|
Activity Number:
|
434
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Government Statistics
|
| Abstract - #300385 |
|
Title:
|
Missing Income Information in Panel Data: Incidence, Imputation, and Its Impact on the Income Distribution
|
|
Author(s):
|
Joachim R. Frick*+ and Markus M. Grabka
|
|
Companies:
|
DIW Berlin and DIW Berlin
|
|
Address:
|
SOEP, Berlin, International, 14195 , Germany
|
|
Keywords:
|
item nonresponse ; imputation ; income inequality ; panel data ; missing data
|
|
Abstract:
|
Based on data from the German Socio-Economic Panel study (SOEP), this paper deals with the selectivity of missing income data due to item-nonresponse and the various longitudinal imputation strategies applied to this panel dataset. We analyze the impact of imputation on the personal income distribution and on income mobility. Comparing income inequality measures based on truly observed data to those derived from all observations, we find an increase in inequality due to imputation. This effect appears to be relevant in both tails of the distribution. Longitudinal analyses show a positive correlation of item nonresponse on income data over time, and also provide evidence of item nonresponse being a predictor of subsequent unit-nonresponse. Applying mobility indicators provide evidence of income mobility being understated using truly observed information only. Multivariate models show that survey-related factors (e.g., interview mode) and indicators for complex income structures are significantly correlated with item nonresponse. In conclusion, our results indicate the selectivity of item nonresponse on income questions in social surveys and push the necessity for adequate imputation.
|
