Zoning,density,and rising housing prices: A case study in Portland,Oregon

Abstract

Chinese

This study uses path analysis models to evaluate the associations between zoning, development density and the sales prices of new and existing single-family homes at the dwelling-unit level in Portland, Oregon. Development density is associated with the sales prices of single-family homes directly by determining land costs and indirectly by correlating with the size and construction costs of structures. A prominent trend in Portland’s and the nation’s real estate markets is that new single-family homes are getting bigger. Another trend is that single-family homes have been built on smaller lots despite their growing floor areas. Because developers tend to build smaller homes on smaller lots, the decline of lot sizes helps to contain the effect of growing home size on housing prices. However, the counter effect of smaller lot sizes is somewhat weak because home sizes have a stronger association with housing prices than lot sizes. Homebuyers in Portland are willing to pay a premium to live in neighbourhoods with higher densities, which further limits the potential of higher density development as a tool to reduce single-family home costs. In addition to its indirect associations with home prices via the determination of lot and home sizes, zoning exhibits a significant and direct association with the prices of existing single-family homes, but not with the sales prices of new single-family homes. Existing single-family homes in higher density zones tend to sell at lower prices, ceteris paribus, suggesting that the concern of future neighbourhood change prevails over the opportunity for redevelopment.

Keywords

density density zoning home affordability housing prices new urbanism

Introduction

Homeownership has been a vital housing policy goal in the United States in the past decades, as it provides the opportunity to build wealth and offers social benefits (Herbert et al., 2013; Linblad and Quercia, 2015). The dream of owning a home, however, is even further out of reach for many Americans as home prices have risen quickly in many cities and middle-class incomes have remained stagnant in the past few decades. Glaeser and colleagues argue that home prices are high in a number of markets mainly because these places have restrictive zoning regulations that constrain new home construction (Glaeser and Gyourko, 2003; Glaeser et al., 2005a, 2005b). In the past decade, an increasing number of empirical studies have drawn similar conclusions: stringent land use controls, particularly low-density zoning, substantially reduce housing supply elasticities and raise housing prices in local markets (Chakraborty et al., 2010; Green et al., 2005; Gyourko et al., 2008; Ihlanfeldt, 2007). Researchers suggest that cities use zoning reform to legalise higher-density development in suburban areas that are now exclusively zoned for low-density single-family homes to increase housing supply and make housing more accessible to low- and middle-income households (Talen and Knaap, 2003).

Most prior studies examine the effects of zoning regulations at aggregate spatial scales (mostly cities, counties and metropolitan areas) by constructing coarse regulatory indices. They offer little information about the heterogeneous patterns of zoning within a local market and the mechanisms through which density zoning influences housing prices (Knaap et al., 2007). This study partially fills this gap by providing a micro-level perspective on how zoning and development density influence single-family home prices in Portland, Oregon. We introduce the three-layer density zoning system, the general trend of development density and the complex patterns of single-family home development within zoning categories in Portland. We use path analysis models to quantify both the direct and indirect associations between zoning, development density (at both the dwelling unit and neighbourhood levels) and the sales prices of new and existing single-family homes. The purpose is to offer some insights into how up-zoning and new urbanist design might influence the costs of single-family homes at the dwelling unit level in a rising market.

The remainder of this article is organised as follows. The next section reviews literature that examines the connections between zoning, density and housing prices. The third section introduces the study area, the city of Portland, and focuses on its zoning regulations, single-family home developments and home prices. In the fourth section, we describe the data and the methods that we use to evaluate the associations between zoning, development density and single-family home prices. The last section summarises our findings and concludes with a discussion of policy implications.

How zoning and development density might influence housing prices

Zoning restricts how a community can utilise its land by dividing the land into zones in which certain activities are prohibited or permitted. One of the most common components of zoning is limitations on the density of new development (Schuetz, 2008). The choices public officials make regarding zoning can generate a variety of intended and unintended consequences. One consequence that researchers have heavily studied and criticised is how zoning regulations, particularly low-density zoning regulations, can inflate housing costs (Ihlanfeldt, 2007).

