.9 (0.5479)[0.0216](0.9239)[0.9241] (0.5134)[−0.0129](0.8382)[0.8383] (0.5080)[−0.0184](0.8720)[0.8722]
−0.6 (0.6859)[0.0609](1.0811)[1.0828] (0.6235)[−0.0015](0.9967)[0.9967] (0.6850)[0.0600](1.0362)[1.0379]
−0.3 (0.8668)[0.0976](1.3092)[1.3128] (0.8265)[0.0573](1.2611)[1.2624] (0.8461)[0.0768](1.2411)[1.2434]
0.0 (1.0005)[0.0005](1.6295)[1.6295] (1.0103)[0.0103](1.6371)[1.6371] (1.0148)[0.0148](1.5526)[1.5526]
0.3 (1.3622)[−0.0663](2.2496)[2.2506] (1.4704)[0.0418](2.3708)[2.3712] (1.3653)[−0.0633](2.1583)[2.1592]
0.6 (2.2510)[−0.2490](3.6965)[3.7049] (2.6373)[0.1373](4.2110)[4.2132] (2.2459)[−0.2541](3.5782)[3.5872]
0.9 (8.1163)[−1.8837](13.551)[13.682] (10.384)[0.3836](17.349)[17.353] (8.1433)[−1.8567](13.022)[13.154]

m ¼ 14; m ¼ 20; N ¼ 485

−0.9 (0.6943)[0.1679](1.0477)[1.0611] (0.5551)[0.0288](0.9621)[0.9625] (0.6045)[0.0782](0.8733)[0.8768]
−0.6 (0.7055)[0.0805](1.2695)[1.2720] (0.6059)[−0.0191](1.1785)[1.1787] (0.6607)[0.0357](0.9952)[0.9958]
−0.3 (0.8508)[0.0815](1.5437)[1.5459] (0.8055)[0.0363](1.4844)[1.4848] (0.7591)[−0.0101](1.2443)[1.2443]
0.0 (0.9627)[−0.0373](1.8300)[1.8304] (1.0029)[0.0029](1.8371)[1.8371] (1.0133)[0.0133](1.4842)[1.4842]
0.3 (1.3411)[−0.0875](2.5335)[2.5350] (1.5446)[0.1160](2.7145)[2.7170] (1.3902)[−0.0384](2.1090)[2.1093]
0.6 (2.1366)[−0.3634](4.2218)[4.2374] (2.7278)[0.2278](4.8890)[4.8943] (2.2631)[−0.2369](3.5772)[3.5851]
0.9 (7.5849)[−2.4151](16.024)[16.205] (10.984)[0.9836](20.976)[20.999] (8.3183)[−1.6817](14.328)[14.427]

m ¼ 20; m ¼ 28; N ¼ 945

−0.9 (0.6389)[0.1126](0.7251)[0.7338] (0.5345)[0.0081](0.6601)[0.6601] (0.5817)[0.0554](0.6393)[0.6417]
−0.6 (0.7261)[0.1011](0.8875)[0.8932] (0.6232)[−0.0018](0.8093)[0.8093] (0.6763)[0.0513](0.7986)[0.8003]
−0.3 (0.8621)[0.0928](0.9955)[0.9998] (0.8014)[0.0321](0.9465)[0.9470] (0.8192)[0.0500](0.9215)[0.9229]
0.0 (1.0264)[0.0264](1.2393)[1.2396] (1.0317)[0.0317](1.2483)[1.2487] (0.9975)[−0.0025](1.1756)[1.1756]
0.3 (1.4071)[−0.0215](1.7296)[1.7297] (1.5683)[0.1397](1.8599)[1.8651] (1.4363)[0.0077](1.6417)[1.6417]
0.6 (2.0415)[−0.4585](2.7426)[2.7806] (2.5226)[0.0226](3.1775)[3.1776] (2.2410)[−0.2590](2.7008)[2.7132]
0.9 (6.8801)[−3.1199](9.9827)[10.458] (10.117)[0.1173](13.173)[13.173] (7.8362)[−2.1638](10.039)[10.269]

(Mean)[Bias](Std.)[RMSE].

Table 8
Total effects: xn,2.

