The role of infrastructure in economic development is always a concern for any government. Despite the growing literature devoted to the correlation between infrastructure and productivity, there is still little plant-level evidence via their causal relationship by employing a natural experiment design. Consequently, in this paper we utilize plant-level data between 1998 and 2003 and employ several econometric models to investigate the causal effects of infrastructure on plant productivity, using the 921 earthquake in Taiwan as a natural experiment. Our results show that a 1% increase in a previous year’s infrastructure expenditure in each city (measured by the expenditure of transportation and other economic services) raises the total factor productivity of each plant located in that city by 0.06% during the present year. Furthermore, this impact is larger for plants with higher total factor productivity compared to those with lower total factor productivity. Our results are expected to have important implications for the benefit measurement of public infrastructure investment policy.