What Happens When Automation Displaces Workers?

Ager, Goni, and Salvanes (2024) study whether gender-biased technological change—specifically the adoption of milking machines—pushed young rural women out of farming and improved their long-term economic outcomes. They use linked Norwegian administrative registry data covering about 725,000 women and men born in rural municipalities, combined with municipality-level agricultural census data on milking machine adoption. They find that milking machines displaced young women from farm work much more than men, increasing female rural-to-urban migration. Although women experienced short-run income losses, a one-standard-deviation increase in machine adoption raised women’s long-run income rank by nearly two percentiles, increased education, and reduced gender gaps in earnings and labor force participation.

Why This Article Was Selected for The Policy Scientist

This article addresses a policy issue of broad and continuing importance: how labor-saving technologies reshape gendered labor markets, mobility, and long-run economic opportunity. As automation and artificial intelligence again raise concerns about displacement, the paper provides historically grounded evidence that technology shocks can alter entrenched labor allocations rather than merely reduce employment. The study is timely because current debates often lack credible long-run evidence on adjustment and reallocation. The authors are well established in the literature on structural change, migration, and labor markets, and this paper builds carefully on classic work on automation and sectoral reallocation. The Norwegian registry data are exceptionally high quality and well suited to the research question. The causal inference strategy is strong, transparent, and persuasive, making the findings plausibly generalizable to other settings with similar institutional features.

Full Citation and Link to Article

Ager, P., Goñi, M., & Salvanes, K. G. (2026). Gender-biased technological change: Milking machines and the exodus of women from farming. American Economic Review, 116(1), 246–286. https://doi.org/10.1257/aer.20240167

Central Research Question

This article asks whether gender-biased technological change in agriculture—specifically the adoption of milking machines in mid-twentieth-century Norway—pushed young rural women out of farming and, if so, whether this displacement improved or worsened their long-run economic outcomes. The authors focus on how automating a task traditionally performed by women altered labor allocation across sectors, migration patterns, education, and lifetime earnings. More broadly, the study examines whether a large, gender-specific automation shock reduced inefficiencies in labor allocation created by high moving costs and rigid gender norms, thereby narrowing gender gaps in income and labor force participation over the long run.

Previous Literature

The paper builds on several established strands of research. First, it contributes to the literature on structural change and agricultural mechanization, which documents how labor-saving technologies drive workers out of agriculture and into other sectors. While this literature has largely focused on aggregate productivity and sectoral shifts, it has paid less attention to gender-specific effects. Second, the study connects to research on automation and labor markets, including work showing that technological change often produces short-run displacement costs but ambiguous long-run welfare effects. Third, it engages with a large literature on female labor force participation and occupational change over the twentieth century, which emphasizes education expansion, sectoral growth in services, and evolving gender norms. The authors also draw on work highlighting migration frictions and misallocation, particularly studies showing that high moving costs can trap workers in low-productivity sectors. Relative to this literature, the paper’s contribution is to show how a narrowly targeted technology shock can disrupt gendered labor arrangements and trigger long-term reallocation that benefits the displaced group.

Data

The analysis relies on exceptionally rich Norwegian administrative and census data. The core sample includes roughly 725,000 women and men born in rural municipalities who were aged 16–25 between 1930 and 1970, the age range most exposed to farm employment decisions. Individual-level registry data allow the authors to track people over decades, observing education, occupation, earnings, migration, labor force participation, and fertility. These data are linked across sources using unique personal identifiers, eliminating concerns about matching error and allowing women to be followed despite name changes after marriage.

These individual records are combined with municipality-level agricultural census data from 1929 through 1969, which report the number of farms, dairy cows, and milking machines. This linkage allows the authors to measure local exposure to milking machine adoption during the period when individuals were young. The quality of the data is unusually high by historical standards: coverage is near universal, measurement error is limited, and outcomes can be observed well into adulthood, making it possible to assess both short- and long-run effects.

Methods

The authors employ a causal inference strategy based on plausibly exogenous variation in exposure to milking machines. Their approach resembles a shift-share design. Local exposure is constructed by interacting the national rollout of milking machines over time with pre-existing municipal differences in dairy intensity, measured by the number of dairy cows per farm in 1930. This design exploits the fact that milking machines automate a task previously performed almost exclusively by young women, while the geographic distribution of dairy farming was largely determined by natural conditions rather than labor market trends.

This exposure measure is used as an instrument for actual milking machine adoption at the municipality level in a two-stage least squares framework. The models include municipality and birth-cohort fixed effects, county-specific cohort trends, and controls for pre-existing agricultural characteristics. Gender interactions allow the authors to estimate differential effects for women and men. The paper devotes substantial attention to validation, including placebo tests using outcomes from earlier cohorts, permutation tests, event-study analyses, and robustness checks addressing alternative specifications and spatial correlation. While the study does not use a randomized controlled trial, the empirical design closely aligns with modern standards for quasi-experimental causal inference.

Findings/Size Effects

The results show that milking machines had large and persistent gender-specific effects. In the short run, young rural women in dairy-intensive areas were pushed out of agricultural employment and experienced income losses associated with job displacement. These effects are consistent with historical accounts of milkmaids losing paid work when farms mechanized.

In the long run, however, the outcomes for affected women improved substantially. A one-standard-deviation increase in milking machines per farm—roughly equivalent to one additional machine per ten farms—raised women’s position in their birth-cohort income distribution by nearly two percentile ranks. Women exposed to higher machine adoption were more likely to migrate out of rural areas, invest in higher education, and work in skilled occupations, particularly in the expanding public sector. Labor force participation among women increased, and long-run earnings rose despite the initial disruption.

Men were also affected by agricultural mechanization, but to a much smaller extent. While some men left farming, many remained in rural areas or moved into other primary-sector or manual jobs that continued to offer returns to existing skills. As a result, gender gaps in income and labor force participation narrowed. The authors also find effects on family outcomes: exposed women delayed childbirth and had fewer children, consistent with higher opportunity costs associated with education and skilled employment.

Conclusion

This study provides compelling evidence that a gender-biased automation shock can reshape labor markets in ways that extend far beyond the immediate task it replaces. By eliminating a key source of female farm employment, milking machines forced young women to reallocate across sectors, overcoming migration frictions and contributing to long-term gains in education and earnings. The findings challenge interpretations of automation that focus narrowly on job loss and instead highlight the importance of adjustment, mobility, and institutional context. Methodologically, the paper stands out for its careful use of high-quality administrative data and a transparent causal inference strategy. Substantively, it offers lessons for understanding contemporary automation debates by showing how technology can alter deeply entrenched gendered labor arrangements and produce long-run distributional effects that differ sharply from short-run impacts.