THE CONTAGION EFFECT OF INFILL DEVELOPMENTS ON LOCAL HOUSING PRICES

Transcription

1 IRES IRES Working Paper Series THE CONTAGION EFFECT OF INFILL DEVELOPMENTS ON LOCAL HOUSING PRICES Joseph T.L. Ooi Thao T.T. Le October 3, 2011

2 THE CONTAGION EFFECT OF INFILL DEVELOPMENTS ON LOCAL HOUSING PRICES Joseph T.L.Ooi National University of Singapore 4 Architecture Drive Singapore rstooitl@nus.edu.sg Thao T.T. Le National University of Singapore 4 Architecture Drive Singapore rstlttt@nus.edu.sg October 3, 2011 Abstract This paper examines the impact of new infill developments, which involve developing vacant or under-used parcels within existing urban areas that are largely developed, on local housing prices. The evidence shows that they have a positive and persistent impact on house prices at the neighborhood level. Timing-wise, we observe that the contagion effect is highest immediately after the new housing development is marketed. A large proportion of the contagion effect can be traced to the overpricing of new homes by developers. Overall, the results indicate that developers act as price leaders and contribute significantly to price discovery in the local housing market. Key words: infill developments, contagion effect, housing. Acknowledgements: This research received the Singapore Ministry of Education s Academic Research Fund (AcRF) Tier 1 funding support: WBS No. R We are grateful to the participants of the 2011 American Real Estate and Urban Economics (AREUEA) mid-year meeting their valuable comments. # Corresponding author. We thank Brent Ambrose, Timothy Riddiough, Eric Rosenblatt, Brent Smith and participants at the American Real Estate and Urban Economics (AREUEA) 2011 mid-year meeting in Washington DC for their valuable comments. The authors received financial assistance from the National University of Singapore for this research project.

3 THE CONTAGION EFFECT OF INFILL DEVELOPMENTS ON LOCAL HOUSING PRICES 1. Introduction Modern cities have grown predominantly by spreading out from the center with new lowdensity homes built at the urban fringe. Common explanations for this pattern of city growth include rising household incomes, lower commuting costs and cheaper land costs in the suburbs. Such decentralization process, nevertheless, gives rise to common problems associated with urban sprawl such as traffic congestion, increased infrastructure costs, and loss of rural and resource lands. Recognizing that such growth is not sustainable, some communities have adopted policies restricting the amount of land in the suburbs that can be used for development and encourage infill developments to accommodate new growth. Infill developments involve developing vacant or under-used parcels within existing urban areas that are largely developed. They include redevelopment opportunities where new and more expensive buildings are constructed in place of old buildings or vacant sites. In addition to curbing urban sprawl, such infill developments also increase a city s tax base through the higher priced new homes. Hence, they tend to be favored by local municipalities. 1 However, an infill development imposes externality effects on nearby properties and thereby affects their values. It is not uncommon for existing residents to resist new developments within their neighborhood for reasons such as visual pollution, increased traffic noise, disruption to local traffic patterns, or loss of a neighborhood s character (Dye and McMillen, 2007). Low-income housing projects also have a negative impact on neighboring property values (Funderburg and MacDonald, 2010). The primary channel that a new development may affect prices of other homes is through the amenity effect. Ellen et al. (2001) argue that a new development refreshes the neighborhood by removing an abandoned boarded-up building or a littered vacant lot, which are 1 The Municipal Research and Services Center (MRSC) of Washington has a good compilation of resources and planning tools for infill developments on its website: The Miami-Dade County in Florida also took a proactive stance by designating infill target areas and establishing programs that provide incentives to encourage redevelopment of vacant, dilapidated or abandoned properties in urban neighborhoods. 1

4 not only visually unappealing but invite unwelcome activities like vandalism and crime. It is therefore not surprising that studies on spillover effects of new developments are conducted primarily in the context of urban regeneration. Several studies have also examined property price appreciation accruing to the spillover effect of developments under the Tax Increment Financing (TIF) scheme, which allows municipalities to designate an area for improvement and then earmark growth in property tax revenues resulting from appreciation to finance economic development within the district (Weber et al., 2003; Smith, 2006; Weber et al, 2007; Smith, 2009). Desalvo (1974) and Schwartz et al. (2006) document significant positive externalities associated with public and subsidized housing projects. Weber et al. (2007), in particular, find that TIFs increase the prices of nearby homes due to amenity improvements. In summary, a new investment could lead to either a positive or negative spillover effect depending on its overall impact on the neighborhood. However, the net effect is more likely to be positive than negative since the new building, by virtue that it is cleaner and more attractively designed, adds to the overall appeal of the neighborhood. This in turn attracts more desirable inhabitants, such as homeowners or higher income occupants, to live in the area. 2 Another channel through which a new development may affect local house prices is the supply mechanism. Since new housing investments generally leads to a net increase in the local housing stock, a valid question to ask is how the additional supply might affect home prices in the neighborhood. An increase in supply, all else being equal, will give rise to new competition and eventually lead to a price decline. Thus, in the case of infill housing developments, existing properties may suffer from increased competition and unfavorable comparison with the newer and better designed buildings (Grenadier, 1996; Newell, 2010). 3 The interaction of these two opposing amenity and supply mechanisms will ultimately dictate how a new infill development might affect local housing prices. In a neighborhood where the existing housing stocks are in relatively good condition, the spillover effect from a new investment may be marginal or even negative. 2 Recent literature on social capital and costs associated with different categories of occupants in a neighborhood argue that homeowners make better citizens (DisPasquale and Glaeser, 1999; Rosenthal, 2008). 3 In his classic analysis on development cascades and overbuilding in the real estate markets, Grenadier (1996) posits that a developer may respond to a decision to a rival s build decision by also choosing to build because the value of an old building diminished by the leader s decision to build a new building. 2

