1 / 1

Mapping the 2nd Industrial Revolution

The Emergence of New Jobs in 19th-Century Britain
H. G. Vipond
Complexity Science Hub
Introduction

Introduction

  • 19th Century: rapid technological change
  • Creative destruction: destroyed jobs, created new jobs (Acemoglu & Restrepo 2019)
  • New Jobs: have been vital (Autor et al, 2025; Kalyani et al 2025)
  • 60% of the jobs that employ Americans today did not exist in the 1950s.
Tesla's magnifying transmitter generating millions of volts
Tesla and his “magnifying transmitter”, generating millions of volts.
Introduction

This Paper: Quantify Jobs Lost & Jobs Created in 2nd IR

  1. Aim: Understand how and where job creation mapped on to job loss
  2. Problem: No systematic quantitative record of job creation and loss in Victorian Britain
  3. Solution: Generate “micro-occupation” data. 800 industries → 10,000 micro-occupations.
  4. Quantify: new and declining occupations in Britain 1851–1911
History

History: 2nd Industrial Revolution

  • 1851: Crystal Palace — Workshop of the World
  • 1870s: Britain — shift to a new phase of industrialization.
  • Traditional crafts sectors start declining
  • Vast numbers of new jobs emerging: electricians, engineers, railway workers
The Great Exhibition at the Crystal Palace, 1851
The Crystal Palace, 1851.
History

History: British Census

  • 1753 Thomas Potter: bill introduced to Parliament. Army, emigration to the colonies, burden of the poor law. Population growth.
  • 1801–1831: Parish (township) register aggregates
  • 1841–1921: Individual Level
Title page of the Census of Great Britain, 1851
History

History: British Census Classification Schema

  • 1841: First individual level census
  • 1851: 18 orders, 90 sub-orders
  • 1861: “The classification of 1851 has been entirely revised”. Dress goes from Services to Textiles. Animal/Vegetable matter go.
  • By 1901: 23 Orders, substantial changes in sub-orders
Census table showing the occupancy of classes of persons
Data

Data: The Integrated Census Microdata Project (ICeM)

  • Decennial English census data from 1851 to 1921 (excluding 1871)
  • 220 million individual level census records, digitized
  • Occupation: at industry level
A page from a census enumerator's book
Data · Task Data

Data Construction 1: “Task” Occupation

Industry Code Occupation — Original Strings New Variable: “Task”
663 (Bootmaker)CORDWAINERCordwainer
663 (Bootmaker)BOOT & SHOE MAKERMaker
663 (Bootmaker)BOOT AND SHOE BINDERBinder
663 (Bootmaker)BOOT & SHOE RIVETTERRivetter
663 (Bootmaker)SEWING MACHINISTSewing Machinist

The table shows examples of original occupation strings from Census records and their corresponding new variable “Task”. I assign 97% of them to “Task”.

Data · Task Data

Data Construction 1: “Task” Occupation

Every spelling of tailor found in the census strings:

…tailor', 'failor', 'tailar', 'tailaress', 'tailer', 'taileres', 'taileress', 'tailering', 'tailers', 'taillor', 'tailloress', 'tailor', "tailor's", 'tailor(journeyman', 'tailor(master', 'tailor,', 'tailor-', 'tailor-maker', 'tailor…', 'tailor:', 'tailore', "tailore's", "tailore'ss", 'tailorees', 'tailorefs', 'tailorer', 'tailorers', 'tailorerss', 'tailores', "tailores's", 'tailoress', "tailoress's", 'tailoress-', 'tailoress…', 'tailoresse', 'tailoresses', 'tailoresss', 'tailorest', 'tailories', 'tailoriess', 'tailoring', 'tailorings', 'tailoris', 'tailoriss', 'tailorist', 'tailorists', 'tailormaker', 'tailorman', 'tailorness', 'tailorous', 'tailorress', 'tailors', 'tailors…', 'tailorss', 'tailory', 'tailorys', 'tailoting', 'tailour', 'tailouress', 'tailours', 'tailress', 'tailroess', 'taioleress', 'tairloress', 'taitor', 'taitoress', 'talior', 'talioress', 'tallor', 'talloress', 'talloring', 'talor', 'talores', 'taloress', 'taloring', 'tayler', 'tayleress', 'taylor', "taylor's", 'taylores', 'tayloress', 'taylorest', 'tayloring', 'tayloris', 'tayloriss', 'taylorist', 'taylors', 'tialor', 'tialoress', 'tilloter', 'tailess'
Data · Task Data

