We propose nonnested tests for competing conditional moment restriction models using the method of conditional empirical likelihood, recently developed by Kitamura, Tripathi, and Ahn (2004) and Zhang and Gijbels (2003). To define the test statistics, we use the implied conditional probabilities from conditional empirical likelihood, which take into account the full implications of conditional moment restrictions. We propose three types of nonnested tests: the moment-encompassing, Cox-type, and efficient score-encompassing tests. We derive the asymptotic null distributions and investigate their power properties against a sequence of local alternatives and a fixed global alternative. Our tests have distinct global power properties from some of the existing tests based on finite-dimensional unconditional moment restrictions. Simulation experiments show that our tests have reasonable finite sample properties and dominate some of the existing nonnested tests in terms of size-corrected powers.

It is well known that the unrestricted bootstrap estimator of the slope parameter in the random walk model without drift converges to a random distribution. This bootstrap failure is commonly attributed to the discontinuity of the limit distribution of the least-squares estimator in the parameter of interest. We demonstrate by counterexample that this type of continuity is not essential for the validity of the bootstrap nor is it essential that the rate of convergence of the estimator remain constant over the whole parameter space.We thank Don Andrews, Shinichi Sakata, Jonathan Wright, and two anonymous referees for very helpful comments. The views expressed in this paper do not necessarily reflect those of the European Central Bank or its members.

This paper develops an asymptotic theory of nonlinear least squares estimation by establishing a new framework that can be easily applied to various nonlinear regression models with heteroscedasticity. As an illustration, we explore an application of the framework to nonlinear regression models with nonstationarity and heteroscedasticity. In addition to these main results, this paper provides a maximum inequality for a class of martingales, which is of interest in its own right.

This paper considers adaptive hypothesis testing for the fractional differencing parameter in a parametric ARFIMA model with unconditional heteroskedasticity of unknown form. A weighted score test based on a nonparametric variance estimator is proposed and shown to be asymptotically equivalent, under the null and local alternatives, to the Neyman-Rao effective score test constructed under Gaussianity and known variance process. The proposed test is therefore asymptotically efficient under Gaussianity. The finite sample properties of the test are investigated in a Monte Carlo experiment and shown to provide potentially large power gains over the usual unweighted long memory test.

The Hodrick–Prescott (HP) filter has been a popular method of trend extraction from economic time series. However, it is impractical without modification if some observations are not available. This paper improves the HP filter so that it can be applied in such situations. More precisely, this paper introduces two alternative generalized HP filters that are applicable for this purpose. We provide their properties and a way of specifying those smoothing parameters that are required for their application. In addition, we numerically examine their performance. Finally, based on our analysis, we recommend one of them for applied studies.

We show that the slope parameter of the linear quantile regression measures a weighted average of the local slopes of the conditional quantile function. Extending this result, we also show that the slope parameter measures a weighted average of the partial effects for a general structural function. Our results support the use of linear quantile regressions for causal inference in the presence of nonlinearity and multivariate unobserved heterogeneity. The same conclusion applies to linear regressions.

Forecasts based on two information sets, one of which includes the other plus additional causal variables are considered. Given a general cost function of forecast errors, it is shown that the expected cost is smaller for the information set that includes the causal variables.

In this paper, we shall derive the asymptotic expansion for the distribution of the likelihood ratio criterion for a Gaussian autoregressive moving average process under a sequence of local alternative hypotheses converging to the null hypothesis with rate of convergence where n is the sample size. Explicit algebraic formulae are presented for certain special cases, including the ARMA(1,1).

This paper compares several tests for linear and loglinear regression models where both the dependent and independent variables are transformed. It is shown that the Lagrange multiplier test proposed by Godfrey and Wickens (1981, Review of Economic Studies 48, 487–496) in the framework of the Box–Cox regression model has the highest asymptotic power of the compared tests. The extended projection test of MacKinnon, White, and Davidson (1983, Journal of Econometrics 11, 53–70), the test of Bera and McAleer (1983, paper presented to the SSRC Econometric Study Group Conference on Model Specification and Testing, Warwick; 1989, Sankhya B 51, 212–224), and the test of Andrews (1971, Biometrika 58, 249–254) are shown to have asymptotically equivalent powers and to have lower powers than the nonnested test of Cox (1961, Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1, 105–123, Berkeley: University of California Press; 1962, Journal of the Royal Statistical Society B 24, 406–424).

