Dynamic Multilevel Covariance Structure Models for Analyzing Rolling Cross-Sectional Tracking Surveys

Citation data:

SSRN Electronic Journal

Publication Year:
2018
Usage 100
Abstract Views 89
Downloads 11
Repository URL:
https://scholar.smu.edu/business_marketing_research/29
SSRN Id:
3127710
DOI:
10.2139/ssrn.3127710
Author(s):
Park, Joonwook; Dillon, William R
Publisher(s):
Elsevier BV
Tags:
Satisfaction Tracking; Performance Metrics; Dynamic Linear Model; Hierarchical Time-Series Cross-Sectional Data; Marketing
article description
Appendix is available at: https://ssrn.com/abstract=3127752 Modeling customer satisfaction tracking data presents some unique challenges. Customer satisfaction tracking data are neither strictly cross-sectional nor strictly longitudinal in a panel sense. We refer to such data as nested repeated (rolling) cross-sectional, reflecting the fact that respondents are nested within brands and that the data are composed of a series of repeated cross-sections. In this study, we develop a new methodology that extends multilevel modeling techniques by considering both temporal variation and individual-level heterogeneity simultaneously in the presence of nested repeated cross-sectional samples. Our modeling approach essentially uses a Dynamic Linear Model (DLM) form and a multilevel structural equation form to “marry” the time series and cross-sectional sources of variation. Our empirical analysis explores the long-term effects of overall customer satisfaction and feature-level satisfaction on brand sales and the salutary benefits, if any, of controlling for individual-level response heterogeneity.