Marketing Auth ors

Time

a n d Pricing

Strategies

Eddie C M . Hui, Joe T.Y. Wong, and K.T. Wong

Abstract

This study examines how overpricing of properties (in terms of above-market price), along with various housing attributes, influence their time-on-market (TOM). The results with the full sample show that only dummy variables depicting years 20032005 and flats located in Kowloon significantly affect TOM. In the sub-period analyses, however, factors such as the abovemarket price, sale price, a flat's tenure status, general property price trend, and changes in unemployment rate have significant impacts on TOM, yet their respective impacts change over time. Specifically, the effectiveness of overpricing in optimizing sellers' returns and TOM depends on economic conditions as well as on the availability of other alternatives on the housing market.

In the valuation world, property appraisers all too frequently relate marketing time to the estimated capital value of a real property. For example, for their definition of "market value" or "open market value," a "reasonable" time period in the market is assumed. Aside from various housing characteristics (Hui, Chau, Pun, and Law, 2007), other important elements that can influence the incentive to buy are the sellers' initial asking prices and their expectations of the selling price (Wong, Hu, Seabrooke, and Raftery, 2005). While housing characteristics generally remain static over a long period of time, list prices, consumers' preferences, and price aspirations and expectations may change over time. Although one would argue that the housing attributes may also undergo dramatic changes, such as urban renewal or redevelopment, it is usually an exception rather than a norm. An optimal asking price and optimal time-on-market (TOM) would maximize the net realizable return from residential sales. From a seller's perspective, one question being asked is whether he/she is able to obtain a higher selling price over a longer period of time, or just settles for a lower selling price over a shorter period of time. The question that follows may be: What is an optimal selling period if the property is sold at the current market value? From a buyer's perspective, he/she would look for the best possible price for a flat, subject to the amount of information and search costs. In other words, how much a seller could gain over the average market price depends on how much a potential buyer is

JRER

V o l .

3 4

N o . 3 - 2 0 1 2

3 76

H u i ,

W o n g ,

a n d W o n g

willing to spend, whether in terms of time or money, on searching for relevant market information and similar flats available [for the respective psychology of buyers and sellers, see Wong and Hui (2008)]. Such considerations dictate the TOM for a fiat for sale and the seller's realized return above market price. If the buyer's search costs exceed the seller's potential gain through a higher list price, the transaction is likely to take place sooner, thus a shorter TOM, and vice versa. The latter scenario could reach a point where transactions fail. For instance, Anglin (1994) reported that 30% to 50% of negotiations ended without a transaction while Anglin, Rutherford, and Springer (2003) noted that about 40% of house listings ended without a sale. But the main problem remains: Could sellers with higher selling costs achieve a faster sale, and at the same time, a maximum realizable return? Alternatively, could sellers benefit from longer marketing time in terms of achieving greater real selling prices? If not, what is the optimal pricing strategy such that a seller could achieve the greatest net present value of the selling price in an optimal marketing time, without prior knowledge of a potential buyer's searching and information costs? This study examines the list price, the expected sale price, and the TOM in the property market under different living tenures, flat sizes, price ranges, over (under)-priced properties, and other physical characteristics. This study focuses on Hong Kong's residential market while previous studies have examined the overseas markets. This paper is organized in four sections. The next section is the literature review on the time-price relationship. The sections that follow present the study methodology, and describe the data source and the empirical results. The last section provides concluding remarks.

Literature Review
Prior to a transaction for real properties, sellers have to decide whether to maximize selling price or to minimize TOM (Trippi, 1977; Miller, 1978). For the former, according to Knight (2002), a seller's search is led by a desire to maximize the discounted present value of realized profits from a transaction through a listing price that balances the marginal costs of continuous searching with the marginal benefits of accepting an offer at present. In the meantime, a buyer faces a similar problem, as he/she searches on the housing market for one particular property that maximizes his/her utility level (Knight, 2002). Based on such interactions between buyer and seller, a logical progression is how to reach that optimal list price for the seller. According to Merlo and Ortalo-Mague (2004), it is typically assumed that a seller faces a trade-off between the rate of arrival of buyers and the sale price. In other words, a lower list price increases the arrival rate of buyers but reduces the chances of high-price transaction (Haurin, 1988). To look at the list price from another perspective, Horowitz (1992) and Yavas and Yang (1995) suggest that it has two different roles. List price is either the seller's reservation

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

7 7

price or an upper bound for the seller's reservation price that will be acceptable wben a prospective buyer matches the offer. Specifically, a seller's choice of an offer price is a function of market price for the seller's home, its characteristics, and the availability of properties in the market (Mayer, 1995). Subsequently, this list price influences the arrival rate of buyers, as well as the distribution of their bids. However, one reason for sellers to set a higher list price is that they have a higher loan-to-value ratio, as discovered in an investigation of the Boston condominium market by Genesove and Mayer (1997). They also have a higher expected TOM and tend to obtain higher transaction prices than owners with proportionately less debt. Eor the latter, TOM has been studied in three different dimensions. The first dimension is the relationship between TOM and price concession ratio; the second is that between TOM and sale price. The last dimension focuses on the how search theory explains TOM (Kalra and Chan, 1994). With regard to the first dimension, it has been shown in Belkin, Hempel, and McLeavey (1976), Kang and Gardner (1989), and a more recent study by Leung, Leong, and Chan (2002) that TOM has a positive correlation with the ratio of list price to sale price. Kalra and Chan (1994) have a similar finding, although the impact of price concession appears to be more noticeable for the transactions of higher-priced homes. However, it is not always the case, as Jud, Winkler, and Kissling (1995) report no significant impact of price concession on TOM. Instead, they find that it is influenced by the standard deviation in sale price. They suggest that a wider distribution of prices may encourage sellers to hold out for higher transaction prices, therefore resulting in a longer TOM. Meanwhile, some other researchers have concentrated their efforts on examining the relationship between TOM and sale price. Yet, the resulting empirical evidence on this matter has been mixed and somewhat inconclusive. Specifically, the sign on TOM has been found to be rather unpredictable (Asabere, Huffman, and Mehdian, 1993). For instance, some studies have discovered a positive correlation between TOM and sale price (e.g., Trippi, 1977; Miller, 1978; Asabere and Huffman, 1993), whereas an inverse relationship between the two was found in Cubbin (1974). Erom a theoretical standpoint, search theory has been frequently used to explain the relationship between sellers' choice for listing price and TOM. Generally speaking, the search theory usually assumes a direct relationship between selling price and TOM. Sellers are willing to wait longer for the probability of a better offer from a buyer (e.g.. Miller, 1978; Turnbull and Sirmans, 1993; Merlo and Ortalo-Magne, 2004). Yet this may not always be the case, as potential buyers might regard properties that remain unsold for a relatively long period of time as defective. According to Taylor (1999), this situation constitutes a phenomenon called negative herding. As such, a longer TOM may result in a lower sales price. A number of papers have also linked sellers' pricing strategies to TOM. It is often found that the more overpriced a flat, the slower it will be sold, as shown on
J R E R V o l . 3 4 N o . 3 - 2 0 1 2

