Individual Agents, Firms, and the Real Estate Brokerage Process

See Benjamin et al. (2000) for a summary of the empirical literature.

Although the commission rate by itself increases average selling price by a greater or less extent as house supply is less or more elastic than demand, it can be shown that lower sellers’ transactions costs (including search costs) by itself lowers average selling price while lower buyers’ transactions costs by itself raises average selling price so that the combined effect on price is ambiguous a priori. See, for example, the housing and brokerage markets model in Turnbull (1996). This provides one rationale for the mixed empirical evidence on price effects of brokers. For example, Jud and Frew (1986) find broker-assistance increases the selling price relative to FSBO but Kamath and Yantek (1982) find no net difference. Nonetheless, regardless of the effect on sales price, the effect of brokerage on economic efficiency is clear: lower transactions costs unambiguously expand the market and improve efficiency.

Following the approach taken for firm level variables, an agent’s total transaction activity is the sum of his or her individual listing and selling activity. Therefore, agents who sell one of their own listings are credited with two transactions, one of each side of the transaction represented.

This does not imply that the two variables are collinear because many houses sold within a firm (Own Firm Sale = 1) are listed and sold by two different agents (Agent Own Sale = 0).

Additional biases in the data can also arise for lengthier sample periods—like changing geographic or functional specializations of agents over time. We found no systematic change in agent specializations during our sample period, which suggests that the 2 year horizon is short enough to minimize this potential problem.

The calculations described earlier in the paper for all of the density and competition variables include all MLS sales for the entire metropolitan area over the period July 1995 through March 1998. Therefore, these variables include all relevant out-of-sample transactions, that is, houses in areas bordering on our sample geographic areas and houses listed before or after the sample time period which overlap with our sample period.

The MLS roster for the time period in our study lists over 1,400 agents for the metropolitan area. The sample area of this study has approximately 400 agents represented. Not all agents are included in this study because of data cleaning issues. For example, all transactions associated with one particular office were removed because all of the resident agent activity was credited to the office manager.

The transactions included in our sample account for approximately 95% of all MLS transactions in the areas covered by our sample. Tests for data pooling imply that the 5% of sold houses with marketing times longer than 6 months should not be pooled with the rest of the sample.

We also investigated alternative specifications with variables distinguishing non-specialists on both sides of the transaction. We further subdivided the omitted dummy variable category of non-Mainly Listing into separate non-specialist and selling specialist dummy variables. A similar construction was applied to the omitted dummy variable category non-Mainly Selling. These additional sets of dummy variables are not significant and have no effect on the other parameter estimates. The results imply that a selling specialist who lists a house (or a listing specialist who sells a house) has the same effect on price and marketing time as the omitted category of non-specialists. We are grateful to an anonymous referee for drawing our attention to this question.

Recall, however, our caveat regarding the fact that broker or agent assistance benefits buyers and sellers even when there are no observable effects on selling price.