Computational development economics Applying ML to the multiple avoidance mechanism problem: willingness to pay to avoid pollution in the Global South

My proposed research agenda would provide a new quantitative method to estimate how much people value policy that improves environmental outcomes at the cost of economic growth. This research agenda is important because it informs how governments should make trade-offs between environmental and economic outcomes when designing environmental policy. Existing research on this topic in the Global South has severely underestimated how much people value environmental regulation because it only considers a single avenue through which people avoid environmental damage. For example, the recent research of Ito & Zhang (2018) and Zhang & Mu (2018) only consider expenditure on HEPA air purifiers and particulate-filtering face masks, respectively, when estimating how much individuals in China would value regulation that limits air pollution. My research would involve producing a machine learning-based estimator that allows us to identify what kind of purchases individuals make to avoid environmental damage. I would then reestimate the study of Ito & Zhang (2018) on air pollution in China using this new approach. I will provide new evidence on people's adaptive behaviours that they employ to avoid air pollution. Firstly, I will provide novel evidence on the relative importance of adaptive behaviours. To avoid the impacts of air pollution, do people spend more on prevention (e.g. buying face masks with HEPA filters) or damage mitigation (e.g. purchasing medication to manage health conditions associated with breathing air pollution)? I expect that I will also materially increase the estimate of how much individuals value environmental regulation. The policy implications of this research are immediate. If the valuation is indeed materially larger than existing studies have shown, this would provide evidence to the Government of the People's Republic of China in favour of more aggressive environmental regulation. Plausibly, the larger impact of the research, though, is in providing a novel method for valuation of environmental outcomes that is applicable across a broad range of environmental outcomes. While I will show application to air pollution regulation, this approach is plausibly relevant in estimating how governments should make trade-offs between economic growth and other environmental outcomes, providing evidence on the value of regulation to avoid other causes of environmental damage. For example, this work might find application to valuing changes in precipitation (rainfall) and temperature associated with climate change. In turn, it would provide evidence on optimal climate change policy.