Replication Package for "Central Signal Premium and Purity-Amplified Reactions to Emerging Industry Initiatives: Evidence from China's Low-Altitude Economy"

This archive is the complete replication package for the manuscript "Central Signal Premium and Purity-Amplified Reactions to Emerging Industry Initiatives: Evidence from China's Low-Altitude Economy", under double-anonymized peer review at a peer-reviewed economics journal. Author identity and affiliation are intentionally withheld from this description during the review stage; both will be disclosed in this metadata record upon formal acceptance. Contents. • The hand-collected Low-Altitude-Economy Policy Event Database (LAE-PED v1.0), containing 47 policy events spanning January 2022 to September 2025, organised into 39 same-trading-day event clusters (33 clean / 14 confounded); • a firm panel of 86 LAE-related A-share listings yielding 3,264 firm-event observations, together with the continuous business-purity score (with tercile and quartile cuts), state-ownership classification, and size/exchange controls; • Python 3.11 code for: market-model event-study estimation, cross-sectional regression with cluster-robust standard errors at the trading-date cluster, generation of all five publication figures, and a setup/integrity-check stage that verifies the exact row counts of every shipped data file; • R 4.3 code for the pre-registered robustness battery: wild-cluster bootstrap with 9,999 Rademacher replications (Roodman et al. 2019), Oster (2019) coefficient-stability bound via the robomit package, and a 200-iteration random-tier placebo; • all main-text tables shipped as CSV, and all five figures (CAAR by tier, interactions, event-time profile, benchmark sensitivity, placebo distribution) shipped as publication-quality PNG; • a pre-registration document dated 15 April 2024 that fixed the three hypotheses (Central-Signal Premium, Purity Amplification, SOE Attenuation), the cross-sectional regression specification, and the primary CAR [0,+1] event window prior to the construction of the firm-event panel; • an amendment ledger logging any post-registration deviations (none to date); • a Dockerfile pinning the full toolchain (Python 3.11.7-slim-bookworm plus R 4.3 with an MRAN snapshot dated 2024-01-15) and a `run_all.sh` script that orchestrates the five-stage pipeline end-to-end from a single command, with a `--skip-r` flag for a Python-only quick check; • a methodological appendix, a stand-alone purity-score methodology document, a data-source pointer file describing a four-tier reproducibility scheme, and a SHA-256 manifest covering every shipped file (44 entries). Running `bash run_all.sh` inside the container rebuilds the entire output/ directory from the shipped inputs in approximately five minutes for the Python stages plus approximately eight minutes for the R robustness battery on a standard laptop, reproducing the numerical results reported in the paper to the precision displayed. Data sources. Raw daily share-price panels can be obtained from either (i) the open-source akshare Python interface, which aggregates publicly reported daily prices from the Shanghai, Shenzhen, and Beijing Stock Exchanges, or (ii) the CSMAR academic-subscription database. The derived firm-event CAR panel used for estimation is bundled in this archive; this is an irreversibly-transformed output that does not re-disclose CSMAR raw records, so that every reported numerical result can be reproduced with or without an independent CSMAR subscription. The patent and capex inputs underlying the purity score require CSMAR Cash-Flow Notes and a CNIPA bulk patent download; see `data/raw_pointers.md` for the full source map. Licence and access. Access is restricted during peer review via a private access token issued exclusively to the editorial office of the journal handling the manuscript. Upon formal acceptance, this archive will be made publicly available under the Creative Commons Attribution 4.0 International (CC-BY-4.0) licence within ten working days, and the corresponding metadata record (including author and affiliation) will be updated accordingly.