3 7 8

H u i ,

Vv'ong,

a n d

V^ong

numerous occasions in the literature such as Miller (1978), Ong and Koh (2000), Anglin and Wiebe (2004), Li (2004), and Merio and Ortalo-Magne (2004). Asabere, Huffman, and Mehdian (1993) report that both the overpricing and underpricing strategies would prevent sellers from achieving the optimal TOMs that correspond to the highest sale prices (in net present values). Besides property prices, housing characteristics are other important determinants of TOM. Miller (1978) points out that, given a housing supply and demand conditions, a property's TOM is a function of its attractiveness (attributes) in comparison to other properties on the market. A key element of a flat's attractiveness is its age. However, there has been no consensus about how age affects TOM. For instance, Zuehlke (1987) reports that the age of a property is an important determinant of TOM, but Kalra and Chan (1994) and Ong and Koh (2000) reach a different conclusion. On the other hand, Haurin (1988) suggests that TOM is directly affected by the atypicality of a house. In other words, it takes more time for properties with more unusual features to market before they can be transacted (Kalra and Chan, 1994). For example, Ong and Koh (2000) found that in condominiums in Singapore, those that were on lower floors tend to have longer TOM. Li (2004) found that property characteristics, such as pool view, number of bedrooms or toilet and private enclosed space, are not significant in explaining TOM; however, properties on higher floors that face south or north, and with more convenient parking will have shorter marketing times. Researchers have also studied the impact of (housing or financial) market conditions on TOM. It has been reported in several studies (e.g., Haurin, 1988; Kalra and Chan, 1994; Yang and Yavas, 1995; Anglin, Rutherford, and Springer, 2003) that TOM is influenced by both local and national economic conditions, as well as being subject to strong seasonal effects. In a study conducted by Leung, Leong, and Chan (2002) in Hong Kong, inflation is another critical factor that influences TOM.

S t u d y Me t h o d o Io g y This study seeks to provide some empirical insight into the effects of various factors on TOM. In particular, we hypothesize that both the above-market price and the sale price have direct influences on TOM. The precise meaning of 'abovemarket price' has to be defined carefully, however. Yavas and Yang (1995) point out that it is potentially problematic to regress TOM on sale price (or likewise, on the ratio of sale price to list price, etc.) because of the simultaneity of the two quantities. [Belkin, Hempel, and McLeavey (1976) and Kang and Gardner (1987) are susceptible to this problem, although some authors like Cubbin (1974) are aware of this fact.) To resolve this simultaneity problem, it is now popular to adopt a two-stage approach. The first stage is to use hedonic regression to obtain the expected value of sale price (or list price if needed). Jud, Seaks, and Winkler (1996), for instance, perform an ordinary least squares (OLS) regression of sale price Fg on certain housing or neighborhood characteristics X:

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 7 9

F, = H'X + s,,.

(1)

xpected sale price P^ is then obtained as the predicted value of the sale

P = H'X.

(2)

In the second stage, the expected prices obtained from the first stage are used as explanatory variables (among other variables) to model TOM. Cox's (1972) proportional hazard model has now become well-known. Suppose TOM is a continuous random variable. Let f(t) denote the probability density function of TOM, F(t) = Fr(TOM < t) its cumulative distribution function and S(t) = Pr(TOM > i) = 1 - E(t) its survival function, then its hazard rate:

h(t) = f(t)/S(t).

(3)

The hazard rate measures the likelihood that a housing unit will be sold at time t, given that it has not been sold at time /. If two houses A and B have hazard rates /i,(0 and h2(t) respectively, with /,() > h2(t) for all t, then house A tends to have a shorter TOM than house B does. Now, in the Cox model, the hazard function of TOM is assumed to take the following form:

h(t) = ho(t)A(X, Z),

(4)

where /^o(0^ called the baseline hazard rate, is a nonnegative function of time and A is a time-independent function that may depend on some housing attributes X (age of property, number of bedrooms, whether it is near the central business district, etc.) some other variables Z that affect both TOM and home prices (months of inventoi^, velocity of sales, P^, P[^, mortgage interest rate, unemployment rate. Consumer Price Index, gross domestic product, etc.). Different definitions of /2o(i) and A give rise to different models of TOM. Eor example, Jud, Seaks, and Winkler (1996) assume that:

A = exp(-/3o - B'X - C'Z),

JRER

Val.

3 4

Na.

3 - 2 0 1 2

3 8 0

H u i ,

V / o n g ,

o n d V\/ong

where the log price difference log(P^/P^) is included in Z as a measure of abovemarket pricing. /zo(i) = pt^'^ is the baseline hazard rate of the Weibull distribution. When p = 1, the baseline hazard rate becomes constant and the density function of TOM decays exponentially. Hence Cox model reduces to the Exponential model in this case. Cox models are not the only hazard models that appear in TOM literature. The semi-log regression in (3), for instance, can be recast as a hazard model with the following density and survival function:

fit) = <p[log{At)/a]/(crt), S(t) = 0[-log(At)/cr],

(7) (8)

where a is the standard deviation of SJOM in 0) ^nd tp, <P are respectively the density function and cumulative distribution function of the standard normal distribution. This model is sometimes called the log-normal model as TOM follows a log-normal distribution in this case. Another model that has appeared in the literature is the log-logistic model (e.g.. Glower, Haurin, and Hendershott, 1998), in which:

Note that these two models are not proportional hazard models, so we do not speak of their baseline hazard rates. At any rate, given that A{X, Z) is fixed, the Cox model, the log-normal model, and the log-logistic model are all oneparameter models. In this study, we adopt the aforementioned two-stage approach to examine the relationship between TOM, above-market price, and list price. However, in the first stage, since the hedonic regression (1) obviously may lead to a negative sale price, we opt to follow Anglin, Rutherford, and Springer (2003) and use a semilog regression in the first stage:

E{\n{P,)) = H'X + sp.

(11)

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 8

The above-market price is defined as:

A = ln(PJ - E{\n{Ps)).

