Common value auctions

1
Common value auctions

• The same value for everyone, but different
  bidders have different information about the
  underlying value

2

• Auction a jar full of coins (asking for an
  estimation)
• average bid will be significantly less than the
  value of the coins (bidders are risk averse)
• winning bid exceeds the value of the jar
• Bazerman Samuelson, 1983, 48 auctions
• True value 8
• Estimated (mean) 5.13
• Bias in estimation risk aversion should work
  against over bidding
• Yet mean winning 10.01

3
THE WINNERS CURSE
4

• Question first raised by Capen, Clapp and
  Campbell 1971.
• This is an example of a problem that comes from a
  field observation before becoming theory.

5

• Winner's curse cannot occur among rational
  bidders (Cox and Isaac 1984).
• Challenge to assumption of rationality.
• But acting rationaly is difficult. Need to
  distinguish between
• expected value of the object, conditioned on
  prior information
• expected value of the object, conditioned on
  winning the auction

6

• Example. You have to advise the takover of firm
  T.
• T knows the true value, you don't
• Assymetry of information.
• Optimal to bid ?
• Cero patatero
• Extreme case of winner's curse.

7

• Experimental evidence (Bazerman Samuelson
  1985) Only 9 bid zero. Majority in 50-75.
• Would learning solve the anomaly? (Weiner,
  Bazerman Carroll 1987).

8

• Each subject (MBA) repeated it 20 times with
  feedback about true value and whether their bid
  was accepted and profit. Of 69 subjects 5 learned
  to bid 1 or less by the end of the 20 rounds.
  Learning seem not to be easy or fast.

9

• Shells boss calls me about a bid
• Calls me again to tell me that the number of
  bidders has increased
• Should I increase or lower my previous bid?

10

• need to bid more aggressively to win
• if you win more likely that you have
  overestimated.
• Solving it not trivial. Do people get it right?

11

• Series of experiments by Kagel, Levin et al. True
  value x varies from trial to trial but always
  between xl and xh.
• Prior is given to each one by drawing xi from
  uniform x?.
• Increase in N -gt more losses.
• Teatments a) change of type of auction (first,
  second), N and ?. Compare results with RNNE.
• This done also with construction firm managers
  (last price) Rules of thumb.

12

• Field data. Oil tracks, Corporate takeovers,
  Publishing auctions. At least prevalence of mild
  form of winner's curse unfulfilled expectations.
• Why is this important? It is part of a general
  problem of after decision blues. Bidders are
  under a cognitive illusion that makes them incur
  in systematic errors.
• What strategy to follow once you have discovered
  the winners curse? Reduce your bids and sell
  short others shares?

13

• Why people succumb to it?
• The value of victory
• Humans assign significant future value to
  victories over humans but not over computer
  opponents, even though such victories may incur
  immediate losses