Damn Lies and Government Statistics

A miracle of modern finance is how the markets respond to government reports that are known to be often inaccurate and heavily revised. The inadequacy of the surveys and methods used to compile these reports (especially the market-moving employment report) is often neglected when the markets respond to the release; the “headline number” is all that matters.

One of the most closely watched figures of late has been the weekly oil output numbers released by the Energy Information Administration. The markets have been desperate to know whether the Saudi price war would succeed in flattening the U.S. industry or at least sharply impairing its growth. And yet according to accomplished oil trader Andrew J. Hall, the published estimates are taken from state agencies and lag by several months. The numbers are “essentially an artifice.”

The chart above shows, in orange, the EIA estimates to which all the attention is usually paid, the weekly output figures. It turns out, however, that the agency itself thinks it may have a better gauge. This is represented by the blue line. (I don’t recall having ever previously seen a graphic presentation of this alternative data.) 

Hall believes the recent fall in output as shown by the blue line is very bullish for the oil price, though it might be noted that the adjustments, though showing a steep fall recently, still show production above the normally consulted weekly output figure.

Bloomberg explains the basis of the different methodology:

A more accurate gauge of U.S. output is an “adjustment” the agency uses, which, in addition to the weekly number, adds up changes to how much oil is in storage, how much was used in refineries and how much was imported and exported, Hall said. 

Paul Sankey, an energy analyst at Wolfe Research, also cited the trend in a report Thursday to investors. The amount of production that isn’t accounted for has fallen “dramatically” in the last two weeks, suggesting U.S. daily output may have fallen by as much as 200,000 to 300,000 barrels in April, he said. “That number seems high to us, but it does support the notion that U.S. production is rolling over at present,” said Sankey, a former analyst at the Paris-based International Energy Agency. 

Imports, refinery demand and storage level data all come from surveys the EIA conducts with oil companies. Export data comes from the U.S. Census Bureau. Production data comes from a variety of sources, including state and federal regulatory agencies. In a perfect world, the supply and disposition would equal each other, said Mike Conner, a petroleum analyst at the EIA. But they often don’t, so the EIA uses the adjustment figure to balance it out. “All we really know for sure is the supply components are not enough to make up for the volume on the disposition side,” Conner said. “We don’t know if the error is in field production, imports, refinery inputs or what have you. Different analysts are going to have different interpretations.”

Bradley Olson and Dan Murtaugh, The Shale Boom Has Already Gone Bust—At Least for Now, May 3, 2015, Bloomberg Business 

* * *

Update, May 12: The question of EIA statistics becomes curioser and curioser. The latest was a blast on Monday against the EIA for gross incompetence in its methods of data collection. The EIA, charged oil analyst Philip Verleger, was overestimating U.S. output by as much as 1.6 million barrels a day. The glut the markets had previously identified was seemingly non-existent, a “phantom.”  Verleger was previously on record as predicting that oil prices would fall back to $50 a barrel by the end of the year, a prediction predicated on the existence of a supply glut. So this was a big reversal for him, and it was for oil markets, if true, a very big deal.  

Then today (Tuesday) came a further twist, with Verleger performing a flip-flop and saying it was all a big mistake. According to the FT, Verleger “said economists at the Federal Reserve had examined his calculations and pointed out an error.” “The effect of the error is to change the sign of the Energy Information Administration’s mis-calculation of US crude oil production,” said Verleger. “It turns out that EIA is underestimating US oil output.”

Got that straight?

There follows the Monday story in the FT, followed by excerpts from its Tuesday report. Wednesday may bring further clarification, but it would seem, on the face of it, that the heat of Verleger’s denunciation on Monday seems rather misplaced given his reversal on Tuesday.

A prominent oil analyst has fired a broadside at the body charged with data collection on the US oil industry, saying it has probably overestimated the country's crude output by as much as 1.6m barrels a day.

Philip Verleger, an independent energy economist, claims the US Energy Information Agency (EIA) has "drastically" overestimated US oil production because it uses outdated techniques rather than data from the field,  reports Neil Hume, FT Commodities editor, in London

Mr Verleger said in a report published on Monday:

This is a large variance that could have enormous implications for the global economy. Prices will be higher because the [supply] glut was phantom. Federal Reserve policy could easily have been different had the error been understood.

The EIA, which acts as the statistical arm of the Department of Energy, declined to comment.

Mr Verleger, who runs consulting firm PKVerleger and advised both President Gerald Ford and Jimmy Carter administrations on economic and energy policy, was one of the few analysts bearish on oil before last year's dramatic collapse in prices.

Last week he forecast a further decline in prices, which have rallied sharply since reaching $40 a barrel in January.

On Monday, Brent, the international marker, was trading at $65 and the US equivalent, West Texas Intermediate, was at $59.50.

Mr Verleger said in his report:

The rise in oil prices… has baffled many who believed global stocks were surging. Global stocks would have done so had the numbers been correct.

Analysts say investors should be careful how they interpret US production statistics, in particular the EIA's weekly short-term supply numbers. This is because they are based largely on modelled estimates, not hard supply data from the field.

In a recent analysis of historical estimates against lagged production data, Citi found short-term EIA forecasts underestimated final data by as much as 200,000 b/d over the last year.

Mr Verleger has come to different conclusion but says the EIA's reliance on estimates and "failure" to make contemporaneous accuracy checks of its numbers is a "dereliction" of responsibility on the EIA's part.

He said in the report:  

Rarely if ever had a US agency charged with collecting data made a miscue of this magnitude. The EIA administrator should be dismissed immediately for gross incompetency.

Neil Hume, US overestimates oil production – top analyst says, FT, May 11, 2015

* * *

From the Tuesday report:

The EIA belatedly responded to Mr Verleger's report on Tuesday, saying the claim of a massive understatement of production was extremely unlikely

"Notwithstanding a widely-reported note alleging that EIA's domestic crude production data are probably overstated by 1.6m barrels a day, EIA management and staff are confident in the quality and validity of the agency's production data. Furthermore, EIA's monthly crude oil production statistics, based on official State oil and gas data have been closely aligned with the weekly estimates reported by EIA," it said in emailed comments.

"Turning to the larger items in the balance equation, EIA directly collects data on refinery crude runs, changes in crude oil stocks, and crude oil imports every week. We are confident in the quality of these data. Unless the reported runs data, the stock data, and the imports data are massively wrong, production cannot be overestimated by 1m to 1.6m b/d as alleged," it added.

Neil Hume, Reverse ferret from oil analyst on crude estimates, FT, May 12, 2015