This three-day workshop covers the Berry-Levinsohn-Pakes (BLP) approach to estimating the statistical relationship between product sales and product characteristics such as prices. As the foundational approach for differentiated products demand estimation in the industrial organization literature, BLP is used by academics, antitrust regulators, and industry professionals to shed light on difficult questions.
- “What is the value of a new good?”
- “Will a merger hurt consumers?”
- “Should we change prices?”
Through a running empirical example, the workshop will use a series of coding exercises to build up practical knowledge for studying these types of questions and more.
This is one of our advanced courses. These courses are designed assuming a solid foundation in the basics of economic models and instrumental variables. Scott's Causal Inference (Part 1) covers instrumental variables.
- History and motivation for BLP (by Ariel Pakes).
- The BLP model.
- Pure logit estimation.
- Price endogeneity.
- Exercise 1: Getting set up with Python and PyBLP, pure logit estimation, and running a price cut counterfactual.
- Preference heterogeneity.
- Mixed logit estimation.
- Numerical best practices.
- Differentiation instruments.
- Exercise 2: Incorporating preference heterogeneity, mixed logit estimation, and evaluating improvements to the price cut counterfactual.
- Micro BLP estimation.
- Choosing micro moments.
- Using more information.
- Exercise 3: Incorporating micro moments, micro BLP estimation, and evaluating improvements to the price cut counterfactual.
