README for:
Revolving versus Convenience Use of Credit Cards: Evidence from U.S. Credit Bureau Data
Scott Fulford and Scott Schuh
Journal of Money, Credit and Banking
April 5, 2022

This file describes the files and programs used to construct the output in the paper.

/**************************/
Data sets
/**************************/

We use three primary data sets:
(1) The Survey of Consumer Payment Choice from 2012-2017. Stata data sets and documentation for this survey are available from: https://www.atlantafed.org/banking-and-payments/consumer-payments/survey-of-consumer-payment-choice.aspx

(2) A subsample from the CFPB Consumer Credit Panel (CCP). This data set is a constructed random consumer level subsample of the CCP rolled up to the consumer not tradeline level. The CCP is not available publicly but is similar to data sets available in the Federal Reserve system or for purchase from a National Consumer Reporting Agency. The major variables used in the analysis are constructed from the raw CCP data and do not necessary share names with the raw CCP:
consumer_nb: A unique anonymous identifier for each consumer
date: The quarter of observation
ccdebt: Total credit card debt from all open lines
cclimit: Total credit card limits from all open credit cards
openCC: The number of open credit card lines.
age: constructed from birthyear

(3) The Making Ends Meet Survey. The survey is associated with the CCP so matches to the same variables in the CCP. MEM is not currently publicly available. The survey is described in:
https://www.consumerfinance.gov/data-research/research-reports/insights-making-ends-meet-survey/
The important variables used are:
revolver: 1 if MEM wave 1 q39 ==1 "Did you have an unpaid balance on any of your cards after making your last
payment?"; 0 if q39 ==0
weight: a weight variable that reweights the data to be nationally representative taking into account sampling and non-response.

/**************************/
Analysis programs
/**************************/

We use Stata version 16/17 for analysis. The directory structure is truncated and does not show the underlying directories. The primary programs are:

*** SCPC_lifecycle.do 
Compares MEM and SCPC to compare revolving and credit care use over the life-cycle

*** CC_EM_5.do
The main EM analysis program. 
Constructs a combined MEM and CCP sample data set.
Defines functions in MATA (Stata's matrix programming language)
Runs the wrapper program regEM_conv_est on different models
Saves the output in CCP_MEM_estimates.dta

*** CC_EM_5_paper_output.do
Takes the output from CC_EM_5.do and prepares figures and tables for the paper.
Prepares ROC curves and distribution curves.
Runs regressions on the CCP sample. 
