Asymmetric buyer information influence on price in a homogeneous housing market

Abstract

Buyers often pay different prices for almost identical houses. One possible explanation is that there are information asymmetries in housing markets. Perhaps, buyers from outside the area have higher search costs and know less about the local market relative to that of current residents. In addition, an out-of-town buyer’s price expectations could be anchored to market prices in their town of origin. This study examines the effect of buyer heterogeneity in the form of geographic location on house prices. We use a new data set to examine the non-local buyer information asymmetry and anchoring hypothesis. Using transaction data from a large development in Chengdu, China, our empirical models are estimated with relatively homogeneous units sold over a short period of time by one seller to minimise possible bias resulting from omitted variables. Our results support the hypotheses that non-local buyers pay higher prices and that high price anchoring occurs.

Keywords

anchoring asymmetric information house price

Background

Extensive research shows that humans can make mistakes when they make judgements. One explanation for these mistakes is the use of heuristics or cognitive short-cuts that simplify decision-making in complex situations (Simon, 1978; Tversky and Kahneman, 1974). Little research on heuristics has focused on the complicated process of buying a house. Heterogeneity in the housing market can be quite substantial. Houses, for example, differ in size, location and quality. In addition, housing markets are decentralised and information can be difficult to obtain.

In general, real estate markets are characterised by infrequent trading, geographically segmented markets and asymmetric information. Buyers must search for desirable properties and the selling price is typically the result of negotiation between the buyer and seller based on the information each possesses. If information asymmetry or search costs vary systematically with any buyer characteristics, then the price paid for any given property should also vary with those buyer characteristics. In particular, the conventional wisdom is that local buyers have more information and lower search costs, so more bargaining power than out-of-town buyers. In addition, non-local buyers may face a time constraint to find a house to enable commuting to a new job or enrolling children in a local school that local buyers do not (Ihlanfeldt and Mayock, 2012). Thus, the bid function of each group of buyers will differ. Faced with the same seller offer function, local buyers would be expected to achieve a lower negotiated price than non-local buyers.

Buyers from areas where real estate values are high or low relative to local prices may also have biased expectations of property values. This bias is known as the ‘anchoring’ effect. Tversky and Kahneman (1974) suggest that individuals use arbitrary reference values (anchors) that influence their estimate of value and that they only conservatively deviate from these reference values. Rather than researching an unfamiliar real estate market as an objective analyst, a potential buyer uses knowledge about a familiar real estate market as an initial estimate of the mean house value in an unfamiliar market. If the initial reference point, the value of a house in a familiar market, is higher or lower than the value for similar houses in the unfamiliar market, then the buyer is likely to reach a biased estimate of the value of a house in the unfamiliar market.

This study tests the impact of the full information assumption on estimated implicit housing prices. We examine the degree to which information asymmetry between buyers impacts bargaining power in the context of hedonic price models, with an application to the market for new condominiums in Chengdu, China. The size and homogeneity of the housing development used in this study allows for greater control over the heterogeneity of housing and neighbourhood characteristics than previous studies, thus eliminating much of the noise, reducing the potential of omitted variables and better isolating the effects of asymmetric information among buyers. This allows us to determine whether a buyer characteristic, living outside the local market area, is related to paying more for the same set of housing characteristics as a local buyer who, we assume, is better informed. The buyers come from a wide geographic area, allowing us to also examine whether prices in the buyer’s origin market create an anchoring effect.

Hedonic models were developed to clarify the manner in which prices are determined in markets for heterogeneous goods (Epple, 1987; Griliches, 1971; Rosen, 1974). This method is commonly used to decompose total expenditure on housing into measurable prices and quantities of its observable components. Housing is viewed as a collection of characteristics, each of which has a well-defined shadow price. The hedonic equation is a regression of expenditures (rents or values) on housing characteristics. The independent variables represent the individual characteristics of the dwelling and the regression coefficients may be interpreted as estimates of the implicit prices of these characteristics. The house price is then estimated by the linearisable combination of the shadow prices and individual characteristics. Bargaining has no effect on price because the characteristic shadow prices are well defined and are known to both buyers and sellers. Under the assumption that bargaining does not affect prices, characteristics of buyers and sellers that serve as proxies for bargaining ability have no role in the model and are omitted from the regression equation.

