When values of regressors are symmetrically disposed, many M-estimators in a wide class of models have a reflection property, namely, that as the signs of the coefficients on regressors are reversed, their estimators' sampling distribution is reflected about the origin. When the coefficients are zero, sign reversal can have no effect. So in this case, the sampling distribution of regression coefficient estimators is symmetric about zero, the estimators are median unbiased and, when moments exist, the estimators are exactly uncorrelated with estimators of other parameters. The result is unusual in that it does not require response variates to have symmetric conditional distributions. It demonstrates the potential importance of covariate design in determining the distributions of estimators, and it is useful in designing and interpreting Monte Carlo experiments. The result is illustrated by a Monte Carlo experiment in which maximum likelihood and symmetrically censored least-squares estimators are calculated for small samples from a censored normal linear regression, Tobit, model.

This paper considers the unit root tests in models with structural change. Particular attention is given to their dependency on the limiting ratios of the subsample sizes between breaks. The dependency is analyzed in detail, and the invariant testing procedure based on a transformed model is developed. The required transformation is essentially identical to the generalized least-squares correction for heteroskedasticity. The limiting distributions of the new tests do not depend on the relative sizes of the subsamples and are shown to be simple mixtures of the limiting distributions of the corresponding tests from the independent unit root models without structural change.

Two issues have come up in the specification of a prior in the Bayesian analysis of time series with possible unit roots. The first issue deals with the singularity that is due to the local identification problem of the unconditional mean of an AR(1) process in the limit of a random walk. This singularity problem is related to the difference between a structural parameterization and the linear reduced form in a standard regression model with fixed regressors. The second is related to the time series nature of the regressor in an AR(1) model. In this paper we will concentrate on the parameterization issue. First, it is shown that the posterior of the autoregressive parameter can be very sensitive to the degree of prior dependence between the unconditional mean and the autocorrelation parameter. Second, the time series nature of the problem suggests a particular form of this dependence.

We prove -consistency and asymptotic normality of a generalized semiparametric regression estimator that includes as special cases Ichimura's semiparametric least-squares estimator for single index models, and the estimator of Klein and Spady for the binary choice regression model. Two function expansions reveal a type of U-process structure in the criterion function; then new U-process maximal inequalities are applied to establish the requisite stochastic equicontinuity condition. This method of proof avoids much of the technical detail required by more traditional methods of analysis. The general framework suggests other -consistent and asymptotically normal estimators.

This paper proposes a general Bayesian framework for distinguishing between trend- and difference-stationarity. Usually, in model selection, we assume that all of the data were generated by one of the models under consideration. In studying time series, however, we may be concerned that the process is changing over time, so that the preferred model changes over time as well. To handle this possibility, we compute the posterior probabilities of the competing models for each observation. This way we can see if different segments of the series behave differently with respect to the competing models. The proposed method is a generalization of the usual odds ratio for model discrimination in Bayesian inference. In application, we employ the Gibbs sampler to overcome the computational difficulty. The procedure is illustrated by a real example.

This paper provides a simple estimation method for an error component regression model with general MA(q) remainder disturbances. The estimation method utilizes the transformation derived by Baltagi and Li [3] for an error component model with autoregressive remainder disturbances, and a standard orthogonalizing algorithm for the general MA(q) model. This estimation method is computationally simple utilizing only least-squares regressions. This is important for panel data regressions where brute force GLS is in many cases not feasible.This estimation method performs well relative to true GLS in Monte-Carlo experiments.

This paper considers the distribution of the Dickey-Fuller test in a model with non-zero initial value and drift and trend. We show how stochastic integral representations for the limiting distribution can be derived either from the local to unity approach with local drift and trend or from the continuous record asymptotic results of Sørensen [29]. We also show how the stochastic integral representations can be utilized as the basis for finding the corresponding characteristic functions via the Fredholm approach of Nabeya and Tanaka [16,17], This “link” between those two approaches may be of general interest. We further tabulate the asymptotic distribution by inverting the characteristic function. Using the same methods, we also find the characteristic function for the asymptotic distribution for the Schmidt-Phillips [26] unit root test. Our results show very clearly the dependence of the various tests on the initial value of the time series.

