when, if ever, WTA measures should be used in CVM studies.
Background
As Adamowicz, Bhardwaj, and Macnab (1993) note, WTP and WTA for the same good
can (and generally are) vastly different. Others have noted that the WTA measure is subject to a
much higher degree of inaccuracy as compared to WTP measures (see e.g. Goldar and Misra,
2001; Coursey, Hovis, and Schulze, 1987; Brookshire and Coursey, 1987; Bishop and Heberlein,
1990). A recent review of 45 studies which contrasted WTA and WTP by Horowitz and
McConnell (2000) found that WTA was often an order of magnitude or more greater than WTP.
One study found that WTA was nearly 3,000 times WTP; on average (dropping this outlier), the
ratio WTA/WTP was about seven. Due to the volatility of these estimates, WTP has generally
been the method of choice in CVM studies; for these and other reasons, the NOAA panel (Arrow
et al. 1993) reviewing CVM strongly recommended against the use of the WTA measure.
Why the disparity between the measures? Strict theory based on compensating and
equivalent variation measures notes that the income effect will cause some difference, though this
should be minimal (see Willig’s seminal 1976 paper). Hanemann (1991: 635) has theorized that
a lack of substitutes for the good in question can lead to disparities between measures, noting
that “the difference between WTP and WTA depends not only on an income effect but also on a
substitution effect.” Empirical support for this position has been mixed (see e.g. Adamowicz,
Bhardwaj, and Macnab, 1993; Shogren et al. 1994; Mantymaa, 1996).
A second suspected source of the difference is survey construction, interviewing
technique, and other issues related to data collection (e.g. Brookshire et al. 1982; Dwyer and
Bowes, 1978). However, the empirical evidence does not strongly support this reasoning
(Horowitz and McConnell, 2000).