
Notes on Programs and Datasets for:

Besley, Persson and Sturm, Political Competition, Policy and Growth: Theory and
Evidence from the United States, Review of Economic Studies, October 2010.

The zip archive bps_final.zip contains all datasets and programs that were used 
for the results reported in this paper. Unpacking this archive and executing the
file bps_master.do in STATA (Version 10.0) 
replicates all results in the paper.

The archive has the following structure. It contains two datasets:

  uscomp.dta: This STATA dataset contains the data that is used for all 
  regressions and Figure 1.

  swingvoters.dta: This STATA dataset contains the data extracted from the NES 
  that is used in the calibration displayed in Figure 2.

The dofile bps_master.do calls three dofiles:

  bps_main_regressions_final.do: This do files uses uscomp.dta and estimates the
  results in Tables 1 to 7 and alternative specifications discussed in the main 
  text. A log file of the results is saved in the subdirectory \logs and .out 
  files of the results are saved in the subdirectory \results.

  bps_figure1_final.do: This do file uses uscomp.dta and creates Figure 1 and 
  saves an .eps copy of the Figure in the subdirectory \results.

  bps_figure2_final.do: This do file uses swingvoters.dta and creates Figure 2 
  and saves an .eps copy of the Figure in the subdirectory \results.

The subdirectory \temp is used by bps_main_regressions_final.do to save 
temporary datasets.

Notes on the data:

(1) Note that the sum of non-farm income and farm income is not exactly equal 
to total state personal income for recent years. These variables come from 
different revisions of the state income data of the Bureau of Economic Analysis.
The farm and non-farm income data were taken from the CD State Personal Income 
1929-2000 published in November 2001 by the Bureau of Economic Analysis. The 
state personal income data was made available by the Bureau of Economic Analysis
in electronic form in February 2005. 

(2) The variable "normdem" which we use to construct our main measure of 
political competition "compnorm" was originally collected by Jim Snyder to 
measure the incumbency advantage. The data exclude elections in which a third 
candidate obtained 10 percent of the vote and elections in which not both a 
Republican and Democratic candidate ran. These exclusions create a few gaps 
in the data for Georgia, Louisiana, South Carolina, Mississippi and Minnesota, 
which we filled in using the ICPSR election data (Study 7757), which is also 
the basic data source for normdem. The raw data behind normdem will in 
future be available online through an elections data project of Jim Snyder.
