CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0521
Title: Exploring the possibility of non-parametric estimation in the first stage of IV2SLS estimation Authors:  Aslam Muhammad - Bahauddin Zakariya University (Pakistan) [presenting]
Sajjad Haider Bhatti - Government College University, Faisalabad-PAKISTAN (Pakistan)
Abstract: The potential bias arising from the endogeneity is common in many empirical econometric relationships. The problem of endogeneity is largely tackled by Instrumental Variables Two Stage Least Squares (IV2SLS) estimation. The idea of estimating the first stage of IV2SLS non-parametrically is presented. The Mincer wage model based on human capital theory is widely used and is typically estimated by using instrumental variables 2SLS to address the possible bias due to endogenous schooling variable. We estimate the Mincer wage function using the French labour force data with recently proposed instrumental variable. We estimate the first stage schooling equation using LOESS method. Our findings show that the estimates from non-parametric first stage estimation in IV2SLS are more efficient compared to those from the traditional IV2LSL approach. Hence, the use of IV2SLS approach with non-parametric first stage is recommended for empirical studies dealing the issue of endogeneity bias.