Online Program

Model-Aided Sampling: An Empirical Review

Keywords: Model-aided sampling, O*NET, occupational survey, sample design, quota sampling, burden to public, estimate bias

Model-aided sampling (MAS) is a hybrid sampling approach that combines probability based sampling with a representative sampling paradigm. MAS is ideally suited for simultaneously sampling multiple target populations and optimizing the sample yield across each population; thereby, reducing the burden to the public while minimizing potential bias in the estimates. The O*NET Data Collection Program (O*NET), a large nationally representative establishment survey of occupations, tested MAS through a simulation study and presented its finding at ICES-III. O*NET simultaneously collects data on over 900 occupations each of equal importance to the study’s objectives. The simulation study indicated that MAS would not substantively bias estimates for an occupation while reducing the level of burden to the study. Since 2007, O*NET has sampled and published estimates for over 300 occupations using the MAS paradigm. This paper presents the results of an empirical study evaluating the effectiveness of MAS to reduce burden to the public without introducing bias to the estimates by comparing the estimates solely obtained through large sampling theory and estimates obtained using the MAS paradigm.