The performance of sample selection estimators to control for attrition bias.
The performance of sample selection estimators to control for attrition bias.
复制标题
DOI：
10.1002/hec.628
复制
发表时间：
2001-07
期刊：
影响因子：
2.1
通讯作者：
Astrid Grasdal
中科院分区：
文献类型：
--
作者：
Astrid Grasdal
Sample attrition is a potential source of selection bias in experimental, as well as non-experimental programme evaluation. For labour market outcomes, such as employment status and earnings, missing data problems caused by attrition can be circumvented by the collection of follow-up data from administrative registers. For most non-labour market outcomes, however, investigators must rely on participants' willingness to co-operate in keeping detailed follow-up records and statistical correction procedures to identify and adjust for attrition bias. This paper combines survey and register data from a Norwegian randomized field trial to evaluate the performance of parametric and semi-parametric sample selection estimators commonly used to correct for attrition bias. The considered estimators work well in terms of producing point estimates of treatment effects close to the experimental benchmark estimates. Results are sensitive to exclusion restrictions. The analysis also demonstrates an inherent paradox in the 'common support' approach, which prescribes exclusion from the analysis of observations outside of common support for the selection probability. The more important treatment status is as a determinant of attrition, the larger is the proportion of treated with support for the selection probability outside the range, for which comparison with untreated counterparts is possible.
