Abstract: Based on Lévy stable distributions, this paper develops a new statistical approach for residence-floor live-load design. The approach involves fours steps: firstly using Lévy stable distributions to fit the cumulative distribution function of live loads, secondly employing Lévy random numbers to determine the maximum distribution of live loads, thirdly using the maximum distributions under a certain quantile to calculate the extreme values of sustained live loads and transient live loads, finally utilizing a weighted method to determine the characteristic value of live loads on a residence floor in a design reference period. Our analysis results show that this new approach does not require a large data set of survey live loads and also remedies the drawback in the empirical Turkstra load combination principle, which is used to determine the characteristic value. Compared with Type I extreme value and Weibull distributions, Lévy stable distributions have the best precision and its parameters can directly depict the skewness and heavy tail of live load distribution. In addition, the simulation method of Lévy random numbers is mathematically simple and easy-to-use for non-expert engineers.