New method reveals hidden biases in regression analysis, impacting policy decisions.
The article reviews a new method called latent instrumental variables (LIV) for dealing with endogeneity in linear regression. LIV does not require observed instruments and can be used to test for endogeneity in regression analysis. The study found that LIV estimation can provide more accurate results compared to ordinary least squares (OLS) and instrumental variable (IV) regression methods in certain cases. In particular, LIV estimation showed less bias in estimating the effects of education on income and campaign expenditures on election outcomes.