ArticlePDF Available

Is Dual Agency in Real Estate a Cause for Concern?


Abstract and Figures

We study dual agency in residential real estate, where the same agent/agency represents both the buyer and seller. We assess the extent to which dual agency suffers from an inherent conflict of interest, where the dual agent furthers the interest of one client at the expense of the other client’s, as well as principal-agent incentive misalignment where the agent furthers her own interest at the expense of one or both clients. And, we examine how these incentive conflicts affect agent behavior and transaction outcomes. To do so, we analyze 10,891 residential real estate transactions in Long Island, NY, from 2004- 2007. Specifically, we (i) identify how dual agency is correlated with house prices and time-to-sale, (ii) describe and assess agent behaviors that could generate these correlations, and (iii) provide some intuition as to the economic effects of prohibiting dual agency in real estate transactions. We find that the incidence of dual agency is uncorrelated with sale price and negatively correlated with time-to-sale. However, on very fast deals, list prices and sale prices are significantly higher on houses sold via dual agency. These findings are consistent with first-resort selling (agents first showing houses to in-house buyer clients) and strategic pricing (agents inducing their seller clients to set a higher list price in anticipation of an internal client agreeing to it) on some deals, in conjunction with agents leaning on sellers to accept a lower sale price on other deals. First-resort selling is indicative of incentive misalignment, while the latter two behaviors reflect a conflict of interest: strategic pricing transfers surplus from the buyer to the seller, and leaning on the seller transfers surplus from the seller to the buyer. Further, our results indicate little difference between dual-agent (same agent) and within-agency (same agency, but different agent) deals. Our findings provide some evidence of distorted outcomes associated with dual agen
Content may be subject to copyright.
Electronic copy available at:
Is Dual Agency in Real Estate a Cause for Concern?
Vrinda Kadiyali, Jeffrey Prince, and Daniel H. Simon
This version: August 2011
Vrinda Kadiyali is at Johnson School of Management, Cornell University. Jeffrey Prince is at Kelley School of
Business, Indiana University. Daniel Simon is at School of Environmental and Public Affairs, Indiana University.
They can be reached at,, and The authors thank
Ellie Cohn, Aaron Moss, David Simon, and Michael Waldman for very helpful discussions, the Johnson School
Dean’s Research Lunch and International Industrial Organization Conference participants for comments, Norman
Mendolsohn for data, and Lars Backstrom for research assistance.
Electronic copy available at:
We examine the effects of the regulation of dual agency, where the same agent or agency represents both
the buyer and the seller, in residential real estate transactions, for 10,888 transactions in Long Island, New
York in 2004-2007. Dual agency involves agents’ potential conflicts of interest and misaligned
incentives, leading some states to prohibit the practice. However, dual agency can also benefit buyers and
sellers, by expanding the pool of houses and buyers that an agency can try to match with each other, and
by offering information and transaction efficiencies. We find that dual agency has an overall null effect
on sale price, but includes two opposing forces where buyer and seller interests might be compromised.
The link between dual agency and timing of sales is less clear. These findings are robust to endogeneity
bias. Overall, we find little evidence that prohibiting dual agency will increase welfare.
Electronic copy available at:
1. Introduction
In this paper, we study the effects of the regulation of dual agency, where the same agent,
or agency, represents both the buyer and the seller, in the residential real estate market. Most
Americans participate in this market as buyers and sellers. Typically, both the buyer and the
seller use an agent, relying on the agent to guide the transaction, make assessments on the value
of houses, and conclude transactions without the usual warranties or return policies commonly
expected for other, far smaller, transactions.
There are several ways in which agents’ incentives may be misaligned in this market (see
Yinger 1981, Salant 1991, Bryant and Epley 1992, Wheaton 1990, Yavas et. al., 2001, etc.).
First, both buyers’ and sellers’ agents are compensated as a percentage of the final sale price.1
While this aligns the seller’s interests with those of her agent, the agent’s interests appear to be at
odds with the buyer’s interests in this compensation scheme.2 Second, the assumption that
agents want to maximize their commission might oversimplify the underlying misalignment of
incentives. Research has shown that agents are more willing than owners to accept a lower price
in order to reduce time to sale (Levitt and Syverson, 2005, Rutherford et. al. 2005, Hendel et. al.
2007). All else equal, faster sales benefit both buyers and sellers. However, the lower
equilibrium sale price is not in the seller’s interest. Similarly, faster time to sale might not benefit
buyers (despite lower prices) if their preferences are compromised to hasten sales.3
In addition to the misaligned incentives that arise in standard, cross-agency deals, policies
that allow dual agency further distort agents’ incentives and their behaviors, in three ways. First,
the agent has an incentive to steer buyers to her own (her agency’s) listings, or to steer sellers to
her own (her agency’s) buyers, rather than steering the buyer (seller) to the house offering the
best value (buyer with the highest willingness to pay). For example, an agent could steer a seller
to internal buyer clients by showing the client’s house to internal buyers before showing it to
other agents’ (agencies’) clients. Similarly, an agent might show buyer clients her own listings
1 An interesting question is why commission-based agency persists in real estate markets when there are known to
be inefficiencies in this system. See Jares, Larsen, and Zorn (2000) for a discussion, and an alternative that has the
seller’s agent buy the property and have the put option of returning the house back to the seller. See also Bernheim
and Meer (2008) for whether brokers perform enough functions to justify their commissions.
2 Laws in some states have attempted to ensure that agents representing buyers clearly owe fiduciary responsibility
to the buyer rather than be an agent for the seller or only loosely represent the buyer’s interest. As expected, such
laws have led to a drop in the sale price (see Curran and Schrag, 2000).
3 See Genesove and Mayer (2001) and Garmaise and Moskovitz (2004) for examples of other (consumer behavior)
suboptimalities in real estate markets.
before showing them listings of other agents (agencies). Second, dual agents (agencies) could
disclose confidential information to one of their clients, helping them in either the search or
negotiation phase of the purchase process. For example, an agent may influence a seller client to
set a higher list price by sharing information with the seller about an internal buyer’s reservation
price, a practice we refer to as strategic pricing. Or, once negotiations have begun, the agent
might provide the buyer (seller) with information about the seller’s reservation price (buyer’s
willingness to pay), providing the buyer (seller) with an advantage in negotiating a lower
(higher) price. Third, in the negotiation phase, a dual agent (agency) may pressure or “lean on”
either the buyer (to accept a higher price) or the seller (to accept a lower price) to hasten the sale.
Despite the conflicting incentives and potential resultant harm (to buyers and sellers)
from dual agency, allowing dual agency also offers a variety of potential benefits. Dual agency
provides information and transactional efficiencies: it might lead to quicker, more efficient
matches by enabling an agent (agents in the same agency) to combine (share) information on
seller and buyer preferences, and paperwork on transactions might conclude faster. Further,
allowing dual agency expands the pool of houses that an agent can show a buyer and the number
of buyers that an agent can try to match with a seller. For example, consider the case in which a
buyer is interested in a house for which the buyer’s agent (agency) is also the listing (seller’s)
agent (agency). If dual agency is prohibited, the agent would not be able to sell the house to the
buyer because of her status as the buyer’s agent.
As the above discussion illustrates, when dual agency is permitted by law, it creates
incentives for agents to engage in behaviors that may affect the sale price and speed of
transaction outcomes. The goal of this paper is to identify the effect of laws regulating dual
agency on the price and speed of transactions. To accomplish this goal, the ideal data set would
be panel data from multiple states, where some states changed the legal status of dual agency,
either legalizing or outlawing dual agency. Using these data, one could compare outcomes for
houses sold when dual agency is legal with those for houses sold when dual agency was not
legal. A difference-in-differences estimate on this panel would provide an estimate of the
average treatment effect of the legal status of dual agency. Lacking such data, we instead
compare outcomes for dual-agency transactions with those for cross–agency transactions in a
market where dual agency is legal. This approach yields an estimate of the effect of a house
being sold via dual agency (“the treatment”) on the houses sold via dual-agency transactions
(“the treated”). We label this the effect of the treatment (dual agency) on the treated (houses sold
via dual agency), or ETT.
There are two requirements for our estimate of ETT to provide an unbiased estimate of
the impact of the legal status of dual agency. First, in order for us to obtain an unbiased estimate
of the ETT, the incidence of dual agency must be uncorrelated with unobservable house
characteristics. That is, we must assume agents are not selecting houses for dual agency based on
house characteristics that we cannot observe in our data. To satisfy this requirement, we use a
very detailed set of observables, and importantly, we use methods in Altonji et al. (2005 & 2008)
for bounding the size of endogeneity bias in our results. Second, dual agency should only affect
transactions concluded via dual agency. If this assumption holds, then an unbiased estimate of
ETT provides an accurate measure of the effect of changing dual agency law. We show that this
requirement is met when estimating the effect of dual agency on sale price, but perhaps less so
for time-to-sale, rendering our results for the latter to be less conclusive.
We conduct our analysis using data from 10,888 completed residential real estate
transactions in Long Island, New York, during the time period 2004-2007. New York state law
permits dual agency as long as the agent (agency) discloses dual representation to both parties.
We focus on two important transaction variables – the sale price and length of time-to-sale.
Two recent papers, Gardiner et. al. (2007) and Evans and Kolbe (2005), have examined
the effects of dual agency. The former paper examines the effect of a law change in Hawaii in
1984 requiring full disclosure of dual agency. The authors find that dual agency reduced the sale
price, but the effect was much smaller after the legislation (8.0% versus 1.4%). In addition, dual
agency reduced the time to sale by about 8.5% pre-legislation and 8.1% post-legislation. Evans
and Kolbe (2005) look at the effect of dual agency on price appreciation for houses that are sold
twice. They find that dual agency in the first transaction has no impact on price appreciation.
They also find very limited evidence that dual agency in the second transaction has a negative
effect on price appreciation.
Our paper differs from these papers on dual agency in three important ways. First, these
other papers have not attempted to identify the effect of a change in the legal status of dual
agency. This is likely due to data limitations, as data surrounding a natural experiment of a
change in legality of dual agency are difficult to find. As discussed above, while we, too, lack
such data, we identify the impact of dual agency through other means (while recognizing
conditions which might prevent us from accurate estimation).
Second, we explore the mechanisms through which dual agency affects transaction
outcomes. To do so, we describe specific agent behaviors that would both increase the likelihood
of dual agency and drive different outcomes in dual-agency transactions. We then investigate
empirically whether these behaviors are present in our data.
Third, while other papers have not made this distinction, in our empirical analysis we
distinguish three types of dual agency: (1) Dual-agent deals, in which the same agent represents
both buyer and seller; (2) Within-branch deals, in which two agents working for the same branch
of an agency represent the buyer and seller; and (3) Within-agency deals, in which two agents
working at different branches of the same agency represent the buyer and seller. As we will
show, these three types of dual agency have different incentive structures and transaction
outcomes. These distinctions have important policy implications because states vary in the types
of dual agency they allow.
We find that, overall, dual agency’s legal status has no net effect on sale prices. However,
through further analysis, we find that this null effect is the combination of two countervailing
effects. The first is agents exploiting information on internal buyer clients’ preferences and
willingness to pay, to help seller clients set a higher list price and ultimately obtain a higher sale
price. The second is agents favoring the buyer over the seller in the negotiation phase, by some
combination of leaning on the seller to accept an offer from an internal buyer and/or disclosing
confidential information to the buyer. We see both types of behaviors in dual-agent deals, while
we only see leaning/confidential information disclosure in the within-branch deals, and we only
see strategic pricing in the within-agency deals. Importantly, our analysis of endogeneity bias
strongly suggests that these findings are not driven by correlation between the incidence of dual
agency and unobserved house characteristics.
