Document View

This paper seeks to empirically determine whether more skilled and productive real estate salespeople, identified as full-payout or 100% commission agents, have a discernable, systematic effect on property selling price and its marketing time. Two types of agents, 100%ers and split-commission agents, are identified and controlled for in hedonic pricing and duration models in order to examine the relationship between incentives and agent performance. The results reveal that 100% agents sell their listed properties faster and at premiums. This paper may also help explain the contradictory findings of earlier research. [PUBLICATION ABSTRACT]

An extensive body of literature has developed on the market for real estate brokerage services. Much of this research has focused on the impact of broker intermediation on selling price and duration. Jud (1983), Jud and Frew (1986), Jud, Seaks, and Winkler (1996), Zumpano, Elder, and Baryla (1996), Elder, Zumpano, and Baryla (1999, 2000), Rutherford, Springer, and Yavas (2005), Huang and Rutherford (2007), Rutherford, Springer, and Yavas (2007), and Turnbull and Dombrow (2007) form a representative sample of these works. To date, however, most of this research has assumed no differences in agent skill levels.

More recently, it has been claimed by some researchers (Munneke and Yavas, 2001; and Allen, Faircloth, Forgery, and Rutherford, 2003) that differences among brokerage firms and or their agents suggest the possibility that some homes sell at premium prices and over a shorter time horizon than is the case with other broker-assisted transactions handled by less skilled or motivated agents. That is, do some types of agents consistently obtain higher prices for clients and customers and sell their properties faster? If so, buyers and sellers may not be indifferent with respect to their choice or use of salespeople and should seek out ways to identify the more skilled and more highly motivated agents.

Unfortunately, in the two papers mentioned above that directly examine this issue, the empirical findings are not reconcilable. In both cases, the way in which the agent is compensated has been singled out as the appropriate productivity marker or signal. Productive salespeople are differentiated from their less productive counterparts by how they are compensated for services rendered. Full-commission firms, such as Remax, are deemed to attract more productive salespeople because they will earn more money than if they shared their commissions with their brokerowners.1

Despite the expectation that full-payout agents would have a discernable effect on market outcomes, Munneke and Yavas (2001) find that the more skilled Remax agents have no long-term impact on selling price or time-on-the-market. Allen, Faircloth, Forgery, and Rutherford (2003), on the other hand, using a similar construct but different data, find that residential properties marketed by "more productive" agents are sold more quickly and at a premium relative to properties sold by "less productive" agents.

In both papers, when the performance of agents in full-payout firms is compared to their counterparts in the more traditional split-commission firms, Remax agents are used to represent all the 100% commission agents in the sample. Their performance is then compared to the performance of agents in all other firms, who are all deemed to be split-commission salespeople. In short, both studies assume that all 100%ers work with Remax.2

This categorization of the sample firms is understandable in light of the difficulty of identifying the specific type of compensation salespeople actually receive. However, many firms besides Remax offer 100% payout packages to their salespeople. Real estate brokerage firms commonly employ a portfolio of agents, some compensated on a split-commissions basis, while others have negotiated fullpayout arrangements, as part of their wealth management strategy. Additionally, practicing broker-owners face periodic fees in exchange for all of their generated commissions. Consequently, segregating the samples by firm rather than at the agent level may result in specification problems. For example, some of the agents working for non-Remax firms are almost certainly full-payout or 100% agents and are most certainly misclassified. Additionally, such a categorization that mixes more productive 100% agents with split agents, most certainly biases the results in the earlier studies towards being statistically insignificant.

This paper empirically reexamines the relationship between agent incentives and performance. In this work's estimations, agents are classified as 100% or splitcommission salespeople based upon an actual determination of each agent's specific compensation arrangement via a survey of qualifying brokers from the sample area. Unlike prior works, the firms they are associated with are not used to separate the sample. Employing comparable sold data from the Montgomery, Alabama market area, the empirical results indicate that 100% agents sell their listed properties faster and at premiums. Sections on incentive and agent performance, data collection and methodology, empirical results, and concluding comments follow in order.

