Two sides of dual agency: evidence from homebuyers and transactions

Abstract

This study gains insights to the motivating causes of dual agency transactions in residential real estate by examining two distinct sources of data. The first is evidence from the National Association of Realtors® (NAR) homebuyers’ survey; the second is multiple listing service (MLS) transaction data. The survey evidence is used to examine buyer characteristics and search methods related to the outcome where a homebuyer is unrepresented by a buyer broker, who would exclusively represent their interests. Results suggest that certain factors contribute to both the likelihood that a buyer will be unrepresented and the incidence of dual agency, including homebuyer experience and inexperience, geographic familiarity by brokers and buyers and the channels used in marketing residential products to generate initial contacts with potential buyers. The matching of results from the empirical analysis using the MLS transaction data and the NAR homebuyer survey data adds confirmation to our findings. Even as dual agency creates opportunities for efficiency gain in the real estate transaction, the empirical results have implications for the consequences of homebuyer involvement in the initiation of dual agency, strategic behaviour recommendations for brokers as well as the possibility to policy adjustments.

Keywords:

• brokerage
• disclosure
• representation
• efficiency

JEL Classification:

• R12
• R31

Acknowledgements

The authors are very grateful to assistance from the National Association of Realtors®, and in particular Dr. Paul Bishop, for providing data for external use from the annual survey of homebuyers and sellers.

Notes

1. The year 2006 represents the midpoint of the sample horizon for the MLS transactional data considered in this study.

2. Responses of ‘Yes, an oral arrangement’ and ‘Don’t know’ are not considered in the estimation of BuyerRep_Written.

3. A third option is to omit responses of ‘Don’t know’, comparing either version of ‘Yes’ with the explicit ‘No’. The results based on this alternative construct to the BuyerRep_Any dependent variable are highly similar and qualitatively unchanged.

4. 26.5% = 1 − exp(−.308), where −.308 is the estimated coefficient of open house in Equation (2). 55.2% = 1 − exp(−.804), where −.804 is the estimated coefficient of for-sale sign in Equation (2).

5. In the construction of the local variable we follow Brastow, Springer and Waller (2011). GIS technology is used to determine whether a broker’s listings are spatially clustered, to define that area, and to identify properties that are listed within it. Local takes on a value of one for listings that occur inside the listing broker’s territory. For brokers with spatially clustered listings, the algorithm fits a 95% confidence ellipse over the listing locations. A minimum volume of 40 listings is required to establish this area.

6. Competition from other listings is measured by the listing density variable, which is a distance-weighted sum of days on market from competing properties following Turnbull and Dombrow (2007).

7. The national brokerage firms include Virginia-based franchises of Assist-2-Sell, Century 21, Coldwell Banker, ERA, Exit Realty, Prudential, RE/MAX, Real Estate 2000, Real Living, The Realty Dot and United Country. Local firms are all others not listed here.

8. Based on the logit model construct, the coefficient can be interpreted as an increase in the odds ratio for dual agency. The change in the odds ratio is calculated as exp(.212) = 1.236.