(12)

Regarding (12), note that the formation of A is somewhat similar to the concept of Degree of Overpricing (DOP) in Anglin, Rutherford, and Springer (2003). However, there are two noticeable differences: (1) the expected transaction/sale price, instead of expected list price, is being used for the computation of A for this study; and (2) as stated earlier, the results derived from the hedonic pricing model are used for the formulation of E(\V{PS)) in this study, in which only housing attributes (X) are included. Meanwhile, the expected list price (in formulation of DOP) in Anglin, Rutherford, and Springer (2003) is subject to both housing attributes (X) and market conditions (Z), which differs from conventional hedonic pricing model specifications. Although there might be issues relating to simultaneity bias, there is an expected relationship between list price and expected sale price (unlike expected list price, this is a realized, transaction price that buyers are able to obtain). This somehow drives the negotiation process and determines TOM. A would be a good measure of the effect upon TOM. The variables in X (housing attributes) are as follows: ln{SIZE) = The natural logarithm of the size (in square feet) of the housing unit; KLN = A dummy variable that the housing unit for sale is located in the Kowloon Peninsula; NT = A dummy variable that the housing unit for sale is located in the New Territories; E, S, W, N, NE, SE, and AW = Dummy variables that represent the orientation of the housing unit (the direction where the living room of an apartment is facing); VACANT = A dummy variable that the housing unit is in vacant possession; OWNER = A dummy variable that the housing unit is owneroccupied; RENTER = A dummy variable that the housing unit is renteroccupied; FLOOR = A dummy variable; 1 if the flat is higher than the 20* floor, 0 otherwise; SPACES = The number of parking spaces entitled to the housing unit; BEDROOMS = The number of bedrooms; CLUBHOUSE = A dummy variable that there is a clubhouse; and POOL = A dummy variable that there is a swimming pool.

J R E R

V o l .

3 4

N o .

3 - 2 0 1 2

3 8 2

H u i ,

V ^ o n g ,

a n d W o n g

The variables in Z (macroeconomic factors) are as follows: RPPI/RPRI = The return (rate of change) of the property price (rental) index, computed by the Rating and Valuation Department, the HKSAR Government, on a monthly basis; UNEMPLOYCH = The changes in unemployment rate in Hong Kong; CPIYYPERCH = The year-to-year growth of the Consumer Price Index; BLRCH = The changes in the best lending rate; and ROCINCOME = The adjustments in median household income. In the second stage, a Cox survival model will be deployed. Both X and Z will be considered explanatory variables and price-related items such as A or ln(P) will be included in Z. Various researchers (e.g., Zuehlke, 1987; Haurin, 1988; Larsen and Park, 1989; Yang and Yavas, 1995; Ong and Koh, 2000; Genesove and Mayer, 2001; Anglin, Rutherford, and Springer, 2003; and Li, 2004) have included different variables X and Z in specifying A. In this study, the proportional hazard assumption is presumed, as Cox (1972) finds that under such assumption, one can actually use a partial likelihood method to estimate the coefficients B and C in (7) without knowing the baseline hazard function. In the Cox model, if the coefficient of a variable is positive, then an increase in the value of the variable would result in a longer TOM for flats; otherwise, an increase in the value of the variable would tend to make TOM shorter.

Sample Data The samples were collected on a cross-sectional basis, including their initial asking prices and eventual sale prices, from January 2003 to June 2006. This was a period when Hong Kong had gradually been recovering from tbe lasting impact of the Asian Financial Crisis in the late 1990s, only to be temporarily hit again due to the SARS epidemic. The number of properties with negative equity reached a historic high of 106,000 by the end of the second quarter of 2003 (Exhibit 1) and unemployment rate was higher than 8% (Exhibit 2). Afterwards, the economy improved gradually, as seen in the price trend and number of transactions of real properties in Exhibit 3. The unemployment rate continued to fall and the interest rate had started climbing in 2005 due to higher demand for mortgage loans, as reflected by the increase in property transactions in 2004 and 2005. Unlike in many western countries. Hong Kong has arguably the largest public housing system in the world, along with Singapore. Yet, as a response to the controversies surrounding the government's assistance homeownership (HOS) program especially during the economic downturn, the Housing Authority (HA) has ceased the construction and sale of HOS flats since 2003. This policy change in some ways has helped accelerate the recovery of Hong Kong's property market. The data set of the study covers residential sales only. Completed transactions from the initial listing price to the sale agreement, including individual marketing time in elapsed calendar days, were included in the data set for empirical study.

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 8 3

Exhibit 1 I Number of Residential Negative Equity Cases

Source: Hong Kong Monetary Authority.

The transactions cover pre-sale units (units under construction), owner-occupied units, renter-occupied units, and vacant-in-possession units. The data refer to domestic flats randomly selected from 168 development estates and individual buildings of various flat sizes from less than 272 sq. ft. to over 3,882 sq. ft. A total of 62, 43, and 63 estate blocks and individual buildings in Hong Kong Island, Kowloon, and the New Territories are covered, respectively. There are 4,010 domestic flat sales records of properties listed as early as January of 2003 and sold as late as June of 2006. , . . The data used for this study, unlike many similar studies on Hong Kong's property market, is not from an official source. The reason is that official data such as the EPRC, based on trading records from the Land Registry of the HKSAR, does not necessarily provide information such as a flat's orientation and its current tenure status due to legal concerns. The dataset used for this study, obtained from Midland Realty of Midland Holding Limited, one of the largest property agencies in Hong Kong, allows us to investigate the possible impacts of aforesaid attributes on property price and TOM, which is something that might not be done via the use of government data. This dataset is systematic and reliable, and generally regarded as a suitable reference for studying Hong Kong's property market.
JRER Vol. 3 4 N o . 3 - 2 0 1 2

3 8 4

H u i ,

W o n g ,

a n d

W o n g

Exhibit 2 I Hong Kong's Unemploymenf Rate and Best Lending Rate

Year/Month Unemployment Rate -Best Lending Rate

Sources: Census & Statistics Department and HSBC.

E x h i b i t 3 I Price Trend and Number of Transactions of Real Propertie

Year Sale and purchase agreements for residential praperties New completions of residential flats Vacancy rate (%)

2003 71,576 26,397 6.8

2004 100,630 26,036 6.2

2005 103,362 17,321 6.0

2006 82,472 16,579 5.9

Note: The saurce is the Rating and Valuation Department.

Summary statistics with regard to the tenure status, price levels, and the distribution of TOM are presented in Exhibits 4 and 5 [see the Appendix for a comparison with the sample deployed in another Hong Kong study on TOM (Leung, Leong, and Chan, 2002)].