However, some research has been undertaken to account for the bargaining power driven by the role of local knowledge, search costs and anchoring and its effect on house prices using a variety of definitions of out-of-town buyer and thresholds for price anchoring. For example, Baryla and Zumpano (1995) find support for the existence of information asymmetry in that house buyers who move more than 25 miles search longer than local buyers. Watkins (1998) concludes that buyers coming from outside the city of Glasgow, Scotland, pay relatively the same price per unit as local buyers for residential properties; however, the in-migrants tend to buy more houses at the cheaper end of the market. Turnbull and Sirmans (1993) find a positive, but insignificant price premium paid by 63 buyers of single-family houses in Baton Rouge, Louisiana subdivisions who came from outside the metropolitan area. These studies have been criticised for the small number of non-local observations and the failure to control for property and location-specific characteristics (Ihlanfeldt and Mayock, 2012; Lambson et al., 2004).

Similarly, Miller et al. (1988) determine that Japanese buyers paid significantly more than other buyers in two Honolulu areas in the late 1980s. Because real estate prices in Japan were substantially higher than in Honolulu, this result can be interpreted as evidence of either higher search costs or anchoring effects. In the latest study on this subject, Ihlanfeldt and Mayock (2012) use both the metropolitan area and distance to identify non-local buyers of houses in Florida. Analysis of their larger data set reveals that buyers from outside the metropolitan area pay higher prices for housing and that the price premium increases with the distance moved. They also find evidence of an anchoring effect when the buyer’s geographic origin area is narrowly defined.¹

We extend the work on whether different groups of buyers pay systematically biased prices through analysis using unique transactions data. The size of new condominium developments in China allows for examination of a homogeneous group of new housing units in a single geographic location that are sold in a relatively short period of time by one developer. This data set with limited variability among housing characteristics addresses the shortcomings identified by Ihlanfeldt and Mayock (2012) in the ability to control for omitted variables that lead to bias in similar studies of more heterogeneous housing markets. We can compare the prices that local and non-local buyers pay for similar units in the same development to determine if out-of-town buyers suffer because of information asymmetry and higher search costs. Local buyers have easier access to information about the neighbourhood, traffic and commuting, and alternative housing developments. Non-local buyers, however, start with less information and have to access information from a distance. Because we know where the out-of-town buyers come from, we can examine relative housing prices among the origin cities to determine whether buyers from higher priced cities are anchoring their reservation price based on their prior experience. Our results show that non-local buyers pay higher prices for condominiums in Chengdu, China. We also find that buyers from cities with higher prevailing prices for houses start negotiations at a higher level tied to their experience in their home market, resulting in their paying an even higher premium than those buyers coming from areas with similar average prices as the local market.

Model and method

From a theoretical perspective, there are a number of reasons why buyers coming from further distances may pay more for a particular property than buyers already located nearby (Ihlanfeldt and Mayock, 2012; Lambson et al., 2004). These include information asymmetry, search costs and biased beliefs (or ‘anchoring’ behaviour). To identify potential price effects due to differential search costs and anchoring behaviour, we employ the standard search model with bargaining to capture the housing market transaction process in the simplest terms. The assumptions of the model require the following: sellers are willing to sell their houses at different prices according to some distribution (e.g. Lambson et al., 2004; Turnbull and Sirmans, 1993); buyers are of different types and we can distinguish between informed and uninformed buyers; and search costs are associated with finding a house to purchase. We also assume that search cost varies between different groups of potential buyers in such way that a buyer who is at an information disadvantage incurs higher marginal search costs. This is a reasonable assumption given that the search cost includes, among other things, the travel cost to visit different houses for sale (the visual inspection cost).

The housing market is such that there is no central marketplace where participants can trade. Sellers must search for buyers and buyers must search for sellers. This is true of both developers of new housing units as well as resellers. Search is costly because the houses being offered for sale in different developments are heterogeneous, requiring a potential buyer to gather extensive information on each offering. The process of identifying new houses on the market and then gathering information and visiting the property or a sales room can be expensive both in terms of money and time. The amount of information that a buyer has about an area will improve bargaining power.

The buyer enters the market with a set of beliefs about the price distribution. An optimal search strategy for the buyer exists such that the buyer will accept any price less than or equal to the reservation price while s(he) will continue searching if the price is higher than the reservation price (for a recent review see Lambson et al., 2004). The buyer will search until the marginal benefit of additional search is less than the expected net cost of the next search. Hence, the optimal search is a trade-off between getting a lower price by searching one more time against cost of continued search. Search costs include the cost of travel, inspection and study of the area. A buyer who does not already live in the local area is unlikely to have as much information as locals about the market to start with. Then they must travel to the area to gather information. Therefore, their search costs to achieve the same level of information will be higher than for local buyers. Higher search costs will result in a buyer setting a higher reservation price. The optimal search strategy suggests that if search costs are higher, the stopping rule will trigger earlier, and thereby, on average, buyers with high search costs will pay more compared with low-search-cost buyers.

Buyers who enter the market with an incorrect belief about the price distribution, but not gaining knowledge and correcting their expectation of the price distribution while searching, will also adjust the ‘stopping’ rule. If the buyer has a higher price expectation, this expectation will trigger the buyer to stop searching for a house earlier.