5 The context of our current study is in a metropolitan city where housing accommodation is primarily in multi-family high-rise buildings. Instead of low-density homes built at the urban fringe, most of the new private housing developments involve high-density infill developments. In total, we identified 275 new high-rise multi-family housing developments built by private developers in Singapore between 1999 and To determine whether proximity to a new infill development is capitalized into local housing prices, we estimate a hedonic model on a sample of 55,887 sale transactions in the secondary housing market. The regression model controls for structural and locational attributes of the individual apartments as well as time-varying effects over the sample period. A potential problem encountered in prior studies examining the spillover effects of low-income and subsidized housing programs is selection bias. Specifically, older neighborhoods with a lower-quality housing stock and lower home values may receive a higher disproportionate share of publicly financed housing projects (see Funderburg and MacDonald, 2010). To address this problem, we employ the difference-in-difference specification to estimate the hedonic models. This involves comparing the prices of properties within a 500-meter radius from a new housing investment with the prices of comparable properties that are located outside this ring but still in the same local district. The impact of the new infill development is estimated as the difference between the change in prices of houses near the new development before and after its introduction and the price appreciation experienced by similar houses outside the ring. We also track the capitalization effect at three critical stages of the development process, namely during the land acquisition, marketing and physical completion of the new project. In general, the whole process from the beginning to the end takes around five years. Figure 1 traces the major stages involved in the construction of a new residential project in Singapore. Like any real estate development projects, the process starts with land acquisition and ends with handingover of the finished apartments to the buyers. In between, a sequence of activities follows from designing and obtaining the necessary planning and building approvals to construction and finally completion of the project when it receives the Temporary Occupation Permit (TOP). Parallel to this, developers apply for a sale license to sell the units before their completion. The practice of selling new homes before their physical completion is a common practice adopted by developers in Asia (Lai, Wong and Zhou, 2004). On average, the time interval from the marketing launch of a new housing development to its physical completion is around three years. 3

6 To preview our results, we observe that infill developments have significant contagion effect on nearby apartments. Specifically, we find apartments in the immediate vicinity of a new infill development recorded higher prices in the range of 1.73% to 1.80% during the land preparation stage of the new development. Their prices increased by another 2.38% to 2.76% when the new development is launched and continue to rise during the construction phase. Prices only slow down after physical completion of the new development, which corresponds to the competitive effect of the new supply on existing homes but contrary to prediction that the spillover effects of a new development would take a longer time to unfold with local property prices only shifting upwards upon completion of the development (Schwartz et al., 2006; Weber et al., 2007). 4 The scale of the new infill development and whether it is built on a vacant or a teardown site also do not appear to have significant impact on local housing prices. In summary, the amenity channel plays a minimal role in our sample, which is not surprising since the existing housing stocks in our sample are young and in relatively good condition. We did not spot any significant difference in the price of houses located inside and outside the 500-meter ring from the infill development prior to the start of the development process. This indicates that on average the contagion effect is not due to developers systematically siting and timing their new projects to take advantage of ex-post price trend in a neighborhood. The evidence instead points to a price discovery process which begins with the inception of the new project, continues throughout the construction phase, and ends with the physical completion of the new development. Specifically, the benefits of an infill development appear to be capitalized soon after the developer signal his commitment to the neighborhood at the land acquisition stage. Interestingly, the steepest capitalization occurs at the marketing stage of the new development and a large proportion of the contagion effect can be traced to the benchmark prices set by the new developments. This corresponds with the price leadership role of property developers who generally have information advantages and greater control over the local housing supply. The individual home sellers, who do not enter the housing market frequently, infer their reservation prices from the prices set by the latest new housing 4 Schwartz et al. (2006) argue that the benefits of blight removal should be felt immediately, while other effects such as those related to occupancy may take longer to unfold. Moreover, during the construction period, negative externalities such as noise and visual pollution and traffic congestion will become more significant. Thus, local property prices may only shift upwards upon completion of the new development. 4

