Conflicts between principals and agents: evidence from residential brokerage☆

Sellers in real estate markets face imperfect information about the market value of their assets and the location of potential buyers. This creates a role for real estate agents who can help sellers both by employing their superior search technology to locate prospective buyers and by using their superior knowledge of the market to suggest an optimal asking price. In return for their services, agents typically receive a percentage (generally 6%) of the transaction price as commission. The significance of the role played by agents is evident: they account for the sale of approximately 81% of single-family dwellings in the housing market (see Federal Trade Commission, 1983). With a sales volume of $1348.2 billion, the estimated total commission income from the sales of existing homes alone exceeded $65.5 billion in 2003 (see www.realtor.org).

The relationship between a home-seller and his agent is a perfect example of a principal-agent problem.¹ Agents have a fiduciary responsibility to represent the interest of their principals (sellers) to the best of their ability, that is, in the same way they would represent themselves. However, whether or not an agent would advance his principal's interest at his own expense is not clear.

The standard agency models in general, and the models of real estate brokerage in particular, argue that the percentage commission system creates an agency problem between an agent and his client because it induces the agent to expend too little effort. A recent study by Williams (1998) questions this result. Williams’ (1998) theoretical work shows that a more realistic modification of standard agency models eliminates any agency problems between agents and their clients under the percentage commission system. This is a significant departure from the earlier models and has important policy implications. The significance of the debate is further augmented by the fact that it applies to most classes of professionals, including accountants, consultants, and lawyers. However, there has been no empirical test put to this debate so far. The principal purpose of this study is to offer empirical evidence on whether the percentage commission structure in the brokerage industry creates any agency problems. The paper also offers a theoretical framework that captures the essential ingredients of the agency problem between the agent and his client under the percentage commission structure.

In order to examine whether the agent acts in the best interests of his client, we compare the effort level and the pricing strategy of an agent for two scenarios, the first being that the agent owns the listed property and the second being that the agent is not the owner. We utilize a data set that is sufficiently comprehensive to allow for the identification of sold properties that were owned by listing agents. Thus, we can directly test for the selling price and marketing time differences that would provide evidence of any agency problems. In a world of perfect information with no agency problems, an agent would seek the same price and use the same effort for a contracted seller as he would in selling his own house. That is, equivalent assets will remain in the market for the same length of time and sell for the same price regardless of whether or not they were sold by an owner-agent.

The agency problem in markets for real assets has been the focus of various recent studies. The general conclusion of the earlier models is that although the percentage commission system ensures the interests of the agent to be in the same direction as those of the client, it fails to align the magnitude of the interests of the agent with those of the client. Geltner et al. (1991), Anglin and Arnott (1991), and Miceli (1991) point out the agency problems with respect to the agent's search effort level and argue that the agent will expend less than the efficient level of effort for the seller. Arnold (1992) studies the pricing aspect of the agency problem and shows that the agent's reservation price for the seller's house will generally be different than the seller's reservation price. The source of the agency problem in these models can be summarized as follows. The seller wants to maximize the selling price while minimizing the time the asset stays on the market. The agent, on the other hand, seeks to maximize his expected commission revenue while minimizing the time on the market. Given that the agent receives a small portion of the transaction price as commission, the agent's goal of maximizing the expected commission may diverge from the seller's goal of maximizing the selling price. Furthermore, given that the targeted selling price will impact the time the asset stays on the market, the agent's desired time on the market may diverge from that of the seller. Since the agent's search efforts are not directly observable and the agent has asymmetric information about the market value of the asset, this divergence can lead to shirking by the agent. As a result, the agent may be motivated to convince the seller to underprice the asset at the listing and/or attempt to induce the seller to accept a suboptimal selling price at the bargaining in order to facilitate a faster sale and obtain a more timely commission. Williams (1998) points out that the existence of such agency problems in earlier models is due to unrealistic assumptions. Specifically, earlier models typically match one agent with one seller and ignore the fact that agents have to search for new clients as well as find buyers for their existing clients. Williams’ model allows for multiple sellers per agent and requires agents to spend time selling the assets of current clients and searching for new clients. His conclusion is supported by some of the theoretical studies on optimal contracting which also show that under certain conditions, the optimal contract between a principal and an agent involves a linear compensation scheme (e.g., Diamond, 1998; Hart and Holmstrom, 1987; Holmstrom and Milgrom, 1987; McAfee and McMillan, 1986). Williams was the first to build a formal equilibrium model of brokerage to prove that percentage commission contract can be incentive compatible.

The current paper offers the first direct empirical test of the agency problem in real estate brokerage. Our unique data set allows us to investigate whether real estate agents treat their own properties differently than they do those of their clients.

The next section of the paper presents the predictions of a simple search model of agent behavior that we formally develop in Appendix A. Our theoretical model suggests that an agent will sell his own house for a higher price than a comparable listed house he does not own. For any given listing price, the agent will also expend more effort for his own house. However, since a higher price makes it less likely for a contacted buyer to purchase the house, the time it takes to sell the agent-owned house may be longer or shorter than that of a house not owned by the agent. Section 3 provides an overview of the data. In Section 4 we discuss the estimation of the models. We estimate duration models to examine the effect on time on the market, probit models to generate inverse Mills ratios to correct for sample selection bias and endogeneity, selling price models corrected for sample selection bias, and selling price models corrected for selection bias and endogeneity using data from 306,869 multiple listing service (MLS) listings from several metropolitan counties in Texas during the years 1999 through 2002. Over the study period, 63% of the listings resulted in completed sales. An important feature of this data set is that over 9,911 (3.23%) of the observations involved agent-owned properties.

The empirical results presented in Section 5 show that agent-owned houses sell no faster than client or non-agent-owned houses, but they do sell at a price premium of approximately 4.5% above the price of a comparable house not owned by a real estate agent. This result that the owner-agent is able to obtain a higher selling price is consistent with the prediction of our theoretical model that an agent's interest does not match the interests of the seller. The final section of the paper, Section 6, summarizes the results and offers some concluding remarks.