London, 5 May 2011: Reuters

General Motors and other American car manufacturers reported strong vehicle sales last month led by sales of smaller and more fuel efficient cars and sport-utility vehicles. While some of the increase may represent a rise in market share at the expense of Japanese automakers whose supply chains have been disrupted by the earthquake, the bulk of the quake effect was yet to feed through in April, and it probably signifies a shift in consumer preferences in response to escalating fuel costs.

"Consumers are continuing to rethink their vehicle choice," according to GM sales head Don Johnson, in remarks referring to the impact of high gasoline prices reported in the Wall Street Journal. Speaking for an automaker, Johnson has no axe to grind in the debate among analysts about whether rising oil prices have begun to ration and destroy demand.

Johnson's comments lend support to the view of analysts at Goldman Sachs, who wrote last month that they saw "nascent signs of oil demand destruction in the United States."

It is a view that has been rejected by leading research teams at Barclays Capital ("it is far too premature to signal that the first signs of demand destruction are already noticeable") and Morgan Stanley (we have "not seen any material evidence to convince us any (gasoline) demand destruction is taking place in the U.S.").

Goldman's thesis got some further support last week when the U.S. Energy Information Administration reported domestic gasoline consumption was running at a rate of just 8.648 million barrels per day in February, unchanged from last year and down more than 2 percent compared with 2009 and 2008.

It was a big downward adjustment from the average 8.9 million barrels per day the agency reported at the time in its weekly consumption estimates.

ELUSIVE CERTAINTY

In an insightful analysis published on Tuesday, my colleagues Jeffrey Kerr and Joshua Schneyer explained why motorists may now be less affected by the sticker shock of $4 gasoline than they were in 2008. But while motorists have had time to get used to previously undreamed of gasoline prices, and the highs of three years ago may be acting as a nominal anchor, it is more likely than not that rising prices are already having both a rationing effect (on discretionary driving activity) and a destruction effect (through new vehicle choice).

In the short-term, rationing has the biggest impact, since destruction is embedded in long-lived technology changes such as size and efficiency of new motor vehicles, which are replaced only very slowly. Destruction is more important in the long-term as technology choices made today under the influence of high prices continue to affect consumption for years to come.

The question is whether there is any evidence of significant rationing and destruction already under way -- beyond anecdotes and isolated data such as new car sales. In practice the data on gasoline consumption is subject to substantial estimating errors ("noisy") and only available with lengthy delays, so there are no unambiguous answers. Moreover, the long-lasting effects of demand destruction take place only very gradually and are usually only visible in retrospect.

In their own ways, Goldman, Barclays and Morgan Stanley are all right. While there may be nascent signs of demand destruction (Goldman) they are surrounded by enough uncertainty it is hard to draw firm conclusions (Barclays, Morgan Stanley).

Searching for signs of demand destruction is a good example of decision-making under uncertainty. What level of certainty and confidence should we expect before reaching a conclusion, even a tentative one?

In theory we should insist on a high level of confidence. Unfortunately, the data is not clean enough to make that realistic. By the time we get something approaching certainty, the data will be so old it is no longer relevant and the market will have moved on.

This is precisely the problem that beset oil analysts in H1 2008, when the market could not agree whether prices had yet risen far enough to cause significant demand restraint. Only after prices had peaked and were falling did the majority of analysts agree that demand had already been falling during the first six months of the year.

GASOLINE SUPPLIED

Gasoline consumption is not measured directly. Instead EIA reports something termed "product supplied" which "approximately represents consumption of petroleum products because it measures the disappearance of these products from primary sources" such as refineries, blending plants, pipelines and bulk terminals.

Gasoline supplied is derived from other data on refinery output, imports, inventories and exports. Product supplied is basically refinery output plus imports minus inventory change minus exports. EIA collects data on the first three (refinery output, imports and stocks) through compulsory weekly and monthly surveys which have a high degree of accuracy.

But there it has no direct data on exports. Instead exports are estimated for the "Weekly Petroleum Status Report" (WPSR) based on recent months extended using prior -year seasonal trends. Definitive export data become available from U.S. Customs and Border Protection about two months later and are then used to provide more accurate estimates for exports and product supplied in "Petroleum Supply Monthly" (PSM).

The problem with this system is that any errors in estimating exports create equal and opposite errors in the estimates for consumption that are not resolved until the definitive export and PSM data become available two months later.

Past experience shows weekly gasoline estimates can vary significantly from the more accurate monthly measure by as much as several hundred thousand barrels per day or 2-6 percent (Charts 1-5).

This would not matter except the changes in demand as a result of rationing or destruction that we are looking for lie in the same range. There is no reliable way to extract changes in gasoline demand from the weekly statistics. Only the monthly stats provide useful guidance -- and those show clear signs that gasoline consumption remains depressed and is not showing anything like the same recovery as distillate supplied.

MORE DATA NEEDED

Estimating exports in the WPSR to calculate product supplied was reasonable when U.S. exports of gasoline and distillate were small and did not change much. But as U.S. refiners have become more export-oriented, and gasoline and diesel exports have surged, estimating errors for both exports and product supplied have grown.

As gasoline exports have risen since 2008, the WPSR has tended to systematically overstate consumption, as it misses the uptrend in exports, only for consumption to be revised lower when the PSM is eventually published. Under the 1974 Federal Energy Administration Act, EIA is required to collate data on coal and petroleum exports (15 USC 784 (a)) but in practice it satisfies the requirements by using data from other agencies such as customs returns (15 USC 784 (d)).

In an ideal world, EIA would obtain its own firm weekly data on exports. But it would need approval from the White House Office of Management and Budget, and it may not be possible at a time when the agency is facing deep cuts to its data collection and analysis as a result of the recent budget deal.

In the meantime, the monthly data are the only useful ones on product supplied. They show consumption lagging the recovery and reduction in unemployment, most likely in part as a result of rising prices through a combination of demand restraint and destruction.

In the battle of the analysts, Goldman looks more likely to be right than wrong.

Ends --

By John Kemp, Reuters market analyst – for Commodities Now. The views expressed here are his own.