Ed
Capen, AAPG member and co-author of the paper that coined the term
"winner's curse," remembers well the birth of Atlantic Refining
Co.'s interest in bidding science.
It was
in the early 1960s. In the Philadelphia and Dallas offices of Atlantic
Refining there was dread spreading. They had got just what they
wanted — they thought.
Atlantic,
which had been active in the Gulf of Mexico since the mid-1950s,
bought virtually everything it had bid on at a Gulf lease sale.
But their bidding "success" put the company in a budgetary bind
for a couple of years. They had overbid.
"Management
did not want to repeat that experience," Capen recalled, "and asked
R&D to see if there were some operations research-type solution
available."
The task
fell to R&D's Bill Campbell, who was applying management science
methodology to exploration decisions. Lawrence Friedman had recently
got the first doctorate in Operations Research using bidding as
his topic. Campbell used this work as his starting place.
Another
Gulf lease sale came along soon — giving R&D a chance to apply
Friedman's theories.
Friedman's
work turned out to be in error, but it did suggest bidding less
than one's value estimate.
"In our
case the rest of the company was also working on the problem in
ignorance of other similar efforts," Capen said. "Everyone knew
that we had to reduce our bids. The geologists lowered their estimates
of success ratio. Geophysicists decreased the size of the targets.
I don't remember, but the bookkeepers may have raised the discount
rate. And, of course, Campbell had already hedged by 30 percent
or so.
"By the
time everyone was through subtracting value, our bids were so low
we didn't buy anything," he said.
After correcting
the error in Friedman's work, Campbell and Bob Clapp, an MIT Sloan
School graduate who joined Campbell about that time, designed a
20,000-trial Monte Carlo model of the bidding process. It worked.
R&D began to "sell" its bidding strategy to operating management.
Reasonable
Projections
Capen had
joined Atlantic's R&D in 1957 as a research geophysicist. A
few years after his arrival at Atlantic he sensed that the repeated
trials and high uncertainty so characteristic of petroleum exploration
made statistics an under-appreciated but very powerful tool.
Capen returned
to school (Southern Methodist University) part time to study statistics.
In 1965 Capen took over Campbell's work, and with Clapp continued
the bidding effort.
To basically
describe Arco's bid strategy, a value is placed on a lease based
on the promise of the geology. Companies should then bid about 30
percent of their value estimate (depending on competition and uncertainty).
Bids are then handed to the government on sale day. On a given tract,
some company value estimates will be too high and others too low.
The winner will tend to be the one who estimated too high. Bidding
only 30 percent of value protects a company that wins
because of an inflated value estimate.
Many managers
do not like this strategy because the reduced bids also lower the
chance of winning. Emotionalism usually carries the day. Managers
would rather buy leases than maximize profit. Often it will be five
to 10 years before the results are known so that the guilty, having
been promoted and/or transferred, are never punished, Capen said.
In 1967
Capen was transferred to operations staff with the principal task
of implementing the new bidding strategy for all Outer Continental
Shelf lease sales. Clapp rejoined Capen a year or so later.
The overbid
difficulty was not confined to Atlantic. Virtually every company
placing the bids was aware of the pitfalls — the industry's OCS
rate of return was miserable.
Atlantic
was careful in releasing information about its system. Over time,
however, Capen was allowed to give oral presentations and finally
a published paper. The lawyers and some in management were reluctant
to share the company's strategy. R&D players saw the obvious
advantages of telling the whole world. The R&D VP recognized
the ploy as "legalized collusion."
"If everyone
lowered their bids to protect from the curse, the entire industry
would be better off," he said. "The sellers would suffer a reduction
in bonuses. The sellers, however, were doing far better than they
should have."
Capen recalled
that when he made an oral presentation to an overflow audience at
the 1970 SPE annual meeting, "I invited the audience to guess my
weight, and then by polling the 600 people I was able to show clearly
that some overestimated and more underestimated. The demonstration
worked perfectly.
"We used
the bidding strategy from 1967 (Atlantic merged with Richfield in
1966) until the merger with BP 30-plus years later precisely to
generate bids," he said. "The work is quantitative, not qualitative.
Arco management was pretty faithful in using the recommendations
made by either Bob Clapp or myself.
"However,
it is Campbell who deserves much of the credit because of his original
assessment that the big leverage in economic decisions would be
in exploration," Capen said. "He was poised to take advantage of
that fact when the bidding opportunity came along. I recall just
one sale where using our bidding strategy saved the company about
$300 million.
"The only
change we made over the next 35 years was a move to numerical integration
rather than Monte Carlo," he continued. "I had reduced the problem
to an expected value integral for the publication in 1971, but we
did not have the computer power for numerical integration. Monte
Carlo was easier, and with enough trials (it) worked very well.
"We had
decided that the lognormal distribution was the proper one to use
in the analyses," Capen said. "We observed it in the bidding, and
we knew the Central Limit Theorem would lead to lognormal. We were
very comfortable with these assumptions. The proof was that we were
able to make reasonable projections of how many leases we would
buy and how much we would spend."
To Coin
a Phrase
It was
in 1971 that Capen, Clapp and Campbell (all began as physicists)
authored the paper "Competitive Bidding in High Risk Situations"
in the Journal of Petroleum Technology. It was that article that
the term "winner's curse" first appeared.
The practical
approach has been applied to situations beyond the oil industry
— including mergers and bidding for baseball players in the free
agent market (see The Winner's Curse by Richard Thaler, Free Press,
1991).
Capen's
work has been published widely in SPE literature. AAPG published
The Business of Petroleum Exploration Treatise publication (1992),
where he presented all the statistical formulae involved with estimating
prospect reserves.
Also he
was part of the team instructing AAPG's "Managing and Assessing
Petroleum Risk" education course for 14 years.
Since retiring
in 1992 from his position as Distinguished Management Adviser at
Atlantic Richfield, he has been an active consultant.
In 2001,
AIME-SPE recognized him with its highest award, Honorary Member
— "For Profound and Lasting Contributions in Advancing the Industry's
Understanding of Risk Assessment." He is now mostly retired and
divides his time between homes in Georgetown, Texas and Durango,
Colo.