Introduction

A surprisingly large number of houses sell above listing or asking price. While the popular press tends to associate the phenomenon with boom markets, such sales occur regularly in both rising and falling markets across the U.S. For example, our data from central Virginia reveal that 34% of properties sell at or above the list price over 2001–2015, ranging from a high of 52% in 2005 near the height of the boom market to a low of 9% in 2012 in the aftermath of the housing market collapse. The question is: why do some houses sell above listing price while nearby similar houses do not? Is it seller under-pricing (whether strategic or accidental), broker acumen, or just luck? This paper offers an empirical method to separate the consequences of seller decisions from the influence of real estate brokers on the likelihood of transactions above list price and provides empirical evidence regarding the roles of each of these factors.

Housing is bought and sold in a search market with high transactions costs and asymmetric information in which most buyers and sellers rarely or at best infrequently buy and sell homes. Not surprisingly, most participants hire agents to guide their decisions and cope with the strategic and legal details that must be overcome to successfully close their transactions.Footnote 1 The result is a drawn out process in which the final transaction price is determined by the negotiation of an initial sales contract between buyer and seller (Han and Strange 2016) with subsequent responses to contingencies or re-negotiation required to push the transaction to a successful closing (Turnbull and Dombrow 2007). An extensive literature exists that looks at how real estate agents influence transactions prices and liquidity. While those questions are essential to understanding price discovery in housing markets in general, our attention instead focuses on the role of sellers, listing agents and luck in driving sales at or above listing price.

The canonical search model used to explain seller strategy assumes rational sellers set their reservation prices to maximize utility, which is increasing in expected selling price and expected liquidity or speed of sale. The seller’s marginal rate of substitution between price and liquidity reflects in part idiosyncratic factors. Households selling their property for divorce settlements, estate resolution, or because of job relocation might be more willing to trade off lower selling prices for quicker sales. These households may set lower listing prices to signal their lower reservation prices to potential buyers (Knight et al. 1994; Barwick et al. 2017). Alternatively, sellers might not fully comprehend current market conditions and as a result set their reservation price and advertise a listing price assuming weaker market demand than actually exists. Whether purposeful or inadvertent, setting unusually low listing prices attracts the attention of potential buyers with market knowledge who understand that offers at or above listing price will likely ensure they obtain the property.

Sellers, however, do not set listing prices unilaterally. They engage agents for a variety of services, one of which includes advice as to the initial listing or asking price as well as their opinion on subsequent offers (Geltner et al. 1991). Of course, the agent’s guidance on these matters may or may not be accepted by the seller or the broker may even encourage under-pricing to make marketing the property easier. Sellers are more apt than not to choose an agent that supports their opinion of value of the subject property, but at the same time, listing agents are more apt to accept a listing contract which they believe to be reasonably priced relative to market values of comparable properties (Anderson et al. 2014). Nonetheless, while it is the ultimate decision of the seller, there is some evidence that listing agents have considerable influence on the sellers’ listing price decisions (National Association of Realtors 2017).

If properties that sell above list price have been mispriced relative to comparable properties, the question remains whether such mispricing is due to motivated informed sellers who value liquidity highly, poorly informed sellers who will not accept the advice of their agents, or agents’ strategies (whether motivated by the desire to collect a quick commission, as a part of a quick-sale strategy undertaken in the interest of their seller, or as Han and Strange (2016) suggest, a strategy to stimulate a bidding war among interested buyers). A novel aspect of our empirical approach is that it removes the agent’s influence on the observed listing price in order to get a better picture of how seller mispricing relates to sales above list. Using this resultant measure of seller intent (or seller market knowledge) along with agent-specific variables in the final stage of the empirical model allows us to estimate separate measures of how each party, seller and agent, influence the likelihood of selling at or above listing price.

