Spatial Econometric STAR Models: Lagrange Multiplier Tests and Monte Carlo Simulations

Size: px
Start display at page:

Download "Spatial Econometric STAR Models: Lagrange Multiplier Tests and Monte Carlo Simulations"

Transcription

1 Spatial Econometric STAR Models: Lagrange Multiplier Tests and Monte Carlo Simulations Valerien O. Pede Department of Agricultural Economics, Purdue University West Lafayette, IN 47907, USA, Raymond J.G.M. Florax Department of Agricultural Economics, Purdue University West Lafayette, IN 47907, USA, Department of Spatial Economics, VU University Amsterdam, The Netherlands Dayton M. Lambert Department of Agricultural Economics, University of Tennessee Knoxville, TN 37996, USA, Matthew T. Holt Department of Agricultural Economics, Purdue University West Lafayette, IN 47907, USA, Abstract This paper investigates non-linearity in spatial process models allowing for gradual regime-switching structures in the form of a smooth transition autoregressive process. Until now, applications of the smooth transition autoregressive (STAR) model have been largely confined to time series analysis. The paper focuses on extending the non-linear smooth transition perspective to spatial process models, in which spatial correlation is taken into account through the use of a weight matrix identifying the topology of the spatial system. We start by deriving a non-linearity test for a simple spatial STAR model, in which spatial dependence is only included in the transition function. Next, we propose a non-linearity test for a spatial STAR model that includes a spatially lagged dependent variable or spatially autocorrelated innovations. Monte Carlo simulations of the various test statistics are performed to examine test power and size. Results indicate that the individual and joint nonlinearity tests show power against traditional spatial lag and error autoregressive processes, and the individual and joint spatial dependence tests have power against STAR nonlinearity. These results complicate the design of an effective specification strategy. Keywords: spatial econometrics, non-linearity, autoregressive smooth transition JEL Classification: C1, C1, C51, O18, R11

2 1. Introduction Over the last decade, nonlinear time series modeling has gained considerable attention in the applied economics literature. Nonlinear models are sometimes more appropriate than linear models in describing economic processes. For instance, the presence of asymmetry and nonlinearities in business cycles can be incorporated in smooth transition autoregressive (STAR) models, which generally leads to a better representation of data generating processes (dgp s) compared to linear regression models. The STAR model is a member of the family of nonlinear models that exhibit regime-dependent or regimeswitching behavior. Smooth Transition Autoregressive (STAR) regressions model the nonlinear dynamics in time series by allowing the dgp to depend on one or multiple regimes that typically follow a first order Markov process. In general, STAR models relax the assumption that parameters associated with the data generating processes are fixed throughout the series, allowing for the endogenous determination of structural discontinuities across time. Regime-dependent or regime-switching STAR models are due to Goldfeld and Quant (1973). Recent innovations of the STAR model have modeled nonlinearities and asymmetric responses in industrial production (Terasvirta and Anderson, 199), the hog-corn cycle (Holt and Craig, 006), exchange rates (Baharumshah and Khim-Sen Liew, 006), interest rates (van Dijk and Franses, 000), and unemployment rates (Skalin and Terasvirta, 00). As implied by its name, the STAR framework allows model parameters to take on different values across regimes, following a potentially smooth transition process. Smooth transition regime switching models are familiar to the time series literature,

3 3 although the general approach has seen frequent use in the biometric sciences (Schabenberger and Pierce, 00). Surprisingly, STAR-type models have received little attention in the spatial econometrics literature. In the spatial econometrics literature, structural breaks across space are typically modeled using dummy variables allowing slopes and intercepts to vary over space (Anselin, 1988; Lambert and McNamara, 009), or semiparametric approaches including Geographically Weighted Regression (GWR) (Fortheringham et al., 00), spatial expansion regression (Cassetti, 197), or random coefficients (Anselin, 1988). Structural instability across space may be caused by different response functions or systematically varying parameters. The central assumption behind these approaches is that regression parameters vary over space due to localized or regional differences in consumer behavior, producer supply, or industry technology. Firms may employ different production technologies, skilled labor may be concentrated in certain locations, or benefits from industry information, demand-side, or supply-side spillovers may occur more frequently in or near agglomeration economies. Ad hoc classifications (e.g., metro versus non-metro, north versus south, or treatment versus control ) may also be sources of spatial heterogeneity. Measurement error due to differences in localized, unobserved factors such as cultural preferences, local knowledge or policy, customs, or social networks may cause structural breaks across space as well. The usual consequences of the statistical validity of models result when these issues are overlooked: inferior forecasts, biased coefficients, and compromised inference. Depending on the problem at hand, the above methods may be integrated into more complicated spatial process models simultaneously allowing for autoregressive lags and correlated error structures.

4 4 Whittle (1954) described a broad class of spatial process models where an endogenous variable is specified to depend on spatial interactions between cross sectional units plus a disturbance term. The interactions are modeled as a weighted average of nearby cross sectional units, and the endogenous variable comprising the interactions is usually referred to as a spatially lagged variable. The weights, grouped in a matrix identifying neighborhood connections, form the distinctive core of spatial process models. The model is termed a first order spatial autoregressive lag model (SAR[1]) (Anselin and Florax, 1995). Whittle s SAR(1) model was popularized and extended by Cliff and Ord (1973, 1981), who further developed models in which the disturbances followed a spatial autoregressive process. The general spatial process model containing a spatially lagged endogenous variable, spatially autoregressive disturbances, and exogenous variables is called a spatial autoregressive model with autoregressive (AR) disturbance of order (1,1) (ARAR) (Anselin, 1988; Kelejian and Prucha, 004, 008; Lee, 00, 007); y = ρw 1 y + Xβ + e, e = λw e + u, u ~ iid(0, Ω), where W 1 and W are (possibly identical) nonstochastic, positive definite, exogenous matrices defining relationships between spatial units, and E[uu ] = Ω. The reduced form version is y = A 1 Xβ + A 1 B 1 u; A = (I ρw 1 ), B = (I λw ) where ρ is the spatial lag regressive term and λ is the spatial error autoregressive term. Spatial processes have often been assumed to be linear, although it is much more likely that spatial dynamics exhibit nonlinear features in a way that is similar to time series models (De Graaff et al., 001). 1 1 The implied spatial spillovers actually follow a process of nonlinear decline over space, given the presence of the spatial multiplier in the spatial lag and the ARAR model (Anselin, 003; Pede et al., 009).

5 5 Spatial analogues of the time series STAR approach have been developed in a handful of studies. For instance, Lebreton (005) developed a spatial version of time series STAR model by incorporating spatial autocorrelation in the transition function and allowing for a smooth transition process. The approach proposed in this paper follows the same principle, but in addition incorporates traditional spatial process models in the STAR framework. A spatial analogue to the STAR model was also presented in Gress (004), Basile and Gress (005) and Basile (008), but these studies considered a nonparametric estimator whereas our approach is parametric. Recently, Dorfman et al. (009) developed a model that is quite similar to the STAR approach using a Bayesian perspective, but their approach models hierarchical rather than contagious autoregressive processes. This paper formally develops a series of tests for identifying nonlinear structural heterogeneity across space, allowing for gradual, endogenous regime switching behavior as a smooth transition autoregressive process in the context of traditional spatial process models. We start by deriving a nonlinearity test for a spatial ARAR(1,1)-STAR model. Next, we derive nonlinearity tests for three nested models: the spatial lag STAR (SAR-STAR), the spatial error STAR (SEM-STAR) and the ARAR model (ARAR- STAR). The tests are developed in the maximum likelihood framework focusing specifically on the Lagrange Multiplier variants. Monte Carlo simulations are used to investigate the finite sample performance of the various tests in terms of size and power.. Family of Spatial STAR Models We modify the spatial STAR concept from the time series perspective to spatial crosssectional processes. We start with a basic specification of the STAR model for spatial

6 6 variables in analogy with the model presented for time series (Terasvirta and Anderson, 199; Holt and Craig, 006; Van Dijk and Franses, 000): y = Xα + Xδ o G(, s γ, c) + ε, (1) where y is an N 1 spatial data series, X is an N k vector of explanatory variables, Gs (, γ, c) is a continuous, potentially smooth, real-valued transition function bounded between zero and one, s is the transition variable, γ and c are (respectively) the slope and location parameters, o is the Hadamard product indicating element-by-element multiplication, and ε an N 1 vector of innovations assumed to be independently and identically distributed with constant variance, or alternatively to exhibit a spatially autoregressive pattern. Note that the interaction between the transition function and the exogenous variables allows for parameter variation across space and between regimes identified through G. In analogy with time series STAR models, a possible candidate for the transition variable could be the spatial lag of the dependent variable y or an explanatory variable x. In this paper we only consider the latter. We therefore define the spatial lag of an independent variable x as Wx, where W represents an exogenously defined weight matrix. 3 The spatial weights matrix is usually a Boolean matrix, with each element taking the value 1 when regions are neighbors and 0 when they are not. The weight matrix is typically row-standardized, such that the cross-product Wx represents the average value We did not consider the spatially lagged dependent variable as the transition variable because of additional complications involved in obtaining a reduced form for the STAR model. 3 Wx is an N 1 vector of observations, and simply represents the cross-product of the weight matrix W and the independent variable x.

