Econometric Assumptions and Critiques in the Acemoglu Frameworks
How do the stylized formulas in this simulator represent the econometric critiques of settler mortality instruments and the macroeconomic bounds on AI productivity?
This interactive workbench demonstrates three central debates in Daron Acemoglu's research: instrumental variable (IV) identification using colonial settler mortality, task automation versus labor reinstatement, and macroeconomic bounds on artificial intelligence. While empirical papers run structural two-stage least squares (2SLS) or calibrate production functions against microdata, this browser tool evaluates simplified closed-form algebraic approximations. It illustrates how introducing direct disease channels (violating the exclusion restriction) or accounting for task complexity ceilings moderates projected causal effects and growth estimates.
The simulator uses stylized linear heuristics rather than estimating 2SLS regressions on colonial historical records or solving full general equilibrium models. Confidence intervals in the mirror table are fixed at ±0.26 rather than computed from empirical variance-covariance matrices. Results demonstrate theoretical trade-offs rather than production econometric forecasts.
Try a worked example
Switch to module 1 by clicking '1. AJR (2001) Settler IV'. Move the 'Exclusion Violation (γ Direct Disease Effect)' slider from 0.12 up to 0.30. In the simulation code, the estimated 2SLS coefficient is computed as baseline_beta - (γ × 1.33) - (noise × 0.20). With the 'Exclude Neo-Europes' baseline (0.84) and noise at 0.15, β̂ drops from 0.65 to 0.41, the IV vs OLS gap contracts to 0.00, and the resilience badge updates to warn of high fragility to disease confounders.
Settler Mortality IVs and Albouy's Measurement Critique
Acemoglu, Johnson, and Robinson (2001) proposed European settler mortality as an instrument for modern institutional quality to address endogeneity and reverse causality. In a major critique, Albouy (2012) examined the underlying historical mortality sources, showing that 36 of the 64 countries were assigned mortality rates from other territories or based on incomparable military campaign records. When these measurement issues are controlled for, first-stage instrument strength weakens and standard errors expand significantly. The Colonial Origins of Comparative Development: An Empirical Investigation: Comment
In the simulator, the 'Settler Mortality Measurement Noise' (σ_e) and 'Exclusion Violation' (γ) sliders illustrate these critique vectors by directly penalizing the estimated institutional coefficient, reflecting how non-zero direct disease effects on modern productivity undermine instrument validity. The Colonial Origins of Comparative Development: An Empirical Investigation: Comment
Macroeconomic AI Productivity Bounds and Task Complexity
In evaluating whether generative AI will spark explosive macroeconomic growth, Acemoglu (2024 NBER Working Paper 32487) uses a task-based framework where aggregate productivity gains follow Hulten's theorem. Under this formulation, aggregate TFP gains depend directly on the fraction of tasks affected and average task-level cost savings. Acemoglu estimates that easily automated tasks account for a modest fraction of labor and that 10-year TFP gains are bounded below 0.66%, falling to under 0.53% when hard-to-learn tasks requiring context-dependent judgment are accounted for. The Simple Macroeconomics of AI
The simulator's Module 3 captures this logic by scaling effective task substitution by (1 - 0.5 × penalty) multiplied by task cost savings π, reproducing the modest 10-year GDP growth ceiling in contrast to unconstrained takeoff projections. The Simple Macroeconomics of AI
Sources and further reading