This record contains the cleaned and cross-platform replication materials for the manuscript “Comprehensive innovation reform pilot zones and provincial green total factor productivity in China”, submitted to Discover Environment. The empirical analysis uses a balanced panel of 30 mainland Chinese provincial-level units observed annually from 2011 to 2021, comprising 330 province-year observations. The package contains the processed analysis panel, separately defined directional distance function (DDF) measures, data documentation, Python analysis code, optional R and Stata diagnostic scripts, code-generated tables and figures, execution logs, metadata and a SHA-256 file manifest. Running python run_all.py from the package root validates and prepares the data, reconstructs the policy-timing variables, generates Tables 1–11 and Tables S1–S4, performs the event-study and manual group-time ATT analyses, regenerates Figures 1–4, and completes output-consistency checks. The manual group-time ATT inference uses 9,999 stratified province-level bootstrap resamples with the fixed random seed 20260801. The optional R did::att_gt() and Stata csdid scripts are retained as diagnostic modules and are not required for the core Python reproduction. This version corrects the treatment-timing documentation by distinguishing the operational year from the binary policy_actual variable. It also reports the global DDF efficiency score, yearly-frontier DDF efficiency score and normalised yearly DDF change index as separate outcomes. Event time −4 in the main event study pools observations at event time −4 or earlier. Figures 1–4 are supplied as EPS vector graphics with embedded Arial typography and as 6600 × 4400 pixel, 1200 dpi PNG files. Regenerating the figures requires a licensed Arial Regular font installed on the operating system or supplied through the CIRPZ_ARIAL_REGULAR environment variable. The Arial font file is not redistributed. The archive contains processed research variables rather than scans or reproductions of third-party statistical yearbooks. The MIT licence applies to the code only and does not transfer rights in underlying third-party source publications. Detailed variable definitions, sources, transformations, quality flags and redistribution notes are provided in data_documentation/variable_dictionary.xlsx. Version 2.0 supersedes the earlier replication archive associated with this record.