Incomplete observations are a common feature of financial applications that use survey response, annual report, and proprietary banking and security issue and pricing data. Finance researchers use a variety of procedures, including deleting offending observations and imputing ad hoc values, that potentially fail to deliver efficient and unbiased parameter estimates. This article examines the application of a statistical framework, multiple imputation methods, that minimizes incomplete data problems if the missingness satisfies certain criteria. When applied to two financial datasets involving severe data incompleteness, the imputation methods outperform the ad hoc approaches commonly used in the finance literature.