335

[ Journal of Law and Economics, vol. 61 (May 2018)] © 2018 by The University of Chicago. All rights reserved. 0022-2186/2018/6102-0012$10.00

Property Rights Restrictions and Housing Prices

Kwan Ok Lee National University of Singapore Joseph T. L. Ooi National University of Singapore

Abstract

Using a natural experiment in Singapore, we examine the economic impact of temporarily restricting owners’ rights to transfer their property. Executive con- dominiums (ECs), introduced to provide affordable housing for middle-class citizens, are subject to restrictions on transferability in the first 10 years, un- like private condominiums (PCs). As per the option theory, EC buyers have the forward-start American put option with the right to sell their properties only after the contract date. Among transacted units matched by location, comple- tion and transaction dates, and complex- and unit-level characteristics, we find that prices of new ECs are about 21 percent lower than those of PCs. After the 10th year, when property rights restrictions are completely removed, the price gap between ECs and PCs narrows to about 3 percent. These results suggest that property rights restrictions and illiquidity generated by the forward-start Amer- ican put option for 10 years results in an 18 percent discount.

1. Introduction

In the context of real estate ownership, private-property rights are often viewed as a bundle of sticks, with each stick representing a right or stream of benefits to the owner. The bundle typically includes the exclusive right to use the asset, the exclusive right to the fruits of the asset, and the freedom to transfer the asset to others (Segal and Whinston 2010). Thus, a property owner has unencumbered rights to occupy, rent, and sell the property. In this study, we examine the eco- nomic effect of restricting property rights to sell or rent a property for a predeter- mined period. We take advantage of a natural experiment that uses the executive

We thank Sam Peltzman, an anonymous reviewer, Geoffrey Turnbull, Wayne Archer, David Ling, Qiang Li, Fan Yi, Tien-Foo Sing, Masaki Mori, participants at the 2018 Allied Social Science Associations– American Real Estate and Urban Economics Association annual conference and the 2017 Association for Public Policy Analysis and Management Fall Research Conference, and sem- inar participants at the University of Hong Kong for helpful comments. We also thank Edward Ti, Yoyie He, and Serene Tay for research assistance, particularly for data collection. Financial support from the National University of Singapore is gratefully acknowledged.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

336 The Journal of LAW & ECONOMICS

condominium (EC) scheme introduced in 1995 to meet the housing aspirations of “sandwiched” middle-class citizens in Singapore.1 The novelty of this scheme is that newly built EC units are sold with a set of temporary restrictions on private- property rights. New EC units are subject to a 5-year minimum occupation pe- riod (MOP): the original buyers cannot rent or sell their condominium units within the first 5 years. They can rent or sell their units after the fifth year, but only to Singaporean citizens or permanent residents (PRs). When EC develop- ments become fully privatized after the 10th year, they attain status equivalent to that of a private condominium (PC) and can be sold to anyone, including foreign buyers.

The EC scheme provides a unique opportunity to examine the impact of re- stricting property rights on housing prices. Our empirical investigation aims to examine the economic value associated with these restrictions by comparing the price of an EC unit against its predicted price if it were sold without the encum- brances. To do so, we identify PCs that are located within a 2-kilometer radius of each EC development and are similar to ECs in terms of land tenure and ame- nities, except for property rights restrictions, and match the ECs with them by complex- and unit-level characteristics, transaction dates, and administrative planning areas (neighborhoods). We also take advantage of the partial relaxation of restrictions after the fifth year and the full relaxation after the 10th year of the EC’s physical completion. Controlling for price movement in the market and in- flation, therefore, we find that the year-by-year comparisons of transacted prices of ECs with those of matched PCs allow us to identify causal counterfactuals of property rights restrictions. After these restrictions are lifted, we would expect the price gap between ECs and their counterpart PCs to converge over time.

We find that the selling prices of new EC units are about 21 percent lower than those of comparable PC units. This discount reflects the upper bound of restrict- ing the owners from renting and selling their units for 10 years. In addition, the sale price of EC units narrows to about 8 percent of the price of equivalent PC units after the fifth year, when buyers of ECs are no longer subject to the restric- tion of selling to Singaporeans but are still prohibited from selling to foreigners. The significant price discount implies that this is still a binding constraint. Fi- nally, we observe that the price gap between the EC and PC units narrows over the next 5 years and eventually becomes about 3 percent after the ECs cross the 10-year milestone, when the remaining restrictions on the transferability of prop- erty rights are removed.

This paper directly contributes to the literature on property rights. The classical literature on property rights has generally focused on their role in transitioning a society toward economic growth and market efficiency (see Coase 1960; Dem- setz 1967; Libecap 1989; North 1990; Mahoney 2005). In the context of real estate ownership, most studies have focused on the effects of restricting owners’ rights

1 When the executive condominium (EC) scheme was first announced in August 1995, the term “sandwiched class” was used to refer to potential home buyers who earned marginally more than S$8,000, which was then the monthly income ceiling for buyers of public flats.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 337

to use a property (McMillen and McDonald 1993; Cannaday 1994; Hughes and Turnbull 1996; Munneke and Slawson 1999; Netusil 2005; Sirmans et al. 2006; Rogers 2006; Lin, Allen, and Carter 2013; Meltzer and Cheung 2014). Ours is the first study that quantifies the economic impact of placing temporary restrictions on owners’ rights to rent and sell their residential properties.

Our results are also useful in understanding the role of illiquidity in asset pric- ing. The current finance literature has mostly relied on data from the financial markets in the context of stock-trading illiquidity and assessed the impact of such illiquidity on asset pricing (for example, Bailey and Jagtiani 1994). Our study contributes to the literature by directly estimating the negative impact of com- plete, temporary illiquidity on the asset pricing of residential properties.

Finally, unlike most previous studies on place-based housing programs using supply-side subsidies, which focus on noneconomic outcomes (Newman and Schnare 1997; Rohe and Freeman 2001; Ellen and Horn 2013; Talen and Kos- chinsky 2014) or their external effects on surrounding communities (Cummings, DiPasquale, and Kahn 2002; Schwartz et al. 2006), we evaluate the effectiveness of using property rights restrictions that are direct economic benefits provided to EC purchasers.

2. Institutional and Scholarly Backgrounds

2.1. Previous Research on Property Rights in Real Estate

Some of the best known examples of property rights restrictions are land use restrictions such as zoning, growth-management laws, minimum lot sizes, and density restrictions. Many studies examine the impact of these restrictions on the prices of properties subject to the restrictions and nearby properties not subject to them (Netusil 2005; Quigley and Rosenthal 2005; Ihlanfeldt 2007; Michael and Palmquist 2010; Munneke et al. 2013). Some studies also investigate the economic impact of restricting other rights to use property, including age restrictions for occupants (Allen 1997; Guntermann and Moon 2002; Guntermann and Thomas 2004; Lin, Liu, and Yao 2010), restrictions on pets in residential dwellings (Cann- aday 1994; Lin, Allen, and Carter 2013), and other covenants and regulations im- posed by private homeowners’ associations (Hughes and Turnbull 1996; Rogers 2006; Meltzer and Cheung 2014). These studies generally conclude that property rights restrictions have a positive impact on housing prices by reducing potential negative externalities and improving the predictability of future characteristics of neighborhoods or by inducing supply constraints. While these studies provide useful insights on the effects of restricting owners’ rights to use property on prop- erty values, none have investigated the economic impact of restricting the own- er’s right to rent and sell a property.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

