This study examines the relationship between national AI capability and labor productivity in low- and middle-income countries (LMICs) through the lens of absorptive capacity theory. Rather than focusing on frontier AI innovation, we conceptualize AI capability as a transformation-stage construct emerging from the interaction between AI-embodied external inflows (e.g., AI-related hardware imports) and domestic complementary conditions (e.g., internet penetration, broadband access, electricity access, and financial inclusion). This framing reflects how most LMICs engage with AI—through embodied technology imports combined with enabling structural conditions. Using a panel dataset of 63 LMICs over the period 2005–2019, we construct a novel AI Capability Index via Principal Component Analysis (PCA), and estimate its association with labor productivity using two-way fixed effects and system GMM estimators. We complement this analysis with interaction-based specifications that separate AI-related inflows and domestic readiness conditions and examine how their interaction relates to labor productivity. Our results reveal a robust and statistically significant positive association: a one standard deviation increase in AI capability is associated with approximately a 9.20–9.37% increase in labor productivity, depending on index construction (flow-based vs. stock-based). Results remain stable across alternative specifications and sensitivity analyses. These findings indicate that productivity gains in LMICs depend not solely on access to AI-related technologies, but also on the complementary institutional and infrastructural conditions that enable their transformation and diffusion. By operationalizing AI capability as a national-level transformation-stage construct, this study extends absorptive capacity theory beyond the firm level and provides a replicable empirical framework for assessing national AI capability formation and its relationship to labor productivity under data-constrained conditions. In doing so, it contributes to ongoing debates on digital transformation, technological adoption, and capability formation in LMICs.
Khan et al. (Mon,) studied this question.