Upload
austyn-couch
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
A Bidding Modelfor Auctions of Offshore Alternative Energy Sites
Prepared for the Eastern Economic AssociationNew York, March 1, 2009
by
Radford Schantz, US Department of the Interior([email protected])
and
Walter Stromquist, Swarthmore College([email protected])
Background
• Energy production sites on public land are auctioned under a variety of authorities and using various auction formats. The focus in this paper is on:
– oil and gas leases, for which (partly) we have an auction model, and,
– alternative energy site leases, where the future auction format is a subject of discussion and planning.
Oil and Gas Lease Auctions
• Offshore: first-price sealed-bid…– … must be auctioned; there is no process of noncompetitive award.– The main authority is the Outer Continenal Shelf Lands Act (1953), 43
USC 1331.
• Onshore: second-price open ascending…– … auctioned when there is competitive interest. Absent that interest,
an onshore lease can be awarded noncompetitively.– The main authority is the Federal Onshore Oil and Gas Leasing
Reform Act (1987).
Alternative energy site auctions
Onshore…• BLM administers wind energy rights-of-way (ROW). The
main authority is the Federal Land Policy and Management Act (1976), 43 USC 1701.
• Three basic types of ROW: energy testing; site testing; and commercial development.
• Sometimes first-come, first-serve noncompetitive process…. competitive process if plan requires it or if more than one applicant shows substantive interest.
Offshore alternative energy
• Section 388 of the Energy Policy Act of 2005 amended the OCS Lands Act and gave Interior authority to grant leases for alternative energy activities.
• “COMPETITIVE OR NONCOMPETITIVE BASIS- …the Secretary shall issue a lease, easement, or right-of- way on a competitive basis unless the Secretary determines after public notice of a proposed lease, easement, or right-of-way that there is no competitive interest.”
First offshore leases planned
Measurement leases: for five years or more, but they cannot be converted into commercial leases. Forty nominations were received…16 areas that we consider top priority.
State Energy Process
New Jersey 6 wind Noncompetitive
Delaware 1 wind Noncompetitive
Georgia 3 wind Noncompetitive
Florida 4 current 3 competitive, 1 noncompetitive
California 2 wave 1 competitive, 1 noncompetitive
Auction models needed
• Offshore oil and gas: we routinely use a lease sale forecasting model, IMODEL, described next…
• Lack models for onshore energies…
• Seeking to develop auction models for offshore alternative energy…
➜ to help to determine the best auction format(s) for this new program
The Bidding Model
• First Price, Sealed Bid Auctions• Common-value model• Public and private information• Symmetric equilibrium strategies• Special features:
- Number of potential bidders is a random variableMay be determined inside the model, from an equilibrium condition
- Role for uninformed bidders In special cases, may determine effective minimum bid for informed
bidders- Flexible private information model
Additive errors, but they apply to a transformed version of value
Rules of the Auction
• Seller announcesMinimum bid, bmin
Bidding fee (ignore the fee for today!)• Anyone can bid• Highest bidder takes the property, pays amount of highest bid.• Bidders do not know how many other bidders there are.
• TODAY’S EXAMPLE:Minimum bid $128 thousandNo bidding fee
The Common Value
• Value of property = V, same for all bidders.• Model V as a random variable• We use an intermediate random variable:
U = “value parameter” V is related to U:
V = v(U)(increasing, but not usually linear)
• Why the extra level of complication?We want to use additive errors, but it isn’t plausible that estimating errors are independent of the size of V. But errors might be independent of some transformed version of V, and that is U.
The Common Value:Public Information
• All public information is represented by a probability distribution on U.(the “prior distribution”).
• Represent the prior by a cumulative distribution function (cdf):
H (u) = Pr ( U ≤ u )
• This determines a prior distribution on V itself.• H may be continuous, discrete, or mixed.• Bidders effectively use H as a prior distribution, and update their
estimates of value based on their private signals and othercircumstances
• H and the value function v(U) are common knowledge.
The common value:Public information
• TODAY’S EXAMPLE:
With probability 0.20, the property is “successful”U = + 1/2V = + $ 1 million
With probability 0.80, the property is “a failure”U = – 1/2 V = – $ 90 thousand
Prior mean value of the tract: $128 thousand
(same as the minimum bid!) H(u) in the
example
How many bidders?
• JARGON ALERT:We call the potential bidders “EVALUATORS.” If they get private signals, they are “INFORMED EVALUATORS.”Otherwise, they are “UNINFORMED EVALUATORS.”