7 7 of x of neighboring locations. 4 We identify spatial regimes using the logistic transition function 5 given as: 1 GWx (, γ, c) =. 1 + exp ( γ( Wx c) σwx ) () Figure 1 depicts an example of how the transition function G operates. Note that two distinct regimes emerge when γ = 100, whereas there are no regimes identified when γ = 0. Observations located along the curve of the function are transitional, not belonging to either regime as denoted by the observation cluster around the (0, 1) values of the transition function along the y-axis. The parameter c functions as a location parameter. That is, the inflection of the transition function is centered on c. The general spatial ARAR STAR model with spatial autocorrelation in the transition function, a spatially lagged dependent variable, and spatially autoregressive errors is: y = ρwy+ Xα + G( Wx, γ, c) o Xδ + ε, ε = λwε + μ, (3) where ρ and λ are scalar spatial autoregressive parameters, μ are independent and identically distributed errors, and the other symbols are as defined before. Three different spatial STAR models are nested within this general ARAR model, depending on the value of the spatial autoregressive parameters. Restricting λ = 0 leads to the spatial lag STAR model; ρ = 0 the spatial error STAR model; and λ = ρ = 0 the simple spatial STAR model in equation (1). Each of these models can be estimated with maximum likelihood 4 Row-standardization implies dividing each element in the row of a matrix by the sum of the elements in the row. Alternative standardization procedures are also possible, but are not treated in detail here (Bivand et al., 008). 5 Other functions are admissible (Schabenberger and Pierce, 00).

8 8 utilizing nonlinear optimization routines. The null hypotheses of parameter stability across regions and/or specific spatial autoregressive processes can be tested with the spatial ARAR STAR, the spatial lag, and the spatial error STAR models as unrestricted models. 3. ML Estimation of Spatial STAR Models and Lagrange Multiplier Tests In this section, we describe maximum likelihood estimation procedures for the family of spatial STAR models. Individual and joint Lagrange Multiplier tests for nonlinear regime-splitting and/or specific spatial autoregressive processes are developed. 3.1 Spatial ARAR STAR Model Consider first the ARAR specification: 1 y = ρwy+ Xβ + ( I λw) μ, (4) where y is an N 1 spatial data series, X an N k matrix of explanatory variables, μ a vector of innovations, λ and ρ are spatial parameters and W an N N spatial weight matrix. A spatial ARAR STAR model can be constructed from (4) by adding a set of coefficients δ for a second regime, and the explanatory variables interacted with a transition function: 1 y= ρwy+ Xβ+ Go Xδ + ( I λw) μ, (5) where G is the logistic transition function defined in (), and all other variables are as before. As in time series analysis, testing directly for nonlinear regime-switching in each model results in unidentified nuisance parameters under the null hypothesis of no spatial

9 9 regimes. This problem can be solved by considering an appropriate Taylor series approximation of the transition function G( Wx, γ, c) around γ = 0, as suggested by Luukkonen et al. (1988). Considering the first-order Taylor series expansion of G(Wx,γ,c) around γ = 0, the spatial ARAR STAR model is approximated as: y ρwy+ Xα + Xδ o η + ηwx + I λw μ (6) 1 ( 0 1 ) ( ), which, after collecting terms, can be rewritten in reduced form as: ( o ) y ( I W) X + X Wx + ( I W) 1 1 ρ ξ ϕ λ μ ρ β + λ μ (7) 1 1 ( I W) Z ( I W), where Z is the N k matrix defined as Z = ( X, Wx o X ), and β is a k 1 vector defined by the original parameters as = ( + ) 0 ξ ηδ α and ϕ = ( ηδ 1 ). Assuming normally distributed errors that are independently and identically distributed with mean 0 and variance σ, the log-likelihood function for the spatial ARAR STAR model is: N N 1 Ly ( ; β, ρλσ,, ) = ln( π) ln( σ ) + lnw + ln A WR εww, R Rε σ (8) where W = I ρw, A WR = I λw and ε = WA y Zβ. are derived as: Lagrange Multiplier tests for spatial dependence and nonlinear regime switching

10 10 LM LM φ ewy ewe = σ σ + LM ( WZβ) M ( WZβ) σ λ= ρ= 0 λ= 0 = epe σ =0,, [ARAR, spatial dependence] (9) [STAR, nonlinear regime switching] (10) epwx ζ ewy ewe LM = σ σ σ + LM + LM ( WXζ ) M ( WXζ ) σ λ= ρ= φ= 0 λ= 0 φ= 0. [STAR ARAR] (11) where e represents the residuals of the model under the null hypothesis estimated using an appropriate generalized least squares (GLS) estimator, σ = ee N, LM λ = 0 is the usual Lagrange Multiplier test for spatial error dependence (Anselin, 1988), P is a projection matrix 1 Z ( Z Z) Z, and M = I P. The tests are asymptotically distributed with, k and k+ degrees of freedom, respectively Spatial Lag STAR Model To derive the spatial lag STAR model we consider first the basic spatial autoregressive lag specification: y = ρwy+ Xβ + μ, (1) with all symbols as previously defined. A spatial lag STAR model is constructed from (1) by adding a set of coefficients δ for a second regime, interacted with a transition function: 6 The mathematical derivation of the tests is provided in Pede et al. (009).

11 11 y= ρwy+ Xβ + Go Xδ + μ, (13) where G is the logistic transition function defined in (), and all other variables are as defined before. Assuming the errors are independent and identically distributed following a normal distribution, the log-likelihood function for the spatial lag STAR model is: N N 1 Ly ( ; β, ρσ, ) = ln( π) ln( σ ) + ln I ρw μμ, σ (14) where μ = ( I ρw) y f( X; θ ), f is the function defined in equation (1) with θ β, δ, γ,, and σ = μμ / N. parameters = ( c) Considering the first-order expansion of G around γ = 0, the Taylor series approximation for the spatial lag STAR model is: y ρwy+ Xα+ Xδo( η + ηwx) + μ, (15) 0 1 which, after collecting terms, can be rewritten in reduced form as: ( o ) [ ] 1 y ( I ρw) Xξ + X Wx ϕ+ μ + 1 ( I ρw) Zβ μ. (16) Noting that the LM test in (10) would also be used to test for detecting nonlinear regime switching in the ARAR model, individual and joint LM tests for the absence of a spatial lag and spatial lag with nonlinear regime splitting are derived as: LM ρ= 0 1 ewy = σ NJ, (17)

12 1 LM epwz β ewy σ σ epwze NJ σ ρ= ϕ=0 = +, (18) where e are the residuals of the model under the null hypothesis estimated using an adequate GLS estimator, and the term NJ is defined as: ( WZβ) M ( WZβ) NJ = tr ( W + W W ) +. σ (19) respectively. 7 The tests are asymptotically distributed with 1 and k+1 degrees of freedom, 3.3 Spatial Error STAR Model We motivate the STAR model with autocorrelated innovations by considering the typical spatial error autoregressive specification: 1 y = Xβ + ( I λw) μ. (0) Unlike the spatial lag model, the marginal effects associated with a (non-uniform) change in (one of) the exogenous variables are stationary across space in the case of the spatial autoregressive error model. A spatial error STAR model is constructed from (0) by adding a set of coefficients δ for a second regime, interacted with a transition function: 1 y = Xβ + Go Xδ + ( I λw) μ. (1) For estimation we once more apply standard maximum likelihood principles for spatial 7 Details about the derivations of the tests are given in Pede et al. (009).

13 13 error process models. Specifically, if the errors are independently and identically distributed following a normal distribution, the log-likelihood function for the spatial error STAR model is: N N 1 L( θ, ρλσ,, ) = ln( π) ln( σ ) + ln I λw μμ, () σ ( ) where μ ( I λw ) y f ( X; θ) =, f is the function in equation (1) with parameters (,,,c) θ = β δ γ, and σ = μμ / N. Since there is no analytical expression forθ, optimization cannot be based on a one-shot maximization of the concentrated likelihood. An iterative feasible generalized least squares approach is, however, appropriate as long as a consistent estimate of λ is attained through optimization of () given μ. In order to construct the appropriate LM tests, we again apply the first-order Taylor series approximation of the transition function, which for the spatial error model is: 1 ( o ) ( ) y Xξ + X Wx ϕ+ I λw μ 1 Zβ + ( I λw) μ. (3) Lagrange Multiplier tests for spatially autoregressive errors and nonlinear regime switching are derived following the general principles outlined in Anselin (1988): LM = LM + LM (4) λ= ϕ= 0 λ= 0 ϕ= 0.

14 14 The tests are asymptotically distributed with 1, k and k+1 degrees of freedom, respectively. 8 The family of LM tests is summarized in Table Monte Carlo Experimental Design Although the asymptotic properties of the derived LM tests are well-known, it is of crucial interest to investigate their performance in small samples in terms of size and power. The performance of the LM tests are therefore examined under various data generating assumptions. The experimental design follows standard practice for Monte Carlo simulations typically used in a spatial econometric context (Anselin and Rey, 1991; Florax and Folmer, 199, 1994, Anselin and Florax, 1995; Anselin et al., 1996; Anselin and Griffith, 1998). 9 We first generate a random spatial pattern where (x, y) coordinate pairs are drawn from a uniform (0, 1) distribution. The sample sizes we investigate are N = 5, 49, 100 and 400. The data generating process is: ( o ) y= I W X + G X + I W 1 1 ( ρ ) β δ ( λ ) μ, (5) where X is an N 3 matrix of explanatory variables consisting of a constant term and two variables drawn from a (0, 1) uniform distribution, μ is a vector of innovations assumed to have a standard normal distribution, and the parameter vectors β and δ are fixed to unity. Assuming that one of these explanatory variables (say x 1 ) governs the regime transition, the transition variable is generated as Wx 1, the spatial lag of x 1. We use 500 Monte Carlo replications, looping over different values for the spatial autoregressive parameters ρ and λ, and the spatial nonlinearity parameters γ and c. Six values of the 8 Details about the derivation of the tests are given in Pede et al. (009). 9 An extensive overview of the available simulation studies is provided in Florax and De Graaff (004).

