Samples
Starting Point Bias in Bidding Games
TABLE 4. Final Bid Outcomes for Different Starting Values: Effect of Excluding Observations
Where Starting Bid Equals Final Bid.
|
Question Format |
Starting Bid |
Number of |
Revised Mean Bida |
|
Willingness to Pay Daily Fuel Tax |
$1 |
16 |
$29 (7) |
|
$800 |
1 |
$39(14) | |
|
Willingness to Accept Payment Not to Fish For Day |
$1 |
8 |
$315(12) |
|
$800 |
15 |
$169(4) | |
|
Willingness to Pay Annual User Fee |
$1 |
19 |
$800 (1) |
|
$8,000 |
1 |
$333 (6) | |
|
Willingness to Accept Payment Not to Fish For Year |
$1 |
2 |
$5,867 (15) |
|
$8,000 |
14 |
$1,783 (3) |
a Number of observations used to calculate revised means given in parentheses.
still alter observed mean final bids in a
statistically significant manner.
All these effects were observed in this
study which incorporated a streamlined
bidding procedure to minimize potential
exasperation effects. This finding suggests
that the information transfer effect of se-
lecting a certain bidding starting point
may be more important than the exasper-
ation effect in influencing final bid out-
comes.
The relationship between starting bid
and refusal rate is an especially complex
issue that remains unresolved. Looking at
Table 1, it is clear that relatively high re-
fusal rates were associated with WTP
questions when large initial starting values
were used. If individuals who refused to
play the bidding game were also those
with low actual valuations of the recrea-
tion experience, then the number of re-
fusals would be inversely related to esti-
mated final mean bid outcomes. Given
that this appears to be the case here, it
can be hypothesized that high starting
values might have somehow served as a
behavioral cue to respondents that influ-
enced their willingness to participate in
the game. This is an acute empirical con-
cern when choice of a starting value
screens out a certain class of respondents
whose valuations are therefore unrepre-
sented in final mean bid calculations.
Conceivably, low starting bids may screen
out high-value users. High initial bids may
discourage low-value respondents. While
this phenomenon is not tied directly to the
information transfer effect, it may never-
theless lead to biased value estimates.
Taken together, the results presented
here suggest that the information transfer
effect works in subtle, yet influential ways.
The effect can perhaps be ameliorated by
obtaining a reasonably accurate estimate
of the final mean bid amount before con-
ducting final field interviews using itera-
tive bidding games. A simple pretesting
procedure using an open-ended question
format could probably suffice to identify
confidence intervals that encompass most
users’ true valuations. The closer the ini-
tial bid is to the final bid outcomes, the
less likely that starting point values will
bias final estimated values because of re-
spondents’ refusal to participate. Pretest-
ing does not, however, overcome the
problem of overly cooperative respon-
dents who accept any initial bid as their
final bid. Finally, in view of the seemingly
positive relationship between final bids and
starting values, use of a range of starting
values within a single survey should prob-
ably be encouraged. In this way, biases
associated with high and low starting
points can perhaps be internally compen-
sated for.
39