The Geography of Old Age in Late-Victorian England and Wales

A User Guide to the Dataset

Dr Tom Heritage, University of Cambridge

August-September 2022

Introduction

This aggregate-level dataset links poor relief data recorded on 1 January 1891 with several  
variables from corresponding 1891 census data, all at the level of the registration district 
(RD). Specifically, the numbers of men and women receiving indoor and outdoor relief in the 
‘nonable-bodied’ category (taken as a proxy of the numbers of older-age men and women on relief) 
are accompanied with a series of socio-economic variables calculated from census data on the 
population aged 60 years and over (our definition of ‘old age’). 

Thus, the dataset fulfils two objectives:

1. To start reconciling poor relief data from the House of Commons Parliamentary Papers 
archive with transcribed Integrated Census Microdata (I-CeM) available at the UK Data 
Service (UKDS). 

2. To capture geographical variations in the proportion of older-age men and women on poor 
relief as well as in several household, occupational and migratory compositions recorded 
in the census, consulting data from 1891 as a pilot study in anticipation of an extended 
project covering all censuses from 1851-1911.

The dataset is based on these existing data: 

The Integrated Census Microdata project (I-CeM). These are readily available for 
download from UKDS. They are transcriptions of raw census enumerators’ books (CEBs) data 
for all populations of England and Wales between 1851 and 1911 (excluding 1871, although 
this period is in elementary format to researchers at the Cambridge Group for the History of 
Population and Social Structure, University of Cambridge, or CAMPOP). Consisting of 162 
million individual records, the data are accompanied with a series of occupational, household 
and migration codes that run to near 100 variables. They are presented in anonymised format 
as well as a secure data format with names. A further enhanced version of I-CeM produced by 
original PI of I-CeM Professor Kevin Schürer, who (at the time of writing) is at CAMPOP, 
corrects numerous significant errors in the original I-CeM, including a far better allocation of 
people to places and corrections to the derived family and household variables (Schürer and 
Higgs, 2020, SN: 7481).

Poor Relief reports. These data are available from the House of Commons Parliamentary 
Papers archive as scans of the original print copies. Biannually, the House of Commons 
produced the numbers of men, women and children receiving poor relief on 1 January and 1 
July since 1850 at county level and since 1857 at Poor Law Union level for all of England and 
Wales. Although Poor Law Union level data cease by 1912, coverage exists in full. Digitised 
transcriptions exist in datasets by Humphrey R. Southall and David R. Gilbert (2020, SN: 4567) 
entitled ‘Great Britain Historical Database: Economic Distress and Labour Markets Data: Poor 
Law Statistics, 1859-1919’. Although embargoed on UKDS, they are released to the public on 
the website ‘A Vision of Britain Through Time’. Also, there is Ian Plewis’ dataset (2020, SN: 
7822) entitled ‘Census and Poor Law Union Data, 1871-1891’. It is open for direct download 
on UKDS and aims to replicate research produced by Victorian contemporary G. Udny Yule 
(1899).

Extracting and Enhancing the Poor Relief Data

Plewis provides data for Poor Law Unions recorded on 1 January 1871, 1881 and 1891, so it 
was decided to make a copy of all relevant variables gathered by Plewis of 1 January 1891. 
However, his data only record Poor Law Unions in England, not those in Wales. Also, the age 
profiles of the older-age population are derived from the published census abstracts produced 
alongside original analysis of the census data by Victorian contemporaries, rather than the more 
refined raw CEB data on which the abstracts extracted their data.

Further coding, checking and cleaning was required to enhance Plewis’ original data. The 
enhancements are: 

1. The addition of Poor Law Unions in Wales.

2. Providing numbers of older people aged 60 years and over from raw CEB data, rather than 
the published census abstracts. 

3. The inclusion of additional variables related to census information on the household, 
migratory and occupational circumstances of older people. 

Extracting Census Data

The series of socio-economic census variables included in this dataset are sourced from a text 
file containing 2,114,020 males and females aged 60 years and over in the 1891 census. This 
was supplied to the author by Professor Alice Reid and contain 47 variables requested by the 
author. The exclusion of RD 302 SCILLY ISLANDS produces an overall sample size of 
2,113,797.

The variables are in the second sheet of the dataset, titled ‘Economic Data’. The occupational 
and household variables derive from the numeric codes ascribed to an occupation and a 
household structure (recorded in the variables ‘Occode’ and ‘HHD’ respectively; information 
on those ‘living on own means’ was gathered from a variable denoting whether one was 
actively pursuing an occupation, ‘Inactiv’). The migratory variables are based on the distance 
between an individual’s parish of enumeration on census night and their parish of birth, 
recorded in the text file under the variable ‘MIN_DIST_KM’, based on an unpublished dataset 
created by Dr Joseph Day (for more information, see Jaadla et al., 2020, pp. 1,550-1,551). 
Information on ‘Occode’ and ‘HHD’ are in the Integrated Census Microdata (I-CeM) 
guidebook (Higgs et al., 2013, pp. 163-183, 234-7). 

These socio-economic variables are not in the original text file itself, but derive from Stata 
commands of tabulations by RD number (known in the text file as the variable ‘RD’) and sex 
(known as ‘Sex’), using the ‘preserve, keep and restore’ functions on Stata to filter out the 
selected codes required for tabulation of key socio-economic compositions. 

To include one example, the original text file data are ‘preserved’ in its original state, before 
we keep all those in the data that are male (command: keep if Sex==“M”). We then keep those 
males with an occupational code of 181 (denoting those working as agricultural labourers; 
command: keep if Occode==181). With a sample size of 100,279, we then tabulate by RD and 
sex (command: tab RD Sex). The numbers by RD are then integrated into our dataset as the 
variable ‘MAgLab’, and the proportion out of the population calculated as:

MAgLabPerC = MAgLab/M60+ x 100 

The data on the text file are restored to return to the original 2,114,020 sample size. The same 
process is applied to the remainder of the socio-economic variables (excluding again data from 
RD 302 SCILLY ISLANDS). 

All the selected socio-economic census data were calculated as a proportion of the population 
aged 60 years and over, using ‘M60+’ and ‘F60+’ as the denominator. While it is possible to 
expand upon this dataset by incorporating more data from the text file, those seen in the dataset
reflect those utilised in two unpublished papers (Heritage, 2022a; Heritage, 2022b). 

Further information on all variables and references are included in the user guide.