15 15 smoothing parameter γ are considered; 0, 1,, 3, 5 and 10. The values for the location parameter c are: 0, the mean of Wx, and one standard deviation above and below the mean of Wx. Seven parameter values are used for the spatiall autoregressive parameters ρ and λ; 0, 0.1, 0., 0.4, 0.6, 0.8 and Three weight matrices were constructed to identify different neighborhood structures, given the random set of spatial coordinates of size N. The first weight matrix is based on the queen contiguity criterion, where all cells sharing a common edge or vertex are considered neighbors. The second neighborhood structure is based on a k- nearest neighbor (KNN) definition. We use the square root of the number of observations in the sample to define the number of nearest neighbors. The weight attributed to each observation is the inverse of the number of nearest neighbors. The third neighborhood definition also follows the KNN criterion, but the weight attributed to each observation is the inverse of the Euclidian distance between neighbors. All weight matrices were rowstandardized such that the sum of each row equaled one. Following the ARAR STAR specification described in equation (5), N observations on the dependent variable were generated from the explanatory variables, the spatial parameters, the transition variable, the smoothing and location parameters, and the standard normal random error vector μ. The matrix of variables Z = ( X, Wx o X ) corresponding with the first-order Taylor expansion of the transition function G was maintained fixed in the 500 replications. The model under the null hypothesis is a classical linear regression model with no spatial dependence: y = Xξ + μ. (6)

16 16 For each replication, the LM tests were computed and compared to their asymptotic critical value at α = We report the proportion that the null hypothesis was rejected, which is defined as the number of times the calculated value of a test statistic exceeds its asymptotic critical value, scaled by the total number of replications. Table 1 summarizes the LM tests evaluated during the Monte Carlo simulation. The Monte Carlo standard errors (MCSE) for a nominal significance level p with M replications are estimated as: MCSE = p(1 p). M (7) Using the 0.05 critical value for p plus or minus two times MCSE and considering 500 replications implies that the 95% confidence interval centered on p = 0.05 includes rejection frequencies between and The power and size of the respective tests are examined, and the simulation results for the Lagrange Multiplier LAG, ERR and ERR+LAG tests are also compared to those obtained in previous simulation studies (Anselin and Rey, 1991; Anselin and Florax, 1995; Anselin et al., 1996) Monte Carlo Simulation Results 5.1 Empirical Size of the Tests To determine the size of the respective tests, we use the simple linear model with no spatial dependence as the restricted model. The rejection frequencies for the null hypotheses of the different tests are obtained from the simulation output for the case 10 R code for the Monte Carlo simulations is presented in Pede et al. (009).

17 17 where the parameters ρ, λ, γ and c are set equal to zero. The rejection frequency of this general null hypothesis, when it is true, is reported in Table for the three weight matrices used in the simulation. It is interesting to observe that all tests except the ERR+LAG test yield rejection frequencies within the 95% confidence interval for the four sample sizes used in the simulation. These results, to the extent that they overlap, correspond to earlier findings reported in Anselin and Florax (1995) and Anselin et al. (1996). 11 The NLIN test overrejects the null hypothesis for N = 100 with the queen weight matrix, but behaves well in all other cases. It should be noted that the empirical size of the tests varies across different weight matrices. Anselin and Rey (1991) and Anselin and Florax (1995) also noted the influence of the choice of weight matrix on the empirical size of the tests. A priori, the influence of the specification of the weight matrix on empirical size may appear counterintuitive because the null hypothesis assumes no spatial dependence, but the weights matrix is obviously used in the definition of the various tests. The ERR+LAG test under-rejects the null hypothesis for all three weights matrices and samples sizes considered. Anselin and Florax (1995) observed similar results for the ERR+LAG in the case of a misspecification in the form of lognormal errors. They stressed that an underrejection of the null hypothesis when no spatial dependence is present does not have any consequences, since the standard estimation results are interpreted as they should be. 11 Anselin and Florax (1995) and Anselin et al. (1996) used 5,000 replications and p = 0.05, which corresponds to rejection frequencies between and for the 95% confidence interval and they observe correct sizes for these tests.

18 18 5. Power of the Tests In this section we discuss the statistical power of the tests. The power of a test is defined as the probability of rejecting the null hypothesis when the null hypothesis is false. The small sample power of the various tests is summarized in Figures through Power against Spatial Autoregressive Processes We first describe the power of the various LM tests against spatial autoregressive processes, and then discuss the small sample power of the tests against nonlinear spatial regimes. With respect to the spatial autoregressive processes, we discuss the spatial error, the spatial lag, and the spatial ARAR data generating processes, successively. In the case of spatial autoregressive error dependence, the small sample characteristics are obtained from the data generating process in which the parameters ρ, γ and c are set to zero. The power functions of the seven tests are shown in Figure for N = 5 and 400 using the queen weight matrix, and in Figure 3 for N = 400 using the KNN(1/N) and KNN(1/d ij ) weight matrices. For all tests the rejection frequencies increase as the magnitude of the spatial autoregressive parameter and sample size increase. In general, the ERR test has the highest power, followed by the LAG test. It should be noted that all tests perform rather poorly in the smallest sample, achieving the 95% rejection mark only for λ > 0.9. For a reasonably large sample size (N = 400) the power of the tests is comparable, but slightly higher for the weights matrices based on the queen and the KNN(1/d ij ) specification as compared to the KNN(1/N) version. To the extent that the setup of the simulations is comparable, these findings are consistent with the simulation results of Anselin and Rey (1991) and Anselin and Florax (1995). 1 For completeness, the rejection frequencies for various data generating processes are listed in Tables 3 through 6 of Pede et al. (009).

19 19 The most noteworthy conclusion from the rejection patterns presented in the respective figures is that all tests have remarkably high power against spatially autoregressive errors, even although most of the tests have a different alternative hypothesis. This is partly due to the similarity between the spatial error and the spatial lag process (Anselin et al., 1996). The NLIN test is least affected, although its power is substantially larger than the nominal significance level associated with the size of the test. As with the LAG and ERR, test this is most likely due to the similarity in the underlying spatial processes, especially since Wx is used as the transition variable. The small sample characteristics in the case where the spatial lag model is the true underlying data generating process are very similar to those described above. Figures 4 and 5 document the simulation outcomes. In accordance with previous simulation work (Anselin and Rey, 1991; Anselin and Florax, 1995) the LAG test appears most powerful. The NLIN test has substantial power against an erroneously omitted spatially lagged dependent variable, and in fact this is more pronounced than in the case of the spatial error model being the true model. The latter result is likely due to the fact that the spatial lag term Wy implicitly includes the transition variable Wx. Figures 6 through 1 provide three-dimensional power functions against the spatial ARAR model for various values of ρ and λ. 13 We again encounter the general pattern identified above. All tests have power against the ARAR process, with the NLIN test s power being the lowest. One should note, however, that the extent to which the NLIN test wrongfully points to the STAR model as the correct alternative is substantial. 13 The surfaces are generated using 36 data points corresponding to the nonzero values for the spatial autoregressive parameters.

20 0 5.. Power against Spatial STAR Nonlinearity The power against a spatial STAR process is obtained by setting the parameters ρ and λ equal to zero and extracting the corresponding rejection frequencies from the simulation output. 14 The power functions of the seven tests against spatial STAR nonlinearity are shown in Figure 13 for N = 5 and 400 using the queen weight matrix, and in Figure 14 for N = 400 using the KNN weight matrices. In each figure, the location parameter is set equal to the mean of Wx. As expected, the individual NLIN test shows the highest power against the null, followed by the joint ERR+NLIN, LAG+NLIN and ERR+LAG+NLIN tests, respectively, in the smallest sample. It is interesting to see that the ERR, LAG and ERR+LAG tests all have very little power against the null in the smallest sample. This is as expected, because no spatial dependence is present. However, in the largest sample, the ERR, LAG and ERR+LAG tests do show considerable power, even although in terms of true parameters ρ = λ = 0. For instance, when N = 400 with a KNN(1/N) weight matrix the power function of all seven tests are almost indistinguishable (see Figure 14). In general, the individual and joint tests for spatial nonlinearity perform well in the largest sample as well as in moderately sized samples, achieving the 95% rejection mark for values of γ greater than 1. In the smallest sample the 95% rejection mark was reached for values of γ greater than 5. We also investigated the power of the tests against spatial STAR nonlinearity in the case where the location parameter is set equal to the values of 0, one standard deviation above the mean of Wx, and one standard deviation below the mean of Wx. The power functions of the seven tests against spatial STAR nonlinearity for various 14 For completeness a numerical overview of the rejection frequencies against STAR nonlinearity are reported in Table 4 of Pede et al. (009).

21 1 configurations of the true parameters, sample sizes and weights matrices are shown in Figures 15 and 16. Figure 17 shows that for a very small sample (N = 5) we again observe the ideal situation that the individual and joint nonlinearity tests have power, whereas the individual and joint spatial dependence tests do not. The power of the NLIN tests is however rather low, and even deteriorating for higher values of γ in the case where the location parameter is set to one standard deviation above the mean of the transition variable. In Figure 16 we observe that the power increases with a larger sample size (N = 400). However, this is the case for the individual and joint tests for spatial dependence processes as well, although in reality ρ = λ = 0. The behaviors described above is also observed when the spatial error or the spatial lag model is taken as the restrictive model. 15 In sum, we are left with a situation that warrants further investigation. All tests have a reasonable size in the case where the non-spatial model is taken as the restricted model. The power of the tests is also quite satisfactory, especially in relatively larger samples (for instance, N = 400). However, the individual and joint nonlinear regime switching tests show power against the traditional spatial autoregressive processes (error and lag), and the individual and joint spatial dependence tests have power against STAR nonlinearity, operationalized with a spatially lagged exogenous variable as the transition variable. Given this behavior, it remains to be seen whether a well-performing specification strategy can be designed. Lundbergh et al. (003) found similar results with the time series STAR. They studied time-varying smooth transition (TV-STAR) models, which describe simultaneously nonlinearity and structural change by using two modeling strategies: the specific-to-general-to-specific procedure which consists of testing for 15 For completeness, Tables 5 and 6 in Pede et al. (009) report the associated rejection frequencies.

