Labor market impacts of AI: A new measure and early evidence
Figure 1: Share of Claude usage by Eloundou et al. task exposure ratingThis figure shows Claude usage distributed across O*NET tasks grouped by their theoretical AI exposure. Tasks rated β=1 (fully feasible for an LLM alone) account for 68% of observed Claude usage, while tasks rated β=0 (not feasible) account for just 3%. Data on Claude usage comes from the previous four Economic Index reports.
Figure 2: Theoretical capability and observed exposure by occupational categoryShare of job tasks that LLMs could theoretically perform (blue area) and our own job coverage measure derived from usage data (red area).
Figure 3: Most exposed occupationsTop ten most exposed occupations using our task coverage measure.
Figure 4: BLS projected employment growth from 2024—2034 vs. observed exposureBinned scatterplot with 25 equally-sized bins. Each solid dot shows the average observed exposure and projected employment change for one of the bins. The dashed line shows a simple linear regression fit, weighted by current employment levels. The small diamonds mark individual example occupations for illustration.
Figure 5: Differences between high and low exposure workers, Current Population Survey
Figure 6: Trends in the unemployment rate for workers in the top quartile of observed exposure and no AI exposure, Current Population SurveyThe top panel shows the unemployment rate for workers in the top quartile of exposure (red line) and the 30% of workers with zero exposure. The bottom panel measures the gap between these two series in a difference-in-differences framework.
Figure 7: New job starts among workers age 22-25 in occupations with high observed exposure and no AI exposure, Current Population SurveyThe top panel shows the percent of young workers starting new jobs in high vs. no exposure occupations. The bottom panel measures the gap between these two series in a difference-in-differences framework.
