Definition Generalized least square (GLS) model

The generalized least square model is a method used in regression analysis alternative to the ordinary least square (OLS) model. Whereas both GLS and OLS models try to explain the relation between one (or many) independent variables and one dependent variable, the GLS model performs better than the OLS one in situations when the independent variable (or some of them) is (or are) somewhat also influenced by the dependent variable.

For instance, a regression analysis can be made to estimate the extent to which having a higher share of the population with tertiary education (independent variable) determines a higher GDP for a country (dependent variable). In this case, however, it is also likely that countries with higher GDPs invest more in education and have more and better universities, and thus the dependent variable in this model influences the independent one. In such cases, the GLS model results in a more accurate estimator than an OLS model.