In this paper, we derive an instrumental variables estimator for the generalized Hausman and Taylor [4] model with a two-way error component. We also extend the work of Breusch, Mizon, and Schmidt [2] to this model and derive two additional instrumental variables estimators. Consistency and asymptotic normality of the parameter estimator is demonstrated, allowing both the number of time periods as well as the number of individuals to increase, their ratio bounded. Consistent estimates of the variance components are discussed.

In this comment on Gurmu and Trivedi's “Variable Augmentation Specification Tests in the Linear Exponential Family,” I show how their generalized linear model (GLM) approach relates to other work in econometrics on specification testing in the linear exponential family. In addition to shedding light on the relationship between the statistics and econometrics literatures on testing in quasi-likelihood frameworks, this comparison reveals some important limitations of GLM as a general framework for devising specification tests.

This paper takes up Bayesian inference in a general trend stationary model for macroeconomic time series with independent Student-t disturbances. The model is linear in the data, but nonlinear in parameters. An informative but nonconjugate family of prior distributions for the parameters is introduced, indexed by a single parameter that can be readily elicited. The main technical contribution is the construction of posterior moments, densities, and odd ratios by using a six-step Gibbs sampler. Mappings from the index parameter of the family of prior distribution to posterior moments, densities, and odds ratios are developed for several of the Nelson–Plosser time series. These mappings show that the posterior distribution is not even approximately Gaussian, and they indicate the sensitivity of the posterior odds ratio in favor of difference stationarity to the choice of the prior distribution.

This paper considers some finite sample properties of the partially restricted reduced form estimators in a general (n + 1) endogenous variable model. In particular, the characteristic functions, density functions, and moments are examined for both the vector of estimators and a linear combination. The approach utilizes both invariant polynomials of matrix argument (see Chikuse and Davis [4] and Davis [6]) and fractional calculus techniques (see Phillips [20,25,26]).

In this paper, we calculate Jeffreys prior for an AR(1) process with and without a constant and a time trend when using the exact likelihood function. We show how this prior can be calculated for the explosive region, even though the unconditional variance of the process is infinite. The calculations lend additional support to the Schotman-van Dijk [6] procedure for restricting the location and the variance of the time trend coefficient. The results show that flat priors are reasonable for the nonexplosive region in an AR(1) without a constant and a time trend where the variance is known and the initial observation is zero, i.e., for the special case studied by Sims and Uhlig [7]. Differences to a flat prior analysis remain in particular for nonzero initial observations, however. For the explosive region, the unconditional prior diverges as the root diverges, supporting findings by Phillips [4]. This paper thus provides a useful perspective as well as some reconciliation for the different stands taken in the literature about priors and Bayesian inference for potentially nonstationary time series.

This paper explores the robustness of minimum distance (GMM) estimators focusing particularly on the effect of intermediate covariance matrix estimation on final estimator performance. Asymptotic expansions to order Op(n−3/2) are employed to construct O(n−2) expansions for the variance of estimators constructed from preliminary least-squares and general M-estimators. In the former case, there is a rather curious robustifying effect due to estimation of the Eicker-White covariance matrix for error distributions with sufficiently large kurtosis.

This paper summarizes recent Bayesian research on unit roots for the applied macroeconomist in the way Campbell and Perron [8] summarized the classical unit roots perspective. The appropriate choice of a prior is discussed. In recognizing a consensus distaste for explosive roots, I find the popular Normal-Wishart priors centered at the unit root to be reasonable provided they are modified by concentrating the prior mass for the time trend coefficient toward zero as the largest root approaches unit from below. I discuss that the tails of the predictive density can be sensitive to the prior treatment of explosive roots. Because the focus of an investigation often is on a particular persistence property or medium-term forecasting property of the data, I conclude that Bayesian methods often deliver natural answers to macroeconomic questions.

This paper treats the theory of identification presented in Haavelmo's classic work, The Probability Approach in Econometrics. This was the first identification theory for stochastic models to be developed in econometrics. The paper presents a detailed commentary on Haavelmo's analysis. It also examines the development of Haavelmo's theory from Frisch's analysis of multicollinearity and also the relationship between Haavelmo's analysis and later work on identification.