Bargaining and informed agent's advice: theory and evidence
OA Version
Citation
Abstract
This dissertation investigates the market effect of informed agents' advice to buyers and sellers. I compare two kinds of informed agents: one who maximizes an expected sales price, the price-maximizing agent, and one who maximizes the probability of trade, the trade-maximizing agent.
The first chapter presents a theoretical model and results. The model is based on a bilateral bargaining game with an informed agent, where the agent has more information than either the buyer or the seller. I show that if there exists a bargaining equilibrium in which a price-maximizing agent disseminates information truthfully, then there also exists a bargaining equilibrium in which a trade-maximizing agent disseminates information truthfully. However, the converse is not true. Thus, for some cases, a trade-maximizing agent is truthful, but a price-maximizing agent is not. In such cases, a trade-maximizing agent leads to a higher trade probability and a higher expected sales price than a price-maximizing agent.
The second chapter studies the model under specific assumptions. Here, both the buyer's and seller's good valuations are uniformly distributed, and the agent knows whether each of the buyer and seller has a high or low valuation. I analyze three cases: an informed agent advises (1) a buyer only, (2) a seller only, and (3) both. Whereas a trade-maximizing agent is always truthful, a price-maximizing agent over-reports a seller's valuation to increase the sales price.
The third chapter empirically tests the model prediction: a trade-maximizing agent leads to a higher trade probability than a price-maximizing agent. I do so using South Korean housing transaction data. I exploit the unique Korean real estate agents' commission scheme to identify the two kinds of agents. I solve the missing listing data problem by applying cluster analysis, which is an unsupervised machine learning algorithm. I show the conditions under which the number of sales is the appropriate proxy for the trade probability even when listing data are missing. The results show that a trade-maximizing agent brings an approximately 0.2% greater number of sales in each ten-day period relative to a price-maximizing agent.