********************************************************************************
* The Growth of Education Differentials in Marital Dissolution in the United States
* Kim McErlean
********************************************************************************

This is the readme file to accompany the Stata code and data for the paper "The Growth of Education Differentials in Marital Dissolution in the United States.
This paper uses the NSFG, cycles from 2006-2019. Data is provided here, but can also be accessed from the NSFG site at https://www.cdc.gov/nchs/nsfg/.

Stata files: (Note: these files were created in Stata version 16).
1. divorce.do: This is the master .do file for the project and runs all necessary code below.

2. setup_nsfg_environment.do: This is the setup file for the project that makes sure you have the necessary packages, are using the right Stata version, sets up macros, etc.

3. setup_kmcerlean.do: This file is called by the above setup file to set the locations of project macros used to store data, results, etc. You should save this file as setup_(computer_username).do and replace the macros with the relevant locations on your computer.

4. divorce_create.do: This files extracts data from the 2006-2010, 2011-2013, 2013-2015 NSFG cycles and keeps only the necessary sample (ever-married women) and does some recodes. Creates a person and person-year file for each cycle.

5. divorce_create_1519.do: This file extracts data from the 2015-2017 and 2017-2019 NSFG cycles and keeps on the necessary sample and recodes. It also appends all cycles together to create one master file. Note: this has to be split into two steps because NSFG changed the data they released starting in 2015, plus some variable names changed.

6. divorce_analysis_mcohort_sts.do: This file finishes adding sample restrictions and recodes, as well as creates life tables estimating divorce and marital dissolution by education and marital cohort (Tables 1-3).

7. divorce_descriptives.do: This file isn't used in the paper, but does create descriptive statistics for the sample.

8. divorce_regression_py.do: This file uses the person-year data to estimate the discrete-time event history models. It primarily feeds into Table 4 and Figure 1, but a series of robustness checks are also conducted here.

Data files:
Cycle 7 (2006-2010): nsfg0610fem.dta
Cycle 8 (2011-2013): Female_NSFG1113.dta
Cycle 9 (2013-2015): 2013_2015_FemRespData.dta
Cycle 10 (2015-2017): 2015_2017_FemRespData.dta
Cycle 11 (2017-2019): 2017_2019_FemRespData.dta

Key variables used in the analsis:
sepdiv: this is an indicator of whether or not the respondent's marriage dissolved (divorce or separation); this is my key dependent variable
div: this is an indicator of whether or not the respondent's marriage ended in divorce, specifically; used in the life table analysis.
mcohort: marital cohort; a recode based on year the female married (independent variable)
educ: this is a recode of HIEDUC to create 4 groups of education (independent variable)
agemar: this was created based on year of birth and year of marriage - I used this in my sample restrictions as well as a control variable
retro: this is an indicator of years between marriage and interview - I used this in my sample restrictions
dur: this is marital duration, calculated as either year of dissolution minus year of marriage, if marriage is dissolved, or year of interview minus year of marriage, if marriage is still intact. This is my key indicator of time in my analysis.
scaled_weight: this is the weight I calculated from the NSFG provided weights, as discussed in my methods section

Control variables:
race: this is a recode of RSCRRACE; I grouped all races outside of White, Black, and Hispanic as "Other"
agemar: age respondent married (see above)
husband_cohab: an indicator of whether or not the respondent and her husband cohabited prior to marriage (recode of LVTOGHX)
OTHERCOHAB: an indicator of whether or not the respondent cohabited with anyone other than her husband prior to marriage (based on LIVEOTH, then start and end dates of cohabitation spells and marriage to understand which cohabitations were prior to marriage)
pmcbirth_mar1: an indicator of whether the respondent had a birth prior to marriage - based off of year married and year(s) of birth of children
METRO: this is an indicator of whether the respondent lives in a rural or urban area