Replication package for "Cross-border media visibility and short-run tourist mobility after Spain's 2024 anti-overtourism protests"

Replication package for a study linking foreign-media visibility of Spain's 2024 anti-overtourism mobilisations to subsequent tourist flows from the same origin markets. The analysis covers seven focal mobilisations that took place between April and July 2024 in the Canary Islands, Eivissa, Palma, Granada, Málaga, Cádiz and Barcelona. For each destination the twenty largest foreign origin markets were frozen from pre-event official tourism data, giving 140 destination-origin dyads across 34 unique origin markets. Exposure is a binary indicator of qualifying coverage by publishers based in each origin market during the window from one day before the mobilisation to seven days afterwards. Outcomes are monthly destination-origin tourist counts from the experimental mobile-phone tourism statistics published by Spain's National Statistics Institute. The package contains the frozen analysis panel, the destination-origin exposure grid, media-measurement validation and retrieval audit tables, the official source snapshots from which outcomes were regenerated, statistical outputs, analysis code, the pre-outcome analysis protocol, an SQLite database, a data dictionary and file-level SHA-256 checksums for all 77 files. Reproduction is exact from the frozen exposure grid and the recovered official source files onward. Two earlier transient objects were not preserved and are not reconstructed: the complete raw-return and candidate archives generated during historical news discovery, and the exposure grid of a superseded content rule. The media-search stage therefore cannot be re-run byte for byte from this package. No replacement records were generated, and this boundary is documented in the package. This version applies the standard finite-sample degrees-of-freedom convention to cluster-robust variances, counting parameters absorbed by the fixed effects in the residual degrees of freedom. Point estimates, the exact destination-level sign-flip diagnostics and the influence decomposition are unaffected by that correction; standard errors, p-values, confidence intervals and joint pre-event test statistics change. Outputs under the superseded convention are retained so that the correction is auditable. To avoid redistributing third-party journalism, the public release omits article body text and archived snippets and removes article titles from the audit tables, retaining derived classifications, publisher domains, dates, cryptographic hashes and, where necessary for auditability, article URLs.