An Investigation of the Imputation of Monthly Earnings for the Survey of Income and Program Participation Using Regression Models

Written by:
Working Paper Number: SIPP-WP-15

Introduction

This study using data from the first four waves of SIPP extends the previous ISDP study.  On each longitudinal record there are twelve months (3 waves) of data for several variables.  In this phase we consider imputation of earnings for each of the four months of the second wave, assuming that all earnings for waves 1 and 3 are reported.  A separate multiple regression model is fit for each month and months 5, 6, 7 and 8 are imputed in that order.  This is not a general study of imputation methods, but an investigation of the applicability of this particular approach.  This work can easily be extended to consider other patterns of missing data.  The study will be reported in several parts:

  • Preliminary analyses of some variables
  • Construction of files for model estimation and imputation
  • Estimation of models
  • Imputation Procedures (using estimated models)
  • Evaluation of imputation
  • Discussion of results
  • Suggestions for future research
Page Last Revised - January 8, 2024