Samples
Starting Point Bias in Bidding Games
bids may do more than just influence final
observed bids. It may also lead to outright
rejection of the hypothetical market sce-
nario by respondents. Another factor in-
fluencing refusal rates was question for-
mat. Questions aimed at determining
annual WTP and WTAC consistently
drew more refusals to answer than ques-
tions concerned with marginal (daily)
WTP or WTAC. Perhaps this is due to the
fact that respondents perceived it less risky
to incorrectly value a day of fishing com-
pared to incorrectly valuing an entire year
of fishing. If so, respondents would tend
to be relatively less interested in playing
the annual WTP or WTAC bidding
games. Furthermore, the perceived real-
ism of valuing a year of fishing is probably
less than trying to value a fishing day.
Statistical results of four bidding game
outcomes are given in Table 2. Mean final
bids were observed to be considerably
higher for WTAC question formats when
compared to WTP responses. This is con-
sistent with findings reported elsewhere by
Rowe et al., Bishop and Heberlein, and
others. As might be expected, mean bids
observed for both WTP and WTAC ques-
tions were lower for Subsample I (daily
use values) than for Subsample II (annual
use values).
To measure the effect of starting bid
values on final bidding outcomes, four
pairwise t-tests were conducted to detect
statistically significant differences be-
tween mean final bids for alternative
starting values. Test results are given in
Table 2. For all four question formats, the
hypothesis was rejected at the 90% confi-
dence level that the mean final bids were
the same regardless of starting bid value.
Significantly higher mean bids were con-
sistently observed for higher initial bid
values. Relative differences in mean bid
outcomes were most perceptible in the
WTP question formats. For these question
formats, mean final bids were 10 to 34
times greater for the higher starting bids.
On the other hand, mean final bids ob-
served for the WTAC questions differed
by a factor of 1.3 to 3.5.
The statistical relationship between ob-
served final bids and starting bid values
was also examined by estimating bid
curves for alternative question formats.
Four separate bid curves were estimated
using ordinary least squares regression
(Table 3). Inclusion of explanatory vari-
ables such as family income, age and fish-
ing frequency permitted a determination
to be made of the impact of starting bid
selection on final bids, holding other po-
tentially relevant factors constant. Regres-
sion results consistently showed positive
coefficients for the starting bid variables.
In all equations except the estimated bid
curve for a daily fuel tax, the starting bid
variable was significant at the 0.05 level.
While analysis of variance results sug-
gests that a significant positive link exists
between starting bids and final bidding
outcomes, the underlying cause of the
phenomenon is not yet evident. Further
data analysis was conducted to investigate
the information transfer effect, and how
it may bias observed bids. Both Thayer
and Rowe et al. have argued that initial
bids may influence respondents’ percep-
tion of legitimate final bids. An extreme
of this situation is where respondents ac-
cept the starting bid as their final bid. The
probability seems low that a randomly se-
lected respondent would have a true val-
uation exactly equal to the starting bid.
Therefore, when a respondent’s stated bid
equals the starting bid, it would seem that
either the respondent is excessively coop-
erative, or the information transfer effect
is profoundly strong.
For all question formats, a count was
made of the bids for which the final bid
equaled the starting bid. These bids were
then removed and the mean bids reesti-
mated for all question formats and start-
ing values. The results showed that out of
a total of 200 bidding games played, 76
games (38%) resulted in final bids equal-
ling starting bids (Table 4). This outcome
37