We also find that allowing dual agency reduces time-to-sale; houses sold via dual agency
sell about eight percent (or six days) faster than houses sold across agencies. We find this effect
for all three types of dual agency. While this is a significant effect, we cannot confidently
maintain all the necessary requirements for us to interpret this as the effect of the legal status of
dual agency. Hence, we treat these findings as suggestive only.
Overall, while we do find evidence of consequences from dual agency regulation, via
strategic pricing and leaning on the seller and/or confidential information disclosure to the buyer,
the net welfare effects of dual agency’s legal status appear largely benign, at least for the
majority of dual agency deals we observe in our data. Consequently, our results provide little
support for the prohibition of dual agency in any form (either within-agency, within-branch, or
dual agent).
The remainder of the paper is organized as follows. In Section 2, we discuss agent
incentives in cross-agency and dual-agency transactions. In Section 3, we describe and assess
legislative and market approaches for regulating dual agency in the housing market. In Section 4,
we describe our data, and in Section 5, we present our empirical results and discussion. We
discuss policy implications of our results in Section 6 and conclude in Section 7.
2. Dual Agency: Incentives and Impact on Transaction Price and Speed
2.1: Cross Agency Incentive Issues
To better isolate the incentives and impact of dual agency in residential housing
transactions, consider the bargaining incentives for agents in a standard, cross-agency
transaction. There are several principal-agent issues in the real estate market (see Yinger 1981,
Salant 1991, Bryant and Epley 1992, Wheaton 1990, Yavas et. al. 2001, etc.). First, both agents
are compensated as a function of the final sale price. This misaligns the buyer agent’s incentives;
the buyers agent (as well as seller’s agent) benefits from a higher sale price, even though this
represents a loss to her clients.4
Second, agents do not simply want to maximize their commission. The seller’s agent
must evaluate the additional commission earned from a higher price against the cost of waiting
longer to conclude the deal. Levitt and Syverson (2005) show that real estate agents wait longer
However, this misalignment of incentives is counterbalanced by
the desire to develop and maintain a good reputation, as a means to attract more clients. This
desire to protect one’s reputation creates a strong incentive for buyers’ agents to bargain
aggressively for a low price (resulting in a low commission) on behalf of their clients. The seller
agent’s incentives are not misaligned; bargaining aggressively on behalf of her clients results in
higher prices and higher commissions.
4 Laws in some states have attempted to ensure that agents representing buyers clearly owe fiduciary responsibility
to the buyer rather than be an agent for the seller or only loosely represent the buyer’s interest. As expected, such
laws have led to a drop in the sale price (see Curran and Schrag, 2000).
to sell, but obtain a higher price, when they sell their own houses compared to selling clients’
Hendel, Nevo, and Ortalo-Magne (2007) find confirming results homeowners who
sell their homes themselves (instead of using a broker) wait longer but also obtain a sale price
premium over houses sold via agents. These results are driven by the fact that agents capture
only a small fraction of the additional proceeds from each sale, while incurring much of the
additional cost of marketing the houses (Levitt and Syverson, 2005, Rutherford et al. 2005).
Therefore, existing research suggests that selling agents benefit from concluding transactions
faster, rather than holding out longer for a possibly higher price. Consequently, misaligned
incentives can have an impact on both the sale price and time-to-sale in standard, cross agency
2.2: How and Why Does Dual Agency Arise?
The process of buying and selling a house may be thought of as a two-stage, search and
negotiation game between buyers, sellers, and their agents. In the first stage, buyers (sellers)
search for homes (buyers), assisted by their agents. In the second stage, after a match is found,
the buyer and seller negotiate through their agents. When dual agency is permitted, the
transaction can conclude either via dual agency or as a standard, cross-agency transaction.
Legalizing dual agency creates incentives for agents to engage in behaviors that increase the
probability of dual agency and distort transaction outcomes in a variety of ways.
An agent can increase the probability of a dual agency transaction by influencing clients’
actions in either of three ways: (1) Manipulating the search process by steering seller (buyer)
clients to internal buyers (sellers); (2) Manipulating either the search or negotiation phase by
disclosing private information; and (3) Manipulating the negotiation process by leaning on the
seller (buyer) to accept the internal buyer’s (seller’s) offer. These agent behaviors can affect our
two outcome variables of interest, the sale price and the time-to-sale.
In (1), an agent uses her influence on her clients’ search process to steer them towards
dual-agency deals. For example, a seller’s agent might steer the seller to internal buyers by
5 A parallel situation is found in financial services markets where brokers are divided into agency and non-agency
brokers. Agency brokers are precluded from buying for themselves, and are only allowed to buy on their client’s
behalf. Non-agency brokers usually offer lower transactions fees but also offer lower prices to sellers because of
their incentive to buy low from sellers and sell high to their buyer clients (Harris, 2003). When non-agency brokers
are better informed than their own clients, they trade on the value of their information. This finding is parallel to
Levitt and Syverson (2005).
showing the house to internal buyers before engaging in broader marketing efforts (e.g. listing
the house on MLS or putting a “For Sale” sign in the front yard). In the case of a buyer’s agent,
steering clients to internal sellers would involve showing internal listings first, and emphasizing
an internal listing’s positive features relative to comparable houses.
In (2), an agent influences the search or negotiation process by disclosing confidential
information to either the buyer or the seller. Agents know more about the preferences (house
characteristics) of their own buyer (seller) clients, while agents in the same agency are likely to
share information about available listings and client preferences.6
In the negotiation phase, the agent (or agency) already represents both sides in a
negotiation. The agent (agency) may have achieved dual status by steering clients toward each
other (as in (1)), or the dual agency may have arisen spontaneously. For example, an agent
(agency) may become a dual agent (agency) if a buyer, who is represented by an agent, discovers
a house listed by that agent (another agent that works for the same agency as the buyer agent). In
the negotiation stage, dual agents (agencies) face a clear conflict of interest due to the zero-sum
nature of price negotiations- it is impossible to simultaneously obtain the lowest possible price
for the buyer and the highest possible price for the seller in the same transaction.
In the search phase, the seller’s
agent might disclose private information about an internal buyer’s preferences to the seller, in
order to influence the list price accordingly - we call this behavior strategic pricing. For example,
an agent representing a client with a three-bedroom, two-bath house with cathedral ceilings, an
eat-in kitchen, and a large back yard, might influence the seller to set a list price at $475,000
rather than $450,000, by disclosing to the seller that she has a client (or is aware of a colleague
with a client) that is interested in finding a house with these particular features, who is willing to
pay as much as $475,000. While information may be available about external buyers, it seems
likely that such information would be more prevalent for internal buyers, making strategic
pricing more likely to occur on deals that ultimately close via dual agency.
In this case, the dual agent can disclose private information to either the buyer or the
seller, giving that client an advantage in the negotiation process. For example, an agent might
inform the buyer that the seller’s reservation price is $270,000, helping the buyer to make a
6 Conversations with real estate agents who have worked at multiple agencies indicate that agencies often hold
regular meetings in which agents describe their available listings and search parameters of their buyer clients to see
if any matches can be made internally. Beyond these formal meetings, agents share information with colleagues
about their clients in an effort to make matches.
lower successful bid. Alternatively, the agent might disclose to the seller that the buyer’s WTP
for the house is $320,000, bolstering the seller’s willingness to hold out for a higher price.
In (3), like in (2), an agent (or agency) is already representing both sides in a negotiation.
Once the agent (agency) has achieved dual status in the negotiation process, she (it) can help
close the dual-agency deal by leaning on the seller or buyer to accept the other’s offer. The dual
agent (agency) may either lean on the seller to accept a lower price from an internal buyer than
what she could expect to obtain from an external buyer, or the dual agent (agency) can lean on
the buyer to accept a higher price from an internal seller than what she could expect to obtain
from an external seller of a similar house.
Agents face a variety of incentives to engage in these behaviors, which increase the
likelihood of a dual agency transaction. First, and perhaps most importantly, dual agency enables
an agent (agency) to capture the commissions for both the buyer’s agent and the seller’s agent, in
one transaction. Although an agent (agency) might achieve the same outcome via two separate
transactions, agents tend to have less exclusive and shorter contracts with their buyer clients.
Therefore, an agent (agency) might not capture both commissions if the buyer does not buy a
house from a seller that is represented by the same agent (agency). Second, by steering clients to
internal buyers/sellers, disclosing confidential information to clients, and by leaning on clients to
accept offers from internal clients, dual agents (agencies) can speed up the search and/or
negotiation process, allowing the agent(s) to earn her (their) commissions sooner. Third, for
similar reasons, dual agency reduces the amount of time and effort agents must spend
researching, marketing, and showing houses to clients. Finally, dual agency allows agents the
opportunity to exploit inside knowledge about their buyers’ preferences and their sellers’ houses
to facilitate a better match, yielding a correspondingly higher sale price (and commission), and a
faster sale.
These incentives for agents to engage in behaviors that increase the likelihood of closing
a dual agency deal are largely the same for all three types of dual agency deals, with one
important exception. In within-branch and within-agency deals, an agent may offer her listings to
fellow-agents of the same branch (agency), in the hopes that they will reciprocate in the future.7
7 Although there may be favor trading across agencies, it is easier to sustain such favor trading within an agency.
Also, there do not appear to be any incentives provided by agencies to encourage more deals in-house than cross-
In dual-agent deals, an agent can directly capture all the benefits herself rather than share them
with a colleague. For this reason, the incentives to engage in dual agency are likely to be greater
for dual-agent deals relative to either type of dual-agency (within-branch or within-agency)
There are also some disincentives for engaging in behaviors that increase the likelihood
of a dual-agency deal. Most importantly, the pool of internal buyers and sellers is much smaller
than the external market. Therefore, selling agents may have to wait longer to find an internal
buyer, or sell at a lower price to attract an internal buyer. Similarly, buyer agents may have to
wait longer to find an internal seller with a house for which the buyer has a high WTP, or settle
for a lower (sale price and) commission. Nonetheless, when the above incentives outweigh the
disincentives, agents are likely to engage in these behaviors to increase the probability of closing
a dual agency deal.
2.3: Price and Speed in Dual Agency Transactions
Agent behaviors can affect the price of dual agency transactions in both the search and
the negotiation phase. If agents steer sellers (buyers) to internal buyers (sellers) during the search
stage, restricting the pool of buyers (sellers) may have a negative effect on the price by creating
sub-optimal matches between buyers and sellers. Because the agent does not look to find the
buyer with the highest WTP or the seller offering the house with characteristics that best match
the buyer’s preferences, instead focusing on finding an internal buyer or seller, it is likely that
such matches will create less value, generating lower sale prices.
In contrast, a higher sale price will result if an agent discloses confidential information
about an internal buyer’s preferences to the seller during the search phase (in order to
strategically set a higher list price). However, this higher sale price is conditional on finding an
internal buyer with preferences that match the seller’s house. In the absence of such a well-
matched buyer, the agent cannot exploit her knowledge of internal buyers’ preferences in this
When a dual agency transaction results from agents leaning on clients or disclosing
information to clients in the negotiation phase, the direction of the impact on sale price depends
on which client the agent chooses to lean. If the agent leans on the buyer to accept the internal
seller’s offer (discloses information to the seller), then this should have a positive effect on price
(and vice versa).
It is important to note that if agents engage in any of these behaviors steering their
clients, disclosing confidential information to them, or leaning on them - in an effort to sell a
house via dual agency, but fail to match an internal buyer with an internal seller, then these
behaviors will have no effect on the transaction price. These agent behaviors only influence the
transaction price when the agent is successful in generating a dual-agency transaction. In
contrast, all of these agent behaviors (steering, disclosing confidential information and leaning)
may affect the speed of both dual-agency and cross-agency transactions, though not always in
the same way. As we will discuss further in Section 5, this lack of clear distinction between
dual-agency and cross-agency transactions will influence our ability to empirically isolate dual
agency effects on time-to-sale. Consider each of these three agent behaviors as they occur in the
search or negotiations phase of the transaction.