Hylleberg, Engle, Granger, and Yoo (1990, Journal of Econometrics 44, 215–238), Beaulieu and Miron (1993, Journal of Econometrics 55, 305–328), Ghysels, Lee, and Noh (1994, Journal of Econometrics 62, 415–442), Smith and Taylor (1998, Journal of Econometrics 85, 269–288; 1999, Journal of Time Series Analysis 20, 453–476; 1999, Discussion paper 99-15 in economics, University of Birmingham), and Taylor (1998, Journal of Time Series Analysis 19, 349–368) have developed a method of testing for seasonal unit roots of zero and nonzero frequencies. They propose to use t- and F-statistics as criteria that are obtained from an auxiliary regression and find their limiting distributions as the number of observations becomes large. Their limiting distributions are expressed by means of Brownian motions. In this paper the moment generating functions associated with the limiting distributions are derived, and it is shown, as in Nabeya (2000, Econometric Theory 16, 200–230), that the limiting distribution of t is well approximated by a distribution given in Gram–Charlier series. The limiting distribution of F is also well approximated by another type of distribution.

This paper considers estimation and inference for varying-coefficient models with nonstationary regressors. We propose a nonparametric estimation method using penalized splines, which achieves the same optimal convergence rate as kernel-based methods, but enjoys computation advantages. Utilizing the mixed model representation of penalized splines, we develop a likelihood ratio test statistic for checking the stability of the regression coefficients. We derive both the exact and the asymptotic null distributions of this test statistic. We also demonstrate its optimality by examining its local power performance. These theoretical findings are well supported by simulation studies.

Building on work of Hansen (1992, Econometric Theory 8, 489–501), we show weak convergence for power transformations of integrated processes, with possibly serially correlated and heterogeneously distributed increments, to stochastic power integrals. The theory is applicable when testing the unit root or cointegration hypothesis in nonlinear systems by regression-based test statistics.

We consider a model Yt = σtηt in which (σt) is not independent of the noise process (ηt) but σt is independent of ηt for each t. We assume that (σt) is stationary, and we propose an adaptive estimator of the density of ln(σt2) based on the observations Yt. Under a new dependence structure, the τ-dependency defined by Dedecker and Prieur (2005, Probability Theory and Related Fields 132, 203–236), we prove that the rates of this nonparametric estimator coincide with the rates obtained in the independent and identically distributed (i.i.d.) case when (σt) and (ηt) are independent. The results apply to various linear and nonlinear general autoregressive conditionally heteroskedastic (ARCH) processes. They are illustrated by simulations applying the deconvolution algorithm of Comte, Rozenholc, and Taupin (2006, Canadian Journal of Statistics 34, 431–452) to a new noise density.

We discuss the existence and uniqueness of stationary and ergodic nonlinear autoregressive processes when exogenous regressors are incorporated into the dynamic. To this end, we consider the convergence of the backward iterations of dependent random maps. In particular, we give a new result when the classical condition of contraction on average is replaced with a contraction in conditional expectation. Under some conditions, we also discuss the dependence properties of these processes using the functional dependence measure of Wu (2005, Proceedings of the National Academy of Sciences 102, 14150–14154) that delivers a central limit theorem giving a wide range of applications. Our results are illustrated with conditional heteroscedastic autoregressive nonlinear models, Generalized AutoRegressive Conditional Heteroskedasticity (GARCH) processes, count time series, binary choice models, and categorical time series for which we provide many extensions of existing results.

This paper deals with statistical inference for the fixed effects panel data seemingly unrelated partially linear regression model. The model naturally extends the traditional fixed effects panel data regression model to allow for semiparametric effects. Multiple regression equations are permitted, and the model includes the aggregated partially linear model as a special case. A weighted profile least squares estimator for the parametric components is proposed and shown to be asymptotically more efficient than those neglecting the contemporaneous correlation. Furthermore, a weighted two-stage estimator for the nonparametric components is also devised and shown to be asymptotically more efficient than those based on individual regression equations. The asymptotic normality is established for estimators of both parametric and nonparametric components. The finite-sample performance of the proposed methods is evaluated by simulation studies.

