Experimental economics Study Guide

📖 Core Concepts Experimental Economics – Uses controlled experiments (lab or field) with real monetary incentives to test economic theories. Cash Incentives – Real payoffs ensure participants treat decisions like real‑world choices. Lab vs. Field – Lab: artificial setting, high control. Field: real market environments (e.g., John List’s field auctions). Natural / Quasi‑Natural Experiments – Observe real events (natural) or mimic them with a designed structure (quasi‑natural). Nash Equilibrium (Coordination Games) – Multiple pure‑strategy equilibria; participants must coordinate on one. Social Preferences – Concerns for others’ welfare: altruism, fairness, reciprocity, spite, equality‑taste. Moral Hazard & Adverse Selection – Contract issues where actions or information are hidden from the other party. Validity – Internal: causal inference within the experiment; External: generalizability to real economies. --- 📌 Must Remember Cash incentives → realistic behavior (no hypothetical bias). Price/Quantity convergence in many lab markets → competitive equilibrium even with imperfect competition. Ultimatum Game: Low offers are often rejected → fairness overrides pure self‑interest. Reinforcement Learning (Roth & Erev): Past high‑payoff actions are repeated; predicts behavior in many games. EWA Model = weighted mix of reinforcement + belief learning; parameters can over‑fit. Field experiments (List) provide higher external validity than pure lab work. Vernon Smith won 2002 Nobel for institutionalizing lab experiments. Social preference games (dictator, trust, public‑goods) are the standard tools for measuring altruism/fairness. --- 🔄 Key Processes Belief Learning Form expectation of others’ actions → update after each round → choose best response. Reinforcement Learning Record payoff of each action → increase attraction for high‑payoff actions → choose action with highest attraction. Experience Weighted Attraction (EWA) Initialize attraction \(A{i}(0)\). After each period \(t\): $$A{i}(t) = \frac{\phi N(t-1)A{i}(t-1) + [\delta + (1-\delta)I{i}(t)]\pi{i}(t)}{N(t-1)+1}$$ where \(\phi\) = decay, \(\delta\) = weight on realized vs forgone payoffs, \(I{i}(t)\) = indicator of chosen action, \(\pi{i}(t)\) = payoff, \(N(t)\) = experience weight. Over‑fitting Mitigation Split data → estimate parameters on training set → test predictions on out‑of‑sample set. Equilibrium Selection in Coordination Games Deductive: Use payoff dominance, risk dominance, or focal point logic. Inductive: Observe early play, update beliefs, converge on a pattern. --- 🔍 Key Comparisons Deductive vs. Inductive Selection – Deductive: Relies only on game structure (payoffs, symmetry). Inductive: Relies on observed play and learning dynamics. Reinforcement vs. Belief Learning – Reinforcement: No modeling of opponents; repeats profitable actions. Belief: Forms explicit expectations about others’ strategies. Lab vs. Field Experiments – Lab: High internal validity, low external validity. Field: Higher external validity, lower control over confounds. Natural vs. Quasi‑Natural Experiments – Natural: No researcher control; purely observational. Quasi‑Natural: Researcher designs a “natural‑like” setting with some control. --- ⚠️ Common Misunderstandings “Lab results are always unrealistic.” – Many lab markets converge to competitive equilibrium despite unrealistic assumptions. “EWA always outperforms other models.” – Over‑fitting can make EWA look better in‑sample; out‑of‑sample tests often reduce its advantage. “Social preference games only measure altruism.” – They also capture fairness, reciprocity, and spite; interpretation must consider all motives. “Field experiments eliminate all validity concerns.” – External validity improves, but internal validity (confounding factors) can still be problematic. --- 🧠 Mental Models / Intuition “Incentive Alignment” – Real money makes participants treat experimental decisions like real economic choices. “Learning as Weighting Past Payoffs” – Imagine a mental “balance scale” that tips toward actions that have historically tipped the scale higher. “Equilibrium Selection as Focal Point” – People gravitate toward the most “obvious” or “salient” equilibrium (e.g., the one highlighted by payoff dominance). --- 🚩 Exceptions & Edge Cases Convergence to competitive equilibrium can fail when participants have strong risk‑averse preferences or when communication is allowed. Cultural variation can flip ultimatum‑game outcomes: some societies reject low offers far more (or less) than others. EWA parameters may need self‑tuning functions when games differ dramatically in payoff volatility. --- 📍 When to Use Which Use Lab Experiments when you need precise control over variables (e.g., testing specific auction rules). Use Field Experiments when testing policy relevance or market mechanisms in real‑world settings. Apply Reinforcement Learning models for games where payoff feedback is immediate and opponents are not strategically modeled. Apply Belief Learning models for games with strategic interdependence and where participants can observe or infer others’ actions. Choose EWA when you suspect both reinforcement and belief processes drive behavior and you have enough data to guard against over‑fitting. --- 👀 Patterns to Recognize Price‑Quantity Convergence → Look for markets that quickly settle near theoretical equilibrium despite limited participants. Fairness Rejection → Ultimatum offers ≤ 20% of pie often trigger rejections across cultures. Learning Curves → Early rounds show high variance; later rounds stabilize as attractions/ beliefs solidify. Multiple Nash Equilibria → Presence of coordination games; check for focal‑point cues (labels, symmetry breaks). --- 🗂️ Exam Traps “All lab experiments lack external validity.” – False; many findings replicate in the field. “EWA is always the best predictive model.” – Wrong if parameters are over‑fit; simpler models may perform better out‑of‑sample. “Only moral hazard is tested in contracts.” – Overlooks adverse selection, hold‑up, exclusive contracting, etc. “Social preference games only measure altruism.” – Ignoring reciprocity, spite, and equality concerns leads to mis‑interpretation. “Natural experiments are the same as field experiments.” – Natural experiments have no researcher‑imposed treatment; field experiments do.