A Simulation Study to Compare Weighting Methods for Survey Nonresponses in the National Survey of Recent College Graduates

Common methods to adjust sampling weights to account for survey nonresponse are the weighting cell technique, response propensity modeling, or a combination of both. Each raises several issues; for examples, which covariates to use to construct the weighting cells or to model response propensities, whether weights are used in modeling, and whether to weight the adjustment factor. To address these issues, we used a simulation based on a data from the National Survey of Recent College Graduates (NSRCG) maintained by the National Science Foundation to evaluate these weighting methods. In the end, we expect that the weighting adjustment will have successfully accounted for possible nonresponse bias.