Listing Prices in Housing Markets

This section discusses key aspects of the literature concerned with the role of listing prices in housing markets. Not surprisingly, pervasive asymmetric information problems in housing markets give rise to wide ranging principal-agent issues. One consequence is the rich literature on principal-agent issues and conflict spanning a wide spectrum including the reallocation of broker efforts as a result of commission rates, dual agency, contract durations, atypical properties and agent inventories. Some of the previous studies that pay particular attention to the role of list prices in these contexts include Haurin (1988), Kang and Gardner (1989), Geltner et al. (1991), Horowitz (1992), Yavas and Yang (1995), Knight et al. (1994, 1998), Arnold (1999), Haurin et al. (2010), Benjamin and Chinloy (2000), Knight (2002), Anglin et al. (2003), Merlo and Ortalo-Magne (2004), Carrillo (2012), De Wit and Van der Klaauw (2013), Anderson et al. (2014), and Han and Strange (2016).

In one of the earliest of these studies, Yavas and Yang (1995) examine the optimal listing price as a decision made jointly between the seller and listing agent. They argue that the listing price is a function of the seller’s valuation of the property, bargaining power, commission rate offered, search costs and the intensity of the seller’s signaling function. The theory is based on the premise that sellers want to sell their property for the highest price in a minimum amount of time, which are conflicting objectives. The authors note that both objectives are dependent to some extent on the asking price as it influences the marketing time required to find a buyer and marketing time feedback effects on the selling price. They offer empirical evidence that overpricing decreases liquidity for mid-priced properties but overpricing has no significant impact on low or higher-end properties.

Glower et al. (1998) use seller survey data to look at seller heterogeneity regarding their motivation for setting list prices and the impact on the sales outcome. The data indicates that properties of sellers with a planned move-out date at time of listing (i.e., more motivated sellers) sell faster than those with no such restrictions. The difference is economically meaningful; highly motivated seller properties sell 30% quicker than other properties. Although the authors are unable to link seller motivation and original list price, they suggest that motivated sellers are more likely to lower list price more often during the marketing period than are less motivated sellers. The study does not address the role of a real estate agent in the process. Arnold (1999) examines the role of listing price in markets involving a search and bargaining process like that in residential real estate whereby the seller uses a broker to market the property and solicit potential buyers. The listing price affects the arrival rate of buyers and serves as an initial offer in the bargaining process. A lower list price increases the arrival rate of buyers but it also weakens the bargaining position of the seller as it reduces the maximum price that the seller can extract from a buyer.

Knight et al. (1994) focus on the role of listing price as a signal of seller motivation to the market. If the seller, in conjunction with the listing broker, sets a listing price that is at or below market value based on property characteristics, then the property is likely to transact sooner relative to a comparable property priced above market value. In this view, lower listing prices might also reflect the listing broker’s desire to earn a commission quickly (Yavas and Yang 1995; Knight 2002; De Wit and Van der Klaauw 2013).

In a different vein, Han and Strange (2016) examine the more recent phenomenon of bidding wars and argue that bidding wars represent an emerging hybrid marketing strategy coupling the search and negotiation process with the auction process. They estimate that bidding wars, while once a rare occurrence at 3–4% of market inventory, almost tripled over 1995–2005, decreasing after the financial and housing crisis but remaining approximately twice that of historical norms. As expected, bidding wars are more common during housing market booms. In their Richmond-Petersburg MSA sample, just under 20% of properties sold above asking price during the 2003–2006 boom period and 7.5% during the declining market of 2007–2010.

Data

The data used in this study are from residential properties listed for sale in the Richmond metropolitan area multiple listing service (MLS) in Virginia over 2001–2015. We use the entire residential dataset including multi-family, single-family attached, and single-family detached transactions to construct several of the variables used in the different stages of the estimation process explained in the next section. These variables include each real estate agent’s listing inventory and the variables that capture local market conditions. After constructing these variables, we limit the sample to single-family detached transactions within the Richmond urban core to estimate the model. Restricting the data to a single property type and narrower geography helps ensure the sample used in the empirical analysis is reasonably homogenous. Finally, we augment the MLS information with data obtained from the Virginia Department of Professional and Occupational Regulation to calculate listing agent experience.