Incentives and Agent Performance

Munneke and Yavas (2001) argue that the more productive agents self-select into 100% firms.3 They consider it a simple adverse selection problem; salespeople base their decision as to which type of compensation arrangement to choose using a straightforward break-even analysis. Highly skilled or motivated salespeople can maximize their earning by choosing a 100% commission arrangement once the commissions generated from their total closed volume, less their periodic desk costs, exceed the amount they would earn if they split their commissions with the firm.4

Using a theoretical model they attempt to show that because a Remax agent receives a larger commission from selling any given listing than does a splitcommission salesperson, the 100% commission agent will expend more effort on each listing than the traditional agent. Because 100% commission agents have greater performance coupled with incentives, they will attract more listings than split-commission agents. However, as these listings increase, the productive agents will not be able to spend as much effort on each of these listings. Hence, their marginal productivity declines as listings increase, with a long-run equilibrium result such that there will be no difference in either property price or selling time between split-commission and full payout agents. In effect, Munneke and Yavas' (2001) full-commission agents, in an effort to maximize their income, dissipate away whatever advantages they might have over split-commission agents, resulting in no discernable differences in agent performance.

If Munneke and Yavas' (2001) theoretical predictions concerning agent performance in the presence of added incentives are correct, then specification for compensation arrangements at the agent rather than at the firm level should prove insignificant into both hedonic and duration estimations.

Data and Methodology

Data

Previous research assumed that only Remax agents were full payout agents when, in point of fact, many non-Remax agents have 100% compensation arrangements. Consequently, what was really looked at was whether selling prices and marketing times were related in a systematic way to working with Remax, not to actual compensation structure. Segregating the sample by firm rather than at the agent level almost certainly resulted in classifying some full-payout agents as splitcommission agents. This could explain the conflicting results of earlier research. The data used here was aggregated at the agent level and is based upon each agent's specific compensation arrangement. Specifically, a survey of qualifying brokers from the Montgomery, Alabama area was conducted with each being asked to name 100% agents and practicing broker-owners in the MLS. Lists were cross-checked and names not universally nominated were individually investigated. This resulted in a final list of agents that face a periodic fee in exchange for all of their earned commissions, making a direct test of agent incentives and performance feasible.

The original data set consists of all conventional residential closings (2,716) that occurred during the calendar year 1998 within a contiguous area in Montgomery, Alabama, a time period consistent with the earlier works on this issue. The Montgomery Area Association of Realtors (MAAR) Multiple Listing Service (MLS) is the principal source of data for this study. Secondary information is provided by the Montgomery County Tax Assessor's Office. MAAR provides the data on selling price, selling time, location, and most of the physical characteristics of listed properties. During the time period sampled, the MAAR MLS listings did not contain any information on the age and square footage of listed properties; therefore, this information is obtained from the County Tax Assessor's office. A survey of local qualifying brokers allowed classification of individual agents on the basis of their compensation arrangement - 100%er versus split-commission.

To insure a complete set of housing characteristics for each sale, observations that do not appear in both the MLS and tax assessor's database are eliminated. Next, obvious data entry errors from the MLS database such as negative time-on-themarket, zero bedrooms or baths, etc. are dropped from the data set leaving 1 ,549 observations for the period in question. Variable definitions are presented in Exhibit 1 while Exhibit 2 provides summary statistics for the entire sample.

Modeling of Property Price and Duration

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

As most of the variables specified in these estimates are included in earlier works, their review adds little. Their discussion is omitted in favor of increased emphasis on the problem of self-selection, the use of the Weibull distribution, and the inclusion of three variables: EFIS, NOMKT, and 100%er.

It may well be the case that 100% agents as a group somehow systematically list properties that sell quicker and close at higher transaction prices than SPLIT agents, making it difficult to determine the marginal effect of 100% agents on price and marketing time. In fact, an examination of Exhibit 2 reveals casual evidence supporting this possibility. In a test for differences in means across agent type, 100%ers appear to list and sell properties that are generally endowed with proxies that favor higher final transaction prices (e.g., SQFT, JD, FP, and DOUBOVN) and shorter marketing times (e.g., BATH and FP). This is commonly referred to as self-selection and in order to test and control for this possibility, an IMR (inverse Mills ratio) is calculated from a first-stage probit for the occurrence of 100% agent listings and specified in Equations (1) and (2). IMR calculations come from a standard two-step estimation technique commonly referred to as the Heckit or Heckman process, which is described in detail in the Appendix.