At the dwelling unit level, zoning can influence housing prices by regulating lot size. Three components usually determine the supply price of building a house: the price of land on which the housing unit sits, the construction cost associated with putting up the structure itself, and the entrepreneurial profit needed to compensate the homebuilder (Glaeser and Gyourko, 2018). In some expensive markets, land accounts for up to 50% of home values (Davis and Heathcote, 2007). In these markets, low-density zoning can significantly inflate housing prices by constraining lot size to be greater than a given amount (Glaeser and Gyourko, 2003). Moreover, low-density zoning can also indirectly increase housing prices via the increase of construction costs when home developers build larger homes on larger lots to recoup the land costs. Figure 1 demonstrates a positive correlation between home size (measured as the total floor area) and lot size for single-family homes in Portland. It confirms that home developers build larger single-family homes on larger lots. The figure indicates a generally rising trend in the positive correlations between home size and lot size: it is around 0.2–0.3 for single-family homes built before 2000 (those built in the 1960s were the exceptions); strengthens in the 2000s; and reaches a historical high (r = 0.48) after 2010.

Figure 1.

Correlation between lot size and single-family home size in Portland, Oregon.

In addition to its impact on housing prices via the determination of lot and home sizes, zoning could affect the price of a house by determining its redevelopment capacity and future neighbourhood change. If future redevelopment permitted by the zoning code is of higher value, properties with a greater potential for future redevelopment will generate a price premium (Pogodzinski and Sass, 1991). To this extent, homebuyers may be willing to pay higher prices for single-family homes that are zoned for higher densities as they see the opportunity to convert them to multifamily homes, assuming multifamily homes command higher prices on the market. On the other hand, single-family homes in high-density zones are more likely to see the conversion of neighbouring properties into multifamily structures, which leads to the devaluation concern (Fischel, 2001). Therefore, homebuyers may associate single-family homes in higher density zones with both greater redevelopment capacities and higher risks of ‘undesirable’ neighbourhood change.

At the neighbourhood level, zoning and development density could also influence home prices through its amenity effects. Higher housing prices in communities with more restrictive land use controls and low densities may not necessarily be a result of supply constraints; instead, they could reflect households’ willingness to pay more for living in a more exclusive community with better public services (Paulsen, 2013). Citing multiple consumer surveys, Gordon and Richardson (1997) argue that low-density settlement is the overwhelming choice for residential living in the United States. According to Fischel (2001), suburban homeowners usually support low-density zoning and development in and nearby their suburban neighbourhoods because they are concerned that denser development could devalue their homes. Aurand (2010), however, shows that neighbourhoods are better equipped to meet the housing needs of low-income households when they have a higher density and a greater variety of housing types.

At the regional level, zoning regulations, particularly those designed for fiscal or exclusionary purposes, could reduce housing supply by constraining new home construction, especially high-density residential developments (Pogodzinski and Sass, 1991). Glaeser and Gyourko find a considerable gap between home prices and construction costs in high-priced areas such as New York and California, and argue that restrictive zoning regulations are the natural explanation for the gap (Glaeser and Gyourko, 2003; Glaeser et al., 2005b). Green, Malpezzi and Mayo (2005) estimate housing supply elasticities for 45 US metropolitan areas and find that housing supply tends to be less responsive to demand change in places that are more heavily regulated. Chakraborty and colleagues test the effects of zoning on multifamily constructions in the suburbs of six US metropolitan areas and find that zoning, as practiced by suburban governments, tends to limit the construction of multifamily housing below market-determined levels (Chakraborty et al., 2010). Zabel and Dalton (2011) estimate the effect of minimum lot size on house prices in the greater Boston area and show that changes in minimum lot size at the zoning district level have a significant and positive impact on housing prices. Saiz (2010) shows that highly regulated areas are also likely to be geographically constrained and that geography is one of the most important determinants of housing inelasticity via the reduction of the amount of land available for development and the increase of land values and incentives for antigrowth regulations. In addition, zoning can also affect land development by influencing the timing of development by reducing the option value of vacant land and reducing investment uncertainty (Cunningham, 2007; McDonald and Siegel, 1986). An empirical study of the effects of the urban growth boundary (UGB) on land development in King County, Washington shows that the net effect of the UGB on reducing development in designated rural areas was significantly weakened because the imposition of the UGB removed the uncertainty about optimum future land use (Cunningham, 2007).

