Item non-response occurs when respondents fail to provide answers to some or all of the questions posed during survey interviews. The standard procedure is to exclude such responses from the econometric analysis. This may be appropriate if the sample included does not differ significantly from those excluded in the analysis. If this is not the case, the econometric analyst faces a sample selection bias problem. The aim of this paper is to provide further evidence using a simple sequential procedure to deal with the problem when using non-randomly selected samples in social science research. The procedure entails different levels of estimation and diagnostic with the Ordinary Least Squares (OLS), Heckman's 2-step and Full Information Maximum Likelihood (FIML) estimators. In the application context, we found the FIML estimator to be more efficient in dealing with sample selection bias than the Heckman's 2-step approach.