Transcript Comparing CM and choice behavior experiments [PDF, 253KB]

Application of choice and
choice behaviour to the
community choices among
option for the development of
the Moranbah township
John Rolfe and Galina Ivanova
Outline of this presentation
• Case study: Mining town
• Methods of assessing community
preferences
• Survey outline
• Results
• Pooling CM and CB data
• Discussion
Background:
Issues for Mining Town
• Mining has been through cyclical patterns of
decline and boom
• Increased use of workcamps and ‘fly-in/flyout’
• Town development is limited by high housing
costs and low levels of local spending
• Other key issues are provision of services
(water) and future proximity of mining
• Limited knowledge what are the priorities in
the community regarding town development
Bowen Basin Area map
Moranbah
• Coal mining town in northern Bowen Basin
• Population of approx. 7,000
• Up to 4,000 non-permanent workforce also
‘cycling’ through the town
• Substantial pressure from local council to
locate more workforce in the town
– Council refuses to allow more workcamps to be
built
– Issues about future development of reserves close
to township
Case study:
Moranbah
• Funded by the State Government
• Objectives of the Study
– Assess views of residents regarding development options
• Use 2 techniques: Choice Modelling (CM) and
Contingent Behaviour (CB)
– focuses attention on the key issues or attributes of importance,
– provides some quantitative feedback about the relative
importance of those issues and attributes (tradeoffs).
• CB vs CM
– CB does not use $
– CB provides insights on how behaviour might change
– CB provides more flexibility with choices
Survey
•
•
•
–
–
–
–
–
–
–
A literature review and extended stakeholder
analysis were used to design a CM / CB survey
Choice Modelling survey
3 choice sets /respondent
3 profiles / choice set describing the alternatives on offer
One of the profiles described a status quo option
The other profiles varied
Contingent Behaviour survey
Reference point – intention to stay in Moranbah (years)
3 profiles describing the alternatives on offer
•
Same profiles as from the CM experiment
Respondent identified the length of residence for each
scenario
Figure 1. Example choice set used in survey
Question 2: Carefully consider each of the following three options. Suppose options A,
B and C were the only options available, which would you choose?
Additional
annual costs to
your
household
Housing and
rental prices
Level of water
restrictions
Buffer for mine
impacts close to
town
Potential Condition in 5 years time
Option A
Growth in
population of 5,000
people
(Options A,B and C)
(Expected outcome under current policy pressures)
No change
Some for
households, town
parks and gardens
are drier than now
No change
None for
households, town
parks and gardens
are drier than now
Slight impacts from
noise, vibration and
dust
4,000 in housing,
1,000 in workcamps
None for
households, town
parks and gardens
are greener than
now
Slight impacts from
noise, vibration and
dust
1,000 in housing,
4,000 in workcamps
$0
Moderate impacts
from noise, vibration
and dust
1,000 in housing,
4,000 in workcamps
Option B
$250
($21/month)
Option C
$1,000
($83/month)
25%
increase
I would choose
Performance of the survey
•
November 2006 - Combined phone and mail-out.
–
–
•
•
•
•
•
Both CM and CB contained in mail-out
Split sample used to test order effect
Random sample of households in Moranbah
community
Each respondent completed three choice sets
The response rate was 41% (131).
