Like moths to a flame: an individual level approach to technological change in 20th century Sweden

Final Seminar

Jonathan Jayes

Supervised by Kerstin Enflo and Jakob Molinder

2025-05-07

📖 Motivation

Why History Matters for the Future of Work

Viral Prediction of Computerization from Frey and Osborne (2017)

I show this figure from the viral paper by Frey and Osbourne from 2017 — they predicted that many tasks in low-skill service jobs were at high risk of automation due to computerisation. Since then, with large language models emerging that can pass professional exams, we must wonder if high skilled office work also contains lots of tasks that can be automated.

So Why does history matters for the debate on the future of work?

• Hisory provides the only way to analyze the full, long-term consequences of major technological transformations on employment, socieyy, and individuals including intergenerational effects

⚡ Context

Sweden’s Electrification

Income shares from Bengtsson and Molinder (2021) and rural electrification rates from Vattenfall (1948)

So why do we care about Sweden as a case study when we think about technological transformations?

there are at least two reasons; 1 - Sweden was a fast adopter of electricity. After Jonas Wenström discovered three phase current in the late 19th century, Sweden rapidly electrified - with the electrification of rural areas advancing rapdily between 1900 and 1921, when the Western line - an important part of the grid, was completed.

Second, Sweden in the early 20th century was undergoing a levelling in terms of inequality; we can see that the share of income going to to the top decile of the income distribution declined - so it is an intersting time with a great levelling taking place.

Further, electrification as a general purpose technology drastically changed that way that we worked and lived; agricultural tasks that used to be very labour intensive were mechanised and then electrified, freeing up labour to move in to factories - which also were completely transformed by the electric motor, improving the quality of the jobs in the factory compared to the risks of belt drives from a central drive shaft, and allowing modular factory production, running only some machines at a time.

⚡ Context

Electrification and the Grid Rollout

Grid rollout in Sweden from Vattenfall (1948)

Location of transformers in Sweden

At this time, I should also speak a little about the quasi-random rollout of the electrifcal grid across sweden, between two large hydrpower plants at Olidan and Alvkarleby - along what was called the “Western Line” - this was a choice by the state at the time to incresase the electrified area in Sweden, and connect these two plants as part of the national grid. You can see that the parishes that lay along this western line, through no action of their own, got access to reliable and affordable electricity prior to their neghbours.

As part of the dissertation, as you can see on the right, I collected detailed information on the location of every power station and transformer in Sweden in the 1920s in order to show carefully that the mechanism of the Western Line did in fact provide these parishes with early access to electricity, and this quasi-random variation is what we use to get causal estimates in paper 2 and paper 3 of the dissertation.

🔬 Scope

Research Aims

• To investigate the impact of technological change, primarily electrification as a General Purpose Technology (GPT), on the Swedish economy in the 20th century.

• To understand how this transition affected labor market outcomes (income, employment, inequality) across different skill groups

• To analyze the role of human capital, particularly high-skilled engineers, and corporate governance in shaping these outcomes.

The central aim of this thesis is to investigate the impacts of new technology on the Swedish economy during the 20th century. I focus on the effects of electricity on the labor market – looking at income, employment, and inequality – across different skill groups. I also examine the roles played by human capital, especially high-skilled engineers, and corporate governance structures in shaping how this technological shift unfolded. The overarching question driving this work is: How does technological change impact labor market outcomes, skill demands, income distribution, and firm-level dynamics

📑 Papers

Characteristic	Paper 1	Paper 2	Paper 3	Paper 4
Title	Has Technology Historically Destroyed Jobs?	Power for Progress	Praise the People or Praise the Place?	Technocrats to Tycoons
Data Type	Economic History Literature	Census Data, Administrative Electrification Records	Biographical Data	Company Annual Reports, Financial Data
Analysis Type	Systematic Review, Narrative Review	Cross-sectional Regression Analysis	Quantitative Historical Analysis, Probit Regression	Panel Data Regression Analysis / Event Study
Authorship	👥	👥	👤	👤
Time Period	1850-1980	1930	1910-1930	1873-1980

I tried to do a few different things in my dissertation in order to improve my skills; so in paper 1 with Ben Schneider I do a big systematic and narrative review on labour dispalcement and labour replacement in order to understand what we do and don’t know about when technology destroys jobs - this is not common in economic history, still trying to work out exactly how to place it.