In the search stage, if agents steer sellers (buyers) to internal buyers (sellers) by showing
listings to internal buyers before external buyers (showing buyer clients internal listings first,
before showing them other agencies’ listings), this will delay time-to-sale because of the smaller
pool of internal buyers (sellers). This delay happens regardless of whether the house is ultimately
sold via a dual agency or a cross-agency transaction. However, the incidence of dual agency may
be correlated with (but will not cause) faster sales, as houses with good unobservables are both
more likely to sell fast and more likely to be sold via dual agency. Similarly, because strategic
pricing will only occur in cases where the agent believes that she has a well-matched buyer and
seller, we expect dual agency to be correlated with, but not cause, faster sales if the internal
match is made (i.e., if the house is sold via dual agency). However, by raising the list price,
strategic pricing delays the time to sale in both dual-agency and cross-agency transactions.
In the negotiation phase, both disclosing information and leaning should reduce the time-
to-sale in dual-agency transactions, as the agent encourages the client to accept an offer rather
than continue negotiations (or continue searching). However, if the agent is not successful in
closing the dual-agency transaction, then providing information to the client or leaning on the
client may delay the time-to-sale if this behavior causes the negotiations to break down,
requiring the search process to continue.
Given that agents can give preferential treatment to either buyers or sellers, by either
providing confidential information to the client or by leaning on the other client, the question
arises: who will agents favor – the buyer or the seller? The vast majority of seller’s agreements
are exclusive. In contrast, the relationship with buyers is looser and less formal, often even
without signed agreements (see Brown and Yingling, 2007). Therefore, we might expect the dual
agent (agency) to exploit this asymmetry by leaning on the seller to accept a lower price from an
internal buyer and by providing confidential information to the buyer that would bolster his
ability to negotiate a lower price. On the other hand, because commissions increase with sale
price, the incentive to lean on a buyer to accept a higher price from an internal seller, and the
incentive to strategically set higher list prices by exploiting information about internal buyers’
preferences, may be greater than the incentive to lean on a seller. Therefore, the direction and
magnitude of the effects of information disclosure and leaning are empirical questions (see
Section 5.3)
Beyond its effect on transaction outcomes through its influence on agent behaviors, dual
agency regulation may also affect transaction prices and speed by allowing for better information
flows and transaction efficiencies. Internal information sharing can result in better and quicker
matches between internal buyers and sellers. This should have a positive effect on sale prices
(independent of strategic list pricing), since available houses are better matched with buyers.
Additionally, this information sharing will have a negative effect on time-to-sale, since the added
information leads to quicker matches. Moreover, once a match is made, dual agency may offer
efficiencies in processing and transferring documents, enabling a faster transaction.
Finally, we can ask whether there is reason to expect any differences in the effects of
dual-agent, within-agency, and cross-agency deals. As mentioned above, it seems clear that,
relative to within-agency and cross-agency deals, agents will feel most strongly any incentives to
pursue dual-agent deals. Consequently, we would expect to find the greatest effects of dual
agency (if any) for these deals. However, we believe this logic is incomplete when it comes to
strategic pricing. Specifically, while a dual agent may have the strongest incentive to engage in
strategic pricing, she likely has very limited opportunity. That is, the likelihood that one of her
buyer clients (a relatively small pool of clients) matches especially well with one of her seller
clients is probably quite small; but the likelihood that one of her colleagues’ buyer clients (a
much larger pool of clients) matches especially well with one of her seller clients is likely much
higher in comparison. Consequently, we might expect to see the most pronounced effects of dual
agency associated with strategic pricing for deals with the largest potential number of matches,
i.e., within-agency deals. We consider the possibility of the different types of dual agency
exerting differential effects in our empirical analysis.
3. Legislative and Market Responses to Dual Agency
There have been a wide range of public and private responses to the challenges posed by
dual agency. As we discuss below, these responses have focused on the inherent conflict of
interest that the dual agent/agency faces in the negotiation stage.
3.1: Legislative Responses to Dual Agency Conflict of Interest
Olazabal (2003) provides a thorough summary of various types of agency relationships in
real estate. As she describes, for most of the twentieth century, sellers listed their property in
multiple listing services with a “listing broker.” A “selling broker” would show the property to
the buyer, but the selling broker did not represent the buyer. Both the listing and selling broker
received their commission from the seller, and neither owed any fiduciary responsibility to the
buyer. This left the seller open to legal liability due to agent misbehavior. While such situations,
by definition, do not have dual agency conflicts, they are certainly subject to misaligned
incentives, as discussed in Section 2.1.
Over time, legislation and market forces have resulted in more buyer representation, as
well as a reduction in seller liability for actions of agents. Despite these changes, dual agency has
persisted. However, responding to criticisms that the resulting system of dual agency was not
transparent and created conflicts of interest, beginning in the early 1990s, many states passed
laws that placed restrictions on dual agency. As a result of these changes in state laws, most
states’ dual agency policies currently fall into one of three categories: dual agency with
disclosure, designated dual agency, and transactional brokerages.
New York is one of several states that allows for dual agency with disclosure. These
states have chosen to allow the practice of dual agency as well as dual agent to continue.
However, all of these states now require agents to disclose their dual-agency status to both the
buyer and the seller. It is not evident that disclosure alone will reduce the conflict of interest,
given the manipulations of the search and negotiation processes might be subtle enough to go
undetected by the harmed party (buyer, seller or both).
A smaller number of states (e.g., Colorado, Maryland) allow for designated dual agency.
This allows for dual agency (within-office or within-agency) but not for dual agents. In
designated agency, a brokerage firm may designate one agent to represent the seller and a
different agent to represent the buyer. However, as discussed in Section 2.2 above, conflicts of
interest and misaligned incentives remain for within-branch or within-agency deals, when favors
are traded by agents of a single agency.
Finally, other states (e.g., Florida) do not allow the buyer and seller agent to be from the
same agency, but allow a new type of entity, a transaction brokerage, to represent both the buyer
and the seller in a transaction. A transaction brokerage provides many of the same services as a
real estate agency, but does not legally represent either party in the transaction. The lack of legal
representation for either party solves the conflict of interest because of the disinterestedness of
the agent, but this does not prevent the agent from looking out for her own self-interest.
Interestingly, the National Association for Realtors refused to endorse such legislation, arguing
that agents owe fiduciary responsibility to clients.
States also vary in their disclosure requirements for dual agents. For example, some
states, including New York, require that agents disclose their dual status in writing, while other
states only require a verbal disclosure. Similarly, states differ in their policies regarding when the
dual agent/agency is obligated to inform the buyer of dual-agency status (Olazabal, 2003). The
later the agent is obligated to reveal to the client that she represents both sides of the transaction,
the more likely it is that either the seller or the buyer will reveal information (e.g. minimum or
maximum acceptable sale price, respectively) that the agent might use to favor the other party
(and herself). In other words, the more lax the disclosure requirements, the more likely it is that
agents can indulge in conflicted interest behavior (and other principal-agent behaviors). In New
York, agents are required to disclose their dual-agency status when substantive contact with a
client is made.
3.2: Market responses to dual agency conflict of interest
Real estate firms have responded variously to any possible conflict of interest that
remains even after the law has attempted to clarify it. First, some real estate firms act as a
“buyer’s broker,” with exclusive fiduciary duty to the buyer, receiving payment from the buyer
(rather than the seller). While designed to protect buyers, this also helps to protect the seller from
any conflict of interest on the part of the seller’s agent. Some agencies have even gone further,
specializing in representing only buyers8
A second response from agents/agencies has been the adoption of a self-imposed policy
of providing referrals, should a dual-agency transaction arise. These referrals are sometimes to
other agents in the same firm (i.e., the equivalent of a designated broker), and sometimes to
agents at other firms (see
. But given the functioning of the real estate market
(e.g. a buyer sees a property that she likes and calls the agent whose contact information is on the
for-sale sign), exclusive buyer’s agents cannot be a complete solution. for examples).
A third response of some agents has been to go further than the law requires and draft
their own contract for dual-agency situations (e.g., see discussion on,
arguing that it is unclear if consumers understand the complex implications of the dual agency
laws (Olazabal, 2003). These contracts try to make clear the conflict of interest in dual-agency
contracts and urge both buyers and sellers to understand their rights. For example, consider
these two clauses (from an agent Greg Swann of Arizona): “The duties of the Licensee(s) in this
transaction do not relieve the Seller or the Buyer from the responsibility to protect their own
interests” and “If you are not completely comfortable with this disclosure of Dual
Representation, you are encouraged to obtain separate representation in this transaction.”
Despite these responses by agents/agencies, it is unclear if agencies can protect buyers
and sellers from the deficiencies of these dual agency laws and the resulting incentive
misalignment and conflict of interest. While these contracts provide additional disclosure and
discussion of dual agency and the resultant conflict of interest, they do not offer any legal
protections to either buyer or seller. This is because the agent behaviors that might harm buyers
and sellers (e.g., disclosure of private information to the other party) are difficult to observe.
Moreover, none of these responses attempts to address the misalignment of incentives in the
search (and possibly negotiation) stage created by permitting dual agency. In the following
sections we empirically assess the impact of dual agency on transactions in order to examine the
8 A discussion on is particularly succinct in its description of issues. One agent posted
his objection to dual agency by likening it to the same attorney representing both parties in a divorce. To which
another agent responded “Who do you think usually comes out ahead in a divorce, the divorcing couple or the
attorneys? If the divorce is amicable, or the couple doesn’t really have any assets or children, do they really need the
additional expense (attorney)? Isn’t divorce, by definition, costly enough?”
effectiveness of current regulation and determine whether more (or less) regulation would be
4. Data
Our dataset comprises 10,888 randomly-selected, single-family, residential home sales in
Long Island, New York, spanning the years 2004-2007. The data come from the Long Island
Board of Realtors, which owns the Multiple Listing Service (MLS) of Long Island, Inc. The
MLS is a clearinghouse where realtors list properties for sale.
This dataset has several advantages. First, the data contain a wealth of information on
house location and house characteristics. Among many other specifics, the data indicate the
number of bedrooms, bathrooms, and other rooms in the house, the number of fireplaces, the
capacity of the garage(s), the presence and type of driveway, the presence of a basement and
whether it is finished, and much more. Transactions also list the names of the seller and buyer
agent, the name of the agency for which each works, the number of days the house was on the
market, the price at which the seller listed the house for sale, and the final sale price9
It is important to note that the MLS data have some flaws, as noted by Levitt and
Syverson, 2005:
requires all sellers’ agents to enter the property within 24-48 hours of reaching an agreement to
list the house. Also, if a property is already under agreement for sale, MLS rules prohibit listing
it as available. This gives us confidence in the measure of time-to-sale in our data.
“The information in the database is entered by the real estate agents themselves. There is
no independent check on the accuracy of the description of the home’s attributes. Also,
there are few restrictions on what agents can type into a field in the database and no
requirement that all fields be completed. As a consequence, there are substantial amounts
of missing data for some variables…, some evidence of obvious errors, and a lack of
uniformity in the way fields are coded.”
To deal with these issues, we drop observations that appear to reflect erroneous data. For
example, we exclude observations where the list price or sale price is less than $50,000 and
9 Merlo and Ortalo-Magné (2004) have a unique dataset from England that includes all offers made on a house
before the final sale and any changes in list price during this period. This allows them to analyze seller and buyer
behavior within a transaction (rather than across). Our dataset only has the original list price and the final sales
price, and the identity of the seller agent and the buyer agent. This suffices for the purposes of our study.
where the sale price is either less than one third of the list price or more than three times the list
price. Similarly, we exclude houses with no bathrooms or no bedrooms.
There are two additional issues relating to the measure of time-to-sale. First, it is
possible that some houses are never listed on MLS because there is already a buyer for the house.