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

A number of researchers have recently expressed concerns with using OLS estimates in any estimation of duration due to numerous inherent problems with this technique. These problems include, but are not limited to, the nonlinear nature of duration estimation, as well as non-normality of the model's error term. As an alternative, survival analysis, using a Weibull hazard function, is often employed to estimate marketing time.7 The flexible nature of the Weibull allows the function to be either monotonically increasing or decreasing. Additionally, a unique quality of the Weibull hazard function occurs when the scale parameter equals one. In this case, the Weibull hazard function reduces to the exponential hazard function. However, marginal interpretation of variables and measuring the explanatory power of models is somewhat easier to accomplish with OLS analysis, which is still favored among many researchers. In the interest of completeness, this work provides both estimation techniques, which are displayed in Exhibit 4.

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Two variables within X^sub i^ not commonly found in earlier works are EIFS and NOMKT. EIFS stands for exterior insulation and finish systems, a synthetic type of stucco that has been associated with water retention, mold, and termite infestations. An earlier work by Johnson, Salter, Zumpano, and Anderson (2001) found EIFS to be associated with both pricing premiums and extended marketing times and is entered into both specifications as a control. NOMKT is a dummy variable that controls for non-traditional broker-market properties. Johnson, Springer, and Brockman (2005) report that properties that are marketed by brokers prior to entering the MLS, but otherwise are recorded in the system ex post as a comparable sale, sell at a premium dictating NOMKT's specification in the pricing model. An additional note on this dummy variable and its specification in the duration model also seems warranted. Since these properties were sold before entering the MLS, they receive extremely truncated marketing times resulting in their artificially accounting for significant amounts of variability in property marketing time. Therefore, NOMKT is specified to control for variability in the duration estimations.8 It is important to understand that NOMKT is not causal when it comes to determining marketing duration, but instead is an artifact of how these properties are recorded in the MLS. As such, the variable is entered strictly as a control and has no economic interpretation.

The variable of interest, 100%er, provides the direct empirical test for this work. 100%er is defined as all agents that pay periodic fees in exchange for 100% of their earned commissions. This includes Remax agents, broker-owners, and agents at companies other than Remax that have the same full-compensation scheme. The analog for 100%er is SPLIT or those agents who work on a more traditional commission split arrangement with their firm. 100%er will reveal whether incentive induced, full-payout agents, when classified at the agent rather than the firm level, systematically impact property price and marketing time.

Empirical Results9

The Hedonic Pricing Model

Exhibit 5 formally reports the first-stage probit from the Heckit process for the hedonic pricing model. From this first stage estimation, the IMRHED is calculated and specified in the operational model for property price formally reported in Exhibit 3. IMRHED is statistically significant, indicating the presence of a selfselection bias among 100% agents and their choice in listings. In addition, it has the opposite sign of the variable of interest as is expected given its orthogonal relationship with 100%er.10

The explanatory power (R^sup 2^) for the model is 72.2%. Also, an F-statistic of 208.65 indicates the specified model is highly significant and an examination of correlation matrix does not suggest any serious colinearity problems. The model's variance inflation factors (VIF) are reported formally in Exhibit 3 as well. Property characteristics and location variables correctly sign and are robust with prior works. Most interestingly, full-payout agents appear to have a positive affect on price, with a marginal contribution of an additional 5.79% in final transaction price.

The Duration Models

Exhibit 6 formally reports the first-stage probit from the Heckit process for the modeling of property duration. IMRDUR is calculated from this first stage and specified in both the OLS and Weibull operational models, which are formally reported in Exhibit 4. The statistical significance of IMRDUR indicates the presence of self-selection. Furthermore, after orthogonalizing IMRDUR in order to avoid issues with colinearity, IMRDUR displays the opposite sign of 100%er as is expected.11

The results of the two duration estimates are robust and basically consistent with earlier studies. From the OLS estimate, the model's explanatory power is extremely high relative to other studies, with an P2 of 48.3%. This seems to be a result of specifying NOMKT. Again, it is worth noting that NOMKT properties should not be viewed as a proximate cause of shorter marketing times but instead as a byproduct of how data are reported to the MLS. Thus, NOMKT aids in the control of variability in the duration estimation but cannot be used for the purpose of drawing inference.

Turning to the variable of interest, 100%er is negative and significant in both models in Exhibit 4. Interpreting, the OLS coefficient reveals that 100% agents, on average, sell their listings over 41% quicker (or approximately 35 days sooner) than SPLIT agents.

Taken as a whole, the empirical results suggest that when controls are instituted across agents rather than at the firm level, differences in skill levels and motivation can impact market outcomes.

