Semiparametric Bayesian inference in multiple equation models

This paper outlines an approach to Bayesian semiparametric regression in multiple equation models which can be used to carry out inference in seemingly unrelated regressions or simultaneous equations models with nonparametric components. The approach treats the points on each nonparametric regression line as unknown parameters and uses a prior on the degree of smoothness of each line to ensure valid posterior inference despite the fact that the number of parameters is greater than the number of observations. We develop an empirical Bayesian approach that allows us to estimate the prior smoothing hyperparameters from the data. An advantage of our semiparametric model is that it is written as a seemingly unrelated regressions model with independent normal–Wishart prior. Since this model is a common one, textbook results for posterior inference, model comparison, prediction and posterior computation are immediately available. We use this model in an application involving a two‐equation structural model drawn from the labour and returns to schooling literatures. Copyright © 2005 John Wiley & Sons, Ltd.     

Forecasting with factor-augmented error correction models

As a generalization of the factor-augmented VAR (FAVAR) and of the Error Correction Model (ECM), Banerjee and Marcellino (2009) introduced the Factor-augmented Error Correction Model (FECM). The FECM combines error-correction, cointegration and dynamic factor models, and has several conceptual advantages over the standard ECM and FAVAR models. In particular, it uses a larger dataset than the ECM and incorporates the long-run information which the FAVAR is missing because of its specification in differences. In this paper, we examine the forecasting performance of the FECM by means of an analytical example, Monte Carlo simulations and several empirical applications. We show that FECM generally offers a higher forecasting precision relative to the FAVAR, and marks a useful step forward for forecasting with large datasets.     

Data-based priors for vector error correction models

We propose two data-based priors for vector error correction models. Both priors lead to highly automatic approaches which require only minimal user input. For the first one, we propose a reduced rank prior which encourages shrinkage towards a low-rank, row-sparse, and column-sparse long-run matrix. For the second one, we propose the use of the horseshoe prior, which shrinks all elements of the long-run matrix towards zero. Two empirical investigations reveal that Bayesian vector error correction (BVEC) models equipped with our proposed priors scale well to higher dimensions and forecast well. In comparison to VARs in first differences, they are able to exploit the information in the level variables. This turns out to be relevant to improve the forecasts for some macroeconomic variables. A simulation study shows that the BVEC with data-based priors possesses good frequentist estimation properties.     

Stability results for nonlinear error correction models

This paper studies analogs of Granger's representation theorem in the context of a general nonlinear vector autoregressive error correction model. The model allows for nonlinear autoregressive conditional heteroskedasticity and the conditional distribution involved can be a mixture distribution of a rather general type. Mixture models of this kind can be thought of as generalizations of threshold models and they have attracted attention in the recent time series and econometrics literature. The paper develops a useful transformation which shows how the nonlinear error correction model can be transformed to a nonlinear vector autoregressive model so that available results on the stationarity or nonstationarity of the latter can be used for the former. The most satisfactory results are obtained in a model in which a specific structural relation between the nonlinearity and equilibrium correction prevails. Without this structural relation only a lower bound for the number of long-run equilibrium relations can explicitly be determined because the exact number depends on properties of the first and second moments of a nonlinear stationary component of the process.     

Bootstrap testing for the null of no cointegration in a threshold vector error correction model

We develop a test for the linear no cointegration null hypothesis in a threshold vector error correction model. We adopt a sup-Wald type test and derive its null asymptotic distribution. A residual-based bootstrap is proposed, and the first-order consistency of the bootstrap is established. A set of Monte Carlo simulations shows that the bootstrap corrects size distortion of asymptotic distribution in finite samples, and that its power against the threshold cointegration alternative is significantly greater than that of conventional cointegration tests. Our method is illustrated with used car price indexes.     

Identification of simultaneous equation models with measurement error

This paper establishes identification conditions for a simultaneous equation model in which some of the exogenous variables are measured with error. It is assumed that observational information is confined to the covariance matrix of the observed variables and that prior information on the structural coefficients and error variances takes the form of zero restrictions. The primary result is an easily-applied assignment condition for checking whether or not there are an adequate number and variety of prior restrictions to identify the structural parameters.     

Residual autocorrelation testing for vector error correction models

In applied time series analysis, checking for autocorrelation in a fitted model is a routine diagnostic tool. Therefore it is useful to know the asymptotic and small sample properties of the standard tests for the case when some of the variables are cointegrated. The properties of residual autocorrelations of vector error correction models (VECMs) and tests for residual autocorrelation are derived. In particular, the asymptotic distributions of Lagrange multiplier (LM) and portmanteau tests are given. Monte Carlo simulations show that the LM tests have satisfactory size properties only if autocorrelation of small order is tested in systems of small dimension. In contrast, portmanteau tests have roughly correct size in small samples only if higher order residual autocorrelation is tested. Their critical values have to be adjusted for the cointegration rank of the system, however.     