It seems likely that such cases would occur more often via dual agency, because agents have
greater knowledge about internal buyers. If this is the case, then our estimates of the relationship
between dual agency and time-to-sale will be biased upwards. Second, as discussed by Levitt and
Syverson (2005), houses sitting on the market for a long time are sometimes withdrawn. They
are then re-listed on MLS, re-setting in the MLS database the days on market to zero (see Tucker
et. al. 2009 for an analysis of the impact of prohibiting this practice). This will not affect our
estimates for role of dual agency unless there is a systematic correlation between dual agency
and re-listing, which seems unlikely. Nonetheless, if dual agency is more likely for re-listed
houses, this would cause a negative bias in our measure of the relationship between list price and
dual agency, because the second list price is likely to be lower than the first. In our results below,
we find a positive relationship between list price and dual agency. Therefore, we are confident
that re-listings are not driving our results.
To measure dual agency we create four indicator variables. AllDualAgency indicates that
the same agency represents both the buyer and the seller in the transaction. DualAgent indicates
that the same agent represents both the buyer and the seller. WithinBranch indicates that the
buyer and seller are represented by different agents who work for the same branch of an agency.
WithinAgency indicates that the buyer and seller are represented by different agents who work
for different branches of the same agency.
Table 1 reports descriptive statistics for all of our variables. We see that nearly half
(48%) of all transactions occur via dual agency10
10 This proportion of dual agency cases may seem high. A likely explanation is that some of the dual agency deals
are instances of subagency. In subagency, the agent listed as the buyer’s agent is actually a subagent of the seller,
with fiduciary responsibilities to the seller only (note that cross-agency deals could also have instances of
subagency). Previous papers on dual agency also face the same issue of misclassification, because MLS data do not
indicate whether the buyer is represented by a buyer’s agent or a subagent of the seller. Subagency is likely to
increase the sale price and increase the time-to-sale relative to transactions where the buyer is represented by a
buyer’s agent (Curran and Schrag, 2000). We find that on average, dual agency has the opposite effect on sale price
and time-to-sale. These findings cannot be explained by the misclassification of subagency.
, with the majority of these cases comprising
dual-agent deals (26%). Within-branch deals comprise 19% of dual agency deals, while within-
agency deals comprise only about 3% of dual agency deals. Table 2 provides sample means by
agency type (dual-agency or cross-agency) for our three key variables: sale price, list price, and
time-to-sale. There are no substantial differences between the dual-agency and the cross-agency
deals, although we do see that sale price and list price are slightly higher on dual agency deals,
while time-to-sale is slightly lower for dual-agency deals.
---Insert tables 1 and 2 here---
Table 3 provides information about the Long Island real estate market. The market is
highly competitive. There are 948 real estate agencies in our sample.11
They range in size from
single-agent firms, to agencies with more than 200 agents. In almost 4000 (more than 36%) of
the transactions, the selling agency has fewer than five agents, while in 2640 transactions (about
24%) the selling agency has more than fifty agents. Only a small number of agencies (51) have
multiple branches; but these agencies tend to be very large, comprising nearly 43% of all the
transactions in our sample.
---Insert Table 3 here---
Before assessing how the regulation of dual agency relates to the price and speed of
transactions, we first examine whether the incidence of dual agency varies with agency size. In
particular, we assess whether there is a greater incidence of within-agency transactions in larger
firms, where agents have a larger network of colleagues with whom to share transactions. The
results of our probit analysis, which we report in Table 4, confirm our intuition: selling agents in
larger agencies are more likely to sell to buyers represented by agents in the same agency. The
estimated marginal effect indicates that a doubling of the number of agents in an agency
increases the probability of a dual-agency deal by approximately 1.05 percentage points
---Insert Table 4 here---
11 Our data are only a random sample of the completed transactions during this time period. We could not obtain all
transactions because of limits on downloads from the MLS website. This limits our ability to accurately measure the
distribution of firm size, and especially its variation over time, since some of the variation that we observe results
from the sampling.
5. Empirical Analysis
5.1. Main Model and Identification Strategy
Our goal in this section is to empirically assess how the regulation of dual agency affects
final sale price and time-to-sale for the set of houses sold via dual agency. To do so, we identify
the effect of the treatment (dual agency) on the treated (houses sold via dual agency), i.e., ETT.
This is our measurement of interest because, by making two basic assumptions, we can: 1) get an
unbiased estimate for this effect and 2) use this measurement to infer the effect of dual agency’s
legal status. As we discuss in subsequent subsections, this approach is particularly appropriate
for our sale-price results, but has some important limitations when applied to our time-to-sale
We begin by presenting our main econometric model:
(1) ln()= +  +
In the above formulation, Yit is a transaction outcome variable (sale price, time-to-sale) for house
i at time t, Xit is a set of observed house characteristics (and agent, agency, and zipcode-year
fixed effects), ABit is a dummy variable equaling one if the agent engages in behaviors designed
to increase the likelihood of (and benefits from) dual agency (e.g., strategic pricing, seller
leaning, steering) henceforth called “agent behaviors”, DAit is a dummy variable equaling one
if the house sold via dual agency, and εit represents the error term (including unobserved house
characteristics). For this model to provide us an unbiased estimate of ETT, we must maintain
two key assumptions:
A1: εit is uncorrelated with Xit and DAit
A2: Agent behaviors only affect outcome variables in transactions conducted via dual-agency.
Assumption A1 is a standard one in empirical studies, i.e., unobservables are uncorrelated
with the regressors.12
Assumption A2 simply implies that ABit only enters the model through an interaction
term with DAit and not by itself. To illustrate this idea, we give a simple example. Suppose
House A has a potential buyer using the same agent (agency) as the seller and House B does not.
Otherwise, the two houses are identical. Assumption A2 states that if the seller agent leans on
her client to accept an offer from her (her agency’s) buyer, or discloses confidential information
to the buyer, but ultimately is not successful in generating a dual agency deal, then the two
houses should sell for the same price at the same speed. In what follows, we will maintain
Assumption A2 for our sale-price analysis, but will relax it (and discuss the consequences) when
conducting our time-to-sale analysis. If Assumption A1 holds, Assumption A2 has two key
merits for our analysis, which we discuss below. First, it allows us to obtain an unbiased
estimate of ETT without actually observing the incidence of agent behaviors. Second, it implies
that our estimate for ETT is also an estimate of the effect of changing the legal status (e.g.,
prohibiting) of dual agency.
It is necessary to obtain an unbiased estimate of ETT. To increase its
plausibility, we include a detailed set of controls (Xs), as discussed in section 5.2. Importantly,
we also test the validity of this assumption, particularly with regard to correlation with DAit,
when we present our results in sections 5.3 and 5.4.
In an ideal setting (e.g., a controlled experiment), we could measure the average
treatment effect (ATE) of dual agency, i.e., the average effect of dual agency on all houses in the
population of interest. This would involve randomly assigning houses into two groups, one
group being sold via dual agency and the other being sold via cross-agency. In our setting, it is
unlikely that assignment of houses into these two groups is random. In particular, the incidence
of dual agency is more likely for houses where agents engage in behaviors designed to increase
the likelihood of (and benefits from) dual agency. As discussed in Section 2, these agent
behaviors can affect outcome variables such as sale price and time-to-sale.
While our data do not allow us to estimate ATE, with Assumption A2 we can get an
unbiased estimate of ETT without observing the incidence of agent behaviors. In equation (1), α
represents the net effect on the outcome variable when there exist agent behaviors (there could
12 In principle, ε should be uncorrelated with AB as well. However, in our final model to be estimated (equation (3)
below), AB will be part of the error term and not a regressor. Consequently, we do not need to assume zero
correlation between ε and AB.
be several) and dual agency ultimately occurs. With our data, we are unable to estimate α
because we cannot directly observe agent behaviors. Further, we are not interested in the
measurement of α per se; we want to know the effect of dual agency on the houses sold via dual
agency (ETT). Recognizing this, we could reformulate the model in the following standard way:
(2) ln()= +() +(  ()) +
Equation (2) is equivalent to equation (1), except by treating (  ())
as part of the error term, we can estimate it using our data. However, if we estimate this model,
our estimated coefficient for DAit would be an estimate for (), which is the average
treatment effect of dual agency. That is, () is the expected change in (the log of) our
outcome variable when the dual agency status of any house is changed. With our model
formulation, a change in the legal status of dual agency (i.e., prohibition in our case) will only
affect houses that were sold via dual agency; therefore the ATE is not our measure of interest
because we are not concerned with what happens when houses that were sold via cross-agency
(despite dual agency being legal) are sold via dual agency. Further, the non-random assignment
of houses to dual agency would cause bias in this estimate. Specifically, the fact that dual
agency status is more likely for houses where agent behaviors occur implies correlation between
DAit and (  ()), the latter being in the error term13
To estimate the effect of dual agency on the houses sold via dual agency (ETT), we now
reformulate equation (1) in the following way:
. This would give us a
biased estimate of the average treatment effect.
(3) ln()= +(| = 1)  +(  (| = 1)) +
In the above formulation, αE(ABit|DAit = 1) is the expected change in (the log of) our outcome
variable associated with a change in dual agency status for the set of houses sold via dual agency.
This is the ETT we desire to estimate, and would give us a proper measure of the effect of
prohibiting dual agency in this market, i.e., the expected change in the outcome variable when
13 We prove this in the Appendix.
we change the dual agency status of the houses currently sold via dual agency. Importantly, if
we assume A1 (no correlation between the error term and our regressors), we can get an
unbiased estimate for αE(ABit|DAit = 1), since DAit is uncorrelated with α(ABitE(ABit|DAit =
1))*DAit by construction (illustrated in the Appendix). In what follows, references to the “causal
effect of dual agency” will be in the context of ETT.14
We conclude this subsection by noting that, when assumption A1 holds, assumption A2
implies that an estimate of ETT is actually an estimate of the effect of prohibiting dual agency.
If dual agency was prohibited, this would both eliminate the treatment (DA) and eliminate agent
behaviors designed to achieve dual agency status (AB). In our model, prohibition would still
only affect houses that were sold via dual agency by setting the interaction term to zero;
consequently, our estimate for ETT is also an estimate for the effect of prohibiting dual agency.
In contrast, if agent behaviors affect the outcome variable directly (i.e., they enter the model by
themselves, in addition to the interaction term with dual agency), prohibiting dual agency would
have an effect that extends beyond just houses sold via dual agency it would affect houses sold
via cross-agency deals as well. As we discuss below, we believe this violation of Assumption
A2 to be more likely in our time-to-sale model.
5.2. Other Variables Affecting Final Sale Price and Time-to-sale
To help in obtaining unbiased estimates for ETT, we consider many observable factors
that are likely to affect final sale price and time-to-sale for residential houses. These factors can
be broadly grouped as: property-specific, agency-specific, agent-specific, time-specific, and
market-specific. Consider first property-specific variables. In our regressions, we include various
hedonic descriptors of the property including number of bedrooms, number of bathrooms,
number of other rooms, fireplaces, capacity of the garage(s), etc. We also include dummies for
various amenities, including an eat-in kitchen, pool, finished basement, etc. In addition, we
include dummies for house style (e.g. Colonial, Cape Cod, etc.), type of fuel (e.g. gas, oil, etc.),
age of the house, seller agent’s assessment of the house’s condition (e.g. excellent, mint, good,
etc.), and other characteristics (see Table 1 for a complete list of house characteristics for which
14 Note that we abstract from general equilibrium implications of this assumption. For example, dual agency on
some houses might lower (or raise) sale price and increase (or decrease) time to sale for cross-agency deals as part
of overall market clearing dynamics. It is hard to speculate what direction and magnitude these general equilibrium
effects might take. Therefore, for simplicity, we abstract from these effects.
we control).15 In some regressions, we also include the list price of the house to control for
features of house quality that are not observable in the data, like southern exposure of the house,
new countertops, etc.16
Next, agencies might differ systematically in price setting and time-to-sale. For example,
ReMax agents keep 100% of the commission of the transaction and pay the agency fixed fees for
office usage and other overheads. In most other agencies, agents give half of their commission to
the agency to cover “desk costs” (Munneke and Yavas, 2001). These compensation differences
might alter price-setting behavior and time-to-sale. Agencies also differ in their market power.
