been attempting to increase the carrying capacity of their land by a variety of methods, including
the clearing of trees and the introduction of non- native grass species. Initially these developments
were limited to patches of more fertile soils. The region now has one of the highest clearing rates in
Australia, with between 4 and 8 of many broad country types being cleared between 1992 and
1995 McCosker and Cox, 1996.
Landholders must gain permission to clear trees from the Queensland Government through the
Department of Natural Resources. In issuing the permits for broadscale tree clearing, the State
Government policy calls for a balance between the benefits of increased productivity most of
which accrue directly to the landholders against the environmental costs of diminished vegetation
cover which are more broadly spread across the regional and national communities. The estima-
tion of these environmental costs is the focus of the CM application described below.
The Queensland Government has recently been revising its tree clearing policies, with the result
that some vegetation communities are now pro- tected. Other vegetation communities can be
cleared to 20 of their original extent on individ- ual properties. The CM exercise is aimed at as-
sessing the environmental costs associated with alternate guidelines.
The use of labels reflecting tree retention levels is thus the most policy relevant way of framing
choice options. In the labelled version of the questionnaire employed in this study, respondents
were asked to choose whether they prefer the current retention level of 20, increased retention
of 30, or a further increase to 50.
5
Note that these labels communicate information regarding
changes in protected vegetation area.
6. Questionnaire design
The questionnaire design phase involved exten- sive background research and two rounds of focus
groups with potential respondents.
6
The first round of focus groups focussed on the identifica-
tion of key decision parameters from the respon- dent’s perspective. This information played an
important role in the selection of attributes. Table
Table 1 Attributes, levels and corresponding variables
Attribute Levels in generic
Levels in alt- spec. model
model Option A: 0
Levy on income Option A: 0
tax base
base Option B: 20,
Options B and C: 40, 60, 80
20, 60, 100, 140
Option C: 80, 100, 120, 140
Option A: 0 Option A: 0
Income lost to re- gion in million
Options B and C: Option B: 5, 7,
5, 10, 15 9
Option C: 11, 13, 15
Option A:0 Jobs lost in region
Option A:0 Options B and C:
Option B: 10, 14, 18, 21, 24
10, 15, 20, 30, 40 Option C: 26,
30, 34, 37, 40 Number of endan-
Option A:18 Option A:18
gered species lost to region
Option B: 10, Option B and C: 4,
8, 12, 16 12, 14, 16
Option C: 4, 6, 8, 10.
Reduction in popu- Option A: 80 Option A:80
lation size of non-threatened
species Option B: 60,
Option B and C: 30, 45, 60,
65, 70, 75 75
Option C: 30, 35, 40, 50
6
The first set of focus groups involved two in Brisbane and a further two in Emerald. The second set involved two groups
in Brisbane.
5
Focus group research indicated that a fourth option per- taining to less than 20 retention was neither policy relevant
nor considered a viable option by respondents.
1 lists the attributes chosen.
7
Note that environmental losses receive equal attention in the choice sets and elsewhere in the
questionnaire to losses in jobs and regional in- come. Inclusion of the latter, ‘developmental’ at-
tributes, is consistent with Portney’s observation Portney, 1994 that people can have nonuse values
for economic and social factors. The balanced environment – economy approach to information
presentation addresses Blamey’s 1996, p. 128 concern that it is often futile to omit or downplay
references to such outcomes ‘in the hope that respondents will not bring their own perceptions of
such factors in as external variables’.
Citizens have general stances on environmental issues that they bring to the valuation situation, and
implicit in these stances is a consideration of the relative importance of environmental factors and
developmental factors. Omitting or downplaying the development side of the story not only leads
some respondents to perceive the questionnaire to be biased, but also results in the elicitation of a
blurred construct. The objective of environmental valuation is to estimate environmental consumer
surplus. An alternative to seeking surplus validity is to seek a form of predictive validity pertaining
to the maximum amount of money citizens are prepared to commit themselves to paying at a
referendum electoral validity. Developmental considerations and payment vehicle protests are
invalid from the former perspective but valid from the latter. Unbalanced information presentation
results in a blurred compromise between these two objectives. In this paper, we have attempted to
maximise predictive WTP validity and to interpret WTP for environmental attributes within this light.
To the extent that losses in regional income and employment are expected to be short-lived, lower
impacts, or zero impacts, can be substituted into the model as an approximation when estimating WTP
and market share. Other stated preference studies that have included developmental benefits include
Lockwood et al. 1994, Morrison et al. 1999.
The second round of focus groups focussed on the refinement of draft questionnaires. Some of the
more important issues that emerged concerned the clarity of information; selection of photographic
stimuli; cognitive burden particularly as it relates to the number of choice sets; perceived bias of
information presented; strategies for choice; and plausibility of attribute-combinations. Particular
attention was also given to whether individuals interpreted information and questions in the way
intended by the researchers. Different ways of introducing and explaining the choice modelling
task were also explored.