We remove incomplete, missing or illogical data suggesting data entry errors as well as outliers.Footnote 2 Foreclosures, REOs, and short sales are also removed because several recent studies highlight the differences between distressed and non-distressed properties with regard to market outcomes. In addition, most banks hire real estate agents to list and market their REOs and some agents specialize in listing foreclosures. Finally, we remove sales by individuals whose professional focus is not in the area of brokerage (that is, investors and developers who hold real estate licenses primarily to sell their own properties) by removing all properties sold by agents whose inventory of listings includes more than 20% of properties identified as agent-owned. The sample used in the analysis comprises 116,129 listed properties of which 101,641 resulted in successful transactions. The MLS reports typical property characteristics such as square footage, bedrooms, bathrooms, exterior type, geographical location, and flooring type. In addition, the MLS data yield information about agent characteristics, listing date, selling date, and listing contract expiration date.

As explained below, the multiple stage estimation process uses the full sample of listings, both sold and unsold, in the first stage estimation and the sample of completed transactions in the final stage estimation. Table 1 contains a comprehensive list of variable definitions. Table 2 provides descriptive statistics for the sample of completed transactions used in the empirical procedure. (The appendix reports descriptive statistics for the full sample of sold and unsold listings and variables specific to each stage of the estimation.) Looking at Table 2, houses in the sample list for an average of $205,764. The average selling price net of seller concessions is $201,167. The average property has 3.4 bedrooms, 1.83 full bathrooms, 1,808 square feet of living area and is approximately 31 years old. Twenty six percent of properties are vacant and 2.8% are tenant occupied while on the market. The sample covers 1,312 brokerage offices and over 8,500 agents with an average agent experience of just under 13 years. Approximately 3.2% of sold properties are agent-owned, 15% are dual agency transactions in which the agent represents both seller and buyer, and about 20% include home warranties. Approximately 6% of listing contracts offer a commission rate between 2 and 2.5% to cooperating brokers, 92% offer a rate between 2.5–3% and less than 1% of contracts offer a 3% or greater commission rate to cooperating brokers. Eight percent of listings require homeowner and/or listing broker notice before being shown to potential buyers.

Table 1 Variable description
Full size table
Table 2 Descriptive statistics
Full size table

Figure 1 shows the percentage of sales occurring at or above listing price by year. The trend for the proportion of properties selling at or above asking price increases through the mid 2000’s. At the onset of the housing crisis in 2007, the percentage of properties selling at or above listing price falls rapidly before increasing again in 2011 and 2012. These sales clearly represent a nontrivial portion of house transactions even during declining and recovering market phases in our sample.

Fig. 1
figure 1

Distribution of properties that sold at/or above list price (by year)

Full size image

Empirical Model

The empirical method for sorting out separate seller and agent influences on the likelihood of a property selling at or above listing price takes into account the fact that the listing agent’s responsibilities include informing the seller about market conditions and offering advice about the appropriate selling strategy and listing price. While the agent may be able to influence seller behavior in this regard, the choice of listing price ultimately is under the control of the seller (Turnbull and Dombrow 2007; Anderson et al. 2014; Hayunga and Pace 2016). Most other aspects of how the seller influences the transaction outcome are not directly observable in our data aside from the seller’s choice of leaving the property vacant or tenant occupied or imposing showing restrictions.

We apply a multi-stage approach to isolate the seller’s primary influence on the probability of selling price at or above listing price. The first stage applies Anglin et al.’s (2003) method to identify the degree of overpricing (DOP) contained in the listing price (LP) for each house. First regress listing price LP for each property on property characteristics X for the full sample of all sold and unsold listed properties

$$ LP=\boldsymbol{\alpha} \boldsymbol{X}+\varepsilon $$
(1)

where the vector X is property characteristics and location and time fixed effects reported in Table 1 (see Appendix for more detail) and ε is the stochastic error. The dependent variable is the initial listing price.Footnote 3

Anglin et al. (2003) defines the degree of overpricing (DOP) for each listing as the prediction residual DOP = LP –LP* where LP* is the predicted listing price from the first stage regression equation above. This listing price residual, DOP, reflects the extent of over- or under-pricing of the listing price for the subject property relative to market norms.

