Ad Tech & Digital Advertising Economics

Establishes that optimal advertising-to-sales ratio equals the ratio of advertising elasticity to price elasticity—the foundation for budget allocation decisions.

Distinguishes search goods from experience goods, showing advertising serves different informational functions for each category.

Formalizes how advertising expenditure and pricing jointly signal quality for experience goods, even when ads contain no direct information.

Nobel Prize-winning work on consumer search theory showing why price dispersion persists even in competitive markets with positive search costs.

Comprehensive survey synthesizing decades of research on how consumers allocate limited attention—the foundation of attention economics.

Establishes that platforms must 'get both sides on board' and that price structure matters due to cross-group network externalities.

Introduces 'competitive bottleneck' concept: when advertisers multi-home but users single-home, platforms gain market power over advertisers.

Nobel Prize-winning introduction of second-price sealed-bid auctions proving truthful bidding is dominant strategy—the basis for historical RTB auction designs.

Nobel Prize-winning characterization of revenue-maximizing auctions introducing virtual valuations—fundamental to reserve price optimization.

Develops affiliated values theory and the linkage principle explaining why open auctions generate higher revenue when values are correlated.

Shows GSP is not truthful but has locally envy-free equilibria. Foundational paper for sponsored search auction economics.

Provides empirical evidence that symmetric Nash equilibrium describes Google's ad prices accurately, with analysis of advertiser behavior.

Endogenizes click-through rates by modeling consumer search behavior, providing theoretical motivation for position-specific reserve prices.

Large-scale Yahoo field experiment validating Myerson's theory in real ad markets, demonstrating optimal reserve price setting.

Meta-analysis of 25 large field experiments showing median confidence intervals on ROI exceeding 100 percentage points, requiring 10M+ person-weeks for informative results.

Compares 15 Facebook RCTs comprising 500M observations against observational methods, documenting that standard approaches fail due to 'activity bias.'

Identifies control-group counterparts of exposed consumers without PSA controls, achieving 5.9-16.4x variance reduction. Meta-study of 432 Google Display experiments.

Introduces Bayesian structural time-series models for measuring intervention effects—the foundation of Google's CausalImpact package.

Develops counterfactual-adjusted Shapley values for theoretically grounded allocation of credit across marketing touchpoints.

Shows attribution isn't just measurement: different rules affect publisher incentives and market equilibrium, with last-touch over-incentivizing late-funnel exposures.

Combines Augmented Synthetic Control with power analysis for geo-holdout tests when user-level randomization is infeasible.

Best Paper Award winner finding average cookie lifetime of 279 days with average value of €2.52 per cookie; restricting to 1 year decreases value by ~25%.

Large Meta experiment with 70K+ advertisers: median cost per incremental customer rose 31% (from $38.16 to $49.93) without offsite data.

Using 3.7B impressions, finds GDPR reduced revenue per click 5.7%, CTR 2.1%, conversion rates 5.4%—though contextual targeting offset ~44% of conversion decline.

NBER review summarizing: GDPR imposed costs, decreased revenue, hurt profitability, and reduced venture funding particularly for data-intensive startups.

Reports U.S. tracking rates dropped 54.73 percentage points (from 72.63% to 17.90%) with advertising revenues decreasing 20.44% for Apple users.

Finds conversion-optimized Meta ads showed 37% CTR reduction post-ATT, with smaller firms facing steeper declines in new customer acquisition.

Shows behavioral and contextual targeting work better in combination, with implications for privacy-constrained advertising strategies.

Formally models how Google's integration across the ad tech stack enables exclusionary strategies harming publishers and advertisers.

Provides comprehensive legal-economic analysis of header bidding's competitive effects and Google's 'last look' advantage through Dynamic Allocation.

Shows conditions under which search engines may allow fraud, formalizing the economics of click fraud in ad-supported platforms.

First large-scale empirical documentation of mobile ad fraud patterns, identifying concealment strategies in daily traffic patterns.

Provides welfare analysis framework showing equilibrium advertising levels may be too high or too low depending on nuisance costs to viewers.

Industry benchmark finding only 36% of spend reached working media in 2023, improving to 68.8% by Q3 2025 as transparency standards gained adoption.

Facebook's influential paper on CTR prediction combining decision trees with logistic regression, establishing key industrial practices.

Google's architecture jointly training linear models for memorization and DNNs for generalization, widely adopted in ad systems.

Huawei's architecture combining factorization machines with deep networks, eliminating need for manual feature engineering in CTR prediction.

Alibaba's introduction of attention mechanisms weighting historical behavior by relevance to candidate ads, now foundational in CTR systems.

Introduces MDP formulation with neural network value functions for RTB, foundational for production bidding systems.

Formulates RTB as constrained optimization with budget pacing, establishing theoretical foundation for DSP bidding algorithms.