Econometrics

Learning Outcomes

• Learning Goals for ECON324:
  
• Understand what the error variance-covariance matrix is
  
• Understand what panel data are
  
• Understand what heteroskedasticity, serial correlation, and cross-sectional correlation are; and their consequences for OLS regression
  
• Match the elements of the error variance-covariance matrix for panel data models to specific types of (i) heteroskedasticity, (ii) serial correlation, and (iii) cross-sectional correlation
  
• Mathematically derive the variance-covariance matrix for the OLS and GLS coefficient estimators using linear algebra and expectation operators
  
• Understand the benefits and costs of using GLS versus OLS when estimating panel data models characterized by different types of heteroskedasticity, serial correlation, and/or cross-sectional correlation
  
• Understand the different panel data estimators available in EViews
  
• Use a 5-step procedure for deciding which EViews panel data estimator you should use in a given situation
  
• Estimate, interpret, and critically analyse results from a wide variety of panel data models, including the interpretation of coefficients and hypothesis tests
  
• Understand what endogeneity bias is, its different sources, and the problems it causes for OLS estimation
  
• Understand what 2SLS and GMM are, and how they address endogeneity bias
  
• Estimate 2SLS and GMM using EViews, and critically evaluate the results from these procedures
  
• Understand the properties of good instrumental variables, the problems of weak instrumental variables, and a procedure for determining whether 2SLS/GMM is likely to be "better" than OLS in a given situation
  
• Understand the different panel data estimators available in STATA, include the PCSE model
  
• Understand the difference between fixed and random effects panel data estimators, determine which to use in a given situation, and use these estimators in STATA
  
• Understand how (i) "differerence-in-differences" and (ii) first differencing can be used to identify the effects of treatments/policies, how they compare with fixed effects models, and how STATA can be used to implement these procedures
  
• Use STATA to obtain predictions, test hypotheses, and do post-estimation analysis of estimated models
  
• Understand what truncation and censoring are and the problems these cause for OLS regression
  
• Use STATA to estimate the Tobit model, count models (specifically, the Poisson Regression model and the Negative Binomial Regression model), the Censored Normal Regression model, and the Truncated Normal Regression model; and how to decide which model to use in a given situation
  
• Derive the likelihood function for the each of the models above
  
• Understand what sample selection bias is, and how it relates to endogeneity bias
  
• Use STATA to estimate the Heckit sample selection procedure and critically evaluate the output

Prerequisites

For further information see Department of Economics and Finance .