One implication of many of the listing price papers discussed earlier is that the observed listing price for a property, hence DOP, is influenced by both the seller and listing agent. Separating the DOP into its components,

$$ DOP={DOP}_{Seller}+{DOP}_{Agent} $$
(2)

While seller characteristics are not directly observable, we can nonetheless observe their selling strategy decisions in their choice of DOPSeller. Since DOPSeller = DOP - DOPAgent we can identify the sellers’ degree of overpricing by removing the agent’s influence on overpricing. The agent’s influence on overpricing is a function of agent characteristics

$$ {DOP}_{Agent}=\boldsymbol{\beta} \boldsymbol{Z}+v $$
(3)

Regressing the calculated DOP for each property on the vector of agent characteristics in Table 1, Z, yields estimated the coefficient vector β*. This equation, too, is estimated using the full sample of sold and unsold properties. The right hand side variables are agent experience, the duration of the listing contract, the size of the listing agent’s contemporaneous inventory of other listings and a set of dummy variables indicating the commission rate being offered to cooperating agents. The estimated agent influence on overpricing for each property is the predicted value from the above regression, β*Z, so that the seller’s influence on overpricing is reflected in the residual

$$ {DOP}_{Seller}= DOP-\boldsymbol{\beta} \ast \boldsymbol{Z} $$
(4)

The variable DOPSeller measures the seller’s degree of mispricing after removing the listing agent’s influence on the choice of listing price. This variable provides our measure of the seller’s propensity to over- or under-price their property after the influence of the listing agent has been removed. The variable reflects seller ignorance of market conditions as well as any strategic efforts to signal their reservation price to potential buyers. We include DOPSeller in the final stage estimation to measure the effects of these aspects of seller pricing strategy on the probability of selling at or above listing price.Footnote 4 Our empirical strategy is designed to separate, as much as possible, the seller and agent influences on the sales outcomes that are filtered through the listing price (hence DOP) strategy. The agent’s influence on outcomes through DOPAgent will be picked up empirically by the agent characteristics vector in the final stage estimates.

The final stage of the analysis assumes that the probability of selling at or above listing price, Pr, is a function of the DOP as well as other aspects of the seller and agent strategies

$$ \mathit{\Pr}={\delta}_0+{\delta}_1 DOP+{\boldsymbol{\delta}}_{\boldsymbol{2}}\boldsymbol{S}+{\boldsymbol{\delta}}_{\boldsymbol{3}}\boldsymbol{Z}+u $$
(5)

where S is the vector of seller strategy variables (see Table 1 and below) and u is the relevant error term. Substituting (2) and (3) for DOP yields

$$ \Pr ={\delta}_0+{\delta}_1{DOP}_{Seller}+{\boldsymbol{\delta}}_{\mathbf{2}}\boldsymbol{S}+\left({\boldsymbol{\delta}}_{\mathbf{3}}+{\delta}_1\upbeta \right)\boldsymbol{Z}+{\delta}_1v+u $$
(6)

Notice that by decomposing DOP into seller and agent components, (6) allows us to estimate the separate influences of each actor on the likelihood of selling above list price. The agent characteristics coefficients now comprise both the direct effect of agent selling strategy on the outcome and the indirect effect through what would be the agent influence on DOP.

The final stage of the empirical approach estimates (6) as a probit model using the indicator I = 1 for sales price (net of seller concessions) at or above listing price and zero otherwise.Footnote 5 Table 3 reports the base model marginal effects estimates and Tables 4 and 5 the expanded models marginal effects estimates.Footnote 6 The explanatory variables in all models include seller motivation or mispricing reflected in DOPSeller, seller strategy reflected in Vacant, Tenant, restrictions placed on buyer offers (Prequalification) or showing the property (Notice to Show), and buyer incentives (Home Warranty). The variables related to listing agent ability and strategy are: listing agent experience (Experience), the time remaining to listing contract expiration (Time to Listing Exp) at the time of sale, the set of dummy variables for commission rates offered to cooperating agents, a dummy variable for a monetary bonus to the listing agent (Bonus), and the number of property photos and agent comments in the listing (Photos and Comments, respectively). It bears repeating that the method used to remove agent influence on DOP to construct DOPSeller creates an omitted variable DOPAgent that is, by construction, correlated with the set of agent characteristics. As a result, the set of agent variables included in the probit model fully account for this purposefully omitted agent effect that would otherwise be channeled through the agent’s influence on the degree of over-pricing.Footnote 7