22 nonlinearity and/or structural change, and the specific-to-general procedure which consists of specifying a model with nonlinearity and/or time-varying properties. Their Monte Carlo results also suggest that neither of the two procedures dominates the other. Moreover, their empirical application on U.S. macroeconomic time series reveals the merit of both modeling strategies. Given the complication in designing an effective specification strategy, further investigation may be needed. In empirical applications, the consistency in model specification could be verified using both approaches. That is, investigate nonlinear structural breaks across space first and then decide on the spatial process, or proceed the other way around. 5. Conclusion In this paper we developed a new method to evaluate and incorporate nonlinear spatial dynamics and regime switching behavior into conventional spatial lag and error process models, based on smooth transition autoregressive models developed in time series research. The spatial STAR model belongs to the family of spatial regression models allowing for parameter variation across space. In particular, the STAR model allows for parameter variation across spatial regimes, where the membership of the spatial regimes is determined endogenously. The spatial STAR model is relatively easy to estimate and substantially more parsimonious than previous spatial regression methods allowing for spatial parameter variation. Moreover, the STAR model proposed here provides the opportunity to incorporate traditional spatial process models such as spatially

23 3 autoregressive errors, a model with a spatially lagged dependent variable, or a combination of lag and error processes in the ARAR model. Four types of spatial STAR models were developed in this study: the basic spatial STAR model, and the spatial error (SEM-STAR), the spatial lag (SAR-STAR) and the spatial ARAR STAR models. Maximum likelihood estimation procedures for each model were presented, and a series of new Lagrange Multiplier tests was developed to investigate nonlinear spatial regime switching and/or spatial dependence. Monte Carlo simulations of these tests reveal that all tests perform well with respect to size. Each tests produce rejection frequencies within a 95% confidence interval under the null hypothesis of an aspatial linear model. The tests also show relatively good power against spatial dependence (spatial error and spatial lag processes) and STAR-type nonlinearity. Almost invariably, the test that was designed for a specific alternative has the highest power. Unfortunately however, the spatial processes associated with the different tests appear to be rather similar in nature. As a result, the individual and joint nonlinear regime tests have power against the traditional spatial error and/or lag processes, and the spatial dependence tests have power against spatial STAR-type nonlinearity. This obviously complicates the design of an effective specification strategy for cross-sectional models with spatial autocorrelation and endogenous regime-switching behavior. The design and investigation of the performance of feasible alternative specification strategies is left for future research, but it is foreseeable that a specific-to-general strategy based on the estimation of the non-spatial model and a subsequent choice for an unrestricted model on the basis of p-values of the different LM tests may very well be effective.

24 4 REFERENCES Anselin, L. (1988). Spatial Econometrics: Methods and Models Kluwer Academic Publishers, Dordrecht. Anselin, L. and Rey, S. (1991). Properties of Tests for Spatial Dependence in Linear Regression Models Geographical Analysis 3: Anselin, L. and Florax, R.J.G.M. (1995). Small Sample Properties of Tests for Spatial Dependence in Regression Models: some Further Results in L. Anselin, R.J.G.M. Florax (Eds), New Directions in Spatial Econometrics, pp 1 74, Springer, Berlin. Anselin L., Bera A.K., Florax R.J.G.M. and Yoon, M.J. (1996). Simple Diagnostic Tests for Spatial Dependence Regional Science and Urban Economics 6: Anselin, L. (006). Spatial Regression Working Paper, Spatial Analysis Laboratory, Department of Geography and National Center for Supercomputing Applications, Urbana-Champaign. Anselin, L. (003). Spatial Externalities, Spatial Multipliers and Spatial Econometrics International Regional Science Review 6: Brunsdon, C., Fotheringham, A. and Charlton, M. (1996). Geographically Weighted Regression: A Method for Exploring Spatial Non-Stationarity Geographical Analysis 8: Burridge, P. (1980). On the Cliff-Ord Test for Spatial Correlation Journal of the Royal Statistical Society B 4: Cho, S.H., Lambert, D.M. and Chen, Z. (009). Geographically Weighted Regression Bandwidth Selection and Spatial Autocorrelation: An Empirical Example using Chinese Agricultural Data Applied Economic Letters 1 6. Cho, S., Lambert, D.M., Roberts, R.K. and Kim, S. (009) Demand for Open Space and Urban Sprawl: The Case of Knox County, Tennessee in A. Páez, J. Le Gallo, R. Buliung, and S. Dall Erba (Eds), Progress in Spatial Analysis: Theory and Computation, and Thematic Applications, Springer, Berlin. Cliff, A.D. and Ord, J.K. (1981). Spatial Processes: Models and Applications London: Pion.

25 5 Ertur, C. and Le Gallo, J. (003). Exploratory Spatial Data Analysis of the Distribution of Regional per Capita GDP in Europe, Papers in Regional Science 8: Florax, R. J. G. M., and Nijkamp, P. (005). Misspecification in Linear Spatial Regression Models Encyclopedia of Social Measurement : , Elsevier, Amsterdam. Florax, R.J.G.M., Folmer, H. and Rey, S.J.(003). Specification Searches in Spatial Econometrics: The Relevance of Hendry s Methodology Regional Science and Urban Economics 33: Florax, R. J. G. M., and De Graaff, T. (004). The Performance of Diagnostic Tests for Spatial Dependence in Linear Regression Models: A Meta-Analysis of Simulation Studies in L. Anselin, R. J. G. M. Florax, and S. Rey (Eds), Advances in Spatial Econometrics: Methodology, Tools and Application, pp 9 65, Springer, Heidelberg. Fotheringham, A.S., Brunsdon, C. and Charlton, M. (1998). Geographically Weighted Regression: a Natural Evolution of the Expansion Method for Spatial Data Analysis Environment and Planning A 30: Fotheringham, A.S., Brunsdon, C. and Charlton, M.E. (00). Geographically Weighted Regression: The Analysis of Spatially Varying Relationships Chichester: Wiley. Gress, B. (004). Semiparametric Spatial Autocovariance Models University of California at Riverside PhD Dissertation. Greunz, L. (004). Industrial Structure and Innovation: Evidence from European Regions Journal of Evolutionary Economics 14: Holt, T.M. and Craig, L. A. (006). Nonlinear Dynamic and Structural Change in the U.S. Hog-Corn Cycle: A Time-Varying STAR Approach American Journal of Agricultural Economics 88: Kelejian, H.H. and Prucha, I.R. (1999). A Generalized Moments Estimator for the Autoregressive Parameter in a Spatial Model International Economic Review 40:

26 6 Kelejian, H.H., Prucha, I.R. and Yuzefovich, Y. (004). Instrumental Variable Estimation of a Spatial Autoregressive Model with Autoregressive Disturbances: Large and Small Sample Results in J. LeSage and R.K. Pace (Eds), Advances in Econometrics, pp , Elsevier, New York. Kelejian, H.H., Prucha, I.R. and Drukker, A. (008). A Spatial Cliff Ord Type Model with Heterokedastcity Innovations: Small and Large Samples Results Department of Economics, University of Maryland, CESIFO Working Paper No 485. Lambert, D.M., K. T. McNamara. (009). Location Determinants of Food Manufacturers in the U.S., : Are Nonmetropolitan Counties Competitive? Agricultural Economics, 40(6): Lebreton, M. (005). The NCSTAR Model as an Alternative to the GWR Model Physica A 355: LeSage, J. P. (004). A Family of Geographically Weighted Regression Models in: L. Anselin, R. J.G.M Florax, S.J. Rey (Eds.), Advances in Spatial Econometrics: Methodology, Tools and Applications, pp 41 64, Springer, Berlin. LeSage, J. P. and Pace, K. (009). Introduction to Spatial Econometrics Taylor & Francis, Group Boca Raton: CRC Press. Lundbergh, S., Terasvirta, T. and van Dijk, D. (003). Time-Varying Smooth Transition Autoregressive Models Journal of Business and Economic Statistics 1: Luukkonen, R., Saikkonen, P. and Terasvirta, T. (1988). Testing Linearity against Smooth Transition Autoregressive Models. Biometrika 75: Pede, O. V., Florax, R.J.G.M. and Holt, M.T. (009). A Spatial Econometric STAR Model with an Application to U.S. County Economic Growth, Working Paper 09-03, Department of Agricultural Economics, Purdue University. Skalin, J. and Terasvirta, T. (00). Modeling Asymmetries and Moving Equilibria in Unemployment Rates Macroeconomic Dynamics 6:0 41. Terasvirta, T. and Anderson, H.M. (199). Characterizing Nonlinearities in Business Cycles Using Smooth Transition Autoregressive Models Journal of Applied Econometrics 7:

27 7 Van Dijk, D. and Franses, P.H. (000). Nonlinear Error-Correction Models for Interest Rates in the Netherlands in W.A. Barnett, D.F. Hendry, S. Hylleberg, T. Teräsvirta, D. Tjøstheim, and A. Würtz (Eds) Nonlinear Econometric Modelling, Proceedings of the 6th EC Meeting, pp 03 7, Cambridge University Press, Cambridge. Van Dijk, D., Terasvirta, T and Franses, P.H. (00). Smooth Transition Autoregressive Models: A Survey of Recent Developments Econometrics Review 1:1 47.

28 8 Table 1: Lagrange Multiplier Tests for Spatial Dependence and/or Nonlinearity. Test ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Test statistic and asymptotic distribution LM 1 ewe = T σ λ = 0 1 ewy σ LM ρ= 0 = NJ LM =0 = epe ϕ ~ χ (3) σ ~ χ (1), ~ (1) χ λ= ϕ= = λ= + ϕ= χ LM 0 LM 0 LM 0 ~ (4) LM LM epwz β ewy ( WZβ) M ( WZβ) ( + ) T= tr W WW σ σ ρ= ϕ= 0 = + T + σ ewy ewe = σ σ + LM ( WZβ) M ( WZβ) σ epwze σ λ= ρ= 0 λ= 0 ~ (4) χ ~ χ () epwx ζ ewy ewe LM = σ σ σ + LM + LM ( WXζ ) M ( WXζ ) σ λ= ρ= φ= 0 λ= 0 ϕ= 0 ~ χ (5)

29 9 Table : Empirical Size of the LM Tests. Queen contiguity KNN (1/N) KNN (1/d ij ) Tests N = 5 N = 49 N = 100 N = 400 N = 5 N = 49 N = 100 N = 400 N = 5 N = 49 N = 100 N = 400 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Note: The 95% confidence interval for the rejection proportion (prop) with 500 replications is < prop <

30 Figure 1: Example of the transition function G(Wx; γ, c), and different levels of the smoothing parameter, γ. Note that two distinct regimes emerge when γ = 100, whereas there are no regimes identified when γ = 0. The parameter c functions as a location parameter. That is, the inflection of the transition function is centered on c. 30

31 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Lambda ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Lambda Figure : Power against Spatial Error Model, N = 5 (top) and 400 (bottom), Queen.

