The purpose of this paper is to describe the performance of generalized empirical likelihood (GEL) methods for time series instrumental variable models specified by nonlinear moment restrictions as in Stock and Wright (2000, Econometrica 68, 1055–1096) when identification may be weak. The paper makes two main contributions. First, we show that all GEL estimators are first-order equivalent under weak identification. The GEL estimator under weak identification is inconsistent and has a nonstandard asymptotic distribution. Second, the paper proposes new GEL test statistics, which have chi-square asymptotic null distributions independent of the strength or weakness of identification. Consequently, unlike those for Wald and likelihood ratio statistics, the size of tests formed from these statistics is not distorted by the strength or weakness of identification. Modified versions of the statistics are presented for tests of hypotheses on parameter subvectors when the parameters not under test are strongly identified. Monte Carlo results for the linear instrumental variable regression model suggest that tests based on these statistics have very good size properties even in the presence of conditional heteroskedasticity. The tests have competitive power properties, especially for thick-tailed or asymmetric error distributions.This paper is a revision of Guggenberger's job market paper “Generalized Empirical Likelihood Tests under Partial, Weak, and Strong Identification.” We are thankful to the editor, P.C.B. Phillips, and three referees for very helpful suggestions on an earlier version of this paper. Guggenberger gratefully acknowledges the continuous help and support of his adviser, Donald Andrews, who played a prominent role in the formulation of this paper. He thanks Peter Phillips and Joseph Altonji for their extremely valuable comments. We also thank Vadim Marner for help with the simulation section and John Chao, Guido Imbens, Michael Jansson, Frank Kleibergen, Marcelo Moreira, Jonathan Wright, and Motohiro Yogo for helpful comments. Aspects of this research have been presented at the 2003 Econometric Society European Meetings; York Econometrics Workshop 2004; Seminaire Malinvaud; CREST-INSEE; and seminars at Albany, Alicante, Austin (Texas), Brown, Chicago, Chicago GSB, Harvard/MIT, Irvine, ISEG/Universidade Tecnica de Lisboa, Konstanz, Laval, Madison (Wisconsin), Mannheim, Maryland, NYU, Penn, Penn State, Pittsburgh, Princeton, Rice, Riverside, Rochester, San Diego, Texas A&M, UCLA, USC, and Yale. We thank all the seminar participants. Guggenberger and Smith received financial support through a Carl Arvid Anderson Prize Fellowship and a 2002 Leverhulme Major Research Fellowship, respectively.

This paper proposes a test of the rank of the submatrix of β, where β is a cointegrating matrix. In addition, the submatrix of β⊥, an orthogonal complement to β, is investigated. We construct the test statistic by using the eigenvalues of the quadratic form of the submatrix. We show that the test statistic has a limiting chi-square distribution when data are nontrending, whereas for trending data we have to consider a conservative test or other testing procedure that requires the pretest of the structure of the matrix. Finite sample simulations show that, although the simulation settings are limited, the proposed test works well for nontrending data, whereas we have to carefully use the test for trending data because it may become too conservative in some cases.I owe special thanks to two anonymous referees, the co-editor, Pierre Perron, and Taku Yamamoto. All errors are my responsibility. This research was supported by the Ministry of Education, Culture, Sports, Science and Technology under grants-in-aid 13730023 and 14203003.