In summary, zoning and development density affect housing prices from both the supply and the demand sides of the real estate market. At the dwelling unit level, zoning regulates development density that could influence the price of a house directly by determining the land cost and indirectly by affecting the construction cost. Low-density zoning encourages developers to build larger homes to recoup the land costs and leads homebuyers to consume more land than needed. Zoning could also affect home prices by influencing their redevelopment capacities and their chances of experiencing significant neighbourhood change. At the neighbourhood level, development density may affect home prices through its amenity and exclusionary effects. At the regional level, rigid zoning regulations constrain the number of homes that can be built on a given amount of land and make home developers less responsive to demand change, which creates a tighter market and increases housing prices.

This study is different from previous studies in at least two ways. First, most existing studies are conducted on large spatial scales (e.g. cities and metropolitan areas) and measure regulatory restrictiveness by constructing coarse composite indexes of regulation or by counting the number of land use regulations in a region. We scale down our units of measurement to the neighbourhood and dwelling unit levels, which allows us to accurately measure zoning restrictiveness and development density as well as their effects on single-family home prices at a fine spatial scale. Second, existing studies assume homogenous development patterns within zoning categories and ignore that heterogeneous development patterns could occur even in a very restrictive regulatory environment. We demonstrate the complex relationship between zoning regulations and residential development density at the micro-level and shed light on the mechanisms through which density zoning and new urbanist design influence single-family home prices.

Zoning and development density in Portland

Flexible residential zoning with density overlay

The city of Portland is the largest city in Oregon with an estimated 2015 population of 632,309. Portland’s zoning contains three layers of zones: base zones, overlay zones and plan districts. Base zoning divides the city into five zones: open space; single-dwelling residential; multi-dwelling residential; commercial; and employment and industrial zones. Overlay zones superimpose base zones as they modify the regulations of base zones and provide special provisions to address specific subjects in particular areas of the city. The plan district provisions modify any portion of the regulations of the base and overlay zones in some sites within the city.

We focus on Portland’s residential zoning, primarily single-family home (SFH) zoning in suburban areas, because researchers cite low-density single-family home zoning as one of the most primary contributors to constrained housing supply (Aurand, 2010; Chakraborty et al., 2010). We do not consider up-zoning single-family to multifamily because it is still rare in Portland. Incremental up-zoning within the single-family category (e.g. from low density single-family residential to medium single-family residential) is much more common and practical.

In 2016, the city of Portland zoned 53% of its land for residential purposes, and zoned 82% of the residential land for single-family homes. We further break down single-family residential land into six specific categories. As shown in Table 1, R5 is the most common single-family zone and comprises about 41% of Portland’s single-family area. R7 and R10 account for about 40% of Portland’s single-family home area. Portland zoned the remaining 19% of the single-family home area for either extremelylow-density development (RF and R20) or high-density single-family home development (R2.5). A distinct feature of Portland’s single-family zoning is that it allows a wide range of densities within each zoning category. For example, although lot size in the R5 zone is expected to be around 5000 square feet (about nine units per acre), it ranges from 3000 square feet (about 15 units per acre) to 8500 square feet (about five units per acre). The wide density range in each zoning category offers flexibility to home developers. It also provides us with a unique opportunity to observe how developers behave when they have such flexibility.

Table 1.

Single-family home zones in Portland.

Zone name	Short name	Base zone			Density overlay
Zone name	Short name	Lot size (sqft)	Area (acre)	Share^a	Area (acre)	Percent^b
ResidentialFarm/Forest	RF	52,000–151,000	3090	7%	0	0
Residential 20000	R20	12,000–34,500	2061	5%	0	0
Residential 10000	R10	6000–17,000	7252	18%	0	0
Residential 7000	R7	4200–12,000	9218	22%	1394	15%
Residential 5000	R5	3000–8500	16819	41%	6077	36%
Residential 2500	R2.5	1600–3000	2949	7%	1742	59%

Notes: ^a The numbers in this column show the share of each zone category in single-family home areas. ^b The numbers in this column show the share of land in each category that is subject to density overlay zoning. Increased density is possible but not guaranteed in density overlay zones. Developments have to meet additional requirements to be built at higher densities.

Source: Compiled by the authors based on the Regional Land-use Information System.