32 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Lambda ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Lambda Figure 3: Power against Spatial Error Model, N = 400, KNN(1/N) (top) and KNN(1/d ij ) (bottom).

33 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Rho ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Rho Figure 4: Power against Spatial Lag Model, N = 5 (top) and 400 (bottom), Queen.

34 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Rho ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Rho Figure 5: Power against Spatial Lag Model, N = 400, KNN(1/N) (top) and KNN(1/d ij ) (bottom).

35 Rho Lambda Figure 6: Power of ERR against ARAR Process, N = 100, Queen Rho Lambda Figure 7: Power of LAG against ARAR Process, N = 100, Queen.

36 Rho Lambda Figure 8: Power of NLIN against ARAR Process, N = 100, Queen Rho Lambda Figure 9: Power of ERR+NLIN against ARAR Process, N = 100, Queen.

37 Rho Lambda Figure 10: Power of LAG+NLIN against ARAR Process, N = 100, Queen Rho Lambda Figure 11: Power of ERR+LAG against ARAR Process, N = 100, Queen.

38 Rho Lambda Figure 1: Power of ERR+LAG+NLIN against ARAR Process, N = 100, Queen.

39 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma Figure 3: Power against Spatial STAR Model, N = 5 (top) and 400 (bottom), with c equal to the Mean of Wx, and the Queen Matrix.

40 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma Figure 34: Power against Spatial STAR Model, N = 400, c equal to the mean of Wx, KNN(1/N) (top) and KNN(1/d ij ) (bottom).

41 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma Figure 15: Power against Spatial STAR Model, N = 5, c equal to Mean Wx+SD (top), Mean Wx SD (bottom), Queen Matrix.

42 ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma ERR LAG NLIN ERR+NLIN LAG+NLIN ERR+LAG ERR+LAG+NLIN Gamma Figure 46: Power against Spatial STAR Model, N = 400, c equal to Mean Wx+SD (top), Mean Wx SD (bottom), KNN(1/N).

Testing Random Effects in Two-Way Spatial Panel Data Models

Testing Random Effects in Two-Way Spatial Panel Data Models Testing Random Effects in Two-Way Spatial Panel Data Models Nicolas Debarsy May 27, 2010 Abstract This paper proposes an alternative testing procedure to the Hausman test statistic to help the applied

More information

Lecture 6: Hypothesis Testing

Lecture 6: Hypothesis Testing Lecture 6: Hypothesis Testing Mauricio Sarrias Universidad Católica del Norte November 6, 2017 1 Moran s I Statistic Mandatory Reading Moran s I based on Cliff and Ord (1972) Kelijan and Prucha (2001)

More information

Using AMOEBA to Create a Spatial Weights Matrix and Identify Spatial Clusters, and a Comparison to Other Clustering Algorithms

Using AMOEBA to Create a Spatial Weights Matrix and Identify Spatial Clusters, and a Comparison to Other Clustering Algorithms Using AMOEBA to Create a Spatial Weights Matrix and Identify Spatial Clusters, and a Comparison to Other Clustering Algorithms Arthur Getis* and Jared Aldstadt** *San Diego State University **SDSU/UCSB

More information

Proceedings of the 8th WSEAS International Conference on APPLIED MATHEMATICS, Tenerife, Spain, December 16-18, 2005 (pp )

Proceedings of the 8th WSEAS International Conference on APPLIED MATHEMATICS, Tenerife, Spain, December 16-18, 2005 (pp ) Proceedings of the 8th WSEAS International Conference on APPLIED MATHEMATICS, Tenerife, Spain, December 16-18, 005 (pp159-164) Spatial Econometrics Modeling of Poverty FERDINAND J. PARAGUAS 1 AND ANTON

More information

Outline. Overview of Issues. Spatial Regression. Luc Anselin

Outline. Overview of Issues. Spatial Regression. Luc Anselin Spatial Regression Luc Anselin University of Illinois, Urbana-Champaign http://www.spacestat.com Outline Overview of Issues Spatial Regression Specifications Space-Time Models Spatial Latent Variable Models

More information

Analyzing spatial autoregressive models using Stata

Analyzing spatial autoregressive models using Stata Analyzing spatial autoregressive models using Stata David M. Drukker StataCorp Summer North American Stata Users Group meeting July 24-25, 2008 Part of joint work with Ingmar Prucha and Harry Kelejian

More information

Testing for Regime Switching in Singaporean Business Cycles

Testing for Regime Switching in Singaporean Business Cycles Testing for Regime Switching in Singaporean Business Cycles Robert Breunig School of Economics Faculty of Economics and Commerce Australian National University and Alison Stegman Research School of Pacific

More information

A SPATIAL ECONOMETRIC STAR MODEL WITH AN APPLICATION TO U.S. COUNTY ECONOMIC GROWTH,

A SPATIAL ECONOMETRIC STAR MODEL WITH AN APPLICATION TO U.S. COUNTY ECONOMIC GROWTH, A SPATIAL ECONOMETRIC STAR MODEL WITH AN APPLICATION TO U.S. COUNTY ECONOMIC GROWTH, 1969 003 by Valerien O. Pede, Raymond J.G.M. Florax and Matthew T. Holt Working Paper # 09-03 March 009 Dept. of Agricultural

More information

Bootstrap Test Statistics for Spatial Econometric Models

Bootstrap Test Statistics for Spatial Econometric Models Bootstrap Test Statistics for Spatial Econometric Models Kuan-Pin Lin Portland State University Zhi-He Long Wu Mei South China University of Technology Current Version: April 22, 2007 Abstract We introduce

More information

ESTIMATION PROBLEMS IN MODELS WITH SPATIAL WEIGHTING MATRICES WHICH HAVE BLOCKS OF EQUAL ELEMENTS*

ESTIMATION PROBLEMS IN MODELS WITH SPATIAL WEIGHTING MATRICES WHICH HAVE BLOCKS OF EQUAL ELEMENTS* JOURNAL OF REGIONAL SCIENCE, VOL. 46, NO. 3, 2006, pp. 507 515 ESTIMATION PROBLEMS IN MODELS WITH SPATIAL WEIGHTING MATRICES WHICH HAVE BLOCKS OF EQUAL ELEMENTS* Harry H. Kelejian Department of Economics,

More information

Econometric modeling of the relationship among macroeconomic variables of Thailand: Smooth transition autoregressive regression model

Econometric modeling of the relationship among macroeconomic variables of Thailand: Smooth transition autoregressive regression model The Empirical Econometrics and Quantitative Economics Letters ISSN 2286 7147 EEQEL all rights reserved Volume 1, Number 4 (December 2012), pp. 21 38. Econometric modeling of the relationship among macroeconomic

More information

Lecture 7: Spatial Econometric Modeling of Origin-Destination flows

Lecture 7: Spatial Econometric Modeling of Origin-Destination flows Lecture 7: Spatial Econometric Modeling of Origin-Destination flows James P. LeSage Department of Economics University of Toledo Toledo, Ohio 43606 e-mail: jlesage@spatial-econometrics.com June 2005 The

More information

Spatial Econometrics

Spatial Econometrics Spatial Econometrics Lecture 5: Single-source model of spatial regression. Combining GIS and regional analysis (5) Spatial Econometrics 1 / 47 Outline 1 Linear model vs SAR/SLM (Spatial Lag) Linear model

More information

Regional Science and Urban Economics

Regional Science and Urban Economics Regional Science and Urban Economics 46 (2014) 103 115 Contents lists available at ScienceDirect Regional Science and Urban Economics journal homepage: www.elsevier.com/locate/regec On the finite sample

More information

Geographically weighted regression approach for origin-destination flows

Geographically weighted regression approach for origin-destination flows Geographically weighted regression approach for origin-destination flows Kazuki Tamesue 1 and Morito Tsutsumi 2 1 Graduate School of Information and Engineering, University of Tsukuba 1-1-1 Tennodai, Tsukuba,

More information

Wan Xu Graduate Research Assistant Food and Resource Economics Department University of Florida Gainesville, FL

Wan Xu Graduate Research Assistant Food and Resource Economics Department University of Florida Gainesville, FL The Impact of Integrated Pest Management Practices on U.S. National Nursery Industry Annul Sales Revenue: An Application of Smooth Transition Spatial Autoregressive Models Wan Xu Graduate Research Assistant

More information

Do Markov-Switching Models Capture Nonlinearities in the Data? Tests using Nonparametric Methods

Do Markov-Switching Models Capture Nonlinearities in the Data? Tests using Nonparametric Methods Do Markov-Switching Models Capture Nonlinearities in the Data? Tests using Nonparametric Methods Robert V. Breunig Centre for Economic Policy Research, Research School of Social Sciences and School of

More information

No

No 28 1 Vol. 28 No. 1 2015 1 Research of Finance and Education Jan. 2015 1 1 2 1. 200241 2. 9700AV 1998-2010 30 F830 A 2095-0098 2015 01-0043 - 010 Schumpeter 1911 2014-09 - 19 No. 40671074 2011 3010 1987-1983

More information

ON THE NEGATION OF THE UNIFORMITY OF SPACE RESEARCH ANNOUNCEMENT

ON THE NEGATION OF THE UNIFORMITY OF SPACE RESEARCH ANNOUNCEMENT MIDWEST STUDENT SUMMIT ON SPACE, HEALTH AND POPULATION ECONOMICS APRIL 18-19, 2007 PURDUE UNIVERSITY ON THE NEGATION OF THE UNIFORMITY OF SPACE RESEARCH ANNOUNCEMENT Benoit Delbecq Agricultural Economics

More information

A Test of Cointegration Rank Based Title Component Analysis.