Meanwhile, the seller also forms an opinion about the reservation price for the house. The seller’s search costs include promotion of the development through a sales display, advertising and sales staff. The seller will search for the highest bidder until the marginal benefit of additional search is less than the expected net cost of the next search. Hence, from the developer’s point of view, the optimal search is a trade-off between getting a higher price by searching one more time against cost of continued search. If uninformed buyers are willing to pay a higher price than informed buyers for the same house, then the seller may increase profits by searching longer for these buyers, but also risks losing those customers to competitors’ properties.

Following Kumbhakar and Parmeter (2010), we present an empirical approach that enables us to introduce an incomplete information setting into a formal hedonic model, such as Rosen (1974). Thus, our method allows for estimation of a model that is consistent with the incomplete information conditions experienced in the housing market. In this scenario, the price a buyer pays is:

P_{m}^{b} = P_{x} + ω

(1)

where P_x represents the lowest price any buyer in the market is willing to pay and ω≥ 0 represents the cost the out-of-town buyer bears for being information deficient. If the traditional price model is shown as price as a function of a vector of house and neighbourhood characteristics as well as market conditions (Z), the resulting equation is:

P = α_{0} + β (Z) + ε

(2)

If borrower characteristics, such as whether the buyer is non-local, affect the price that a group of buyers pays for the same house, then such influential borrower characteristics B must be included in the model to produce:

P = α_{0} + β (Z) + γ (B) + ε

(3)

In our case, if out-of-town buyers enter the market with different beliefs about the price distribution, then the prices they are willing to pay may be systematically higher than those of local buyers. Their perceived price distribution is anchored to the higher price levels from their home areas.

The price paid for a housing unit is a function of the characteristics of the dwelling, its neighbourhood, and access to services and facilities. Most hedonic studies that identify the structural characteristics of the dwelling that contribute to price are based on single-family house sales in the USA. Some of these housing characteristics are irrelevant for condominiums and others may not be appropriate for the Chinese housing market. The common physical characteristics included in hedonic equations to estimate value of condominiums include age, size, number of rooms, floor within a multi-storey building, amenities and views (Chen et al., 2011; Choy et al., 2012; Liao and Wang, 2012; Ong and Koh, 2000; Shimizu et al., 2010; Wong et al., 2011; Wu et al., 2013). Consumers are expected to pay higher prices for newer and larger units in complexes that offer better amenities. The effect of building height, floor location and direction the unit faces on housing unit price may be ambiguous. High-rise buildings may represent prestige and offer better views; however, taller buildings also create higher population density that may result in overcrowding of common areas and resident stress (Wong et al., 2011). Higher floors within a building may provide better views, but with reduced access. Living on the top floor of a building ensures that no neighbours are located above the unit, but the resident is more susceptible to any problems with the roof or equipment located on top of the building. Living on the bottom or ground floor of a building offers ease of access, but also loss of privacy. Beautiful sunsets are accompanied by afternoon heat and northern exposures escape the summer sun, but are darker in the winter. In addition to the building, neighbourhood characteristics and location affect condominium values (Li, 2004; Liao and Wang, 2012; Ong and Koh, 2000). The relevant characteristics include neighbouring property uses, proximity to services, transportation options, and access to schools and workplaces.

It has been shown that house prices exhibit a seasonal pattern (Harding et al., 2003; Kaplanski and Levy, 2012), calling for time controls for seasonality. Because many homebuyers do not pay the purchase price in cash, the use of mortgages or alternative financing that increases the overall cost of the home may affect the price a buyer is willing to pay in a trade-off between a large present capital outlay versus payments with interest over time. In addition, a payment method that gives the seller the entire purchase price at the time of the transaction is worth more than a down payment plus periodic instalment payments, so the seller may be willing to give a cash discount. Thus, type of financing may affect final negotiated price (Turnbull and Sirmans, 1993).

Because this analysis uses data from one large new condominium development, the site, amenities, access and neighbourhood characteristics are uniform across all observations. Similarly, all units are of the same age, are in similar condition and vary little in terms of unit finishes or amenities. Therefore, the baseline empirical model, equation (4) for this analysis states that the price paid for a house is a function of the size and location of the unit within the development and financing used by the borrower with control for seasonality.

\begin{matrix} Ln PRICE & = α_{0} + α_{1} SIZE \\ + α_{2} TOP + α_{3} BOTTOM \\ + α_{4} HIGHRISE \\ + \sum_{i = 2}^{3} φ_{i} VIE W_{i} \\ + α_{5} FINANCE \\ + \sum_{i = 2}^{4} γ_{i} QUARTE R_{i} + ε \end{matrix}

