This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

How Adaptable Are American Workers to AI-Induced Job Displacement?

Manning, Aguirre

2026NBER Working Paper Series
Exposure / measurement
LLM / Generative AIAugmentation vs. substitutionJunior / entry-levelSenior / older workersGender
DOI: 10.3386/w34705
Summary

Manning and Aguirre construct an occupation-level adaptive capacity index combining net liquid wealth, skill transferability, geographic density, and age across 356 U.S. occupations to study which workers are best positioned to navigate job transitions if AI exposure leads to displacement.

Main Finding

The authors find a positive correlation (r=0.502) between AI exposure and adaptive capacity: of 37.1 million workers in top-quartile AI exposure, 26.5 million have above-median adaptive capacity, but 6.1 million workers (4.2% of workforce) face both high exposure and low adaptive capacity, concentrated in clerical and administrative roles.

Primary Datasets

O*NET

Other

SIPP

Labor outcomes

OES/OEWS

Labor outcomes

O*NET Skills and Work Activities; SIPP (Survey of Income and Program Participation); BLS OEWS

Secondary Datasets

ACS

Labor outcomes

CPS

Labor outcomes

ACS (American Community Survey); CPS (Current Population Survey)

Key Methods
Growth-weighted skill transferability using O*NET cosine similarity weighted by BLS employment projections; KNN imputation for missing O*NET data; employment-weighted quartile analysis; composite index construction via Z-score normalization and percentile transformation
Sample Period
2018-2023
Geographic Coverage
US
Sample Size
356 occupations covering 147.9 million workers (95.9% of U.S. workforce); SIPP requires minimum 15 observations per occupation pooled across 3 years
Level of Analysis
Occupation, Region
Occupation Classification
Modified SIPP (MSIPP)
Industry Classification
None
Replication Package
Yes
Notes
NBER Working Paper 34705. Constructs an occupation-level adaptive capacity index covering 356 occupations (95.9% of US workforce). Finds 6.1 million workers in high-exposure, low-adaptive-capacity occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations. [Claude classification]: Uses KNN imputation with k=10 for approximately 15% of occupations missing O*NET skill data. Modified SIPP classification system created to harmonize across data sources. Extensive robustness checks across 57 alternative specifications. Bootstrap CI for correlation: [0.353, 0.624]. High-vulnerability occupations 81.3% female versus 48.0% in other occupations.