As highly related to the investors' earnings expectations and trading decision-making behavior, securities transaction tax (STT) has long been regarded as a typical regulatory mechanism exploited by policy makers. However, neither theoretical analysis nor empirical studies reach consensus about the role and policy effect of the securities transaction tax. Within the framework of agent-based computational finance, this paper presents a new artificial stock market model with heterogeneous agents, which allows us to assess the impacts of varying STTs on market behavior to come to robust conclusions. First we investigate the dynamics of benchmark market with no tax levied, and then market behaviors with different STTs are thoroughly checked. The results show that a modest transactions tax does contribute to stabilize markets by reducing market volatility, but its negative effects on market efficiency cannot be ignored at the same time. The findings suggest that regulatory authorities should introduce STT discreetly to strike a balance between stability and efficiency.