MLE RGMME BE1

m ¼ 5; m ¼ 15; N ¼ 486
−0.9 (0.7405)[0.2142](1.0495)[1.0711] (0.5400)[0.0137](0.9383)[0.9384] (0.7143)[0.1880](0.9527)[0.9711]
−0.6 (0.6426)[0.0176](1.1832)[1.1833] (0.5452)[−0.0798](1.0945)[1.0974] (0.6336)[0.0086](1.0625)[1.0625]
−0.3 (0.7893)[0.0200](1.4192)[1.4194] (0.7363)[−0.0330](1.3769)[1.3773] (0.7930)[0.0238](1.2671)[1.2673]
0.0 (1.1352)[0.1352](1.7735)[1.7786] (1.1425)[0.1425](1.7978)[1.8034] (1.1187)[0.1187](1.6122)[1.6166]
0.3 (1.3130)[−0.1155](2.3382)[2.3410] (1.3641)[−0.0644](2.5104)[2.5112] (1.3039)[−0.1247](2.1258)[2.1295]
0.6 (2.1650)[−0.3350](3.8321)[3.8467] (2.2367)[−0.2633](4.8739)[4.8810] (2.2058)[−0.2942](3.5488)[3.5610]
0.9 (8.8208)[−1.1792](14.870)[14.917] (8.983)[−1.0173](21.194)[21.219] (8.4111)[−1.5889](13.663)[13.755]

m ¼ 7; m ¼ 21; N ¼ 974
−0.9 (0.6375)[0.1112](0.9676)[0.9740] (0.5371)[0.0108](0.8787)[0.8788] (0.6472)[0.1209](0.9089)[0.9169]
−0.6 (0.6904)[0.0654](1.1364)[1.1383] (0.6286)[0.0036](1.0510)[1.0510] (0.6710)[0.0460](1.0818)[1.0828]
−0.3 (0.7488)[−0.0205](1.3616)[1.3618] (0.7127)[−0.0566](1.3102)[1.3114] (0.7570)[−0.0122](1.2826)[1.2826]
0.0 (0.9979)[−0.0021](1.7208)[1.7208] (0.9924)[−0.0076](1.7300)[1.7301] (0.9843)[−0.0157](1.6311)[1.6312]
0.3 (1.3588)[−0.0697](2.3514)[2.3525] (1.4002)[−0.0283](2.4762)[2.4764] (1.3683)[−0.0603](2.2376)[2.2384]
0.6 (2.2487)[−0.2513](3.8327)[3.8409] (2.4129)[−0.0871](4.3412)[4.3421] (2.3032)[−0.1968](3.6830)[3.6883]
0.9 (7.9484)[−2.0516](14.125)[14.273] (9.1202)[−0.8798](17.435)[17.457] (8.2548)[−1.7452](13.476)[13.588]

m ¼ 14; m ¼ 20; N ¼ 485

−0.9 (0.4541)[−0.0722](1.1198)[1.1221] (0.4928)[−0.0335](1.0320)[1.0326] (0.4934)[−0.0329](0.9253)[0.9259]
−0.6 (0.6735)[0.0485](1.3686)[1.3694] (0.6521)[0.0271](1.2722)[1.2725] (0.6550)[0.0300](1.0590)[1.0594]
−0.3 (0.7766)[0.0074](1.6621)[1.6621] (0.7352)[−0.0340](1.5977)[1.5981] (0.8281)[0.0588](1.3254)[1.3268]
0.0 (1.0455)[0.0455](1.9743)[1.9749] (1.0062)[0.0062](1.9803)[1.9803] (0.9899)[−0.0101](1.5811)[1.5812]
0.3 (1.3500)[−0.0785](2.7266)[2.7277] (1.3245)[−0.1041](2.8995)[2.9014] (1.3690)[−0.0596](2.2282)[2.2290]
0.6 (2.2405)[−0.2595](4.5546)[4.5620] (2.3083)[−0.1917](5.1855)[5.1890] (2.3677)[−0.1323](3.7946)[3.7969]
0.9 (8.3963)[−1.6037](17.399)[17.473] (9.4221)[−0.5779](22.578)[22.585] (8.8920)[−1.1080](15.193)[15.233]