In paper 2, with my supervisors, we make use of detailed individual level census data where we observe income in order to see what happens on the labour market in places that by luck get access to electricity prior to their neighbours,

In paper 3 I use new ML methods to structure biographic information from free text into a structured format that can be augmented with e.g. coordinates of places that engineers and business leaders lived, and I use this data to try and determine if it was the people or the place that was more important to early electrifying areas that benefitted, as shown in paper 2.

Then in paper 4, I use the method proposed by Acemoglu, le Matire and He (2023) to see what the impact of a board member bringing specific kinds of education and experience (a technology or way of doing things) to Swedish companies in the 20th century.

🖨️ Methods

More on Methods

A lot of my dissertation is built around new tools that allow us as economic historians to compile a large amount of data to analyze in ways that were not possible before.

i’ve spent quite a long time working with these, and paper 3 goes into some detail on how you as a researcher can use the new tools like Large Langauge Models to structure data that is in a very time consuming format to work with by hand into a database that you can augment and then analyze.

While I won’t go into detail about those methods in this presentation; there is a lot of information available in my public github profile that you can copy.

Paper I

Has Technology Destroyed Jobs?

A Systematic and Narrative Review of Historical Labor Displacement

Jonathan Jayes

Benjamin Schneider

Research Question: What does the historical literature reveal about the conditions under which technological change has led to labor displacement, and what are the key gaps in our quantitative understanding of these episodes?

Results: Largest effects occur when:

• Technology directly substituted for existing manual or craft labor (e.g., agricultural mechanization, automated bottle making)

• Technology automated routine tasks (e.g., telephone operation, early data processing).  

Geographic Focus: Global (published in English)

Data Sources: Published Studies

Books

1850

1980

1990

Great - on to the papers!

Before diving into the Swedish case, Paper 1 (co-authored with Benjamin Schneider) takes a broader historical view. We systematically reviewed the literature on technological labour displacement before 1980. The key question was: Under what conditions has technology historically replaced workers, and where are the quantitative gaps? These gaps we discovered from the narrative review. The idea was to look to history journals and books by historians to find instances of technological change that destroyed jobs, but for which there have not been many quantitative studies.

Paper I

Systematic Review Findings

Technological change repeatedly caused significant labor displacement and this was most pronounced when:  

• Technology directly substituted for existing manual or craft labor (e.g., agricultural mechanization, automated bottle making)

• Technology automated routine tasks (e.g., telephone operation, early data processing)

Narrative Review Findings

Significant technology-driven shifts occurred in sectors like domestic service, cultural industries (e.g., musicians), pastoral farming, and early ‘white-collar’/clerical work, but these are often poorly documented in quantitative data.

Points to where future research could focus

Settings of studies in the reviews

We found clear evidence of displacement, especially when tech directly substituted for manual or craft skills – think agricultural mechanization or early factory automation. However, the outcomes were highly context-dependent. Importantly, the review also highlighted significant episodes, like in domestic service or cultural industries, where displacement likely occurred but lacks rigorous quantification - for example the replacement of live orchestras in early cinemas, and automatic piano players in hotels; these might seem trivial, but on aggregate, we just don’t know how many musicians lost their jobs as a result of this; and it points to where future economic history could research.

One limitation that I want to point out is that our study only looked for studies published in English - hence we see a lot of studies on Western Europe, the United States, and the British Colonial offshoots like Australia, South Africa, and Canada. This is not to say that historical labour displacement did not take place in say, South America or other Sub—Saharan Africa just that we don’t find these studies in our search - which I can talk a little more about later.