A Test of Cointegration Rank Based Title Component Analysis. A Test of Cointegration Rank Based Title Component Analysis Author(s) Chigira, Hiroaki Citation Issue 2006-01 Date Type Technical Report Text Version publisher URL http://hdl.handle.net/10086/13683 Right

More information

Spatial Statistics For Real Estate Data 1

Spatial Statistics For Real Estate Data 1 1 Key words: spatial heterogeneity, spatial autocorrelation, spatial statistics, geostatistics, Geographical Information System SUMMARY: The paper presents spatial statistics tools in application to real

More information

Spatial Regression. 9. Specification Tests (1) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Spatial Regression. 9. Specification Tests (1) Luc Anselin.   Copyright 2017 by Luc Anselin, All Rights Reserved Spatial Regression 9. Specification Tests (1) Luc Anselin http://spatial.uchicago.edu 1 basic concepts types of tests Moran s I classic ML-based tests LM tests 2 Basic Concepts 3 The Logic of Specification

More information

Obtaining Critical Values for Test of Markov Regime Switching

Obtaining Critical Values for Test of Markov Regime Switching University of California, Santa Barbara From the SelectedWorks of Douglas G. Steigerwald November 1, 01 Obtaining Critical Values for Test of Markov Regime Switching Douglas G Steigerwald, University of

More information

Switching Regime Estimation

Switching Regime Estimation Switching Regime Estimation Series de Tiempo BIrkbeck March 2013 Martin Sola (FE) Markov Switching models 01/13 1 / 52 The economy (the time series) often behaves very different in periods such as booms

More information

Least Absolute Value vs. Least Squares Estimation and Inference Procedures in Regression Models with Asymmetric Error Distributions

Least Absolute Value vs. Least Squares Estimation and Inference Procedures in Regression Models with Asymmetric Error Distributions Journal of Modern Applied Statistical Methods Volume 8 Issue 1 Article 13 5-1-2009 Least Absolute Value vs. Least Squares Estimation and Inference Procedures in Regression Models with Asymmetric Error

More information

W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS

W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS 1 W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS An Liu University of Groningen Henk Folmer University of Groningen Wageningen University Han Oud Radboud

More information

Financial Econometrics

Financial Econometrics Financial Econometrics Nonlinear time series analysis Gerald P. Dwyer Trinity College, Dublin January 2016 Outline 1 Nonlinearity Does nonlinearity matter? Nonlinear models Tests for nonlinearity Forecasting

More information

Economic modelling and forecasting

Economic modelling and forecasting Economic modelling and forecasting 2-6 February 2015 Bank of England he generalised method of moments Ole Rummel Adviser, CCBS at the Bank of England ole.rummel@bankofengland.co.uk Outline Classical estimation

More information

BANDWIDTH SELECTION FOR SPATIAL HAC AND OTHER ROBUST COVARIANCE ESTIMATORS. Working Paper # October Dept. of Agricultural Economics

BANDWIDTH SELECTION FOR SPATIAL HAC AND OTHER ROBUST COVARIANCE ESTIMATORS. Working Paper # October Dept. of Agricultural Economics BANDWIDTH SELECTION FOR SPATIAL HAC AND OTHER ROBUST COVARIANCE ESTIMATORS by Dayton M. Lambert, Raymond J.G.M. Florax and Seong-Hoon Cho Working Paper # 08-10 October 2008 Dept. of Agricultural Economics

More information

Appendix A: The time series behavior of employment growth

Appendix A: The time series behavior of employment growth Unpublished appendices from The Relationship between Firm Size and Firm Growth in the U.S. Manufacturing Sector Bronwyn H. Hall Journal of Industrial Economics 35 (June 987): 583-606. Appendix A: The time

More information

VAR Models and Applications

VAR Models and Applications VAR Models and Applications Laurent Ferrara 1 1 University of Paris West M2 EIPMC Oct. 2016 Overview of the presentation 1. Vector Auto-Regressions Definition Estimation Testing 2. Impulse responses functions

More information

Volume 03, Issue 6. Comparison of Panel Cointegration Tests

Volume 03, Issue 6. Comparison of Panel Cointegration Tests Volume 03, Issue 6 Comparison of Panel Cointegration Tests Deniz Dilan Karaman Örsal Humboldt University Berlin Abstract The main aim of this paper is to compare the size and size-adjusted power properties

More information

splm: econometric analysis of spatial panel data

splm: econometric analysis of spatial panel data splm: econometric analysis of spatial panel data Giovanni Millo 1 Gianfranco Piras 2 1 Research Dept., Generali S.p.A. and DiSES, Univ. of Trieste 2 REAL, UIUC user! Conference Rennes, July 8th 2009 Introduction

More information

Spatial Autocorrelation and Interactions between Surface Temperature Trends and Socioeconomic Changes

Spatial Autocorrelation and Interactions between Surface Temperature Trends and Socioeconomic Changes Spatial Autocorrelation and Interactions between Surface Temperature Trends and Socioeconomic Changes Ross McKitrick Department of Economics University of Guelph December, 00 1 1 1 1 Spatial Autocorrelation

More information

Application of eigenvector-based spatial filtering approach to. a multinomial logit model for land use data

Application of eigenvector-based spatial filtering approach to. a multinomial logit model for land use data Presented at the Seventh World Conference of the Spatial Econometrics Association, the Key Bridge Marriott Hotel, Washington, D.C., USA, July 10 12, 2013. Application of eigenvector-based spatial filtering

More information

Combining Regressive and Auto-Regressive Models for Spatial-Temporal Prediction

Combining Regressive and Auto-Regressive Models for Spatial-Temporal Prediction Combining Regressive and Auto-Regressive Models for Spatial-Temporal Prediction Dragoljub Pokrajac DPOKRAJA@EECS.WSU.EDU Zoran Obradovic ZORAN@EECS.WSU.EDU School of Electrical Engineering and Computer

More information

SPATIAL ECONOMETRICS: METHODS AND MODELS

SPATIAL ECONOMETRICS: METHODS AND MODELS SPATIAL ECONOMETRICS: METHODS AND MODELS STUDIES IN OPERATIONAL REGIONAL SCIENCE Folmer, H., Regional Economic Policy. 1986. ISBN 90-247-3308-1. Brouwer, F., Integrated Environmental Modelling: Design

More information

Spatial heterogeneity in economic growth of European regions

Spatial heterogeneity in economic growth of European regions Spatial heterogeneity in economic growth of European regions Paolo Postiglione 1, M.Simona Andreano 2, Roberto Benedetti 3 Draft version (please do not cite) July 4, 2015 Abstract This paper describes

More information

Vector Auto-Regressive Models

Vector Auto-Regressive Models Vector Auto-Regressive Models Laurent Ferrara 1 1 University of Paris Nanterre M2 Oct. 2018 Overview of the presentation 1. Vector Auto-Regressions Definition Estimation Testing 2. Impulse responses functions

More information

G. S. Maddala Kajal Lahiri. WILEY A John Wiley and Sons, Ltd., Publication

G. S. Maddala Kajal Lahiri. WILEY A John Wiley and Sons, Ltd., Publication G. S. Maddala Kajal Lahiri WILEY A John Wiley and Sons, Ltd., Publication TEMT Foreword Preface to the Fourth Edition xvii xix Part I Introduction and the Linear Regression Model 1 CHAPTER 1 What is Econometrics?

More information

TESTING FOR NORMALITY IN THE LINEAR REGRESSION MODEL: AN EMPIRICAL LIKELIHOOD RATIO TEST

TESTING FOR NORMALITY IN THE LINEAR REGRESSION MODEL: AN EMPIRICAL LIKELIHOOD RATIO TEST Econometrics Working Paper EWP0402 ISSN 1485-6441 Department of Economics TESTING FOR NORMALITY IN THE LINEAR REGRESSION MODEL: AN EMPIRICAL LIKELIHOOD RATIO TEST Lauren Bin Dong & David E. A. Giles Department

More information

Spatial Effects in Convergence of Portuguese Product

Spatial Effects in Convergence of Portuguese Product Spatial Effects in Convergence of Portuguese Product Vítor João Pereira Domingues Martinho Instituto Politécnico de Viseu 2011 Working paper nº 79/2011 globadvantage Center of Research in International

More information

LM threshold unit root tests

LM threshold unit root tests Lee, J., Strazicich, M.C., & Chul Yu, B. (2011). LM Threshold Unit Root Tests. Economics Letters, 110(2): 113-116 (Feb 2011). Published by Elsevier (ISSN: 0165-1765). http://0- dx.doi.org.wncln.wncln.org/10.1016/j.econlet.2010.10.014

More information

Spatial Regression. 10. Specification Tests (2) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Spatial Regression. 10. Specification Tests (2) Luc Anselin.  Copyright 2017 by Luc Anselin, All Rights Reserved Spatial Regression 10. Specification Tests (2) Luc Anselin http://spatial.uchicago.edu 1 robust LM tests higher order tests 2SLS residuals specification search 2 Robust LM Tests 3 Recap and Notation LM-Error

More information

The Davies Problem: A New Test for Random Slope in the Hierarchical Linear Model

The Davies Problem: A New Test for Random Slope in the Hierarchical Linear Model The Davies Problem: A New Test for Random Slope in the Hierarchical Linear Model Rutger van Oest Department of Marketing, BI Norwegian Business School, Oslo, Norway Philip Hans Franses Econometric Institute,

More information

Instrumental Variables/Method of

Instrumental Variables/Method of Instrumental Variables/Method of 80 Moments Estimation Ingmar R. Prucha Contents 80.1 Introduction... 1597 80.2 A Primer on GMM Estimation... 1599 80.2.1 Model Specification and Moment Conditions... 1599