m ¼ 20; m ¼ 28; N ¼ 945

−0.9 (0.5204)[−0.0059](0.7638)[0.7638] (0.5129)[−0.0134](0.6969)[0.6970] (0.5136)[−0.0127](0.6664)[0.6665]
−0.6 (0.6675)[0.0425](0.9379)[0.9389] (0.6339)[0.0089](0.8553)[0.8554] (0.6437)[0.0187](0.8319)[0.8321]
−0.3 (0.7681)[−0.0011](1.0509)[1.0509] (0.7348)[−0.0345](0.9988)[0.9994] (0.7654)[−0.0038](0.9627)[0.9627]
0.0 (0.9787)[−0.0213](1.3102)[1.3104] (0.9722)[−0.0278](1.3183)[1.3186] (1.0084)[0.0084](1.2270)[1.2271]
0.3 (1.2626)[−0.1659](1.8237)[1.8312] (1.2965)[−0.1320](1.9505)[1.9550] (1.3216)[−0.1069](1.7183)[1.7216]
0.6 (2.2583)[−0.2417](2.9123)[2.9223] (2.4962)[−0.0038](3.3585)[3.3585] (2.3318)[−0.1682](2.8320)[2.8370]
0.9 (8.2405)[−1.7595](10.521)[10.667] (10.063)[0.0634](13.813)[13.813] (8.9324)[−1.0676](10.441)[10.495]

(Mean)[Bias](Std.)[RMSE].
20 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21

Appendix C. Variable definitions and summary statistics

This section contains variable definitions and descriptive statistics for the data sets used in the empirical illustrations.
Table 9
Hedonic-housing price equation, N = 506.
Variables Definition Min Max Mean Std

Crime Per capita crime by town 0.0063 88.9762 3.6135 8.6015

Zoning Proportion of residential land zoned for lots over 25,000 ft2 0 100.0000 11.3636 23.3225
Industry Proportion of non-retail business acres per town 0.4600 27.7400 11.1368 6.8604
Charles river Charles River dummy variable 1 if tract bounds river 0 1 0.0692 0.2540
Nox Nitric oxides concentration parts per 10 million 0.3850 0.8710 0.5547 0.1159
Rooms Average number of rooms per dwelling 3.5610 8.7800 6.2846 0.7026
House age Proportion of owner-occupied units built prior to 1940 2.9000 100 68.5749 28.1489
Distance Weighted distances to five Boston employment centers 1.1296 12.1265 3.7950 2.1057
Access Index of accessibility to radial highways 1.0000 24.00 9.5494 8.7073
Tax rate Full-value property tax rate per $10,000 187.0000 711.0000 408.2372 168.5371
Pupil/teacher Pupil–teacher ratio by town 12.6000 22.0000 18.4555 2.1649
Black pop. 1000(Bk–0.63)2 where Bk is the proportion of blacks by town 0.3200 396.9000 356.6740 91.2949
Lower class Lower status of the population 1.7300 37.9700 12.6531 7.1411
House price Median value of owner occupied house price in $1000 5.0000 50.0000 22.5328 9.1971

Source: This data set is available in the Kelley Pace's Spatial Statistics Toolbox.

