Frequently faced with sharp criticism from industry and WallStreet for its estimates on everything from supply and demand tonatural gas prices, the Energy Information Administration earlierthis month publicly admitted there were serious errors over morethan a decade in its gas and oil drilling data.

During a recent effort to update well completion estimates withactuals and check for mistakes, EIA detected some “unusualpatterns” in its results. “Discrepancies” were uncovered thatturned out to be much more serious than first thought. For totalwells by year, underestimates ranged up to more than 2,100 wells,11% of the 1995 total. And the impact of the errors extended backin time to at least the early 1980’s, EIA said in its March NaturalGas Monthly.

“Users of the EIA drilling activity data are therefore advisedthat the drilling activity data which were published or otherwisedistributed by EIA prior to February 1998 are substantially inerror,” a note in the publication warns.

“I expect there were a lot of financial decisions based on[these incorrect numbers],” said EIA’s Phil Shambaugh, co-author ofthe report that explains the errors. “That’s why we wanted to digin and make sure we got all of the wrinkles out of it and thenpublicize the fact that, hey, there have been problems with thesenumbers.”

Revisions show some striking differences. It turns out themid-1990s drilling decline was not as steep as previouslyindicated. The drop in prices did lead to fewer wells by 1995, butthey had been underestimated by 11%, EIA said. Further in the past,the corrections show drilling activity did attain a peak level in1981, but the peak was higher than first reported: 91,496 wells asopposed to 90,034. Gas and oil exploratory wells were greatlyunder-reported in the post-1985 period with more than half of thewells missing in certain years. And success rates measured as theshare of successful gas and oil wells relative to the total wellsimproved greatly as early as 1986 rather than by the mid-1990s.

“I was troubled quite a bit by the difference in the successrates. That’s the sort of thing that is so fundamental for ourunderstanding of the industry and how it reacted to the economicsof the late 1980s when the industry was going through some veryhard times,” said William Trapman, an EIA economist and co-authorof the report on the errors. “So our understanding of the impactsof economics as well as the adoption of alternate techniques to tryto enhance industry performance – 3-D seismic was adopted in thatperiod – [was not quite correct]. If you look at the success ratesusing the old data, it looks as though nothing really happeneduntil just a few years ago even though we know anecdotally andotherwise that these technologies were being applied on awidespread basis.. The data understated by a substantial margin thefavorable impact of some of these techniques that the industry wasapplying to try to deal with the very severe conditions that theyfound themselves in.”

A Simple Misunderstanding?

EIA is quick to point out it doesn’t collect the data.Houston-based Petroleum Information/Dwights has been doing the workfor EIA since 1994 for about $15,000/year. Prior to that,PI/Dwights provided the data to the American Petroleum Institute,which then shipped it to EIA.

In trying to avoid placing the blame on its only source ofdrilling data, Shambaugh attributed the errors to miscommunicationparticularly following the switch to PI/Dwights. He said “not fullyunderstanding how the private companies were collecting the dataand what each variable meant in its full entirety got us into thisproblem.”

“It was just a misunderstanding, I think,” said PI/Dwights PeteMcClintock, manager of statistics. “I’m not so sure they understoodthat the data that they were getting, the data that we were sellingto the API was filtered by API specifications.. There were somewells that API didn’t want delivered to them. For instance, theydidn’t want any workovers. They didn’t want any wells with lessthan 50 feet of drilled footage. They didn’t want what we called onour system pending wells, wells that didn’t have completeinformation on them. EIA would have wanted those wells if they hadrealized they weren’t getting them. I got the impression after Ispoke with the DOE at length that they assumed they were gettingall the wells.

“They apparently didn’t look at what they got or understood whatthey got. I wasn’t quite sure that these guys had a handle onprocessing this data and an understanding how this data reallyworks,” said McClintock, noting EIA’s modeling methodologies areoften criticized even by EIA personnel. EIA has to use anestimating model on drilling data because of the time lag in wellcompletion reporting. Despite EIA’s claims to the contrary,McClintock said he’s not sure the problem has been correctedcompletely “or if it’s a case of even the blind squirrel finds anut sometimes.”

While noting the good work PI/Dwights does, EIA’s WilliamTrapmann said the biggest problems occurred simply because “we didnot get all the records” from PI/D.

“I thought there was a programming error involved. I’m not surewhat the misunderstanding would have been… There were a couplemonths in the 1995-96 period in which it was something like halfthe records weren’t present. That was the worst case. And that’snot a misunderstanding about recompletions.”

EIA said the problems would not have occurred had it done itsown data collection all along. Using an outside source “does notallow complete familiarity with the data, and it hampers efforts toinvestigate concerns with the data. Second, the well completiondata are inherently quite erratic, so even sizable fluctuations arenot prima facie a cause for concern.” And third, reported data areoften inflated because they contain wells that were completedmonths and years prior to the reporting date. Perhaps worst of all,there is no other timely source for well completion data, EIA said.

But the time and expense of conducting well completion surveysmake it unlikely if not impossible for EIA to do the job itself,Shambaugh said.

“We are certainly monitoring the data as we receive it and thenumbers that we generate to try to make sure that if somethingunfortunate does happen then we’ll catch it more quickly than thispast problem,” said Trapmann. “This is an unfortunate thing andkind of embarrassing, but it had to be taken care of. We’re hopingwe’re on top of it now.”

©Copyright 1998 Intelligence Press Inc. All rights reserved. Thepreceding news report may not be republished or redistributed, inwhole or in part, in any form, without prior written consent ofIntelligence Press,Inc.