More information

SIMULATION AND APPLICATION OF THE SPATIAL AUTOREGRESSIVE GEOGRAPHICALLY WEIGHTED REGRESSION MODEL (SAR-GWR)

SIMULATION AND APPLICATION OF THE SPATIAL AUTOREGRESSIVE GEOGRAPHICALLY WEIGHTED REGRESSION MODEL (SAR-GWR) SIMULATION AND APPLICATION OF THE SPATIAL AUTOREGRESSIVE GEOGRAPHICALLY WEIGHTED REGRESSION MODEL (SAR-GWR) I. Gede Nyoman Mindra Jaya 1, Budi Nurani Ruchjana 2, Bertho Tantular 1, Zulhanif 1 and Yudhie

More information

A two-step approach to account for unobserved spatial heterogeneity 1

A two-step approach to account for unobserved spatial heterogeneity 1 A two-step approach to account for unobserved spatial heterogeneity 1 Anna Gloria Billé ᵃ*, Roberto Benedetti b, Paolo Postiglione b ᵃ Department of Economics and Finance, University of Rome Tor Vergata

More information

DEPARTMENT OF ECONOMICS

DEPARTMENT OF ECONOMICS ISSN 0819-64 ISBN 0 7340 616 1 THE UNIVERSITY OF MELBOURNE DEPARTMENT OF ECONOMICS RESEARCH PAPER NUMBER 959 FEBRUARY 006 TESTING FOR RATE-DEPENDENCE AND ASYMMETRY IN INFLATION UNCERTAINTY: EVIDENCE FROM

More information

A SPATIAL CLIFF-ORD-TYPE MODEL WITH HETEROSKEDASTIC INNOVATIONS: SMALL AND LARGE SAMPLE RESULTS

A SPATIAL CLIFF-ORD-TYPE MODEL WITH HETEROSKEDASTIC INNOVATIONS: SMALL AND LARGE SAMPLE RESULTS JOURNAL OF REGIONAL SCIENCE, VOL. 50, NO. 2, 2010, pp. 592 614 A SPATIAL CLIFF-ORD-TYPE MODEL WITH HETEROSKEDASTIC INNOVATIONS: SMALL AND LARGE SAMPLE RESULTS Irani Arraiz Inter-American Development Bank,

More information

Identifying Nonlinearities in Spatial Autoregressive Models

Identifying Nonlinearities in Spatial Autoregressive Models Identifying Nonlinearities in Spatial Autoregressive Models Nicolas Debarsy and Vincenzo Verardi June 2009-First Draft Abstract In spatial autoregressive models, the functional form of autocorrelation

More information

Spatial groupwise heteroskedasticity and the SCAN approach

Spatial groupwise heteroskedasticity and the SCAN approach Spatial groupwise heteroskedasticity and the SCAN approach Coro Chasco, Julie Le Gallo, Fernando A. López and Jesús Mur (*) Universidad Autónoma de Madrid. email: coro.chasco@uam.es (**) Université Bourgogne

More information

Spatial Analysis 2. Spatial Autocorrelation

Spatial Analysis 2. Spatial Autocorrelation Spatial Analysis 2 Spatial Autocorrelation Spatial Autocorrelation a relationship between nearby spatial units of the same variable If, for every pair of subareas i and j in the study region, the drawings

More information

Reasons for Instability in Spatial Dependence Models

Reasons for Instability in Spatial Dependence Models 1 Reasons for nstability in Spatial Dependence Models Jesús Mur () Fernando López () Ana Angulo () () Department of Economic Analysis University of Zaragoza Gran Vía, -4. (55). Zaragoza. Spain. e-mail:

More information

AJAE Appendix: The Commodity Terms of Trade, Unit Roots, and Nonlinear Alternatives

AJAE Appendix: The Commodity Terms of Trade, Unit Roots, and Nonlinear Alternatives AJAE Appendix: The Commodity Terms of Trade, Unit Roots, and Nonlinear Alternatives Joseph V. Balagtas Department of Agricultural Economics Purdue University Matthew T. Holt Department of Agricultural

More information

News Shocks: Different Effects in Boom and Recession?

News Shocks: Different Effects in Boom and Recession? News Shocks: Different Effects in Boom and Recession? Maria Bolboaca, Sarah Fischer University of Bern Study Center Gerzensee June 7, 5 / Introduction News are defined in the literature as exogenous changes

More information

Econometric Analysis of Cross Section and Panel Data

Econometric Analysis of Cross Section and Panel Data Econometric Analysis of Cross Section and Panel Data Jeffrey M. Wooldridge / The MIT Press Cambridge, Massachusetts London, England Contents Preface Acknowledgments xvii xxiii I INTRODUCTION AND BACKGROUND

More information

Estimation and Hypothesis Testing in LAV Regression with Autocorrelated Errors: Is Correction for Autocorrelation Helpful?

Estimation and Hypothesis Testing in LAV Regression with Autocorrelated Errors: Is Correction for Autocorrelation Helpful? Journal of Modern Applied Statistical Methods Volume 10 Issue Article 13 11-1-011 Estimation and Hypothesis Testing in LAV Regression with Autocorrelated Errors: Is Correction for Autocorrelation Helpful?

More information

Economics 536 Lecture 7. Introduction to Specification Testing in Dynamic Econometric Models

Economics 536 Lecture 7. Introduction to Specification Testing in Dynamic Econometric Models University of Illinois Fall 2016 Department of Economics Roger Koenker Economics 536 Lecture 7 Introduction to Specification Testing in Dynamic Econometric Models In this lecture I want to briefly describe

More information

RAO s SCORE TEST IN SPATIAL ECONOMETRICS

RAO s SCORE TEST IN SPATIAL ECONOMETRICS RAO s SCORE TEST IN SPATIAL ECONOMETRICS Luc Anselin Regional Economics Applications Laboratory (REAL) and Department of Agricultural and Consumer Economics University of Illinois, Urbana-Champaign Urbana,

More information

On the econometrics of the Koyck model

On the econometrics of the Koyck model On the econometrics of the Koyck model Philip Hans Franses and Rutger van Oest Econometric Institute, Erasmus University Rotterdam P.O. Box 1738, NL-3000 DR, Rotterdam, The Netherlands Econometric Institute

More information

Generated Covariates in Nonparametric Estimation: A Short Review.

Generated Covariates in Nonparametric Estimation: A Short Review. Generated Covariates in Nonparametric Estimation: A Short Review. Enno Mammen, Christoph Rothe, and Melanie Schienle Abstract In many applications, covariates are not observed but have to be estimated

More information

Spatial Regression. 11. Spatial Two Stage Least Squares. Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Spatial Regression. 11. Spatial Two Stage Least Squares. Luc Anselin.  Copyright 2017 by Luc Anselin, All Rights Reserved Spatial Regression 11. Spatial Two Stage Least Squares Luc Anselin http://spatial.uchicago.edu 1 endogeneity and instruments spatial 2SLS best and optimal estimators HAC standard errors 2 Endogeneity and

More information

Omitted Variable Biases of OLS and Spatial Lag Models

Omitted Variable Biases of OLS and Spatial Lag Models Omitted Variable Biases of OLS and Spatial Lag Models R. Kelley Pace and James P. LeSage 1 Introduction Numerous authors have suggested that omitted variables affect spatial regression methods less than

More information

If we want to analyze experimental or simulated data we might encounter the following tasks:

If we want to analyze experimental or simulated data we might encounter the following tasks: Chapter 1 Introduction If we want to analyze experimental or simulated data we might encounter the following tasks: Characterization of the source of the signal and diagnosis Studying dependencies Prediction

More information

Steven Cook University of Wales Swansea. Abstract

Steven Cook University of Wales Swansea. Abstract On the finite sample power of modified Dickey Fuller tests: The role of the initial condition Steven Cook University of Wales Swansea Abstract The relationship between the initial condition of time series

More information

GARCH Models Estimation and Inference. Eduardo Rossi University of Pavia

GARCH Models Estimation and Inference. Eduardo Rossi University of Pavia GARCH Models Estimation and Inference Eduardo Rossi University of Pavia Likelihood function The procedure most often used in estimating θ 0 in ARCH models involves the maximization of a likelihood function

More information

Testing Restrictions and Comparing Models

Testing Restrictions and Comparing Models Econ. 513, Time Series Econometrics Fall 00 Chris Sims Testing Restrictions and Comparing Models 1. THE PROBLEM We consider here the problem of comparing two parametric models for the data X, defined by

More information

Choice of Spectral Density Estimator in Ng-Perron Test: Comparative Analysis

Choice of Spectral Density Estimator in Ng-Perron Test: Comparative Analysis MPRA Munich Personal RePEc Archive Choice of Spectral Density Estimator in Ng-Perron Test: Comparative Analysis Muhammad Irfan Malik and Atiq-ur- Rehman International Institute of Islamic Economics, International

More information

MODELING ECONOMIC GROWTH OF DISTRICTS IN THE PROVINCE OF BALI USING SPATIAL ECONOMETRIC PANEL DATA MODEL

MODELING ECONOMIC GROWTH OF DISTRICTS IN THE PROVINCE OF BALI USING SPATIAL ECONOMETRIC PANEL DATA MODEL Proceeding of International Conference On Research, Implementation And Education Of Mathematics And Sciences 015, Yogyakarta State University, 17-1 May 015 MODELING ECONOMIC GROWTH OF DISTRICTS IN THE

More information

Network data in regression framework

Network data in regression framework 13-14 July 2009 University of Salerno (Italy) Network data in regression framework Maria ProsperinaVitale Department of Economics and Statistics University of Salerno (Italy) mvitale@unisa.it - Theoretical

More information

A SPATIAL ANALYSIS OF A RURAL LAND MARKET USING ALTERNATIVE SPATIAL WEIGHT MATRICES

A SPATIAL ANALYSIS OF A RURAL LAND MARKET USING ALTERNATIVE SPATIAL WEIGHT MATRICES A Spatial Analysis of a Rural Land Market Using Alternative Spatial Weight Matrices A SPATIAL ANALYSIS OF A RURAL LAND MARKET USING ALTERNATIVE SPATIAL WEIGHT MATRICES Patricia Soto, Louisiana State University