(4)

We employ a semi-logarithmic equation where LnPRICE is the log of the price paid by the buyer for a condominium. Some physical characteristics can vary among units within the development, so are included in the equation. SIZE is size in square metres. HIGHRISE indicates whether the building is a high-rise building of 18 floors or a mid-rise building containing 11 floors. TOP and BOTTOM represent the location within the building, bottom floor or top floor (which is floor 11 in mid-rise and 18 in high-rise buildings). VIEW indicates which direction the unit faces and is represented by a series of dummy variables (WEST, SOUTH, OTHER). Method of payment, FINANCE, can be cash one-time payment or financed, which includes the options of a down payment and instalment payments via a plan provided by the developer or a down payment plus mortgage payments arranged through a bank. Time dummies are used to control for seasonality in terms of the QUARTER in which the purchase takes place.

To determine if a price premium is paid by non-local buyers, we estimate a log-log model as follows:

\begin{matrix} Ln PRICE & = α_{0} + α_{1} SIZE \\ + α_{2} TOP + α_{3} BOTTOM \\ + α_{4} HIGHRISE \\ + \sum_{i = 2}^{3} φ_{i} VIE W_{i} \\ + α_{5} FINANCE \\ + \sum_{i = 2}^{4} γ_{i} QUARTE R_{i} \\ + α_{6} Ln DISTANCEOUTSIDE + ε \end{matrix}

(5)

The variable of interest is LnDISTANC-EOUTSIDE. DISTANCEOUTSIDE is a continuous variable for the distance between the non-local buyer’s origin city and Chengdu measured as shortest highway mileage in kilometres from city centre to city centre based on baidu map.

In addition, equation (6) is used to examine whether the level of prices in the buyer’s home city influences the price the buyer pays in Chengdu. Here, we create a new variable called LnAVEPRICEANCHOR, which represents if the buyer lives outside of Chengdu at the time of purchase in a city in which the average housing price is greater than the subject location, Chengdu. AVEPRICEANCHOR takes the value of average price per square metre for housing in the non-local buyer’s city of origin. The model now becomes:

\begin{matrix} LnPRICE & = α_{0} \\ + α_{1} SIZE + α_{2} TOP \\ + α_{3} BOTTOM \\ + α_{4} HIGHRISE + \sum_{i = 2}^{3} φ_{i} VIE W_{i} \\ + α_{5} FINANCE + \sum_{i = 2}^{4} γ_{i} QUARTE R_{i} \\ + α_{6} LnDISTANCEOUTSIDE \\ + α_{7} LnAVEPRICEANCHOR + ε \end{matrix}

(6)

Data

The data for this study come from the sale of new condominiums in a development in Chengdu, the capital of Sichuan Province. Located in the centre of the province, Chengdu covers 12,121 km² and has a population of more than 11.63 million people (Chengdu Statistical Yearbook, 2011). Rapid urbanisation and a tradition of homeownership have created ongoing demand for new housing in the major cities of China. The volume of new private construction had reached more than 610 million m² annually by 2010. Because of the large number of units constructed in recent decades, new units account for more than one-half of the floor area of all private housing units sold in China (Wu et al., 2013). Thus, data for these types of transactions represent an important market segment in China more so than in many other countries.

Chengdu is representative of the national market with an average new apartment price close to the national average (Deng et al., 2012). Housing prices vary tremendously across the country. According to China Statistical Yearbook (2011), the average housing price in China in 2010 was 4725 Yuan/m². Meanwhile, the average prices among the 35 large and medium-sized cities reported by the National Bureau of Statistics of China ranged from 17,151 Yuan/m² in Beijing and 14,290 in Shanghai to 3196 Yuan/m² in Xining. The average price for Chengdu, where the development used in this study is located, was 5827 Yuan/m² in 2010. Thus, buyers in different cities would have vastly different average house price reference points.

As in many Asian countries, housing developers in China usually sell units of their projects prior to completion (i.e. pre-sale). In China, pre-sale is allowed only when certain criteria are met. These include the developer obtaining consent for pre-sale from the government, preparing a construction schedule and date of completion for the project, and investing at least 25% of the total cost. The developer creates a list of all the planned units along with listing prices. Chinese developers normally offer several payment options to buyers, including a one-time payment of the entire purchase price, an instalment plan financed by the developer and down payment in combination with a mortgage obtained from a bank. The down payment is usually at least 30% of the unit’s purchase price.

The new house market in Chengdu is dominated by large-scale condominium developments offering relatively homogeneous units within the complex. There is considerable competition amongst homebuilders in Chengdu as new developments are constantly starting construction (almost 20 million m² of new housing space started construction in 2010 while 13 million m² were completed). Thus, developers are sensitive to the actions of their competitors in the pre-sale market, similar to sellers in Singapore (Li, 2004), and there is no motive for discrimination among buyers.