Paper II

Power for progress

The impact of electricity on individual labor market outcomes

Jonathan Jayes

Kerstin Enflo

Jakob Molinder

Research Question: How did access to electricity impact individual labor market outcomes (income, employment, inequality) in Sweden?

Results:

• Early electricity access significantly raised incomes, by 30-45%.

• Gains concentrated among lower-income individuals and those with primary education, reducing local inequality.

• No evidence of increased unemployment; instead, occupational upgrading observed.

Geographic Focus: 🇸🇪

Data Sources: Swedish Censuses

Electrification Reports

1850

1930

1980

Paper 2 (co-authored with Kerstin Enflo and Jakob Molinder) zooms into the Swedish case - this paper is conditionally accepted at Explorations, which we are excited about. We use the rollout of the ‘Western Line’ power grid between 1909 and 1921 as a quasi-natural experiment.

Crucially, we use newly digitized 1930 census data which, uniquely for this period, includes individual income and education, allowing a granular look at who was affected and how. THE OUTCOME WE MEASURE IS IN 1930.

Our first key finding from Paper 2 is striking. This new technology did not destroy jobs in the early electrifying parishes; and instead individuals born in parishes with early electricity access earned significantly higher incomes by 1930 – around 30-45% higher on average than comparable individuals in control areas.

Paper II

Income gains were largest at the bottom of the income distribution

Coefficient plot from quantile regression

But critically, as this quantile regression plot shows [Point to visual], these gains weren’t evenly distributed. The largest positive income effects occurred in the lower parts of the income distribution, meaning electrification actually reduced local income inequality. This contrasts with standard Skill-Biased Technical Change narratives that are studied in the United States; which saw a hollowing out of the middle of the income distribution as there was more demand for low-skilled labour.

Paper II

Electricity related jobs emerged in early electrification parishes

Share of jobs across three job categories

More on Paper II

How do we explain these gains? What is really nice is that we can show the mechanism through which these gains occurred; here we plot the largest 3 occupations in non-electricity jobs, indirect electricity jobs, and direct electricity jobs - classifications based on the individual’s occupation. What is neat is that we can see in the parishes which got early access to electricity, more individuals hold these emerging electricity related jobs like electricians, workers on electricity generators and linesmen, as well as factory workers.

So what is our takeaway? workers in the middle of the skill distribution really benefited from this new technology; and they didn’t need specific education in order to do so; they also didn’t move that far away from home when these new jobs came along; the median distance from birthplace to place of residence for these workers in 1930 was 24 km.

Paper III

Praise the people or praise the place?

Upper tail human capital in electrifying Sweden

Jonathan Jayes

Research Question: To what extent were engineers relocating to early electrification areas distinguishable by their educational or network backgrounds compared to peers?

Results:

• High-skilled engineers were highly geographically mobile, vs medium-skilled workers.
• Migrants moving into these parishes were more likely than locals to have KTH degrees and general overseas experience, indicating active recruitment of specific external human capital.

Geographic Focus: 🇸🇪

Data Sources: Biographical Dictionaries

Industrial Catalogue

1850

1900

1930

Vem är Vem

1980

So, electrification benefited the broad, less-educated workforce locally. But who were the highly skilled individuals – the engineers and business leaders – actually designing, implementing, and managing these new systems? Paper 3 asks: Were the pioneers active in these early electrifying ‘Places’ distinctive ‘People’?.

To answer this, I constructed a novel dataset from roughly 66,000 entries in the ‘Vem är Vem?’ (‘Who is Who?’) Biographical dictionaries using Large Language Models and Natural Language Processing to extract detailed career and background information. I should say here that the Vem är Vem biographical dictionaries were published between 1945 and 1969; but contain information on what the individuals who are included were doing at each stage of their career, and crucially, betwen 1900 and 1930 when the early electrifying parishes had their advantage.