7 development in the neighborhood. 5 House prices in the locality continue to increase during the build-up phase of the new development, which is consistent with the developers practice of increasing the sale price gradually as more units are sold over time (Sirmans, Turnbull and Dombrow, 1997; Lai, Wang and Zhou, 2004). However, after the project is completed, developers may reduce prices for the unsold units due to additional carrying costs, which accounts for the slowdown in home prices in the neighborhood. The finding is also consistent with the notion that pre-sale developments are only viewed as competitors for tenants and owneroccupiers after their completion (see Ooi and Le, 2011). The remainder of this paper is organized as follows. Section 2 presents the methodology used to estimate the spillover effect of new infill housing developments. Section 3 describes the data. Section 4 discusses the estimation results and robustness checks. Section 5 concludes the paper with a summary of the key findings and implications. 2. Methodology The value of a property is determined by its attributes, which are often categorized into location, structural, and neighborhood factors (Rosen, 1974). In conventional hedonic price models, the neighborhood factor is capitalized into housing values through two types of spatial externalities: neighborhood effects due to shared neighborhood characteristics and amenity, such as accessibility, schools, shopping, crime; and spillover effects due to adjacency with the immediate neighboring properties (Can, 1990). Empirically, an estimate of the spillover effects can be obtained from a comparison of the change in prices for the nearby properties before and after the event relative to the rate of price changes for properties in the control group over the same period. However, a major challenge is finding a comparable neighborhood or property as a control sample to measure the differences in their property value, and thereby isolate the spillover effects. Prior studies on the impact of subsidized housing have generally employed basic hedonic regressions, but a limitation of the technique is the inability to establish whether subsidized sites are systematically located in weak/strong neighborhoods, or whether subsidized housing actually 5 See Markham (1951) and Ono (1982) for a sample of the classic literature on the theory of price leadership. 5

8 leads to the neighborhood decline/improvement. For example, older neighborhoods with a lowerquality housing stock and lower home values tend to receive a disproportionate share of publicly financed housing projects (Schwartz et al., 2006; Funderburg and MacDonald, 2010). To illustrate the problem, we begin with a simple model that provides a direct estimate of the effect of a new infill development: P it = α + θ L ijt + ε it (1) where P it is the sale price of existing unit i at time t; α is a permanent component of home price. The binary variable L ijt captures both the spatial and time dimensions. It has the value of unity if unit i is located within a 500-meter ring from a new development j and sold after the new development was launched. The coefficient θ provides a simple estimate of the effect of a new development on the sale prices of nearby properties. The underlying assumption of this model is that in the absence of the launch, θ would be 0; that is, there is no systematic difference in the mean prices of those properties with L ijt = 1 and properties with L ijt = 0. This condition can also be written as E 0 it L ijt. An unbiased estimate of θ can be obtained as: ˆ P P P (2) 1 0 where P 1 and P 0 denote the mean prices of properties with L ijt = 1 and properties with L ijt = 0 respectively. A key assumption of the model is that the new developments are randomly located. However, if developers are able to strategically choose strong areas to locate their projects, the prices of existing properties in these areas should be consistently higher than properties in other weak areas even in the absence of a new development. As a result, the price differences in these two areas might be attributed to some neighborhood characteristics unobserved or not measurable, such as crime rate or better air quality. We cannot reasonably rule out the possibility of such selection bias. A simple unbiased estimate of the new development effect can nevertheless be obtained from a difference-in-difference estimator. Essentially, we compare the prices of properties within a 500-m ring of a new development (in-ring properties) with the prices of comparable properties that are outside this ring (out-of-ring properties). We then compare the magnitude of this difference before and after the launch of the new development. This approach ensures that any 6

9 systematic difference in house prices related to the selection bias, if present, will not affect the estimated effect. 6 The difference-in-difference estimator can be derived from: P it = α + μt + γnear ij + θl ijt + ε it (3) where T = 1 if unit i is sold after the new development is launched; NEAR ij = 1 if unit i is located within a 500-m ring of a new development, and the definition of other variables are as before. Hence, γ reflects the baseline difference in sale price of homes located inside and outside of the 500-meter ring from a new housing development, while μ reflects the time trend. The true effect of a launch is again measured by θ which is a difference in differences (refer to Table 1), which can be represented as follows: ˆ P P ( P P ) ( P ) (4) 1 0 1,1 1,0 0,1 P0,0 where the subscripts of P denoting the value of T t and NEAR ij respectively. The first bracket represent the difference in sale price of properties inside and outside a 500-m ring of a new development after it is launched, while the second bracket is the price difference of the two groups before the new development is launched. The underlying assumption of this model is that the price trends of the two groups are similar in the absence of the event. This assumption is less strict than the underlying assumption of Equation (1). We can reasonably expect it to hold in reality, if changes in economic conditions affect all properties in similar ways. Finally, in order for ˆ to be the true estimated effect of a new development we need to ensure the properties being compared are similar in all aspects except for the presence of a nearby development. The heterogeneity of housing units can be controlled for by simply adding a vector of housing characteristics into Equation (3), as in a standard hedonic model. We thus expand our model and re-define the variables as follows: P idt = α + β i X i + δ d D d + ρm t + μt + γnear ij + θnpl ijt + λ LPOST ijt + ε it (5) where P idt is the log of the sales price of property i in district d and in quarter t. X i is a vector of property-related characteristics and locational attributes, which is represented by AREA (the unit s floor area), AREA_SQ (the unit s floor area squared), LEVEL (the level of the unit), 6 See Mayer (1995) for a discussion of the methodology. Prior studies have employed the difference-in-difference specification to examine the impact on property values of new railway stations (Gibbons and Machin, 2005), schools (Bogart and Cromwell, 2000) and new subsidized housing investment (Ellen et al., 2001; Schwartz et al., 2006). 7