CM data analysed with logistic regression models
CB data analysed with multiple regression models
Results of the Choice Modelling Experiment
Coefficient
Standard Error
-0.599
0.937
-0.001***
0.000
0.284**
0.119
0.218*
0.114
Buffer for Mine Impacts
0.248**
0.118
Population in Work Camps
1.583**
0.144
Female
1.243***
0.259
Number of Children
0.261***
0.098
Income
0.000**
0.000
Age
0.037**
0.015
Length of residence
-0.100*
0.053
0.212*
0.125
Spending in Moranbah
-0.010**
0.005
Improved services less travel
0.025***
0.007
Constant
Cost
Housing and Rentals
Water Restrictions
Enjoy living in Moranbah
Number of observations
Log likelihood function
R-sqrd
420
-316.4385
.31
*** = significant at the 1% level, ** = significant at the 5% level, * = significant at the 10% level.
Results of the Choice Modelling Experiment
Part worth,
expected
Confidence intervals for Part worth
lower CI
Higher CI
-$582
-$2,691
$1,300
Housing and Rentals
$276
$51
$601
Water Restrictions
$212
-$6
$483
Buffer for Mine Impacts
$241
$12
$541
$1,540
$1,048
$2,636
Constant
Population in Work Camps
Example of Choice Behaviour survey
Question 2. If the scenario below summarised the key changes in Moranbah in the
next five years, would it change how long you think you would live in Moranbah?
Option 1
Housing and rental prices
No change
Level of water restrictions
Some for households, town parks and gardens are drier
Buffer for mine impacts close to town
Slight impacts from noise, vibration and dust
Growth in population of 5,000 people
4,000 in housing, 1,000 in workcamps
Please circle how many years from now you think you will live in Moranbah if this is how it
develops (Remember your answer in Question 1)
less than one year
1 - 2 years
2 - 3 years
3 - 4 years
4 - 5 years
6 - 10 years
10 - 15 years
over 15 years
unsure
1
2
3
4
5
6
7
8
9
Planned stay in Moranbah
30.0%
25.0%
20.0%
15.0%
10.0%
5.0%
0.0%
Less than 1-2 years
1 year
2-3 years
3-4 years
4-5 years 6-10 years
Years
10-15
years
over 15
years
Conducting the CB analysis
• Calculated a dependent variable from the
responses
– Change in intention to stay
– Difference between ‘Planned years of stay’ and
‘Years of stay’ for each profile
– Variable indicates the level of behaviour response
to each profile
• Used as dependent variable in regression
Model 1
Coefficient
Constant
CB Housing and Rentals
CB Water Restrictions
CB Buffer for Mine Impacts
CB Population in workcamps
Model 2
Coefficient
5.845***
10.451***
-0.294
-0.644
0.123
0.029
-0.501*
-0.701*
-1.198***
-1.335***
Female
-1.268*
Younger than Primary School
kids
-0.293
Primary school kids
-0.365
Secondary school kids
0.481*
Age
-0.048
Adjusted R Square
0.045
0.070
Choice Modelling vs. Contingent Behavior
•
Results are similar but not identical
–
–
•
•
Priority of the variables is similar
Relative weighting of variables is more
extreme with CB data
Results suggest that responses may
vary to some extent between CM and
CB formats
Potential reasons
–
–
–
Different formats (closed vs open-ended)
Categorical responses in CB
Brevity of the response format
Testing for order effects
• Tested CM models to identify if coming
before or after CB sets affected values
• No significant effect on CM models
Pooling the data
• Exploratory test to combine data sets
• Converted CB data to binary format
– Assumed that respondents implicitly
considered a status quo option relative to
each CB choice on offer
– Set the status quo option and the choice
alternative
– Stacked CB data with CM dataset
Pooled data exercise
Cost
Change Years of stay
Housing and Rentals
Water Restrictions
Buffer for Mine
Impacts
Population in Work
Camps
Constant – Altern. 1
Constant – Altern. 2
0.254**
0.037
16.734
-0.206^
0.099
Joint model
Coeff.
-0.001***
15.908
0.082
0.092
0.171^
-0.275**
-0.095
1.332***
0.799***
0.382**
-0.482***
-0.979***
0.615***
-0.219^
0.493**
CM model
Coeff.
-0.001***
CM model
Coeff.
•
Discussion
Choice modelling:
–
–
•
Contingent Behaviour:
–
–
•
Promising tool for prioritising development
options
different benefits can be compared to costs of
implementation
Largely supportive of the CM results
Different format to engage people in choice
exercises
Pooled data set
–
Difficult to add value from pooling exercise