Then I use the same early electrifying parishes as a subsample to ask, to what extent were engineers relocating to early electrifying areas distinguishable by their educational or network backgrounds compared to their peers?

Paper III

Findings on the upper tail

• Engineers and managers in early electrification parishes were more likely to have KTH degrees and overseas experience in the USA, indicating active recruitment of specific external human capital.

Map of Engineers’ and Managers’ US experience

Unlike the relatively immobile workers in Paper 2, these high-skilled individuals were highly geographically mobile, often moving hundreds of kilometers for education and work, especially in electricity-related sectors; where electrical engineers move more than 300km on average, compared to the 24 for the median Sweden in 1930.

We can also see that these engineers and business leaders working in the early electrifying parishes were more likely to have studied at KTH than the other engineers and managers, and specifically were likely to have more experience in the United States, often at leading firms like General Electric or Ford. This US experience, involving exposure to cutting-edge technology and management, acting as a key channel for knowledge transfer. This is in line with Per Olof Groenberg’s research on four engineering firms in Sweden - who found that importing skills back from the USA was key to their success; here thanks to new ways to gather data, we can expand the analysis to the nearly the entire population of engineers and managers in Sweden at the time, about 2 percent of the population of the entire country.

Paper III

Probit coefficient plot for Western Line workers

More on Paper III

Finally, we can show that the skilled workers in early electrifying parishes were disproportionately from agricultural or production backgrounds, suggesting different pathways into the elite compared to the rest of the country. it appears that in addition to decreasing income uneqaulity, the early electrifying areas allowed a different class of individuals to enter the upper tail of the skill distribution.

Paper IV

Technocrats to Tycoons

The Shift in Swedish Corporate Leadership and Its Economic Consequences in the 20th century

Jonathan Jayes

Research Question: What were the firm-level consequences of having board members with specific business and technical education, and international (U.S.) experience in the 20th c?

Results:

• Event studies show limited/no significant impact of first appointments of directors with Business, Technical, or US Experience backgrounds on key outcomes (Profitability, Employment, Labor Share Proxy).

• Findings differ from contemporary studies (e.g., AHLM 2022)

Geographic Focus: 🇸🇪

Data Sources:

Annual Reports (SSE)

Biographical Dictionaries

1850

1873

1980

Finally, Paper 4 shifts focus to the firm. Did the type of leader matter for firm-level outcomes? Specifically, did appointing directors with engineering backgrounds (especially US experience) versus those with business/finance training lead to different results in firm performance or how gains were shared with employees?

To answer this, I constructed a long-run panel dataset for 71 firms from 1873-1980 using novel multimodal LLM techniques to extract data from historical annual reports, and matched directors to biographies in order to get information about their education and experience.

Paper IV

Setup: comparing results of Acemoglu, He, and leMaire (2022) with the the setting of Sweden in the 20th century.

Hypothesis: different time period and content of the education mean we will not find the same results as in the US and Denmark post 1980.

Acemoglu, He, and leMaire (2022)

• Acemoglu, He, and leMaire (2022) study the effect of managers with a business degree (“business managers”)on firm performance in the USA and Denmark.

• They find managers with a business degree reduce their employees’ wages, but do not improve firm performance in profits or sales.

• Match 15,301 directors in my Who is Who dataset with the 71 firms with at least 30 years of data in the SSE company reports archive

• Event study; what is the impact of appointing a director with a) business degree, b) technical degree, c) US experience on the firm performance?

Paper IV

Directors’ education over time

Paper IV

Event study showing impact of appointing a director with business education on return on assets

indeed we do see the share of directors with business or engineering education varying throughout the time period, but when we look at the event study plots, there is no pre-trend and no effect of business education on return on assets or employment.

Paper IV

Event study showing impact of appointing a director with business education on number of employees

🔑 Takeaways

• Electrification’s Positive Distributional Impact in Sweden (Challenging Universal SBTC).

• Heterogeneity Matters: Different Impacts/Mobility for Medium vs. High Skills.

