Changes

'''24/07/2017''' - Review of propensity score matching and the Hochberg and Fehder (2015) paper.

'''25/07/2017''' - Review of instrumental variables in panel data methods.

'''26/07/2017''' - in order to eliminate any endogeneity in the categorical variable "hub", a potential instrument is real estate taxes. If the real estate taxes are higher, there might be a higher demand for Hubs. Data on mortgage based real estate taxes was downloaded from the census data, using the American Communities Survey. A fixed effects IV regression shows that taxes don't perform as satisfactory IVs in this model.