Why AI Economic Forecasts Vary Wildly: It Comes Down to a Few Key Parameters

Economists studying AI's impact on growth reach vastly different conclusions — from modest productivity gains to explosive acceleration — not because they use different data, but because they assign different values to a handful of critical parameters. Frameworks developed by researchers like Acemoglu, Restrepo, and Jones model how automation shifts tasks from labour to capital, and how AI might substitute for researchers themselves, potentially loosening the constraints that keep growth steady. The single most consequential variable is the elasticity of substitution, which determines how easily capital replaces labour across tasks and largely separates conservative forecasts from dramatic ones. Other key factors include the actual share of tasks that can be automated, the cost savings realised per task, and how quickly AI capabilities translate into widespread economic deployment. Experts stress that large aggregate effects require multiple parameters to align simultaneously, and that historical diffusion of general-purpose technologies has consistently been slower than anticipated.