Prelims

Half Title Page

THE ECONOMETRICS OF COMPLEX SURVEY DATA: THEORY AND APPLICATIONS

Series Page

ADVANCES IN ECONOMETRICS

Series editors: Thomas B. Fomby, R. Carter Hill, Ivan Jeliazkov, Juan Carlos Escanciano, Eric Hillebrand, Daniel L. Millimet, Rodney Strachan, David T. Jacho-Chávez, Alicia Rambaldi

Previous Volumes

Title Page

ADVANCES IN ECONOMETRICS VOLUME 39

THE ECONOMETRICS OF COMPLEX SURVEY DATA: THEORY AND APPLICATIONS

EDITED BY

KIM P. HUYNH

Bank of Canada, Canada

DAVID T. JACHO-CHÁVEZ

Emory University, USA

GAUTAM TRIPATHI

University of Luxembourg, Luxembourg

United Kingdom – North America – Japan – India – Malaysia – China

Copyright Page

Emerald Publishing Limited

Howard House, Wagon Lane, Bingley BD16 1WA, UK

First edition 2019

Copyright © 2019 Emerald Publishing Limited

Reprints and permissions service

Contact: permissions@emeraldinsight.com

No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. Any opinions expressed in the chapters are those of the authors. Whilst Emerald makes every effort to ensure the quality and accuracy of its content, Emerald makes no representation implied or otherwise, as to the chapters’ suitability and application and disclaims any warranties, express or implied, to their use.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-1-78756-726-9 (Print)

E-ISBN 978-1-78756-725-2 (Online)

ISBN: 978-1-78756-727-6 (Epub)

ISSN: 0731-9053 (Series)

List of Contributors

Introduction

The assumption of simple random sampling is widely used in applied research in the social, behavioral and biomedical sciences, as well as in empirical public policy analysis. However, this assumption is seldom true in practice. Stratified and cluster sampling are routinely used by most statistical agencies in the world, and because of budgetary reasons, the actual sampling process may be even more complicated. Correct statistical analysis therefore requires a careful consideration of these complex survey designs when performing estimation and inference.

The papers in this volume of Advances in Econometrics were presented at the “Econometrics of Complex Survey Data: Theory and Applications” conference organized by the Bank of Canada, Ottawa, Canada, from October 19 to 20, 2017. The editors would like to acknowledge the generous financial support provided by the Bank of Canada.

Below is a brief overview of the papers accepted in this volume, grouped into the following four categories: (1) sampling design; (2) variance estimation; (3) estimation and inference and (4) business, household and crime surveys.

Sampling Design

“Can Internet Match High Quality Traditional Surveys? Comparing the Health and Retirement Study and Its Online Version” by Marco Angrisani, Brian Finley and Arie Kapteyn revisit the question of comparability of online and more traditional interview modes by studying differences across Internet-based, face-to-face and phone-based surveys. They find little evidence of mode effects when comparing various outcomes providing support for internet-based surveys.

“Effectiveness of Stratified Random Sampling for Payment Card Acceptance and Usage” by Christopher S. Henry and Tamás Ilyés uses the universe of merchant cash registers in Hungary to assess the effect of stratified random sampling on estimates of payment card acceptance and usage. It compares county, industry, and store size stratifications to mimic the usual stratification criteria for standard merchant surveys. By doing this, they can quantify the effect on estimates of card acceptance for different sample sizes.

Variance Estimation

“Wild Bootstrap Randomization Inference for Few Treated Clusters” by James G. MacKinnon and Matthew D. Webb proposes a bootstrap-based alternative to randomization inference, which mitigates problems of over- or under-rejection in t tests in pure treatment or difference-in-differences settings when the number of clusters is very small.

“Variance Estimation for Survey-weighted Data Using Bootstrap Resampling Methods: 2013 Methods-of-Payment Survey Questionnaire” by Heng Chen and Q. Rallye Shen proposes a bootstrap-resampling method to estimate variability when sampling units are selected through an approximate stratified two-stage sampling design. Their proposed method allows for randomness from both the sampling design and the raking procedure.

Estimation and Inference

“Model Selection Tests for Complex Survey Samples” by Iraj Rahmani and Jeffrey M. Wooldridge extends Vuong’s model selection test (“Likelihood Ratio Tests for Model Selection and Non-Nested Hypothesis,” Econometrica, 1989) to allow for complex survey samples. By using an M-estimation setting, their test applies to general estimation problems including linear and nonlinear least squares, Poisson regression and fractional response models. With cluster samples and panel data, they show how to combine the weighted objective function with a cluster-robust variance estimator, thereby expanding the scope of their test.

“Inference in Conditional Moment Restriction Models When There is Selection Due to Stratification” by Antonio Cosma, Andre V. Kostyrka and Gautam Tripathi shows how to use a smoothed empirical likelihood approach to conduct efficient semiparametric inference in models characterized as conditional moment equalities when data are collected by variable probability sampling.

“Nonparametric Kernel Regression Using Complex Survey Data” by Luc Clair derives the asymptotic properties of a design-based nonparametric kernel-based regression estimator under a combined inference framework involving multivariate mixed data. It also proposes a least squares cross-validation procedure for selecting the bandwidth for both continuous and discrete variables. Simulation results show that the estimator is consistent and that efficiency gains can be achieved by weighting observations by the inverse of their inclusion probabilities if the sampling is endogenous.

“Nearest Neighbor Imputation for General Parameter Estimation in Survey Sampling” by Shu Yang and Jae Kwang Kim studies the asymptotic properties of the nearest neighbor population imputation estimator of population parameters when handling item nonresponse in survey sampling. When estimating a variance, the authors propose a replication variance estimator.

Business, Household and Crime Surveys

Last but not least, “Improving Response Quality with Planned Missing Data: An Application to a Survey of Banks” by Geoffrey R. Gerdes and Xuemei Liu reports a “random blocking” approach to shortening the questionnaires for individual respondents when collecting data on noncash payments by type, cash withdrawals and deposits, and related information in a survey of a population of depository institutions in the United States. Their approach is a special case of multiple matrix sampling and an extension of a split questionnaire or planned missing value design. They find that the proposed blocking approach helped increase unit-level and item-level response for smaller institutions.

“Does Selective Crime Reporting Influence Our Ability to Detect Racial Discrimination in the NYPD’s Stop-and-frisk Program?” by Steven F. Lehrer and Louis-Pierre Lepage uses data from the New York City’s Stop-and-Frisk program to assess the presence of crime type heterogeneity in racial bias and police officer decisions of reported crime type. They find evidence that differences in biases across crime types are substantial while accounting for sample-selection which may arise from conditioning on crime type.

“Survey Evidence on Black Market Liquor in Colombia” by Gustavo J. Canavire-Bacarreza, Alexander L. Lundberg and Alejandra Montoya-Agudelo uses a unique national survey on illegal liquor commissioned by the Colombian government to estimate the determinants of the demand for smuggled and adulterated liquor. To address unit and item nonresponse, they implement a multiple imputation procedure with chained equations.

Kim P. Huynh

David T. Jacho-Chávez

Gautam Tripathi