The data used in this analysis comes from one typical development located in Wenjiang district in Chengdu.² The development covers about 60,000 m² on a construction area of approximately 210,000 m², which is the average size for a development in Chengdu. The development is comprised of 18 condominium buildings separated into two subdivisions. Subdivision 1 contains 958 units and subdivision 2 contains 1180 units. All the units within the development share the same location and neighbourhood attributes as well as building characteristics such as age and materials. They may differ as to unit-level characteristics in terms of size, floor level and orientation.

Owing to missing values for important variables on transactions in subdivision 1, our study only uses data from subdivision 2. Subdivision 2 offers both one- and two-bedroom units in ten buildings that contain either 11 or 18 storeys each. Initial sales of units in subdivision 2 took place from December 2009 through July 2011. The units were marketed and pre-sold by the developer.

A total of 1145 transactions are recorded for subdivision 2. Removal of observations with missing values on variables in the model results in 940 transactions in the analysis.³ Summary statistics for the sample are shown in Table 1. Panel A presents the descriptive statistics on the continuous variables for the entire sample as well as for the 561 buyers from within Chengdu and the 379 buyers from outside Chengdu, the definition of non-local buyer in equation (5). Panel B provides frequencies of the binary variables used in the analysis.

Table 1.

Sample summary statistics.

Panel A: Continuous variables
Variable	Mean		Std. Dev.		Minimum	Maximum
All buyers (940 observations)
PRICE (Yuan)	355,268		112,023		178,515	691,909
SIZE (m²)	86.68		24.89		46.70	133.26
Inside Chengdu buyers (561 observations)
PRICE (Yuan)	328,732		104,304		178,515	628,577
SIZE (m²)	80.93		23.65		46.70	133.26
Outside Chengdu buyers (379 observations)
PRICE (Yuan)	394,548		111,642		181,696	691,909
SIZE (m²)	95.19		24.27		46.86	133.26
Panel B: Binary variables
Variable	All Buyers		Inside Chengdu buyers		Outside Chengdu buyers
	Number	Per cent	Number	Per cent	Number	Per cent
Total observations	940	100	561	100	379	100
FLOOR
TOP	63	6.70	33	5.88	30	7.92
BOTTOM	36	3.83	27	4.81	9	2.37
OTHER	841	89.47	501	89.30	340	89.71
HEIGHT
MID-RISE	229	24.36	100	17.83	129	34.04
HIGH-RISE	711	75.64	461	82.17	250	65.96
VIEW
WEST	177	18.83	111	19.79	66	17.41
SOUTH	104	11.06	54	9.63	50	13.19
OTHER	659	70.11	396	70.59	263	69.39
FINANCE
CASH	234	24.89	107	19.07	127	33.51
INSTALMENT OR MORTGAGE	706	75.11	454	80.93	252	66.49
QUARTER
1ST QUARTER	192	20.43	80	14.26	112	29.55
2ND QUARTER	54	5.74	25	4.46	29	7.65
3RD QUARTER	330	35.11	243	43.32	87	22.96
4TH QUARTER	364	38.72	213	37.97	151	39.84

The average price paid by buyers from outside Chengdu (394,548 Yuan) is higher than the average for local buyers (328,732 Yuan) with an overall price range of 513,394 Yuan, so prices do vary considerably among the units and between local and non-local buyers. One reason is size. The units range in size from 46.70 to 133.26 m² and buyers from outside the city bought larger units, on average, than local buyers. Non-local buyers are also much more likely to make a one-time cash payment than local buyers. Local buyers seem to favour high-rise buildings. Among the non-local buyers, 47 come from cities with a higher average housing price than Chengdu in 2010, including Beijing, Shanghai, Guangzhou and Shenzhen.

Results

Table 2 reports the model estimates for our sample. As indicated in the first columns, the base model, model (1), specifies the natural log of sales price as a function of unit characteristics and payment/financing method. The coefficients indicate that buyers pay higher prices for larger units as expected. Units on the top floor sell at a premium, as in Shimizu et al.’s (2010) Tokyo study. In the buildings in this development, the units on the top floor own the space above the ceiling, so they have additional space as well as prestige and the lack of an upstairs neighbour. Units on the ground floor also sell at a premium (in contrast to the discount found by Shimizu et al. (2010) and Wong et al. (2011)). Ground floor units of this development usually have access to a storage room or garden, so the additional amenities appear to outweigh the lack of privacy.

Table 2.

Regression results for condominium prices.

