Stated preference analyses seeking to determine the public's value for air quality improvements often estimate willingness to pay (WTP) for days at a specified minimum quality threshold (e.g., days with clean air), but do not consider the temporal distribution of pollution levels below this threshold. This paper develops a choice experiment designed to evaluate WTP for a more complete distribution of air quality improvements, including the number of days per year at multiple air quality levels. The model is applied to a case study of air quality improvement in the core districts of Xi'an City, China. Results from a linearly constrained mixed logit model demonstrate that average household WTP for improving a lightly polluted, moderately polluted, heavily polluted, or severely polluted day to a clean air day is 7.42, 8.90, 13.06, and 24.28 RMB per year, respectively. These results show that WTP depends not only on the total number of clean air days, but on the total distribution of pollution levels across all days of the year. Results are directly relevant to the development of clean air policies in China, for which benefit estimates are currently unavailable.