338 The Journal of LAW & ECONOMICS

2.2. Institutional Framework

The EC scheme was introduced in 1995 as a new housing solution for the sand- wiched class of citizens in Singapore.2 Riding on citizens’ growing aspirations to upgrade from public housing to private residential properties, the EC scheme was envisioned as a form of private housing to be developed, priced, and sold by pri- vate developers. Designed and built to imitate PCs, ECs feature gated commu- nities and common facilities such as swimming pools, gyms, tennis courts, and clubhouses. Similar to PCs, ECs are sold with strata titles, meaning that common property such as parking lots and recreational facilities are owned collectively by the units’ owners. Buyers of ECs also have to seek financing from commercial banks, just like PC buyers.3

As with PCs, private developers and their marketing agents handle the sales and pricing of ECs. However, only Singaporean households whose combined monthly income is less than S$14,000 at the time of application are eligible to buy ECs directly from developers.4 In contrast, the sale of PCs is not subject to these eligibility conditions. On obtaining a temporary occupation permit for EC devel- opment, the developer invites buyers to take possession of their EC units. Buyers of newly built EC units are subject to a set of restrictions. One is a minimum occupation period: buyers of ECs cannot sell or rent their units within the first 5 years of ownership.5 After the fifth year, EC units can be sold on the open market. Buyers of ECs resold in the open market are not bound by most of the eligibil- ity conditions, except that they must be Singaporean citizens or PRs. After the 10th year, all restrictions are lifted, and the EC is then considered a fully privat- ized condominium. This means that foreigners can buy ECs on the open market beginning 11 years after the EC project’s completion. During the 5-year MOP, EC owners cannot invest in private residential property in Singapore or over- seas. The rationale for imposing these restrictions on owners is to ensure that the scheme is effective in accommodating the aspirations of the sandwiched middle class to own and live in affordable condominiums.

The basic mechanism for ECs is the government’s land use planning and land

2 The income ceiling for EC flats, originally set at S$10,000, was subsequently raised to S$12,000 in August 2011 and to S$14,000 in August 2015.

3 Similar to home buyers in the public and private housing markets, buyers of EC units may use savings from the Central Provident Fund—a mandatory social security savings scheme in Singa- pore funded by contributions from employers and Singaporean employees—to make the 15 percent down payment for a property purchase (every buyer has to secure a minimum 5 percent cash down payment). A first-time home buyer can also receive a housing grant of up to S$30,000, depending on household income.

4 Besides income and nationality, other eligibility conditions for purchasing an EC include age, nuclear-family status, and existing property ownership.

5 Although buyers cannot rent a unit to anyone during the minimum occupation period, they may rent out bedrooms. Since February 1, 2010, owners must obtain prior approval from the Hous- ing Development Board before renting out bedrooms during the minimum occupation period.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 339

sales.6 In Singapore, 90 percent of land is owned by the government, and land par- cels are sold to private developers through a first-price sealed-bid auction mech- anism called the Government Land Sales Program (Ooi, Sirmans, and Turnbull 2006, 2011; Chow and Ooi 2014). Developers who bid for parcels designated for ECs know that they can build only ECs and are likely to bid less than they would for similar parcels designated for PCs. Therefore, from the government’s perspec- tive, it must forgo the revenue from land sales of ECs compared with similar land sales of PCs.7

2.3. Economic Intuitions

The option theory is particularly useful in understanding the underlying eco- nomic mechanisms of ECs. As shown in Figure A1, let us assume two types of put options: a put option to sell the property to Singaporean residents and a put option to sell to foreigners.8 Buyers of PCs have the American options for both: AP(S), which is the option to sell the property to Singaporean residents at any time, and AP(F), which is the option to sell the property to foreigners at any time. On the other hand, buyers of newly built EC units have the forward-start Ameri- can put options for both at the time of the purchase: FAP(S), which is the option to sell the property to Singaporeans after year 5, and FAP(F), which is the option to sell the property to foreigners after year 10. In other words, buyers of ECs can sell their properties at any time after the specified contract years (fifth year for FAP[S] and 10th year for FAP[F]) before the lease expires (99th year).

The forward-start contracts should impose temporary property rights restric- tions on buyers of ECs and reduce the option price of EC units compared with counterpart PCs, which have regular American put options. Buyers of ECs know of this property rights restriction at the time of purchase, and we believe that this

6 One may question why the government has not chosen alternative policy measures such as rental assistance or a first-time buyer’s tax credit. First, Singapore has a long legacy of promoting homeownership since its independence. Along with the government’s commitment to homeown- ership, most Singaporean residents have very high aspirations for homeownership, and affordable- homeownership programs such as those by the Housing Development Board and ECs are popular among Singaporean residents. Next, from the government’s perspective, forgoing revenue from land sales may be politically more viable than providing large tax credits to a specific group of house- holds. Moreover, the advantage of ECs is that the government can control the minimum occupancy of buyers to prevent speculation activities while promoting the real estate development and con- struction that contribute to the economy. Finally, while the government does not own any EC units, the land will be returned to the government after the expiration of the 99-year lease. This leasehold scheme applies to both ECs and private condominiums (PCs) in our sample.

7 Therefore, the amount of subsidy for an EC scheme would be the loss of revenue on the sale of development sites designated for ECs instead of PCs. For example, the sale price per square meter (PSM) of gross floor area (GFA) was S$3,133 for one EC site (Watercolours), while the sale prices PSM of GFA for PCs located in the adjacent block and sold the same year (Searstrand, Sea Esta, and Ripple Bay) ranged from S$3,604 to S$4,332.

8 Presale is a dominant form of initial sales in Singapore, and these transactions usually start 3 years before the completion of the project. Once purchased as presales, EC units cannot be trans- acted again until specified years (fifth and 10th years) after completion of the project.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

340 The Journal of LAW & ECONOMICS

should contribute to discounts in the initial sales of ECs.9 Discounts in the op- tion price associated with forward-start contracts should diminish over time and eventually disappear when a given EC unit passes the contract date (fifth year for FAP[S] and 10th year for FAP[F]). We can observe these dynamics of discounts of ECs for initial sales and for resales between the fifth and 10th years, when EC units can be sold only to Singaporean residents, and after the 10th year, when ECs can be sold to anyone.

3. Data and Methods

3.1. Data

The main source of data on transactions of individual units is the Real Estate Information System, a database maintained by the Urban Redevelopment Au- thority (URA), Singapore’s national land use planning authority. The transaction data include the sale price, contract date, area (in square meters), floor, and ad- dress of the units. The transaction price is the agreed purchase price of the prop- erty and excludes stamp duties and legal and agency fees, which are fairly stan- dard in Singapore. We obtained project-level data from the quarterly release of data from Property Market Information, a database that is also maintained by the URA. The quarterly release contains the date of written permission, building approval, grant of sale license, marketing launch, and completion of the develop- ments. Finally, we use a geographic information system (GIS) to compute infor- mation about the location, such as distance to the closest subway station and to the central business district (CBD). Table A1 presents the variables used in our analyses and their definitions and sources.