More information

Purchasing power parity: A nonlinear multivariate perspective. Abstract

Purchasing power parity: A nonlinear multivariate perspective. Abstract Purchasing power parity: A nonlinear multivariate perspective Frédérique Bec THEMA, University of Cergy-Pontoise and CREST, France Mélika Ben Salem OEP, Paris-Est University and LEA-INRA (PSE), France

More information

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley Review of Classical Least Squares James L. Powell Department of Economics University of California, Berkeley The Classical Linear Model The object of least squares regression methods is to model and estimate

More information

Modeling Spatial Externalities: A Panel Data Approach

Modeling Spatial Externalities: A Panel Data Approach Modeling Spatial Externalities: A Panel Data Approach Christian Beer and Aleksandra Riedl May 29, 2009 First draft, do not quote! Abstract In this paper we argue that the Spatial Durbin Model (SDM) is

More information

GARCH Models Estimation and Inference

GARCH Models Estimation and Inference GARCH Models Estimation and Inference Eduardo Rossi University of Pavia December 013 Rossi GARCH Financial Econometrics - 013 1 / 1 Likelihood function The procedure most often used in estimating θ 0 in

More information

Predicting bond returns using the output gap in expansions and recessions

Predicting bond returns using the output gap in expansions and recessions Erasmus university Rotterdam Erasmus school of economics Bachelor Thesis Quantitative finance Predicting bond returns using the output gap in expansions and recessions Author: Martijn Eertman Studentnumber:

More information

SPACE Workshop NSF NCGIA CSISS UCGIS SDSU. Aldstadt, Getis, Jankowski, Rey, Weeks SDSU F. Goodchild, M. Goodchild, Janelle, Rebich UCSB

SPACE Workshop NSF NCGIA CSISS UCGIS SDSU. Aldstadt, Getis, Jankowski, Rey, Weeks SDSU F. Goodchild, M. Goodchild, Janelle, Rebich UCSB SPACE Workshop NSF NCGIA CSISS UCGIS SDSU Aldstadt, Getis, Jankowski, Rey, Weeks SDSU F. Goodchild, M. Goodchild, Janelle, Rebich UCSB August 2-8, 2004 San Diego State University Some Examples of Spatial

More information

Introduction. Introduction (Contd.) Market Equilibrium and Spatial Variability in the Value of Housing Attributes. Urban location theory.

Introduction. Introduction (Contd.) Market Equilibrium and Spatial Variability in the Value of Housing Attributes. Urban location theory. Forestry, Wildlife, and Fisheries Graduate Seminar Market Equilibrium and Spatial Variability in the Value of Housing Attributes Seung Gyu Kim Wednesday, 12 March 2008 1 Introduction Urban location theory

More information

Tests forspatial Lag Dependence Based onmethodof Moments Estimation

Tests forspatial Lag Dependence Based onmethodof Moments Estimation 1 Tests forspatial Lag Dependence Based onmethodof Moments Estimation by Luz A. Saavedra* Department of Economics University of South Florida 4202 East Fowler Ave. BSN 3111 Tampa, FL 33620-5524 lsaavedr@coba.usf.edu

More information

1 Overview. 2 Data Files. 3 Estimation Programs

1 Overview. 2 Data Files. 3 Estimation Programs 1 Overview The programs made available on this web page are sample programs for the computation of estimators introduced in Kelejian and Prucha (1999). In particular we provide sample programs for the

More information

Cross-sectional space-time modeling using ARNN(p, n) processes

Cross-sectional space-time modeling using ARNN(p, n) processes Cross-sectional space-time modeling using ARNN(p, n) processes W. Polasek K. Kakamu September, 006 Abstract We suggest a new class of cross-sectional space-time models based on local AR models and nearest

More information

Chapter 2 Linear Spatial Dependence Models for Cross-Section Data

Chapter 2 Linear Spatial Dependence Models for Cross-Section Data Chapter 2 Linear Spatial Dependence Models for Cross-Section Data Abstract This chapter gives an overview of all linear spatial econometric models with different combinations of interaction effects that

More information

Spatial Effects in Convergence of Portuguese Product

Spatial Effects in Convergence of Portuguese Product Spatial Effects in Convergence of Portuguese Product Vitor João Pereira Domingues Martinho Unidade de I&D do Instituto Politécnico de Viseu Av. Cor. José Maria Vale de Andrade Campus Politécnico 354-51

More information

Functional Form. Econometrics. ADEi.

Functional Form. Econometrics. ADEi. Functional Form Econometrics. ADEi. 1. Introduction We have employed the linear function in our model specification. Why? It is simple and has good mathematical properties. It could be reasonable approximation,

More information

On Autoregressive Order Selection Criteria

On Autoregressive Order Selection Criteria On Autoregressive Order Selection Criteria Venus Khim-Sen Liew Faculty of Economics and Management, Universiti Putra Malaysia, 43400 UPM, Serdang, Malaysia This version: 1 March 2004. Abstract This study

More information

GMM Estimation of Spatial Error Autocorrelation with and without Heteroskedasticity

GMM Estimation of Spatial Error Autocorrelation with and without Heteroskedasticity GMM Estimation of Spatial Error Autocorrelation with and without Heteroskedasticity Luc Anselin July 14, 2011 1 Background This note documents the steps needed for an efficient GMM estimation of the regression

More information

Econometrics of Panel Data

Econometrics of Panel Data Econometrics of Panel Data Jakub Mućk Meeting # 6 Jakub Mućk Econometrics of Panel Data Meeting # 6 1 / 36 Outline 1 The First-Difference (FD) estimator 2 Dynamic panel data models 3 The Anderson and Hsiao

More information

Lectures on Structural Change

Lectures on Structural Change Lectures on Structural Change Eric Zivot Department of Economics, University of Washington April5,2003 1 Overview of Testing for and Estimating Structural Change in Econometric Models 1. Day 1: Tests of

More information

GMM Based Tests for Locally Misspeci ed Models

GMM Based Tests for Locally Misspeci ed Models GMM Based Tests for Locally Misspeci ed Models Anil K. Bera Department of Economics University of Illinois, USA Walter Sosa Escudero University of San Andr es and National University of La Plata, Argentina

More information

Departamento de Economía Universidad de Chile

Departamento de Economía Universidad de Chile Departamento de Economía Universidad de Chile GRADUATE COURSE SPATIAL ECONOMETRICS November 14, 16, 17, 20 and 21, 2017 Prof. Henk Folmer University of Groningen Objectives The main objective of the course

More information

On the Power of Tests for Regime Switching

On the Power of Tests for Regime Switching On the Power of Tests for Regime Switching joint work with Drew Carter and Ben Hansen Douglas G. Steigerwald UC Santa Barbara May 2015 D. Steigerwald (UCSB) Regime Switching May 2015 1 / 42 Motivating

More information

Spatial Regression. 6. Specification Spatial Heterogeneity. Luc Anselin.

Spatial Regression. 6. Specification Spatial Heterogeneity. Luc Anselin. Spatial Regression 6. Specification Spatial Heterogeneity Luc Anselin http://spatial.uchicago.edu 1 homogeneity and heterogeneity spatial regimes spatially varying coefficients spatial random effects 2

More information

A TIME SERIES PARADOX: UNIT ROOT TESTS PERFORM POORLY WHEN DATA ARE COINTEGRATED

A TIME SERIES PARADOX: UNIT ROOT TESTS PERFORM POORLY WHEN DATA ARE COINTEGRATED A TIME SERIES PARADOX: UNIT ROOT TESTS PERFORM POORLY WHEN DATA ARE COINTEGRATED by W. Robert Reed Department of Economics and Finance University of Canterbury, New Zealand Email: bob.reed@canterbury.ac.nz

More information

ECONOMICS 7200 MODERN TIME SERIES ANALYSIS Econometric Theory and Applications

ECONOMICS 7200 MODERN TIME SERIES ANALYSIS Econometric Theory and Applications ECONOMICS 7200 MODERN TIME SERIES ANALYSIS Econometric Theory and Applications Yongmiao Hong Department of Economics & Department of Statistical Sciences Cornell University Spring 2019 Time and uncertainty

More information

Knowledge Spillovers, Spatial Dependence, and Regional Economic Growth in U.S. Metropolitan Areas. Up Lim, B.A., M.C.P.

Knowledge Spillovers, Spatial Dependence, and Regional Economic Growth in U.S. Metropolitan Areas. Up Lim, B.A., M.C.P. Knowledge Spillovers, Spatial Dependence, and Regional Economic Growth in U.S. Metropolitan Areas by Up Lim, B.A., M.C.P. DISSERTATION Presented to the Faculty of the Graduate School of The University

More information

A Robust LM Test for Spatial Error Components

A Robust LM Test for Spatial Error Components A Robust LM Test for Spatial Error Components Zhenlin Yang January 2009 Paper No. 04-2009 ANY OPINIONS EXPRESSED ARE THOSE OF THE AUTHOR(S) AND NOT NECESSARILY THOSE OF THE SCHOOL OF ECONOMICS & SOCIAL

More information

Birkbeck Working Papers in Economics & Finance

Birkbeck Working Papers in Economics & Finance ISSN 1745-8587 Birkbeck Working Papers in Economics & Finance Department of Economics, Mathematics and Statistics BWPEF 1809 A Note on Specification Testing in Some Structural Regression Models Walter

More information

Size and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach

Size and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach Size and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach Ghazi Shukur Panagiotis Mantalos International Business School Department of Statistics Jönköping University Lund

More information

Tests of the Present-Value Model of the Current Account: A Note

Tests of the Present-Value Model of the Current Account: A Note Tests of the Present-Value Model of the Current Account: A Note Hafedh Bouakez Takashi Kano March 5, 2007 Abstract Using a Monte Carlo approach, we evaluate the small-sample properties of four different

More information