Data Construction 2: Census Linking

  • Context: Census linking is indispensable
  • Concern: False positives pose threat to inference:
    1. Migration: O'Grada et al. (2025), 68–77% vs 14%
    2. Occupation: Helgertz et al. (2023)
  • Problem: Without ground truth, how to check?
  • Solution: Leverage statistical regularities of false matches: litmus test

Methods

Methods

1. Overview

  1. 6.5 million unique strings describing occupation
  2. Around 800 occupation codes
  3. The census gives us some help: enumerators worked to written instructions
    1. Most important job first
    2. second instruction — to come
Methods

2. Three Main Challenges

Methods

3. AI-Assisted Human Labeling

1Pre-LLM
Incremental machine learning model for topic assignment: a deterministic track feeds an ML-hybrid loop that auto-assigns confident cases, asks the user about uncertain ones, and updates the model iteratively.
Human in the loop
2Early LLM era
The same incremental machine learning model, with the user decisions now taken by Claude rather than by a person.
Claude in the loop
3100% Human Labelling (*kind of)
Block A
  • load and count
  • cleaning
  • top 100
  • draft from the dictionary
1 I decide
Block B
  • magnet
  • LLM confirms
2 I decide
Block C
  • priority rules
  • fragment + industry-only
  • residue pass
3 I sign off
Final
  • deliverables — five files
Methods

4. Current Pipeline

SCRIPT MODEL HUMAN DELIVERABLE
19 occodes · 5,510,622 records · 99.59% with a Task 1
Methods

5. Output

occode 663 · Shoe and boot makers (and repairers)
34,685 distinct strings · 1,300,892 records · 99.64% with a Task 1
stringrecordscum % task 1task 2product
SHOEMAKER204,19215.8%shoemaker—shoe
SHOE MAKER168,66528.9%shoemaker—shoe
BOOT MAKER95,59436.3%bootmaker—boot
CORDWAINER77,45842.3%cordwainer——
BOOTMAKER50,83046.2%bootmaker—boot
SHOE BINDER33,70348.9%binder—shoe
SHOE FINISHER31,23151.3%finisher—shoe
BOOT MACHINIST28,03653.4%machinist—boot
BOOT & SHOE MAKER24,39155.3%bootmakershoemakerboot
BOOT FINISHER24,32457.2%finisher—boot
BOOT REPAIRER23,76959.1%repairer—boot
SHOE MACHINIST18,99760.5%machinist—shoe
BOOT BINDER17,43461.9%binder—boot
BOOT CLOSER15,21663.1%closer—boot
SHOE RIVETTER14,88964.2%rivetter—shoe
SHOE FITTER12,72265.2%fitter—shoe
BOOT AND SHOE MAKER12,05666.1%bootmakershoemakerboot
BOOT FITTER10,40066.9%fitter—boot
SHOE CLICKER9,08967.6%clicker—shoe
CLICKER8,95568.3%clicker——
The twenty most frequent strings, 68% of all records in the occode. One row per distinct string; the map runs to every string of ten records or more.

Descriptive Evidence

Descriptive Evidence

1. Occupational Orders over Time

Descriptive Evidence

2. Orders ranked by growth, 1851–1911

Showing the growth in different sectors of the economy over the period. Sectors shown in blue are growing more rapidly than average population growth.

