ABSTRACT
This chapter teaches consequences and solutions of measurement errors and causes and solutions of the endogeneity problem. It discusses two-stage least squares estimation method and conditions required for the instrumental variables. The chapter examines the consequence of measurement errors for ordinary least squares estimation. The correlated-omitted-variable problem indicates that important (strongly significant and economically meaningful) factors should not be omitted from a model. It is because any models which omit significant variables would produce biased results. The 2SLS uses the OLS method both in the first and in the second stage. For efficiency, it requires the disturbances to be homoscedastic and uncorrelated between them. In contrast, the GMM incorporates heteroscedasticity and serial correlation in estimating the parameters and their standard errors. Most cross-sectional observations are collected independently from observational units, the 2SLS method is often employed in addressing the endogeneity problem for cross-sectional data.