The city of Portland subjects parts of its medium- and high-density single-family zones (R7, R5 and R2.5) to alternative design density overlay zoning (Table 1). The purpose of Portland’s alternative design density overlay zoning is to encourage higher-density infill development in areas that are well served by existing public services. In R5 zones, density can be up to 2500 square feet per dwelling unit. In R2.5 zones, attached single-family or low-density multifamily residential development may be allowed in areas that are further controlled by alternative design density zones.

The trend of residential density in Portland: Bigger homes on smaller lots

We use lot size as an indicator of residential density at the parcel level in single-family home zones. Figure 2 shows the median lot sizes and floor areas of single-family homes that were built in Portland and the United States in different periods. The median lot size in Portland for single-family homes built before 1930 is about 5000 square feet. The median lot size of Portland’s single-family homes started to grow after 1930 and peaked in the 1960s at about 7600 square feet. The median lot size of new homes has been declining steadily since the 1970s, falling below 4000 square feet after 2009. To this extent, Portland’s single-family homes zones have been experiencing densification since the 1970s.

Figure 2.

Medium lot and home sizes of single-family homes built in different time periods in Portland.

Portland’s single-family homes built before 1940 have a varied area ranging from 1512 square feet (in the 1920s) to 1791 square feet (in 1930s). Single-family homes from the 1940s and 1950s are quite small, with median sizes of 1407 square feet and 1360 square feet respectively. The floor areas of single-family homes have been growing since the 1970s and exceeded 2000 square feet after 2009. Therefore, a prominent trend in Portland’s single-family housing market since the 1970s is the building of bigger homes on smaller lots.

A comparison between Portland and the entire nation shows that Portland’s single-family homes are smaller than the national average by floor area and lot size. However, a similar trend can be seen throughout the nation: the median lot size of single-family homes has been shrinking while their median floor area has been growing. In other words, American developers are building bigger single-family homes with smaller yards.

Recent single-family home development

We identify 2785 new single-family homes that were built in the City of Portland between January 2012 and June 2016. Most of these new developments are in areas that are already well served by public facilities because Portland is already built up and has few large-size vacant residential lots for new development. The spatial distribution of these new single-family homes follows the patterns of existing single-family homes. The median distance from these new homes to the central business district (CBD, represented by the location of the city hall) is 4.2 miles, which is close to the median distance from all the existing single-family homes to the CBD (4.4 miles). The median lot size of new single-family homes built between 2012 and 2016 is 3995 square feet and their median floor area is 2220 square feet. Due to missing information, home transaction records are only available for 86.5% these new single-family homes (2408 out of 2785). The median and mean prices of these 2408 new homes are US$389,501 and US$412,771 (both in 2010 value) respectively.

About a half of these new single-family homes are in R5 zones. About 30% of them are in the high-density single-family home zone (R2.5) or multifamily home zones (R1, R2 and R3). Less than 10% are in low-density single-family home zones (RF, R20 and R10).

There is a significant amount of heterogeneity in the development density within each single-family home zone. Builders constructed about two thirds (67%) of new single-family homes in R5 zones on relatively small lots that are within 2900–5100 square feet (we use cut-off values 2900 and 5100 square feet instead of 3000 and 5000 square feet to accommodate measurement inaccuracy in digital maps). Builders took advantage of Portland’s alternative density overlay zoning and built almost one-fifth (19%) of new single-family homes on lots smaller than 2900 square feet. Only 18% of the developments in R5 zones are built on lots larger than 5100 square feet, although it is legal to build residential structures on lots up to 8500 square feet in R5 zones. This suggests that Portland’s developers build single-family homes in higher densities when they have the opportunity to do so.

Zoning, density and single-family home prices

We evaluate the relationships between zoning, density and single-family home prices based on the sales data of Portland’s new and existing single-family homes that were traded between January 2012 and June 2016. Home prices in the Portland metropolitan area hit bottom in early 2012 and have since recovered (FRED, 2017). This study period thus reflects the recovery and expansion phases of the real estate cycle in Portland.

