The Missing Inventory

While I was on vacation, a potentially important energy threshold was reached without much media attention, outside of Robert Rapier's energy blog. As gasoline prices declined steadily for the last month, except in the Midwest, gasoline inventories were dropping, too. Last week, for the first time since at least 1991, they fell to a level equivalent to twenty days' supply at current consumption rates. That ought to signal that gas prices will soon head back up, even though we're now past the traditional end of the peak driving season. However, a key factor is missing from these figures. Ethanol may only amount to 3.5% of our gasoline consumption, but its volume is growing large enough to affect our view of traditional industry statistics.

The energy industry produces a bewildering variety of data, but analysts pay particular attention to crude oil and refined product inventories, because they reflect the outcomes of the interaction of all current supply and demand factors. When inventories fall, prices generally rise, and vice versa. We certainly saw that earlier this year, when gasoline inventories dropped by 15% from the end of January to end-April, and retail gasoline prices rose by 35% in response. Something unusual has been affecting these inventories throughout the entire 2007 driving season. Total US gasoline inventories have remained below their seasonal-average range since March. While this is partly due to refining problems in the Midwest, inventories have been quite low on the East Coast and Gulf Coast, too. Something else is at work.

The chart below, taken from last week's edition of the DOE's This Week in Petroleum, expresses these inventories in terms of the equivalent number of days of gasoline consumption. This provides an even more focused view of inventories, because it takes the current level of demand into account. It shows that days' supply normally ranges between 21.5 and 25 days, with a strong seasonal pattern. When the EIA reports that this measure has fallen to 20 days for the first time since they've been monitoring it, we should take notice and possibly worry. In fact, the curve for this year's data is consistently 0.5-1 day below that for 2006. Besides refining problems, what else could explain this shift? I think ethanol provides at least part of the answer.

Consider how the growth of ethanol blending has changed the gasoline business, driven by a combination of gasoline reformulation, the phaseout of MTBE, and the introduction of a national renewable fuel standard. Since gasoline containing ethanol can't be transported in pipelines, refiners must produce a pre-ethanol blend for shipment, to which ethanol is added at distribution terminals close to the point of sale. In just the last two years, the share of total gasoline inventories made up of this "RBOB" pre-blend has increased from 10.7% to 20.2%, reflecting the rapid increase in ethanol production. But while RBOB is counted in gasoline inventory, the ethanol required to turn it into merchantable gasoline isn't. That means that EIA inventory data is missing at least 2% of the volume that will end up in total gasoline sales, the equivalent of 4 million barrels, or nearly a half-day's supply at the national level. A quick look at ethanol inventories, which stood at nearly 9 million barrels in April, validates this logic.

In other words, when we add in the ethanol held in storage for blending into gasoline that's also in storage, we find a volume sufficient to eliminate most of the days'-supply gap between the 2006 and 2007 data. Before we relax too much, however, we should recognize that even 21 days of gasoline inventory represents a pretty slim cushion, relative to typical historical levels. And if most of that uncounted ethanol is sitting in rail cars or in places where it far exceeds the 10% limit for blending into gasoline, rather than E-85, then it wouldn't do much good in a pinch--such as after another big Gulf Coast hurricane. But what seems pretty clear is that, as our reliance on ethanol grows, our use of inventory and other industry benchmarks must adjust to the reality of a non-petroleum blending component significant enough to distort the old patterns.