References
Abadir, Karim M., Magnus, Jan R., 2005. Matrix algebra. Econometric Exercises.Cambridge
University Press, New York9780521822893.
Albert, James H., Chib, Siddhartha, 1993. Bayesian analysis of binary and polychotomous
response data. J. Am. Stat. Assoc. 88.422, 669–679 (ISSN 01621459).
Anderson, James E., van Wincoop, Eric, 2003. Gravity with gravitas: a solution to the bor-
der puzzle. Am. Econ. Rev. 93.1, 170–192.
Anselin, Luc, 1982. A note on small sample properties of estimators in a first-order spatial
autoregressive model. Environ. Plan. A 14.8, 1023–1030.
Anselin, Luc, 1988. Spatial Econometrics: Methods and Models. Springer, New York.
Anselin, Luc, 2006. In: Mills, T.C., Patterson, K. (Eds.), Spatial Econometrics, vol. I. Palgrave
Macmillan, Basingstoke.
Arraiz, Irani, Drukker, David M., Kelejian, Harry H., Prucha, Ingmar R., 2010. A spatial Cliff-
ord-type model with heteroskedastic innovations: small and large sample results.
J. Reg. Sci. 50.2, 592–614.
Badinger, Harald, Egger, Peter, 2011. Estimation of higher-order spatial autoregressive
cross-section models with heteroskedastic disturbances. Pap. Reg. Sci. 90.1,
213–235 (ISSN 1435–5957).
Baltagi, Badi H., 2008. Econometric Analysis of Panel Data. John Wiley & Sons.
Behrens, Kristian, Ertur, Cem, Koch, Wilfried, 2012. ‘Dual’ gravity: using spatial economet-
rics to control for multilateral resistance. J. Appl. Econ. 27.5, 773–794.
Casella, George, George, Edward I., 1992. Explaining the Gibbs sampler. Am. Stat. 46.3,
167–174.
Chib, Siddhartha, 2001. Chapter 57 Markov chain Monte Carlo methods: computation and
inference. In: Heckman, J.J., Leamer, E. (Eds.), Handbook of Econometrics. Handbook
of Econometrics, vol. 5. Elsevier, pp. 3569–3649.
Chib, Siddhartha, Greenberg, Edward, 1995. Understanding the Metropolis–Hastings al-
gorithm. Am. Stat. 49.4, 327–335.
Doğan, Osman, Taşpınar, Süleyman, 2013a. GMM estimation of spatial autoregressive
models with autoregressive and heteroskedastic disturbances. Working Paper. City
University of New York Graduate Center. http://dx.doi.org/10.2139/ssrn.2227163.
Doğan, Osman, Taşpınar, Süleyman, 2013b. GMM estimation of spatial autoregressive
models with moving average disturbances. Reg. Sci. Urban Econ. 43.6, 903–926.
Drukker, David M., Egger, Peter, Prucha, Ingmar R., 2013. On two-step estimation of a spa-
tial autoregressive model with autoregressive disturbances and endogenous regres-
sors. Econ. Rev. 32.5-6, 686–733.
Elhorst, J. Paul, Lacombe, Donald J., Piras, Gianfranco, 2012. On model specification and
parameter space definitions in higher order spatial econometric models. Reg. Sci.
Urban Econ. 42.1-2, 211–220.
Feenstra, Robert C., 2004. Advanced International Trade. Princeton University Press,
Princeton, NJ.
Geweke, John, 1993. Bayesian treatment of the independent student-t linear model.
J. Appl. Econ. 8, S19–S40.
Geweke, John, Keane, Michael, 2001. Chapter 56 computationally intensive methods for
integration in econometrics. In: Heckman, J.J., Leamer, E. (Eds.), Handbook of econo-
metrics. Handbook of Econometrics, vol. 5. Elsevier, pp. 3463–3568.
Gilley, Otis W., Kelley Pace, R., 1996. On the Harrison and Rubinfeld data. J. Environ. Econ.
Manag. 31.3, 403–405.
Griffiths, William E., 2007. Heteroskedasticity. In: Baltagi, Badi H. (Ed.), A Companion to
Theoretical Econometrics. Blackwell Publishing Ltd, pp. 82–100.
Harrison, David, Rubinfeld, Daniel L., 1978. Hedonic housing prices and the demand for
clean air. J. Environ. Econ. Manag. 5.1, 81–102.
Hepple, Leslie W., 1979. Bayesian analysis of the linear model with spatial dependence. In:
Bartels, Cornelis P.A., Ketellapper, Ronald H. (Eds.), Exploratory and Explanatory Statis-
tical Analysis of Spatial Data. Martinus Nijhoff Publishing, Boston, Massachusetts.
Hepple, Leslie W., 1995a. Bayesian techniques in spatial and network econometrics: 1.
model comparison and posterior odds. Environ. Plan. 27.3, 247–469.
Hepple, Leslie W., 1995b. Bayesian techniques in spatial and network econometrics: 2.
computational methods and algorithms. Environ. Plan. 27.4, 615–644.
Hepple, Leslie W., 2002. Analytical simplifications for marginal distributions in
Bayesian spatial econometrics. School of Geographical Sciences, University of
Bristol, Working Papers Series. University of Bristol, School of Geographical
Sciences.
Hepple, Leslie W., 2003. Bayesian and maximum likelihood estimation of the linear model
with spatial moving average disturbances. School of Geographical Sciences, Universi-
ty of Bristol, Working Papers Series. University of Bristol, School of Geographical
Sciences.
Hoogerheide, Lennart F., van Dijk, Herman K., van Oest, Rutger D., 2009. Simulation