Both generic and alternative-specific choice set configurations were trialed, and in some cases, in
the same group.
8
Whilst some participants thought alternative-specific labels were a good idea, others
were not so sure: I think this puts a whole different spin on
things....I think it makes the decision more realistic. E1
If you have that [label] on every single one, then people might just look at that… They
might just go ‘I’ll pick the one with the most trees’. B2
A common tendency was to select the option with the label that most closely coincided with one’s
environmental attitudes, and to see if significant reasons existed not to choose this alternative. The
following statements illustrate how some individu- als used the labels to help structure their
evaluations:
Straight away I thought yeah I do want to do something to increase it [the minimum permissi-
ble tree retention] and I don’t mind paying some money. B3
8
The particular group to which the verbatim quotes listed below correspond is indicated at the end of the citation for
each quote, using the abbreviations B1 – 4 and E1 or E2. B2 thus refers to the second focus group conducted in Brisbane.
7
Inclusion of an attribute pertaining to the percentage of tree retention did not appear to be necessary: participants
appeared to be more focussed on final outcomes such as the number or percentage of species and jobs lost or saved. This
suggests that the omission of an ‘area’ attribute is unlikely to result in significant omitted variable bias. More likely, inclu-
sion of the above-mentioned policy names in the second version of the questionnaire may prompt considerations that
would not otherwise occur.
Putting the top line [labels] up there makes it a bit easier. B3
The final questionnaires were presented in the form of a small booklet with a colour cover,
and a colour insert containing an attribute glos- sary for use when completing the choice sets. A
map on the cover indicated the location of the Desert Uplands in Queensland, and proximity to
nearby towns. A graphic artist finalised the pre- sentation of the questionnaire and pamphlet.
The levels assigned to the attributes listed in Table 1 were chosen such that the resultant at-
tribute-space encompassed the vast majority of policy-relevant tree clearing options. Information
regarding the ecological effects of different tree clearing options in the Desert Uplands, and
consequent
implications for
humans, is
ex- tremely
limited. Information
that could
be gained regarding likely outcomes for jobs, re-
gional income and threatened and non-threat- ened species was obtained from a variety of
sources summarised in Rolfe et al., 1997 and in consultation with experts. The high level of
uncertainty regarding attributes such as impact on endangered species meant that the range of
levels chosen was wider than one would expect to be the case with most policy options.
Because most
respondents would
expect greater tree clearing to involve worse environ-
mental outcomes and better economic outcomes, attribute values in the labelled treatment were
selected from alternative-specific sets of values. In other words, different policy labels were asso-
ciated with different sets of outcomes. Table 1 lists the attribute levels applied to each option
in each treatment and Fig. 1 illustrates the main differences between the two treatments.
9
The ca- pacity to incorporate associations between at-
tribute values and labels within the design of the experiment can be considered an advantage of
the alternative-specific approach to choice set design. Implausible combinations of attributes
and labels are minimised as a result.
10
To ensure that the attributes varied indepen- dently of one another, such that their individual
effect on respondents’ preferences can be iso- lated, an orthogonal experimental design was
used to assign attribute levels to alternatives. Fractional factorial designs were used to reduce
the number of alternatives to a manageable level. Choice sets were constructed in such a
way that orthogonality both between and within alternatives was ensured. To reduce implausibil-
ity problems whilst at the same time increasing the balance between environmental and eco-
nomic variables, a correlation between jobs lost and income lost was introduced by creating a
composite 8 level attribute. Sixty-four choice sets were allocated to eight blocks of eight
choice sets in each of the two versions, produc- ing a total of 16 versions of the questionnaire.
An eight-level, orthogonal blocking variable was included as part of the experimental design, and
used to generate the eight different blocks or versions of the survey. The purpose of the
blocking factor is to insure that the eight blocks feature a balanced distribution of levels across
all
attributes, which
ordinarily cannot
be achieved using random assignment.
In both versions of the questionnaire, state- ments were included in the scenario with the
purpose of further diffusing perceptions of im- plausible attribute combinations that may give
rise to problematic response strategies. Respon- dents were told to ‘consider carefully the impli-
cations of each tree-clearing option by looking at the numbers in the table. To keep matters
simple, we do not describe how each option would work. Some implications which may seem
a little odd are in fact quite possible…You will find some questions easier than others.’
10
Whether or not alternative-specific attribute levels are required clearly depends to a large extent on the nature of the
information communicated by the labels. For example, labels based on biogeographic differences such as ‘The Desert Up-
lands’ and ‘The Brigalow’ may not require alternative-specific attribute levels if respondents do not have strong a priori
expectations regarding the relative magnitude of attribute lev- els for these regions.
9
Every attempt was made to keep the two main versions of the questionnaire as similar as possible. Minor wording differ-
ences were, however, required in two paragraphs prior to the choice sets in order to bring the scenarios in line with them.
Fig. 1. a An example of a generic choice set. b An example of an alternative-specific choice set.
7. Survey logistics