Insights from Economic Experiments
Plott et al (2006): Testing Rational Choice
This study explores how experimental markets can test rational individual choice. Theories of markets where rational individual choice is a necessary component either challenge its predictions or cannot be applied.
Design & Implementation
- Based on Induced Value Theory.
- Suppliers and buyers had marginal cost and demand functions.
- Middlemen bought from suppliers and sold to buyers, repeated each period.
- Inventories could not be carried forward to the next period.
- Double anonymity used.
- Total of 8 periods.
Key Results
- Both markets converged to the predicted equilibrium price after 5 periods.
- Predicted volumes converged close to actual volumes.
- Profits of middlemen dissipated to near zero.
The market becomes accurate after a process of convergence, a finding not suggested by current models.
Huck et al (2004): Firms and Cournot Competition
This meta-analysis and experiment investigates how the number of firms impacts Cournot competition.
Design & Implementation
- Meta-Analysis of Cournot oligopolies.
- Market demand modeled according to the inverse demand function p=max{100-Q,0}.
- Subjects chose quantities between 0 and 100 with a 0.01 step.
- Varied the number of firms (n=2, 3, 4, 5).
- Experiment recruited via telephone and email at a London university.
- Used a neutral frame, changing terminology.
Key Results
- For the meta-analysis, r (a measure of convergence) increases with the number of firms.
- For the experiment, total quantity increases with the number of firms, significant for rounds 17-25.
- The neutral frame led to outcomes closer to the Nash Equilibrium (NE).
- Collusive behavior was not feasible with two firms.
- Rivalistic behavior was more prevalent in markets with more firms.
Kirchkler et al (2004): Causes of Asset Market Bubbles
This research examines the causes of “bubbles” in experimental asset markets. The perception of stocks as constantly increasing or constant can contribute to price bubbles.
Design & Implementation
- 10 subjects traded a dividend-paying stock for experimental currency over 10 periods.
- Included 6 treatments:
- Declining Fundamental Value (FV), Increasing Cash/Asset (C/A) ratio.
- Constant FV, Increasing C/A ratio.
- Declining FV, Constant C/A ratio.
- Constant FV and C/A ratio.
- Changed wording to “stocks of depletable gold mine”.
- Robustness check for Treatment 5.
Key Results
- Found high mispricing in treatments with declining FVs.
- Overvaluation emerged only when a declining FV was coupled with an increasing C/A ratio.
Voslinsky and Azar (2021): Incentive Impact on Experiments
This study explores how different uses of incentives impact experimental results.
Key Findings on Incentives
- Incentives help recruit subjects.
- Performance-based incentives can motivate careful decision-making.
- Experiments without incentives do not adversely affect behavior, but may make it noisier.
- Paying for all rounds versus only one round in multi-round experiments yielded inconclusive results.
- Paying all participants versus a subset was at least as effective as paying all.
- Conversion rates of experimental currency to real money influence behavior towards nominal vs. real money.
- Discusses non-monetary incentives and incentives in experiments generally.
Holt and Laury (2002, 2005) and Harrison et al (2005): Incentives and Risk Behavior
These studies investigate how order and scale effects on incentives shape risk behavior.
Holt and Laury (2002) Findings
- Subjects are risk-averse for most stakes.
- Risk aversion increases with stakes.
- No significant effect from using hypothetical payment.
However, the 2002 experiment had a possible order effect.
Design & Implementation (Follow-up)
- Tested stakes of x1, x10, or x1 then x10.
Key Results (Follow-up)
- Showed that there is both an order and scale effect that increases subjects’ risk aversion.
Rydval & Ortmann (2004): Incentives, Ability, and IQ
This research examines how the effects of financial incentives and cognitive abilities can be approximated by a measure of subjects’ IQ.
Design & Implementation
- Used individual-level data from Gneezy and Rustichini.
- Plotted curves ranking IQ performance with different levels of stakes.
Key Results
- Found high within-treatment variation.
- Performance curves for high-incentive treatments were virtually identical.
- The low-incentive treatment yielded lower performance than the no-payment treatment.
Thus, this suggests motivational problems for lower IQ individuals and indicates that financial incentives have a non-monotonic effect.
Della Vigna & Pope (2016): Behavioral Interventions and Effort
This study investigates how behavioral interventions induce effort and whether this aligns with behavioral expert’s expectations.
Design & Implementation
- Designed a large experiment comparing 18 treatments.
- Treatments included standard incentives, non-monetary psychological inducements, and behavioral factors differing in one paragraph of instructions.
- Measured beliefs of academic experts about the effectiveness of treatments.
- Each treatment had 550 subjects.
- Provided benchmark results for 3 treatments to help experts calibrate participant effort.
Key Results
- Monetary incentives have a strong motivating effect.
- Psychological inducements have a moderate effect.
- Behavioral factors were consistent with behavioral models.
- Experts correctly anticipated the effectiveness of psychological inducements but overestimated the effectiveness of very-low-pay treatments, gift exchange, and probabilistic payments.