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 8 5

E x h i b i t 4 I Summary Statistics

Sample No. Tenures Pre-sale Vacant in Possession Renter-occupied Owner-occupied Unknown i..' . Overall Sale Prices $l-$1.99mil $2 mil-$3.99 mil $4 mil-$5.99 mil $6 mil and over Overall Note: "Excludes cases with missing values. 150 2037 323 1493 7 4010 1625 1117 %

3.7
50.8 ^' '
8.1

37.2 0.2 100 40.6 27.9 10.3 21.2 100 '

. ' '

414 851
4007

E x h i b i t 5 I Disfribution of Marketing Time (TOM)

P'Yeor Sold 1 st Qtr 2nd Qtr 3rd Qtr 4th Qtr 2"'' Year 3'^ Year After 3rd Year 1096-1141 1 0.03 4010 100 Total

No. of days 1-90 N 2175 54.2

91-180 181-273 274-365 366-730 731-1095 820 20.4 461 11.5 244 6.1 277 6.9 32 0.8

Note that about half (51.0%) of the transacted properties are vacant. Statistics also show that about 40% of them are in the low end of the market (the most common flat type in Hong Kong), with a list price below HK$2 million. A closer look at the marketing periods of individual properties provides some insight to the speed of sales. Exhibit 6 shows the overall distribution of TOM. Slightly over half (54.16%) of the properties were transacted within three months after they had been put on sale, followed by 20.47% within the second quarter. In other words, about three-quarters (74.63%) were transacted in half a year's time. Only about 8% were sold over one year's time. In short, overall sales were
JRER Vo 3 4 No. 3 - 2 0 1 2

3 8 6

Hui,

W o n g ,

a n d

W o n g

E x h i b i t 6 I Hedonic Regression of Natural Log Sale Price I

Variables Constant lnS;Z

Coeff. 5.472"' 1.487'" -0.154'" -0.527"* -0.020" -0.307"* 0.188"* 0.194'" 0.025* 0.048" 0.043'" 0.053"* 0.069"' -0.188' 0.043* 0.090'" 0.020'

Std. Dev. 0.112 0.020 0.013 0.014 0.010 0.017 0.014 0.015 0.014 0.019 0.013 0.014 0.013 0.111 0.025 0.018 0.010

T-value 48.765 73.542 -11.549 -37.836 -1.969 -18.047 13.642 12.948 1.782 2.561 3.150 3.921 5.251 -1.700 1.679 5.049 1.957

Sig. 0.000 0.000 0.000 0.000 0.049 0.000 0.000 0.000 0.075 0.010 0.002 0.000 0.000 0.089 0.093 0.000 0.050

KLN NT
BEDROOMS DINING CLUBHOUSE POOL SPACES

W NE SE NW
UNKNOWNTEN PRSALE RENTER OWNER

Notes: N = 3,998, excluding cases with missing values. The adjusted R-square = 0.900. 'Significant at the 5% level. ** Significant at the 1% level. "Significant at the 10% level.

effected relatively quickly, despite the fact that there is a wide range of TOM from 1 day to 1,141 days.

Results a n d Findings Stage 1 : Hedonic Pricing Regression Model The results of the hedonic regression, using the full sample, are shown in Exhibit 6. Statistically insignificant variables have already been removed through a stepwise selection process in the regression model, and only those with 10% significance or less remain. Four tenure variables {RENTER, OWNER, PRSALE, and UNKNOWNTEN), a number of orientations {NE, NW, SE, and WO, two location variables {NT and KLN), and other property attributes {inSIZE, POOL,

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s 3

E x h i b i t 7 1 Correlation between Selected Economic Variables between January 2003 and June 2006

RPPi RPPI BLRCH UNEMPLOYCH CPIYYPERCH RPRI ROCINCOME Note: 1

BLRCH -0.065 1

UNEMPLOYCH -0.259 -0.008 1

CPiYYPERCH -0.091 0.370* -0.129 1

RPRI 0.362* 0.197 -0.531** 0.316*

ROCINCOME 0.443** 0.165 -0.403" 0.159 0.346*

* Significant at the 5% level. ''Significant at the 1% level.

DINING, SPACES, CLUBHOUSE, and BEDROOMS) are significant at least at the 10% level. In regards to fiat location, those located on Hong Kong Island are more costly than others in Kowloon (KLN) and the New Territories (NT), primarily due to its better accessibility to tbe Central Business District. Moreover, tbe correlation between transaction price and the RENTER variable is the highest among the four significant variables, which points to a relatively higher transaction price for a renter-occupied flat than for others that have identical attributes. The reason is that it generates a higher level of transaction cost for the owner (mostly in terms of time) in order to release the flat from the current leasing contract. In terms of the impact of property attributes on transaction prices, the hedonic model suggests that the presence of a swimming pool (POOL) inside a housing estate produces the most positive effect on flat prices, followed by clubbouse (CLUBHOUSE) and carpark spaces (SPACES). Also, the negative relationship between both DINING/BEDROOMS and transaction price suggests, albeit indirectly, that the size of the rooms play a role in the final transaction price. Lastly, condominiums with living rooms facing northwest, southeast, west, and northeast have positive correlations with their resultant transaction prices. Overall speaking, the selected housing attributes and economic variables explain 90% of the variations of the nominal sale price (in natural log). Stage 2: Estimation of TOM with Cox Survival Model Cox Survival Model Results (Eull Sample). Before conducting the Cox survival model, a correlation test is carried out for the variables in the Z vector. The reason for doing so is to avoid the possible issue of multicollinearity among independent variables in a regression analysis. The results in Exhibit 7 show that many of these variables are somewhat highly correlated.

JRER

Vol.

34

No.

