Abstract: Parisian option is a complex path-dependent option extended from the barrier options, which allows the holder buy or sell a certain underlying asset at a pre-specified price under the condition that underlying asset price above or below a given level of a continuous or cumulative occupation time before maturity. The numerical methods for pricing Parisian option include binomial tree method, finite difference method and Monte Carlo method. Compared with other numerical methods, Monte Carlo method is more flexible and easy to implement and improve; moreover, its estimation error and convergence speed has stronger independence with the dimensions of the problem to be solved, and thus can solve the target variable of high-dimensional derivative securities pricing better.In this paper the Parisian option is priced using the Monte Carlo method, and improves the standard Monte Carlo algorithm is improved to multi-level Monte Carlo algorithm. Our research results show that under the given accuracy, multi-level Monte Carlo algorithm can reduce the calculation costs from O(ε^-3) to O(ε^-2(logε)²) comparing with the standard Monte Carlo method. On the other hand, under given calculation cost, multi-level Monte Carlo method can converge to the true value faster comparing with standard Monte Carlo method. Applying this method to Parisian option pricing not only expanses the choice scope of Parisian options' numerical algorithms, but also improves the precision of Parisian option pricing, and lays a certain foundation for Parisian options' application in the domestic market.

Key words: Parisian option, standard Monte Carlo method, multi-level Monte Carlo method, computation cost