Pre-registration: local effects of non-food ultra-hard-discount store openings on independent retailers in France (SIRENE, 2008-2025)

# Pre-registration: local effects of non-food ultra-hard-discount openings on independent retailers (France, SIRENE) Deposited on 26 October 2026, **before any outcome was computed**, as part of the revision of manuscript admsci-4609094 (Administrative Sciences). Author: Jorge Anibal Restrepo-Morales (ORCID 0000-0001-9764-6622). ## Design fixed in advance - Data: INSEE SIRENE StockEtablissement and StockEtablissementHistorique, release of 1 October 2026; INSEE commune membership table 2026 (bassins de vie 2022) and commune correspondence table 2003-2026. - Treatment 1: year of creation of the first Action establishment (SIREN 753308238, NAF division 47, including closed establishments) in the unit; the previous year is treated as possible anticipation. Treatment 2: first opening of any chain listed in `formato_rotulos.txt`. - Main unit: bassin de vie 2022 x year, 2008-2025. Gradient: rings of 0-1, 1-3 and 3-10 km (Lambert coordinates). Municipality as robustness. Commune codes harmonised to COG 2026. - Independent retailer in year t: legal unit (SIREN) with exactly one active establishment in NAF division 47 in that year, excluding the brands in `franquicias_excluidas.txt`. - Outcomes: stock, entries and exits of independent establishments by NAF class (`solapamiento_naf.csv`). - Main specification: triple difference (high-overlap minus placebo, within bassin de vie and year), fixed effects bassin x activity and region x year; control for same-year openings of large grocery stores (47.11D/47.11F) and of other format chains. Estimators: Callaway and Sant'Anna (2021), not-yet-treated controls; Sun and Abraham (2021) as independent re-computation. Inference clustered by departement; permutation placebo of opening dates. - Hypotheses: H1 local decline of high-overlap independents; H2 operating mainly through fewer entries; H3 no effect on placebo activities and decay with distance. Bazaars (47.19B) reported separately. - Minimum detectable effect computed with placebo dates before estimation. - Pre-specified tests that could overturn the thesis: non-zero pre-trends; equal effects for placebo activities; no distance gradient; opening-date mismatch (re-run with +/-1 year window); results excluding Paris and communes above 100,000 inhabitants. ## Files - `solapamiento_naf.csv`: overlap classification of NAF classes (high, low, placebo, separate). - `formato_rotulos.txt`: brands defining the format (treatment 2). - `franquicias_excluidas.txt`: franchise and affiliate networks excluded from "independent". - `PLAN_FASE0_v2_1_DECISIONES.md`: design decisions (Spanish), with the external review that motivated them. - `SHA256SUMS.txt`: hashes of the files above. Version 1.1.0 (26 Oct 2026): Amendment 1 added (ENMIENDA_2026-10-26.md): expected signs for the ring-level flow tests and additional controls, deposited before those outputs were opened. All version 1.0.0 files unchanged.