This dissertation aims to study, by means of economic framework, the optimal numbers of surveillance points and the rate of fines on drunk drivers of which could decrease social loss from "drunk driving" behavior. We consider the model, which assumes that drunk drivers response to the setting of surveillance points and the level of fine rate. The police finance the expenses for setting the surveillance points from the fines and/or subsidy from the government. We collected data and parameters from various sources to investigate and provide implications for policy makers to reduce the accidents throughout our simulation analysis. Our findings show that; Firstly, drunk driver decision to consume alcohol is based on the cost of drinking which is comprised of probability charged and fines. Secondly, the surveillance points reduce the accidental losses more than fines do but the cost of the surveillance points is higher. Thirdly, nowadays the number of surveillance points and fine rate are too low actually so that surveillance points in Bangkok could be 11 points per week as well as the optimal fine rate is Baht 22,500. Finally, police can set more surveillance points from government subsidy. By increasing fine rate allocation to police government spend the lowest expenditure, compare with lump-sum subsidy and proportion in police expenditure In addition we make questionnaire with the drivers who were arrested in "drunk driving" behavior to confirm the fact that the increase of surveillance points and fine rate could reduce drunk driving behavior statistically.