Agency differences in market power and size might cause systematic differences in the incidence
of dual agency. For example, a larger agency might expect to be able to set a higher list price and
obtain a higher sale price, and at the same time have a higher incidence of dual agency because
of more agents working for the firm. In addition, because of differences in inventories of unsold
houses, agencies might face varying pressure to sell houses. We control for such agency
differences by including seller agency fixed effects in our models.
Agents’ differences might also affect price and time-to-sale. For example, agents might
differ in their reputation, ability to bargain, expertise, discount factors etc. (Palmon and
Sopranzetti, 2008). These differences might be systematically correlated with dual agency. For
example, a seller who hires a reputed agent might set a high list price with the expectation of
getting a higher sale price. This reputed agent might have a longer list of buyer clients than other
agents and therefore end up being a dual agent more often. To control for the influence of these
agent-specific factors on sale price and time-to-sale, we include fixed effects for each seller
agent in our models.17
Finally, time-varying and time-invariant market-specific variables are also likely to
influence our key variables. Some of these market-specific variables include school district
quality, tax rates, current interest rates, expected future interest rates, employment rate, inventory
of unsold houses and new home construction, trends in the industry (e.g., internet penetration,
15 The data also contain a variable measuring square footage; however, in many cases, lot size was recorded in the
data instead of interior square footage. Moreover, this variable is missing in many cases. Nonetheless, our results are
generally robust to inclusion of this variable.
16 Taylor (1999) suggests that a house with a low list price sitting on the market for a long time can be viewed as a
lemon. Genesove and Mayer (1997) show that sellers with lower equity positions built in the house set higher list
prices and receive higher sale prices. Our inclusion of list price captures these effects too.
17 Note that agency and agent fixed effects are not collinear because agents change agencies within the sample.
which gives buyers information beyond that provided by agents), and concentration of real estate
We include zip code-year fixed effects to account for these variations.
5.3. Dual Agency and Sale Price
We first estimate the following equation for the sale price for house i listed in year t:
(4) ln()= + + +
In this equation, X1 contains house characteristics, and X2 contains fixed effects for zip code-
years, seller agencies, and seller agents, as described in Section 5.2. Following our discussion in
section 5.1, if Assumption A2 holds, then δ represents the effect of dual agency for houses sold
via dual agency (ETT) and u contains both unobserved house characteristics (ε) and deviations in
agent behaviors from their conditional mean interacted with dual agency
((  (| = 1))), the latter being uncorrelated with DA by construction (see
equation (3)).
It is important to note that, for our sale price analysis, we believe Assumption A2 is
particularly plausible. As we discuss in Section 2, we believe that none of the three types of
agent behaviors should affect the sale price in cross-agency transactions; they should only affect
the price when a house is sold via dual agency.
We estimate equation (4) (and all succeeding equations that include dual agency as a
covariate) two different ways: (a) including the DualAgent, WithinBranch, and WithinAgency
variables and (b) using the AllDualAgency variable, which indicates any of three categories of
dual agency. The results for this model are in Table 5. The results in the first column show that
while dual agent and within-branch deals have no effect on sale price, within-agency deals have
a positive and statistically significant effect on price, increasing the price by more than three
percent. However, in the second column we find that, overall, dual agency has no effect on sale
18 Hsieh and Moretti (2003) show that the low cost of entry in to the residential real estate agent market causes the
number of agents to be positively related to the cost of land (and hence the size of the commission in any
transaction) in the market. Our zip code-year fixed effects control for both market-specific land price differences,
and competitive structure differences.
---Insert Table 5 here---
While our regression includes a detailed set of controls, we recognize that dual agency
could still be endogenous (i.e., assumption A1 may not hold), particularly if the selection of
houses into the treatment group is non-random due to correlation with unobserved house
characteristics. Hausman-type instruments are typically used in such situations (e.g. Nevo 2001).
In our case, incidence of dual agency in other markets (e.g., other zip codes) is one such possible
instrument, relying on the assumption that there exist zip code-level variables that influence the
likelihood of dual agency but not the sale price of the focal house. However, this identification
strategy will not work if house-specific unobservables are driving agent behaviors designed to
achieve dual agency.
Therefore, we instead follow the approach in Altonji et. al. (2005 & 2008) to estimate the
magnitude of endogeneity bias as a function of the relative strength of dual agency’s selection on
unobservables, as compared to observables. This technique identifies the likelihood of finding a
non-zero effect of a variable when its true effect is zero (e.g., finding dual agency raises prices
when, in fact, it has no effect); however, we will also use it to establish the likelihood of finding
a zero effect when the true effect is non-zero. To illustrate, consider the following equation:
(5) (|)(|)
where λ is the measure of relative selection on unobservables, DA is one of our four dual agency
dummy variables (dual agent, within-branch, within-agency, and any of the preceding three
types) and X is our full set of regressors. Note that this equation only considers ε, and not u
our full error term. This is because, as discussed in section 5.1, only ε (unobserved house
characteristics) can be correlated with DA in our general equation (3). Consequently, while our
analysis below will involve using residuals from the sale price equation (equation (4)), which
approximate the full error term, any bias from selection is coming entirely from ε.
In equation (5), the left-hand side represents selection on unobservables and the right-
hand side (except λ) represents selection on observables. Note that, given a null hypothesis for
the true effect of DA (ETT), we can get estimates for each piece of this equation from the data19
except for λ and E(ε|DA=1) E(ε|DA=0). Altonji et al. (2008) note that the bias in our estimate
for the effect of DA is:
(6) ()=()
)[(| = 1)(| = 0)]
where 
is the predicted value of DA from a regression of DA on X.
The first part of equation (6) can be estimated directly from the data. Then, given a value
for λ, we can use equation (5) to get an estimate for the second piece of equation (6). Following
Altonji et al. (2008), we use λ = 1 as our benchmark (i.e., equal selection on observables and
unobservables). This seems quite conservative given the detailed set of controls and fixed
effects that we include in our regressions.
In the first two columns of Table 5, and in subsequent results, we have instances where
our parameter estimates are notably different from zero, and instances where they are essentially
zero. The Altonji et al. (2005) method is directly applicable for assessing the likelihood that a
non-zero estimate is driven by endogeneity bias. Here, the null hypothesis is that the variable’s
causal effect is zero (δ = 0). In column 3, we report the size of the bias in our estimate for this
case, assuming equal selection = 1). In column 6, we report the size of λ necessary to fully
explain the non-zero parameter estimate as due to endogeneity bias. Following Altonji et al.
(2005), if the latter is greater than one, we conclude that the measured effect is unlikely to be
explained by endogeneity bias.20 In Table 5, the only variable that appears to have a non-zero
effect is WithinAgency.21
19 If the hypothesized effect of DA is δ, we need only regress Y-δ*DA on X to get consistent estimates for β and the
For that variable, Table 5 shows that (assuming equal selection) the
endogeneity bias is 0.0249; to fully explain the estimated effect of WithinAgency by endogeneity
bias would require λ = 1.3, leading us to conclude its effect truly is non-zero.
20 Our set of controls is comparable in breadth to theirs, and for their analysis they claim “the ratio of selection on
unobservables relative to selection on observables is likely to be less than one.”
21 When we break dual agency into three components, we effectively have three possible “treatments” rather than
one. In assessing the bias for any one of these three treatments (say DA1), if the hypothesized effect of DA1 is δ, we
regress Y-δ*DA1 on X only and on X, DA2, DA3 to get estimates for β and the ε’s. In all cases, our results are
substantively unchanged.
The above method is not especially useful in assessing the validity of parameter estimates
that indicate no effect. The concern there is that the apparent null effect is a biased estimate of
an effect that is actually non-zero (positive or negative). To assess whether this is occurring for
our estimates near zero, we consider alternative hypotheses of a modest positive effect (δ =
0.005) and a modest negative effect = -0.005). By choosing these cutoffs, we are effectively
treating any effect less than 0.5% as a virtual zero effect. We then determine the implied
endogeneity bias under each hypothesis (still assuming equal selection), as well as the size of λ
necessary to fully explain the parameter estimate as due to endogeneity bias when the null is
true. We report the latter measure only for the relevant null (e.g., if the parameter estimate is
.002 and the bias is .001, the estimate can only be non-zero and explained by the bias if the null
is δ = -0.005). In Table 5, the remaining three dual agency variables all have near-zero
estimates, and all three would require λ >> 1 to attribute these estimates to a true non-zero effect
coupled with endogeneity bias.
Summarizing, we find a positive effect on sale price for within-agency deals, but a null
effect for all dual agency deals, and these effects are not caused by endogeneity. As discussed in
Section 2, there are three agent behaviors leaning, steering and information disclosure that
might underlie these results. We now try to parse these sale price effects to isolate these three
If strategic pricing (i.e. setting a higher list price on the basis of an internal buyer’s WTP)
is occurring in our data, the higher starting point of negotiation ultimately leads to a higher sale
price. We check this by running the following regression:
(7) ln()= + + +
In this regression, we do not believe dual agency (DA) has a direct effect on list price. In fact, it
cannot have an effect, given that dual agency status is determined after the list price. However,
we may believe the “true” equation is as follows:
(8) ln()= + + +
where SPit is a dummy variable representing the existence of strategic pricing. By estimating
equation (7), DAit can serve as a proxy for strategic pricing. If its estimated coefficient is
positive, this indicates that strategic pricing occurs and is more prevalent on dual agency deals.
The estimation results from equation (7) are in Table 6, and show a strong, positive
correlation between dual agency and list prices. Moreover, consistent with the results in Table 5,
we find that this correlation is largest for within-agency deals, suggesting that strategic pricing is
driving the higher sale prices we find for these types of deals. We also find a notable correlation
for dual-agent deals and a modest one for within-branch deals (though not statistically
significant). We again use the Altonji et al. (2005 & 2008) approach to determine the magnitude
of any bias on these estimates. In all cases, we find that λ must be greater than one to explain the
measured relationships. In the case of dual-agent deals, the bias is actually toward zero, so no
level of selection can explain the result. Overall, these findings suggest that strategic pricing
utilizing inside information is occurring.
---Insert Table 6 here---
The results in Table 6 suggest that strategic pricing is occurring, particularly on deals that
close as within-agency and dual-agent deals. As mentioned in Section 2.3, the likely reason we
see more strategic pricing on within-agency deals is because when agents exploit cross-branch
information networks, they are able to greatly expand the pool of internal buyer clients for whom
they can access information about preferences. Consequently, while the incentives for strategic
pricing are highest on dual-agent deals, the opportunity for strategic pricing is highest on within-
agency deals.
Because strategic pricing raises the starting point for negotiation, we expect that it will
yield higher sale prices. If we control for list price, we are effectively controlling for any type of
strategic pricing that is occurring (whether it utilizes inside information or otherwise).
Therefore, by controlling for list price, we should be able to identify the effect of dual agency
stemming from the combined effect of all other agent behaviors associated with dual agency
(other than strategic pricing). Of course, list price not only controls for strategic pricing, it also
captures other house unobservables that may impact final sale price. However, because we have
established that unobservables are likely not causing any substantial bias in our estimate of the
effect of dual agency, the inclusion of list price should only affect our estimate for δ in an
important way if it is capturing part of dual agency’s effect on sale price (through the strategic
pricing effect).