Variable	Baseline model (Model 1)		Non-local buyers identified (Model 2)		Non-local buyers and buyers from higher priced cities identified (Model 3)
	Coefficient	t-statistic	Coefficient	t-statistic	Coefficient	t-statistic
SIZE	0.013	94.589***	0.013	90.788***	0.013	91.196***
FLOOR
TOP	0.046	4.720***	0.046	4.617***	0.051	4.864***
BOTTOM	0.089	8.587***	0.092	8.683***	0.090	8.478***
HIGHRISE	0.131	16.333***	0.130	16.250***	0.129	16.170***
VIEW
SOUTH	0.075	10.369***	0.074	10.108***	0.075	10.382***
WEST	−0.038	−6.830***	−0.039	−6.977***	−0.037	−6.490***
FINANCE	−0.020	−3.683***	−0.233	−4.175***	−0.025	−4.357***
LnDISTANCEOUTSIDE			0.002	2.588***	0.003	3.313***
LnAVEPRICEANCHOR					0.014	2.609***
QUARTER	Yes		Yes		Yes
Constant	11.573	615.554***	11.575	614.102***	11.443	226.858***
Adjusted R²	0.950		0.951		0.951
Number of observations	940		940		906

Notes: This table presents estimates from a regression with log of condominium transaction price as the dependent variable. LnDISTANCEOUTSIDE is the log of origin city distance to Chengdu (measured in km), and LnAVEPRICEANCHOR is the log of average price in origin city in 2010. *, **, and *** represent significance at the 10%, 5% and 1% level, respectively.

All mid-rise buildings are located in the centre of the complex surrounded by high-rise buildings, so they tend to have poorer views, lighting and ventilation. This is reflected in the positive and statistically significant coefficient on the indicator variable for HIGHRISE rather than the premium Wong et al. (2011) found for low-rise buildings in Hong Kong. West-facing units sell for significantly less than units on the other sides of the building (similar to Li, 2004), while south-facing units garner the highest prices (similar to Shimizu et al. (2010) and Choy et al. (2012)). The west-facing units suffer from heat from the afternoon sun without the benefit of air conditioning. The south-facing units receive light, but less heat.

In addition to the physical characteristics, prices paid for condominium units vary depending on the method of payment. Developers prefer an all-cash sale to an instalment payment or mortgage financing and appear to agree to a price discount for cash.

Our variable of interest in model (5) is LnDISTANCEOUTSIDE. The coefficient is positive and significant at the 1% level. Thus, the greater the distance a buyer lives from Chengdu, the higher the price paid for a housing unit, supporting the assertion that search costs are greater for buyers located further from the market. Our findings support the conclusions of Ihlanfeldt and Mayock (2012) that non-local buyers pay higher prices for houses.

While our LnDISTANCEOUTSIDE variable picks up the price differential paid by possibly less informed and time-constrained buyers from outside of the city, it does not allow us to differentiate the price variation associated with price anchoring. We assume that because only movers from more expensive cities would have a biased assessment of local prices, their bids should be more affected by prices in their origin city than bids made by movers from cheaper cities (those coming from cheaper cities would search longer and eventually learn about local prices according to Ihlanfeldt and Mayock (2012)). The possible anchoring effect of higher average housing prices in the buyer’s home city, LnAVEPRICEANCHOR, on the final transaction price is shown in model (6).⁴ The coefficient on LnAVEPRICEANCHOR is positive and significant at the 1% level. The results indicate buyers from higher priced cities are paying significantly higher prices than local buyers. The coefficient indicates an additional more than 1% price premium for units sold to the buyers who come from cities with higher average housing costs, again supporting Ihlanfeldt and Mayock’s (2012) findings. At the same time the coefficient on LnDISTANCEOUTSIDE is still positive and significant at the 1% level. The two coefficients together point to significant price premiums for properties sold to the outside buyers as a result of both information asymmetry and origin price anchoring.

As a robustness test, we create additional specifications. We apply a Box-Cox transformation to the dependent variable (PRICE) and the continuous explanatory variables (SIZE, DISTANCEOUTSIDE and AVEPRICEANCHOR) in model (7). The transformations are, respectively, 0.405, 0.915, 0.435 and 0.400. The coefficients on the other variables in the equation show little sensitivity to the specification of the measure of distance outside of Chengdu or average price in origin city (Table 3). After transformation, the coefficients have the same sign as in the previous results; however, DISTANCEOUTSIDE is significant at the 5% level and AVEPRICEANCHOR is significant at the 10% level when employing a one-tail test.

Table 3.

Regression results for condominium prices computed from Box-Cox transformation of dependent variable and independent variables.