3 - 2 0 1 2

3 8 8

H u i ,

VV'ong,

a n d V / o n g

Therefore, in order to avoid the bias stemmed from multicollinearity of variables in Vector Z, only RFFI and UNEMFLOYCH [which is also included in, for example, Jud, Seaks, and Winkler (1996)] are included in the Cox regression model. The Cox model results of the full sample are illustrated in Exhibit 8. Interestingly, the findings point out that the only significant variables are the location variable KLN and the three YEAR dummy variables (YEAR2003, YEAR2004, and YEAR2005). The negatively significant relationship between KLN and TOM indicates that flats located in Kowloon appeared to be transacted slightly sooner than those in other parts of Hong Kong. Of the three YEAR dummy variables, YEAR2004 has the most significantly negative correlation with TOM, indicating that transactions in 2004 in general tended to take place relatively sooner than those in other years. On the other hand, as all three YEAR variables are negative, it means that it took relatively longer for flats to be transacted in 2006 (the reference variable). The reason why the level of TOM was comparatively low in 2004 is that the Hong Kong economy began to recuperate from the adverse effects brought by the SARS epidemic a year before. The unemployment situation was gradually improving. Property price level was escalating, which remarkably helped reduce the amount of negative equity flats in the territory (Exhibit 1). In addition, the mortgage rate, which is directly related to the best lending rate, was at a stable low level in 2004 (Exhibit 2). These factors provided the conditions for a new demand for residential properties. With decreasing negative equity cases, more flats were available on the market, which reduced search costs, resulting in faster transactions (i.e., lower TOM) in general. In the meantime, the insignificance of other independent variables in both vectors X and Z insinuates that, due to changing economic conditions during the study period, the impacts of these housing attributes and macroeconomic factors might have been fluctuating and inconsistent over time (i.e., large standard deviations), to the point where statistically significant estimates are not possible to obtain when the full sample is deployed. In order to illustrate such year-specific impacts, four sub-period investigations were conducted (2003, 2004, 2005, and 2006). Unlike in many previous studies, the first two years were regarded as a post-crisis (or recovering) period when Hong Kong's economy just began to recover from the effects of the SARS epidemic. In contrast, the next two years were similar to the situations depicted in other studies, as Hong Kong's economy gradually recovered. Sub-period investigations allow for comparisons of the effects (and adjustments) of different variables on TOM in various years, which could contribute to the existing literature considering the nature of Hong Kong's economy during the period. The respective Cox model results for each sub-period are shown in Exhibit 9. Cox Model Results (Sub-period Samples). The findings suggest that many variables are statistically significant when sub-period samples are used for the investigations. This is noticeably different from that with the full sample as seen

M a r k e t i n g

Time

a n d P r i c i n g

S t r a t e g i e s

3 8 9

E x h i b i t 8 I The Cox M o d e l Results for the Full Sample

Variables A ' -

Coeff. -0.136* 0.327** \ : > . -?.

S.E. 0.077 0.135 0.591 0.020 0.047 . . J. -,.- 0.059 0.072 0.071 0.074 0.073 0.058 0.058 . . 0.057 0.087 ' . 0.088 0.100 0.053 0.033 0.049 ' -'i- , 0.053 0.006 0.131 0.035 0.064 . 0.053 0.051

HPs] [iniPsW

0.454 -0.015
-0.087*

KLN NT E
S
W

-0.066 -0.076

.'

'

. .

-0.011 0.057 0.055

N
NE

'

I
...:.

>
') '

0.015 0.031 -0.014

SE NW VACANT OWNER
RENTER SPACES

^l

-0.113 ' . ' -0.089 -0.099 0.008

./.: . . .1

, ' ;

BEDROOMS CLUBHOUSE POOL RPPI UNEMPLOYCH FLOOR YEAR2003 YEAR2004 YEAR2005 ... ' . ' .^

0.000 -0.015 -0.024 0.007 -0.179 -0.014 -0.194*** -0.356*** -0.215***

Notes: The number af observations is 4,009. The 2 log-likelihood is 58471.432. The chi-square is 74.179. * Significant at the 5% level. **Significantatthe 1% level. *** Significant at the 10% level.

J R E R

I V o l .3 4

N o .

3 - 2 0 1 2

3 9 0

Hui,

V / o n g ,

and

W o n g

E x h i b i t 9 I The Cox Model Results for the Sub-period Samples

Variables

2006

S.E.

2005

S.E.

2004

S.E.

2003

S.E.

A A^
In(Ps) [ln{PsW

2.792*" 0.290

1.653 *** 0.160 -0.537*** 0.139 -3.860*** 0.259

-1.767*** 0.496 -0.844 *** 0.286 - 1 . 8 6 1 * * * 0.349 -2.369*** 0.437 3.468** -0.110** -0.102 -0.139 -0.096 -0.121 -0.010 0.360** 0.124 -0.003 0.071 1.588 2.624 ** 1.094 2.051* 1.070 0.035 -2.434 0.061 1.484 0.050 ; A

0.052 -0.082 ** 0.119 0.150 0.184 0.183 0.185 0.168 0.149 0.151 0.146 0.114 0.196 0.027 0.100

0.036 -0.064* 0.083 -0.100 0.104 0.079

KLN NT E S W N NE SE NW
VACANT OWNER RENTER ; SPACES BEDROOMS CLUBHOUSE POOL RPPI UNEMPLOYCH FLOOR

0.084 -0.432*** 0.110 0.102 -0.678*** 0.135 0.127 -0.085 0.121 0.065 0.158 0.163 0.179 0.167 0.132 0.135 0.135 0.220 0.222 0.266 0.131 ;

0.130 -0.199 0.128 -0.080

0.364 *** 0.124 -0.109 0.198 0.203 " 0.113 0.213 ** 0.130 -0.095 0.103 -0.126 0.104 0.035

0.136 -0.104 0.134 0.045

0.100 -0.098 0.097 -0.126 0.097 -0.339** 0.119 -0.159

0.101 -0.078

1.378*** 0.416 1.323*** 0.419 1.135" -0.137 0.038 0.076 -0.003 0.442 0.136

0.512 *** 0.185 - 0 . 6 1 1 * * * 0.461 ** 0.515 ** 0.052

0.187 -0.473*** 0.120 -0.143 0.203 -0.368** 0.092 0.021 0.142 -0.540** 0.090 0.058 0.255*

0.083 -0.114 * 0.128 -0.076 0.139 -0.008

0.058 -0.031 0.087 -0.073 0.093 -0.110

0.238'** 0.080 0.109 0.118 :.

0.085 -0.020 0.095 0.012

-0.127*** 0.017 -0.029 *** 0.010

2.242*** 0.317

0.130*** 0.011 1.683*** 0.211 0.099 0.061

-0.083*** 0.014 -0.142 0.045 0.228 0.086

-4.638*** 0.268 0.060

-0.152*

0.089 -0.123 **

Notes: The number of observations for 2006 = 624, 2 log-likelihood = 8,934.525, and chisquare = 363.783. The number of observations for 2005 = 1,345, 2 log-likelihood = 19,086.480, and chi-square = 482.157. The number of observations for 2004 = 1,307, 2 loglikelihood = 18,408.744, and chi-square = 334.677. The number of observations for 2003 = 733, 2 log-likelihood = 9,937.397, and chi-square = 997.106. * Significant at the 5% level. ** Significant at the 1% level. *** Significant at the 10% level.