3.2. Identification Strategies

In total, 32,817 transactions were recorded in 41 EC developments between June 1996 and June 2016. They include both initial sales from developers and re- sales in which the original buyers sold their units after satisfying the 5-year MOP. Of the 6,775 resales, 39.2 percent involved the sale of EC units that had achieved full privatization status. Property rights restrictions on ECs and their subsequent relaxation over time provide an opportunity to investigate empirically the treat- ment impact of restrictions on the rights to rent and sell a property and to calcu- late the discount associated with forward-start contracts. To quantify the impact, we identify a comparison group that shares attributes with the treatment group of ECs: PCs are good candidates, as they are very similar to ECs in terms of ame- nities but without the restrictions on property rights discussed above. Hence, we use GIS to identify and select PC projects located within a 2-kilometer radius of the ECs. We draw a 2-kilometer circle around each EC and select PCs within this

9 In other words, the expected discount in the option price for ECs relative to that for compara- ble PCs is likely to affect potential buyers’ utilities attached to EC units. Therefore, they should be willing to pay less than what they would have paid for similar PCs. Hence, we hypothesize that the market price for ECs will clear at a lower price than that of PCs in their initial sales.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Fi gu

re 1

. Lo

ca tio

ns o

f e xe

cu tiv

e an

d pr

iv at

e co

nd om

in iu

m s

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

342 The Journal of LAW & ECONOMICS

circle (see Figure 1). We further restrict the potential counterfactuals to PCs that were launched in the same period as the ECs (between June 1996 and June 2016) and have a 99-year leasehold.10 This results in a pool of 106 PC developments with 53,358 sale transactions between 1996 and 2016 for the potential compari- son group.

To ensure a causal interpretation of our analysis results, we use a propensity- score-matching procedure to select matches for each EC transaction from the pool of 53,358 PC transactions. We use logistic regression to estimate the pre- dicted values to be used as a propensity score. The dependent variable is the EC dummy (which equals one for EC units and zero for PC units), and the inde- pendent variables are transaction year, completion year, area (in square meters), floor, type of sale (namely, presale, initial sale, or resale),11 number of units in the development, distance to the nearest subway station, distance to the CBD, and planning area (neighborhood).12 We match each treatment observation to one comparison observation closest in propensity score within a caliper’s width of .003 without allowing replacement, to improve the covariate balance and reduce bias.13

3.3. Empirical Models

To analyze the difference in price between ECs and PCs in initial-sale trans- actions, we include several important variables in the standard hedonic model. Controlling for all physical attributes such as a unit’s area (in square meters), floor, and age (in years, from the date of the building’s completion to the date of sale), the binary variable EC measures the difference in transacted prices between EC units (the treatment group) and comparable PCs (the comparison group). To control for unobserved spatial heterogeneity, we include a set of binary variables representing the planning areas (neighborhoods) where condominium proj- ects are located (τk). We also include the quarterly Private Residential Property Price Index for nonlanded properties (PPIt), a set of fixed effects for the year of

10 All of the EC developments are built on sites with a 99-year leasehold tenure, while the PCs are sold on either leasehold or freehold tenure. To account for potential differences in prices and quality by land tenure, freehold PCs are not included in the potential comparison group.

11 In our matched sample, more than 95 percent of the initial sales are presales. Still, our propensity- score matching accounts for the potential pricing differences between presale (sale be- fore completion) amounts and spot-sale (sale of competed units) amounts. It includes the sale type with the three categories presale, initial sale, and resale and both transaction year and completion year, which together account for the duration between them and annual interest-rate changes. To further address the potential remaining difference, we redid the matching, including the duration between the transaction year and completion year and the average annual interest rate during this duration. Regression results are almost identical.

12 To facilitate urban planning, the Urban Redevelopment Authority divides Singapore into 55 planning areas. Each planning area has a population of about 150,000 and is served by a town center and several neighborhood commercial and/or shopping centers. We consider these to be neighbor- hoods in our analysis.

13 During this matching process, we lose 9,905 EC observations from our treatment group, as no transactions in the pool of the comparison group had a propensity score within a caliper’s width of .003 of them, and therefore they were left unmatched.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 343

transaction (φt), and neighborhood-by-year fixed effects (τk × φt) to control for time-varying market conditions. Our base regression equation is thus

log EC PPI( ) ,Pikt k ik i t k t k t ikt= + + + + + + +×α π β δ τ ϕ τ ϕ εx (1)

where Pik is the selling price of the ith condominium unit in the kth neighbor- hood in year t, xik is a vector of the explanatory variables containing physical at- tributes and locational characteristics of unit i and neighborhool k, εit is the inde- pendently and identically distributed error term, α is the intercept, and β is the estimated coefficient of our treatment, which has a value of one if the unit is in an EC development. Here β reflects the initial discount for EC units compared with matched PC units, and we expect the sign of this coefficient to be negative.

Equation (1) is the specification used for the matched sample comprising only initial sales. Next we estimate our models for the full sample combining both re- sales and initial sales. In this specification, we add a dummy variable Initial Saleit in the regression model to control for potential differences in the price discovery process in initial sales (units purchased directly from the developer) and resales (units sold by owners).14 We also add the variable Age to account for physical de- preciation of the development’s structure over time, which would influence the transacted prices of both ECs and PCs. Finally, to examine the changes in the differences in price between ECs and PCs over time, as EC units move toward full privatization with no property rights restrictions, we include a set of dummies for the number of years after completion of ECs (EC × Year 5, . . . , EC × Year 10):15

log Initial Sale EC Age EC Year 5

Pikt k ik it i it i

( ) = + + + + + × +

α π β λ ρ ρ

x

1 22 3

4 5 68 9 EC Year 6 EC Year 7

EC Year EC Year EC i i

i i i

× + × + × + × + ×

ρ ρ ρ ρ YYear

PPI 10

+ + + + × +γ τ ϕ τ ϕ εt k t k t ikt .

(2)

In equation (2), the coefficient of EC, β, reflects the initial discount for EC units compared with matched PC units. Controlling for this, we use a set of coeffi- cients, ρ, to indicate the change in resale price of ECs sold in a given year. Hence, this model allows us to test how the resale prices of EC units converge with those of matched PCs as ECs move closer to full removal of property rights restrictions. Linking to economic intuitions presented in Section 2.3, ρ1 represents the for- gone discount associated with the delayed option to sell to Singaporeans (FAP[S], that is, the value of the partial removal of transferability restrictions), and ρ6 rep- resents the full difference in values between American put options and forward- start American put options (that is, the value of the full removal of transferability

14 One potential concern with the initial-sales data is that they may not reflect market- clearing prices. In recent sales practices in Singapore, developers do not release their offer price beforehand and quickly adjust it with demand. Hence, we think that even initial residential sales involve more dynamic rather than static pricing, and thus the prices of initial sales in our sample would be close to market-clearing prices. The sales of earlier groups in our sample may still be subject to developers’ mispricing, which is the limitation of our data.

15 If EC properties have passed the 10-year milestone, the variable EC × Year 10 has a value of one.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

344 The Journal of LAW & ECONOMICS

restrictions). We hypothesize that the discount for resale prices of ECs compared with those of counterpart PCs after the 10th year, given by β + ρ6, will be smaller than the discounts for ECs in earlier years.

4. Results

4.1. Summary Statistics

Table 1 presents the summary statistics of the sample before and after we per- form the propensity-score-matching procedure. The descriptive statistics of the sample before matching are based on the initial sample of 86,175 sale transactions between 1996 and 2016. All monetary values are expressed in Singaporean dol- lars.16 The mean selling prices of EC and PC units are S$736,019 and S$845,323, respectively, which indicates a 12.9 percent discount for EC units. However, EC units tend to be larger and slightly farther from the CBD and subway stations. The average PC development is larger than the average EC development.

After matching, we have 22,912 EC unit transactions and 22,912 PC unit trans- actions. According to the covariate balance of this matched sample, EC and PC units are statistically homogeneous with respect to the transaction and comple- tion years, area and floor, sale type (initial or resale), number of units in the de- velopment, distance to the nearest subways and the CBD, and neighborhood (see Table A2).17 The average treated EC unit and counterpart PC has an area of 114 square meters and is between the eighth and ninth floors. In terms of location, the average distance to the CBD and the closest subway station are 14.5 kilome- ters and 1 kilometer, respectively. Given that Singapore is a small city-state, this suggests that the EC and PC units in our final matched sample tend to be outside the prime residential district in the central region.

