Replication package for "Lifetime and intergenerational consequences of poor childhood health"

Replication package for “Lifetime and intergenerational consequences of
poor childhood health”
================
Krzysztof Karbownik and Anthony Wray,
Journal of Human Resources

You can cite this replication package using Zenodo, where an archival
version of this repository is stored.
[![DOI](https://zenodo.org/badge/542135759.svg)](https://zenodo.org/badge/latestdoi/542135759)

## Overview

This replication package contains code, output, raw data, and Stata
packages. The code in this replication package will replicate all tables
and figures from raw data using Stata and a few unix shell commands.
Instructions for accessing restricted-use data are provided below.

The entire replication takes 7 and a half days to run on a 2.9 GHz
Intel-based Windows PC (64-bit x86-64) with 24 GB of RAM or 4 and a half
days on a machine with 64 RAM.

## Data availability statement

Replication materials are provided under a BSD 3-Clause License.

The hospital inpatient admissions data from St. Bartholomew’s and Guy’s
Hospitals that support the findings of this study are included with the
replication package.

The hospital inpatient admissions data from Great Ormond Street Hospital
are publicly available from the Historic Hospital Admission Records
Project (HHARP) (<https://hharp.org/>). The HHARP data used in this
study were kindly shared by Dr. Sue Hawkins (<drsuehawkins@gmail.com>).
They cannot be posted online or included with the replication materials.

The anonymized complete-count data (UK Data Archive study number
[7481](https://beta.ukdataservice.ac.uk/datacatalogue/studies/study?id=7481#!/details))
for the Censuses of England and Wales, 1881 to 1911, are publicly
available from the Integrated Census Microdata (I-CeM) project
(<https://icem.data-archive.ac.uk/#step1>). Since data access is
[safeguarded](https://ukdataservice.ac.uk/find-data/access-conditions/),
the data files used for this study cannot be included with the
replication materials.

“Integrated Census Microdata (I-CeM) Names and Addresses, 1851-1911:
Special Licence Access” (UK Data Archive study number
[7856](https://beta.ukdataservice.ac.uk/datacatalogue/studies/study?id=7856#!/details))
is a safeguarded dataset that contains names and addresses from the
Integrated Census Microdata (I-CeM) dataset of the censuses of Great
Britain for the period 1851 to 1911. These data can only be accessed by
obtaining a Special License agreement with the UK Data Archive and the
data distributor, FindMyPast Ltd.

Prior versions of the complete count 1881 Census of England and Wales
were accessed from the North Atlantic Population Project
(<https://www.nappdata.org/napp/>) in 2013. The NAPP data are now
distributed through the IPUMS International data system
(<https://international.ipums.org/international/>). At the time we
accessed the data, the residential address variable was available for
download from NAPP, but it is now only available as part of the
restricted-use dataset “Integrated Census Microdata (I-CeM) Names and
Addresses, 1851-1911: Special Licence Access.” Thus, we cannot include
our original extracts from NAPP as part of the replication materials.

All other data are contained in the Zenodo repository.

### Public-use raw data links

-   Integrated Census Microdata (I-CeM) project data documentation and
    supporting materials, including parish and place list dictionary
    files are available
    [here](https://www.essex.ac.uk/research-projects/integrated-census-microdata).
-   An archived version of a website for an index of streets by
    registration district from the 1891 Census of England and Wales is
    located
    [here](https://webarchive.nationalarchives.gov.uk/ukgwa/+/http://yourarchives.nationalarchives.gov.uk/index.php?title=Place:Altrincham_Registration_District,_1891_Census_Street_Index).
    The authors accessed and scraped data from the website in 2013. As
    the original website was a “wiki”-style page, the archived version
    may differ from the version accessed by the authors.

### Limited-use data links and instructions

Researchers interested in accessing the hospital inpatient admission
records from the Historic Hospital Admission Records Project (HHARP)
(<https://hharp.org/>) will need to create an account on the HHARP
website. Data download is limited to 200 observations. The search
results page indicates how many entries match the search criteria.
Downloaded data files are in `.csv` format. Researchers who require
larger volumes of data should contact Dr. Sue Hawkins
(<drsuehawkins@gmail.com>).