ij ^

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 9

in the previous section [cf., Leung, Leong, and Chan (2002) presented the opposite results]. As seen in Exhibit 9, a number of housing characteristics as well as macroeconomic factors are significant at least at the 10% level. Effects of Property Price & Pricing Strategy. Assuming a non-linear relationship between TOM and property price variables, the above-market price {A), sale price ln(P5), and their respective square-terms are included in Z as explanatory variables. Since \n{P^ is highly correlated with ln(5/Z'), the latter is removed from Z. Exhibit 8 gives parameter estimates for the Cox model. As one can see, the abovemarket prices A and A- are significant in the pooled sample and the four subperiods, indicating that overpricing does have a non-linear impact on the TOM for a residential property. Across the sub-periods, the findings further suggest that TOM was relatively shorter in 2003 and 2004 than in 2005 and 2006. In other words, for two properties with the same price mark-up, the one in 2003 is expected to be sold faster than that in 2006. Why were properties able to be transacted sooner in 2003/2004 than in 2005/ 2006 as the economy gradually improved? Considering the amount of properties with negative equities in 2003 and 2004 (Exhibit 1), it was more costly for potential buyers to obtain better bargains as housing supply was lower. As a result, it was entirely possible for sellers to overprice their flats to obtain additional premiums yet able to sell them sooner during these two years. Yet, from 2005 onwards, as Hong Kong's economy was gradually recovering, the number of negative equity flats drastically decreased (Exhibit 1). This means more flats were available on the secondary property market. Besides, potential buyers were subject to higher interest rates. This reduced demand for housing. In such circumstance, it was the usual selection of either a quicker turnover at a lower list price, or a higher list price but with a longer TOM. In summary, the effects of overpricing on TOM change overtime, upon the supply of alternatives on the market and the opportunity cost incurred for potential buyers. Similarly, the impact of sale price on TOM is also non-linear. The coefficient of sale price (in natural log) is significantly positive and its square-term is significantly negative, from 2004 to 2006. This means that for more costly properties, TOM would increase at a decreasing rate. This finding adds to knowledge in that previous studies only found linear relationships between TOM and sale price (e.g., Cubbin, 1974; Trippi, 1977; Miller, 1978; Asabere and Huffman, 1993). Geographical Location. Geographical location of the housing unit is shown be significant only in 2003, which suggests that flats in Kowloon and the New Territories on average took less time to sell than properties in Hong Kong Island. This contrasts with the result of Ong and Koh (2000), who find that housing units near the central business district tend to have shorter TOMs. Orientation. The orientation of a residential flat does not have very significant impact on TOM, aside from the NW variable in 2003 (negative), W/NE/NW variables in 2005 (all positive), and the N variable in 2006 (positive), all with SW

JRER

Vol.

3 4

N o .

3 - 2 0 1 2

3 9 2

H u i ,

W o n g ,

a n d W o n g

being the control variable. The findings are slightly different from that in Li (2004). An underlying reason is that the listed price of flats with relatively inferior orientations (e.g., less popular ones such as those facing north and west) have been already adjusted accordingly in response to the demand. Tenure. Zuehlke (1987) shows that TOM can be influenced by the vacancy of a housing unit. In this study, we further examine how TOM would be affected by other kinds of occupancies. With the exception of 2003, all three tenures (vacant, owner-occupied, and renter-occupied) are significant at the 1% level relative to the omitted category (pre-sale flats). Their impacts on TOM were negative in the first two periods, yet positive in the last two. This suggests that pre-sold residential units (the control variable) on average have the longest TOMs in 2003 and 2004, while having the shortest TOMs in the next two periods. Meanwhile, the impact of owner-occupied, renter-occupied, and vacant fiats are very close. A probable reason for the relatively lower TOM for vacant properties is that, although the number of residential negative equity cases was greatly reduced in 2004, the property price level was not high enough for homeowners to obtain desirable profits from a transaction, considering the expenses involved (e.g., searching for another flat and/or terminating the leasing contract with renters). Owners of vacant flats, however, were able to take advantage of this situation and sold their flats faster, due to lower transaction costs and the instant availability of their flats. This resulted in a rather noticeable drop in vacancy rate of Hong Kong's residential properties, from 6.8% in 2003 to 6.2% in 2004 (Exhibit 4). In a sense, our findings regarding vacancy on TOM extend Zuehlke's (1987) conclusion. Other Housing Attributes. Similar to the findings in Ong and Koh (2000), the findings indicate a significantly negative relationship between TOM and the dummy variable ELOOR. However, this only applies to 2005 and 2006. In other words, in these two sub-periods, residential properties higher than the 20* floor of a building were sold faster than those lower. Another factor that has significant impact on TOM is the amount of carpark space, but only negatively in 2006. Except these two, the other selected housing attributes are shown to have no significant effect on TOM. It should be noted tbat the insignificance of the number of bedrooms on TOM is in line with the findings in Anglin, Rutherford, and Springer (2003). Macroeconomic Variables. Changes in unemployment rate (UNEMPLOYCH) first show a significantly positive relationship with TOM in 2004, then negative in 2005, and eventually positive again in 2006. In general, an increase in unemployment rate, reflecting worse economic conditions, tends to lengthen TOM, as potential buyers would become a lot more cautious when it comes to costly decisions such as home purchases. Yet in 2005, a negative impact between the change in unemployment rate and TOM is found. It was the time when the unemployment situation had been gradually improving (Exhibit 2). Even though the economy was recovering, homeowners were expecting even better economic

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 9 3

conditions in the future. This insinuates a higher price (and better return) for their flats. In other words, the noticeably longer TOM recorded under an improving economy in 2005 is due to the expectation from sellers for even better returns in the near future. With regard to the correlation between TOM and general property price trend (RFFI), the impacts of RRFI on TOM are significantly negative in 2003, 2005, and 2006, but significantly positive in 2004. The reason is that, with the expectation of an even higher price for properties in the near future, as reflected in RPPI, a buyer will purchase a property of his/her choosing as soon as possible although the list price may be higher than the buyer's expected market price. This constitutes faster transactions and generally a lower TOM. The positive correlation between TOM and RFFI in 2004 can be attributed to a much larger supply of second-hand properties on the market. The amount of negative equity cases had rapidly attenuated from more than 100,000 cases in 2003 to only 19,000 cases by the end of 2004 (Exhibit 1), indicating that sellers were more willing to sell their flats. Meanwhile, new completions of flats in 2003 and 2004 were more or less identical (Exhibit 4). Because of these factors, the search cost incurred for buyers to look for better bargains had significantly reduced. As a result, a longer TOM arises as sellers face higher competition for offers.