Conclusion

Research is evolutionary, and the identification of inconsistent results with prior accepted relationships is part of the regular paradigm building process. The findings in this work fill that role. Specifically, this work provides results that are inconsistent with Munneke and Yavas (2001) but consistent with Allen, Faircloth, Forgey, and Rutherford (2003). Both of these earlier works seek to determine the impact of 100% agents on property price and marketing time with membership as a participating agent in a Remax franchise being held out as the proxy of productivity. Unfortunately, many agents that pay their firm a periodic fee in exchange for 100% of their commissions work across many real estate firms rendering the empirical results in these two works suspect. However, these two works yield great insight into the issue of incentives and agent performance.

By controlling for compensation arrangements at the agent rather than the firm level, this work avoids the specification problem associated with earlier research and perhaps explains the differing empirical results. Also, it is important to note that this work specifies all agents (Remax members, 100%ers that work with other companies, and broker-owners) that pay a periodic fee in exchange for 100% of their commission as the variable of interest and as such provides a direct test of incentives' affect on performance. At this precisely defined level and in the presence of controls for self-selection, these 100% agents sell properties faster (approximately 35 days) and at premiums (approximately 5.8%) relative to traditional split agents. Furthermore, these results strongly suggest that intermediation effects can be induced with proper incentives.

Therefore, has a compensation arrangement been identified that appears to systematically induce favorable market outcomes for sellers in terms of price and marketing time? Preliminarily, the answer seems to be yes. However, a few questions, which are beyond the scope of this purely empirical work, remain before a definitive answer is provided.

First, an extension to Munneke and Yavas' (2001) agent's profit maximization function that includes listing servicing and maintenance costs could reveal an optimal number of listing, which could induce different theoretical predictions consistent with both the empirical findings presented here and those reported in earlier works. Specifically, when there are listing maintenance costs such as servicing problems associated with larger listing portfolios, an optimal number of listing may exist per agent type that is less than, equal to, or greater than the number of listings at which a 100% agent dissipates away their skill advantage, as predicted in Munneke and Yavas. Thus, in some settings 100%ers will dissipate away their skill advantage, while in others it would not be in their profit maximizing interest to take on additional listings. If this is the case, then identification of agent compensation arrangement as a marker of greater agent productivity and an ability to affect favorable market outcome for sellers devolves to an empirical test for each geographic market.

Second, part of the problem with empirical analysis is that it is often difficult to isolate the impact that market conditions may have on the statistical significance of explanatory variables, which is often the case with cross-section data. If real estate agent productivity assessments are undertaken when markets are expanding, the results may be very different than if the same estimations are undertaken during a depressed housing market. Only additional study, possibly with the use of panel data, or additional estimations under differing market conditions will ultimately resolve the question.

Finally, some new works are beginning to investigate the probability of sale (see Anglin, Rutherford, and Springer, 2003; Huang and Rutherford, 2007; Johnson, Benefield, and Wiley, 2007; and Rutherford, Springer, and Yavas (2005, 2007). Perhaps the inclusion of this additional information concerning marketing failure in future analysis on agent incentive structure and brokerage intermediation will better highlight performance differences, if any.

Wooldridge (2000) distills these works into a straightforward two-stage process (sometimes referred to as the Heckit or Heckman process).

Enlarge 200% Enlarge 400%

Construction of the first-stage probit and the subsequent calculation of its Inverse Mills Ratio, are designed to create a proxy for unobservable variability associated with the effect of interest. As such collinearity between the IMR and the effect of interest is quite often problematic. A simple solution, however, is to orthogonalize the IMR rendering it indecipherable for the purpose of economic interpretation but maintaining its ability to load self-selection variability. Finally, it should be noted that the purpose of the first-stage probit is to construct the IMR and not to build a model that predicts the occurrence of the effect of interest. As such, economic interpretation of this first stage is typically meaningless.

The Heckit process can now be summarized as follows:

1. Estimate the first-stage probit employing all the elements from the structural equation plus at least one element that is a significant predictor of the effect, i.e., Equation (A).

2. Specify the IMR from the first-stage probit in second-stage estimation, i.e., Equation (B).

3. If IMR proves significant, the selection process associated with the effect is present and properly controlled. If necessary, orthogonalize the ratio, and then interpret the marginal impact of the effect of interest accordingly.

4. If IMR proves insignificant, the selection process associated with the effect is not present and there is no need for a control. Drop the IMR from the specification. Re-estimate the structural equation and interpret the marginal impact of the effect of interest accordingly.