based Bayesian econometric inference: principles and some recent computation-
al advances. In: Belsley, David A., John Kontoghiorghes, Erricos (Eds.), Handbook
of Computational Econometrics. John Wiley & Sons, Ltd, Chichester, UK,
pp. 215–281.
Kakamu, Kazuhiko, Wago, Hajime, 2008. Small-sample properties of panel spatial
autoregressive models: comparison of the Bayesian and maximum likelihood
methods. Spat. Econ. Anal. 3.3, 305–319.
Kelejian, Harry H., Prucha, Ingmar R., 1998. A generalized spatial two-stage least squares
procedure for estimating a spatial autoregressive model with autoregressive distur-
bances. J. Real Estate Financ. Econ. 17.1, 1899–1926.
Kelejian, Harry H., Prucha, Ingmar R., 1999. A generalized moments estimator for the
autoregressive parameter in a spatial model. Int. Econ. Rev. 40.2, 509–533.
Kelejian, Harry H., Prucha, Ingmar R., 2007. Specification and Estimation of Spatial
Autoregressive Models with Autoregressive and Heteroskedastic Disturbances. De-
partment of Economics, University of Maryland (July).
Kelejian, Harry H., Prucha, Ingmar R., 2010. Specification and estimation of spatial
autoregressive models with autoregressive and heteroskedastic disturbances.
J. Econ. 157, 53–67.
Kelley Pace, R., Barry, Ronald, 1997. Quick computation of spatial autoregressive estima-
tors. Geogr. Anal. 290 (3), 232–247.
Koop, Gary, 2003. Bayesian Econometrics. John Wiley and Sons Ltd., West Sussex,
England.
Koop, Gary, Poirier, Dale J., Tobias, Justin L., 2007. Bayesian Econometric Methods.
Cambridge University Press, New York, USA.
Lee, Lung-fei, 2003. Best spatial two-stage least squares estimators for a spatial
autoregressive model with autoregressive disturbances. Econ. Rev. 22.4, 307–335.
Lee, Lung-fei, 2004. Asymptotic distributions of quasi-maximum likelihood estimators of
spatial econometrics models. Econometrica 72, 1899–1926.
Lee, Lung-fei, 2007a. GMM and 2SLS estimation of mixed regressive, spatial
autoregressive models. J. Econ. 137, 489–514.
Lee, Lung-fei, 2007b. The method of elimination and substitution in the GMM estimation
of mixed regressive, spatial autoregressive models. J. Econ. 140, 155–189.
Lee, Myoung-Jae, 2010. Micro-econometrics: Methods of Moments and Limited Depen-
dent Variables. Springer, New York, USA.
Lee, Lung-fei, Liu, Xiaodong, 2010. Efficient GMM estimation of high order spatial
autoregressive models with autoregressive disturbances. Econ. Theory 26.01,
187–230.
LeSage, James P., 1997. Bayesian estimation of spatial autoregressive models. Int. Reg. Sci.
Rev. 20.1-2, 113–129.
LeSage, James P., 1999. Spatial Econometrics. Regional Research Institute, West Virginia
University.
LeSage, James P., Pace, Robert K., 2009. Introduction to Spatial Econometrics Statistics: A
Series of Textbooks and Monographs. Chapman and Hall/CRC, London.
Lin, Xu, Lee, Lung-fei, 2010. GMM estimation of spatial autoregressive models with un-
known heteroskedasticity. J. Econ. 157.1, 34–52.
Liu, Chuanhai, Rubin, Donald B., 1995. Ml estimation of the t distribution using EM and its
extensions, ECM and ECME. Stat. Sin. 5, 19–39.
 O. Doğan, S. Taşpınar / Regional Science and Urban Economics 45 (2014) 1–21
Liu, Xiaodong, Lee, Lung-fei, Bollinger, Christopher R., 2010. An efficient GMM estimator
of spatial autoregressive models. J. Econ. 159.2, 303–319.
McCallum, John, 1995. National borders matter: Canada–U.S. regional trade patterns. Am.
Econ. Rev. 85.3, 615–623.
Newey, Whitney K., McFadden, Daniel, 1994. Large sample estimation and hypothesis
testing. Handb. Econ. IV, 2111–2245.
Oliveira, Victor De, Song, Joon Jin, 2008. Bayesian analysis of simultaneous autoregressive
models. English Sankhyā: The Indian Journal of Statistics, Series B (2008), 700 (2),
pp. 323–350.
Pace, Robert K., LeSage, James P., Zhu, Shuang, 2012. Spatial dependence in regres-
sors and its effect on performance of likelihood-based and instrumental vari-
able estimators. In: Terrell, Daniel Millimet Dek (Ed.), 30th Anniversary
21
Edition. Advances in Econometrics, vol. 30. Emerald Group Publishing Limited,
pp. 257–295.
Parent, Olivier, LeSage, James P., 2007. Bayesian model averaging for spatial econometric
models. University of Cincinnati, Economics Working Papers Series 2007–02. Univer-
sity of Cincinnati, Department of Economics.
Parent, Olivier, LeSage, James P., 2008. Using the variance structure of the conditional
autoregressive spatial specification to model knowledge spillovers. J. Appl. Econ.
23.2, 235–256.
Tierney, Luke, 1994. Markov chains for exploring posterior distributions. Ann. Stat. 22.4,
1701–1728.
White, Halbert G., 1980. A heteroskedasticity-consistent covariance matrix estimator a
direct test for heteroskedasticity. Econometrica 48, 817–838.