Conclusion

This study examined how overpricing of properties, along with a variety of housing attributes and market conditions, can affect their TOM. The duration of a listed property on the market is important because it represents the liquidity of real assets, as well as the respective trade-offs for buyers and sellers. The study deploys a new two-stage methodology. In the first stage, the above-market price (overpricing) is measured by the difference between list price and expected sale price, which the latter is regressed through a hedonic pricing model. Then in the second stage, the effects of various factors including above-market price on TOM are tested by a Cox model. Our study utilizes a first-hand database that covers about 4,000 transactions of residential units on the Hong Kong property market under different living tenures, flat sizes, price ranges, and other physical characteristics. This is believed to be more reliable and comprehensive to other secondary data available. The empirical results show that TOM has significant relations only with abovemarket price, the location variable KLN, and the three YEAR dummy variables when the sample is analyzed as a whole. The significance of overpricing (or the above-market price) on TOM is non-linear. This finding proffers an extension to the existing literature (e.g.. Miller, 1978; Ong and Koh, 2000; Anglin and Wiebe, 2004; Li, 2004; and Merlo and Ortalo-Magne, 2004) concerning how such pricing strategy influences a property's TOM.

J R E R

V o l .

3 4

N o .

3 - 2 0 1 2

3 9 4

H u i ,

W o n g ,

a n d

W o n g

Nonetheless, the sub-period investigations offer new, in-depth insights into the dynamics of TOM and overpricing that are not observable when it is investigated as a whole. In these investigations, factors such as above-market price, sale price, changes in unemployment rate, general property price trend (RPPI), and an apartment's tenure status are all significant on TOM. These findings extend the existing literature in two different ways. Firstly, as a response to Asabere, Huffman, and Mehdian (1993), this study elucidates that whether to overprice or underprice a property mainly lies in the market/economic situations of a specific period, as well as the supply/availability of alternatives on the housing market. Secondly, our findings further extend the literature (e.g., Haurin, 1988; Kalra and Chan, 1994; Yang and Yavas, 1995; Anglin, Rutherford, and Springer, 2003) which states that TOM itself is directly influenced by local and national economic conditions. Actually, such economic conditions play an indirect, but important, role in the effectiveness of overpricing as a strategy in optimizing TOM as well that is particularly so when the economy either is during or starts recovering from a crisis (i.e., 2003 and 2004). People are uncertain about future market situations. To make situations worse than it already is, housing prices are so much lower than that when homeowners bought their flats. A transaction during this period most likely means losses for them (i.e., negative equities), regardless of their pricing strategies. These factors provide the conditions for lower supply on the secondary property market. Worse, the significance of RPPI on TOM, which is another new finding in this study, shows that expectations of a downward property price trend in the near future cause buyers to postpone their buying decision in hopes of even better prices in the future. It results in fewer transactions on the property market. This, in addition to the already-low supply of housing, incurs higher cost (both financial and time) for buyers to obtain information for better deals than usual. Thus, it is easier for homeowners to sell their flats sooner in that period, provided the same amount of overpricing (i.e., a lower TOM). By contrast, as the economy gradually recovers (i.e., 2005 and 2006), property prices soars accordingly. More flats are available on the secondary property market as the number of negative equity flats significantly decreases. Information of better bargains is much easier to obtain. In other words, the search costs for buyers attenuates. Thus, overpricing as a strategy would no longer be as effective as it was before in optimizing TOM for sellers. Similar arguments can be used to explain the effects of variables, such as RPPI, ELOOR, RENTER, OWNER, and VACANT, on TOM. Similar to above-market price, the significant, non-linear relationship between sale price (with the exception of 2003) and TOM extends the literature (e.g., Trippi, 1977; Miller, 1978; Asabere and Huffman, 1993; Cubbin, 1974). This finding shows that luxury residential properties that generally occupy larger space have longer TOMs, or lower turnover rates. Similar to other previous studies, this study primarily concentrates on overpricing strategy and its impact on a property's TOM in a growing economy. Yet, a constantly growing economy is not always the case, as the sub-prime crisis in the

M a r k e t i n g

T i m e

dP r i c i n g

S t r a t e g i e s

3 9 5

U.S. has shown the world the impact of various economic crises on the real estate market. As the world has become more globalized, a more fluctuating market environment has ensued. In other words, both the stock market and the real estate market become increasingly more susceptible to demand shocks, both from within and from the outside. As a result, future studies should focus on how property owners respond to such demand shocks through their pricing strategies (e.g., the possibility of underpricing to ensure faster turnover) in such market conditions.

Appendix The table below compares the samples deployed in another Hong Kong study on TOM (Leung, Leong, and Chan, 2002). It is worth noting that the statistics of TOM as recorded in Leung, Leong, and Chan (2002) is much higher than that of our sample. In addition, this study uses a first-hand database provided by one of the biggest real property agents in Hong Kong, whereas in Leung, Leong, and Chan (2002) a secondary data set was utilized. Furthermore, our study primarily investigates how overpricing, among other housing attributes and macroeconomic variables, influence TOM in a period which the economy was gradually recovering from a low point in 2003 when Hong Kong was under the onslaught of the SARS epidemic. On the other hand, Leung, Leong, and Chan (2002) focused on a period in which Hong Kong's economy had been more fluctuating.

Sample Period Sample Size Information Source % of transaction from the second-hand market Distribution of Market Rime Mean Median Std. Dev. Mox. Min.

Leung et al. (2002) Jan. 1993-Dec. 1999 11,612 EPRC 100%

This Paper Jan. 2003-June 2006 4,010 Midland Realty 96.3% ^i

301.67 81 460.73 3,002 1

132.50

78

151.27
1,141

1
1
1

End notes
' Being a listed company in Hong Kong, as at January 31, 2007, Midland had 426 branches in Hong Kong, mainland China, and Macau. Its transactions cover all real properties across Hong Kong amounting to over one-third of the market share in terms of transaction value.
JRER Vol. 3 4 N o . 3 - 2 0 1 2

3 9 6

H u i ,

W o n g ,

a n d

W o n g

The conditions included in market value definitions of the Uniform Standards of Professional Appraisal Practice (USPAP) in the U.S. generally fall into three categories. One of them is: the conditions of sale (e.g., exposure in a competitive market for a reasonable time prior to sale). (Definitions: USPAP 2005.) Secondly, one of the conditions included in open market value definitions of the Royal Institution of Chartered Surveyors (RICS) Red Book 5* edition in the U.K. assumes that, prior to the date of valuation there had been a reasonable period (having regard to the nature of the property and the state of the market).

