kimenjuivy/kenya-development-index

# Kenya County Development and Opportunity Index (KCDOI) Which Kenyan counties are falling behind in development, along which dimensions, and by how much? This project builds a county-level socioeconomic development index for all 47 counties, starting from raw KNBS government publications rather than a pre-cleaned dataset. ## The core finding A substantial development gap exists between ASAL and non-ASAL counties. Non-ASAL counties score **0.84 points higher** on average on the composite Development Score than ASAL counties — a large, statistically significant effect (**Cohen's d = 1.13, p < 0.001**), confirmed by both a parametric t-test and a non-parametric Mann-Whitney U test, so the result doesn't depend on a normality assumption that might not hold at n=47. That gap isn't just visible in the composite score — it shows up as a consistent gradient across every independent domain measured: | Development tier | Counties | Avg poverty | Avg electricity access | Avg GDP per capita | |---|---|---|---|---| | High Development | 4 | 22.4% | 86.5% | KES 506,058 | | Emerging | 16 | 35.0% | 50.0% | KES 275,750 | | Lagging | 18 | 45.0% | 27.6% | KES 187,000 | | Critical Need | 9 | 67.1% | 22.4% | KES 134,574 | ## Exploratory analysis Before building anything, the correlation structure across domains needed checking — nine independently-sourced datasets merging into a coherent story isn't guaranteed, it has to be verified. Poverty, education, electricity, financial inclusion, internet access, and GDP per capita all move together in the expected direction — the strongest pairwise correlations (education-electricity, r=0.82; mains electricity-internet, r=0.80) survive Benjamini-Hochberg FDR correction across all 45 pairwise tests, with 73% of pairs remaining significant even after that correction. One variable stood apart: unemployment rate correlated with almost nothing else in the cluster, a real finding rather than noise, since "seeking work" measures something str …