Preregistration: Learning or incumbency? Persistence in competitive R&D grant allocation to firms

Using open application-level data that include rejections, we compare approval within the same competition cell for three groups: first-time applicants, firms previously only rejected, and firms previously funded. We are not aware of earlier work that makes this three-way comparison for firms. A dynamic random-effects model then asks whether an advantage of previously funded over previously rejected firms survives conditioning on unobserved firm heterogeneity; whether the design separates learning from incumbency depends on that model. We also provide a descriptive counterpart, in a second country, to the reapplication analysis of Bignone et al. (2026). All data are public and the pipeline is reproducible. The focal outcomes (approval or rejection of the analysed applications) are sealed in a separate file that has not been opened since creation; its SHA-256 is 91a6db6b009b66335ddf2d4efdf18e1a75c193aa91e63eaa14dda8bf9a110b83.