Exact tests in single equation autoregressive distributed lag models

For hypotheses on the coefficient values of the lagged-dependent variables in the ARX class of dynamic regression models, test procedures are developed which yield exact inference for given (up to an unknown scale factor) distribution of the innovation errors. They include exact tests on the maximum lag length, for structural change and on the presence of (seasonal or multiple) unit roots, i.e. they cover situations where usually asymptotic and non-exact t, F, AOC, ADF or HEGY tests are employed. The various procedures are demonstrated and compared in illustrative empirical models and the approach is critically discussed.     

Estimation and model selection based inference in single and multiple threshold models

This paper evaluates the properties of a joint and sequential estimation procedure for estimating the parameters of single and multiple threshold models. We initially proceed under the assumption that the number of regimes is known á priori but subsequently relax this assumption via the introduction of a model selection based procedure that allows the estimation of both the unknown parameters and their number to be performed jointly. Theoretical properties of the resulting estimators are derived and their finite sample properties investigated.     

Asymptotic distribution of the cointegrating vector estimator in error correction models with conditional heteroskedasticity

This paper explores the asymptotic distribution of the cointegrating vector estimator in error correction models with conditionally heteroskedastic errors. Asymptotic properties of the maximum likelihood estimator (MLE) of the cointegrating vector, which estimates the cointegrating vector and the multivariate GARCH process jointly, are provided. The MLE of the cointegrating vector follows mixture normal, and its asymptotic distribution depends on the conditional heteroskedasticity and the kurtosis of standardized innovations. The reduced rank regression (RRR) estimator and the regression-based cointegrating vector estimators do not consider conditional heteroskedasticity, and thus the efficiency gain of the MLE emerges as the magnitude of conditional heteroskedasticity increases. The simulation results indicate that the relative power of the t-statistics based on the MLE improves significantly as the GARCH effect increases.     

Identification of simultaneous equation models with measurement error: a computerized evaluation

Rank conditions for identification in structural models are often difficult evaluate. Here we consider simultaneous equation models with measurement error and we show that previously published rank conditions for identification are not well-suited for evaluation. An alternative rank condition is derived and a computer algebra program is presented that takes care of both the construction and the computation of the rank of the relevant Jacobian matrix. It uses the parameter restrictions as input in order to characterize the identification situation of the individual parameters in the output.     

Testing structural equation models: the impact of error variances in the data generating process

Yet another paper on fit measures? To our knowledge, very few papers discuss how fit measures are affected by error variance in the Data Generating Process (DGP). The present paper deals with this. Based upon an extensive simulation study, this paper shows that the effects of increased error variance differ significantly for various fit measures. In addition to error variance the effects depend on sample size and severity of misspecification. The findings confirm the general notion that good fit as measured by the chi-square, RMSEA and GFI etc. does not necessarily mean that the model is correctly specified and reliable. One finding is that the chi square test may give support to misspecified models in situations with a high level of error variance in the DGP, for small sample sizes. Another finding is that the chi-square test looses power also for large sample sizes when the model is negligible misspecified. Other results include incremental fit indices as NFI and RFI which prove to be more informative indicators under these circumstances. At the end of the paper we formulate some guidelines for use of different fit measures.     

Bootstrap inference for impulse response functions in factor‐augmented vector autoregressions

In this study, we consider residual‐based bootstrap methods to construct the confidence interval for structural impulse response functions in factor‐augmented vector autoregressions. In particular, we compare the bootstrap with factor estimation (Procedure A) with the bootstrap without factor estimation (Procedure B). Both procedures are asymptotically valid under the condition , where N and T are the cross‐sectional dimension and the time dimension, respectively. However, Procedure A is also valid even when with 0 ≤ c < ∞ because it accounts for the effect of the factor estimation errors on the impulse response function estimator. Our simulation results suggest that Procedure A achieves more accurate coverage rates than those of Procedure B, especially when N is much smaller than T. In the monetary policy analysis of Bernanke et al. (Quarterly Journal of Economics, 2005, 120(1), 387–422), the proposed methods can produce statistically different results.     