Table 3 Marginal probability effects of key explanatory variables in probit model of sale above list price
Full size table
Table 4 Marginal probability effects of key explanatory variables in expanded probit model of sale at or above list price
Full size table
Table 5 Marginal probability effects of key explanatory variables in expanded probit model of sale at or above list price during different housing market cycles
Full size table

The probit model also includes measures of nearby sales at or above listing price to capture localized effects of similar sales. Sold at or Above is a dummy variable indicating another property sold at or above listing price within a one mile of the subject property and Sales at or Above measures the number of such sales in the surrounding neighborhood. The marginal effects of these variables on the probability of selling above listing price may pick up possible momentum effects at the neighborhood level. Guren (2018) shows that sellers’ decisions can drive price momentum when listing prices are strategic complements in a price-setting game. Applying the rationale more broadly to sellers’ overall marketing plans, positive Sold at or Above and Sales at or Above marginal probability effects are consistent with the notion that individual sellers behave as if aspects of their marketing plans that are unobserved in the data are strategic complements with plans adopted by other sellers in the neighborhood; negative marginal effects are consistent with strategic substitutes.

Finally, all models include the complete set of property characteristics listed in Tables 1 and 2 along with census block fixed effects as location controls.Footnote 8 The probit models also include year and quarter fixed effects to control for market cycle and seasonal market conditions.

Empirical Results

Table 3 reports the final stage probit estimated marginal effects of the listed variables for base models over the entire sample period. The first specification includes the seller degree of overpricing variable DOPSeller directly; the second specification interacts this variable with the market segment dummy variables in order to ascertain the extent to which seller mispricing effects vary across price segments.

Looking first at the variables related to market conditions, the dummy variables for price segment in the first model show some variation in marginal price effects, with houses in the below $140,000 listing price categories least likely to sell above list and houses listed above $250,000 most likely to do so. The remaining market variables capture the extent to which neighboring properties sold at or above listing price in the previous year. For both models the dummy variable or Above, both have significantly positive marginal effects on the probability of the subject house selling above list. One nearby sale at or above listing price is associated with a significantly greater probability of the subject property selling above listing price and increasing the number of such sales further increases the probability. This indicates that sales above list is to some extent a neighborhood phenomenon and that such sales feed off of other nearby sales. When taken at face value, houses lucky enough to be near other sales above list are more likely to sell above list themselves. But, as noted earlier, this pattern may also reflect spatially localized momentum effects similar to the price momentum that can arise when each seller pricing strategy is set in response to surrounding sellers’ strategies (Guren 2018). The pattern may also reflect that buyers to some extent rely on nearby transactions when formulating their offers.Footnote 9

Turning to the variables related to seller behavior, DOPSeller has a significant negative effect on the probability of the house selling at or above listing price in both models. Recall that this variable is constructed to remove the influence of the listing agent on the listing price set by the seller; it therefore captures mispricing by the seller and not the agent. The negative coefficient is consistent with the notion that sellers who insist on under-pricing their property (i.e., DOPSeller < 0) significantly increase the probability of selling their houses at or above list price. Buyers (or the agents working with them) appear to take advantage of obvious mispricing with offers that meet or exceed sellers’ advertised thresholds yet may not exceed fair market value. Of course, a selling price above listing price may also reflect the outcome of buyer bidding wars for some underpriced properties, but we have no way of identifying such situations in our data. In the second model, the DOPSeller interaction with the price segments dummy variables reveals that seller mispricing effects vary across market segments, with the strongest effect in the $140,000–250,000 listing price range.

