over the sample period. For the fixed effect and panel FGLS models, the marginal effect of
ethanol production on the crack spread is estimated to be -0.000073 and -0.000077, respectively.
Regional Analysis
Pooling cross-sectional and time-series information provides more accurate estimation results.
However, it is instructive to analyze the time-series data of each region individually. Each PADD
region has unique supply and demand conditions of crude oil and refinery products, different
market structures, and different ethanol production and usage. The effects of explanatory
variables may differ considerably because of region-specific factors.
We apply regular OLS regression on individual region’s monthly data series over the period
1995 to 2007. The estimation results for the relative gasoline price and 3:2:1 crack spread are
summarized in Tables 3 and 4, respectively.
From the estimation results for the relative gasoline price, ethanol production has a significant
negative effect on gasoline prices in all regions. And the magnitude of the effect varies with
PADD regions, ranging from -0.000041 to -0.000095. As expected, in PADD II, the Midwest
region, ethanol production has the largest impact on the gasoline price with a coefficient
of -0.000095. The substitution effect is highly significant and reduces the gasoline price by 39.50
on average over the sample period. The West Coast and East Coast experience similar negative
ethanol impacts with estimates of -0.000056, which means that the corresponding gasoline price
is lowered by 23.30. The Gulf Coast region, PADD III, has a slightly higher coefficient estimate
of -0.000059, or, equivalently, a 24.60 reduction in gasoline prices. The Rocky Mountain region,
or PADD IV, experienced the smallest downward gasoline price change, at 17.10, probably
because of its comparatively low total gasoline consumption. These results tell us what would
have happened had we removed the entire ethanol industry at the mean of the data set, and they
are not marginal effects of removing one unit of ethanol capacity in each region.
From the estimation results of the profit margin for individual regions, effects of some
explanatory variables differ considerably across regions. In PADD regions III and V, the HHI
has a significant positive effect on refinery profit. This result suggests that higher market
12