U.S. Gasoline Demand Strong Despite Poor Jan Weather

Unexpectedly weak gasoline consumption in the United States reported for the month of January has been blamed by some market participants for the continued slide in oil prices on Tuesday.

Gasoline consumption fell almost 0.6 percent in January compared with the same month a year earlier, according to the U.S. Energy Information Administration ("Petroleum Supply Monthly", EIA, April 4).

The EIA's monthly estimate for gasoline consumption contrasted with more recent weekly data showing strong year-on-year growth in gasoline demand.

The monthly estimate has been seized on by bearish traders and analysts as evidence crude and gasoline prices have risen too far too fast over the last two months.

But on closer inspection there is nothing surprising about the monthly reading for January, which is in line with the low weekly consumption numbers reported during that month.

Consumption was likely disrupted by the extensive flooding across the Midwest at the start of the month and the snow blizzards which hit the East Coast at the end of January.

Gasoline consumption only began to ramp up from the middle of February and has remained strong since then according to the weekly data (http://tmsnrt.rs/1RBCJ3c).

Since the end of February, U.S. gasoline consumption has been running at a seasonal record rate ("Weekly Petroleum Status Report", EIA, April 1).

Monthly estimates published by the EIA on Monday do not alter this picture of gasoline consumption significantly or indicate that it needs to be reassessed.

DATA RELIABILITY
Like almost all government statistics, every number published by the EIA is an estimate, in this case based on a (compulsory) survey of refiners, traders, tank farm operators, blenders and pipeline companies.

Like any estimate, EIA data is subject to a variety of sampling and non-sampling errors, but because the surveys capture almost all the relevant firms and are highly standardised, errors tend to be fairly small.

The monthly survey is based on a slightly more comprehensive sample so should be more accurate but the differences between the two surveys turn out to be minor.

There is considerable noisiness in the weekly survey caused by both sampling and non-sampling errors (http://tmsnrt.rs/1RBCOnk).

Taking a four-week average removes most of the noise and leaves the weekly survey tracking the monthly numbers closely (http://tmsnrt.rs/1RBCN2Q).

The EIA's Weekly Petroleum Status Report showed gasoline consumption averaging 8.715 million barrels per day in the four weeks ending on Jan. 29.

The EIA's monthly estimates now show gasoline consumption averaging 8.670 million bpd for the month of January, a difference of 45,000 bpd (0.5 percent), which is insignificant.

The four-week average turns out to be a very timely and accurate predictor of the monthly estimates which are published two months later.

Difference between the four-week average and the monthly data are generally small. The average difference between the two surveys is less than 0.5 percent (http://tmsnrt.rs/1RBCS6H).

Some of the residual difference can be explained by the fact that there are not exactly four weeks in each month so cut off points for the weekly and monthly surveys do not match exactly.

The EIA's weekly surveys correctly identified the weakness of gasoline consumption in January as a result of poor weather.

More recently, the weekly surveys have pointed to very strong gasoline consumption thanks to a combination of better weather, cheap fuel, strong job growth and wage gains.

Based on the prior performance of the weekly and monthly survey data, there is no reason to think gasoline consumption over the last two months has been anything other than very strong.

STATISTICAL ISSUES
Like any other statistical exercise, the EIA surveys focus on what can be measured even if it is not quite the same as what we actually want to know about.

In this case, we would like to know how much gasoline is actually consumed by motorists each month. But with more than 113 million registered cars and 260 million vehicles of all types on U.S. roads that is impossible ("Highway Statistics", Federal Highway Administration, 2015).

Instead the EIA calculates the volume of gasoline supplied to the domestic market based on domestic production plus imports minus exports minus reported stock changes.

The EIA collects weekly and monthly data on production, imports and stock changes, and monthly data on exports, and then estimates the amount of product supplied.

In effect, product supplied measures the amount of gasoline that disappears from the reporting system as a proxy for consumption.

Statistical errors in measuring any of the other components also show up as an error in the estimate for the amount of product supplied.

Exports have been notoriously problematic because the agency has only had access to monthly rather than weekly estimates, although this is now changing.

In the past, short-term changes in export volumes caused errors in estimates for the amount of product supplied.

Sudden shifts in the composition of demand between domestic and foreign consumers account for some of the noisiness in the weekly statistics.

But there is no reason to think they are causing larger than usual errors at the moment or to doubt domestic gasoline demand is very strong.

By John Kemp