Other variables related to seller strategy enter importantly in both models. Vacant houses and Tenant occupied houses are less likely to sell above listing price. The literature has established that vacant houses and tenant occupied houses sell at discounts and are less liquid relative to owner occupied houses (Turnbull and Zahirovic-Herbert 2011, 2012), so these results are not surprising. Vacant or tenant occupied houses may not show well or, in the case of tenant occupied houses, may have excessive visible wear and tear. At the same time vacant houses have higher holding costs that may signal weaker seller bargaining power.

Other seller strategy decisions have mixed effects on the probability of selling above listing price. Requiring buyers to Prequalify has no significant effect but requiring Notice to Show significantly reduces the probability. The latter result is intuitively appealing, since putting limitations on showing the property likely reduces buyer traffic. Interestingly, offering a Home Warranty to buyers significantly lowers the probability as well.

Now consider variables related to agent incentives, capabilities and behavior. The effect of number of photos published in the listing is significantly negative in both models in Table 3. The number of agent Comments in the listing increases the likelihood of selling above list in both models. This pattern suggests that listing agents use extensive comments to draw the attention of agents working with high value informed buyers and is consistent with conclusions by Goodwin et al. (2014) and Luchtenberg et al. (2018). Greater listing Agent Experience increases the probability of selling above listing price in both models. An alternative specification using a measure of successful sales in the previous year as an indicator of agent experience yields the same results. It appears that experienced agents use more effective marketing strategies in this regard than do their less experienced counterparts.Footnote 10

The commission rate offered to cooperating agents also affects the probability that a property sells above list price. The omitted category for this set of dummy variables is for buyer agent commissions below 2%. The only robust result here is that offering a commission rate of less than 2% is associated with a significantly greater probability of selling above listing price. This pattern is somewhat surprising in light of the incentives being offered cooperating brokers. Also surprising, the selling Bonus dummy variable marginal effect is not significant in either model. It appears that this type of agent incentive does not lead to selling at list or above. These results are, however, consistent in that they indicate that motivating cooperating brokers is not the key to sales above list and other factors, including luck, drive the process.

Anderson et al. (2014) and Clauretie and Daneshvary (2008) argue that the structure of the listing contract influences agent sales effort. The time until listing contract expiration variable captures these effects in our models. The positive Time to Listing Exp coefficient indicates that houses sold with greater time until listing contract expiration have a greater probability of selling above list. This implies that sales above listing price tend to occur early in the contract period, the period in which listing agents have the strongest incentive to procrastinate and not aggressively promote the property (Clauretie and Daneshvary 2008; Geltner et al. 1991). This is one more indicator that luck is an important element at work driving sales above listing price; it looks like high value buyers tend to show up early in the marketing period when agents have the weakest incentives to aggressively push the property.Footnote 11

Atypical Properties, Agent Owned Properties, and Dual Agency Effects

Table 4 reports key parameter estimates for the expanded models taking into account the effects of atypical properties, agent owned properties, and dual agency. Before turning to these additional variables, note that none of the key conclusions presented above fundamentally change in the extended models.

Consider first the Atypical house effect on the probability of selling at or above list once seller and agent effects are removed from the likelihood. Our construction of the Atypical variable follows the approach established by Haurin (1988) and Haurin et al. (2010). The marginal effect estimates on this variable in models (3) and (4) in Table 4 indicate that atypical properties have a lower likelihood of selling above listing price. This is consistent with the body of evidence that atypical properties generally sell at a discount or take longer to sell.

We next consider whether listing agents who adopt sales strategies that lead to a greater likelihood of being the sole agent involved in the transaction also influence the probability of selling above listing price. The Dual Agency marginal effect in models (5) and (6) is significantly negative; it appears that dual agency transactions are less likely to sell at or above listing price.

