When considering the advancement of artificial intelligence (AI), a general-purpose technology, opinions vary regarding the future value and demand for expertise in human labour. Techno-optimists believe AI will increase productivity to such an extent that labour requirements will diminish, necessitating a universal basic income grant to sustain human consumption. Techno-pessimists believe it will automate many white-collar jobs entirely, as surveyed in Ernst et al. (2019). Both views predict that inequality will increase and the labour force will continue to bifurcate – high-skilled workers designing AI systems will become rich, while middle-skilled workers will be pushed into poorly paid service jobs – echoing previous technological junctures brought about by the adoption of general-purpose technologies. Research on robotisation of the workplace tells us that the routinisation of workers’ tasks worsened pay and working conditions, resulting in increased wage inequality (Acemoglu and Restrepo 2020), and that the adoption of robots decreased the employment share of occupations intensive in routine tasks (Wacker et al. 2016).
However, three recent studies have found that AI has helped novice or lower-skilled workers catch up to high-skilled workers across various white-collar tasks. Noy and Zhang (2023a, 2023b) found significant improvements in writing speed and quality among those using ChatGPT, with the biggest quality improvements concentrated at the bottom of the skill distribution. Brynjolfsson et al. (2023) found that access to AI tools increases productivity by 14% on average, with a 34% improvement for novice and low-skilled workers but minimal impact on experienced and highly skilled workers. Dell’Acqua et al. (2023) found that consultants at the Boston Consulting Group who used AI completed more tasks more quickly, with gains concentrated among those at the bottom of the skill distribution. Overall, Filippucci et al. (2024) conclude that “emerging micro-level evidence [on AI] shows substantial improvements in firm productivity and worker performance”.
One way to inform the debate is to look for similar examples of rapid technological change in history. In a recent paper, we study the rapid introduction of electricity in early 20th century Sweden (Jayes et al. 2024). We find that this transformative technology benefitted those at the bottom of the income distribution, resulting in higher incomes, lower inequality, and the emergence of new occupations accessible to workers with just primary education.
Our findings suggest that the impact of technological change on labour markets depend crucially on the features of the technology itself and the institutional context in which it is adopted.
The Swedish electrification experience
Sweden’s electrification during the early 20th century provides a unique opportunity to study the impact of rapid technological change on labour markets. The Second Industrial Revolution, marked by advancements in steel production, electricity, and assembly-line manufacturing, gained momentum during this period (Frey 2019). Electricity played a key role in reforming production methods through mechanisation, revolutionising the organisation of factory floors and farming practices (Alexopoulos and Cohen 2016).
Our study focuses on the rollout of the main electricity network between two large hydroelectric power stations, Olidan and Älvkarleby, during the first decades of the 20th century in Sweden. The location of these power stations was selected due to their superior natural capacity for hydroelectricity generation, and the linking grid, called the Western Line (Centralblocket), was completed in 1921. The parishes along this line are shown in orange in Figure 1. This grid served as the initial pillar of ‘the electric mainline system’ and connected Olidan and Älvkarleby in a straight-line manner, independent of previous local conditions.
Figure 1 Map of the Western Line connecting hydropower plants in Sweden, observed parishes in 1930 census, right
By examining areas strategically situated along this line, we can pinpoint regions that gained access to stable and reliable electricity, providing them with an advantage over other comparable areas. Between 1915 and the completion of the Western Line in 1921, the share of electrified rural households increased rapidly, reaching over 40% by 1930, as shown in Figure 2. Subsequently, in the post-war era, electrification expanded rapidly, eventually reaching nearly the entire population. Concurrently, there was an inflection in the income distribution. The share of income going to the top 10% fell fast, and individuals between the 50th and 90th percentiles of the income distribution became better off.
Figure 2 Evolution of income shares in Sweden and rural electrification rate
The Swedish electrification experience provides a unique setting to study the impact of new technology as an equalising force. We examine individual outcomes in the labour market, as well as the role of institutions such as education and local union presence in shaping these outcomes (Bengtsson 2019, Acemoglu and Johnson 2023).
