Surveys provide a unique opportunity to collect data that is not available through official registries. However, responders are not always happy to share all the information requested in questionnaires. Resulting missing data problem poses a threat to further empirical work, decreasing the available sample size and introducing potential sample selection. A frequently avoided answer concerns wages, particularly at the right tail of their distribution.  The missing wages are usually imputed based on statistical models that use other observed factors to predict the expected value. It is very important to obtain accurate predictions for reliability of studies using the imputed data.

In order to find potential improvements in predicting missing wages in CPS (Current Population Survey) survey, a best prediction competition has been announced by Minnesota Economics Big Data Institute. I am proud to announce that my submission (as a joint work with Egor Malkov) has been distinguished among the competition winners.