The paper traces Trygve Haavelmo's training and early career as an econometrician from graduation in economics at the University of Oslo in 1933 until his departure for the United States in 1939. The overwhelming influence on Haavelmo in this period was Ragnar Frisch, whose econometric laboratory at the University of Oslo was Haavelmo's workplace and training ground. In the latter part of the period Haavelmo traveled in Europe, mostly within the network of econometricians Frisch had been instrumental in establishing. Haavelmo's work with Frisch, his interaction with other econometricians and statisticians, and his own scholarly work are set out in some detail, allowing assessment to be made of the development of Haavelmo's econometric ideas. Of particular interest is how far his ideas had evolved by 1939. This paper deals with Frisch and his research program in the early 1930s. Haavelmo's activities are narrated by and large chronologically. A sequel to this paper will deal with Haavelmo's scientific activities while in the United States from 1939 to 1944.I have benefited greatly from extensive, generous, and very insightful advice—and prodding—from three anonymous referees on earlier drafts. I am also most grateful for advice and encouragement from the editor. To one of the anonymous referees I also owe the title. I absolve the referees and take full responsibility for all remaining errors and shortcomings. The paper draws on correspondence and documents from the Rockefeller Archive Center, Tarrytown, New York, and the Frisch Correspondence Files at the National Library of Norway and in addition from the Frisch and Haavelmo archives, currently at the Department of Economics, University of Oslo. I am most grateful to professor emeritus Tore S. Thonstad, who has done a great job of organizing the Frisch and Haavelmo archives. I have also benefited from the comprehensive Frisch bibliography prepared by professor emeritus Kåre N. Edvardsen. I thank J.J. Polak for reminiscences, Yngve Willassen for information, E.S. Jansen and J. Kiviet for a push to submit, and Hilde Bojer and Inger Bjerkodden for encouragement.

The exact mean integrated squared error (MISE) of the nonparametric kernel density estimator is derived for the asymptotically optimal smooth polynomial kernels of Müller (1984, Annals of Statistics 12, 766–774) and the trapezoid kernel of Politis and Romano (1999, Journal of Multivariate Analysis 68, 1–25) and is used to contrast their finite-sample efficiency with the higher order Gaussian kernels of Wand and Schucany (1990Canadian Journal of Statistics 18, 197–204). We find that these three kernels have similar finite-sample efficiency. Of greater importance is the choice of kernel order, as we find that kernel order can have a major impact on finite-sample MISE, even in small samples, but the optimal kernel order depends on the unknown density function. We propose selecting the kernel order by the criterion of minimax regret, where the regret (the loss relative to the infeasible optimum) is maximized over the class of two-component mixture-normal density functions. This minimax regret rule produces a kernel that is a function of sample size only and uniformly bounds the regret below 12% over this density class.

The paper also provides new analytic results for the smooth polynomial kernels, including their characteristic function.This research was supported in part by the National Science Foundation. I thank Oliver Linton and a referee for helpful comments and suggestions that improved the paper.

This issue of Econometric Theory marks 20 full years of publication of the journal ET and the beginning of our third decade. To celebrate the occasion, I invited a group of econometricians, including all the present and past co-editors of ET, to reflect on the future of econometrics. To focus discussion, I selected a theme that relates to a recent trend in the practical use of econometric methodology, one that seems destined to play an increasing role in the future of the subject—automated modeling and inference. Accordingly, this issue contains contributions on various aspects of the theme of automation, introducing the notion of automated discovery, analyzing the validity of selection procedures, and discussing the practical utility of these automated methods in econometric modeling exercises.

This paper studies the asymptotic properties of a nonstationary partially linear regression model. In particular, we allow for covariates to enter the unit root (or near unit root) model in a nonparametric fashion, so that our model is an extension of the semiparametric model analyzed in Robinson (1988, Econometrica 56, 931–954). It is proved that the autoregressive parameter can be estimated at rate N even though part of the model is estimated nonparametrically. Unit root tests based on the semiparametric estimate of the autoregressive parameter have a limiting distribution that is a mixture of a standard normal and the Dickey–Fuller distribution. A Monte Carlo experiment is conducted to evaluate the performance of the tests for various linear and nonlinear specifications.We thank Bruce Hansen, Roger Koenker, Helmut Lütkepohl, Peter Phillips, three referees, and participants of the 8th World Congress of the Econometric Society and the 10th Midwest Econometrics Group Meeting for helpful comments on an earlier version of this paper. This investigation was supported by the University of Kansas General Research Fund allocation 2301789-003.