Given that the overall effect of dual agency is a virtual zero, we would expect the effect
after controlling for list price to be negative if strategic pricing does raise sale price. We verify
this by estimating the following equation:
(9) ln()= + + + +
The results are in Table 7, and confirm our intuition. The results show that, controlling
for list price (and hence, any strategic pricing effects), dual agency lowers sale price, specifically
for dual-agent and within-branch deals. Further, the measured effects are quite precise and
unlikely to be the result of endogeneity bias. The very large numbers in the last column are the
result of these regressions having very high R-squareds (driven by the inclusion of list price as a
regressor), meaning there are few remaining unobservables (thus requiring enormous selection to
explain away the results).
The results in Table 7 are consistent with agent behaviors favoring the buyer, whether in
the form of seller leaning (rather than buyer leaning) or information disclosure to the buyer in the
negotiation phase to attain a dual agency deal. These results suggest that the weaker contracts
with buyers cause agents to favor buyers, even though this results in a lower sale price (and
lower commission). Moreover, consistent with our discussion in Section 2.3, the effect is more
negative for dual-agent deals than for within-branch or within-agency deals. This is consistent
with the idea that agents have stronger incentives to effectively cut prices in order to capture both
sides of the commission than to curry favor with a colleague.
---Insert Table 7 here---
5.4. Dual Agency and Time-to-Sale
In this subsection, we attempt to estimate the effect (ETT) of dual agency on time-to-sale.
We begin with the following equation:
(10) ln()= + + +
The results for this equation are in Table 8. These results indicate that dual agency deals
close about seven-eight percent (on average, about six days) faster than cross-agency deals.22
Moreover, we see that the effect is a little bigger for within-branch and within-agency deals
compared to dual-agent deals. Again, the final column suggests that these findings are unlikely
due to endogeneity bias (λ > 1 for all).
---Insert Table 8 here---
As discussed above, the results in Table 8 are only meaningful with regard to ETT and
the effect of prohibiting dual agency if Assumption A2 holds. While we believe this assumption
is reasonable for sale price, this may not be the case for time-to-sale. As we discuss in Section
2.3, it seems likely that agent behaviors designed to achieve dual agency may affect time-to-sale,
regardless of whether the deal concludes via dual agency. Consequently, equation (3) now
(11) ln()= + + +(  (| = 1)) +,
where δ = αE(ABit|DAit = 1) and Yit is time-to-sale.
The key difference here is that we now have another variable (ABit) in the “error” term,
and this variable is likely (positively) correlated with dual agency. The presence of this variable
implies a bias in our estimate for δ. Moreover, the bias is likely positive, because agent
behaviors likely slow down cross-agency transactions (they only speed up transactions if dual
agency is achieved), implying γ > 0, and agent behaviors and dual agency are likely positively
correlated. Hence, the effect of taking a house sold via dual agency and instead selling it cross-
agency likely would slow the sale by more than 7%.
22 Note that the marginal effects that we report are approximations. The marginal effect in a semilog specification,
like ours, equals exp(β)-1 (Thornton and Innes, 1989). Nonetheless, in our results, the estimated coefficients and the
marginal effects on the dual agency variables have negligible differences. Therefore, we simply report the estimated
However, this measure can no longer be interpreted as the effect of prohibiting dual
agency. If dual agency is prohibited, agent behaviors designed to achieve dual agency are
eliminated, and according to equation (11), this will affect houses sold via dual-agency and via
cross-agency deals. For example, if δ truly is -0.07 and γ is 0.05, then prohibiting dual agency
would speed up the sale of houses that were sold cross agency by 5% but slow down the sale of
houses that were sold dual agency by 2% (5% - 7%).
Given the above discussion, we can only infer the effect of prohibiting dual agency if we
utilize the upper bound on δ and make some assumption as to the magnitude of γ. To that end,
given approximately half of the houses in our data are sold via dual agency, then using the upper
bound for δ of -0.07, and assuming γ < ½|δ|, prohibiting dual agency would increase time-to-sale
on average + ½δ < 0). Put another way, assuming γ < ½|δ| implies that agent behaviors, on
average, increase time-to-sale on all houses by less than 3.5%, and since agent behaviors
decrease time to sale by 7% on the roughly 50% of houses sold via dual agency, the net effect of
allowing dual agency is faster sales on average; conversely, prohibiting dual agency will result in
slower sales.
6. Economic Harm from Dual Agency
Dual agency can influence agent behaviors, affecting welfare through price and/or timing
effects. However, as discussed in the introduction, there are several factors that may prevent
these issues from causing economic harm. It is likely that buyers (because of their looser
affiliation with their agents) and sellers (despite or because of signing exclusive contracts with
agents) are savvy enough to prevent dual agents from exploiting them. Moreover, due diligence
by buyers and sellers has been made easier by the internet, especially in the time period we are
studying (2004-2007). In addition, it is possible that agents care enough about their reputations to
forego engaging in these behaviors. Barriers to entry in the real estate industry are low, with
very few educational requirements for becoming a realtor, and therefore the industry is likely to
be quite competitive (Han and Hong, 2010).
Although the aspects of the real estate industry listed above are likely to mitigate
incentive conflicts, our results show there are indeed some economic consequences of allowing
dual agency. We find that allowing dual agency has competing effects on prices. On the one
hand, it facilitates strategic pricing, causing higher list and ultimately higher sale prices. On the
other hand, allowing dual agency also induces favoritism to the buyer when the seller is
represented by the same agent (agency), causing lower prices. On net, the effect of changing the
legal status of dual agency on sale prices is a virtual zero in our data. Because pricing is a zero-
sum game, the competing effects on prices offer little direct evidence of welfare effects.
If allowing dual agency does, in fact, accelerate time-to-sale, this could reflect a welfare
gain or loss. This may reflect a loss if allowing dual agency yields worse matches than what
would result if dual agency were prohibited. Steering buyers (sellers) to internal clients and/or
leaning on them to accept the offer of an internal seller (buyer) could lead to these clients settling
on an inferior house (buyer), given their preferences, than they would have found in a broader
search with no steering by their agent. On the other hand, buyers and sellers benefit if faster
transactions come from improved transactional or information efficiencies of dual agency. There
is no way for us to determine from the data if the faster transactions reflect welfare-improving
transaction/information efficiencies or welfare-reducing steering/pressuring. However, the
evidence we find for strategic pricing provides at least some support for the existence of
information efficiencies.
7. Concluding Remarks
We find that on net, allowing dual agency, has virtually no effect on housing prices. This
null effect is likely a combination of strategic pricing (which tends to raise prices) and favoritism
toward the buyer, including seller leaning and disclosing sellers’ confidential information to
buyers (which tends to lower prices). We also find that allowing dual agency reduces time-to-
sale by up to eight percent; however, its true effect may be lower. The behaviors associated with
dual agency have welfare ramifications; however, the net effect appears to be small.
Overall, our results suggest that the regulation of dual agency has relatively benign
welfare implications for both price and time-to-sale. Therefore, prohibiting dual agency is likely
an overreaction, especially as prohibition reduces choices for both buyers and sellers.
Additionally, it is not clear that legal requirements for practices like Chinese wall provisions for
dual-agency transactions are required, given the benign effects we find in our data. Finally, the
differences we find across dual agency types are also important from a policy perspective. For
example, in some states like Colorado and Maryland, only within-agency deals are permitted.
Our findings suggest that this distinction between types of dual agency might have welfare
consequences in the form of higher prices due to strategic pricing, which generally are not
mitigated by buyer favoritism (e.g., seller leaning), as is the case on dual-agent deals.
Certain features of the real estate market might explain the mildness of our net pricing
results and the differences across dual agency types. These features might not be present in other
industries or the same industry in other locations. A useful avenue for further research is to
examine whether our results from the Long Island, NY market generalize to other markets and
industries where dual agency is permitted.
Altonji, J., Elder, T. and C. Taber (2008), “Using Selection on Observed Variables to Assess
Bias from Unobservables when Evaluating Swan-Ganz Catheterization”, American
Economic Review, 98(2), 345-350.
Altonji, J., Elder, T. and C. Taber (2005), “Selection on Observed and Unobserved Variables:
Assessing the Effectiveness of Catholic Schools”, Journal of Political Economy, 113(1),
Bernheim, B.D. and J. Meer (2008), “How Much Value Do Real Estate Brokers Add? A Case
Study”, NBER Working Paper 13796
Brown, T. and K. Yingling (2007), “Antitrust and Real Estate: A Two-Sided Approach”,
Competition Policy International, 3(1), 225-235
Bryant, J. and D. Epley, (1992), “The conditions and perils of agency, dual agency and
undisclosed agency”, Real Estate Law Journal, 117.
Coy, P. (2007), “New Listing (Sort of)”, Business Week, January 22, pp 44.
Curran, C. and J. Schrag (2000) “Does It Matter Whom an Agent Serves? Evidence From Recent
Changes in Real Estate Markets”, Journal of Law and Economics, April, pp265-284
Evans, R. and P. Kolbe (2005), ‘Homeowners' Repeat Sales Gains, Dual Agency and Repeated
Use of the Same Agent”, Journal of Real Estate Research, 27(3), pp 267-92.
Gardiner, J, J. Heisler, J. Kallberg and C. Liu (2007), “The Impact of Dual Agency”, Journal of
Real Estate Financial Economics, 35, pp 39–55
Genesove, D. and C. Mayer (1997), “Equity and Time-to-sale in the Real Estate Market”,
American Economic Review, 87(3), 1997, pp. 255-69.
Genesove, D. and C. Mayer (2001), “Loss Aversion and Seller Behavior: Evidence from the
Housing Market”, Quarterly Journal of Economics, November 2001, pp1233-1260.
Han, L. and S. Hong (2010), “Testing Cost Inefficiency under Free Entry in the Real Estate
Brokerage Industry”, working paper, University of Toronto
Harris, L. (2003), Trading & Exchanges: Market Microstructure for Practitioners, Oxford
University Press
Hendel, I, A. Nevo and F. Ortalo-Magné (2008), “The Relative Performance of Real Estate
Marketing Platforms: MLS versus”, American Economic Review
Papers and Proceedings, May, 151-156
Hsieh, C. and E. Moretti (2003), “Can Free Entry Be Inefficient? Fixed Commissions and Social
Waste in the Real Estate Industry”, Journal of Political Economy, 111(5), pp 1076-1122.
Jares, T., J. Larsen, and T. Zorn (2000), “An Optimal Incentive System for Real Estate Agents”,
Journal of Real Estate Research, 20(1-2), (July-Oct): pp 49-59.
Levitt, S. and C. Syverson (2008), “Market Distortion When Agents Are Better Informed: The
Value of Information in Real Estate Transactions”, Review of Economics and Statistics,
90(4), 599-611
Merlo, A. and F. Ortalo-Magné (2004), “Bargaining over Residential Real Estate: Evidence from
England”, Journal of Urban Economics, 56(2), pp 192-216.
Munneke, H. and A. Yavas (2001), “Incentives and Performance in Real Estate Brokerage”,
Journal of Real Estate Finance and Economics, 22, pp 5-21.
Nevo, A (2001), “Measuring Market Power in the Ready-to-Eat Cereal Industry,” Econometrica,
69(2), 307-342.
Olazabal, A. (2003), “Redefining Realtor Relationships and Responsibilities: The Failure of
State Regulatory Responses”, Harvard Journal on Legislation, 40, pp 65-132.
Palmon, O. and B. Sopranzetti (2008). “Brokers, Information, and Transaction Outcomes: Evidence
from the Real Estate Market”, working paper, Rutgers University
Rutherford, R., T. Springer, and, A. Yavas, (2005), “Conflicts between principals and agents:
Evidence from residential brokerage”. Journal of Financial Economics, 76, pp 627665.
Salant, S. W. (1991), “For sale by owner: When to use a broker and how to price the house”,
Journal of Real Estate Finance and Economics, 4, pp 157–173.
Streitfeld, David (2008), “Feeling Misled on Home Price, Buyers Are Suing Their Agent”, New
York Times, January 22
Taylor, C. (1999), “Time-on-the-Market as a Sign of Quality”, Review of Economic Studies,
66(3), 1999, pp 555-78.