The Owner Agent dummy variable marginal effects in models (7) and (8) measure the extent to which owner-agents are less likely to sell their own houses properties at or above the listing price relative to their clients. Rutherford et al. (2005), Levitt and Syverson (2008) and a growing number of subsequent studies offer evidence that agents adopt different selling strategies for their own properties than for their clients’ properties. In our sample, the probit marginal effects estimates for Owner Agent indicate that agents selling their own properties are less likely than their clients to sell at or above listing price. Our estimates are consistent with the notion that agents follow different strategies or exert different effort when selling their own properties relative to what they do for their clients’ properties.

Market Cycle Effects

Our data period runs from 2001 through 2015, a period encompassing rising, falling and recovering market phases. Table 5 reports the final probit marginal effects estimates for the different phases identified using the FHFA price index for the Richmond-Petersburg MSA: rising market (2001–2007), market decline (2008–2010), and recovery (2011–2015). It is interesting to note that the full sample qualitative conclusions also hold for the rising market phase as reported in the first column in Table 5. Many of the results also hold in the recovering market phase reported in the last column, but with some important differences. The declining market results in the second column differ considerably from the other sets of estimates, which is not surprising in light of empirical evidence that the most recent catastrophic collapse impeded price discovery in other housing markets (Turnbull et al. 2018; Zabel 2015).

One difference between the rising market and the declining and recovering market results is that nearby sales above listing price no longer increase the likelihood of selling above list in the declining and recovering markets; there is no stable relationship in the declining market and nearby sales above list decrease the probability of selling above list in the recovering market. The differences in results across market phases emphasizes that the price discovery process is influenced by the broader market context. In contrast, the marginal effect of the key seller variable, DOPSeller, is significantly negative in all market phases, the same as in the pooled sample. The marginal effect, however, is significantly weaker in the recovering market than in the rising and declining phases.

The market phase clearly matters for agent abilities and strategies. Agent experience and comments in the listing increase the probability in the rising and recovering market phases but have no effect in the declining market. Similarly, including a greater number of pictures in the listing lowers the probability of selling above listing price in all market phases, but the effect is less significant in the recovering market. And some marketing strategies simply appear not to matter. For example, offering a bonus to buyers’ agents has no significant effect in any market phase.

Other differences across market phases are evident as well. Atypical houses have lower probabilities of selling at or above listing price in the rising and recovering market phases, but exhibit no significant effect in the declining market. The Owner Agent marginal effects indicate that agents are less likely to sell above listing price in the rising market, as in the pooled sample, but more likely in the declining market. There appears, however, to be no significant difference between agent-owned and client properties in the recovering market. That agents behave differently with their own property than when selling clients’ properties fits with evidence of principal/agent effects in the broader literature. What is different here is the absence of such evidence with respect to selling above listing price in the recovering market.

Conclusion

A surprisingly large number of houses sell above listing prices in a wide range of markets and in all market conditions. The question is: why do some houses sell above listing price while neighboring similar houses do not? This paper argues that this occurs because sellers misprice their property at the outset, they work with real estate brokers who are particularly skilled at bringing in high value buyers, or they are simply lucky to have high value buyers show up during the marketing period. We find evidence of all three factors at work.

This paper makes two contributions to the literature dealing with sales above listing price. First, it offers an empirical framework to isolate seller and agent influences on the likelihood of selling above listing price. Second, it offers a complete empirical analysis of the seller, agent and market determinants of sales above listing price in all market phases. The results show that sellers and their agents both drive the likelihood of such outcomes but an element of luck enters importantly as well. To the extent sellers do not follow their agents’ guidance when setting the listing price, whether due to ignorance or because they wish to signal their impatience to potential buyers, greater under-pricing systematically increases the likelihood of selling above list. Specific marketing strategies and agent incentives also appear to influence the likelihood of selling above listing price, but not always as expected. Seller decisions to put vacant or tenant occupied houses on the market reduce the probability. Agent experience also increases the probability of selling above list in rising and recovering market phases but exhibits no significant effect in the declining market. And agent incentives can have counter-intuitive effects. For example, a seller offer of a cash bonus does not influence the probability of sale above list in either boom or declining markets. Perhaps more surprising, the listing broker’s offer of larger commissions to cooperating brokers for bringing buyers to the table either reduces or has no effect on the probability of selling above list across all market phases.