Model conceptualisation: Path analysis

We test the effect of zoning and development density on the prices of single-family homes through path analysis, an extension of multiple regression. Unlike multiple regression in which a variable can either be an independent variable or a dependent variable, path analysis examines situations in which there are several dependent variables and those in which there are chains of influence (e.g. variable A influences variable B, which in turn affects variable C) (Streiner, 2005). In path analysis, variables are either exogenous, meaning their causes lie outside the model, or endogenous, meaning that other variables in the model determine them. A path analysis can be conducted based on a series of sequential multiple linear regressions, one for each endogenous variable. The model estimates all the regressions simultaneously and yields overall goodness of fit measure.

In our path analysis models (Figure 3), we assume that a bundle of services and (dis)amenities at both the dwelling-unit and neighbourhood levels determines the price of a single-family home. This includes the following: building structure measured as floor area; land lot measured as lot size; neighbourhood environment (physical and social); public services associated with the location of the house; and zoning.

Figure 3.

Model conceptualisation (path analysis).

We assume that zoning is exogenous because it is determined at the municipal level and is unlikely influenced by individual development projects. In addition to its indirect effects on home prices via the determination of lot and home sizes, zoning could directly affect property values by shaping their redevelopment capacities and their chances of experiencing significant neighbourhood change in the future. The two impacts may run in opposite directions and the overall effect depends on which one dominates. To quantify the direct effect of zoning on housing prices, we categorise studied single-family homes into three zones based on their zoning codes: low density zones, medium density zones and high density zones. In the model, we use single-family homes in medium density zones as the reference to which single-family homes in the other two zones are compared.

We measure density at both the dwelling-unit level (represented by lot size) and the neighbourhood level (measured by the net residential density). We assume the effect of lot size is endogenous as it is affected by other variables in the model such as land use density at the neighbourhood level and zoning codes. Lot size can directly influence housing prices by determining the cost of land; it can also indirectly affect housing cost by affecting the size of the structure built on it. Land use density at the neighbourhood level, however, is assumed exogenous because builders and homebuyers at the dwelling unit level are unlikely to be able to change neighbourhood density, particularly given that most residential developments in Portland exist in built-up areas. Neighbourhood density influences housing prices in two ways, directly as a neighbourhood amenity and indirectly as a factor that influences lot sizes of individual homes within the neighbourhood.

New and existing single-family home sales and control variables

We present the descriptive statistics of the single-family home units (new and existing) that were traded in Portland from 2012 to the first half of 2016 in Table 2. We collect the data from three primary sources including the Regional Land-use Information system (Portland Metro, n.d.), the 2010–2014 American Community Survey (Manson et al., 2018) and the School Report Cards data set (Oregon Department of Education, n.d.).

Table 2.

Characteristics of new- and existing-home sales (January 2012–June 2016).

Variable	New-home sales			Existing-home sales
Variable	Median	Mean	Std. Deviation	Median	Mean	Std. Deviation
Sale price (in US$1000, 2010 value)	392.56	412.33	177.43	286.36	329.83	217.75
Floor area (in 1000 sqft)	2.22	2.33	0.72	1.59	176.81	0.82
Lot size (in 1000 sqft)	3.77	4.39	0.41	5.07	6.91	6.79
Density zoning:
Low-density zoning (yes = 1)	0.00	0.06	0.23	0.00	0.07	0.26
Medium-density zoning (yes = 1)	0.00	0.38	0.48	0.00	0.49	0.50
High-density zoning (yes = 1)	1.00	0.57	0.50	0.00	0.44	0.50
Neighbourhood environment:
Net residential density (units/acre)	9.16	9.75	5.03	8.51	9.39	6.63
Mixed land use:
Low level mixed use (yes = 1)	0.00	0.21	0.41	0.00	0.26	0.44
Medium-low level mixed use (yes = 1)	0.00	0.24	0.43	0.00	0.25	0.43
Medium-high level mixed use (yes = 1)	0.00	0.33	0.47	0.00	0.25	0.43
High level mixed use (yes = 1)	0.00	0.22	0.42	0.00	0.25	0.43
Elementary school quality (%)	68.13	65.14	18.61	67.97	65.99	18.35
Median household income (US$1000)	58.19	61.76	23.08	58.02	64.27	27.88
Percent of White population (%)	76.44	73.79	14.16	76.73	73.67	15.16
Home age	–			68.00	68.01	30.34
Demolition and redevelopment	0.00	0.28	0.45	–
Within 1/2 mile of light rail transit (yes = 1)	0.00	0.18	0.38	0.00	0.18	0.38
Employment accessibility by auto	3.62	3.65	0.27	3.66	3.61	0.28
Distance to CBD (in mile)	4.29	4.43	1.55	4.43	4.68	1.86
Submarket:
North	0.00	0.18	0.39	0.00	0.13	0.34
Northeast	0.00	0.25	0.43	0.00	0.31	0.46
Northwest	0.00	0.02	0.12	0.00	0.03	0.16
Southeast	0.00	0.45	0.50	0.00	0.39	0.49
Southwest	0.00	0.11	0.31	0.00	0.14	0.35
Transaction year:
2012	0.00	0.14	0.35	0.00	0.16	0.36
2013	0.00	0.19	0.39	0.00	0.19	0.39
2014	0.00	0.23	0.42	0.00	0.21	0.41
2015	0.00	0.29	0.45	0.00	0.26	0.44
2016	0.00	0.16	0.36	0.00	0.18	0.38
Number of observations (N)	2329			35341

