DOES GOVERNMENT AI READINESS IMPROVE GOOD GOVERNANCE? EVIDENCE FROM A 97-COUNTRY PANEL (2021–2024)

Abstract

Amid rapid public-sector AI deployment and tightening governance frameworks, it remains unclear whether greater national AI readiness translates into measurably better government performance. This study aims to investigate the empirical relationship between government AI readiness and good governance across countries, and to distinguish between cross-sectional correlations and within-country effects over time. Using a balanced panel of 97 countries (N = 388) for 2021–2024, we estimate fixed-effects (country and year) and random-effects models for the Chandler Good Government Index (y) on the Government AI Readiness Index (x), apply a Hausman specification test, and verify inference with country/time clustering, CR2 and Driscoll–Kraay standard errors. The random-effects model indicates a strong positive association (β = 0.003859, p < 2.2e−16), but the Hausman test rejects RE consistency (χ² = 774.72, p < 2.2e−16), implying correlation between AI readiness and unobserved country effects. In the preferred two-way fixed-effects model, the coefficient is small and negative (β = −0.000846) and becomes statistically insignificant under robust inference (country-clustered HC1 p = 0.198, CR2 p = 0.228, Driscoll–Kraay **p = 0.203). The evidence suggests that the positive cross-country correlation between AI readiness and governance does not translate into statistically significant improvements within countries over the period 2021–2024.