• Skills & Knowledge Transfer: Role of Mobile Engineers, US Experience.

• Leadership & Context: Board Impact Seems Context-Dependent (Historical Sweden vs Modern Denmark and USA).

• Methodological Contributions (AI/LLMs, New Data).

• Electrification’s positive distributional impact in Sweden (challenging universal SBTC).

• Heterogeneity matters: Different impacts/mobility for medium vs. high skills.

• Skills & Knowledge Transfer: Role of mobile engineers, US experience.

• Leadership & Context: Board impact seems context-dependent (historical Sweden vs modern).

• Methodological contributions (AI/LLMs, new data).

📷

Appendices

More context

Stocking Frame Knitters

Paper I Appendix

Paper I Appendix

Setup: Systematic and Narrative Review

Looking specifically for instances of labour displacement and labour replacement due to technological change.

Sankey Diagram of Search Process

Paper I Appendix

Table organizing the technologies examined in the systematic review:

Technology Category/Theme	Specific Examples of Technologies Examined
1. Steam Power & Industrial Mechanization	Coke smelting, Blast furnaces, Bessemer & Siemens-Martin processes (iron/steel), Steam Power (general industry adoption), Steam Engines (shipping)
2. Mechanization & Automation in Agriculture	Tractors (general purpose, pre-harvest), Mechanical Cotton Pickers, Wire Fencing, Poison (dingo control - complementary to fencing), Windmills/Windpumps, Milking Machines (bucket, pipeline, parlour), Set-field agriculture (vs. slash-and-burn)
3. Industrial Revolution & Factory Automation	Factory machinery (general), Looms (power looms) & Spinning tech (Jenny, Water-Frame, Mule), General “Industrial Rev. tech”, Factory system, General Mechanization (replacing hand methods)
4. Electrification & Telecommunications Automation	Electrification (general manufacturing impact), Telephone Automation (Mechanical Switching), Computers (utility plant monitoring/control), Nuclear Power (enabling larger units), Mechanized Line Work (utilities), Computer Process Control (petroleum refining), Advanced Refining Tech (Cracking, Desulfurization, Reforming), Ring Spinning & Automatic Looms (textiles, in electrification context)

Paper I Appendix

Technology Category/Theme	Specific Examples of Technologies Examined
5. Labour-Saving Factory Technology	Mass production methods, Threshing Machines, Glass Machinery (Owens automatic bottle machine, bulb/tube/window glass machines), Printing Machines (Linotype, Rotary presses, Autoplate), Tire Machines (Banbury mixers, tire-building machines, vulcanizers, conveyors), Knitting Machines (automatic hosiery), Hosiery Machines (general automation), Can Lines (two-piece, welded), Laundry Tech (steam tunnel finishing, automated washing/handling), Apparel Tech (refined cutting/sewing, NC), Footwear Tech (synthetics, water jet/laser cutting, NC stitching, molding, lasting/bottoming methods), Auto Tech (computers, NC machines, transfer lines, robots), Box Machinery (automated diemaking, faster presses, auto stripping/gluing/packing), Semi-automatic bottle machines
6. Automation in Service Industries	Household Technologies (appliances like stoves, vacuums, washers; prepared goods), EDP/Computers (general office, banking, insurance, government), MICR (banking), Retail Tech (Computers, POS, Self-service, UPC scanners, microfilm)
7. Automation in Primary Industries	Coal Mining Tech (Mechanical loaders, Continuous miners, Strip mining equipment, Longwall systems, Computers), Oil & Gas Extraction Tech (Offshore platforms/drill ships, Enhanced recovery methods, Computers, Advanced exploration), Computerized Sawmills
8. Automation in Transportation	Rail Technology/Automation (Dieselization, Mechanized track maintenance, Centralized Traffic Control (CTC), Office automation/Computers, Automated classification yards), Pipeline Automation (Computerized scheduling/control, Improved pipes/pumps, Automatic metering)