Descriptive Evidence

3. Occupational Industries: Growth and Decline

Showing growth by industry over the period. Note that the extreme outliers are primarily in industries which were very small or non-existent in 1851.

New occupations (new to the 19th century) Established occupations

Population doubled over the period: any industry growing more than 100% outpaced population growth, industries which grew less lagged.

Descriptive Evidence

4. Micro-Occupations: Growth and Decline

Each industry is itself made up of many different jobs — micro-occupations.

New Evidence

New Evidence

1. Map of specific new jobs

Where did specific new occupations emerge across the country?

New jobs of the 19th century

Ranked by number of workers, 1911.

Occupation Workers, 1911
Electrician55,447
Motor driver44,947
Typist41,927
Bicycle maker33,461
Telegraph33,129
Telephonist19,230
Photographer17,134
Asphalter1,612
Anaesthetist353
X-ray operatorAlso known as…
radiographer · skiagraphist · roentgenologist · x-rayist · x-ray man (even railway xrayman) · x-ray attendant · x-ray photographer		77
New Evidence

2. Mapping of management jobs

An initial mapping of the rise of management jobs in the UK, by county.

Percentage Share of the Working Male Population
New Evidence

3. Map of mechanization

An initial mapping of the emergence of new technologies in the UK, by county.

Mechanization rises on both measures. In share and absolute numbers, by 1911 there are 1.5 million more people working with machines.

Percentage share of male population
New Evidence

4. Mapping of the apprenticeship system

The apprenticeship system declines everywhere between 1851–1911. The decline is more rapid after 1881. Less urban areas seem to retain more of the system than elsewhere.

But this is a decline in share, not in absolute numbers. Summed across the three roles, the system actually grows slightly — from about 252,000 in 1851 to about 261,000 in 1911, a net rise of roughly 9,000. Apprentices themselves increase substantially, by almost 90,000; the real decline is among journeymen (down about 65,000) and masters.

Total Participation

% Share of Male Population

Role Breakdown

% Share of Male Population in Role

Results

A: Workforce in Transition
B: Social Mobility
C: Fertility
Results · A: Workforce in Transition

1.1. Measuring displacement

2 × 2 difference-in-differences, over two stacked transition windows: 1851–1861 and 1861–1881
exit rate, occupation i, county c, census τ
county fixed effects
treated occupation
after mechanization
the coefficient of interest
covariates, interacted with

Picking out an outlier

  • I run this regression 231 times.
  • shows how much an occupation’s exit rate moves vis-à-vis the average.
Results · A: Workforce in Transition

1.2. Who was displaced

Male occupation placebo DiD distribution A density curve formed from 231 large control occupations. Hollow circles mark each occupation. Bootmakers are highlighted in red at 0.68 percentage points, the 77.9th percentile.
Results · A: Workforce in Transition

1.3. Did contraction push workers out?

Change in exit
Results · A: Workforce in Transition

1.4. Or did it stop new ones coming in?

Change in entry
Results · A: Workforce in Transition

1.5. More unusual on exit, or on entry?

Each occupation is ranked twice against the same universe of 235 large male occupations. Once on exit rate increasing, once on how much entry collapsed.

Exit percentile against entry percentile
  • Everything is extreme on entry. All 18 sit above the 71st percentile of entry collapse.
  • Exit is what separates them.

Results

B: Social Mobility
Results · B: Social Mobility

2.1. Which industries had managers?

Occupations totaling 50% of all managers by census year
  • In 1851, seven occupations held half the managers.
  • By 1901 it took thirty-one.
  • Supervision spread out across the economy.
Results · B: Social Mobility

2.2. How many managers?

  • 25,000 managers in 1851.
  • 282,000 by 1911, an elevenfold rise.
  • More supervisors, spread over more trades.
Results · B: Social Mobility