Variable	Box-Cox transformation (Model 7)
	Coefficient	t-statistic
SIZE	3.173	84.664***
FLOOR
TOP	9.104	4.406***
BOTTOM	16.341	7.197***
HIGHRISE	12.615	7.772***
VIEW
SOUTH	8.717	5.750***
WEST	−7.284	−7.172***
FINANCE	−3.430	−3.105***
DISTANCEOUTSIDE	0.048	2.071**
AVEPRICEANCHOR	0.052	1.588^a
QUARTER	Yes
Constant	230.317	53.734******
Adjusted R²	0.940
Number of observations	906

Notes: This table presents estimates from a regression with condominium transaction price as the dependent variable. DISTANCEOUTSIDE is the origin city distance to Chengdu (measured in km), and AVEPRICEANCHOR is the average price in origin city in 2010. Model (7) gives the estimates after Box-Cox transformation of dependent variable (PRICE), SIZE, DISTANCEOUTSIDE, and AVEPRICEANCHOR. *, **, and *** represent significance at the 10%, 5% and 1% level, respectively.

While not significant when a two-tailed test is employed, AVEPRICEANCHOR is significant at the 10% level when a one-tailed test is employed.

We create OUTSIDE as a dummy variable for whether a buyer comes from outside Chengdu. The coefficient is positive and significant at the 1% level as shown in model (8) in Table 4. It indicates that buyers who come from outside of Chengdu pay a price premium of more than 1%. This translates into more than a 4000 Yuan price premium at the mean unit price of 355,268 Yuan. We create ANCHOR, a dummy variable that represents whether a buyer from outside Chengdu comes from a city in which the average price of housing was higher than in Chengdu in 2010 and add it in model (9). It is positive and significant, indicating that there is a relationship between the level of house prices in the buyer’s origin city and the price paid for a unit in Chengdu, providing further support for the hypothesis that buyers coming from higher priced markets anchor their reservation price to the origin city price levels. Thus, the overall results indicating a differential in prices paid by local versus non-local buyers as well as anchoring to higher prices among buyers coming from a higher priced market are robust to different specifications.

Table 4.

Robustness test of regression results for condominium prices.

Variable	Non-local buyers identified (Model 8)		Non-local buyers and buyers from higher priced cities identified (Model 9)
	Coefficient	t-statistic	Coefficient	t-statistic
SIZE	0.013	90.499***	0.013	90.426***
FLOOR
TOP	0.046	4.624***	0.045	4.616***
BOTTOM	0.092	8.694***	0.091	8.566***
HIGHRISE	0.130	16.278***	0.128	15.942***
VIEW
SOUTH	0.074	10.112***	0.074	10.077***
WEST	−0.039	−6.979***	−0.038	−6.936***
FINANCE	−0.023	−4.122***	−0.023	−4.150***
OUTSIDE	0.013	2.456***	0.010	1.864*
ANCHOR			0.019	2.175**
QUARTER	Yes		Yes
Constant	11.565	613.683***	11.576	613.019***
Adjusted R²	0.951		0.951
Number of observations	940		940

Notes: This table presents estimates from a regression with log of condominium transaction price as the dependent variable. OUTSIDE is whether a buyer is located outside of Chengdu. ANCHOR is whether a non-local buyer is from a city with higher average house price than Chengdu in 2010. *, **, and *** represent significance at the 10%, 5% and 1% level, respectively.

Conclusions

The inefficiencies of the real estate market lead to groups of disadvantaged participants paying higher prices for equivalent properties compared with better-informed buyers. The disadvantage often arises from asymmetric information among potential buyers. Experienced local buyers hold more information about the market and values. Buyers entering the market from other geographic areas hold less information and experience higher search costs than local buyers. In addition, the non-local buyers’ knowledge about their home market may bias their beliefs about the value of properties in other markets.

Our empirical tests of the influence of relative geographic location of buyers determine that non-local buyers pay higher prices for condominiums in a typical development in Chengdu. The results are robust to various specifications. Thus, local buyers use their information and search advantages to negotiate lower prices than non-local buyers for similar housing units. We also find that buyers from areas with higher prevailing prices for similar houses start negotiations at a higher level tied to their experience in their home market, resulting in their paying an even higher premium than those buyers coming from areas with similar or lower average prices as the local market. Limiting the study to one large typical development of relatively homogeneous housing units sold by one owner over a relatively short period of time removes much of the noise that may have interfered with the results in some earlier studies that presented conflicting results regarding these questions. We can reduce the possibility that buyers coming from higher priced markets may pay more because they obtain greater unobserved housing utility from the units they choose because of omitted variables as cautioned against by Ihlanfeldt and Mayock (2012).

