Country-week panel handoff

Output file: country_week_panel_2020_2022.csv

Sources

• oxcgrt_compact_national_v1.csv: national OxCGRT policy data.
• owid_compact.csv: OWID COVID outcome data.

Columns

• country: OWID country name.
• week: Monday start date for a complete Monday-Sunday week.
• H1: weekly mean of OxCGRT H1_Public information campaigns.
• H2: weekly mean of OxCGRT H2_Testing policy.
• H7: weekly mean of OxCGRT H7_Vaccination policy.
• outcome: weekly sum of OWID new_cases_per_million, interpreted as new confirmed COVID-19 cases per million people during the week.

Choices made

• Used 2020-01-01 through 2022-12-31 as the source date window, then kept only complete Monday-Sunday weeks. The resulting panel runs from week 2020-01-06 through week 2022-12-19.
• Selected 15 countries by best overlapping data coverage across the two files: full daily OxCGRT national coverage, no missing H1/H2/H7 values, full daily OWID records, and no missing outcome values after filling leading pre-reporting OWID case blanks as zero. Among complete-data countries, population was used as the tie-breaker.
• Selected countries: Bangladesh, China, Democratic Republic of Congo, Egypt, Ethiopia, Indonesia, Iran, Japan, Mexico, Nigeria, Pakistan, Philippines, Russia, United States, Vietnam.
• Averaged daily H1/H2/H7 values within each week so mid-week policy changes are represented as fractional weekly values.
• Excluded partial edge weeks rather than scaling them, so weekly outcomes are comparable across the panel.

Validation

• Balanced panel: 15 countries x 155 weeks = 2,325 rows.
• Output columns are exactly country, week, H1, H2, H7, outcome.
• No missing values in the final panel.

The build is reproducible with build_country_week_panel.py.