Date of Award


Degree Type


Degree Name

Doctor of Philosophy



Major Professor

Luiz R. Lima

Committee Members

Jacob S. LaRiviere, Scott J. Holladay, Edward T. Yu


This dissertation consists of two chapters. Chapter 1 examines the effect of transportation costs of shipping ethanol on retail gasoline prices over space. The Renewable Fuel Standard (RFS) of 2007 legislated a new market into existence in the U.S. by mandating that ethanol be blended with petroleum in retail gasoline markets. Using a quantile difference-in-differences econometric approach to analyze weekly retail gasoline price data for over 200 cities from 2007 to 2014, we find evidence that the mandate differentially impacted gasoline prices across the U.S. Specifically, we find that cities farther from ethanol production centers paid higher retail gasoline prices than cities close to ethanol production centers. We argue that the observed retail price differences are driven by market frictions associated with transportation costs for ethanol which, unlike petroleum, cannot be shipped via pipeline. This effect has been exacerbated due to the run-up in ethanol RIN (renewable identification numbers) prices starting in 2013. Importantly, the effect of this market friction on retail gasoline prices varies with the relative prices of ethanol and petroleum blendstock. Our results highlight the spatial incidence associated with the mandated ethanol market. While unanticipated, we argue that these market frictions are not surprising.

In Chapter 2 we investigate the forecasting performance of a variety of individual models found in empirical literature and their linear combinations in the context of carbon dioxide emissions. We conduct out-of-sample forecasting exercise by using state-level data for carbon dioxide emissions in the U.S. Forecast error and tests of predictive accuracy are compared both for individual models and their linear combinations. Consistent with reported results on the application of forecast combinations, we show that the forecast combination technique generally improves forecast accuracy. The best performing combination outperforms all the individual models as the forecast horizon increases. More importantly, forecast accuracy from the best performing individual model is not significantly better than that of the best combination forecast. Among the class of forecast combinations considered in this paper, bias-corrected average forecast performs relatively well.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."