10 LEVEL_SQ (the level of the unit squared), FH (a binary variable for units in development that has a freehold tenure), CONDO (a binary variable for condominium units); AGE (age of the unit), CBD (distance to CBD), and MRT (distance to the closest subway station). D d is a series of fixed effects to control for unobserved time-invariant features of each planning district in Singapore. M t is a time-variant component to capture house price movements in the overall market. The time dummy T in Equation (3) is expanded to a set of 58 dummy variables T t, reflecting the time of sale, to control for the time effect over our study period 1997:Q1 2011:Q2. ε it is a serially uncorrelated transitory component of prices. Unbiased estimates of the coefficients in Equation (5) can be obtained using OLS. The definition of NEAR ij remains the same, which reflects the baseline differences in sale price of homes located inside and outside of the 500-meter ring from a new housing development. The binary variable L ijt in Equation (3) is split into two terms, NPL ijt and LPOST ijt. The first term NPL ijt has the value of one if a new infill development was launched within a 500- meter ring in the past three months; thus, the coefficient for NPL ijt provides the simplest estimate of the immediate impact of a new housing development on nearby houses. If the new development dampens the prices of nearby houses, the θ coefficient would be negative and statistically significant. However, if the new project enhances the prices of nearby houses, the θ coefficient would be positive and statistically significant. In addition, we include LPOST ijt to monitor the persistency and dynamics of the spillover effect after the new development is launched. It is an interactive variable of NEAR ij with N ij, which is a continuous variable indicating the number of quarters between the date of sale in the ring and the new project launch date. If after the launch, house prices in the ring continue to rise (or decline) relative to house prices outside the ring, the coefficient λ will be positive (negative). Finally, we add two more interacting variables in to control for the land acquisition stage and completion of construction. They represent two important milestones marking the start and end of the new development project. LAND ijt is a binary variable to indicate if the sale in the ring occurs within 8 quarters preceding a nearby project launch; that is, if the in-ring sale occurs after the land is acquired but before the project is marketed. 7 LPOST ijt is also interacted with TOP ijt, 7 We were unable to obtain the actual dates of site acquisition and preparation for a number of the sites. As a proxy, we use the average gestation period between site acquisitions/preparation and marketing of the new project, which is eight quarters (see Figure 1). 8

11 which is a dummy for a sale transaction made after the physical completion of a nearby new development, to examine whether nearby properties respond to the physical completion of the infill development. If the amenity effect associated with the completion of an infill development, such as population moving into the neighborhood, is capitalized into house prices, we should see a significant positive coefficient for LPOST*TOP. To summarize, our basic estimation model is represented as follows: P idt = α + β i X i + δ d D d + ρm t + μt + γnear ij + ηland ijt + θnpl ijt + λ LPOST ijt + φlpost ijt *TOP ijt + ε it (6) 3. Data Before discussing the empirical results, it is useful to provide a quick review and description of the data. Most of the information is extracted from REALIS, a database maintained by the government agency responsible for urban planning and development control in Singapore. Since both the amenity and supply mechanisms require the new investment to be large enough to create significant externalities, we only include infill developments that have more than 100 apartment units in the study sample. In total, 275 new residential developments were launched between 1999:Q1 and 2010:Q1. Representing new housing investments over the sample period, they are scattered throughout the main residential areas in Singapore. The new developments are further categorized according to the height of the tallest building in the new development relative to the height of the existing building (HEIGHT), the project size in terms of number of units (SUPPLY) and whether it is built on a teardown site (REDEV). 8 The average development in our sample has 298 new dwelling units. 63.3% of the new projects are built on teardown sites and the average height of the new structure is just below 19 storeys. Our empirical analysis involves 55,887 resale transactions between 1997 and 2011 in the city-state of Singapore. For each transaction, we collect the following information: sale price, date of sale, floor area, floor level, age, type of development, and land tenure. Spatial distance to 8 The Singapore land market is a dual market. While there is privately owned vacant land available for development, the government controls a large part of the supply of developable vacant land in the economy (see Ooi, Sirmans and Turnbull, 2011). 9

