The occasional ＂strike hard＂ campaigns against crime launched by the Chinese government provide an opportunity to isolate the separate effects of severity and certainty of punishment on the crime rate. The ＂strike hard＂ campaigns increase the severity of the punishment hut keep the certainty of the punishment unchanged. We use provincial panel data from 1988 to 2015 to examine the impacts of the two strategies on the crime rate with pooled mean group models. The empirical results show that a significant decrease in crime rates is associated with greater certainty of detection, hut greater severity has no significant effect. A 1% increase in the detection rate (a measurement of certainty) predicts about 2.7% lower crime rate. The results are robust even after considering the endogenous nature of punishment policies and controlling for the measurement error in the officially reported data.