4.2. Effects of Property Rights Restrictions on Housing Prices

Table 2 reports the parameter estimates for the hedonic models, which are es- timated using ordinary least squares regression. Model 1 employs the matched subsample of transactions in the initial-sale market, while model 2 employs the matched full sample of transactions in both the initial-sale and resale markets. Model 3 employs the matched subsample of projects completed before 2009, which belong to the pioneering group and have all units eligible for resale. With this more balanced sample, we can examine the potential change in the economic effect of restrictions on property rights depending on the vintage of ECs and test the stability of our main results from model 2. Model 1 uses equation (1), while models 2 and 3 use equation (2).

Unsurprisingly, the movements in the price of residential properties at the broad market level (PPI) have a significant influence on condominium prices.

16 The exchange rate in January 2017 was S$1 to US$.71. 17 This contrasts with some heterogeneities between ECs and PCs in the initial sample, as stated

above.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Ta bl

e 1

Su m

m ar

y St

at is

ti cs

V ar

ia bl

e

Be fo

re M

at ch

in g

A fte

r M at

ch in

g

EC PC

EC PC

M ea

n SD

M ea

n SD

M ea

n SD

M ea

n SD

Pr ic

e (S

$) 73

6, 01

9 17

7, 77

5 84

5, 32

3 1,

84 5,

18 9

72 2,

53 2

16 5,

83 4

85 6,

61 5

29 1,

93 3

Tr an

sa ct

io n

Ye ar

20 08

.5 2

6. 24

20 08

.0 5

5. 67

20 08

.3 6

6. 12

20 08

.2 2

5. 49

C om

pl et

io n

Ye ar

20 08

.4 2

7. 50

20 08

.4 4

7. 07

20 08

.2 3

7. 44

20 08

.1 7

6. 87

Fl oo

r A re

a (m

2 ) 11

5. 54

23 .4

3 10

7. 23

40 7.

43 11

4. 05

22 .4

1 11

2. 79

39 .2

6 Fl

oo r L

ev el

8. 51

4. 96

8. 28

5. 79

8. 47

5. 13

8. 51

5. 98

Sa le

T yp

e 1.

21 .4

0 1.

33 .5

7 1.

23 .4

2 1.

22 .4

1 D

ev el

op m

en t S

iz e

48 6.

81 17

0. 80

55 6.

81 18

1. 80

50 9.

49 15

8. 19

50 9.

6 15

7. 15

A ge

(y ea

rs )

1. 85

3. 81

1. 62

3. 50

1. 93

3. 74

1. 97

4. 09

D ist

an ce

S ub

w ay

(k m

) 1.

01 .5

1 .8

9 .5

7 .9

8 .5

3 .9

9 .6

D ist

an ce

C BD

(k m

) 15

.0 2

2. 47

13 .5

6 2.

53 14

.4 9

2. 45

14 .4

6 2.

4 N

ei gh

bo rh

oo d

12 .2

4 4.

13 9.

37 5.

16 11

.4 6

4. 22

11 .5

1 4.

83 PP

I 12

3. 06

25 .1

0 12

0. 75

25 .7

7 12

3. 38

24 .4

3 12

0. 96

26 .1

5 Tr

an sa

ct io

ns 32

,8 17

53 ,3

58 22

,9 12

22 ,9

12 N

ot e.

V al

ue s a

re fo

r p riv

at e

co nd

om in

iu m

s ( PC

s) lo

ca te

d w

ith in

a 2

-k ilo

m et

er ra

di us

o f e

xe cu

tiv e

co nd

om in

iu m

s ( EC

s) . F

or S

al e

Ty pe

, 1 =

in

iti al

sa le

(i nc

lu di

ng p

re sa

le ),

2 =

re sa

le , a

nd 3

= su

bs al

e ( sim

ila r t

o re

sa le

b ut

u su

al ly

o cc

ur rin

g be

fo re

co m

pl et

io n

of th

e b ui

ld in

g) . S

al e T

yp e

is us

ed fo

r m at

ch in

g.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Table 2 Regression Results Using the Matched Sample

Variable Model 1 Model 2 Model 3 Constant 11.9955** 11.9977** 12.0606**

(.20872) (.12736) (.14175) Floor Area .01267** .01195** .00960**

(.00064) (.00051) (.00050) Floor Area2 −.00002** −.00002** −.00001**

(.00000) (.00000) (.00000) Floor Level .01226** .01036** .00949**

(.00172) (.00185) (.00193) Floor Level2 −.00034** −.00028* −.00026**

(.00009) (.00009) (.00008) Development Size −.00007+ −.00009* −.00003

(.00004) (.00003) (.00005) Distance Subway −.05859** −.05356** −.02637

(.01286) (.01126) (.02049) Distance CBD −.01104 −.01393+ −.01819*

(.00975) (.00668) (.00613) PPI .00875** .00854** .00893**

(.00169) (.00076) (.00095) EC −.22839** −.23270** −.21541**

(.02099) (.02073) (.03072) Age −.01403** −.01626**

(.00137) (.00257) Initial Sale .05580** .08608**

(.01652) (.02388) EC × Year 5 .12167** .07931**

(.02192) (.02582) EC × Year 6 .10210** .07791**

(.02346) (.02319) EC × Year 7 .12518** .08764**

(.02493) (.02211) EC × Year 8 .13791** .10083**

(.02633) (.02488) EC × Year 9 .15696** .12544**

(.02262) (.02317) EC × Year 10 .16596** .12308**

(.02166) (.03242) Year × Neighborhood Yes Yes No N 35,599 45,824 23,086 Adjusted R2 .924 .914 .878 Note. Robust standard errors clustered by neighborhood are in parentheses. Model 1 includes initial sales. Model 2 includes all sales. Models 3 includes only projects completed before 2009. All regres- sions include year and neighborhood dummies.

+ p < .10. * p < .05. ** p < .01.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 347

Most of the structural variables also have the expected signs. Sale price is related positively to the condominium unit’s size, but its influence increases at a decreas- ing rate. Units on upper floors are also popular, as supported by the higher prices they fetched. Distances to subway stations are related negatively to the sale price, contrary to the findings of Munneke et al. (2011), Ooi and Le (2013), and Ooi, Le, and Lee (2014). A possible explanation is that our sample covers only entry-level condominiums whose buyers may have a tighter budget for car ownership. For them, the positive externality of easy access to mass transportation is stronger than the countervailing negative externality from pedestrian congestion or noise from a metro station.

The results in Table 2 reveal a negative and statistically significant coefficient for the variable EC across all specifications. Controlling for observed attributes of units and projects, the restriction on transferability imposed on ECs for 10 years is associated with an initial price reduction of about 21 percent (models 1 and 2) compared with similar PCs with no restrictions.18 When we limit our sample to the older group of ECs (model 3), the initial price discount is about 19 percent. This price discount, caused by restricting the transferability of property rights, is comparable to the 17.7–23.1 percent decrease in residential condominium prices due to age restrictions for occupants (Carter et al. 2013) and larger than the 11 percent price premium for condominiums with a policy that does not restrict pets (Lin, Allen, and Carter 2013).