12 the CBD and to the closest subway station is also measured. 9 Hedonic regressions on the sales transactions are then estimated to determine whether the launch of a new development nearby is capitalized into house prices in addition to the usual structural, location, neighborhood and market-wide factors. The dependent variable in the hedonic price model is the log of transaction price, which is the agreed transaction price of the property between the buyer and the seller excluding stamp duties, legal and agency fees, and other professional fees. Table 2 contains the definitions and descriptive statistics of the variables in the basic empirical model. The average price of a residential unit in our sample is slightly over S$ 1 million. 10 The average unit has a floor area of m 2 and is located on the 8 th floor. The apartments in our transaction sample are fairly new with the average age falling just above 5 years old. Thus, potential spillover effect through the amenity channel is likely to be negligible. As Singapore has a small physical area, the greatest distance to the nearest subway station is only 6.1 km. Similarly, the furthest distance away from the CBD is only 21.4 km. 4. Empirical Results 4.1 Base model Table 3 reports the parameter estimates for the hedonic model with the difference-in-difference specification. Model (1), which is the base model, controls for time of sale and the unit s district 2 through a set of fixed effects. The equation is significant at the 5% level. The high R value of indicates that the factors in our hedonic model explain most of the variations observed in house prices. Most of the control variables also behaved as expected. House price is related positively with the unit s floor area and level; the influence of floor area on house price increases at a decreasing rate but the influence of floor level on house price increases at an increasing rate. 9 There is a large literature on the effect of a transportation node on surrounding properties with mixed results. McMillen and McDonald (2004), Gibbons and Machin (2005) and Amstrong and Rodriguez (2006) find a positive impact, whilst Dornbusch (1975), Forrest et al. (1996) and Gatzlaff and Smith (1993) find proximity to subway station has a negative effect on property values. 10 Monetary values are expressed in Singapore dollars throughout this paper. The exchange rate for US$1 is equivalent to S$1.22 in July

13 This indicates a preference for dwelling units located on the higher floors. The coefficient for freehold tenure (FH) is positive and statistically significant in all the three models. Specifically, apartments on freehold tenure fetched 10.9% more than equivalent apartments on leasehold tenure. A condominium unit, which comes with more landscaping and luxury facilities, is priced 17.5% higher than a private apartment unit. House age and distances to the central business district (CBD) are negatively correlated to sale price, indicating that older apartments which are located further away the central business district are sold at lower price. On the other hand, apartments located near to subway stations are sold at a discount, which suggests that the negative externality from pedestrian congestion, noise, or other externalities from a metro station is stronger than the countervailing positive externality of easy access to mass transportation (Munneke, et al., 2010). The performance of the residential homes at the broad market level (MARKET) also has a significant influence on prices in the neighborhood. Focusing on the spillover effects of new infill developments, the coefficients for LAND, NPL and LPOST are positive and significant. Collectively, they indicate that new infill developments have a significant contagion effect on nearby apartments. The coefficient for NPL shows that, all else equal, existing apartments fetched a higher price if a new infill housing development was launched within a 500-meter radius in the previous quarter. Judging from the magnitude of the coefficient, the immediate wealth effect of a new property launch in the neighborhood is +2.38%. In other words, the average apartment in our sample, which is priced at S$1.06 million, would increase by S$25,228 following the launch of a new infill development nearby. The positive coefficient for LAND indicates that the capitalization effect starts earlier, probably around the time when the developer shows his commitment by buying the land or submitting the planning application. The positive coefficient LPOST, which measures for the capitalization effect from the time the project launch until its physical completion, suggests prices of nearby apartments continue to increase at the rate of 0.56% every quarter. On the basis that the average infill development takes three years to complete, the total increase over the construction period would be in the region of 6.72%. Thereafter, prices of nearby apartments continue their upward drift but at a slower pace due to the dampening effect of competition when the newly completed houses compete with existing properties for tenants and owner occupiers. Meanwhile, the coefficient for NEAR is insignificant, both statistically and economically, 11

14 implying that prices of apartments located inside and outside the 500-meter radius were more or less the same prior to the new infill development. Overall, the results suggest that a new infill development has a positive and persistent spillover effect on local house prices. Its contagion effect on local house prices is strongest at the point in time when the new property is launched. However, this effect is dampened by the competition effect when the new housing stocks are physically completed. To illustrate, Figure 2 charts the contagion effect of an average new infill development on local housing prices over different stages of its development process. In total, a new infill development induces a 10.82% price increase of surrounding properties, and the effect appears steepest at the marketing stage of the new project. The contagion effect is persistent but becomes insignificant following the completion of the new development. 4.2 The Amenity and Supply Channels Prior literature has identified two possible channels through which a new development can affect the value of nearby properties. First, the supply channel argues that increased competition from the new housing supply creates a downward pressure on local housing values. 11 Second, the amenity channel could lead to either a positive or a negative price effect depending on its overall impact on the neighborhood. However, as the newer building are typically of better quality, their impact through the amenity channel is more likely to be positive than negative. To further investigate the role of the amenity and supply mechanisms, we examine how the characteristics of the new development may impact the spillover effects. Specifically, we analyze if the observed positive spillover effects are influenced by three characteristics of the new development, namely the nature of the development site, the intensity of the new building, and the scale of the development. First, a new development could be built on an existing plot of vacant land or by tearing down and redeveloping an existing building. The former involves constructing a new structure on a previously vacant site, whilst the latter involves a new structure replacing an old structure. Their impact on surrounding properties may differ. Particularly, an 11 Harding et al. (2009) used the same logic to explain why foreclosed properties, by increasing the supply of homes in the neighborhood, place a downward pressure on local property prices. 12

