Search

We propose a novel statistic for testing the structural parameters in Instrumental Variables Regression. The statistic is straightforward to compute and has a limiting distribution that is pivotal with a degrees of freedom parameter that is equal to the number of tested parameters. It therefore differs from the Anderson-Rubin statistic, whose limiting distribution is pivotal but has a degrees of freedom parameter that is equal to the number of instruments, and the Likelihood based, Wald, Likelihood Ratio and Lagrange Multiplier, statistics, whose limiting distributions are not pivotal. We analyze the relationship between the statistic and the concentrated likelihood of the structural parameters and show that its' limiting distribution is not affected by weak instruments. We discuss examples of the non-standard shapes of the asymptotically pivotal confidence sets that can be constructed using the statistic and investigate its power properties. To show its importance for practical purposes, we apply the statistic to the Angrist-Krueger (1991) data and find similar results as in Staiger and Stock (1997).

We propose finite sample tests and confidence sets for models with unobserved and generated regressors as well as various models estimated by instrumental variables methods. The validity of the procedures is unaffected by the presence of identification problems or \"weak instruments\", so no detection of such problems is required. We study two distinct approaches for various models considered by Pagan (1984). The first one is an instrument substitution method which generalizes an approach proposed by Anderson and Rubin (1949) and Fuller (1987) for different (although related) problems, while the second one is based on splitting the sample. The instrument substitution method uses the instruments directly, instead of generated regressors, in order to test hypotheses about the \"structural parameters\" of interest and build confidence sets. The second approach relies on \"generated regressors\", which allows a gain in degrees of freedom, and a sample split technique. For inference about general possibly nonlinear transformations of model parameters, projection techniques are proposed. A distributional theory is obtained under the assumptions of Gaussian errors and strictly exogenous regressors. We show that the various tests and confidence sets proposed are (locally) \"asymptotically valid\" under much weaker assumptions. The properties of the tests proposed are examined in simulation experiments. In general, they outperform the usual asymptotic inference methods in terms of both reliability and power. Finally, the techniques suggested are applied to a model of Tobin’s q and to a model of academic performance., Nous proposons des tests et régions de confiance exactes pour des modèles comportant des variables inobservées ou des régresseurs estimés de même que pour divers modèles estimés par la méthode des variables instrumentales. La validité des procédures proposées n’est pas influencée par la présence de problèmes d’identification ou d’instruments faibles, de sorte que la détection de tels problèmes n’est pas requise pour les appliquer. De façon plus spécifique, nous étudions deux approches différentes pour divers modèles considérés par Pagan (1984). La première est une méthode de substitution d’instruments qui généralise des techniques proposées par Anderson et Rubin (1949) et Fuller (1984) pour des problèmes différents, tandis que la seconde méthode est fondée sur une subdivision de l’échantillon. La méthode de substitution d’instruments utilise directement les instruments disponibles, plutôt que des régresseurs estimés, afin de tester des hypothèses et construire des régions de confiance sur les \"paramètres structuraux\" du modèle. La seconde méthode s’appuie sur des régresseurs estimés, ce qui permet un gain de degrés de liberté, ainsi que sur une technique de subdivision de l’échantillon. Pour faire de l’inférence sur des transformation générales, possiblement non-linéaires, des paramètres du modèle, nous proposons l’utilisation de techniques de projection. Nous fournissons une théorie distributionnelle exacte sous une hypothèse de normalité des perturbations et de régresseurs strictement exogènes. Nous montrons que les tests et régions de confiance ainsi obtenus sont aussi (localement) \"asymptotiquement valides\" sous des hypothèses distributionnelles beaucoup plus faibles. Nous étudions les propriétés des tests proposés dans le cadre d’une expérience de simulation. En général, celles-ci sont plus fiables et ont une meilleure puissance que les techniques traditionnelles. Finalement, les techniques proposées sont appliquées à un modèle du q de Tobin et à un modèle de performance scolaire.

Large shareholders may play an important role for firm policies and performance, but identifying an effect empirically presents a challenge due to the endogeneity of ownership structures. However, unlike other blockholders, individuals tend to hold blocks in corporations that are located close to where they live. Using this fact, we create an instrument – the density of wealthy individuals near a firm’s headquarters – for the presence of a large, non-managerial individual shareholder in a public firm. We show that these shareholders have a large impact on firms. Consistent with theories of large shareholders as monitors, we find that they increase firm profitability, increase dividends, reduce corporate cash holdings, and reduce executive compensation. Consistent with the view that there exist conflicts between large and small owners in public firms, we uncover evidence of substitution toward less tax-efficient forms of distribution (dividends over repurchases). In addition, our analysis shows that large shareholders reduce the liquidity of the firm’s stock.

Mendelian randomization (MR) is a method of exploiting genetic variation to unbiasedly estimate a causal effect in presence of unmeasured confounding. MR is being widely used in epidemiology and other related areas of population science. In this paper, we study statistical inference in the increasingly popular two-sample summary-data MR design. We show a linear model for the observed associations approximately holds in a wide variety of settings when all the genetic variants satisfy the exclusion restriction assumption, or in genetic terms, when there is no pleiotropy. In this scenario, we derive a maximum profile likelihood estimator with provable consistency and asymptotic normality. However, through analyzing real datasets, we find strong evidence of both systematic and idiosyncratic pleiotropy in MR, echoing the omnigenic model of complex traits that is recently proposed in genetics. We model the systematic pleiotropy by a random effects model, where no genetic variant satisfies the exclusion restriction condition exactly. In this case we propose a consistent and asymptotically normal estimator by adjusting the profile score. We then tackle the idiosyncratic pleiotropy by robustifying the adjusted profile score. We demonstrate the robustness and efficiency of the proposed methods using several simulated and real datasets., Comment: 59 pages, 5 figures, 6 tables