Modern Econometric Analysis: Surveys on Recent Developments PDF
Preview Modern Econometric Analysis: Surveys on Recent Developments
Modern Econometric Analysis Surveys on Recent Developments Olaf Hiibler • Joachim Frohn (Editors) Modern Econometric Analysis Surveys on Recent Developments With 8 Figures and 11 Tables Springer Professor Dr. Olaf Hiibler University of Hannover Empirical Economic Research and Econometrics Konigsworther Platz 1 30167 Hannover Germany [email protected] Professor Dr. Joachim Frohn University of Bielefeld Universitatsstrafie 25 33615 Bielefeld Germany [email protected] First published in Allgemeines Statistisches ArchiVy Vol 90, Issue 1, 2006 ISBN-10 3-540-32692-8 Springer Berlin Heidelberg New York ISBN-13 978-3-540-32692-2 Springer Berlin Heidelberg New York Cataloging-in-Publication Data Library of Congress Control Number: 2006923354 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illus trations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springeronline.com © Springer Berhn • Heidelberg 2006 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publica tion does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Hardcover-Design: Erich Kirchner, Heidelberg SPIN 11680970 42/3153-5 4 3 2 1 0 - Printed on acid-free paper Preface The importance of empirical economics and econometric methods has greatly in creased during the last 20 years due to the availability of better data and the improved performance of computers. In an information-driven society such as ours we need quickly to obtain complete and convincing statistical results. This is only possible if the appropriate econometric methods are applied. Traditional econometric analysis concentrates on classical methods which are far from suitable for handling actual economic problems. They can only be used as a starting point for students to learn basic econometrics and as a reference point for more advanced methods. Modern Econometrics tries to develop new approaches from an economic perspective. A consequence is that we have less of a unified econometric theory than in former times. Specific branches which require specific methods have been established. Nowadays, nobody has complete knowledge of every area of econometrics. If someone is interested to learn more about a field, relatively unknown to them, they will require support. This volume is intended to be both an introduction to new econometric methods and a reference for professional econometricians and graduate students. The book provides concise surveys have been carefully selected of many relevant new develop ments in econometrics. Nevertheless, the contributions are selective. We have asked leading German econometricians to write papers about their specific research fields with emphasis on their own area of interest. Nearly all of them have responded with enthusiasm and we are very pleased with the outcome. Hopefully, the reader will share this positive impression. The papers cover the methods used in simultaneous equation models over modern time series, the use of duration and panel data analy sis, microeconometrics, and specific data problems. The reader will find discussion of the benefits and pitfalls as well as the statistical properties of the methods pre sented. The authors emphasize the advantages and disadvantages. They outline the progress which has been made is made in the last few years and the problems which remain unsolved. It is not our intention to demonstrate the workability of the methods by specific applications. For this purpose the reader should consult other publications. Nevertheless, in some chapters advices can be found. All of the contributions in this book are also published in the Journal of the German Statis tical Society (Allgemeines Statistisches Archiv), Volume 90 (2006), Issue 1. We would Hke to thank all the authors who have produced a superb series of papers and the referees who have contributed to a lot of improvements. The support of the German Statistical Society, particularly of Karl Mosler, the president of the society, is gratefully acknowledged. Last but not least we thank the Springer Pub- VI Preface lishers and especially Werner A. Miiller for their cooperation, and their prompt and professional editorial assistance. Olaf Hiibler, Hannover Joachim Prohn, Bielefeld March 2006 Contents 1 Developments and New Dimensions in Econometrics 1 Olaf Hiibler, Joachim Prohn 2 Large Scale Simultaneous Structural Models 7 Pu Chen, Joachim Prohn 3 Dynamic Factor Models 25 Jorg Breitung, Sandra Eickmeier 4 Unit Root Testing 41 Jiirgen Welters, Uwe Hassler 5 Autoregressive Distributed Lag Models and Cointegration 57 Uwe Hassler, Jiirgen Welters 6 Cointegrated Structural VAR Analysis 73 Helmut Liitkepehl 7 Econometric Analysis of High Frequency Data 87 Helmut Herwartz 