Paper I Appendix

Table organizing the technologies examined in the narrative review:

Narrative Theme	Examples of Technologies Discussed Qualitatively
1. Mechanization of Workshop Industries	Early Textile Machinery (Spinning Jenny, Water Frame, Mule, Power Loom), Factory System, Framework Knitting Machines, Steam Press, Mechanized Papermaking, Stereotyping, Linotype/Monotype, Machine Binding, Mechanization of Printing/Shoemaking/Bootmaking (General), Sewing Machines, Cutting Machines, Calculating Engines, Portsmouth Block-making Machinery
2. Reshaping Production Lines	Continuous-process Operations, Advanced Factory Mechanization (General), Owens Automatic Bottle Machine, Coal-cutting Machines, Bandsaws, Rationalization Techniques (linked to tech adoption), Automation (e.g., Flash Removal), Capital-intensive Moulding Presses
3. Transformation of ‘White-Collar’ & Service Work	Early Office Machinery (Typewriter, Adding Machines, Calculators), Early Computers & Data Processing (EDP), Punched-card Machines (Hollerith), Automatic Elevators, Cash Registers, Automation in Ticketing/Scheduling (Transportation)
4. Cultural Industries & Other Specific Cases	Mechanical Musical Instruments (Player Pianos, Automatic Organs), Sound Synchronization (Cinema), Projector Technology Simplification, Containerization (Shipping), Satellite Communications (GMDSS), Mechanization (General, e.g., Dairying), Automation (Electronics - e.g., Printed Circuitry)
5. Agriculture & Primary Resources (Narrative Focus)	Wire Fencing & Windmills/Windpumps (where displacement delayed/complex), Set-field Agriculture (vs. slash-and-burn), Tractors & Cotton Pickers (impediments/social context), Central Mills (Sugar), Sawmill Automation (where linked to recession/income loss focus)

Paper I Appendix

Screening Tool built for the systematic review in Streamlit

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper II Appendix

Paper III Appendix

Paper III Appendix

Example of biographic data explorer

Paper III Appendix

Link to Github repo with guide to scraping biographies and structuring the data

Paper III Appendix

Paper III Appendix

Paper III Appendix

Paper III Appendix

Paper III Appendix

Paper III Appendix

Paper III Appendix

Paper IV Appendix

Paper IV Appendix

Paper IV Appendix

Paper IV Appendix

Top 15 Institutions of Business education in Sweden (1900-1930)
Institution	Count
Göteborgs Handelsinstitut	625
Stockholm	483
Lund	439
Uppsala	388
Handelshögskolan	331
Stockholms Handelshögskola	263
Universitet	226
Handelshögskolan i Göteborg	221
Utländsk	200
Göteborg	167
Lunds Universitet	163
Uppsala Universitet	146
Stockholms universitet	143
Handelshögskolan i Stockholm	136
Stockholms Högskola	133
Tekniska Högskolan	130
Sverige	123

References

Acemoglu, Daron, Alex Xi He, and Daniel leMaire. 2022. “Eclipse of Rent-Sharing: The Effects of Managers’ Business Education on Wages and the Labor Share in the US and Denmark.” Working Papers, no. 22-58 (December).

Bengtsson, Erik, and Jakob Molinder. 2021. “What Happened to the Incomes of the Rich During the Great Levelling? Evidence from Swedish Individual-level Data, 1909–1950.” What Happened to the Incomes of the Rich During the Great Levelling? Evidence from Swedish Individual-Level Data, 1909–1950, Lund Papers in Economic History,.

Frey, Carl Benedikt, and Michael A. Osborne. 2017. “The Future of Employment: How Susceptible Are Jobs to Computerisation?” Technological Forecasting and Social Change 114 (January): 254–80. https://doi.org/10.1016/j.techfore.2016.08.019.

Vattenfall. 1948. “Procentuellt Antal Elektrifierade Hushåll.” Vattenfall.