Researchers based outside the UK who are interested in accessing the
complete count records for the Censuses of Great Britain (including the
Censuses of England and Wales) from the Integrated Census Microdata
(I-CeM) project (<http://icem.data-archive.ac.uk/#step1>) will need to
create an account with the UK Data Service
(<https://beta.ukdataservice.ac.uk/myaccount/login>) by first requesting
a username
[here](https://beta.ukdataservice.ac.uk/myaccount/credentials). If you
are in the UK and from an institution of higher or further education or
your organisation is part of the UK Access Management Federation (UKAMF)
and on
[this](http://www.ukfederation.org.uk/content/Documents/AccountableIdPs)
list of federation members, you can use the username and password issued
to you by your organisation to login/register with the UK Data Service.
Data download is limited to 1,000,000 observations. Downloaded data
files are in `.csv` pipe-delimited format. Authors converted `.csv`
files to `.dta` files using Stat/Transfer and the shell script
`shells/csv_to_dta.sh`.

### Restricted-use data instructions

The dataset that contains names and addresses from the complete count
Censuses of England and Wales is restricted use, but may be accessed by
obtaining a Special License agreement with the UK Data Archive and the
data distributor, FindMyPast Ltd. Instructions can be found
[here](https://beta.ukdataservice.ac.uk/datacatalogue/studies/study?id=7856#!/access-data).
It can take several months to be approved and gain access to the data.
Data are provided as `.txt` files. Authors converted `.txt` files to
`.dta` files using Stat/Transfer and the shell script
`shells/csv_to_dta.sh`.

## Computational requirements

### Software requirements

-   Stata 14. Some sections of code are set to version 12.
    -   All user-written Stata programs used in this project can be
        found in the `scripts/code/programs` directory

Portions of the code use shell commands, which may require Unix. The
analysis takes 7 and a half days to run on a 2.9 GHz Intel-based Windows
PC (64-bit x86-64) Microsoft Windows 10 Professional operating system,
24GB of RAM, and 8-core Stata/MP version 17.0. The run time will be 3
days shorter on a computer with 64 RAM.

At least 150GB of spare hard drive space is required for the raw data
files and at least 500 GB of spare hard drive space is required for the
intermediate files if you want to build the processed data from raw
data.

## Instructions for replicators

1.  Unzip, download, or clone the replication folder.
2.  Set the global `PROJ_PATH` on line 29 in `0_run_all.do`, which
    points to the replication folder.
3.  Download the I-CeM complete count census data to the folder
    `raw/icem` according to the description above. The I-CeM data
    downloaded as `.csv` files can be converted to `.dta` files using
    the shell script `csv_to_dta.sh` in the `shells/` folder. Save each
    county and year to a separate file in the `raw/icem` folder
    following the name convention `England_YYYY_County`. The names of
    the counties should correspond to the lists of counties in lines 13
    to 19 of `01.01_extract_icem_self_variables.do`.
4.  Add the Special License data with names to the folder `raw/icem_sl`.
    Convert the `.txt` files to `.dta` files using the shell script
    `csv_to_dta.sh`. The `.dta` files should be named
    `ICEM_NAMES_EW_YYYY`.
5.  Obtain the HHARP hospital data and save to `raw/hharp`.
6.  The complete count census data from NAPP used in this study include
    the variable `GB81A_ADDRESS` that is no longer publicly available.
    The extracts of NAPP data used in lines 10 and 118 of
    `scripts/code/01_build/04_extract_hospital_data/04.03_clean_residential_addresses.do`
    need to be replaced with the I-CeM Special License data that include
    the address variable from the 1881 census of England and Wales. The
    variable names may need to be changed.
7.  Save and run `0_run_all.do`.

## Acknowledgements

This code is based on AEA data editor's [readme guidelines](https://aeadataeditor.github.io/posts/2020-12-08-template-readme). Some content on this page was adapted from [Alex Hollingsworth](https://hollina.github.io/).