15 infill development built on a vacant site may lead to the loss of a green field, existing infrastructure being stretched, and more congestion in the neighborhood. Conversely, an infill development involving tearing down of an old building and replacing it with a new one is more likely to improve amenities in the neighborhood. Second, we control for the height of the new building vis-à-vis the height of the existing buildings in the surrounding area. Whilst a taller building may impede the view of existing buildings in the vicinity, it may also be indicative that a higher floor area ratio is permitted in the area; thus, increasing property values in the neighborhood. Third, we examine whether the spillover effect is dictated by the scale of the new development. Intuitively, larger projects should have a bigger impact. We measure scale by the number of dwelling units in the new infill development. This variable also serves as a control for the supply channel in terms of the number of new units added to the local market. If the supply effect dominates the amenity effect, a larger project will add more supply and thereby create a stronger downward pressure on local property prices. In Model (2), NPL is interacted with three characteristics of the new housing development, namely redevelopment (REDEV), number of units (SUPPLY) and relative height (HEIGHT). The coefficients of the individual interaction variables will provide a direct estimate of how that particular attribute of the new development may affect the prices of nearby properties. Sale transactions that are located near to two new infill developments with opposing characteristics are dropped from the sample. The estimation results, as reported in the last column of Table 3, are robust. In particular, the coefficients for LAND, NPL and LPOST remain positive and statistically significant. However, none of three new interaction terms are statistically significant. With respect to the characteristics of the new development, we observe that the coefficient for NPL*REDEV is positive as expected but it is not statistically significant. This suggests that that the spillover effect of a new development is not dictated by whether it is built on a vacant piece of land or a teardown site. We also find a positive but statistically insignificant coefficient for NPL*HEIGHT. The coefficient for NPL*SUPPLY is not statistically significant, albeit negative as theory predicts, further suggesting that the size of the project has no effect on the spillover effect. This corresponds with our prior results that the competition and amenity channels do not play significant role during the marketing stage of the project. 13

16 4.3 The Information Channel As noted earlier, the spillover effect of a new infill development is strongest at the point in time when the new project is marketed. In this section, we explore the possibility of new information being revealed and capitalized at this stage of the of the development cycle. Specifically, we study the relationship between the contagion effect and the pricing strategy adopted by the developer. We first benchmark the price level of units in the new development relative to the prices of resale houses in the vicinity by deriving the degree of mispricing for each unit sold in the new development. 12 We do this by running a hedonic regression on all existing units within the same area as the new development and computing the degree of overpricing for a unit as the difference between the fitted values derived from the hedonic model and their actual selling prices. We then take the average (and median) for all the units sold in the same development to represent the degree of overpricing for a specific development. All else being equal, we find that new units are sold at a higher price than existing units in the neighborhood with the average (median) price premium set by the developers is 4.8% (3.1%). The developer s ability to set new benchmark prices could be attributed to either dominance over the local housing supply or ability to pick winning sites (by predicting ex-ante of a latent demand for housing in a specific location). The former is consistent with a dominant oligopolistic price leadership, while the latter is consistent with a barometric price leadership A simple and quick way is to find the ratio of the average price of new units sold in the development over the average price of the existing units sold in the vicinity around the same period. Using this method, we find that the new residential projects are, on average, priced at 31.5% higher than nearby properties. This approach, however, does not control for the structural attributes, such as tenure, size, age, and floor level of the individual units. Clapp, Dolde and Tirtiroglu (1995) employed a similar approach to measure excess return of housing in a locality compared with the average metropolitan area. 13 The classical literature on price leadership focuses on an oligopolistic industry structure with a dominant firm acting as the price setter, and many smaller firms, which own their own cannot produce enough output to influence price. The alternative model of price leadership is barometric leadership where the leader serves as a barometer of current market conditions for the other firms in the industry (Stigler, 1947; Markam, 1951). As modeled by Cooper, (1996), barometric price leadership is due to informational setting. Specifically, the well-informed firm will become a leader in an environment of asymmetric information. A number of studies have considered the ability of developers to influence property prices within a submarket (see Somerville, 1999; Wang and Zhou, 2006; Ott, Hughen and Read, 2011). 14

