Replication archive for "Do Firms Report More University Knowledge Sourcing When Informed About Engagement Channels? Evidence from a Field Experiment"

This archive contains the Stata code and an experimental-design crosswalk required to reproduce the tables reported in: Angenendt, D. and Salter, A. (2026). "Do Firms Report More University Knowledge Sourcing When Informed About Engagement Channels? Evidence from a Field Experiment." Research Policy. The paper embeds a question-order experiment in a multi-wave survey of UK firms (administered 2020–2021). It finds that exposing firms to a comprehensive list of university engagement channels before asking about knowledge sourcing increases the reported probability of citing universities as a knowledge source by about 12 percentage points, a 34% relative increase over the control group. Contents. 00_master_public.do — main Stata do-file that installs dependencies, runs the analysis, and writes all output tables. 01_main_analysis_public.do — the analysis do-file. _helpers.do — shared subroutines. data/experimental_crosswalk.csv — the author-supplied crosswalk keyed on record_id. Contains only experimental-design and response-timing variables (treatment assignment, wave labels, early-responder indicators, response-time bucket). No firm-level size, age, or financial information. README_public.docx — full documentation. reconciliation.md — cell-by-cell comparison of the archive's output to the published tables. LICENSE.txt — MIT licence for the code. External dependency. The archive uses the CBR Survey of Knowledge Exchange Activity with Universities by UK Companies, 2017–2021, deposited separately at the UK Data Service (Study Number 9037; doi:10.5255/UKDA-SN-9037-1). Users register once with the UK Data Service, download the Stata-13 file, and place it in code_public/data/. The UKDA End User Licence does not permit us to redistribute the data file. Scope. The archive reproduces 14 of the 18 tables in the paper, either exactly (for specifications using only the experimental treatment) or up to a small covariate substitution for specifications that originally used three variables that are absent from the UKDA public release: continuous employment, start-up age, or two-year asset growth. Four tables (B.1, C.1, C.5, C.6) require the full firm-level sampling frame of ~337,000 firms or response-propensity data on non-responders, and cannot be reproduced from public data alone. Details are in README_public.docx. Software. Stata 17 or later. The master script auto-installs the required user-contributed packages (estout, psmatch2, ebalance, psweight, balancetable, stddiff, tost) on first run. Contact. David Angenendt ([e-mail]), TUM School of Management and Centre for Business Research, University of Cambridge. Ammon Salter ([e-mail]), Warwick Business School.