This paper provides two main new results: the first shows theoretically that large biases and variances can arise when the quasi-maximum likelihood (QML) estimation method is employed in a simple bivariate structure under the assumption of conditional heteroskedasticity; and the second demonstrates how these analytical theoretical results can be used to improve the finite-sample performance of a test for multivariate autoregressive conditional heteroskedastic (ARCH) effects, suggesting an alternative to a traditional Bartlett-type correction. We analyze two models: one proposed in Wong and Li (1997, Biometrika 84, 111–123) and another proposed by Engle and Kroner (1995, Econometric Theory 11, 122–150) and Liu and Polasek (1999, Modelling and Decisions in Economics; 2000, working paper, University of Basel). We prove theoretically that a relatively large difference between the intercepts in the two conditional variance equations, which leads to the two series having correspondingly different volatilities in the restricted case, may produce very large variances in some QML estimators in the first model and very severe biases in some QML estimators in the second. Later we use our bias expressions to propose an LM-type test of multivariate ARCH effects and show through simulations that small-sample improvements are possible, especially in relation to the size, when we bias correct the estimators and use the expected hessian version of the test.Both authors thank H. Wong for providing us with the Gauss program to simulate the Wong and Li (1997) model. We also thank three anonymous referees for extremely helpful comments, and we are grateful for the comments received at seminars given at Cardiff University, Michigan State University, Queen Mary London, University of Exeter, and University of Montreal. We acknowledge gratefully also the financial support from an ESRC grant (award number T026 27 1238). A previous version of this paper appeared as IVIE Working paper 2004-09.

A vector autoregression with deterministic terms and with no restrictions to its characteristic roots is considered. Strong consistency results for the least squares statistics are presented. This extends earlier results where deterministic terms have not been considered. In addition the convergence rates are improved compared with earlier results.Comments from S. Johansen are gratefully acknowledged.

A Bayesian reference analysis of the cointegrated vector autoregression is presented based on a new prior distribution. Among other properties, it is shown that this prior distribution distributes its probability mass uniformly over all cointegration spaces for a given cointegration rank and is invariant to the choice of normalizing variables for the cointegration vectors. Several methods for computing the posterior distribution of the number of cointegrating relations and distribution of the model parameters for a given number of relations are proposed, including an efficient Gibbs sampling approach where all inferences are determined from the same posterior sample. Simulated data are used to illustrate the procedures and for discussing the well-known issue of local nonidentification.The author thanks Luc Bauwens, Anant Kshirsagar, Peter Phillips, Herman van Dijk, four anonymous referees, and especially Daniel Thorburn for helpful comments. Financial support from the Swedish Council of Research in Humanities and Social Sciences (HSFR) grant F0582/1999 and Swedish Research Council (Vetenskapsrådet) grant 412-2002-1007 is gratefully acknowledged. The views expressed in this paper are solely the responsibility of the author and should not be interpreted as reflecting the views of the Executive Board of Sveriges Riksbank.

Our subject is the notion of automated discovery in econometrics. Advances in computer power, electronic communication, and data collection processes have all changed the way econometrics is conducted. These advances have helped to elevate the status of empirical research within the economics profession in recent years, and they now open up new possibilities for empirical econometric practice. Of particular significance is the ability to build econometric models in an automated way according to an algorithm of decision rules that allow for (what we call here) heteroskedastic and autocorrelation robust (HAR) inference. Computerized search algorithms may be implemented to seek out suitable models, thousands of regressions and model evaluations may be performed in seconds, statistical inference may be automated according to the properties of the data, and policy decisions can be made and adjusted in real time with the arrival of new data. We discuss some aspects and implications of these exciting, emergent trends in econometrics.The first version of this paper was written in April 2004 for the 20th Anniversary Issue of Econometric Theory. Helpful comments by the co-editor, Oliver Linton, Benno Pötscher, Brendan Beare, and two referees on the first draft are gratefully acknowledged.

This paper develops an asymptotic theory for integrated and near-integrated time series whose range is constrained in some ways. Such a framework arises when integration and cointegration analyses are applied to time series that are bounded either by construction or because they are subject to control. The asymptotic properties of some commonly used integration tests are discussed; the bounded unit root distribution is introduced to describe the limiting distribution of the sample first-order autoregressive coefficient of a random walk under range constraints. The theoretical results show that the presence of such constraints can lead to drastically different asymptotics. Because deviations from the standard unit root theory are measured through two noncentrality parameters that can be consistently estimated, simple measures of the impact of range constraints on the asymptotic distributions are obtained. Generalizations of standard unit root tests that are robust to the presence of range constraints are also provided. Finally, it is shown that the proposed asymptotic framework provides an adequate approximation to the finite-sample properties of the unit root statistics under range constraints.Partial financial support from Italian MIUR grants is gratefully acknowledged. I thank, without implicating, Pentti Saikkonen (the co-editor), an anonymous referee, Attilio Gardini, Martin Jacobsen, Robert de Jong, Paolo Paruolo, Anders Rahbek, and participants at the 58th European Meeting of the Econometric Society, Stockholm, August 21–24, 2003, for helpful comments. I also thank the Bank of International Settlements for providing the European monetary system exchange rate data.