Consistent with housing depreciating with age, models 2 and 3 show that the coefficient for Age is negative and statistically significant. The depreciation rate for our matched sample, ranging from 1.4 percent to 1.6 percent per annum, is slightly less than the 2.5 percent annual depreciation rate recorded for the US housing market by Harding, Rosenthal, and Sirmans (2007). We believe that this is because land is more limited and the contribution of land to housing value is higher in Singapore than in the United States. As described in Section 3.3, the purpose of adding the interaction between EC and Year in models 2 and 3 is to separate the impact of aging among ECs from that among PCs. Therefore, in con- trast to the negative coefficient for Age, the coefficients for EC × Year are posi- tive. For example, ECs sold after the fifth year have a premium of approximately 13 percent over the initial sales of EC units (model 2). In other words, these units have only about an 8 percent discount compared with the 21 percent discount on newly built EC units. The 13 percent price recovery for these units is associated with the partial removal of property rights restrictions, as initial buyers of ECs can now sell their units to Singaporeans. The coefficients of EC × Year remain positive and become larger as the 10th year approaches. The ECs sold after the 10th year recover about 18 percent of their initial discount, with only an approx- imately 3 percent discount over PCs (model 2). Controlling for observed market changes, we find that this 5 percent additional recovery in price is associated with

18 Following Kennedy (1981), the initial price discount for EC units is calculated as exp[ var( )] ,b b 1 12 1 1- – exp[ var( )] ,b b 1 12 1 1- – where var( )b



1 is the square of the standard error for b 

1. For more accurate estimation results, we use this method in all reports of the price discount and recovery in this section.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

348 The Journal of LAW & ECONOMICS

the full removal of the remaining property rights restriction on selling to foreign- ers.

Figure 2 shows the price discount of ECs compared with their counterfactual PCs from when they are sold in the presale market (age = 0) to just after the 10- year milestone. When sold in the initial-sale market, the price discount for the EC unit is at its highest, at 21 percent.19 After the fifth year, EC units can be sold on the resale market to Singaporean buyers, and EC buyers are subject to only partial illiquidity. At that point, the EC price discount drops to about 8 percent. The discount is widened slightly and temporarily after the sixth year, potentially because of a surge in the supply of ECs in the resale market. Then the price gap continues to narrow after the fifth year and as ECs move toward full privatiza- tion. After the 10th year, EC owners are free to sell their units to anyone, includ- ing foreigners, and are not subject to any illiquidity constraints. The price gap between ECs and PCs decreases to about 3 percent after the 10th year.

When limiting the sample to the older group of ECs (Table 2, model 3), we still find a significant decrease in the price discount of ECs—from an initial discount of 19 percent to 11 percent after the fifth year and 6 percent after the 10th year. The price recovery from removing the restriction of transferring property to Sin- gaporeans after the fifth year is smaller at 8 percent compared with 13 percent in model 2, while the additional price recovery from full privatization after the 10th year remains similar at about 5 percent. The EC units without property rights restrictions have a permanent discount of about 6 percent compared with a 3 per-

19 Initial buyers of ECs cannot purchase private residential property in Singapore or overseas for the first 5 years, which may be a forgone investment opportunity. As most initial buyers of ECs subject to this restriction do not have the financial capacity to purchase another unit without selling their EC unit, given the income eligibility ceiling, we do not think this opportunity cost would be too large. In other words, transferability of the EC unit would be the precondition for another property investment. Anyone not subject to this precondition, not being able to account for the restriction on buying additional private residential property, yields an upper bound for the true treatment effect.

Figure 2. Simulated price gaps (%) between executive condominiums and private condo- miniums.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 349

cent discount in model 2. These results suggest that older ECs experienced lower levels of price discount and recovery in their forward-start put options, poten- tially because markets were still uncertain about the outcomes for ECs.

Finally, we compare these regression results with option price values directly simulated using the option-pricing formula (Barone-Adesi and Whaley 1987; Rubinstein 1991). Table 3 shows the option price values directly simulated using the option-pricing formula.20 As discussed in Section 2.3, AP(S) is the option to sell the property to Singaporean residents at any time, and AP(F) is the option to sell the property to foreigners at any time. Unlike these options given to the owners of PCs, newly built EC units have FAP(S), which is the option to sell the property to Singaporeans starting after year 5, and FAP(F), which is the option to sell the property to foreigners after the 10th year. In our analysis, we assume that the only difference between FAP(S) and FAP(F) is the contract date.

The results are strikingly consistent with our regression results in Table 2. The initial option values of ECs with two forward-start American put options (FAP(S) and FAP(F) with contract dates in the fifth and 10th years, respectively) are about 15.9 percent lower than the initial option values of the counterpart PCs with American put options (AP[S] and AP[F]). When one of the forward-start contracts reaches after the fifth year, the gap in the option values reduces to about 4.6 percent. It means that the forgone discount (that is, recovery from the EC’s initial price) associated with reaching the contract end date for FAP(S) and the half milestone for FAP(F) and being free from complete illiquidity is about 11.3 percent. This falls between 8 percent of model 3 and 13 percent of model 2 (Table 2). As an FAP(F) moves toward the contract’s end date after the 10th year, the option prices of ECs and PCs converge and finally become the same. In other words, the forgone discount associated with moving from the 5-year to the 10- year milestone of the contract date for an FAP(F) and being free from all illiquid- ity constraints is about 4.6 percent. This is consistent with the 5 percent price recovery from full privatization in our regression results in Table 2 (models 2 and 3) and Figure 2.

The magnitude of the discount associated with temporary restrictions on for- eign transferability and the partial illiquidity from these restrictions is compara- ble to the 10 percent discount observed by Bailey and Jagtiani (1994) on stocks in the Thai Main Board, which are subject to permanent restrictions on foreign ownership.21 While the theoretical calculation indicates that ECs’ option prices completely converge with the prices of PCs after contract end dates, our regres-

20 We calculate the American put option attached to PCs and values of the forward-start Ameri- can put option attached to ECs over time. For this calculation, we assume that the current prices for ECs and PCs are the same, as we control for all other characteristics of units and developments (that is, the average price of all initial sales in our matched sample, S$797,169), and their strike prices are the same as the current prices. We use the average annual change in price for the volatility (15.39 percent), the market interest rate for the risk-free rate (2 percent), the market yield of condomini- ums for the dividend (2.54 percent), and the lease’s duration for the maturity (99 years).

21 As the 5 percent forgone discount from our analysis does not account for the discount associ- ated with reaching the 5-year milestone of the contract date for selling to foreigners, the total dis- count associated with selling to foreigners should be larger than 5 percent.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Ta bl

e 3

O pt

io n

Pr ic

es B

as ed

o n

th e

O pt

io n-

Pr ic

in g

Fo rm

ul a

Ye ar

0 Ye

ar 1

Ye ar

2 Ye

ar 3

Ye ar

4 Ye

ar 5

Ye ar

6 Ye

ar 7

Ye ar

8 Ye

ar 9

Ye ar

1 0

EC s:

FA

P( S)

22 3,

09 5

23 4,

24 8

23 8,

88 7

24 3,

61 8

24 8,

44 2

25 3,

36 1

25 3,

40 9

25 3,

45 2

25 3,

48 8

25 3,

51 8

25 3,

54 2

FA

P( F)

20 2,

53 9

21 2,

63 5

21 6,

83 3

22 1,

11 4

22 5,

48 0

22 9,

93 1

23 4,

47 1

23 9,

10 0

24 3,

82 0

24 8,

63 3

25 3,

54 2

To ta

l 42

5, 63

4 44

6, 88

3 45

5, 72

0 46

4, 73

2 47

3, 92

2 48

3, 29

3 48

7, 88

0 49

2, 55

2 49

7, 30

8 50

2, 15

2 50

7, 08

3 PC

s:

A P(

S) 25

2, 92

7 25

3, 11

8 25

3, 18

6 25

3, 24

9 25

3, 30

8 25

3, 36

1 25

3, 40

9 25

3, 45

2 25

3, 48

8 25

3, 51

8 25

3, 54

2

A P(

F) 25

2, 92

7 25

3, 11

8 25

3, 18

6 25

3, 24

9 25

3, 30

8 25

3, 36

1 25

3, 40

9 25

3, 45

2 25

3, 48

8 25

3, 51

8 25

3, 54

2

To

ta l

50 5,

85 5

50 6,

23 5

50 6,

37 1

50 6,

49 8

50 6,

61 5

50 6,

72 2

50 6,

81 9

50 6,

90 4

50 6,

97 6

50 7,

03 7

50 7,

08 3

EC d

isc ou

nt (%

) −

15 .9

− 11

.7 −

10 .0

− 8.

