35.1 Discrete Choice Models and Continuous Heterogeneity
Discrete and Continuous Heterogeneity
Price Customization
Targeting
Models:
Random Coefficients Logit
Purchase Incidence
Brand Choice Models
(P. K. Chintagunta and Nair 2011)
Goals of demand analysis (which affect model-form)
Forecast: (not much causal inference)
Examples:
Aggregate: (Bass 1969); (Dekimpe and Hanssens 1995)
Individual: (Peter M. Guadagni and Little 1983)
Descriptive model in stable environments
Structural for radically different counterfactuals
Measurement
usually used under experiments and causal inference
Structural models
Reduced-form, causal effects models
Testing
- Reduced-form, causal effects models
Demand, supply, and marketing mix are endogenously determined.
Best case: find exogenous shocks to the system to estimate
Impose supply model into the demand estimation step (p. 980)
Counterintuitive to assume utility maximization for estimating consumer-level models, instead of firms. But we observe evidence of well-fitting model for the demand-side, but not yet in the supply side. But lack thereof evidence still does not mean that it’s wrong, it’s just mean we need more development.
Building blocks of individual-level demand models
Direct utility specification of demand system
Indirect utility specification of demand systems.
(Lehmann, McAlister, and Staelin 2011)
Tradeoff between rigor (sophistication) and relevance
Basic discipline migrated and viewed as more sophisticated, which lead to arms race. (cascade more and more sophisticated)
Execution rigor > idea quality. We should view analytical rigor and substantive content equally.
Impact:
Citation
Breadth and reach (to other disciplines)
Game the system: cite reviewers.
A good research paper should be (p.162)
reasonably realistic/general
relatively simple and robust
insightful
reasonably communicable
More complex methods are only appropriate when (p. 163)
(P. Chintagunta, Dubé, and Goh 2005)
Try to understand the role of potential weekly brand-specific characteristics that influence consumer choices, but they are unobserved
Endogeneity
Inclusion of the UBC
UBC: they are the first guys to do it in the dis aggregate model.
(Dong, Manchanda, and Chintagunta 2009)
What is the benefit of individual-level targeting in the presence of strategic behavior by other firms?
Setting
Pharmaceutical industry
Individual-level targeting to physicians
Targeted ad (i.e., detailing)
Model
Physician response: capture the responsiveness each physician to targeting
Firm detailing choices: firms strategically target and how much ad