The scale and homogeneity of residential developments in Chinese cities present the opportunity for a cleaner examination of questions that have been confounded by the heterogeneous nature of houses in most cities. Further research could isolate other consumer characteristics that may influence the bargaining power of consumers when bidding on houses. Additional analysis of non-local buyers could also assist in determining how much of the premium non-local buyers are paying is due to search costs versus time constraints. Where sufficient observations exist, a distance gradient might be examined to further delineate the boundary beyond which the information disparity develops to classify bidders as local versus non-local buyers. In addition, a sufficiently large and geographically diverse set of buyers in a homogeneous development would allow exploration of what differential threshold in average market prices must exist before anchoring significantly affects bidding behaviour.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Baryla

Zumpano

(1995) Buyer search duration in the residential real estate market: The role of the real estate agent. Journal of Real Estate Research 10: 1–13.

Chen

Clapp

Tirtiroglu

(2011) Hedonic estimations of housing demand elasticity with a markup over marginal costs. Journal of Housing Economics 20: 233–248.

Chengdu Statistical Yearbook (2011) Chengdu Statistical Bureau, Beijing: China Statistical Publishing House.

China Statistical Yearbook (2011) State Statistical Bureau, Beijing: China Statistical Publishing House.

Choy

LHT

WKO

Mak

SWK

(2012) Housing attributes and Hong Kong real estate prices. Construction Management and Economics 20(5): 359–366.

Clauretie

Thistle

(2007) The effect of time-on-market and location on search costs and anchoring: The case of single-family properties. Journal of Real Estate Finance and Economics 35(2): 181–196.

Deng

(2012) Mortgage rate and the choice of mortgage length: Quasi-experimental evidence from Chinese transaction-level data. Working Paper 12/02, BBVA Research.

Epple

(1987) Hedonic prices and implicit markets: Estimating demand and supply functions for differentiated products. Journal of Political Economy 95(1): 59–80.

Griliches

(1971) Price Indices and Quality Change. Cambridge, MA: Harvard University Press.

10.

Harding

Rosenthal

Sirmans

(2003) Estimating bargaining power in the market for existing homes. The Review of Economics and Statistics 85(1): 178–188.

11.

Ihlanfeldt

Mayock

(2012) Information, search, and house prices: Revisited. Journal of Real Estate Finance and Economics 44: 90–115.

12.

Kaplanski

Levy

(2012) Real estate prices: Seasonality’s sentiment effect. Journal of Empirical Finance 19(1): 123–146.

13.

Kumbhakar

Parmeter

(2010) Estimation of hedonic price functions with incomplete information. Empirical Economics 39: 1–25.

14.

Lambson

McQueen

Slade

(2004) Do out-of-state buyers pay more for real estate? An examination of anchoring-induced bias and search costs. Real Estate Economics 32(1): 85–126.

15.

W-F

(2004) The impact of pricing on time-on-market in high-rise multiple-unit residential development. Pacific Rim Property Research Journal 10(3): 305–327.

16.

Liao

W-C

Wang

(2012) Hedonic house prices and spatial quantile regression. Journal of Housing Economics 21: 16–27.

17.

Miller

Sklarz

Ordway

(1988) Japanese purchases, exchange rates and speculation in residential real estate markets. Journal of Real Estate Research 3(3): 39–49.

18.

Ong

Koh

(2000) Time on-market and price trade-offs in high-rise housing sub-markets. Urban Studies 37(11): 2057–2071.

19.

Rosen

(1974) Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy 8(1): 34–55.

20.

Shimizu

Takatsuji

Ono

. (2010) Structural and temporal changes in the housing market and hedonic housing price indices. International Journal of Housing Markets and Analysis 3(4): 351–368.

21.

Simon

(1978) Information-processing theory of human problem solving. In: Estes

(ed.) Handbook of Learning and Cognitive Processes: Vol. 5: Human Information Processing. Hillsdale, NJ: Erlbaum, pp. 271–295.

22.

Simonsohn

Loewenstein

(2006) Mistake #37; The effect of previously encountered prices on current housing demand. The Economic Journal 117: 175–199.

23.

Turnbull

Sirmans

(1993) Information, search and house prices. Regional Science and Urban Economics 23(4): 45–57.

24.

Tversky

Kahneman

(1974) Judgement under uncertainty: Heuristics and biases. Science 185(4157): 1124–1131.

25.

Watkins

(1998) Are new entrants to the residential property market informationally disadvantaged? Journal of Property Research 15(1): 57–70.

26.

Wong

Chau

Yau

. (2011) Property price gradients: The vertical dimension. Journal of Housing and the Built Environment 26: 33–45.

27.

Deng

Liu

(2013) House price index construction in the nascent sousing market: The case of China. Journal of Real Estate Finance and Economics DOI: 10.1007/s11146013-9416-1.