AI summary by TwinMind
Lecture 02: Monopoly Pricing and Durable Goods, Part 2
MIT OpenCourseWare
1h 21m 51s
Monopoly Information Design Theory Recap
- Continuation from previous lecture on monopoly information design
- Example: Clover sandwich with consumer values uniform on 0-18, cost $6
- Full information: monopoly profit at price 12 (halfway between cost 6 and max value 18)
- No information (bagged sandwich): all consumers value at 9, monopoly price 9, higher profits but zero consumer surplus
- Partial information: monopoly price 10, both consumers and firm better off than full information
- Practical implementation examples:
- Omniscient firms (Amazon, Google, Apple) knowing consumer preferences better than consumers
- Revealing partial product characteristics (e.g., cauliflower base ingredient)
General Information Design Framework
- Russler and Zenus paper framework
- Values distributed according to F on [0,1], zero cost, μ = mean consumer value
- Firm chooses any joint distribution on V and s (signals)
- Consumers willing to pay E[V|s] after receiving signal s
- Constraints: E[s] = μ and F must be mean-preserving spread of signal distribution G
Profit vs Consumer Surplus Maximization
- Profit maximization: Simple - tell consumers if value > C or < C, charge E[V|V>C], fully extract surplus
- Consumer surplus maximization: More complex optimization problem
- Goal: get everyone to buy at lowest possible price
- Optimal distribution: concentrated mass at low values with flat section
- Creates monopolist indifference over price range, forces low pricing
Empirical IO Historical Context
- Pre-1960s: Industry-specific thick books, descriptive work, light on theory
- 1960s-70s: Cross-section regressions across industries
- Structure-conduct-performance literature
- Markup = α₀ + α₁(four-firm concentration ratio) + α₂(advertising/sales ratio)
- Used Census of Manufacturers data for 150+ industries
- Critiques leading to decline:
- Industry-specific factors omitted (750 variables for 150 observations)
- Accounting vs economic costs measurement error
- Endogeneity issues (profitable industries attract entry)
New Empirical Industrial Organization (NEIO)
- 1980s revolution: Return to single-industry studies
- Emphasis on well-measured variables (prices, quantities, market shares vs accounting data)
- Heavy focus on identification and endogeneity
- Theory-guided empirical work
- Structural evolution: From equilibrium objects to primitive parameters estimation
- Counterfactual policy simulations capability
- Disconnect from theoretical models due to empirical constraints
Paper Topic Selection Strategies
- Applied/policy questions with data search
- Simple industries for complex model testing (cardboard boxes, ketchup)
- Data-driven opportunities (startup connections, rich micro data)
- Recent revival of "big questions" driven by macroeconomists
Textbook Market Study (Chevalier & Goolsbee)
- Policy questions: Why $300 textbooks? Why frequent new editions making used books worthless?
- Theoretical question: Are consumers forward-looking in durable goods markets?
- Industry advantages:
- Static purchase decisions (semester timing)
- Limited substitution (assigned textbooks only)
- No lemons problem (observable quality)
- Large sample (subjects × semesters × textbooks)
Background Facts
- New editions every 3-5 years
- Fixed textbook prices (no decline over time despite rental value logic)
- Used/new price ratio: 50-75% standardized
- Buyback at 50% of cover price
- Market shares: ~30% new, ~20% used, ~50% library/other
- Hazard rate peaks at year 3 for revisions
Empirical Strategy
- Step 1: Estimate revision probability using hazard model
- Function of book age, subject (economics/psychology/biology), level
- Peak probability ~35% at 5 semesters for economics books
- Step 2: Test forward-looking behavior via demand response to revision probability
Demand Estimation
- Discrete choice model: utility from new book, used book, no book
- U(new) = X_jt β - α(P_jt - δμP_jt(1-death_prob)) + ε
- Net price accounts for expected resale value discounted by δμ
- Logit transformation: ln(share_new/share_none) = utility specification
- IV for price: nonprofit publisher status, competitor concentration
Results
- Estimated λ = δμ = 0.61
- Given μ ∈ [0.4, 0.75], implies δ close to 1 (highly forward-looking consumers)
- Strong demand response to revision probability
- Amazon validation: used book prices drop dramatically when new edition announced
Media Bias Study (Gentzkow & Shapiro)
- Research question: Profit-maximization vs ideological motivations for media bias
- Setting: Newspaper industry (pre-internet era with multiple local papers)
- Approach: Test optimal product quality theorem for news slant
Bias Measurement Innovation
- Automated text processing of newspaper content
- Identified 1000+ two/three-word phrases differentiating Democratic vs Republican speech patterns
- Congressional speech analysis using Bush vote share in districts
- Newspaper slant = coefficient from regressing phrase usage on Republican indicator
- Examples: "Iraq war" vs "war on terror", "estate tax" vs "death tax"
Demand Model
- Utility = average quality - γ(newspaper_slant - ideal_slant_z)² + ε
- Ideal slant in zip code z = α + β(Republican_share_z)
- Logit transformation to estimable form
- Key interaction: slant × Republican_share coefficient identifies profit-maximizing relationship
Results
- Strong positive relationship: right-slanted news sells better in Republican areas
- Newspaper slant correlates with zip code Republican share
- Owner effects analysis:
- Owner fixed effects double R² (0.08 to 0.44)
- But owner effects not significant when including state fixed effects
- Owner political contributions explain only 5% variance vs 20% for zip code politics
- Conclusion: Profit-maximization appears to dominate ideological motivations
Methodological Concerns
- Heroic assumptions: no heterogeneity within zip codes, no error in ideal slant equation
- Phrase selection may capture non-political content (agriculture vs urban topics)
- Strong functional form assumptions for profit-maximization claims
