Overview

In this course, you will discover models and approaches that are designed to deal with challenges raised by the empirical econometric modelling and particular types of data. You will: – Explore the motivations of each approach by means of graphs, preliminary statistics and presentation of economic theories – Discuss the problem of identification of the parameters, and how to address this problem by modelling simultaneous equations and causality in economics. – Examine the key features of panel data, and highlight the advantages and disadvantages of working with panel data rather than other structures of data. – Learn how to choose what econometric specification to adopt by introducing the test for poolability and the Hausman tests. – Discuss models for probability that are used where the variable under investigation is qualitative, and needs to be treated with a different approach. – Learn how to apply this approach to building an Early Warning system to forecast systemic banking crises using data from the World Bank. It is recommended that you have completed and understood the previous two courses in this Specialisation: The Classical Linear Regression Model and Hypothesis Testing in Econometrics. By the end of this course, you will be able to: – Respond appropriately to issues raised by some feature of the data – Resolve address problems raised by identification and causality – Resolve problems raised by simultaneous equation and instrumental variables models – Resolve problems raised by longitudinal data – Resolve problems raised by probability models – Manipulate and plot the different types of data.

Syllabus

Random Regressors

This module presents models and approaches that are designed to deal with challenges raised by the regressors being random variables. We address the problems raising when regressors are correlated with the error term, ad when this problem is likely to raise. We look at modelling simultaneous equations and discuss causality in economics, with an application to returns to schooling. We will discuss the problem of identification of the parameters, and how to address this problem. We finally estimate a model for the demand and supply of fish.

Panel Data Models: The Basics

We describe the problems as raising when repeated observations are present in the sample, and it is possible to deal with unobserved heterogeneity in the sample. We analyse key features of panel data and highlight the advantages of working with panels of data instead of other structures of data. We analyse fixed effects models and estimators associated with this approach. A full example where the neoclassical growth model is estimated is presented and discussed using the PWT tables data. We see how to choose between fixed effects models and pooled models by introducing the test for poolability.

Further Analysis of Panel Data Models

This week we study random effects on models of panel data. We analyse the Hausman test, that helps study whether we should be adopting the fixed effects models of the random effects models. The two approaches are compared using the Solow growth model. We also analyse the role of time in panel data models by presenting the between estimator, the two ways estimators, where we have time effects, and we finally look at dynamic panel data models, where the lagged dependent variable enters the set of regressors.

Probability Models

We discuss models for probability, that are used where the variable under investigation is qualitative, and needs to be treated with a different approach. We analyse the difficulties raised by linear models when the dependent variable is binomial. We study logit and probit estimators. We apply probability models to the problem of building an Early Warning system to forecast systemic banking crises using data from the World Bank.