8 Using Quantile Regression for Duration Analysis 103 Bernd Fitzenberger, Ralf A. Wilke 9 Multilevel and Nonlinear Panel Data Models 119 Olaf Hiibler 10 Nonparametric Models and Their Estimation 137 Goran Kauermann 11 Microeconometric Models and Anonymized Micro Data 153 Gerd Ronning 12 Ordered Response Models 167 Stefan Bees, Rainer Winkelmann 13 Measurement Error Models and Methods 183 Hans Schneewei£, Thomas August in VIII Contents 14 Microeconoraetric Estimation of Treatment Effects 199 Marco Caliendo, Reinhard Hujer 15 Survey Item Nonresponse and its Treatment 215 Susanne Rassler, Regina T. Riphahn 1 Developments and New Dimensions in Econometrics Olaf Hiibler"^ and Joachim Prohn^ ^ Institute of Empirical Economic Research, University of Hannover [email protected] ^ Faculty of Economics, University of Bielefeld [email protected] Summary. This book presents 14 papers with surveys on the development and new topics in econometrics. The articles aim to demonstrate how German econo- metricians see the discipline from their specific view. They briefly describe the main strands and emphasize some recent methods. 1.1 Introduction 75 years ago on the 30th of December in 1930 the Econometric Society was founded and two years later the Society has decided to establish its own journal 'Economet- rica'. This was the birth of econometrics, but the roots lie in the mathematical and statistical economics of the nineteenth century. The 1930s were determined by the enthusiasm of an international group of young people. National style and theoreti cal allegiance seemed to matter less than their common methodological programme. Ragnar Frisch (1970, p. 152), the first editor of Econometrica, describes this feeling when he talks about the First European Meeting of the Econometric Society in 1931 in Lausanne: 'We, the Lausanne people, were indeed so enthusiastic all of us about the new venture, and so eager to give and take, that we had hardly time to eat when we sat together at lunch or at dinner with all our notes floating around on the table to the despair of the waiters'. The first years can definitely be characterized by mathematical models and methods to describe macroeconomic problems and a lack of suitable data sets. Since this time we have observed a rapid and sustainable development. Was it a success story or a chain of frustrations? Both is true. On the one hand the enthusiasm has vanished, the belief that large econometric models can completely explain the economic development has given way to a more skeptical view. The strategy of the Cowles Commission was criticized and alterna- 1 Developments and New Dimensions in Econometrics tive approaches were presented as 'general to specific modelling' by Hendry (1980), 'sensitivity analysis' by Learner (1985) or 'reduced form vector antoregressive mod els' by Sims (1980). Forecasts are not always satisfactory. Many researchers believe that it is not really possible to estimate structural models. Summers (1991), for example, is convinced that applied econometrics is ineffective, not more than ex plorative data analysis can be done by econometrics. Nevertheless, many econome- tricians still have the objective to estimate structural models. Some of the following contributions demonstrate the progress in this field, especially the first and the fifth paper. On the other hand the data situation has significantly improved. Not only are aggregated yearly time series available, but quarterly, monthly, weekly and even daily or tick-by-tick data can be used. Furthermore, large cross-sectional, panel, duration and multilevel data are the basis of many applied econometric studies. Computers give us the chance to process such large information. This was and is a good motivation to develop estimation methods and testing procedures for these specific data constellations. In particular progress in time series analysis has been made, since this field is considered in the perspective of econometric regression models. Nowadays, unit roots and co-integration investigations are standard in single equations models and vector antoregressive models. This means non-stationary in contrast to traditional time series are feasible; the bridge is made between economic theory of equilibria and the econometricians, whose models concentrated on the short-run dynamics, and poor performance of simultaneous macroeconometric models has improved. Recent developments tackle the problem for structural multivariate models, but seasonal models are also considered and structural shifts are in discussion. Econometricians also try to benefit from traditional multivariate statistical methods, like factor analysis and combine these with modern time series methods. While in the 1930's and over a period of 40 years macroeconometric models with many equations