Trends in U.S. air pollution: A time series analysis

This dissertation consists of three essays, in which I utilize time-series econometric techniques to study air pollution emissions trends in the United States. In the first essay, which is written jointly with Eric Zivot, we use an array of unit root and structural break tests to determine whether emissions of Nitrogen Oxides and Volatile Organic Compounds are trend-stationary or difference-stationary and whether there was a break in the trends of these emissions around the same time the Clean Air Act of 1970 was passed. We find Volatile Organic Compounds emissions to be trend-stationary with a break at the time the Clean Air Act of 1970 was passed and we do not find a trend break in the Nitrogen Oxides emissions. In addition, results regarding trend-stationarity of Nitrogen Oxides emissions depend on the methods used. Thus, in the second essay, I conduct a Monte Carlo experiment to study the size and power of two unit root tests that allow for a structural change in the trend at a known time using the data-generating process calibrated to the actual pollution series. I demonstrate that finite sample properties of the Perron test are better than the Park and Sung Phillips-Perron type test. Severe size distortions in the Park and Sung test can explain the rejection of a unit root in the Nitrogen Oxides emissions. In the third essay, I study the dynamic relationship between air pollution emissions and macroeconomic variables, such as output, energy consumption and energy intensity, in the United States using vector-autoregression methodology. I find that energy consumption and energy intensity Granger cause emissions of Sulfur Dioxide and Carbon Dioxide. I also find that there is no Granger causality between output and emissions and that Granger causality between output and energy consumption depends on the methods and sample size used.