Advanced Econometrics I

1.5 Credits · Level: Graduate

The linear regression model (also known as the OLS regression model) assumes that the dependent variable is continuous, such as income, or population of a town or country. However, many dependent variables in applied empirical research are not continuous. Some are binary, such as answers to “yes/no” questions in surveys, others are discrete – i.e. they take on a limited number of integer values – and distinctly bounded, such as a (non-negative) number of terrorist attacks, and even unordered, such as items a household purchases from a basket of goods. Sometimes, the dependent variable equals zero for many observations, such as the number of farm animals owned by urban dwellers. In these circumstances, the linear model is usually not well-suited. This class will familiarize you with alternatives to the linear model better suited for non-continuous data, such as probit, logit, multinomial, count, and duration models. We will study their theoretical underpinnings, as well as read applied research employing these methods.