In this paper, we prove the validity of an Edgeworth expansion to the distribution of the Whittle maximum likelihood estimator for stationary long-memory Gaussian models with unknown parameter . The error of the (s − 2)-order expansion is shown to be o(n−(s−2)/2)—the usual independent and identically distributed rate—for a wide range of models, including the popular ARFIMA(p,d,q) models. The expansion is valid under mild assumptions on the behavior of the spectral density and its derivatives in the neighborhood of the origin. As a by-product, we generalize a theorem by Fox and Taqqu (1987, Probability Theory and Related Fields 74, 213–240) concerning the asymptotic behavior of Toeplitz matrices.

Lieberman, Rousseau, and Zucker (2003, Annals of Statistics 31, 586–612) establish a valid Edgeworth expansion for the maximum likelihood estimator for stationary long-memory Gaussian models. For a significant class of models, their expansion is shown to have an error of o(n−1). The results given here improve upon those of Lieberman et al. in that the results provide an Edgeworth expansion for an asymptotically efficient estimator, as Lieberman et al. do, but the error of the expansion is shown to be o(n−(s−2)/2), not o(n−1), for a broad range of models.

Strong consistency and asymptotic normality are established for the quasi-maximum likelihood estimator for a class of ARCH(q) models. The conditions are that the ARCH process is geometrically ergodic with a moment of arbitrarily small order. Furthermore for consistency, we assume that the second-order moment exists for the nondegenerate rescaled errors and, similarly, that the fourth-order moment exists for asymptotic normality to hold. Contrary to existing literature on (G)ARCH models the parameter space is not assumed to be compact; we only impose a lower bound for the constant term in our parameterization of the conditional variance. It is demonstrated that the general conditions are satisfied for a range of specific models.We are grateful to the editor and the referees for their very helpful and detailed suggestions, which, we believe, improved the paper substantially. We thank Søren T. Jensen for stimulating discussions and Jonathan Dennis for helpful research assistance. Rahbek acknowledges continuing financial support from the Danish Social Sciences Research Council. Kristensen received funding from the Danish Research Agency and the Financial Markets Group, LSE, during this research.

In frontier analysis, most of the nonparametric approaches (free disposal hull [FDH], data envelopment analysis [DEA]) are based on envelopment ideas, and their statistical theory is now mostly available. However, by construction, they are very sensitive to outliers. Recently, a robust nonparametric estimator has been suggested by Cazals, Florens, and Simar (2002, Journal of Econometrics 1, 1–25). In place of estimating the full frontier, they propose rather to estimate an expected frontier of order m. Similarly, we construct a new nonparametric estimator of the efficient frontier. It is based on conditional quantiles of an appropriate distribution associated with the production process. We show how these quantiles are interesting in efficiency analysis. We provide the statistical theory of the obtained estimators. We illustrate with some simulated examples and a frontier analysis of French post offices, showing the advantage of our estimators compared with the estimators of the expected maximal output frontiers of order m.We thank J.P. Florens for helpful discussions and C. Cazals for providing the post office data set. We also are very grateful to the referees for useful suggestions.

Model selection has an important impact on subsequent inference. Ignoring the model selection step leads to invalid inference. We discuss some intricate aspects of data-driven model selection that do not seem to have been widely appreciated in the literature. We debunk some myths about model selection, in particular the myth that consistent model selection has no effect on subsequent inference asymptotically. We also discuss an “impossibility” result regarding the estimation of the finite-sample distribution of post-model-selection estimators.

