Analysts wonder if US gasoline consumption data may be overstated

By John Kemp

LONDON, April 11 (Reuters) -- Strong growth in US gasoline consumption has been one of the most important factors supporting oil prices in 2016, but some analysts question whether the official data is overstating the strength of gasoline demand.

US gasoline consumption has been running at a record seasonal rate since February, according to weekly data published by the Energy Information Administration (EIA) in its "Weekly Petroleum Status Report" (WPSR).

But the consumption figures are so high that some analysts question whether the gasoline is actually being exported rather than consumed domestically.

The EIA's oil-data collection program is designed to track the flow of petroleum from the well head to the final consumer.

Under authority that dates from the energy crisis of the 1970s, the agency collects information on domestic production, imports, exports, refining and stockpiles.

The major gap in the system is consumption. Consumption is almost impossible to measure in real time because there are too many end-users (there 260 million registered vehicles in the US).

So the EIA uses the disappearance of refined products from its primary reporting system as a proxy for the amount of fuel consumed by end-users.

The volume of "product supplied" is calculated as the residual of domestic refinery production plus imports minus exports minus the change in stockpiles.

Because product supplied is calculated as a residual, any errors measuring production, imports, exports and stockpiles flow through directly into errors in implied consumption.

Estimated exports are the largest source of uncertainty about consumption in the short term. If the EIA over-estimates exports, it will under-estimate domestic consumption, and vice versa.

The agency collects its own comprehensive data on production, imports and stockpiles from refiners, importers, pipeline companies and motor fuel blenders.

But it relies on data collected by US Customs and the Census Bureau to measure exports, and these are only available with an average lag of around 8 weeks.

In the meantime, the agency is forced to estimate the volume of refined fuels which are exported, and any errors in those estimates flow through directly into errors in implied consumption.

The EIA estimates exports in the WPSR and then replaces the estimates with hard data from the Census Bureau when the monthly numbers are published in "Petroleum Supply Monthly" (PSM).

The EIA's weekly and monthly surveys both showed domestic gasoline consumption was subdued in January, most likely because of bad weather affecting large parts of the country at the beginning and end of the month.

But consumption reportedly rebounded to record seasonal rates in weekly surveys for February and March leading some analysts to question whether the EIA is under-estimating exports.

This is certainly possible. The EIA's estimating procedure has a backward-looking component so it can miss sudden changes in the rate of exports and only responds with a lag.

The question is how large these errors in estimating exports tend to be and how much of an error they impart to estimates of domestic consumption.

US gasoline exports are very small in relation to domestic consumption so even relatively large errors in estimating exports have only a small impact on consumption estimates.

In 2015, the country exported an average of 475,000 bpd of finished gasoline and another 150,000 bpd of blending components compared with domestic consumption of nearly 9.2 million bpd.

The difference between exports and domestic consumption is more than an order of magnitude: the amount of gasoline exported amounted to less than 7% of domestic consumption, according to the EIA.

The agency would need to make some truly enormous errors in estimating exports to have more than a marginal impact on its estimates for consumption.

A comparison between the agency's weekly estimates for exports (smoothed with a four-week average) and the actual volume exported as reported later by the Census Bureau suggests errors are mostly within +/- 125,000 bpd and at most about +/- 250,000 bpd.

The resulting error in implied gasoline consumption is normally +/- 1%, or at most +/- 2.5%, which is a fairly a high degree of accuracy for an economy-wide statistic.

Estimating errors are potentially much more serious for distillate fuel oil, where exports averaged almost 1.2 MMbpd in 2015, compared with domestic consumption of almost 4.0 MMbpd.

The EIA reported that finished motor gasoline exports averaged 445,000 bpd for the four weeks ended Feb 26, while domestic consumption averaged 9.256 MMbpd.

In fact, we now know from customs data, finished gasoline exports averaged about 720,000 bpd during the month of February, according to the US International Trade Commission.

If exports were understated by 275,000 bpd, then domestic consumption was overstated by a similar amount and was probably running at around 8.981 MMbpd in February.

Even adjusting for higher exports, domestic gasoline demand was still running at near-record rates in February and likely climbed even further in March.

Government statistics are often reported to several decimal places or significant figures, which can make them seem more precise than they really are.

Every statistic, especially for large estimates like economic output, is an estimate and subject to a range of measurement errors.

Government data should always be treated as approximations to the underlying reality and analysed with an appropriate degree of caution.

Provided the data show big trends, that is generally good enough; analysts should be wary of ascribing too much significance to small changes that could just be the result of measurement errors and noise.

Josiah Stamp, a British industrialist, tax inspector and statistician, gave a warning about the limits of government statistics, which has become known as Stamp's Law.

"The government are very keen on amassing statistics. They collect them, add them, raise them to the nth power, take the cube and prepare wonderful diagrams. But you must never forget that every one of these figures comes in the first instance from the village watchman, who just puts down what he damn pleases".

Stamp's Law is worth recalling when analyzing the weekly and monthly petroleum supply estimates published by the EIA.

The EIA's data collection efforts have their origin in the energy crisis of the early 1970s, when a sudden shortage of gas and oil, worsened by the Arab oil embargo, caught politicians and the public unprepared.

Congress ordered the government to start collecting and analyzing its own statistics rather than rely on data from industry associations.

The Federal Energy Administration, forerunner to the US Department of Energy, was given power to demand detailed information from energy producers and consumers.

In the current financial year, the EIA has a budget of $122 million. As a result, more is known about the supply and disposition of oil, gas, coal and electricity in the US than anywhere else in the world.

Even so there are limitations that can wrong-foot the unwary. In the case of gasoline exports and domestic consumption, however, the errors appear to be relatively small.