17 To examine how the price setting behavior of developers influence the reaction of house prices in the neighborhood, we interact the new variable, PRICING, with NPL and incorporate it as an additional explanatory variable in Model (3). Its coefficient provides a direct measure of the economic impact of the degree of overpricing in a new development on the contagion effect. Intuitively, we expect it to have a positive impact on local property prices. The regression results, which are reported in Table 4, are revealing. While the results for the other variables are robust to the inclusion of the new interaction variable, we observe that the economic impact of a new property launch is now weaker, both statistically and economically. Specifically, the coefficient for NPL lost its statistical significance, even at the 0.10 level. The magnitude of the contagion effect is also lower at 0.99%, as compared to 2.76% previously. This implies that new developments that are launched at similar prices to existing units in the neighborhood do not have any significant spillover effect on nearby property values. Interestingly, the coefficient for the new interaction variable is strongly significant, indicating that the pricing of the new development has a big impact on the spillover effect. Overall, the combined results indicate that the positive spillover effects on nearby properties following a new property launch is dictated to a large degree by the developer s pricing strategy. The spillover effects are larger if the new apartments are sold at a higher price. An average project in our sample, which is overpriced by 4.8%, leads to a 1.71% rise in prices of surrounding homes. This implies a price discovery process at work in the local housing market, whereby the better informed developer sets the pricing for his new apartments. This is subsequently reflected in the resale prices of existing apartments in the area when the less informed individual sellers follow the developer s pricing. 4.4 Additional Robustness Test Lastly, we examine whether the contagion effect is dictated by market timing. 14 Figure 3 tracks the aggregate number of new housing units launched quarterly between 1999 and The 14 Several theories have been proposed in the finance literature to explain the volume fluctuations in the IPO market. Lowry (2003) consolidates and formalizes a number of these models into three major hypotheses: the Capital Demand Hypothesis, the Information Asymmetry Hypothesis, and the Investor Sentiment Hypothesis. Buttimer, Hyland and Sanders (2005) examined their applicability to explain REIT IPO waves and pricing. 15

18 spread of the new units launched is obviously not random, suggesting that developers time their launches to take advantage of favorable conditions in the market. Employing the 75 percentile value as the cut-off, we define the quarters in which there are 2,952 or more new units launched as hot periods for launching a new project. 15 Out of our sample of 275 new infill developments, 108 projects (39.3%) were launched in a hot market, while only 44 (16%) projects were launched in a cold market, which is defined as periods where there are 1,433 or less units launched in the same quarter. The remaining 123 (44.7%) projects were launched in normal market situation. Not unexpected, we observe that developments which were launched in hot markets are on average overpriced by 8.57%, while those launched in cold market were only marginally underpriced by 0.04%. This suggests that while developers are quick to raise their selling prices in a hot market, they are less forthcoming in terms of taking the lead in reducing prices in a cold market. In Model (4), we interact HOT_MKT and COLD_MKT with NPL to differentiate the contagion effect of new property launches during a hot market and a cold market. The estimation results are reported in Table 4. The coefficient and statistical significance for all the variables are robust. Again, the coefficient for NPL is positive but statistically and economically insignificant. The coefficients for the new interactive variables, NPL*HOT_MKT and NPL*COLD_MKT are not statistically significant. This indicates that the new developments impact on nearby properties is not affected by market sentiment, even though the developers timing of the launches is not random. The coefficient for the overpricing variable remains statistically significant. 5. Conclusion In this paper, we seek to examine whether the wealth of existing homeowners actually decreases or increases when an infill housing development is built in the same neighborhood. The literature has traditionally focused on two possible channels on how a new development may affect local property values. The first channel is the supply effect, in which the new supply increases the 15 As an additional check, we also defined HOT_MKT as periods in which there are 1,873 or more (median value) new units launched. The results are robust and thus not reported in this paper for brevity reason. 16

19 area s housing stock and thereby depresses property values. The second channel is the amenity effect, in which the new development removes an eyesore and refreshes the neighborhood. This makes housing in the neighborhood more desirable and expensive. Thus, the new development would lead to price appreciation in the local market. The amenity channel, in particular, presumes the existence of a gap in the quality of the new and old housing stock. Since new housing investments are frequently channeled into stable neighborhoods where the existing housing stock is in a relatively good condition, it is not clear whether the new development would have similar impact on the nearby properties. In addition to these two conventional channels, we also explore the possibility of a third channel, namely the revelation and capitalization of new information through the developers pricing of the new development. The empirical results show that despite the diminished role of the amenity effect, new housing investment in a good neighborhood still has a positive spillover effect. Specifically, we find a sustained positive impact on the values of nearby homes. The contagion effect seems to be concentrated around the period when the new development is launched and our analysis reveals that the source of economic gains comes from the benchmark price established in the area by the new development. This corresponds with the price leadership role of property developers who generally have information advantages and greater control of supply in the local housing markets. They set prices of new units in the primary housing market, which then spillovers to existing units through the resale housing market with the strength of the spillover dependent on the benchmark price set by the developer. The results have important policy implications on the wealth effects of infill developments, which commonly face neighborhood oppositions. Our study shows that new infill housing developments generate a positive impact on local property values, and hence the wealth of the existing homeowners. 17

