denizo254/kenya-digital-divide

# Kenya's digital divide is not a hardware problem Data analysis of the 2019 Kenya Population and Housing Census showing that the gender gap in **mobile phone ownership** (0.6 percentage points) is dwarfed by the gender gap in **internet use** (5.0 points) — and that in 19 of Kenya's 47 counties, women own more phones than men and still use the internet less. **→ Read the article: `article.md`** ## Findings | | Men | Women | Gap | |---|---|---|---| | Owns a mobile phone | 47.6% | 47.0% | **+0.6 pts** | | Used the internet | 25.1% | 20.1% | **+5.0 pts** | - The internet gender gap is **8.3×** the phone-ownership gap. - In **19 of 47 counties** women out-own men on phones; in **all 19**, men still lead on internet use. - County internet use ranges **7.6×**, from Nairobi City (52.4%) to Turkana (6.9%). - Only 11 counties sit above the national rate of 22.6%; **~64%** of Kenyans live in a county below it. - Phone ownership and internet use correlate at **r = 0.87**, but conversion from one to the other varies nearly threefold across counties (Nairobi 76%, Tana River 28%). ## Reproducing this ```bash pip install pandas requests matplotlib python src/01_download.py # fetch source CSVs, record SHA-256 checksums python src/02_analyse.py # clean, merge, compute every quoted figure python src/03_charts.py # render the three charts python src/04_verify_claims.py # assert all 31 article claims against the data ``` `04_verify_claims.py` exits non-zero if any figure in the article drifts from the source data, so the prose and the dataset cannot silently diverge. ## The parsing problem worth knowing about The census tables put two levels of geography in one column: a county header row followed by its sub-county rows. Names repeat across levels — **Kiambu is both a county and a sub-county within Kiambu** — so filtering by name silently double-counts, and Nairobi is missed entirely because it is listed as "Nairobi City". `src/parse.py` identifies rows …