Skip to main content
Have a personal or library account? Click to login
Misspecification Effects in the Analysis of Panel Data Cover

Misspecification Effects in the Analysis of Panel Data

Open Access
|May 2016
Read online

Abstract

Misspecification effects (meffs) measure the effect on the sampling variance of an estimator of incorrect specification of both the sampling scheme and the model considered. We assess the effect of various features of complex sampling schemes on the inferences drawn from models for panel data using meffs. Many longitudinal social survey designs employ multistage sampling, leading to some clustering, which tends to lead to meffs greater than unity. An empirical study using data from the British Household Panel Survey is conducted, and a simulation study is performed. Our results suggest that clustering impacts are stronger for longitudinal studies than for cross-sectional studies, and that meffs for the regression coefficients increase with the number of waves analysed. Hence, estimated standard errors in the analysis of panel data can be misleading if any clustering is ignored.

Language: English
Page range: 487 - 505
Submitted on: Jul 1, 2014
Accepted on: Jan 1, 2016
Published online: May 28, 2016
Published in print: Jun 1, 2016
Published by: Sciendo
In partnership with: Paradigm Publishing Services

© 2016 Marcel de Toledo Vieira, Peter W.F. Smith, Maria de Fátima Salgueiro, published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.