2.3. Where did managers come from?

Results · B: Social Mobility

2.4. Fathers’ occupations in occupation space

Results · B: Social Mobility

2.5. Fathers’ occupational status

    Results · B: Social Mobility

    2.6. Where the new bootmaking jobs were

      Results · B: Social Mobility

      2.7. Growing up in a mechanizing county

      Were boys raised in Leicestershire or Northamptonshire more likely to become bootmaking foremen thirty years later?

      boy i, raised in county c, is a bootmaking foreman as an adult
      raised in a mechanizing county — Leicestershire or Northamptonshire
      the difference that childhood county made
      cohort and age fixed effects
      • 0.45pp more likely to become bootmaking foremen.
      • Unchanged with age and cohort controls.
      • All boys, so entry and advancement together — not promotion within the trade.

      Results

      C: Fertility
      Results · C: Fertility

      3.1. Sons entering their father’s occupation

      Results · C: Fertility

      3.2. Where the new bootmaking jobs were

        Results · C: Fertility

        3.3. Comparing brothers

        brothers in bootmaker-father households where some sons entered the trade and some did not
        adult outcome for son i in family f
        family fixed effects
        the son became a bootmaker
        local bootmaking fell by 20% or more; Growth counties are the reference, the middle is dropped
        age and birth order; errors clustered by county
        • Family FE holds parents, place and upbringing constant, and therefore comparison is between brothers.
        • is the gap between the son who followed his father and the son who did not.
        • is how much wider that gap is where the trade was collapsing.
        Results · C: Fertility

        3.4. The cost of persistence: children under 18

        Bootmaker sons Carpenter sons — placebo
        1851–18811861–1891 1851–18811861–1891
        Entered the trade 0.448***0.261* −0.0110.659**
        (0.107)(0.137) (0.270)(0.243)
        × Decline county −0.240*−0.220 0.351−0.348
        (0.137)(0.161) (0.276)(0.256)
        Total gap in Decline 0.208**0.041 0.340***0.312***
        [0.082][0.096] [0.060][0.073]
        Observations 3,3292,816 3,3562,920
        R2 0.4880.487 0.4680.499
        • Bootmaker sons had more children than their brothers, and Decline takes about half of that back.
        • No carpenter interaction is distinguishable from zero.
        Family fixed effects throughout; controls are age and birth order. Standard errors clustered by county in parentheses, and for the linear combination in square brackets. * p<0.10, ** p<0.05, *** p<0.01.
        Results · C: Fertility

        3.5. Age at marriage in old and new bootmaking

        Conclusion

        • From roughly 800 industries to about 10,000 micro-occupations.
        • Micro-occupations are necessary to track the emergence of new work and the decline of older jobs.
        • Rise of management jobs and factory work.
        • Decline of the apprenticeship system, in terms of shares. In absolute numbers, it increases by about 9,000.
        • Implications: how society absorbs technological shocks; who gets the new jobs.
        • Next steps: finalise boundaries and definitions of new jobs.

        Appendix

        Appendix

        1. The shape of the workforce: Orders, sub-Orders, and occupations

        Every occupation nests inside a sub-Order, and every sub-Order inside one of the 22 Orders. The circles below pack that whole structure, with each circle's area proportional to its 1911 workforce. Click any bubble to zoom in; click the background to zoom back out.

        Appendix

        2. Changing Taxonomies: Census Waves 1851–1911

        The grey treemap is the 1911 classification: Orders, their sub-Orders and the occupations within them. Hovering an occupation in an earlier census lights up the others it was lumped with — the more scattered, the more that census cut across the modern Orders.

        Census year: hover an occupation
        Appendix

        3. Migration within Orders, 1851 → 1861

        Zoom in one level. Even within a single Order, the census kept reorganising. Here the 22 Orders stay fixed as the outer bubbles, and inside each one the occupations are grouped by their real census category for the chosen year. Flip between 1851 and 1861 to watch occupations split apart, merge together, or reshuffle within their Order. Unchanged occupations stay grey.

        Census year:
        Split apart Merged together Reshuffled Unchanged