We consider the statistical properties of the variance ratio statistic in the context of testing for market efficiency defined by uncorrelated returns. The statistic is then the ratio of the variance of K-period returns to the variance of one-period returns scaled by K. We use a continuous-time asymptotic framework whereby we let the sample size increase to infinity keeping the span of the data fixed. We also let the aggregation parameter K increase such that K/T → κ as T, the sample size, increases to infinity. We consider the limit of the statistic under the null hypothesis and under three alternative hypotheses that have been popular in the finance literature. Our analysis permits us to address size and power issues with respect to κ and the sampling interval used. Our theoretical and simulation results show that power is initially increasing as κ increases but then decreases with further increases in κ. This shows that for any given alternative there exists a value of K relative to T that will maximize power. We thus investigate the properties of a test that is the maximal value of the variance ratio over a range of possible values for K. The importance of the trimming to define this range is highlighted.This paper is drawn from chapter 2 of Vodounou's Ph.D. dissertation at the Université de Montréal (Vodounou, 1997). We thank two referees for useful comments and Tomoyoshi Yabu and Xiaokang Zhu for research assistance. Perron acknowledges financial support from the Social Sciences and Humanities Research Council of Canada (SSHRC), the Natural Sciences and Engineering Council of Canada (NSERC), the Fonds pour la Formation de Chercheurs et l'Aide à la Recherche du Québec (FCAR), and the National Science Foundation (grant SES-0078492).

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.

This article investigates the statistical inference problem of whether a measurement equation is self-consistent in the logarithmic realized GARCH model (log-RealGARCH). First, we provide the sufficient and necessary conditions for the strict stationarity of both the log-RealGARCH model and the log-GARCH-X model. Under these conditions, strong consistency and asymptotic normality of the quasi-maximum likelihood estimators of these two models are obtained. Then, based on the asymptotic results, we propose a Hausman-type self-consistency test for diagnosing the suitability of the measurement equation in the log-RealGARCH model. Finally, the results of simulations and an empirical study are found to accord with the theoretical results.

In this paper we derive the necessary and sufficient conditions for the t-ratio to be Student's t distributed. In particular, it is demonstrated for a special case that under conditions of nonnormality characterized by elliptical symmetry, the t-ratio remains Student's t distributed provided that the random vector forming the t-ratio has a diagonal covariance structure. Our results also show that the findings of Magnus (2002, in A. Ullah, A.T.K. Wan, & A. Chaturvedi (eds.), Handbook of Applied Econometrics and Statistical Inference, 277–285) on the sensitivity of the t-ratio remain invariant in the elliptically symmetric distribution setting. Extension to the linear model is considered. Exact results giving finite sample justification for the t-statistic under nonnormal error terms are derived. Furthermore, the distribution of the F-ratio assuming elliptical errors is examined. Our results reject the argument by Zaman (1996, Statistical Foundations for Econometric Techniques) that nonnormality of disturbances has in general no effect on the F-statistic.The authors thank Anurag Banerjee, Judith Clarke, Kazuhiro Ohtani, and Guohua Zou for their helpful suggestions and advice during the course of this work. In particular, we are very grateful to Judith Clarke for bringing to our attention the unpublished work of D.H. Thomas and to Guohua Zou for his careful reading of an earlier version of this paper. Thanks also go to the co-editor Benedikt Pötscher and two anonymous referees for their valuable comments. The second author gratefully acknowledges financial support from the Hong Kong Research Grant Council.

We revisit in-sample asymptotic analysis extensively used in the realized volatility literature. We show that there are gains to be made in estimating current realized volatility from considering realizations in prior periods. The weighting schemes also relate to Kalman-Bucy filters, although our approach is non-Gaussian and model-free. We derive theoretical results for a broad class of processes pertaining to volatility, higher moments, and leverage. The paper also contains a Monte Carlo simulation study showing the benefits of across-sample combinations.