In high-dimensional (HD) sparse linear regression, parameter selection and estimation are addressed using a constraint $l_0$ on the direction of the parameter vector. We begin by establishing a general result that identifies this direction through the leading generalized eigenspace of specific measurable matrices. Using this result, we propose a novel approach to the selection of the best subsets by solving an empirical generalized eigenvalue problem to estimate the direction of the HD parameter. We then introduce a new estimator based on the RIFLE algorithm, providing a non-asymptotic bound for the estimation risk, minimax convergence, and a central limit theorem. Simulations demonstrate the superiority of our method over existing $l_0$ -constrained estimators.

This article proposes and studies two Huber-type estimation approaches, namely, the Huber instrumental variable (IV) estimation and the Huber generalized method of moments (GMM) estimation, for a spatial autoregressive model. We establish the consistency, asymptotic distributions, finite sample breakdown points, and influence functions of these estimators. Simulation studies show that compared to the corresponding traditional estimators (the two-stage least squares estimator, the best IV estimator, and the GMM estimator), our estimators are more robust when the unknown disturbances are long-tailed, and our estimators only lose a little efficiency when the disturbances are short-tailed. Moreover, the Huber GMM estimator also outperforms several robust estimators in the literature. Finally, we apply our estimation method to investigate the impact of the urban heat island effect on housing prices. A package is published on GitHub for practitioners to use in their empirical studies.

A general asymptotic theory is established for sample cross moments of nonstationary time series, allowing for long-range dependence and local unit roots. The theory provides a substantial extension of earlier results on nonparametric regression that include near-cointegrated nonparametric regression as well as spurious nonparametric regression. Many new models are covered by the limit theory, among which are functional coefficient regressions in which both regressors and the functional covariate are nonstationary. Simulations show finite sample performance matching well with the asymptotic theory and having broad relevance to applications, while revealing how dual nonstationarity in regressors and covariates raises sensitivity to bandwidth choice and the impact of dimensionality in nonparametric regression. An empirical example is provided involving climate data regression to assess Earth’s climate sensitivity to CO $_2$ , where nonstationarity is a prominent feature of both the regressors and covariates in the model. To our knowledge, this application is the first nonparametric empirical analysis to assess potential nonlinear impacts of CO $_2$ on Earth’s climate while allowing for nonstationarity in both the regressors and covariates.

We develop a methodology for conducting inference on extreme quantiles of unobserved individual heterogeneity (e.g., heterogeneous coefficients and treatment effects) in panel data and meta-analysis settings. Inference is challenging in such settings: only noisy estimates of heterogeneity are available, and central limit approximations perform poorly in the tails. We derive a necessary and sufficient condition under which noisy estimates are informative about extreme quantiles, along with sufficient rate and moment conditions. Under these conditions, we establish an extreme value theorem and an intermediate order theorem for noisy estimates. These results yield simple optimization-free confidence intervals (CIs) for extreme quantiles. Simulations show that our CIs have favorable coverage and that the rate conditions matter for the validity of inference. We illustrate the method with an application to firm productivity differences across areas of varying population density. By analyzing the left tails of the productivity distributions, we find no evidence of stronger firm selection in more densely populated areas.

Pretesting for exogeneity has become routine in many empirical applications involving instrumental variables (IVs) to decide whether the ordinary least squares or IV-based method is appropriate. Guggenberger (2010a, Econometric Theory , 26, 369–382) shows that the second-stage test – based on the outcome of a Durbin-Wu-Hausman-type pretest in the first stage – exhibits extreme size distortion, with asymptotic size equal to 1 when the standard critical values are used. In this paper, we first show that both conditional and unconditional on the data, standard wild bootstrap procedures are invalid for two-stage testing. Second, we propose an identification-robust two-stage test statistic that switches between OLS-based and weak-IV-robust statistics. Third, we develop a size-adjusted wild bootstrap approach for our two-stage test that integrates specific wild bootstrap critical values with an appropriate size-adjustment method. We establish uniform validity of this procedure under conditional heteroskedasticity or clustering in the sense that the resulting tests achieve correct asymptotic size, regardless of whether the identification is strong or weak. Our procedure is especially valuable for empirical researchers facing potential weak identification. In such settings, its power advantage is notable: whereas weak-IV-robust methods maintain correct size but often suffer from relatively low power, our approach achieves better performance.

This article studies the principal component analysis (PCA) estimation of weak factor models with sparse loadings. We uncover an intrinsic near-sparsity preservation property for the PCA estimators of loadings, which comes from the approximately (block) upper triangular structure of the rotation matrix. It suggests an asymmetric relationship among factors: the sparsity of the rotated loadings for a stronger factor can be contaminated by the loadings from weaker ones, but the sparsity of the rotated loadings of a weaker factor is almost unaffected by the loadings of stronger ones. Then, we propose a simple alternative to the existing penalized approaches to sparsify the loading estimators by screening out the small PCA loading estimators directly, and construct consistent estimators for factor strengths. The proposed estimators perform well in finite samples, as shown by a set of Monte Carlo simulations.

Panel data often contain stayers and slow movers . The literature proposes an estimator for the average partial effects (APEs) for this setting without a formal theory. The literature is also silent about inference in the presence of stayers and many slow movers. We contribute to this state of the art. First, we develop an asymptotic theory to guarantee that such an estimator is consistent in the presence of stayers and slow movers. Second, we propose its standard error. Third, we relax the existing assumption to allow for “many” slow movers. Fourth, we generalize the existing estimator. Fifth, we establish that this generalized estimator can achieve larger extents of bias reduction and hence faster convergence rates. Simulation studies demonstrate that the conventional 95% confidence interval covers the true value of the APE with 37%–93% frequencies whereas our proposed one achieves 93%–96% coverage frequencies. Using the U.S. Panel Study of Income Dynamics, we find that estimates of the marginal propensity to consume based on our generalized estimator remarkably differ in values from those of the existing estimators. Moreover, the generalized estimator achieves more than three times as small standard errors as those of the existing robust estimator.

Abstract This article provides a general asymptotic theory for mildly explosive autoregression. We confirm that Cauchy limit theory remains invariant across a broad range of error processes, including general linear processes with martingale difference innovations, stationary causal processes, and nonlinear autoregressive time series, such as threshold autoregressive and bilinear models. Our results unify the Cauchy limit theory for long memory, short memory, and anti-persistent innovations within a single framework. Notably, we demonstrate that in the presence of anti-persistent innovations, the Cauchy limit theory may be violated when the regression coefficient approaches the local-to-unity range. Additionally, we explore extensions to models with varying drift, which is of significant interest in its own right.

This article develops a new test to detect changes in generalized autoregressive conditionally heteroscedastic (GARCH(1,1)) processes without imposing a stationary assumption. Specifically, the procedure tests the null hypothesis of a GARCH process with constant parameters, either in (strictly) stationary or explosive regimes, against the alternative hypothesis of parameter changes. We derive the limiting distribution of the test statistics and establish their asymptotic consistency. Monte Carlo simulations show that the proposed test has good size control and high power. We demonstrate a prototype application on a small group of stocks and report a further extensive application to more than ten thousand U.S. stocks.