Air pollution from PM 2. 5 affects many cities worldwide, causing both health impacts and mood depression. One of the obstacles to implementing environmental regulations for PM 2. 5 reduction is that there are limited studies of PM 2. 5 welfare loss and few investigations of mood depression caused by PM 2. 5. This article describes a survey study conducted in Beijing, China to estimate the welfare loss due to PM 2. 5. In total, 1709 participants completed either a face-to-face or online survey. A contingent valuation method was applied to elicit people's willingness to pay to avoid PM 2. 5 pollution and willingness to accept a compensation for such pollution. The payment/compensation was evaluated for two outcome variables: perceived health impacts and mood depression caused by PM 2. 5 pollution. This is one of few papers that explicitly studies the effects of PM 2. 5 on subjective well-being, and to the authors' knowledge, the first to estimate welfare loss from PM 2. 5 using a random forest model. Compared to the standard Turnbull, probit, and two-part models, the random forest model gave the best fit to the data, suggesting that this may be a useful tool for future studies too. The welfare loss due to health impacts and mood depression is CNY 1388.4/person/year and CNY 897.7/person/year respectively, indicating that the public attaches great importance to mood, feelings and happiness. The study provides scientific support to the development of economic policy instruments for PM 2. 5 control in China.