Suppose you are hired as an Economist Consultant for American Airlines. On your first day at the job, your first task is to estimate the elasticity of airfare to passengers demand (i.e. the % change in passengers demand given a 1 % change in airfare). You have the following data available: # of passengers, airfare, and route concentration ratios for 1149 routes within the U.S. for the years 1997 through 2000, you begin by estimating a simple demand model:

where you allowed for separate year intercepts. The variable "dist" is the route distance in miles, naturally it does not change over time. Applying what you have learnt in your Econometric class, you decide to estimate this model using FE, RE, Fixed effects with IV and Random effects with IV.

You convinced Rene, that you should use and IV for airfare, the IV you proposed is "concen" (concentration) which is the fraction of route traffic accounted for by the largest carrier. For the IV estimation, we can think of the first stage as being:

which is (implicitly) estimated by RE or FE depending on whether eq (1) is estimated by RE or FE.

You finish running all your regressions, and it's time to present your results to the AA CEO:

The CEO ask you some relevant questions:

iv. Is it significantly different the elasticity of passengers with respect to airfare if I use the FE or RE coefficients? What does this tell you about the lambda of RE?

v. Why does the distance variable (dist) drops out from the Fixed Effects and Fixed Effects with IV regression?

vi. The CEO ask you, why do you think "concentration" is a good IV? What would you respond?

ix. What other IVs would you propose for airfare?