New single-family homes account for a small portion (7%) of total home sales in Portland between 2012 and 2016. Among the 2785 new single-family homes we examine, we are not able to include 456 of them (16%) in the model due to missing variables. Compared with the 2329 homes included, the 456 homes that we dropped are more likely to be built on larger lots (median size 5000 vs. 3769 square feet) and slightly further away from the CBD (median distance 4.5 vs. 4.3 miles). There were 36,245 existing single-family home transactions in Portland from 2012 to 2016. We remove 904 of them (2.5%) due to missing variables or suspicion of non-arm’s length transactions (sale prices below US$50,000). We thus include 35,341 observations in the model for existing single-family home sales.

New single-family homes tend to be more expensive when measured by median sales price (US$392,562 vs. US$286,364 in 2010 value), larger when measured by median floor area (2220 vs. 1590 square feet) and denser when measured by median lot size (3769 vs. 5070 square feet) than existing homes. This finding is consistent with the finding above that home developers in Portland and more broadly the United States are building larger homes on smaller lots. The construction of 28% of new single-family homes involves the demolition of preexisting building structures. In 2016, the median age of traded existing homes was about 68 years old. New- and existing-home sales are quite similar regarding their relative location to CBD (4.3 vs. 4.4 miles). Precisely the same share of them (18%) are located within a half mile of rail transit stations.

We use the 2010 census block groups to represent neighbourhoods in Portland. We develop two variables, net residential density and mixed land use, to represent their physical environment. We calculate net residential density as the number of housing units per acre of residential land in a neighbourhood. We measure mixed land use as a ratio between the number of retail and food-related employment and the total number of residents in the neighbourhood. Because the impact of mixed land use on home prices may not be linear, we categorise neighbourhoods into four roughly even groups based on the levels of mixed use: zero or low, medium-low, medium-high and high. To measure job accessibility, we develop an employment accessibility index at the traffic analysis zone level based on the negative exponential travel impedance formula.

We use three variables to indicate the socioeconomic environment of each neighbourhood: median household income, elementary school quality and racial composition. We use median household income as reported by the 2010–2014 American Community Survey. We use the average percentage of students in an elementary school who met or exceeded state standards of maths, reading and science in the academic year 2012–2013 as the elementary school quality measure. Table 2 shows that new and existing single-family home sales are quite similar regarding their neighbourhood environment. Therefore, significant differences exist at the dwelling unit level, such as floor area, lot size and building age.

We include Portland’s five housing submarkets and the geographic coordinate of each dwelling unit as the two variables to control for the potential spatial correlation between neighbouring dwelling units. We also include transaction year fixed effects in our model to capture the time trend because housing prices are in recovery from the 2008 housing crisis during the study period.

Modelling results

We present the results of the two path analysis models for new and existing single-family home prices in Table 3. Each model contains three dependent variables: floor area, lot size and sales price. We report both the original and standardised coefficients. We remain consistent with previous hedonic home pricing models and transform sale price, floor area and lot size into a natural logarithm. We also transform neighbourhood density into a natural logarithm to remain consistent with this natural logarithm transformation and to make interpretation easier.

Table 3.

Path analysis model results.