20 Bibliography Armstrong, R. J., Rodriguez, D. A An Evaluation of the Accessibility Benefits of Commuter Rail in Eastern Massachusetts Using Spatial Hedonic Price Functions. Transportation 33, Bogart, W.T., Cromwell, B.A How much is a neighborhood school worth? Journal of Urban Economics 47, Buttimer, R.J., Hyland, D.C. and Sanders, A REITs, IPO waves and long-run performance. Real Estate Economics 33(1), Can, A Measurement of neighborhood dynamics in urban house prices. Economic Geography 66, Cooper, D.J Barometric price leadership. International Journal of Industrial Organization 15, Clapp, J.M., W. Dolde, D. Tirtiroglu Imperfect information and investor inferences from housing price dynamics. Real Estate Economics 23(3), DiPasquale, D., Glaeser, E Incentives and social capital: Are homeowners better citizens? Journal of Urban Economics 45, Desalvo, J. S Neighborhood upgrading effects of middle-income housing projects in New York city. Journal of Urban Economics 1, Dornbusch, D.M BART-Induced Changes in Property Values and Rents, in Land Use and Urban Development Projects: Phase I, BART Impact Study, U.S. Department of Housing and Urban Development and U.S. Department of Transportation. Dye, R.F., McMillen, D.P Teardowns and land values in the Chicago metropolitan area. Journal of Urban Economics 61, Ellen, I., Schill, M., Susin, S., Schwartz, A Building homes. Reviving neighborhoods: spillovers from subsidized construction of owner-occupied housing in New York City. Journal of Housing Research, 12 (2), Forrest, D., Glen, J., Ward, R The impact of a light rail system on the structure of house prices: a hedonic longitudinal study. Journal of Transport Economics and Policy 30(1), Funderburg, R., MacDonald, H Neighborhood valuation effects from new construction of lowincome housing tax credit projects in Iowa: A natural experiment. Urban Studies 47(8), Gatzlaff, D. H., Smith, M. T The impact of the Miami Metrorail on the value of residences near station locations. Land Economics 69, Gibbons, S., Machin, S Valuing rail access using transport innovations. Journal of Urban Economics 57,

21 Grenadier, S.R The strategic exercise of options: development cascades and overbuilding in real estate markets. The Journal of Finance 51(5), Harding, J.P., Rosenblatt, E., Yao, V.W The contagion effect of foreclosed properties. Journal of Urban Economics 66, Lai, R.N., Wang, K. Zhou, Y Sale before completing of development: pricing and strategy. Real Estate Economics 32(2), Lowry, M Why does IPO volume fluctuate so much? Journal of Financial Economics 67(1), Markham, J.W., The nature and significance of price leadership. American Economic Review 41(5), Mayer, B.D Natural and quasi-experiments in economics. Journal of Business and Economic Statistics, 13(2), McMillen, D. P., McDonald, J Reaction of House Prices to a New Rapid Transit Line: Chicago s Midway Line, Real Estate Economics 32(3), Munneke, H.M., Ooi, J.T.L., C.F. Sirmans, Turnbull, G Sequential Sales of Similar Asset: The Law of One Price and Real Estate. Journal of Regional Science (forthcoming). Newell, T Development and neighborhood revitalization: The effects of residential investment on property values in Durham, NV. The Michigan Journal of Business 3(2), Ono, Y Price leadership: A theoretical analysis. Economica 49, Ooi, J.T.L., Sirmans, C.F., Turnbull, G Government Supply of Land in a Dual Market. Real Estate Economics 39(1), Ooi, J.T.L and Le, T.T.T New Supply and Price Dynamics in the Singapore Housing Market. Urban Studies (forthcoming). Ott, S.H., Hughen, W.K. and Read, D.C Optimal phasing and inventory decisions for large-scale residential development projects. Journal of Real Estate Finance and Economics (forthcoming). Rosen. S Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy 82(1), Rosenthal, S.S Old homes, externalities, and poor neighborhoods. A model of urban decline and renewal. Journal of Urban Economics 63, Schwartz, A.E., Ellen, I.G., Voicu, I., Schill, M.H The external effects of placed-based subsidized housing. Regional Science and Urban Economics 36(6), Sirmans, C. F., Turnbull, G.K., and Dombrow, J Residential Development, Risk, and Land Prices. Journal of Regional Science 37(4), Smith, B.C The Impact of Tax Increment Financing Districts on Localized Real Estate: Evidence from Chicago s Multifamily Markets. Journal of Housing Economic 15,