Respondents

1

• Respondents use of information in choice
  modelling surveys

Choice Modelling Workshop Brisbane, 1-2 May 2008
Bill Kaye-Blakea, Walt Abellb, and Eva
Zellmana aAgribusiness and Economics Research
Unit bApplied Computing Group
2
Decision-making

• How people use information
• Theory Simon, others
• Evidence
• Logit models derived from neoclassical paradigm
• Key question does it make a difference?

3
Early attempt

• Paper survey with GM and non-GM alternatives
• Analysed choices
• MNL, etc. (McFadden)
• Boundedly rational model (Simon)
• Results not brilliant
• Needed different data

4
Latest attempt

• Started from Hensher Rose (debrief questions)
  and earlier BR work
• Teamed up with computer programmer and
  psychologist
• Computerised survey to capture information access
• Follow-up questions

5
Choice set display
6
Use of information

Unopened Opened Percentage unopened Group
Texture 770 1840 29.5 a
Price 368 2242 14.1 b
Colour 815 1795 31.2 a
Production syst. 451 2159 17.3 c
Nutrition 434 2176 16.6 c
Country of origin 614 1996 23.5 d
Similar within groups and statistically different between groups. Determined by chi-square tests, and are significant at the 0.01 level. Similar within groups and statistically different between groups. Determined by chi-square tests, and are significant at the 0.01 level. Similar within groups and statistically different between groups. Determined by chi-square tests, and are significant at the 0.01 level. Similar within groups and statistically different between groups. Determined by chi-square tests, and are significant at the 0.01 level. Similar within groups and statistically different between groups. Determined by chi-square tests, and are significant at the 0.01 level.
7
(No Transcript)
8
Models

• Two models
• Assumed full information
• Only accessed information
• Mixed logit
• Random parameters based on significance
• Panel models
• For unused information coded in Limdep as -888

9
ML Assumed full information
Parameters Parameters Derived St dev Derived St dev
Value St. err. Value St. err.
Floury -0.348 0.144 0.869 0.160
Organic 0.349 0.178 0.972 0.186
GM -2.46 0.331 2.14 0.339
Omega-3 0.898 0.217 1.23 0.220
Low-GI 0.383 0.191 0.814 0.228
Australia -0.613 0.172 0.670 0.253
China -1.43 0.219 1.16 0.255
Price -1.37 0.144
Pink -0.000 0.000
Yellow 0.245 0.126
10
Partworths Full information
Attribute Partworth (NZ/kg) Standard deviation
Floury -0.252 0.373
Organic 0.262 0.419
GM -1.84 0.923
Omega-3 0.666 0.526
Low-GI 0.276 0.349
Australia -0.447 0.291
China -1.05 0.497
Price -1.00 0.0873
Pink -2.75E-05 9.98E-05
Yellow 0.179 0.0545
11
ML Accessed information
Parameters Parameters Derived St dev Derived St dev
Value St. err. Value St. err.
Floury 2.52E-03 1.00E-03 3.07E-03 1.37E-03
Organic 0.677 0.0684 9.06E-04 2.61E-03
GM -0.676 0.0684 7.32E-04 3.21E-03
Omega-3 0.0918 0.0592 8.82E-04 1.51E-03
Low-GI -0.0886 0.0591 6.26E-04 1.64E-03
Australia 0.307 0.0770 1.13E-05 1.61E-03
China -0.305 0.0770 2.81E-05 1.61E-03
Price 2.38E-03 6.42E-04
Pink -0.203 0.0573
Yellow 0.206 0.0573
12
Findings

• All available information not used
• Controlling for information use changes
  parameters
• Australia ?
• Price ? 0
• Unexplained heterogeneity largely disappears