Objective: To explore, post hoc, whether a large decline in smoking estimates between the 2005 and 2006 New Jersey Adult Tobacco Surveys is real or spurious given various methodological and environmental changes between the 2 time periods of data collection.
Methods: Using multiple data sources, we explored survey timing, poststratification approach, midinterview terminations, wireless substitution, and question order.
Results: Changes in question order were likely responsible for the majority of the unexpected decline in smoking prevalence; to a lesser degree, wireless substitution and midinterview terminations also likely contributed to an artificially exaggerated decline.
Conclusion: Methodological changes can artificially affect trends in prevalence estimates.