Explanatoryvariable	New home sales: Dependent variable									Existing home sales: Dependent variable
	Ln(lot size)			Ln(floor area)			Ln(home price)			Ln(lot size)			Ln(floor area)			Ln(home price)
	coef.	std. coef.	p-value	coef.	std. coef.	p-value	coef.	std. coef.	p-value	coef.	std. coef.	p-value	coef.	std. coef.	p-value	coef.	std. coef.	p-value
Endogenousvariable
Ln(floor area)	–	–	–	–	–	–	0.553	0.323	0.000	–	–	–	–	–	–	0.498	0.432	0.000
Ln(lot size)	–	–	–	0.332	0.581	0.000	−0.024	−0.024	0.323	–	–	–	0.307	0.370	0.000	0.112	0.117	0.000
Exogenousvariable
Density zoning:
Medium density(reference)	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Low density	0.515	0.239	0.000	−0.001	−0.001	0.961	−0.006	−0.003	0.884	0.498	0.249	0.000	0.024	0.014	0.009	0.046	0.024	0.000
High density	−0.215	−0.211	0.000	−0.036	−0.061	0.000	−0.032	−0.032	0.094	−0.085	−0.082	0.000	−0.052	−0.060	0.000	−0.049	−0.049	0.000
Ln(neighbourhooddensity)	−0.266	−0.228	0.000	–	–	–	0.101	0.088	0.002	−0.241	−0.221	0.000	–	–	–	0.106	0.102	0.000
Mixed land use:
No or low level (reference)	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Medium-low level	–	–	–	–	–	–	0.012	0.010	0.606	–	–	–	–	–	–	0.017	0.015	0.000
Medium-high level	–	–	–	–	–	–	0.037	0.036	0.113	–	–	–	–	–	–	0.029	0.025	0.000
High level	–	–	–	–	–	–	−0.028	−0.023	0.288	–	–	–	–	–	–	−0.008	−0.007	0.090
Home age/10	–	–	–	–	–	–	–	–	–	0.214	1.260	0.000	−0.157	−1.114	0.000	−0.027	−0.167	0.000
(Home age/10)²	–	–	–	–	–	–	–	–	–	−0.014	−1.062	0.000	0.011	0.984	0.000	0.002	0.162	0.000
Redevelopment	0.262	0.233	0.000	–	–	–	0.018	0.017	0.295	–	–	–	–	–	–	–	–	–
Light rail transit(LRT)	–	–	–	–	–	–	−0.251	−0.195	0.000	–	–	–	–	–	–	−0.099	−0.077	0.000
LRT*Ln(distanceto CBD)	–	–	–	–	–	–	0.153	0.170	0.002	–	–	–	–	–	–	0.040	0.045	0.000
Ln(distanceto CBD)	0.058	0.042	0.047	−0.269	−0.340	0.000	−0.286	−0.212	0.000	0.158	0.135	0.000	−0.321	−0.330	0.000	−0.320	−0.285	0.000
Employment accessibility	–	–	–	–	–	–	0.062	0.034	0.316	–	–	–	–	–	–	0.111	0.063	0.000
Median income	–	–	–	0.002	0.178	0.000	0.003	0.135	0.000	–	–	–	0.003	0.212	0.000	0.003	0.162	0.000
School quality	–	–	–	–	–	–	0.000	−0.009	0.688	–	–	–	–	–	–	0.001	0.044	0.000
White population	–	–	–	–	–	–	0.004	0.124	0.000	–	–	–	–	–	–	0.003	0.099	0.000
Fixed effects variable
Transaction year	Yes			Yes			Yes			No			Yes			Yes
Submarket	Yes			Yes			Yes			Yes			Yes			Yes
Coordinate	Yes			Yes			Yes			Yes			Yes			Yes
Model goodness-of-fit
R-square	0.484			0.596			0.492			0.433			0.326			0.720
CFI	0.982									0.984
RMSEA	0.047									0.041
SRMR	0.007									0.006
N	2329									35341

Note: Significant coefficients in bold and italic .

