Abstract: The economic theory basis and research progress of evaluating the value of a statistical life (VSL) with contingent valuation method (CVM) are reviewed respectively. Some biases, such as hypothetical bias, strategic bias and information bias, which may be brought into the evaluation of VSL with CVM, are analyzed. Willingness to pay and willingness to accept, question formulation and risk variability, which may affect the validity of the VSL results, are discussed. According to the results of literature analysis, and combining the previous empirical study of using CVM to evaluate the VSL of employees, the scheme about how to improve the validity and reliability of VSL with CVM is designed. Through optimized design on the overall framework, working environment risk perception survey, design of the working risk of willingness to pay, socio-economic information variable settings, readiness survey and pre-survey, sample size and sample design, data statistics and analysis, etc., the assessment result can be ensured.