[Reference]  »  View reference page with links

References

Anglin, P.M., R.C. Rutherford, and T.M. Springer. The Trade-off between Selling Price of Residential Properties and Time-on-the-Market: The Impact of Price Setting. Journal of Real Estate Finance and Economics, 2003, 26, 95-11.

Allen, M.T., S. Faircloth, F. Forgey, and R.C. Rutherford. Salespersons Compensation and Performance in the Housing Market. Journal of the Academy of Finance, 2003, 1, 62-71.

Cubbin, J.S. Price, Quality, and Selling Time in the Housing Market. Applied Economics, 1974, 6, 171-87.

Elder, H.W., L.V. Zumpano, and E.A. Baryla. Buyer Search Intensity and the Role of the Residential Real Estate Broker. Journal of Real Estate Finance and Economics, 1999, 3, 351-68.

_____. Buyer Brokers: Do They Make a Difference? Their Influence on Selling Price and Search Duration. Real Estate Economics, 2000, 28, 337-62.

Greene, W.H. Econometric Analysis. Third Edition, New Jersey: Prentice-Hall, 1997.

Heckman, J.J. The Common Structure of Statistical Models of Truncation, Sample Selection and Limited Dependent Variables and a Simple Estimator of Such Models. Annals of Economic and Social Measurement, 1976, 5, 475-92.

_____. Sample Selection Bias as a Specification Error. Econometrica, 1979, 47, 153-61.

Huang, J. and R.C. Rutherford. Who You Going to Call? Performance of Realtors and Non-Realtors in a MLS Setting. Journal of Real Estate Finance and Economics, 2007, 35, 77-93.

Johnson, K.H., J.D. Benefield, and J.A. Wiley. Issues of Property Price, Duration, and Probability of Sale for Architectural Review Boards. Working Paper, Florida International University.

Johnson, K.H., T.H. Springer, and C.M. Brockman. Price Effects of Non-Traditionally Broker-Marketed Properties. Journal of Real Estate Finance and Economics, 2005, 31, 331-43.

Johnson, K.H., S.P. Salter, L.V. Zumpano, and R.I. Anderson. Exterior Insulation and Finish Systems: The Effect on Residential Housing Prices and Marketing Time. Journal of Real Estate Research, 2001, 22, 289-311.

Jud, G.D. Real Estate Brokers and the Market for Residential Housing. Journal of the American Real Estate and Urban Economics Association, 1983, 11, 69-82.

Jud, G.D. and J. Frew. Real Estate Brokers, Housing Prices, and the Demand for Housing. Urban Studies, 1986, 23, 21-31.

Jud, G.D., T.G. Seaks, and D.T. Winkler. Time on the Market: The Impact of Residential Brokerage. Journal of Real Estate Research, 1996, 12, 447-58.

Kiefer, N.M. Economic Duration Data and Hazard Functions. Journal of Economic Literature, 1988, 26, 646-79.

Knight, J.R. Listing Price, Time on Market, and Ultimate Selling Price: Causes and Effects. Real Estate Economics, 2002, 30, 213-37.

Miller, N.G. Time on the Market and Selling Price. Journal of the American Real Estate and Urban Economics Association, 1978, 6, 164-74.

Munneke, H.J. and A. Yavas. Incentives and Performance in Real Estate Brokerage. Journal of Real Estate Finance and Economics, 2001, 22, 5-21.

Rutherford, R.C., T.M. Springer and A. Yavas. The Impact of Contract Type on Broker Performance. Real Estate Economics, 2001, 29, 389-409.

_____. Conflicts Between Principals and Agents: Evidence from Residential Brokerage. Journal of Financial Economics, 2005, 76, 627-65.

_____. Evidence of Information Asymmetries in the Market for Residential Condominiums. Journal of Real Estate Finance and Economics, 2007, 35, 23-38.

Turnbull, G.K. and J. Dombrow. Individual Agents, Firms, and the Real Estate Brokerage Process. Journal of Real Estate Finance and Economics, 2007, 35, 57-76.

Wooldridge, J.M. Introductory Econometric: A Modern Approach. First Edition, Massachusetts: South-Western College Publishing, 2000.

Yavas, A. and S. Yang. The Strategic Role of Listing Price in Marketing Real Estate: Theory and Evidence. Real Estate Economics, 1995, 23, 347-68.

Zumpano, L.V., H. Elder, and E.A. Baryla. Buying a House and the Decision to Use a Real Estate Broker. Journal of Real Estate Finance and Economics, 1996, 13, 169-81.