Pricing and Market Design
Optimization, learning, and incentives in modern markets
what codex thinks I will be teaching
Course Description
Every market must decide who gets what, and on what terms. This course asks:
When do simple prices work, when do they fail, and what replaces them?
Across the course, prices will play a leading role, as tools to extract surplus, ration scarcity, decentralize allocations, correct externalities, and experiment and learn. We will combine ideas from operations research, economics, and computer science to study demand, scarce capacity, customer choice, strategic behavior, private information, and institutions such as auctions and matching. The questions we will study are increasingly important in digital platforms, where algorithms – and now increasingly autonomous agents – both learn from markets and change the data the markets generate.
This is intended to be a mathematically substantive, model-driven course for senior OR/CS undergraduates and master’s students. That said, I will aim to keep our organizing principle as economic question first, mathematical machinery second: we will introduce optimization, probability, learning, or game-theoretic tools when a market-design problem demands it. Most lectures will begin with concrete market questions, develop a model, and try to extract reusable principles rather than presenting the math in isolation.
Course Information
- Instructor: Sid Banerjee, email
- Lectures: Tuesday/Thursday, 1:25–2:40 p.m., CIS 142
- Office: 229 Rhodes Hall
Detailed dates, assessment information, and course policies will be posted when finalized.
References
There is no required textbook; however we will assign readings from three main references (all available online through Cornell Library):
- Rakesh Vohra and Lakshman Krishnamurthi, Principles of Pricing.
- Tim Roughgarden, Twenty Lectures on Algorithmic Game Theory.
- Kalyan Talluri and Garrett van Ryzin, The Theory and Practice of Revenue Management.
Selected course notes and papers will supplement these references, particularly for learning, online allocation, reputation, and matching.
Lectures and Notes
The plan below is tentative. Topics, dates, and the division between lectures may change with pace. Lecture-note links will be activated as materials are posted.
Unit 1: Pricing, Demand, and Learning
Lecture 1 — Aug. 25: Pricing with full information: surplus, market clearing, and congestion tolls
LP Toolkit [notes]
Lecture 2 — Aug. 27: From values to demand: quantiles, virtual values, and elasticity
Lecture 3 — Sept. 1: From optimal pricing to learning: markup, greedy failure, and regret
- Lecture notes: [Lec 3]
- Recommended Reading:
- Slivkins, Ch. 1, §§1.1–1.3 [Slivkins]
Lectures 4–5 — Sept. 3 and Sept. 8: Learning to price: confidence, exploration, and UCB
- Lecture notes: [Lecs 4–5]
- Recommended Reading:
- Slivkins, §§1.3.1–1.3.3 Slivkins
- Bandit Lab: [game] — compare learning policies in an interactive multi-armed bandit simulator
Unit 2: Scarcity, Scale, and Online Allocation
Lecture 6 — Sept. 10: Scarcity and the value of capacity: Littlewood’s rule
Lecture 7 — Sept. 15: Static versus dynamic revenue management
- Topics: single-resource RM, dynamic programming, static policies, and square-root scaling
- Lecture notes: [Lecs 7–8]
- Suggested Reading:
- T&vR, §§2.5.1–2.5.2 T&vR
Lecture 8 — Sept. 17: Static versus dynamic revenue management, continued
- Lecture notes: [Lecs 7–8]
Lecture 9 — Sept. 22: Fluid models and Jensen’s inequality
- Topics: fluid benchmarks, concavity, and Jensen’s inequality
- Lecture notes: [Lecs 9–10]
- Suggested Reading:
- T&vR, §3.1.2.3, §§3.2.2–3.2.5, and §3.3.1 T&vR
Lecture 10 — Sept. 24: Network RM: fluid LPs, bid prices, and static controls
- Topics: network RM, LP duality, bid prices, and static randomized controls
- Lecture notes: [Lecs 9–10]
Lecture 11 — Sept. 29: Bayes Selector and compensated coupling
- Format: Online (Zoom)
- Topics: hindsight prediction, compensated coupling, and satisfying actions
