households in a sample of size N. The exponential functional form guarantees the predicted WTP will be
non-negative. Given expression (3), the probability of purchasing the pass is represented as
Pi = Pr(yt = 1) = Pr(WTPi ≥ ln($b.)) = Φ(-1(Xβ - ln($bɔ)), (4)
σ
where Φ is the standard normal cumulative density function, yi = 1 if the household purchased the pass,
yi = 0 if the household did not purchase the pass, and bi is the proposed price of the NRP. The initial bids
are chosen at random from the following bid vector:
b= ($25, 45, 65, 85, 105, 125, 145, 165). (5)
Table 3 shows the percent of the RDD and NPF samples that respond “YES” to the initial bids. As
expected, the percent responding “YES” generally declines as the bids increase in value.
We selected the range of bids based on focus groups and comparisons to other recreation pass
programs. As described in Section 4, we then ask a follow-up bid which is randomly selected from either
bH = (b+$20, 2*b-$5) if the respondent accepts the bid or bL = (b-$20, 0.5*(b+$5)) if the respondent
declines.4 If a respondent answers “NO” to both bids, we ask a follow-up question with a bid equal to
zero. The relevant probabilities of purchasing a pass within each possible range of prices are
P1 = Pr(-∞ < WTP1 < 0) = Φ(l(X'β)) (6.1)
σ
Pi2 = Pr(0 < WTP1 < bL ) = Φ(-1-(Xβ - ln(bL ))) - P1i (6.2)
σ
P13 = Pr(bL < WTPi < bl ) = Φ(!(X'β- ln(bl )))-Φ(i(X'β- ln(bL ))) (6.3)
σσ
4 In other words, if the initial bid price is accepted, the second price is (randomly) either $20 higher, or else (with
equal probability) it is approximately double the initial bid amount. If the initial price is declined, the second price
is either $20 lower, or else it is approximately half the initial amount. Adjustments of $5 (when doubling or halving
the initial amount) ensure that all bids are multiples of 5, to avoid any confounding that might arise if multiples of
10 have a different psychological resonance with respondents.
11