2 cumulative citations
View corpus contextOutside Sweden’s cities, occupations exposed to automation risk saw weaker job and pay growth from 2011–2021, but bigger local markets and closely related local skills softened the hit; urban occupations — richer in non-automatable tasks — showed no such decline.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
2 cumulative citations
View corpus contextThis paper examines how local labour market skill relatedness and size moderate the impacts of automation on occupations across Swedish local labour markets. Using administrative data and a spatially explicit risk of automation measure that accounts for regional differences in occupational task contents, it finds a negative association between automation and employment growth and wage income growth for non-metropolitan occupations between 2011 and 2021. Skill relatedness and labour market size mitigate these negative relationships. In contrast, no negative associations are found for metropolitan occupations. Due to their higher shares of non-automatable tasks, they are more resilient to adverse automation effects.
Summary
Main Finding
Automation is associated with lower employment growth and wage income growth for occupations in non‑metropolitan Swedish local labour markets (2011–2021). Higher local labour market skill relatedness and larger market size attenuate these negative effects. Metropolitan occupations show no such negative associations, likely because they contain a higher share of tasks that are less automatable.
Key Points
- The negative impact of automation on employment and wages is heterogeneous across regions: present in non‑metro areas, absent in metro areas.
- Skill relatedness — the extent to which an occupation’s skill set is similar to other local occupations — mitigates automation’s adverse effects.
- Larger local labour markets also buffer occupations from automation‑driven declines.
- Metropolitan areas are more resilient because their occupational task mixes include more non‑automatable tasks.
- Results highlight the importance of regional task composition and agglomeration in shaping automation outcomes.
Data & Methods
- Geographic and temporal scope: Sweden, 2011–2021, analysis at the occupation × local labour market level.
- Data: Swedish administrative employment and wage records (administrative microdata) combined with task/occupation content information.
- Automation exposure: a spatially explicit risk-of-automation measure that accounts for regional differences in occupational task contents (i.e., the same occupation can face different automation risk depending on local task mixes).
- Empirical approach: regression analyses linking local automation risk to subsequent employment growth and wage income growth, with tests for moderation by (a) local skill relatedness and (b) labour market size; comparisons between metropolitan and non‑metropolitan labour markets.
- Controls and identification: standard covariates and fixed-effects strategies implied to isolate associations (paper emphasizes robustness across specifications).
Implications for AI Economics
- Heterogeneity matters: aggregate estimates of automation impact can mask strong regional variation; models and forecasts should incorporate local task composition and urban/rural differences.
- Local complementarities and skill networks are protective: policies that foster skill relatedness (retraining that builds on existing local skills) may reduce displacement costs.
- Agglomeration economies moderate automation risk: larger labour markets and metropolitan agglomerations provide more non‑automatable task opportunities and occupational diversity, suggesting urbanization changes the distributional effects of AI.
- Policy targeting: rural and smaller labour markets may require more active labor‑market interventions (reskilling, mobility support, industry diversification) to cope with automation.
- Research agenda: incorporate spatially varying task content into structural models of automation; evaluate the causal role of reallocation, job creation, and local spillovers in mediating AI impacts.
Assessment
Claims (7)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| The analysis uses administrative data and a spatially explicit risk of automation measure that accounts for regional differences in occupational task contents. Other | positive | spatially explicit risk of automation / measurement approach |
Reading fidelity
high
Study strength
high
|
not reported
|
| There is a negative association between automation and employment growth for non-metropolitan occupations between 2011 and 2021. Employment | negative | employment growth |
Reading fidelity
high
Study strength
medium
|
not reported
|
| There is a negative association between automation and wage income growth for non-metropolitan occupations between 2011 and 2021. Wages | negative | wage income growth |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Skill relatedness mitigates the negative relationships between automation and labour-market outcomes (employment and wage growth). Skill Acquisition | positive | moderation of automation effects on employment and wage growth |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Labour market size mitigates the negative relationships between automation and labour-market outcomes (employment and wage growth). Labor Share | positive | moderation of automation effects on employment and wage growth |
Reading fidelity
high
Study strength
medium
|
not reported
|
| No negative associations between automation and employment or wage growth are found for metropolitan occupations over the study period. Employment | null_result | employment growth and wage income growth |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Metropolitan occupations have higher shares of non-automatable tasks and are therefore more resilient to adverse automation effects. Automation Exposure | positive | share of non-automatable tasks / resilience to automation |
Reading fidelity
high
Study strength
medium
|
not reported
|