AI summary by TwinMind

Lecture 21: Bounded Rationality, Part 2

MIT OpenCourseWare

1h 19m 58s

Bounded Rationality and Rule of Thumb Models

  • Different approaches to boundedly rational models in economics
    • Psychology-economics style: behavioral models where consumers behave irrationally and firms exploit them
    • Rule of thumb approach: write down models resembling actual consumer behavior rather than complex rational models
  • Critique of rational modeling approach
    • Complex 8-page mathematical arguments for top journal publication
    • Full rationality assumptions often unbelievable
    • Rational modeling can still involve "cheating" through assumption selection
    • Sociology of economics: complicated calculations needed to show academic accomplishment

Smallwood and Conlisk Model (Classic Rule of Thumb Example)

  • Research question: Do superior products win out in markets with imperfect information?
  • Cereal box example: consumers face complex information processing
    • Only tasted small fraction of available cereals
    • Must infer quality from limited information (prices, others' purchases, advertising, shelf placement)
    • Rational approach would require prior over century's worth of cereal purchases and network effects
  • Model assumptions:
    • Products differ in quality (breakdown probability BK)
    • Consumers continue buying same product until breakdown occurs
    • After breakdown, buy product K with probability MK(T)^σ (market share raised to sigma power)
  • Sigma parameter interpretations:
    • σ = 0: random selection from available products
    • σ = 1: proportional to market share (watching one person ahead)
    • σ > 1: explicit reasoning about popularity correlation with quality
    • σ → ∞: always buy most popular brand

Smallwood-Conlisk Results

  • Main theorem: If σ = 1, most popular product dominates in T → ∞ limit
    • Product with smallest BK (highest quality) achieves MK(T) → 1
    • Best product takes over, inferior products die out
  • σ < 1: All products remain active with positive market shares
    • Superior products have larger shares but don't reach monopoly
  • σ > 1: Inferior products can dominate through initial popularity advantages
  • Key insight: Consumers must be fairly naive (σ = 1) for market selection to work optimally
    • Put some weight on popularity but not too much

Banerjee-Fudenberg Extension (1995)

  • More utility-framework approach to Smallwood-Conlisk
  • Consumers receive utility with product-specific means and idiosyncratic preferences
  • Period-specific quality differences (e.g., traffic routes, restaurant quality, insurance coverage)
  • Product 1 better with probability P
  • Consumer behavior:
    • Buy same product with probability 1-α (inertia)
    • With probability α, ask n friends about recent experiences
    • Choose product with better average utility among friends
  • Results: Optimal social learning occurs when n is fairly small
    • When α close to 1 (frequent questioning), need 3-6 friends
    • When α small, asking 1 friend works well (similar to Smallwood-Conlisk σ=1)

Spiegler's Market for Quacks Model

  • Extends word-of-mouth models to include firm pricing
  • Application: health supplements/medications advertised online
  • Model setup:
    • Large number of products giving good outcome with probability α
    • Outside option also gives good outcome with probability α at zero cost
    • Utility from product K: (1-PKT) with probability α, -PKT with probability (1-α)
  • Consumer behavior: Ask one user of each product about experience, buy product giving highest utility if positive
  • Results similar to Stahl price dispersion model
    • Mixed strategy equilibrium with price distributions
    • Price inversely related to product quality
    • High-quality products face more competition (more positive reviews), leading to lower prices
    • Low-quality products face less competition (fewer positive reviews), enabling higher prices

Journal Refereeing Process Model

  • Academic publishing as rule of thumb application
  • Paper quality has two dimensions:
    • Q: fundamental idea/insight (unchangeable)
    • R: craftsmanship quality (improvable through revision)
  • Referee learning process:
    • Referees don't know true α (weight on Q) and Z (quality threshold)
    • Try to learn from experience with own papers and observed acceptances/rejections
  • Overconfidence bias: Everyone thinks their papers are ε better than reality
  • Dynamics lead to perpetual disequilibrium:
    • Referees set standards too high
    • Editors accept papers below referee recommendations to fill journals
    • Creates cognitive dissonance leading to preference for R over Q
  • Empirical evidence: Publication times increased from 6-8 months (1970s) to 2-3 years (2000s)
    • Median time for top IO empirical papers: 7 years from job market to publication

Advertising Theory Overview

  • Three main approaches to modeling advertising effects:
    1. Informational content: tells consumers about product existence, prices, attributes
    2. Signaling: signals quality through money-burning
    3. Preference manipulation: directly changes consumer preferences

Butters Model (1977) - Informational Advertising

  • Setup: n firms selling identical goods, unit mass of identical consumers
  • Consumers get utility V-P, firms have cost C
  • Advertising technology: spending A(x) reaches fraction x of consumers
  • Equilibrium features:
    • Price dispersion (similar to Stahl model)
    • Consumers buy from cheapest firm they heard about
    • Mixed strategy pricing: P = V for monopoly consumers, lower prices for competitive situations
  • Key result: Advertising levels are socially efficient
    • Private return to reaching marginal consumer equals social return
    • Marginal benefit = (V-C) × Pr(consumer saw no other ad)
    • No distortion from business stealing or consumer surplus effects

Milgram-Roberts Model (1980s) - Signaling Advertising

  • Two-quality products (high H, low L) with common prior
  • Firms observe own quality, choose price P and advertising level A
  • Consumers observe A, update beliefs about quality, make purchase decisions
  • Profits depend on both true quality Q and perceived quality Q̂
  • Separating equilibrium possible:
    • High-quality firms: high price + positive advertising
    • Low-quality firms: low price + zero advertising
  • Requirements for equilibrium:
    • Repeat purchases must be important
    • Word-of-mouth learning effects must be small
  • Advertising is pure money-burning but can be socially valuable through improved matching

Behavioral Advertising Model (Shapiro - Unpublished)

  • Based on memory manipulation and confusion
  • Psychological evidence: famous person experiments, orange juice taste tests
  • Consumer types θ get benefit V with probability θ when consuming good
  • Consumers have n prior experiences but recall only one for decision-making
  • Two advertising mechanisms:
    1. Memory replacement: substitute fake memory for real one
    2. Memory priming: make positive memories more likely to be recalled
  • Even with rational consumers aware of manipulation, firms may still advertise for information design reasons
  • Advertising effectiveness depends on product quality (less beneficial for truly high-quality products)

Marketing Field Relationship

  • Marketing departments have ~1000 professors in US with dedicated journals
  • Citations in marketing papers: 6% to economics (up from historical levels)
  • IO theory and empirical work heavily cited in marketing literature
  • Substantial overlap in topics: pricing, demand estimation, advertising campaign effects

TwinMind

Get summaries like this for your own meetings

TwinMind takes notes in the background during meetings, lectures, and calls, then writes the summary for you.

Free plan to get started

No bot joins your calls

Phone, desktop & Chrome apps

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome