• Problem and Context• Survey Tool• What the Future May Bring: Model Estimation
Empirically Approaching Destination Choice Set Formation A. Horni, IVT, ETH Zürich
Robustness of Estimated Parameters
Thesis Schuessler (2010):Route choice
Pellegrini et al. (1997): Shopping destination choice
Problem for operational model
The Deterministic Approach
robserved
rt
csreal(t) if any!threshold rt = f(rt)
cs formation criteria (exogenous)
To date: specification of exogenous factors for destination choice set formation rather ad hoc and more like a proof of concept.
The Probabilistic Approach
cs formation criteria
Speed-ups e.g. → convergence to deterministic approch
endogenous!
But: combinatorial complexity
Conclusion in the words of Pagliara and Timmermans (2010):
„Even though the inclusion of latent stochastic thresholds and the simultaneous estimation of thresholds and utility functions represents an important step forward in discrete choice analysis, forecasting results still depend on the researchers’ specification of the choice set.“
Decision Horizon: e.g., Grocery Shopping
meat for dinner
vegetables for dinner
dinner for cat
→ relevant choice between
and
… and not in cs immediately prior to choice!
choice set immediately prior to choice
context!
Decision Horizon – Generation of PS
Habitual „decisions“/ Routine response behavior
preferred set of stores → relevant for transport planning
extensive decisions
impulsivedecisions
non-compensatory decision behavior→ rule-based
learning process
e.g. grocery shopping
Decision Horizon: Sets Involved in the Decision Process (a First Step)
Unawareness set
Awareness set= cs(t –t)
Inept set (-)(Inert set (0))
cs(t)
Narayana and Markin 1975
Evoked set (+) (Inert set (0))
Purely Statistical Approach vs. Behavior-Based Approach
Homo oeconomicus → universal choice set
Thresholds where parameters stabilize
Lacking research
Computationally infeasibleNot explicative
inconsistentproductive?
Behavior-based criteria for cs formation
Lacking research
Allora, …
EmpiricalMethodologial
• Decison horizon• Statistical vs. behavioral model
• Preferred set- characteristics- frequencies
• Sets involved in decision process• Core area within STP• Reasons for NOT visiting a store• Trip chaining
Model estimation
MATSim model
Providing a research (survey) tool
Survey „Tool“
• Web-based • Google street view
• Grocery shopping• 300 stores, partly manually collected• future: attributes of stores
Model Estimation
Observed choice
Awareness set?
Preferred set
Choice set
Requirements:1. Easy to survey and generate in op. models
2. Actually plays a well defined role in decision process
„New“ model
Concluding Remarks
Empirical basisTTB for time-geography
Input to discussion on decision horizon and extent of behavioral basis of discrete choice models (vs. purely statistical)
Survey tool
Pretest
- Game-like traits appreciated → less fatigue
- Dominance of closest Coop or Migros (not deliberated)