Data and Replication Files for "The Complications of Controlling Agency Time Discretion: FDA Review Deadlines and Drug Safety" (American Journal of Political Science 56 (1) (January 2012) 98-114).

The replication archive includes the data and code used in Carpenter et al.  The data and code files are summarized the Data and Code boxes below, respectively. The main data replication files -- as well as other files -- are available directly from the Dataverse replication archive for this study. A file containing technical details of the estimations as well as results from many auxiliary tests and estimations is available here

The replication data sets, found in the Dataverse replication archive for this study, include two main replication files, as well as ancillary files that were constructed to verify the robustness of the results to coding decisions. 



DeadlinesData73009v2 with regime indicator.csv

A data set containing each drug used in the analysis, with indicators of time of approval on the PDUFA or FDAMA clock.  This was the primary dataset used for the post-market regulatory event analysis.


A STATA version of DeadlinesData73009v2 with regime indicator.csv


A data set containing observations for each drug-month.  The time series for each drug month is censored either at the final month of the dataset or at the month of final approval.  Used in Cox proportional hazards analyses.

Additional files

These files adopt different coding decisions, including changing the pre-deadline approval window and adjusting how the deadline is calculated in days. 


Both the STATA and R code are written to operate on Windows machines, and references to working directories apply to the local machine on which the code was originally run.  Some code modification will be necessary to make the code run on another machine.



STATA code to replicate Table 3.doc

STATA code for PMRE analysis.  This code behaves differently under different STATA versions while yielding approximately the same results.  As a robustness check, these analyses were rerun using comparable functions in R; this file provides the STATA code only.


R code to double-check the STATA routine in

STATA code for the Cox models.


Robustness check of the STATA-based models conducted in R.


Batch file to run all matching analyses found in the appendix.  (The other R scripts in the /code folder pertaining to matching are called directly or indirectly by this script.)

As a robustness check, several coding decisions were revisited and incorporated in later versions of these data.