Being ‘Too’ Smart Doesn’t Always Pay Off; Truth-Telling in a Strategy-Proof Mechanism

Abstract

In strategy-proof mechanisms truth-telling is a dominant strategy yielding optimal results for all participants, yet according to experimental evidence truthful revelation rates are worryingly low. This thesis is based on the growing realm of literature finding that participants tend not to play their dominant strategy, i.e. tell the truth, within matching markets even when given strategy-proof mechanisms. I design an experiment using the well-known Gale-Shapley Deferred Acceptance (DA) algorithm that aims to test whether using tutorials helps participants better understand that truth-telling is their optimal strategy within strategy-proof mechanisms. These tutorials allow participants to practice their approach and observe the outcome of their choices before having to commit to a final strategy. I hypothesise that participant engagement in tutorials will increase truth-telling rates in matching mechanisms and predict that given advice alone reduces rates of truthful preference revelation, participant engagement in tutorials can effectively mitigate the misleading power of advice. This thesis finds that the proportion of participants that told the truth was 3% lower if they received the tutorials compared to when no additional information was presented. However, tutorials improve on truth-telling rates compared to advice as 31% more people told the truth when given tutorials and no advice compared to if they were presented with advice alone. Additionally, participants engaging with both advice and tutorials together had truth-telling rates 11% higher than those given solely advice, which further supports the effectiveness of tutorials in overcoming the misleading effect of advice.