A series of earlier studies established, on World Bank appraisals and on named West African projects, that the rate of return of a public investment suffers from an optimism bias, and derived from it a reference-class correction that de-biases the announced return. The present article places this corrected efficiency at the core of a dynamic stochastic general equilibrium model, calibrated on a West African Economic and Monetary Union economy, Mali. Public capital raises the productivity of private factors, is built with a lag, and is acquired only to the extent of its efficiency, a parameter that the reference-class correction precisely allows to estimate. The model is tested over two horizons. In the long run, a permanent infrastructure programme of two points of gross domestic product raises steady-state output by 12.3 percent under announced efficiency but by only 7.8 percent under corrected efficiency, and a low efficiency additionally raises the tax burden required to stabilise the debt. In the medium run, the impact multiplier is near zero, a consequence of the construction lag, the output response peaks around the thirteenth year, and the cumulative multiplier is markedly higher under announced efficiency. The lesson for the practice of infrastructure advisory is that the quality of project appraisal and execution, that is, efficiency, governs the macroeconomic return and the sustainability of a programme more than its volume, and that the lever is structural and quantifiable.