2 −

6. 5

− 4.

6 −

3. 7

− 2.

8 −

1. 9

− 1.

0 0

N ot

e. P

ric es

a re

in S

in ga

po re

an d

ol la

rs . E

C s =

e xe

cu tiv

e co

nd om

in iu

m s;

FA P(

S) =

fo rw

ar d-

st ar

t A m

er ic

an p

ut o

pt io

n to

s el

l t he

p ro

pe rt

y to

S in

ga po

re an

s; FA

P( F)

= fo

rw ar

d- st

ar t A

m er

ic an

p ut

o pt

io n

to se

ll th

e p ro

pe rt

y to

fo re

ig ne

rs ; P

C s =

p riv

at e c

on do

m in

iu m

s; A

P( S)

= A

m er

ic an

p ut

o pt

io n

to se

ll th

e p ro

pe rt

y to

S in

ga po

re an

s; A

P( F)

= A

m er

ic an

p ut

o pt

io n

to se

ll th

e pr

op er

ty to

fo re

ig ne

rs .

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 351

sion results using the transaction data suggest a remaining price differential be- tween them, even after the removal of property rights restrictions. We believe that the remaining 3 percent discount for ECs may be due to negative perceptions of place-based subsidized housing (for example, Cummings and Landis 1993; Galster, Tatian, and Smith 1999; Briggs, Darden, and Aidala 1999) or the physical inferiority of ECs.

4.3. Robustness Checks

First, one potential competing hypothesis is that the significant discount for ECs is not from the property rights restrictions but from their significantly lower quality and lack of amenities. Thus, we collect information about the construction quality and projects’ amenities of ECs and PCs in our matched sample and incor- porate those variables into our regression model as control variables.22 Model 1 in Table 4 adds the variables for construction quality and projects’ amenities to the model specification used in model 2 of Table 2.23 It uses the matched sample of the treatment group (ECs) and comparison group (PCs) within a 2- kilometer radius of ECs after excluding observations without data on construction qual- ity. The results demonstrate that the initial price discount of ECs compared with their counterpart PCs is consistent with that shown in model 2 in Table 2, even after controlling for quality and amenities.24

Second, to address the possibility that changes in price are distinctive at only particular milestone years, we include only two variables, EC × Year 5 and EC × Year 10 and rerun our regression. Model 2 in Table 4 removes several inter- actions of EC and Year between the 5- and 10-year milestones from the model specification used in model 2 of Table 2. It uses the matched sample with the comparison group of PCs within a 2-kilometer radius of ECs. The results suggest that the initial discount of ECs is robust at around 20.7 percent. The coefficients for EC × Year 5 and EC × Year 10 are positive and statistically significant. The magnitudes of price recovery after the fifth and 10th years are 13.6 percent and

22 We use the Building and Construction Authority’s (BCA’s) Construction Quality Assessment System (CONQUAS) data for construction quality. To collect the CONQUAS data, BCA regula- tors independently assess the workmanship and construction quality of new buildings in Singapore throughout their construction. As CONQUAS score information is publicly published, there is no information asymmetry between developers and buyers (Chau and Choy 2011). For more informa- tion on CONQUAS, see Ooi, Le, and Lee (2014). From the t-test results (see Table A3), we find little evidence that the construction quality and amenities of PCs are superior to those of ECs. Amenities of the development projects are from PropertyGuru (http://www.propertyguru.com.sg), which is equivalent to Zillow (http://www.zillow.com) in the United States.

23 Since CONQUAS scores may be correlated with attributes of the development and market, we try to employ both raw and orthogonalized CONQUAS scores as alternative proxies for construc- tion quality. The latter are essentially the residuals from regressing the CONQUAS scores on a set of development characteristics and a set of fixed effects for planning districts and time. Results are very similar. We present results using the raw CONQUAS scores.

24 We also find a consistent pattern for the decrease in discount over time for ECs. The initial dis- count for ECs is about 20 percent, and the forgone discount after the fifth year is about 11 percent. After the 10-year milestone, the price gap between EC and PC units is about 1.5 percent, which is even less than the 3 percent of our main results (Table 2, model 2).

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

352 The Journal of LAW & ECONOMICS

17.8 percent, respectively, and the permanent discount for ECs is about 3 per- cent. These are consistent with the results in model 2 in Table 2. To address the remaining confounding changes in market conditions even after controlling for neighborhood- year fixed effects, we also attempt to use different sample cuts and test the stability of our results.25

Finally, we repeat our analysis by restricting the potential comparison group to within a 1-kilometer radius of each EC development. The matching process is identical to that described in Section 3.2 (caliper = .003 with no replacement), and the matched sample is composed of 19,278 EC unit transactions and 19,278 PC unit transactions. Model 3 in Table 4 demonstrates that the results using this new matched sample are robust with our main results shown in model 2 of Table 2. We consistently find a significant initial discount of ECs and a price recovery over time.26

25 We limited one sample to transactions between 2001 and 2009 and another to transactions after 2009 and reran regressions with the specification used in model 2 of Table 2. Results (not shown) suggest that while the initial discount is slightly larger for the post-2009 sample at 22.5 percent, the pattern of the price recovery and a 2–3 percent permanent discount for ECs are consistent for both samples.

26 We also performed other robustness tests, such as using a bounding technique to test the stabil- ity of our estimation results and ensure nonzero effects of the set of interactions of EC and Year, us-

Table 4 Robustness Checks

Model 1 Model 2 Model 3 EC −.22826** −.23200** −.23137**

(.02850) (.02095) (.02398) EC × Year 5 .10958** .12741** .11059**

(.02538) (.01863) (.02275) EC × Year 6 .10970** .09206**

(.03423) (.02264) EC × Year 7 .13755** .14766**

(.03007) (.01887) EC × Year 8 .15600** .15190**

(.03198) (.02405) EC × Year 9 .16466** .15926**

(.03349) (.03183) EC × Year 10 .17440** .16418** .16292**

(.03422) (.02154) (.02078) Construction Quality .00052*

(.00164) Project amenities Yes No No N 33,214 45,824 38,556 Adjusted R2 .918 .914 .920 Note. Robust standard errors clustered by neighborhood are in parentheses. Model 2 includes milestone years only. Model 3 uses private condominiums within a 1- kilometer radius of executive condominiums. All regressions include unit characteristics and year, neighborhood and year-neighborhood dummies.

