You are on page 1of 1

This file is the “Readme file” for the posted programs (in Stata and R) that generate the

results reported
in “Sales Taxes and Internet Commerce,” by Einav, Knoepfle, Levin, and Sundaresan.

The data used for the paper is proprietary, and therefore cannot be posted online. However, we hope that
interested researchers could understand the code (which we hope is self-explanatory) and make requests
for further analysis. We will do our best to accommodate such requests within the limits of our non-
disclosure agreement.

These are the files included in the posted zip file:

Code for “location surprise” estimation (Section I, Tables 1-4, A1-3):


 location-surprise.estimation.do
 location-surprise.estimation.R for substitution IV bootstrap

Code for gravity model estimation (Section II, Table 5, Table A4):
 gravity.estimation.R (models (a)-(e))*
 gravity.estimation.do (models (a)-(d) only; Stata chokes on monthly version)

Code for cross-sectional total online purchasing regressions (Section II, Table 6 Panel A):
 cross-section.estimation.do

Code for rate-change total online purchasing regressions (Section II, Table 6 Panel B):
 rate-change.estimation.R*

Supporting code for R is in common.functions.R.

* rate-change.estimation.R and gravity.estimation.R use a custom R package ‘ppois’ for panel Poisson
regression with sparse matrices; the code is in directory ‘ppois’ and can be installed in R from the
supplied archive with the command install.packages(“ppois_0.1.tar.gz”)

You might also like