The Need for Reliable Energy Data

I'm back at my desk after some business travel, and the item in this morning's batch of news that caught my eye concerns the reliability of the oil industry data collected by the Energy Information Agency of the US Department of Energy. The article in today's Wall St. Journal (subscription may be required) described EIA's methods for tallying oil inventories and other industry data as "antiquated and out-of-date." Nor is the Journal the first to draw attention to this issue. Last year US News & World Report published a story that reached a similar conclusion as the Journal: the EIA doesn't have enough money in its budget to do both the work expected of it and improve its processes. Yet I can't help wondering whether the real issue we ought to be focusing on is improving the accuracy of the oil data, or getting the data for other, increasingly important energy sources up to at least the same level of timeliness, comprehensiveness and accuracy as those for oil.

Before writing this, I had a quick conversation with one of the experts at the American Petroleum Institute who is involved in reviewing and analyzing the weekly industry statistics API puts out to subscribers. Although gathered independently and on a voluntary, rather than government-mandated basis, API's reports generally reflect the same underlying data and sources as EIA's. The last time I was actually involved in submitting EIA/API data from an operating facility was in the early 1980s, when everything was faxed in and compiled manually. I was surprised to hear that some of the data still comes in that way, though most of it is apparently gathered electronically, either though electronic data interchange or via email. What he emphasized to me, though, was that regardless of how the data is actually assembled and reviewed, it actually represents an extremely accurate survey, covering something like 85-90% of the industry, with non-filers' results estimated from less frequent census-type reports. That's much more comprehensive than the sampling rate for many of the other economic statistics on which the market depends--and to which it sometimes reacts violently.

One of the problems with any such system involves how the information is used. As long as traders focus so keenly on week-to-week changes, rather than the totals, this will tend to amplify the impact of any errors that creep in. For example, in last week's EIA statistics, the entire US commercial inventory of crude oil stood at 344 million barrels, reflecting a 1 million barrel increase from the previous week. An error of just 2 million barrels in either direction--or 0.3% of the total--could have increased that inventory build to 3 million barrels or swung it to a 1 million barrel drop, with very different outcomes for oil prices. While it would be nice to think errors of that magnitude could be avoided entirely, should the market be so sensitive to such changes, knowing that no assessment like this can ever be made 100% accurate, no matter how precisely it is assembled?

While the system might lend itself to improvements such as requiring electronic data submission by all participants and adding more analysts to scrutinize the filings for errors and omissions, I suspect the more urgent priority is expanding its scope to encompass all of the energy sources on which we now depend. After all, when the current national energy information system was first devised petroleum-based fuels were essentially the whole game for transportation energy, while still accounting for a significant portion of the input to fossil fuel power plants. Today ethanol satisfies roughly 8% of US gasoline demand, and the 14-16 million barrels of inventory that the ethanol industry keeps on hand is the energy equivalent of about 7% of the 200-230 million barrels of gasoline and blending components the oil industry has at any point. Those percentages are mandated by law to grow significantly in the next decade, as biofuels displace petroleum products.

How much longer should we be satisfied with production and inventory data for biofuels that are weeks or months out of date, when we require accurate weekly updates on petroleum and its products? And consider that this picture will only become more complicated as an increasing proportion of our needs are satisfied by various renewable and distributed energy sources. If we can spend billions improving the management and storage of health data, wouldn't it be worth widening our net and spending an extra few million to get a better handle on the energy flows and stocks upon which the entire economy depends?