Thornton, R., J. Innes. (1989), “Interpreting Semilogarithmic Regression Coefficients in Labor
Research”, Journal of Labor Research, X(4), 443-447.
Tucker, C., J. Zhang and T. Zhu (2009), “Days on Market and Home Sales”, working paper,
Wheaton, W. (1990), “Vacancy, search, and prices in a housing market matching mode”, Journal
of Political Economy, 98(6), 1270–1291.
Yavas, A., T. Miceli, and C. Sirmans, (2001), “An experimental analysis of the impact of
intermediaries on the outcome of bargaining games”, Real Estate Economics, 29, pp 251–
Yinger, J. (1981), “A search model of real estate broker behavior”, American Economic Review,
71(4), pp 591–604.
Correlation between DAit and (  ()):
,  ()     () ()
  ()
    () ()   ()=
( ) () ()  ()( ) () ()=
  1 () (,)
The last term above is non-zero since α ≠ 0, E(DAit) < 1, and Corr(DAit,ABit) ≠ 0.
Correlation between DAit and ( (| = 1)):
(,(  (| = 1))) (  (  (| = 1)))
() ((  (| = 1)))
(  (  (| = 1))) () ((  (| = 1)))=
( ) (| = 1) ()
  ()( ) () (| = 1)= 0
The last equality relies on the fact that DA and AB are binary variables, so Pr(DA = 1) = E(DA) and
Pr(DA = 1 and AB = 1) = E(DA*AB).
Table 1: Descriptive Statistics
Mean (Std Deviation)
568,825 (361,502)
596,981 (400,118)
83.11 (69.68)
0.48 (0.50)
0.26 (0.44)
0.19 (0.40)
0.03 (0.16)
7.37 (1.62)
3.53 (0.87)
2.05 (0.81)
0.52 (0.50)
0.90 (0.31)
0.87 (0.34)
0.42 (0.61)
0.76 (0.43)
0.90 (0.30)
0.87 (0.33)
0.67 (0.47)
1.08 (0.70)
0.97 (0.17)
0.09 (0.28)
1945 (19)
0.03 (0.16)
* In addition to these variables, we also include dummy variables for different categories of house style, type of
construction (14 categories), type of fuel used (4 categories), type of heating system (6 categories) and general
condition/appearance (17 categories).
Table 2: Comparing Means for Dual Agency and Cross Agency Deals
Dual Agency
Cross Agency
Sale price ($)
List price ($)
Time-to-sale (days)
Table 3: Description of Long Island Real Estate Market
Mean (Median)
Agency size (agents)
41.54 (9)
5.51 (1)
Zip code-years
Table 4: Dual Agency as a Function of Firm Size
Dependent variable: Dual Agency (Within-branch and within-agency only)
Estimation method: Probit
Marginal Effect (std error)
Ln(Agency size)
0.015 (0.003)**
** Significant at .01.
Standard errors, clustered by agent, are in parentheses.
This analysis include all house characteristics and zip code-year fixed effects.
Table 5: The Relationship Between Sale Price and Dual Agency
Dependent variable: Ln(Sale price)
Estimation method: OLS
1 2
Implied Bias
under Equality
of Selection
(Null = 0)
Implied Bias
under Equality
of Selection
(Null = -0.005)
Implied Bias
under Equality
of Selection
(Null = 0.005)
Selection to
Explain Point
under Null
Dual agent
branch deals
agency deals
All dual
( 0.0037)
Within R-
0.72 0.72
+ Significant at 0.10; * Significant at 0.05; ** Significant at 0.01.
Standard errors, clustered by agent, are in parentheses.
All regressions include all house characteristics and seller agent, seller agency, and zip code-year fixed effects, as
described in Section 5.2.
Table 6: The Relationship Between List Price and Dual Agency
Dependent variable: Ln(List price)
Estimation method: OLS
1 2
Implied Bias
under Equality
of Selection
(Null = 0)
Selection to
Explain Point
under Null
Dual agent
Can’t Explain
branch deals
agency deals
All dual
( 0.0037)
Within R-
0.73 0.73
+ Significant at 0.10; * Significant at 0.05; ** Significant at 0.01.
Standard errors, clustered by agent, are in parentheses.
All regressions include all house characteristics and seller agent, seller agency, and zip code-year fixed effects, as
described in Section 5.2.
Table 7: The Relationship Between Sale Price and Dual Agency Controlling for List Price
Dependent variable: Ln(List price)
Estimation method: OLS
1 2
Implied Bias
Equality of
(Null = 0)
Implied Bias
under Equality
of Selection
(Null = -0.005)
Implied Bias
under Equality
of Selection
(Null = 0.005)
Selection to
Explain Point
under Null
Dual agent
( 0.0014)
All dual
( 0.0011)
Within R-
0.98 0.98
+ Significant at 0.10; * Significant at 0.05; ** Significant at 0.01.
Standard errors, clustered by agent, are in parentheses.
All regressions include all house characteristics and seller agent, seller agency, and zip code-year fixed effects, as
described in Section 5.2.
Table 8: The Relationship Between Time-to-Sale and Dual Agency
Dependent variable: Ln(Time-to-sale)
Estimation method: OLS
1 2
Implied Bias
Equality of
(Null = 0)
Selection to
Explain Point
under Null
Dual agent
( 0.0270)
branch deals
agency deals
Can’t Explain
All dual
( 0.0216)
Within R-
0.17 0.17
+ Significant at 0.10; * Significant at 0.05; ** Significant at 0.01.
Standard errors, clustered by agent, are in parentheses.
All regressions include all house characteristics and seller agent, seller agency, and zip code-year fixed effects, as
described in Section 5.2.
Table 9: The Relationship Between Time-to-Sale and Dual Agency Controlling for List
Dependent variable: Ln(Time-to-sale)
Estimation method: OLS
1 2
Implied Bias
Equality of
(Null = 0)
Selection to
Explain Point
under Null
Dual agent
( 0.0269)
Within branch deals
Within agency deals
Can’t Explain
All dual agency
( 0.0216)
Ln(List price)
Within R-square
+ Significant at 0.10; * Significant at 0.05; ** Significant at 0.01.
Standard errors, clustered by agent, are in parentheses.
All regressions include all house characteristics and seller agent, seller agency, and zip code-year fixed effects, as
described in Section 5.2.
... The problems caused by information asymmetry in existing home market can be classified into two categories: problems of hidden information and problems of 1 The Real Estate Brokerage Act imposes a duty on the broker to register related data with the Real Estate Information Network System (REINS) when the exclusive mediation agreement is concluded and prohibits the broker from intentionally concealing information from the other party of the contract or monopolizing it. 2 In the US, as a result of state's own legislation, there are (1) several states that, while permitting various types of dual agency, oblige the broker to disclose to both the seller and the buyer that he or she is a dual agent (e.g., New York), (2) a small number of states permit designated dual agency, i.e., they prohibit dual agency by the same person, but permit the dual agency within the same branch or the same firm (e.g., Colorado, Maryland), and (3) other states that, instead of prohibiting dual agency, permit transaction brokerage that does not legally represent either the seller or the buyer (e.g., Florida) (Kadiyali et al. 2014). The effectiveness of the obligation by the state laws is secured through the possibility of nullification or suspension of real estate agents' or brokers' licenses that are prescribed through such state laws (Olazabal 2003, pp. ...
... The mitigation of information asymmetry between the seller and the buyer by promoting real estate transactions through mediation has merits for both parties. In particular, in a dual-agency deal, as noted by Kadiyali et al. (2014), the broker knows such as the preferences of both sides which can accelerate and improve the efficiency of negotiations during the transaction 5 . However, Kakoikomi occurs if dual agency is permitted and the broker is able to intentionally hide potential sellers and buyers. ...
... However, this requires following conditions: (1) the broker shall take a stance that is not biased toward either the seller or buyer and (2) there shall be no difference in the negotiation power between the selling and buying brokers. Kadiyali et al. (2014) argue that whether dual agency affects sale price to be higher or lower depends on which side (i.e., the seller side or the buyer side) is taken by the broker, so its impact is not necessarily definitive. If taking the seller side, the sale price will be higher and the commission received increases proportionally, whereas if the buyer side is taken, the sale price will be lower in the sense that the broker fails to mediate the transaction with the most desirable customer for the broker. ...
Full-text available
In residential real estate market, agents have an incentive to steer their clients to their own listings or buyers rather than offering the best value transaction, which is derived from allowing dual agency and information asymmetry among buyers, sellers, and agents. We estimated the commission levels and sale prices of real estate brokers through a questionnaire survey and found that seven out of ten brokers are closing dual-agency deals and lowering sale prices. We could not find any effects of the number of employees, location of office, and major types of contract on dual agency.
... It has for instance, empirically shown that when the DMS is practiced, it is likely to create sub-optimal outcomes in the SA REI. This finding confirms a string of other findings (see amongst others: Kaplan, 2004;Kadiyali, Prince & Simon, 2014;Deutsch, Keil & Laamanen, 2011;Evans & Kolbe, 2005). On the ideology of real estate agents, it has been shown that real estate agents who hold strong ideological viewpoints are likely to produce less than optimal results for their principals. ...
... The outcomes may be those associated with perceptions of fairness in transactions, confidence in the REA, the perception of quality of service and possibilities of future business. This may arise from REAs disclosing sensitive information to one party or pressuring either the seller (to accept a lower price) or the buyer (to accept a higher price) to hasten the sale (Kadiyali, et al., 2014;Izzo, et al., 2003). The study has shown that brokers have an impact on property price and that buyers usually lose out (Zietz & Newsome, 2002). ...
... The following recommendations naturally flow out of this study: State intervention in the form of disallowing agents from representing two opposing interests (Nanda & Pancak, 2010). This would be in line with practices in other countries Like New Zealand and the UK (Gardiner, Heisler, Kalberg & Liu, 2007;Mersel, 1996;Kadiyali, et al., 2014). Furthermore, it is imperative for agents to not only avoid exploiting the inherent COI in their dealings with both seller and buyer principals, they also need to make consumers aware of their position, and to explain the possible effect on the broker's obligation to represent them (Carpenter, 2014;Coretz, 2017). ...
Full-text available
The impact of the: mandate type used, real estate agents' ideological outlook and compromising of principals' on the ultimate outcomes to buyers and sellers have largely been ignored in the South African real estate industry. This study attempted to bridge that gap by investigating the influence of the dual mandate system, the ideology of real estate agents and the compromising of the interests of one or all the principals on the outcomes of any real estate transaction. Stratified random sampling was used for information gathering. Data were collected using face-to-face filling in of the survey instrument and 150 participants agreed to take part in the study. Confirmatory factor analysis (CFA) The Johannesburg real estate industry characteristics: … 8 was employed to assess the reliability and validity of the results. The results reveal that Dual mandate system and Ideological persuasion of actors in the real estate industry does positively impact on the suboptimal outcomes to consumers. Furthermore, the results also conclusively showed that the principal whose interests are compromised usually gets less than optimal results. Keywords: South African real estate industry, dual mandate system, ideological outlook, compromised principal and suboptimal results.
... Price distortions can also be caused by dual agency relationships, defined as the seller and buyer agents being employed by the same real estate firm. In this regard, Kadiyali et al. (2014) and Johnson et al. (2015) investigate the effect of dual agency on sale price. The study by Gardiner et al. (2007) suggests that dual agency reduces sale prices and decreases the time a house is listed on the market. ...