We discuss the effects of temporal aggregation on the estimation of cointegrating vectors and on testing linear restrictions on this vector. We adopt a discrete time approach and demonstrate, in contrast with the findings of Chambers (2003, Econometric Theory 19, 49–77), who adopts a continuous time approach, that in some situations, when the regressand must be aggregated, systematic sampling is preferable to average sampling for estimation purposes. Like Chambers, we show that the best aggregation scheme for regressors, in terms of asymptotic estimation efficiency, is always average sampling. We also show that different types of aggregation have no influence on the relative size of tests of linear restrictions on the cointegration vector.We thank Soren Johansen, Niels Haldrup, Raquel Waters, the associate editor, and two anonymous referees for their helpful comments. Of course, any remaining error is the responsibility of the authors. The first author gratefully acknowledges the financial support of a Marie Curie Fellowship of the European Community Programme “Improving the Human Research Potential and the Socio-Economic Knowledge Base” under contract HPMF-CT-2002-01662 and the Danish Research Council. The second author gratefully acknowledges the financial support of the Spanish Ministry of Science and Technology SEC2002-01512.

The sampling window method of Hall, Jing, and Lahiri (1998, Statistica Sinica 8, 1189–1204) is known to consistently estimate the distribution of the sample mean for a class of long-range dependent processes, generated by transformations of Gaussian time series. This paper shows that the same nonparametric subsampling method is also valid for an entirely different category of long-range dependent series that are linear with possibly non-Gaussian innovations. For these strongly dependent time processes, subsampling confidence intervals allow inference on the process mean without knowledge of the underlying innovation distribution or the long-memory parameter. The finite-sample coverage accuracy of the subsampling method is examined through a numerical study.The authors thank two referees for comments and suggestions that greatly improved an earlier draft of the paper. This research was partially supported by U.S. National Science Foundation grants DMS 00-72571 and DMS 03-06574 and by the Deutsche Forschungsgemeinschaft (SFB 475).

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 analyze the effects of a very general class of time-varying variances on well-known “stationarity” tests of the I(0) null hypothesis. Our setup allows, among other things, for both single and multiple breaks in variance, smooth transition variance breaks, and (piecewise-) linear trending variances. We derive representations for the limiting distributions of the test statistics under variance breaks in the errors of I(0), I(1), and near-I(1) data generating processes, demonstrating the dependence of these representations on the precise pattern followed by the variance processes. Monte Carlo methods are used to quantify the effects of fixed and smooth transition single breaks and trending variances on the size and power properties of the tests. Finally, bootstrap versions of the tests are proposed that provide a solution to the inference problem.We are grateful to Peter Phillips, a co-editor, and two anonymous referees whose comments on an earlier draft have led to a considerable improvement in the paper.

In this article, starting from continuous-time local level unobserved components models for stock and flow data we derive locally best invariant (LBI) stationarity tests for data available at potentially irregularly spaced points in time. We demonstrate that the form of the LBI test differs between stock and flow variables. In cases where the data are observed at regular intervals throughout the sample we show that the LBI tests for stock and flow data both reduce to the form of the standard stationarity test in the discrete-time local level model. Here we also show that the asymptotic local power of the LBI test increases with the sampling frequency in the case of stock, but not flow, variables. Moreover, for a fixed time span we show that the LBI test for stock (flow) variables is (is not) consistent against a fixed alternative as the sampling frequency increases to infinity. We also consider the case of mixed frequency data in some detail, providing asymptotic critical values for the LBI tests for both stock and flow variables, together with a finite-sample power study. Our results suggest that tests that ignore the infraperiod aspect of the data involve rather small losses in efficiency relative to the LBI test in the case of flow variables but can result in significant losses of efficiency when analyzing stock variables.We are grateful to co-editor Jeff Wooldrige, two anonymous referees, Marcus Chambers, Andrew Harvey, Rod McCrorie, Neil Shephard, and seminar participants at Nuffield College, Oxford, for their helpful comments and suggestions on earlier versions of this paper. We are especially indebted to Marcus Chambers, who suggested the formulation of the observation equations used in our continuous-time unobserved components models for stock and flow variables. All errors are ours.