This paper examines how policy choices affect the fatality rates in the construction sector across states in the US.The fatality data are drawn from the Census of Fatal Occupational Injuries and the years between 1992 and 2014 are studied.In addition to a simple OLS model regressing fatality rate on policy variables as well as a set of control variables,we also examined models without fixed effect,models with fixed effect for either years or states and models with fixed effect for both factors.The results indicate that larger number of employees,higher wages,and more frequent inspections are associated with lower fatality rate.The effect of union protection seems to be greater in larger states.The penalty variable is found to be significantly positive in the regression,which may be an indicator of non-compliance and low credibility of the employers.Extended waiting periods are more associated to non-fatal injuries.The workers compensation effect cannot be clearly specified because of the different severity distribution across states.In addition,regressing the fatality rates only on the control variables,we conducted a comparison between these predicted rates and the real fatality rates.States with significant residuals can become the focus of further research.