Highly digitalizable sectors did not generate net job gains during the COVID-era but paid more and went remote: wages rose by about €0.52/hr (≈4.6%) and remote work surged by ~41 percentage points compared with less-digitalized activities.

Main Finding

Using the COVID-19 shock as a quasi-natural experiment and a difference-in-differences design on quarterly EU-27 sectoral data (2018–2024, N = 36,685), the study finds that higher sectoral digitalization potential (classified by telework feasibility and digital intensity) had: - No detectable effect on aggregate employment levels (DiD = 0.06, p ≈ 0.98). - A statistically significant positive effect on average hourly wages (DiD = €0.52/hour, p < 0.001) — reported as ≈ a 4.6% increase in the wage gap favoring highly digitalized activities. - A very large positive effect on remote work incidence (DiD = 40.74 percentage points, p < 0.001); remote work rose from 17.6% → 82.1% in highly digitalized sectors versus 1.3% → 6.6% in less digitalized sectors.

Key Points

• Treatment definition: sectors were binary-classified as highly vs. less digitalized based on telework feasibility (Dingel & Neiman 2020 adapted to NACE Rev.2) and digital intensity.
• Outcomes analyzed: employment level (thousand persons), average hourly wages (€/hour), and share of employees working remotely (%).
• Timing: Pre-crisis = 2018–2019; 2020 treated as the shock year (excluded from main DiD); Post-crisis = 2021–2024.
• Identification: Difference-in-differences with country fixed effects; the interaction (DigitalGroup × Post) yields the ATET.
• Robustness checks: pre-trend analysis and placebo tests were conducted (authors report testing parallel trends and running placebo periods).
• Data: Quarterly panel for 27 EU Member States (EU-27), sourced from Eurostat. Analysis performed in Python (pandas, statsmodels, numpy; visualized with matplotlib/seaborn).
• Descriptives: Highly digitalized sectors showed higher wages and much greater pre/post shares of remote work; less digitalized sectors accounted for larger employment volumes but greater heterogeneity.
• Sector examples (telework potential ranges shown): Information & Communication (70–85%), Financial & Insurance (60–80%), Professional & Scientific (50–70%) classified as highly digitalized; Manufacturing, Construction, Retail, Accommodation & Food Services classified as less digitalized.

Data & Methods

• Sample: 27 EU Member States; quarterly N = 36,685 observations covering 2018–2024.
• Treatment construction: Binary sectoral classification combining telework feasibility and digital intensity (NACE Rev.2 mapping).
• Econometric model (summary): Y_it = α + β D_i + γ Post_t + δ (D_i × Post_t) + θ X_it + ε_it, where δ is the DiD estimator of interest; country fixed effects included.
• Outcomes: employment (thousands), hourly wages (€), remote work share (%).
• Estimation: OLS DiD regressions with country fixed effects. Pre-trend checks and placebo treatments used to probe identification.
• Software: Python (pandas, statsmodels, numpy); figures with matplotlib/seaborn.
• Limitations acknowledged by authors: binary sector classification may mask within-sector heterogeneity; exclusion of the UK to maintain EU-27 consistency; potential confounding from concurrent shocks (energy transitions, geopolitical events) although country fixed effects and pre-trend checks were used to mitigate bias.

Implications for AI Economics

• Reallocation over aggregate job loss: The absence of a measurable aggregate employment effect suggests digitalization (as proxied here) operated more through reallocation and changes in job composition than through net employment destruction during and after the COVID shock. For AI economics, this aligns with scenarios where automation and digital tech shift tasks and occupations rather than producing immediate large net job losses at the sectoral aggregate level.
• Wage dispersion and skill premia: The observed positive wage effect in more digitalized sectors implies a premium for digital-capable workers. AI adoption is likely to amplify demand for advanced digital, cognitive, and complementary skills, increasing wage dispersion unless matched by upskilling and education policy.
• Remote work and productivity/location effects: The huge rise in remote work in digitalized sectors underscores how digital technologies change work organization and spatial labor markets. For AI, increased remote-capable tasks may interact with AI-enabled collaboration and monitoring tools, affecting productivity, labor supply choices, and urban/rural labor market dynamics.
• Distributional concerns and policy priorities: The results highlight inequality risks—workers in less digitalized activities faced limited remote-work opportunities and smaller wage gains. AI-specific policy should prioritize targeted reskilling, lifelong learning, and redistribution mechanisms to manage transition costs.
• Measurement and attribution cautions: The study’s digitalization proxy blends telework feasibility and digital intensity; it does not isolate AI-specific adoption. AI economics research should use more granular measures (firm- or occupation-level AI adoption indicators) to identify AI’s distinct effects from broader digitalization.
• Research directions for AI economists:
  • Use occupation- and firm-level data to trace task-level substitution/complementarity between AI and human labor.
  • Examine heterogeneous effects of AI across worker skill groups, genders, and regions.
  • Study dynamic effects on firm productivity and job reallocation (short vs. medium/long run).
  • Assess policy interventions (training subsidies, portable benefits, taxation) that can mitigate inequality associated with tech-driven reallocation.
  • Investigate interactions between AI adoption and remote/hybrid work models on labor supply, matching, and wages.

Overall, the paper provides causal evidence that sectoral digitalization significantly increased wages and, especially, remote work during the COVID-19 shock while leaving aggregate employment roughly unchanged — findings that are directly relevant to how economists expect AI and advanced digital technologies to reshape wages, work organization, and distributional outcomes.