Replication Material (US Cosponsorship Data) used in
If you are interested in the Argentine Cosponsorship data click here.
The Cosponsorship data for the US Congress was collected by James Fowler (UCSD). Here we just reshaped his data to run the code provided below.
To download the Datasets for the US Congress
1. Zip files containing separate lower house (House 93rd to 108th files) and upper house data (Senate 93rd to 108th files). Wide files Stata 9 format.
NOTE: The individual congress files names are in Spanish ("DIPUTADOS") and ("SENADORES") but the data is the correct one :).
To Replicate our Analysis
If you are interested in loading the data in R 2.7 and retrieving all the estimates as explained in our article use this replication code.
To run the replication code, create a folder “C:\usa” and unzip the House and Senate files.
Install R 2.6 or higher.
Copy and Paste the code from the replication file all at once.
If you decide to copy and paste in steps (to see what the code is doing), make sure to copy and paste together the section:
##
## Loop over all 16 House datasets or all 16 Senate datasets
##
for (i in 1:Z){
for (i in 1:Z){
…..
…..
….. vcolor[[i]]<-color
}
##
## Finish the loop. The Data has been created
## To see Ideal Point Estimates
Why using cosponsorship data to study Legislatures around the world?
There are many reasons to use cosponsorship data to retrieve ideal point estimates:
First, while in many legislatures it is difficult to obtain roll call data to estimate the preferences of legislators, cosponsorship data is relatively abundant and readily available.
Second, cosponsorship data provides information about the preferences of legislators prior to floor level voting pressures.
Third, in contrast with roll call, data which is generally available only for a small subsample of bills, cosponsorship data generally includes most bill initiatives.
What type of problems should be resolved to take advantage of the cosponsorship data?
Most bill initiatives have only a handful of cosponsor. Consequently, cosponsorship data is saturated with zeros.
To deal with this problem there are a number of social network models we can use.
Alternatively, we can compute agreement matrices from the raw data and estimate ideal points using any number of Item Response Models.
In the article “Comparing Cosponsorship and Roll Call Ideal Points” we use Principal Component Analysis on the log transformed agreement matrices.
However, there are a large number of statistical models than can be used to retrieve ideal point estimates from properly transformed cosponsorship data.