New Method Estimates Time Spent on Nearly 18,000 U.S. Job Tasks

A recent study published on arXiv introduces a systematic approach to estimating time shares for nearly 18,000 tasks that encompass almost all jobs in the U.S. This research fills a significant void in the task-based economic framework, which views occupations as collections of tasks and serves as the primary method for analyzing the impact of technology on labor. Previous studies often depended on arbitrary or poorly supported choices for task weights, with some recent proposals suggesting that task weights should reflect time spent. However, existing time shares were either derived from broad ONET data not designed for this analysis or estimated using opaque language models. The proposed method incorporates a task's time based on its expected frequency from ONET and the duration required to complete a single task instance, solving a constraint to estimate the latter. The paper, titled 'Estimating time spent on work tasks,' can be found on arXiv under identifier 2608.05172.