Structural Estimation

Introduces market share inversion mapping observed shares to mean utilities, enabling IV estimation with aggregate data.

The canonical 'BLP' paper: random coefficients logit with unobserved product characteristics and endogenous prices.

Essential implementation guide clarifying BLP algorithms and computational practice.

Rigorous nonparametric identification foundations establishing when demand is identified.

Modern best practices with open-source Python implementation; addresses numerical stability and optimal instruments.

Extends discrete choice to allow complementarity between products; estimates substitution between print and online news using Washington Post data. Essential for digital goods where products may be complements rather than substitutes.

BLP-style random coefficients nested logit for iOS and Android app stores; estimates $33.6 billion annual consumer surplus from mobile apps and quantifies in-app purchase and advertising effects.

Foundational ordered probit entry model using threshold ratios to infer competitive effects from market structure.

Methods for estimating discrete games with multiple equilibria; foundation for airline entry applications.

Authoritative survey covering static and dynamic entry models with identification and estimation methods.

Inference methods for entry games with multiple equilibria using partial identification.

Structural model of competition between 'flexible' peer suppliers (Airbnb) and 'dedicated' sellers (hotels). Finds Airbnb generated $41 consumer surplus per room-night with welfare gains concentrated during capacity-constrained periods.

Develops computational methods for large-scale entry games; tractable approach for markets with many potential entrants relevant to platform expansion analysis.

Foundational paper introducing nested fixed point (NFXP) algorithm and conditional independence assumptions.

Revolutionary CCP approach expressing value functions as functions of observable choice probabilities.

Extends CCP to dynamic games; introduces nested pseudo-likelihood addressing computational challenges.

Integrates unobserved heterogeneity into CCP estimators using EM algorithm.

Excellent practical guide bridging theory and implementation.

Studies consumption dynamics using individual listening histories across streaming platforms; finds streaming adoption increases quantity and diversity of music consumption.

Separates structural state dependence from heterogeneity using household scanner data; essential for understanding subscription stickiness and switching costs in digital services.

The foundational 'GPV' paper introducing nonparametric bid inversion to recover private value distributions.

Establishes nonparametric identification conditions for major auction formats.

Comprehensive handbook chapter; authoritative survey of identification and estimation.

Addresses unobserved auction heterogeneity using multiplicative structure; widely applied in procurement.

Applied survey covering auction theory testing and structural estimation in practice.

Foundational analysis of GSP auction mechanism used by Google and Yahoo; characterizes equilibria and proves revenue equivalence under complete information. Required reading for ad auction design.

Structural econometric methods for estimating bidder valuations in sponsored search auctions; bridges GSP theory with empirical estimation of ad auction data.

Foundational control function approach using investment to proxy for unobserved productivity.

Uses intermediate inputs as proxy; more widely applicable than OP due to data availability.

Resolves functional dependence problems in OP/LP; the 'ACF' estimator is now standard practice.

Combines production function and markup estimation; highly influential applied methodology.

Resolves identification issues with gross output using nonparametric IV methods.

Canonical theoretical framework defining two-sided markets where price structure (not just level) matters; establishes concepts of membership fees, usage fees, and indirect network effects.

First rigorous structural estimation of cross-side network effects in two-sided markets; estimates simultaneous equations for consumer demand, advertiser demand, and publisher pricing.

General theory of monopoly platform pricing showing platforms internalize only marginal users' externalities (Spence distortion). Essential for understanding market power measurement in platform markets.

Provides necessary and sufficient conditions for identifying peer effects using network structure; shows how intransitive networks solve Manski's reflection problem.

Uses 27.4 million Yahoo users to separate peer influence from selection; shows previous methods overestimate influence by 300-700%. Essential methodology for network effect identification.

Early empirical paper distinguishing network externalities from learning spillovers using geographic variation; finds effects tied to email/internet use, supporting communication network hypothesis.

Full industry structural model combining viewership, demand, pricing, bundling, and Nash bargaining between distributors and content providers. Finds unbundling raises negotiated input costs 103%.

Uses Spotify growth to estimate streaming's market impact; finds 137 streams displace 1 track sale but streaming also displaces piracy, making it approximately revenue-neutral.

Estimates willingness-to-pay for digital songs and simulates revenue under alternative pricing (bundling, two-part tariffs); finds bundling raises revenue 16-33% over uniform pricing.

Most-cited digital piracy paper using matched file-sharing and sales data; uses German school holidays as instrument. Established the empirical agenda for piracy research.

Authoritative synthesis reviewing piracy displacement effects, creative incentives, and enforcement effectiveness. Essential methodological guide for digital content research.

First rigorous estimation of 'Long Tail' welfare gains; shows Amazon variety benefits were 7-10x larger than price competition gains. Foundational paper quantifying why e-commerce matters.

Explains the price dispersion puzzle—why dispersion persists despite low online search costs. Shows retailers strategically make search harder through add-on pricing and complex fees.

Uses clickstream data to directly observe search sequences; rejects standard sequential search in favor of fixed sample size model. Methodological benchmark for structural search estimation.

Equilibrium model of search and price competition on eBay; quantifies platform design trade-offs between match quality and price competition using detailed browsing data.

Explains eBay's shift from auctions to posted prices by estimating demand trade-offs between price discovery and convenience using millions of seller experiments.

Foundational methodology for backing out search cost distributions from equilibrium prices; underlies virtually all subsequent empirical search estimation.

Uses Expedia field experiment to causally identify position effects ($1.92/position); gold-standard paper on ranking effects in digital platforms. Bass Outstanding Dissertation Award winner.

Integrates costly search with endogenous consideration set formation; shows advertising shifts awareness rather than preferences, increasing competition.

Canonical consumer-side structural model of sponsored search using Bing data; finds 51% more clicks would occur without ad competition. Essential for understanding attention allocation.

Foundational framework for using ML to estimate nuisance parameters while preserving valid inference on structural parameters through orthogonal scores and cross-fitting.

Develops causal forests for heterogeneous treatment effect estimation with valid confidence intervals. Critical for personalization, targeting, and A/B test analysis.

Deep learning framework for instrumental variables; two-stage neural network approach for demand estimation with price endogeneity. Bridges deep learning with core BLP problem.

Rigorous theoretical foundations proving deep nets can be used as first-step estimators in semiparametric inference; provides convergence bounds validating neural networks for causal inference.