This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.

The Future of Employment: How Susceptible are Jobs to Computerisation?

Frey, Osborne

2016Technological Forecasting and Social Change7,908 citations

Exposure / measurementInterdisciplinaryTheoretical model

Automation / RobotsMachine Learning (pre-LLM)AI (General)Routine task changeGeneral automation

View Repository DOI: 10.1016/j.techfore.2016.08.019

Summary

Frey and Osborne develop a novel methodology using Gaussian process classification and O*NET data to estimate the probability of computerisation for 702 US occupations, examining how susceptible jobs are to automation based on engineering bottlenecks related to perception/manipulation, creative intelligence, and social intelligence.

Main Finding

Approximately 47 percent of total US employment is at high risk of computerisation, with automation risk exhibiting a strong negative relationship with wages and educational attainment, predicting a truncation of labor market polarization as computerisation primarily affects low-skill, low-wage jobs rather than middle-income routine jobs.

Primary Datasets

O*NET

Other

O*NET

Secondary Datasets

OES/OEWS

Labor outcomes

BLS OES

Key Methods: Gaussian process classification with hand-labeled training data (70 occupations) to predict computerisation probability for 702 occupations based on O*NET task characteristics; identifies engineering bottlenecks to automation
Sample Period: 2010
Geographic Coverage: US
Sample Size: 702 detailed occupations covering 138.44 million jobs (97% of total US employment)
Level of Analysis: Occupation
Occupation Classification: SOC 2010
Industry Classification: N/A
Replication Package: Partial

Notes

702 occupation scores [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper. [Claude classification]: Seminal paper establishing automation exposure measurement methodology. Uses Gaussian process classifier as methodological tool (not studying ML itself). Theoretical model is simple task-based production function distinguishing susceptible vs non-susceptible labor inputs. Hand-labeled 70 occupations for training. Published version of widely-cited 2013 working paper.