The infrastructure investment effects on GDP growth have been of interest to economists and policy makers in the past 20 years. Around thirty studies have investigated the effect of infrastructure investment on the growth of GDP, resulting in 346 point estimates of the effect. Meta-regression analysis is used to combine these disparate estimates. On average, infrastructure investment exerts positive effects on GDP growth. However, this result is conditional upon the explanatory variables in the different models, which are sources of pronounced heterogeneity in this literature. The main findings are that: a) in this literature there is evidence of Type I publication bias (directional, in this case positive), b) there is also evidence of Type II publication bias (favouring statistical significance regardless of direction) and c) the two types of publication bias are combined with an absence of authentic empirical effect in this literature. For the practice of MRA we make a case for checking the robustness of the results with respect to the estimation technique, model specification and sample.