Working Paper: NBER ID: w27704
Authors: Steven T. Berry; Philip A. Haile
Abstract: A recent literature considers the identification of heterogeneous demand and supply models via "quasi-experimental'' variation, as from instrumental variables. In this paper we establish nonparametric identification of differentiated products demand when one has "micro data'' linking characteristics of individual consumers to their choices. Micro data provide a panel structure allowing one to exploit variation across consumers within each market, where latent demand shocks are fixed. This facilitates richer demand specifications while substantially softening the reliance on instrumental variables, reducing both the number and types of instruments required. Our results require neither the structure of a "special regressor'' nor a "full support'' assumption on consumer-level observables.
Keywords: differentiated products; demand identification; micro data; nonparametric methods; instrumental variables
JEL Codes: C14; C26; C3; D12; L0
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| micro data (C81) | nonparametric identification of differentiated products demand (D10) |
| availability of micro data (C81) | reduction in identification complexity (C30) |
| within-market variation in consumer choice problems (D11) | identification of latent demand shocks (J23) |
| identification of latent demand shocks (J23) | identification of demand (R22) |