Replication code for "When software evaluates: Performance evaluation and the AI wage premium in the UK, with comparative evidence from the EU"
Abstract
Python code and frozen result files reproducing all tables, figures and supplementary tables of the article "When software evaluates: Performance evaluation and the AI wage premium in the UK, with comparative evidence from the EU" (submitted to PLOS ONE). The analysis uses the UK Skills and Employment Survey 2024 (UK Data Service SN 9542), the EU AIM-WORK survey (European Commission Joint Research Centre) and the European Working Conditions Survey 2024 (UK Data Service SN 9511), together with country-level ICTWSS v2.0 indicators (OECD/AIAS). Estimates include weighted wage regressions with occupation fixed effects, wild-cluster bootstrap inference (1,999 replications), Gelbach decompositions, Oster bounds and interval regression on banded earnings. The survey microdata are licensed third-party data and are NOT included. README.md explains how to obtain each dataset, where to place the files, the run order, and which script produces each table and figure. A full rerun against the frozen results matched 2,594 of 2,604 values; the remaining 10 differ only at the sixth decimal place.