* p < .05. ** p < .01.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 353

5. Conclusion

This study contributes to the literature by offering empirical evidence of the economic impact of restricting property rights, namely, the right to rent or sell a property. Using a carefully matched sample of 22,912 ECs and 22,912 PCs trans- acted between 1996 and 2016 in Singapore, we find that EC units are sold at prices that are 21 percent lower than those of otherwise identical PCs during the initial launch. The magnitude of the discount provides an upper limit to the impact of imposing restrictions on the transferability of property rights for 10 years. The discount could also be attributable to the eligibility conditions that filter out po- tential buyers who do not meet the conditions, negative perceptions of subsidized housing, or unobserved differences in housing quality between ECs and PCs. We observe that the price discount narrows to 8 percent after the fifth year, when ini- tial EC buyers are allowed to sell their units on the resale market to Singaporean residents. The price gap reduces over the next few years to about 3 percent after the 10th year, when the restrictions on transferability of property rights are com- pletely removed. The remaining discount is likely to be a permanent discount due to unobserved lower quality of ECs and the negative perceptions of them.

Our empirical evidence provides an important implication for affordable- housing programs: restricting the right to transfer property for a temporary pe- riod can significantly reduce initial housing prices, which makes EC units more affordable for initial purchasers. For illustration, consider the average PC unit in our sample, which has a transaction price of S$845,323. The same unit would be worth approximately S$152,158 less if the two types of property rights restric- tions were imposed on the initial home buyer.27 If the initial buyers of ECs occupy their units beyond the contract end dates, the discount associated with forward- start contracts disappears, and the price difference between EC units and com- parable PCs are reduced significantly over time. In other words, EC buyers pay lower prices in the beginning and achieve higher capital gains than do buyers of com parable PCs. Most of these economic benefits of ECs are transferred to middle- class citizens and PRs who are eligible for initial purchase of ECs. We be- lieve that our findings will be of policy interest, particularly in many global cities that have experienced heavy financial and human inflow from foreign countries and, in turn, a significant reduction in affordable homes for domestic middle- income families.

Our results also suggest a significant role of illiquidity to asset pricing in the context of real estate and other assets. We demonstrate that complete and partial

ing a one-to-many matching procedure instead of one-to-one matching, adding project fixed effects to control for unobserved heterogeneity by individual developments, and using a smaller geograph- ical boundary (geographic sectors; Singapore is divided into 118 geographic sectors for computing development charges) for the neighborhood fixed effects. Our main regression results are robust to all of these tests. The results are available from the authors on request.

27 From the perspective of the market, these discounts could be a temporary dissipation. As ex- plained in Section 2.2, this is directly associated with the cost borne by the government in forgoing the revenue from its land sales. The government does so to achieve its policy goal of providing af- fordable housing.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

354

illiquidity generated by the forward-start American put options during the 10- year period leads to a discount of approximately 18 percent in the option prices during this period. Compared with complete illiquidity in the first 5 years, partial illiquidity caused by restrictions on selling the property to foreigners has a re- duced but still negative impact on assets’ pricing. These results provide useful im- plications for other financial markets that restrict resales of assets or foreign own- ership. For example, restricting major shareholders from selling their shares after an initial public offering and imposing nontradable shares among state-owned enterprises in China are likely to have a significant price impact on these shares. Our findings are also relevant to the economic impact of foreign-ownership re- strictions in stock markets in countries such as China and Thailand.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

355

Appendix

Additional Figures and Tables

Fi gu

re A

1. O

pt io

n- pr

ic e

ap pl

ic at

io n

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

356

Fi gu

re A

2. C

om pa

ris on

o f h

ist og

ra m

s: fu

ll un

m at

ch ed

sa m

pl e

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

357

Table A1 Description of Variables and Data Sources

Description Source Price Sale price in 2014 Singaporean dollars REALIS Transaction Year Year of sale of the unit PMI Completion Year Year of the project’s completion PMI Floor Area Area of the unit (m2) REALIS Floor Level Floor on which the unit is located REALIS Initial Sale Equals one for initial sales, including presales, an d zero for

resales REALIS

Development Size Number of units in the development PMI Age Years between project’s completion and the unit’s sale; equals 0

for initial sales and presales PMI

Distance Subway Distance to the nearest subway station (km) GIS Distance CBD Distance to central business district (km) GIS Neighborhood One of 55 planning areas of Singapore delineated by the URA PMI PPI Private Residential Property Price Index (quarterly value) URA Sources. Urban Redevelopment Authority (URA), Real Estate Information System (REALIS) (https:// spring.ura.gov.sg/lad/ore/login/index.cfm); URA, Property Market Information (PMI) (https://www .ura.gov.sg/realEstateIIWeb/transaction/search.action). Note. GIS = geographic information system.

Table A2 Summarized Quality of the

Matched Sample

Bias Variance

Ratio Transaction Year 2.4 1.24 Contract Year .9 1.17 Floor Area 4.3 .33 Floor Level −.7 .73 Sale Type 2.5 1.04 Development Size −.1 1.01 Distance Subway −2.2 .79 Distance CBD 1.1 1.05 Neighborhood −1.1 .76 Note. All results pass Cochran’s rule of thumb, which indicates that the mean difference be- tween the variable and the matched sample is smaller than one-quarter of a standard deviation for the variable and suggests that good balance is achieved after matching, following Cochran (1968) and Ho et al. (2007). The detailed calcu- lation process is available from the authors on request.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

358 The Journal of LAW & ECONOMICS

References

Allen, Marcus T. 1997. Measuring the Effects of “Adults Only” Age Restrictions on Con- dominium Prices. Journal of Real Estate Research 14:339–46.

Bailey, Warren, and Julapa Jagtiani. 1994. Foreign Ownership Restrictions and Stock Prices in the Thai Capital Market. Journal of Financial Economics 36:57–87.

Barone-Adesi, Giovanni, and Robert E. Whaley. 1987. Efficient Analytic Approximation of American Option Values. Journal of Finance 42:301–20.

Briggs, Xavier de Souza, Joe T. Darden, and Angela Aidala. 1999. In the Wake of Desegre- gation: Early Impacts of Scattered-Site Public Housing on Neighborhoods in Yonkers, New York. Journal of the American Planning Association 65:27–49.

Cannaday, Roger E. 1994. Condominium Covenants: Cats, Yes; Dogs, No. Journal of Ur- ban Economics 35:71–82.

Carter, Charles C., Zhenguo Lin, Marcus T. Allen, and William J. Haloupek. 2013. An- other Look at Effects of “Adults-Only” Age Restrictions on Housing Prices. Journal of Real Estate Economics and Finance 46:115–30.

Chau, K. W., and Lennon H. T. Choy. 2011. Let the Buyer or Seller Beware: Measuring Lemons in the Housing Market under Different Doctrines of Law Governing Transac- tions and Information. Journal of Law and Economics 54:S347–S365.

Chow, Yuen Leng, and Joseph T. L. Ooi. 2014. First-Price Sealed-Bid Tender versus En- glish Open Auction: Evidence from Land Auctions. Real Estate Economics 42:253–78.

Coase, R. H. 1960. The Problem of Social Cost. Journal of Law and Economics 3:1–44.