- Lecture notes: [Lecs 11 and 13; Lec 14 planned]
- Suggested Reading:
- Bayes Selector Lab: [simulation] — compare static fluid, re-solved fluid, and Bayes selection against hindsight
Oct. 1 — Class canceled (no lecture)
Lecture 13 — Oct. 6: Compensated coupling: Uniform[0,1] example and regret bounds
- Lecture notes: [Lecs 11 and 13; Lec 14 planned]
- Suggested Reading:
- Banerjee, The Compensated Coupling, Simons Institute talk, slides 14–20 and 24–34
- Vera–Banerjee, §§3–4 [paper]
Lecture 14 — Oct. 8 (planned): Bayes Selector (discrete types); introduction to the Spiral-Down Effect
- Topics: discrete examples and regret bounds; demand learning, endogenous data, and spiral down
- Bayes Selector notes: [combined notes]
- Spiral-Down draft: [notes]
- Interactive demo: The Spiral-Down Lab
- Suggested Reading:
Unit 3: Customer Choice and Assortment
Choice models and assortment optimization follow the Spiral-Down discussion; their meeting dates will be updated after Lecture 14.
Unit 4: Auctions, Game Theory, and Mechanisms
Oct. 13: Fall Break — no class
The remaining sequence is tentative; topics and their allocation across class meetings may change.
- Posted prices versus auctions
- Game theory, auction formats, and strategic bidding
- Truthful allocation in single-parameter environments and Myerson’s lemma
- Monopoly reserves and simple near-optimal auctions
Unit 5: Segmentation and Richer Pricing
- Observable and hidden customer types: segmentation and screening
- Menus and self-selection: versioning and nonlinear pricing
- Multidimensional values: bundling and multi-product pricing
Unit 6: Allocation, Competition, and Information
- Multi-item allocation, combinatorial auctions, and VCG
- Pricing under competition
- Adverse selection, reputation, and information in markets
Nov. 26: Thanksgiving Break — no class
Unit 7: Matching, Platforms, and Synthesis
- Matching and markets without money
- Platforms and two-sided markets
- Course synthesis and review
Possible extensions, as time permits, include overbooking, finite-inventory dynamic pricing, proper scoring rules, censored-demand estimation, multi-parameter revenue maximization, and auction extensions.
Assignments
- Assignment 1 [questions] [solutions]: due on September 8 at 1:00 p.m. ET
- Assignment 2 [questions]: due on October 1 at 1:00 p.m. ET
Learning Goals
By the end of the course, students should be able to:
- formulate and solve basic pricing and allocation models using buyer values, demand, choice, and resource constraints;
- interpret LP dual variables and dynamic marginal values as prices or opportunity costs;
- analyze demand learning using concentration bounds, regret, optimism, and the feedback between decisions and observations;
- distinguish exact, fluid, and clairvoyant benchmarks, and explain how scale affects performance and tractability;
- model substitution using random-utility models and optimize simple assortments; and
- analyze strategic and informational problems and compare prices, auctions, mechanisms, reputation, and matching as market-design interventions.
Prerequisites and Background
Required background
Comfort with linear optimization, basic probability, calculus, and mathematical modeling, approximately at the level of ORIE 3300 and ORIE 3500 (or equivalent). You should know (or be willing to learn) to use LP duality and complementary slackness; random variables, expectation, conditional probability, and common distributions; and elementary calculus.
Helpful but not required
Prior exposure to economics, game theory, stochastic processes, or algorithms. Some assignments may involve computation or simulation, so familiarity with Python or a comparable language will be helpful.
Siddhartha Banerjee
Associate Professor
Sid Banerjee is an associate professor in the School of Operations Research at Cornell, working on topics at the intersection of data-driven decision-making, market design, and algorithms for large-scale networks.