Financial Well-Being in an Urban Setting: An Application of Multiple Imputation

Many studies delete incomplete data prior to model estimation, resulting in less efficient and potentially biased parameter estimates. Multiple imputation provides a model-based method of simultaneously estimating missing values for several variables, conditioned on the observed values. The technique is applied to financial well-being data collected by survey from householders in Oklahoma County, Oklahoma. Ordered logistic models are estimated for both complete cases and multiply imputed data. Estimates from the complete case model are somewhat biased and less efficient compared with the multiple imputation model.