References Anglin, P.M. A Summary of Some Data on Buying and Selling Houses. Working paper. University of Windsor, Canada, 1994. Anglin, P.M., R. Rutherford, and T.M. Springer. The Trade-off between the Selling Price of Residential Properties and the Time-on-the-Market: The Impact of Price Setting. Journal of Real Estate Finance and Economics, 2003, 26:1, 95-111. Anglin, P.M. and R. Wiebe. Pricing in an Illiquid Real Estate Market. 2004. http://www. uwindsor.ca/Palutnglin. Asabere, P.K. and RE. Huffman. Price Concessions, Time on Market, and the Actual Sale Price of Homes. Journal of Real Estate Finance and Economics, 1993, 6:2, 167-74. Asabere, P.K., E.E. Huffman, and S. Mehdian. Mispricing and Optimal Time-on-Market. Journal of Real Estate Research, 1993, 8:1, 149-56. Belkin, J., D.J. Hempel, and D.W. McLeavey. An Empirical Study of Time on Market Using Multidimensional Segmentation of Housing Markets. Journal of the American Real Estate and Urban Economics Association, 1976, 4:2, 57-75. Cox, D.R. Regression Models and Life-tables (with discussion). Journal of Royal Statistical Society, 1972, B 34, 187-220. Cubbin, J.S. Price, Quality and Selling Time in the Housing Market. Applied Economics, 1974, 6:3, 171-87. Genesove, D. and C.T. Mayer. Equity and Time to Sale in the Real Estate Market. American Economic Review, 1997, 87:3, 255-69. Glower, M., D.R. Haurin, and PH. Hendershott. Selling Price and Selling Time: The Impact of Seller Motivation. Real Estate Economics, 1998, 26:4, 719-40. Haurin, D.R. The Duration of Marketing Time of Residential Housing. Journal of the American Real Estate and Urban Economics Association, 1988, 16:4, 396-410. Horowttz, J.L. The Role of the List Price in Housing Markets: Theory and an Econometric Model. Journal of Applied Economics, 1992, 7:2, 115-29. Hui, CM., C.K. Chau, L. Pun, and M.Y. Law. Measuring the Neighboring and Environmental Effects on Residential Property Value: Using Spatial Weighting Matrix. Building and Environment, 2007, 42, 2333-43. Jud, G.D., D.T. Winkler, and G.E. Kissling. Price Spreads and Residential Housing Market Liquidity. Journal of Real Estate Finance and Economics, 1995, 11:3, 251-60. Jud, G.D., T.G. Seaks, and D.T. Winkler. Time-on-Market: the Impact of Residential Brokerage. Journal of Real Estate Research, 1996, 12:3, 447-58.

M a r k e t i n g

T i m e

a n d

P r i c i n g

S t r a t e g i e s

3 9 7

Kalra, R. and K.C. Chan. Censored Sample Bias, Macroeconomic Factors and Time on Market of Residential Housing. Journal of Real Estate Research, 1994, 9:2, 253-62. Kang, H.B. and M.J. Gardner. Selling Price and Marketing Time in the Residential Real Estate Market, Journal of Real Estate Research, 1989, 4:1, 21-35. Knight, J.R. Listing Price, Time on Market, and Ultimate Selling Price: Causes and Effects of Listing Price Changes. Real Estate Economics, 2002, 30:2, 213-37. Larsen, J.E. and W.J. Park. Nonuniform Percentage Brokerage Commissions and Real Estate Market Performance. Real Estate Economics, 1989, 17:4, 422-38. Leung, C.K.Y., Y.C.F. Leong, and I.Y.S. Chan. TOM: Why Isn't Price Enough? International Real Estate Review, 2002, 5:1, 91-115. Li, W.F. The Impact of Pricing on Time-on-Market in High-rise Multiple-unit Residential Developments. Pacific Rim Property Research Journal, 2004, 10:3, 305-27. Mayer, C.J. A Model of Negotiated Sales Applied to Real Estate Auctions. Journal of Urban Economics, 1995, 38, 1-22. Merlo, A. and F. Ortalo-Magne. Bargaining over Residential Real Estate: Evidence from England. Journal of Urban Economics, 2004, 56:2, 192-216. Miller, N.G. Time on Market and Selling Prices. Journal of the American Real Estate and Urban Economics Association, 1978, 6:2, 164-74. Ong, S.E. and Y.C. Koh. Time-on-Market and Price Trade-offs in High-rise Housing Submarkets. Urban Studies, 2000, 37:11, 2057-71. Rating and Valuation Department (various years). The Property Review (various issues). Rating and Valuation Department, the Government of Hong Kong Special Administrative Region, PR China. Taylor, C. Time on the Market as a Sign of Quality. Review of Economic ^iwdies, 1994, 66:3, 555-78. Trippi, R.R. Estimating the Relationship between Price and Time to Sale for Investment Property. Management Science, 1977, 23:8, 838-42. Turnbull, G.K. and C.F. Sirmans. Information, Search, and House Prices. Regional Science and Urban Economics, 1993, 23:4, 545-57. Wong, J.T.Y., CM. Hui, W. Seabrooke, and J. Raftery. A Study of the Hong Kong Property Market: Housing Price Expectations. Construction Management and Economics, 2005, 23, 757-795. Wong, J.T.Y. and CM. Hui. The Myth of Property Prices: On the Psychology of Sellers and Buyers. Property Management, 2008, 26:3, 171-90. Yang, S.X. and A. Yavas. Bigger is not Better: Brokerage and Time-on-Market. Journal of Real Estate Research, 1995a, 10:1, 23-33. Yavas, A. and S.X. Yang. The Strategic Role of Listing Price in Marketing Real Estate: Theory and Evidence. Real Estate Economics, 1995, 23:3, 347-68. Zuehlke, T.W. Duration Dependence in the Housing Market. Review of Economics and Statistics, 1987, 69:4, 701-04.

This project was funded by the Hong Kong Polytechnic University (Project Numbers: 8-ZZlZ and G-YH86). The research team wishes to thank Midland Realty for their

J R E R

V o l .

3 4

N o .

3 - 2 0

3 9 8

H u i ,

W o n g ,

a n d

W o n g

generous supply of the transaction data for this study and their insighiful comments on the performance of the property market in Hong Kong. The team also thanks Dr Hankel Fung and Mr. Ka-hung Yu for their assistance.

Eddie CM. Hui, Hong Kong Folytechnic University, Hung Hom, Kowloon, Hong Kong or bscmhui @ inet.polyu. edu. hk. Joe T.Y. Wong, Hong Kong Folytechnic University, Kowloon, Hong Kong, China or bstywong @ inet.polyu. edu. hk. K.T. Wong, Hong Kong Folytechnic University, Kowloon, Hong Kong, China or Hong Kong Folytechnic University, Kowloon, Hong Kong, China or ktwong(ieee.org.

Copyright of Journal of Real Estate Research is the property of American Real Estate Society and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.