Table A3 Comparisons of Construction Quality and Project Amenities

of the Matched Sample

EC PC

DifferenceMean SD Mean SD Construction quality 83.11 7.27 85.38 6.46 .112 Project-level amenities: Swimming pool .974 .161 .998 .042 .000 Children’s pool .611 .487 .504 .500 1.000 Barbecue pit .940 .237 .912 .283 1.000 Gym .944 .231 .807 .394 1.000 Mini-mart .048 .214 .023 .149 1.000 Pavilion .172 .378 .160 .367 .999 Playground .884 .321 .853 .354 1.000 Sauna .533 .499 .484 .500 1.000 Clubhouse .645 .479 .646 .478 .415 Exercise area .657 .475 .626 .484 1.000 Basketball court .118 .323 .100 .300 1.000 Tennis court .885 .319 .892 .310 .006 Lounge .098 .298 .044 .204 1.000 Game room .109 .312 .070 .255 1.000 Note. Differences are for P-values: mean of private condominium (PC) > mean of executive condominium (EC). For construction quality, N = 16,713 for ECs and 16,501 for PCs; N = 22,912 for project amenities.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Housing Prices 359

Cochran, William. G. 1968. The Effectiveness of Adjustment by Subclassification in Re- moving Bias in Observational Studies. Biometrics 24:295-313.

Cummings, Jean L., Denise DiPasquale, and Matthew E. Kahn. 2002. Measuring the Con- sequences of Promoting Inner City Homeownership. Journal of Housing Economics 11:330–59.

Cummings, Paul M., and John D. Landis. 1993. Relationships between Affordable Hous- ing Developments and Neighboring Property Values. Working Paper No. 599. Univer- sity of California, Institute of Urban and Regional Development, Berkeley.

Demsetz, Harold. 1967. Toward a Theory of Property Rights. American Economic Review 57:347–59.

Ellen, Ingrid Gould, and Keren Mertens Horn. 2013. Creating a Metric of Educational Op- portunity for Assisted Households. Washington, DC: BiblioGov.

Galster, George C., Peter Tatian, and Robin Smith. 1999. The Impact of Neighbors Who Use Section 8 Certificates on Property Values. Housing Policy Debate 10:879–917.

Guntermann, Karl L., and Seongman Moon. 2002. Age Restriction and Property Values. Journal of Real Estate Research 24:263-278.

Guntermann, Karl L., and Gareth Thomas. Loss of Age-Restricted Status and Property Values: Youngtown Arizona. Journal of Real Estate Research 26:255-276.

Harding, John P., Stuart S. Rosenthal, and C. F. Sirmans. 2007. Depreciation of Housing Capital, Maintenance, and House Price Inflation: Estimates from a Repeat Sales Model. Journal of Urban Economics 61:193–217.

Ho, Daniel E., Kosuke Imai, Gary King, and Elizabeth A. Stuart. 2007. Matching as Non- parametric Preprocessing for Reducing Model Dependence in Parametric Causal Infer- ence. Political Analysis 15:199-236.

Hughes, William T., Jr., and Geoffrey K. Turnbull. 1996. Uncertain Neighborhood Effects and Restrictive Covenants. Journal of Urban Economics 39:160–72.

Ihlanfeldt, Keith R. 2007. The Effect of Land Use Regulation on Housing and Land Prices. Journal of Urban Economics 61:420–35.

Kennedy, Peter E. 1981. Estimation with Correctly Interpreted Dummy Variables in Semi- logarithmic Equations. American Economic Review 71:801.

Libecap, Gary D. 1989. Contracting for Property Rights. New York: Cambridge University Press.

Lin, Zhenguo, Marcus T. Allen, and Charles C. Carter. 2013. Pet Policy and Housing Prices: Evidence from the Condominium Market. Journal of Real Estate Finance and Economics 47:109–22.

Lin, Zhenguo, Yingchun Liu, and Vincent W. Yao. 2010. Ownership Restriction and Housing Values: Evidence from the American Housing Survey. Journal of Real Estate Research 32:201–20.

Mahoney, Joseph T. 2005. Property Rights Theory. Pp. 109–38 in Economic Foundations of Strategy. Thousand Oaks, CA: Sage.

McMillen, Daniel P., and John F. McDonald. 1993. Could Zoning Have Increased Land Values in Chicago? Journal of Urban Economics 33:167–88.

Meltzer, Rachel, and Ron Cheung. 2014. How Are Homeowners Associations Capitalized into Property Values? Regional Science and Urban Economics 46:93–102.

Michael, Jeffrey A., and Raymond B. Palmquist. 2010. Environment Land Use Restriction and Property Values. Vermont Journal of Environmental Law 11:437–64.

Munneke, Henry J., Joseph T. L. Ooi, C. F. Sirmans, and Geoffrey K. Turnbull. 2011. Se- quential Sales of Similar Assets: The Law of One Price and Real Estate. Journal of Re-

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

360 The Journal of LAW & ECONOMICS

gional Science 51:355–70. Munneke, Henry J., C. F. Sirmans, Barrett A. Slade, and Geoffrey K. Turnbull. 2013. Hous-

ing Regulation, Externalities, and Residential Property Prices. Real Estate Economics 42:422–56.

Munneke, Henry J., and V. Carlos Slawson. 1999. A Housing Price Model with Endoge- nous Externality Location: A Study of Mobile Home Parks. Journal of Real Estate Fi- nance and Economics 19:113–31.

Netusil, Noelwah R. 2005. The Effect of Environmental Zoning and Amenities on Property Values: Portland, Oregon. Land Economics 81:227–46.

Newman, Sandra J., and Ann B. Schnare. 1997. “. . . And a Suitable Living Environment’’: The Failure of Housing Programs to Deliver on Neighborhood Quality. Housing Policy Debate 8:703–41.

North, Douglass C. 1990. Institutions, Institutional Change, and Economic Performance. Cambridge: Cambridge University Press.

Ooi, Joseph T. L., and Thao T. T. Le. 2013. The Spillover Effects of Infill Developments on Local Housing Prices. Regional Science and Urban Economics 43:850–61.

Ooi, Joseph T. L., Thao T. T. Le, and Nai-Jia Lee. 2014. The Impact of Construction Qual- ity on House Prices. Journal of Housing Economics 26:126–38.

Ooi, Joseph T. L., C. F. Sirmans, and Geoffrey K. Turnbull. 2006. Price Formation under Small Numbers Competition: Evidence from Land Auctions in Singapore. Real Estate Economics 34:51–76.

———. 2011. Government Supply of Land in a Dual Market. Real Estate Economics 39:167–84.

Quigley, John M., and Larry A. Rosenthal. 2005. The Effects of Land Use Regulation on the Price of Housing: What Do We Know? What Can We Learn? Cityscape 8:69–137.

Rogers, William H. 2006. A Market for Institutions: Assessing the Impact of Restrictive Covenants on Housing. Land Economics 82:500–512.

Rohe, William M., and Lance Freeman. 2001. Assisted Housing and Residential Segrega- tion: The Role of Race and Ethnicity in the Siting of Assisted Housing Developments. Journal of the American Planning Association 67:279–92.

Rubinstein, Mark. 1991. Pay Now, Choose Later. Risk 4(2):13. Schwartz, Amy Ellen, Ingrid Gould Ellen, Ioan Voicu, and Michael H. Schill. 2006. The

External Effects of Place-Based Subsidized Housing. Regional Science and Urban Eco- nomics 36:679–707.

Segal, Ilya, and Michael D. Whinston. 2010. Property Rights. Pp. 100–158 in Handbook of Organizational Economics, edited by Robert Gibbons and John Roberts. Princeton, NJ: Princeton University Press.

Sirmans, Stacy G., Lynn MacDonald, David A. MacPherson, and Emily Norman Zietz. 2006. The Value of Housing Characteristics: A Meta Analysis. Journal of Real Estate Fi- nance and Economics 33:215–40.

Talen, Emily, and Julia Koschinsky. 2014. The Neighborhood Quality of Subsidized Hous- ing. Journal of the American Planning Association 80:67–82.

This content downloaded from 150.108.161.220 on November 30, 2019 12:03:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).