have dominated our discipline, in the 1970's the situation changed. The beginning of this new phase is characterized by the new interest in microe- conomic questions and the availability of large cross-sectional data sets, but in applied econometrics the old methods were still used. A little later a revolution of new approaches started in many fields and is still continuing. First, the range of estimation methods extends enormously. Besides the traditional least squares, max imum likelihood methods and methods of moments we find several further methods in the literature: instrumental variable methods, generalized methods of moments, quasi maximum likelihood, quantile estimators, extremum estimator, generalized estimating equations, simulation based estimators. Non- and semi-parametric pro cedures are developed. Second, researchers pay more attention to testing. Especially the spectrum of specifications tests is extended. Incorrect omission of regressors, inclusion- of other variables that are proxies for the omitted variables and a false functional form of the econometric relationship are leading sources of biased estima tions. Third, nonlinear approaches become more relevant. Specific microeconomic data lead to binary, multinomial and ordered probit or logit, count data, duration and tobit models. All of them have a specific nonlinear character and the problem of unobserved heterogeneity has to be solved. Fourth, econometricians analyze specific data problems in more detail. These include measurement errors, anonymization of 1.2 Contributions data and missing values. Fifth, evaluation methods of poUcy intervention and the measurement of treatment effects are more important in economically difficult pe riods. In Germany, this phase started some years after the German reunification. The main problem is to compare two situations where only one is observable. 1.2 Contributions This book contains 14 contributions. Most of the new developments are more or less intensively reviewed. The first paper, presented by Pu Chen and Joachim Prohn, is in some sense a reha bilitation of large scale structural macro econometric models. The authors emphasize the relevance for empirical economics. Apart from traditional aspects like identifica tion and estimation of simultaneous approaches the paper integrates non-stationary variables and co-integration. In detail, statistical inference in large systems is dis cussed. The paper argues that until now the modern statistical inference is not fully integrated into the simultaneous structural approach. Chen and Prohn describe the state of the art of large scale simultaneous econometric models, identify the major unsolved problems and suggest a combined data-theory strategy to look for the specification of the models. The second paper also focusses on empirical macroeconomic models. Jorg Breitung and Sandra Eickmeier discuss large dimensional dynamic factor models which have recently become popular in this field, especially in macroeconomic policy analy sis. The basic idea is to reduce the dimensionality. The large number of available macroeconomic variables is combined with factors that should be interpret able. The paper gives a short survey on factor models and considers different procedures for determining the number of factors. The main new aspect is the introduction of dy namic factor models and procedures to estimate the innovations (error terms) of the factors where VAR models are used. Furthermore, the authors present an overview of economic applications and add their own empirical contribution to co-movements in Europe. Jiirgen Wolters and Uwe Hassler address problems of unit root testing, which nowa days is the starting point of most empirical studies of time series. First, the most popular approach, the Dickey-Fuller and the augmented Dickey-Fuller test includ ing the distribution of the test statistics is presented. Additionally, the authors describe the selection of the lag length, the treatment of the deterministic part in the augmented Dickey-Fuller regression and problems with the variation of time span and frequency of observations. Most important is the overview to unit root testing under structural shifts. Consequences of ignoring breaks, suggestions of cor rections under situations with breaks, smooth and multiple breaks are the content of this section. Co-integration in single equation models within an autoregressive distributed lag framework is the topic of the next contribution. Uwe Hassler and Jiirgen Wolters start their paper with different representations of autoregressive distributed lag model including the error correction model. Inference to a co-integration vector and CO-integration testing follows using error correction models. Several approaches and
The list of books you might like
