Air pollution has plagued cities around the world for years. Major metropolitan areas often install air pollution regulations that vary in stringency over time. Regulations on polluters are lax on days when the air quality is good. For days with high ambient air pollution, governments impose stricter measures to avoid exacerbating the already poor air quality. This study focuses specifically on the Heavy Air Pollution Emergency Plan (HAPEP) of Chengdu, China. During the winter months, Chengdu frequently faces periods of sustained, high levels of ambient particulate matter (PM). Importantly, these high levels of ambient PM are not ruled by temporary increases in the flow of PM being emitted each day. Instead, it is driven largely by the weather conditions. For example, during periods of stagnant wind conditions, the stock of PM grows in the air. This accumulation process continues until the weather conditions change -- i.e., the wind picks up -- and the stock of PM in the air is cleared out. This pollution pattern is representative of the cities which are in a closed basin terrain. The Chengdu HAPEP requires monitoring short-term PM forecasts. If the predicted PM exceeds certain concentration and duration thresholds, an air pollution alert is issued. Subsequently, a set of measures intended to lessen air pollution are imposed, including driving restrictions and production suspensions.
In the first chapter, I develop a dynamic optimization framework taking into account the characteristics of PM pollution by directly modeling the stock nature of pollutants and the cleanup process of pollutants. The model leads to an important conclusion: the optimal amount of pollution emission should always increase over time if there is no pollutant dispersion. Although in reality pollution dissipation is never absolutely zero, this conclusion indicates an incentive to delay the pollution for social welfare improvement. However, the variations in expected pollution dissipation should also be incorporated when deciding the optimal action.
In the second chapter, I empirically estimate the dynamic process of pollution dissipation by identifying how daily weather conditions drive the change in ambient PM levels from one day to the next. I also show that the weather conditions in the near future can be forecasted with high accuracy. By altering the timing of historical interventions according to both existing conditions and expected upcoming weather patterns, a 12.1% more PM pollution reduction and a 25.5% more bronchitis hospital visit reduction can be achieved within a period in 2018.
In the third chapter, I estimate the health effects of PM pollution. I find that pollution exposures up to six days ago are associated with contemporaneous bronchitis hospitalization. I also assess the curvature of the response function of respiratory hospitalization to PM pollution. I fail to find any convincing evidence that supports the presence of nonlinearity. One caveat of this conclusion is that I only consider possible non-linear effects of the contemporaneous pollution concentration but ignore the possible non-linear effects of the lagged pollution concentration, due to the complexity of the specification and the difficulty in identification if both of them are incorporated. Although this simplification might introduce a bias in an unclear way, it is of less concern since prediction instead of specific point estimates is the focus of this analysis.