Subsampling inference in threshold autoregressive models

This paper discusses inference in self-exciting threshold autoregressive (SETAR) models. Of main interest is inference for the threshold parameter. It is well-known that the asymptotics of the corresponding estimator depend upon whether the SETAR model is continuous or not. In the continuous case, the limiting distribution is normal and standard inference is possible. In the discontinuous case, the limiting distribution is non-normal and it is not known how to estimate it consistently. We show that valid inference can be drawn by the use of the subsampling method. Moreover, the method can even be extended to situations where the (dis)continuity of the model is unknown. In this case, the inference for the regression parameters of the model also becomes difficult and subsampling can be used again. In addition, we consider an hypothesis test for the continuity of a SETAR model. A simulation study examines small sample performance and an application illustrates how the proposed methodology works in practice.     

Indirect inference in structural econometric models

This paper considers parametric inference in a wide range of structural econometric models. It illustrates how the indirect inference principle can be used in the inference of these models. Specifically, we show that an ordinary least squares (OLS) estimation can be used as an auxiliary model, which leads to a method that is similar in spirit to a two-stage least squares (2SLS) estimator. Monte Carlo studies and an empirical analysis of timber sale auctions held in Oregon illustrate the usefulness and feasibility of our approach.     

Specification and inference in linear models

In the framework of I.I.D. sampling, a general class of linear models is analyzed. Incidental parameters are shown to naturally arise in this class of models. More fundamentally, special attention is paid to the high dimensionality of the parameter space. The objective of the paper is to offer a strategy for progressively specifying a model within that class of linear models. By so doing, we aim at displaying the precise role of each assumption, at offering alternatives to unnecessarily restrictive specifications, and, thereby, at improving the robustness of the inference procedures we discuss. Decompositions of the inference process are obtained through a systematic use of (Bayesian) cuts. Maximum Likelihood Estimation and Bayesian Inference are discussed.An objective of the progressive specification is to preserve the computational tractability and the interpretability of the procedures we develop by relying on known properties of the usual multivariate regression model.     

Bootstrap J tests of nonnested linear regression models

The J test for nonnested regression models often overrejects very severely as an asymptotic test. We provide a theoretical analysis which explains why and when it performs badly. This analysis implies that, except in certain extreme cases, the J test will perform very well when bootstrapped. Using several methods to speed up the simulations, we obtain extremely accurate Monte Carlo results on the finite-sample performance of the bootstrapped J test. These results fully support the predictions of our theoretical analysis, even in contexts where the analysis is not strictly applicable.     

A sequential Monte Carlo approach to inference in multiple‐equation Markov‐switching models

Vector autoregressions with Markov‐switching parameters (MS‐VARs) offer substantial gains in data fit over VARs with constant parameters. However, Bayesian inference for MS‐VARs has remained challenging, impeding their uptake for empirical applications. We show that sequential Monte Carlo (SMC) estimators can accurately estimate MS‐VAR posteriors. Relative to multi‐step, model‐specific MCMC routines, SMC has the advantages of generality, parallelizability, and freedom from reliance on particular analytical relationships between prior and likelihood. We use SMC's flexibility to demonstrate that model selection among MS‐VARs can be highly sensitive to the choice of prior.     

Likelihood inference in some finite mixture models

Parametric mixture models are commonly used in applied work, especially empirical economics, where these models are often employed to learn for example about the proportions of various types in a given population. This paper examines the inference question on the proportions (mixing probability) in a simple mixture model in the presence of nuisance parameters when sample size is large. It is well known that likelihood inference in mixture models is complicated due to (1) lack of point identification, and (2) parameters (for example, mixing probabilities) whose true value may lie on the boundary of the parameter space. These issues cause the profiled likelihood ratio (PLR) statistic to admit asymptotic limits that differ discontinuously depending on how the true density of the data approaches the regions of singularities where there is lack of point identification. This lack of uniformity in the asymptotic distribution suggests that confidence intervals based on pointwise asymptotic approximations might lead to faulty inferences. This paper examines this problem in details in a finite mixture model and provides possible fixes based on the parametric bootstrap. We examine the performance of this parametric bootstrap in Monte Carlo experiments and apply it to data from Beauty Contest experiments. We also examine small sample inferences and projection methods.     

Consistent inference in fixed-effects stochastic frontier models

The classical stochastic frontier panel data models provide no mechanism to disentangle individual time invariant unobserved heterogeneity from inefficiency. Greene (2005a, b) proposed the so-called “true” fixed-effects specification that distinguishes these two latent components. However, due to the incidental parameters problem, his maximum likelihood estimator may lead to biased variance estimates. We propose two alternative estimators that achieve consistency for $n\to \infty$ with fixed $T$. Furthermore, we extend the Chen et al. (2014) results providing a feasible estimator when the inefficiency is heteroskedastic and follows a first-order autoregressive process. We investigate the behavior of the proposed estimators through Monte Carlo simulations showing good finite sample properties, especially in small samples. An application to hospitals’ technical efficiency illustrates the usefulness of the new approach.