(It seems wrong to call them all “bidders,” since they might choosenot to bid.)
• The number of INFORMED EVALUATORS is a random variable, N.
• TODAY’S EXAMPLE: N is Poisson, with mean m = 3.q(n) = Pr (N = n) = (3n/n!) exp(-3) for n = 0, 1, 2…
How many bidders?
• In the model, N is always Poisson. The user can specify the mean, m.
• OR: The user might let the model determine m internally.
User specifies an “information cost,” kAll evaluators may decide whether to “buy” a private signal
(and so become informed evaluators)At equilibrium, they each choose to do so with some common
probability p, chosen so that the expected profit for informed evaluators is exactly k.
The consequence: N is Poisson, with mean m determined by k.WE’RE NOT DOING THIS TODAY.
Uninformed evaluators
• Do uninformed evaluators matter?
Not usually.
Sometimes, yes. Suppose that the lowest possible value of V is$200 thousand, but the minimum bid is $128 thousand.
Then uninformed evaluators can rationally bid $200 thousand,
and somebody surely will do so.
This “uninformed bid” replaces the legal minimum as thestarting point for the informed bidders.
The uninformed evaluators have no effect in today’s example.
PRIVATE SIGNALS
• Each informed evaluator receives private information in the form of a PRIVATE SIGNAL. The i-th evaluator’s signal is Xi. We model each signal as a random variable.
• ADDITIVE ERRORS: Xi = U + Ri
Now Ri is an ESTIMATING ERROR. The evaluator knows its own Xi, but not Ri or the other evaluators’ signals.
• We assume that estimating errors are INDEPENDENT and NORMALLY DISTRIBUTED. (They are independent of U, V, N, and each other.)
• The standard deviation of the private signals is the same for all evaluators. It is a model parameter, . TODAY WE’LL USE = 2.
PRIVATE SIGNALS (example)
• Left curve: distribution of private signals if U = - 1/2• Right curve: distribution of private signals if U = +1/2
• Evaluators update their “success probabilities” based on their private signals (and other circumstances).
5 5 10
0 .05
0 .10
0 .15
0 .20
Summary of the inputs
• Inputs to the model:
Bmin and fee Rules of the auctionv ( ) Defines value, V = v(U)H ( ) Public (prior) distribution for Um Mean number of informed evaluators
(or specify information cost, k) standard deviation for estimating errors
• All of these are entered by the user, and are assumed to be common knowledge.
Analyzing the Game
• This is a non-cooperative game.
• Let’s find a symmetric equilibrium strategy:
x* = smallest private signal that justifies a bid g(x) = optimal bid by an evaluator with signal x
(assume that g(x) is increasing, and defined when x x*.)
• This model does not look for asymmetric equilibriums.
• This is a hard calculation! But not that hard. Standard theory applies.
The symmetric equilibrium strategy
• IN TODAY’s EXAMPLE:
x* = 1.67814(…so it takes an encouraging signal to justify a bid,but not an exceptionally rare signal.)
Here is g:
1 2 3 4 5 6
120 000
140 000
160 000
180 000
200 000
220 000
240 000
How does bidding depend on m?
1 2 3 4 5 6
50 000
100 000
150 000
200 000
250 000
Equilibrium bidding strategy for m = 1, 2, 3, 4(m = mean number of informed evaluators)
More competition -> Be more reluctant to bid at all But bid higher if you have a very favorable signal.
Distribution function for the High Bid
Probability of no bid at all: 0.597
Probability of bid above $ 250 thousand: 0.026
140 000 160 000 180 000 200 000 220 000 240 000
0 .2
0 .4
0 .6
0 .8
1 .0
Outputs from the model (in the example)• Expected number of bidders:
Each evaluator bids if its private signal is at least x*.
If the tract is a success, this requires x* = 1.67814so estimation error is 1.17824which happens with probability 0.278.
If the tract is a failure, it happens with probability 0.138.
So, with three informed evaluators, the expected number of bids is(3)(0.278) = 0.834 (if success)(3)(0.138) = 0.414 (if failure) or 0.498 (averaged over cases).
In this example, we should expect to get about ½ of a bid (on average).
More outputs from the model
• Expected number of bids• Probability of getting at least one bid• Expected value of highest bid, if there is a bid• Expected revenue to seller
• The model can also give the distribution of these variables.