The Relative Empirical Validity of Dependent and Independent Data Collection in a Panel Survey

Daniel H. Hill

Abstract:
Data from the 1985 and 1986 SIPP panels are used to assess the relative empirical validity of occupational and industrial change measures obtained with independent and dependent data collection methods. These assessments are made in the context of simple descriptive and more elaborate event history models using two distinct definitions of change. The results suggest that for unreliable (“noisy”) measures, like occupation and industry of employment, dependent measurement methods in panel surveys result in a net improvement in data quality. While it may be true that dependent measurement methods miss some true change, they also provide vastly lower amounts of spurious change. As a result, the dependent data collection method appears to improve the empirical validity of noisy measures in panel surveys.

Keywords:
Panel surveys; change measures; measurement error.