When Environmental Warnings Create Spillovers: Alert Delivery Design and Influenza Transmission in South Korea

Health Outcomes</p><p><br></p><p dir="ltr">**Provider:** National Health Insurance Service (NHIS), South Korea</p><p><br></p><p dir="ltr">Daily influenza outcomes are constructed from medical-utilization records containing ICD-10 diagnosis codes **J09–J11**. The analysis uses regional information and, where applicable, age, sex, and inpatient/outpatient dimensions. The main outcome is daily influenza incidence per 1,000 residents.</p><p><br></p><p dir="ltr">The paper also uses selected non-influenza health outcomes in placebo analyses.</p><p><br></p><p dir="ltr">### 2.

Air Pollution and PM Warning Records</p><p><br></p><p dir="ltr">**Provider:** AirKorea, Korea Environment Corporation</p><p><br></p><p dir="ltr">AirKorea data are used to construct:</p><p><br></p><p dir="ltr">- PM10 and PM2.5 pollution measures;</p><p dir="ltr">- PM advisory and warning indicators;</p><p dir="ltr">- warning timing and post-warning exposure variables; and</p><p dir="ltr">- supplementary pollution and warning-threshold measures.</p><p><br></p><p dir="ltr">PM10 and PM2.5 warnings are combined into a common warning indicator because both generate the public warning signal studied in the paper.

The July 2018 revision to the PM2.5 warning criterion is retained and examined in robustness analyses.</p><p><br></p><p dir="ltr">### 3. Weather</p><p><br></p><p dir="ltr">**Provider:** Korea Meteorological Administration (KMA)</p><p><br></p><p dir="ltr">The analysis uses daily:</p><p><br></p><p dir="ltr">- temperature;</p><p dir="ltr">- precipitation;</p><p dir="ltr">- wind speed; and</p><p dir="ltr">- wind direction.</p><p><br></p><p dir="ltr">Weather data are harmonized geographically with the regional health and pollution data.</p><p><br></p><p dir="ltr">### 4.

Regional Economic Characteristics</p><p><br></p><p dir="ltr">**Provider:** Korean Statistical Information Service (KOSIS)</p><p><br></p><p dir="ltr">KOSIS data are used for regional economic controls, including **GRDP per capita**. Annual regional values are merged to the daily regional panel.</p><p><br></p><p dir="ltr">### 5. Influenza Vaccination</p><p><br></p><p dir="ltr">**Original provider:** Korea Disease Control and Prevention Agency (KDCA) </p><p dir="ltr">**Survey:** Community Health Survey (지역사회건강조사) </p><p dir="ltr">**Access platform:** Korean Statistical Information Service (KOSIS)</p><p><br></p><p dir="ltr">Regional influenza vaccination rates from the **KDCA Community Health Survey**, accessed through KOSIS, are used as annual regional controls and merged to the daily analysis panel.</p><p><br></p><p dir="ltr">### 6.

Search Behavior</p><p><br></p><p dir="ltr">**Provider:** Naver DataLab</p><p><br></p><p dir="ltr">Daily search indices for 2016–2019 are used as high-frequency proxies for public attention and activity preferences around PM warnings.</p><p><br></p><p dir="ltr">The analysis groups search terms into the following categories:</p><p><br></p><p dir="ltr">- home activities;</p><p dir="ltr">- indoor activities;</p><p dir="ltr">- outdoor activities;</p><p dir="ltr">- pollution-related searches;</p><p dir="ltr">- protective behavior / masks; and</p><p dir="ltr">- air purification.</p><p><br></p><p dir="ltr">Naver DataLab reports normalized search indices rather than raw search counts.

The exact search terms used in each category are documented in the paper and its appendix.</p><p><br></p><p dir="ltr">## Data Construction</p><p><br></p><p dir="ltr">The replication workflow should reproduce the analysis datasets through the following steps:</p><p><br></p><p dir="ltr">1. Clean and aggregate NHIS health outcomes to the jurisdiction-day level.</p><p dir="ltr">2. Construct daily regional pollution, advisory, and warning measures from AirKorea data.</p><p dir="ltr">3.

Merge daily weather controls from the Korea Meteorological Administration.</p><p dir="ltr">4. Merge annual GRDP per capita from KOSIS.</p><p dir="ltr">5. Merge annual regional influenza vaccination rates from the KDCA Community Health Survey accessed through KOSIS.</p><p dir="ltr">6.

Construct post-warning and post-advisory treatment histories.</p><p dir="ltr">7. Construct the Emergency Alert-era treatment interactions and alternative treatment-timing definitions.</p><p dir="ltr">8. Construct age, sex, geographic, timing, and other heterogeneity samples.</p><p dir="ltr">9.

Construct the Naver DataLab daily search series used in the behavioral-mechanism analysis.</p><p dir="ltr">10. Run the analysis code to reproduce the paper's tables and figures.</p><p><br></p><p dir="ltr">## Redistribution and Third-Party Data</p><p><br></p><p dir="ltr">This repository redistributes processed analysis data and source files only to the extent permitted by the terms of the original data providers.</p><p><br></p><p dir="ltr">Where redistribution of original third-party source files is restricted or uncertain, the original files are **not** included.

Instead, this repository documents the source, variables used, and the steps needed to obtain and process the original data.</p><p><br></p><p dir="ltr">Users are responsible for complying with the terms and conditions of the respective data providers when obtaining or reusing source data.</p><p><br></p><p dir="ltr">## Reproducibility</p><p><br></p><p dir="ltr">To reproduce the paper:</p><p><br></p><p dir="ltr">1. Place any source files that cannot be redistributed in `data/raw_not_redistributed/`.</p><p dir="ltr">2.

Place the final processed analysis datasets in `data/processed/`.</p><p dir="ltr">3. Place data-construction and analysis scripts in `code/`.</p><p dir="ltr">4. Update this README with the exact execution order and software/package versions.</p><p dir="ltr">5.

Run the scripts in the documented order.</p><p dir="ltr">6. Replicated tables and figures should be written to `output/tables/` and `output/figures/`.</p><p><br></p><p dir="ltr">## Citation</p><p><br></p><p dir="ltr">If you use this replication package, please cite the associated paper and the Zenodo record.</p><p><br></p><p dir="ltr">**Park, Jonghoon. “When Environmental Warnings Create Spillovers: Alert Delivery Design and Influenza Transmission in South Korea.”**</p><p><br></p><p dir="ltr">After the Zenodo deposit is published, add the final Zenodo DOI here.</p><p><br></p><p dir="ltr">## Contact</p><p><br></p><p dir="ltr">Jonghoon Park </p><p dir="ltr">Department of Economics </p><p dir="ltr">Auburn University</p><p><br></p>