How electricity reshaped the labour market
We find that individuals born in parishes along the main electricity grid, and thus exposed to new technology early, earned more than 30% higher incomes by 1930. They were also more likely to be employed in occupations augmented by electrification, both directly and indirectly.
In Figure 3, we show the three most common occupations in three job categories. There were fewer farm workers in the parishes with early access to electrical technology and more farmers. In jobs indirectly related to electricity, higher rates of holding these occupations were found in parishes with early access to electricity. Similarly, while the absolute rates of direct electricity job holding were low, there were more electricians, workers on generators and linesmen in early access parishes. These groups of parishes were comparable in terms of occupational structure prior to the electrification (in 1890 and 1900).
Figure 3 Share of employment in selected jobs, with birth parish type denoted by column and job type denoted by row
Our findings suggest that electrification replaced mainly unskilled occupations in agriculture while creating new positions that were filled by people with no more than a primary education.
Examining impacts across the income distribution, we find that gains were concentrated in the lower end of the spectrum, with the most pronounced effects among those with only a primary education. Figure 4 displays the coefficient on being born in a parish that lay along the Western Line in quantile regression. We show that early access to electricity resulted in a large income premium for those at the bottom of the income distribution in 1930, after controlling for observables such as age, gender, schooling, marital status, and class of occupation.
Figure 4 Coefficient plot for quantile regression, showing log-income premium for individuals ranked by income deciles
Implications for contemporary debates
Our findings from early 20th century Sweden have implications for current discussions on technology, labour, and inequality. Electrification in Sweden led to the creation of new jobs accessible to workers with primary education, contributing to reduced inequality and inclusive growth. Our study suggests that the impact of technological change on labour markets may depend on the features of the technology itself. Similar to early findings on AI that suggest that the technology provides a catch-up effect, new technology worked as a boon to novices and those workers in the bottom half of the skill distribution.
However, as our understanding of AI technology improves, policymakers should pay close attention to the kinds of new jobs that are created. We may yet need expanded social safety nets. Moreover, our findings underscore the importance of investing in high-quality basic education and skills development to ensure that workers can adapt to and benefit from technological change.
References
Acemoglu, D, and S Johnson (2023), Power and progress: Our thousand-year struggle over technology and prosperity, PublicAffairs.
Acemoglu, D, and P Restrepo (2020), “Robots and jobs: Evidence from US labor markets”, Journal of Political Economy 128(6).
Alexopoulos, M, and J Cohen (2016), “The medium is the measure: Technical change and employment, 1909 – 1949”, Review of Economics and Statistics 98(4): 792–810.
Bengtsson, E (2019), “The Swedish Sonderweg in question: Democratization and inequality in comparative perspective, c. 1750–1920”, Past and Present 244(1): 123–61.
Brynjolfsson, E, D Li, and L R Raymond (2023), “Generative AI at work”, NBER Working Paper 31161.
Ernst, E, R Merola, and D Samaan (2019), “Economics of artificial intelligence: Implications for the future of work”, IZA Journal of Labor Policy 9(1).
Filippucci, F, P Gal, C Jona-Lasinio, A Leandro, and G Nicoletti (2024), “Should AI stay or should AI go: The promises and perils of AI for productivity and growth”, VoxEU.org, 2 May.
Frey, C B (2019), The technology trap: Capital, labor, and power in the age of automation, Princeton University Press.
Jayes, J, J Molinder, and K Enflo (2024), “Power for progress: The impact of electricity on individual labor market outcomes”, CEPR Discussion Paper 18973.
Noy, S, and W Zhang (2023a), “Experimental evidence on the productivity effects of generative artificial intelligence”, Science 381: 187–92.
Noy, S, and W Zhang (2023b), “The productivity effects of generative artificial intelligence”, VoxEU.org, 7 June.
Wacker, K M, E Gentile, S Miroudot, and G de Vries (2016), “Robots replace routine tasks performed by workers”, VoxEU.org, 29 October.
