It is a well-documented phenomenon that coal combustion is a significant cause of air pollution, while few documents have revealed the process. Based on the panel dataset covering 30 Chinese provinces (excluding Tibet) in China during the period 2000–2012, this study investigates the actual contribution of centralized and scattered coal consumption to PM2.5 concentrations through a panel cointegration model and panel vector error-correction model (PVECM). The empirical results indicate that both scattered and centralized coal consumption have significant positive correlations with PM2.5 concentrations in the long term. Specifically, the PM2.5 pollution caused by scattered coal combustion is approximately 1.38 times that of the same amount of centralized coal combustion. Furthermore, in the short-term, centralized coal consumption has a significant positive relationship with PM2.5 pollution, while scattered coal has no significant effect. Therefore, in the short term, the consumption of centralized coal should be reduced, whereas in the long term, scattered coal control policies should be the primary focus of the government.