Full-text available
We focus on the housing market and examine why nonlocal home buyers pay 12% more for houses than local home buyers. We established a database on the residential housing market for Lafayette and West Lafayette, Indiana, that includes house transactions from 2000 to 2020. The dataset contains highly detailed information on individual buyers and house characteristics. We explain the price differential controlling for arguments such as imperfect information on prices, wealth effects, heterogeneous buyer preferences, and differential search and travel costs across buyers, among others. We estimate a housing demand model that returns heterogeneous marginal willingness to pay parameters for housing attributes. Our results show that nonlocal home buyers are willing to pay more for specific housing attributes, especially for house size, school quality, and house age. We also find that arguments such as gratification, reward, and imperfect price information explain the price differential to a large extent. Search and travel cost arguments have an adverse effect on nonlocal buyers’ house spending.
... Sirmans, Turnbull, and Dombrow (1995) offer results consistent with this hypothesis, given that 3.8% of the houses in the data set were sold before being exposed to the MLS, which is typically required within 48 hours of listing. Kadiyali, Prince, and Simon (2009) find that houses sold in a dual agency situation have higher list prices and sell faster. The authors suggest that their results are consistent with ''first resort selling'' where agents show properties to their own clients first, as well as ''strategic pricing,'' which may lead to higher list prices when a sale to an internal client is anticipated. ...
This study empirically examines what factors increase the probability of a quick sale, an event where a property sells quicker than what is considered a normal marketing duration. This investigation is the first to explore the determinants of the probability of a quick sale utilizing a Probit methodology. Findings of the study suggest that overpricing, increased housing inventory, and increased listings by the listing broker decrease the probability of a quick sale. Dual agency, rising interest rates, and vacant properties increase the probability of a quick sale.
... The literature on residential real estate brokerage is expansive. Studies range from examining barriers to entry in the brokerage business (Hsieh and Moretti 2003), to how hard brokers work (Benjamin et al. 2004;Goodwin et al. 2012), to potential conflicts of interest when brokers act as principals (Bian et al. 2017;Kadiyali et al. 2014) or dual agents (Han and Hong 2016;Johnson et al. 2015), to how commissions rates are determined (Shy 2012;Wiley et al. 2014). Many other brokerage topics have been explored in the residential real estate setting as well, often with conflicting results. ...
Full-text available
Using a sample of CCIM designees and candidates in an experimental setting, this study examines the impact of broker signaling in commercial real estate transactions. It also explores the effect of certainty of closure in commercial real estate transactions. Findings suggest brokers are able to influence transaction pricing. Moreover, detailed analysis reveals that when a signal is above a reference point implied by previous transactions, the strength of the signal matters; privately communicated signals from reliable sources have significantly greater impact than signals which are made widely available. Additionally, we find an approximately 10% premium in transactions with lower certainty of closure than one with high certainty. The latter result varies by transactional participant type; owner/developers require a larger premium than institutional sellers.
... In general, they conclude that, for their sample agents and homebuyers and repeat-sale methodology, there was insufficient evidence to indict agents of failing to represent their clients. Kadiyali, et al. (2009) find that dual agency distorted some outcomes, especially on the sale of properties that occurred very quickly. For their overall sample, dual agency did not have a significant effect on sales price but was associated with higher list prices and shorter time on market. ...
Full-text available
The same real estate broker representing both the seller and buyer in the same real estate transaction is a situation known as dual agency and introduces obvious incentive conflicts. Previous studies have focused on the effect of dual agency on transaction outcomes such as selling price and time on market. This study examines whether the misalignment of agent incentives or agent specialization are significant factors in promoting dual agency transactions. Various definitions or areas of specialization are examined including geographical, property type, and price range specialization. Similarly several areas of misalignment of agent incentives are examined including strategic reservation pricing and first resort selling. Preliminary logit results indicate that dual agency is more likely to occur when the property is: sold within three days of listing, located within a geographic area where the listing agent has specialized, properties that listed for less than $125,000 or more than $350,000, is new construction, or listed by a broker with fewer active listings. These results suggest that dual agency is at least partially caused by a mixture of information advantages and transactional efficiencies resulting from agent specialization as well as evidence supporting principal agent conflict resulting from the misalignment of agent incentives.
This study is the first to examine dual agency sales over the listing contract between seller and listing agent. We test hypotheses about the timing of dual agency and its effects on sales price and time on market. Probit results indicate that dual agency sales are more likely to occur near the beginning or the end of a listing contract. Three stage least squares results demonstrate that dual agency affects sales price in both periods and that dual agency sales have shorter marketing times. Results support the conclusion that dual agency sales result from both incentives and informational efficiencies.
In residential real estate market, the agent has an incentive to steer her clients to her own listing or buyers rather than offering the best value transaction, which derives from allowing dual agency and information asymmetry among buyers, sellers and agents. We estimated commission levels and sale prices of real estate brokers by questionnaire survey, and found that seven out of ten brokers are making dual-agency deals and lowering sale prices. We couldnʼt find any effects of the number of employees, location of office and major types of contract on dual agency.
Aim: To explore critical care nurses' decisions to seek help from doctors. Background: Despite their well-documented role in improving critically ill patients' outcomes, research indicates that nurses rarely take decisions about patients' treatment modalities on their own and constantly need to seek advice or authorization for their clinical decisions, even for protocol-guided actions. However, research around the factors related to, and the actual process of, such referrals is limited. Design: A grounded theory study, underpinned by a symbolic interactionist perspective. Methods: Data collection took place in a general intensive care unit between 2010 - 2012 and involved: 20 hours of non-participant and 50 hours of participant observation; ten informal and ten formal interviews; and two focus groups with ten nurses, selected by purposive and theoretical sampling. Data analysis was guided by the dimensional analysis approach to generating grounded theory. Findings: Nurses' decisions to seek help from doctors involve weighing up several occasionally conflicting motivators. A central consideration is that of balancing their moral obligation to safeguard patients' interests with their duty to respect doctors' authority. Subsequently, nurses end up in a position of dual agency as they need to concurrently act as an agent to medical practitioners and patients. Conclusion: Nurses' dual agency relationship with patients and doctors may deter their moral obligation of keeping patients' interest as their utmost concern. Nurse leaders and educators should, therefore, enhance nurses' assertiveness, courage and skills to place patients' interest at the forefront of all their actions and interactions.
This chapter surveys the literature on the microstructure of housing markets. It considers one-sided search, random matching, and directed search models. It also examines the bargaining that takes place once a match has occurred, with the bargaining taking various forms, including two-party negotiations of different types and multiparty housing auctions. The chapter reviews research on real estate brokers as intermediaries as well, focusing on the role of brokers in the matching and bargaining process, the nature of competition and entry in the brokerage industry, and the incentive issues that are present. The chapter also considers the inefficiencies that pervade the brokerage industry and the related policy debates. These are important issues both because of the inherent importance of housing and brokerage and because of the importance of housing to macroeconomic dynamics.
Full-text available
Labor economists frequently misinterpret coefficients of variables in semilogarithmic regression equations. The proportional rates of change in the dependent variable that are implied by these coefficients are often erroneously assumed to be valid over arbitrarily large intervals. This note provides mathematical and empirical evidence on how serious the error can be. A simple formula is developed for making correct interpretations of semilog regression coefficients.
There are relatively few direct tests of the economic effects of asymmetric information because of the difficulty in identifying exogenous information measures. We propose a novel exogenous measure of information based on the quality of property tax assessments in different regions and apply this to the U.S. commercial real estate market. We find strong evidence that information considerations are significant. Market participants resolve information asymmetries by purchasing nearby properties, trading properties with long income histories, and avoiding transactions with informed professional brokers. The evidence that the choice of financing is used to address information concerns is mixed and weak.
This book is about trading, the people who trade securities and contracts, the marketplaces where they trade, and the rules that govern it. Readers will learn about investors, brokers, dealers, arbitrageurs, retail traders, day traders, rogue traders, and gamblers; exchanges, boards of trade, dealer networks, ECNs (electronic communications networks), crossing markets, and pink sheets. Also covered in this text are single price auctions, open outcry auctions, and brokered markets limit orders, market orders, and stop orders. Finally, the author covers the areas of program trades, block trades, and short trades, price priority, time precedence, public order precedence, and display precedence, insider trading, scalping, and bluffing, and investing, speculating, and gambling.
Brokers and dealers are involved in transactions that span a large variety of markets. However, there is neither a clear consensus as to the impact of brokers on transaction outcomes, nor a common understanding as to the nature of the services that brokers provide and the mechanism by which these services add value. We use the residential real-estate market to demonstrate that transaction outcomes, including the time on the market and the list and transaction prices, are associated with the extent of a broker's information. We find that the accuracy of the list price is also associated with the broker's information. We conclude that well-informed brokers help market participants by obtaining and conveying accurate and credible information, and thereby enhance an asset's value, transparency, and liquidity. Lastly, our paper implies that previous research which uses a binary variable for the presence of a broker may suffer from an omitted variable problem.
For much of the twentieth century, residential real estate transactions conformed to a 'traditional' mode - the seller engaged a broker, who listed the home in a multiple listing service, where it was noticed and purchased by a buyer, with a commission paid to the broker by the seller from the sale proceeds. While the listing/selling broker model endured for decades, it was laden with problems - it left the buyer unrepresented, created agency relationships that were counterintuitive to the parties, and often left both the consumer and realtor unsure of the precise nature of their legal relationship. Over the last fifteen years, state legislatures have set out to address these ills, enacting legislation to increase disclosure requirements, create new realtor roles, and redefine existing ones. While these reforms have added consumer choice and flexibility to the marketplace, they have not done enough to alleviate consumer confusion. After providing a comprehensive survey of state reforms, this Article argues that new laws must focus on imposing concrete duties upon licensees, most notably, other-party duties, in order to provide meaningful consumer protection. Indeed, rather than relying on increased disclosure requirements and broader consumer choice, states must enact laws that proactively protect buyers and sellers in order to eliminate the confusion produced by both the traditional model and a diverse and complicated set of reforms.
From 1996 to 2005, the residential real estate industry witnessed the greatest run-up in prices ever seen. But to hear most residential real estate agents tell it, the boom passed them by.
This article provides an empirical framework to study entry and cost inefficiency in the real estate brokerage industry. We develop a structural entry model that exploits individual level data on entry and earnings to estimate potential real estate agents' revenues and reservation wages, thereby recovering costs of providing brokerage service. We estimate the model, using the Census data. Based on our cost estimates, we find strong evidence for cost inefficiency under free entry, particularly attributable to wasteful non-price competition. We further use the estimated model to evaluate welfare implications of the rebate bans that currently persist in some U.S. states. We find that removing rebate bans would decrease the equilibrium number of real estate agents by 5.14% and reduce total brokerage costs by 8.87%.
A conflict of interest exists when a party to a transaction can gain by taking actions that are detrimental to its counterparty. This paper examines the growing empirical literature on the economics of conflicts of interest in financial institutions. Economic analysis shows that, although conflicts of interest are omnipresent when contracting is costly and parties are imperfectly informed, there are important factors that mitigate their impact and, strikingly, it is possible for customers of financial institutions to benefit from the existence of such conflicts. The empirical literature reaches conclusions that differ across types of conflicts of interest but are overall more ambivalent and certainly more benign than the conclusions drawn by journalists and politicians from mostly anecdotal evidence.
When a homeowner uses an agent to sell his property, he may have less information than his agent and be disadvantaged in price setting and negotiating. This study examines whether the percentage commission structure in real estate brokerage creates agency problems. We investigate whether agents are able to use their information advantage to either sell their own property faster or for a higher price than their clients’ properties. The empirical results confirm our theoretical predictions of agency problems, as we find that agent-owned houses sell no faster than client-owned houses, but they do sell at a price premium of approximately 4.5%.
This paper presents a new data set of individual residential property transactions in England. The main novelty of the data is the record of all listing price changes and all offers made between initial listing and sale agreement. We establish a number of stylized facts pertaining to the sequence of events that occur within individual property transaction histories. We assess the limitations of existing theories in explaining the data and discuss alternative theoretical frameworks for the study of the strategic interactions between buyers and sellers.