The fit indices, as reported in Table 3, suggest a strong representation of the data: the comparative fit indices (CFI) are well above 0.90 (0.982 and 0.984); the standardised root mean square residuals are much lower than 0.05 (0.007 and 0.006); the root mean square errors of approximation (RMSEA) are below 0.05 (0.047 and 0.041). The R-square values of the two sales price equations indicate that the two models can explain 49% of the variability of new home sales prices and 72% of the variability of existing single-family home sales prices. Measured by the magnitude of the standardised coefficients, the four most important predictors of the sales prices of both new and existing single-family homes are floor area, distance to CBD, median household income in the neighbourhood and the share of the White population in the neighbourhood.

Zoning does not seem to show a significant and direct association with new single-family home sales prices; however, it does demonstrate a significant and direct association with the sales prices of existing homes. Existing single-family homes in high-density zones tend to sell at lower prices, ceteris paribus. This finding suggests that the concern of future neighbourhood change rather than redevelopment capacity dominates the association between high-density zoning and the sales prices of existing single-family homes.

The associations between parcel-level density (measured by lot size) and sales prices are different between new and existing single-family homes. Lot size does not seem to have a significant and direct association with the prices of new single-family homes. Lot size is indirectly associated with the sales prices of new single-family homes through its positive correlation with floor area. For existing single-family homes, lot size is significantly associated with sales prices both directly and indirectly. Because we use home price, lot size and floor area in a natural logarithm, we interpret their coefficients as elasticities. Everything else being equal, a 10% increase in lot size is indirectly associated with a 1.59% increase in the sales prices of new homes, mainly through its correlation with floor area. A 10% increase in lot size is associated with a 2.65% increase in the sales prices of existing single-family homes, with almost 60% of the increase due to an increase in floor area. The model results confirm the existence of our hypothesised direct and indirect associations between neighbourhood density and home prices, but the direct association dominates. In general, the direct association between neighbourhood density and single-family home prices is positive in both models, meaning that single-family home prices tend to be higher in denser neighbourhoods, after controlling for the distance to the CBD. Everything else being equal, a 10% increase in neighbourhood density is associated with a 1.1% increase in the sales prices of new single-family homes and a 0.8% increase in the sales prices of existing single-family homes.

As indicated by standardised coefficients, floor area is the most important predictor of the sales prices of both new and existing single-family homes. A 10% increase in floor area is associated with about a 5.5% increase in new single-family home prices and a 5.0% in existing single-family home prices. For existing single-family homes, about a third of the association between floor area and home price reflects the indirect association between lot size and home price.

Building age displays a non-linear relationship with home prices: everything else being equal, home prices decline with building age except for single-family homes that are more than 135 years old (built in 1880 or earlier). Proximity to a light rail station is not always positively associated with single-family home prices. The association between light rail transit and single-family home prices depends on their relative locations to the city centre: the association is positive for new single-family homes that are at least 5.2 miles away from the CBD and existing single-family homes that are at least 11.9 miles away from the CBD. This is probably because access to rail transit is more valuable in suburban areas where the density and quality of public transit service are low. Proximity to CBD shows a strong and positive association with the sales prices of both new and existing single-family homes.

Several other neighbourhood characteristics also show significant associations with single-family home prices. As expected, mixed land use exhibits a non-linear association with the sales prices of existing single-family homes. Mixed land use is positively associated with existing single-family home sales prices when the level of mixed use is moderate. This positive association disappears once the level of mixed use is high and its negative externality starts to dominate. The association between mixed land use and the sales prices of new single-family homes, however, is not statistically significant. Better employment accessibility is positively associated with the sales prices of existing homes but its association with the prices of new single-family homes is not statistically significant. Not surprisingly, single-family homes in neighbourhoods with higher incomes tend to be larger and more expensive. Higher shares of the White population predict higher single-family home prices. School quality demonstrates a positive and significant association with the sales prices of existing single-family homes, but its association with new single-family home prices is not statistically significant after controlling for other variables.

Discussion and conclusion

Before proceeding to the discussion of our findings, it is important to note that the scope of our study is limited in several ways. First, the context of this study represents real estate markets that have high demand but tight supply. Portland was bouncing back quickly from the 2008 housing recession during the study period and experienced one of the fastest home price growths in the nation. Portland has little large-size vacant land and most of the newest developments occur in built-up areas. Second, this study focuses on market-rate single-family homes. Our findings may not apply to the rental and multifamily home markets, which could have even larger potentials to provide affordable homes. Third, unlike the majority of existing studies, we did not design our study to examine whether and how zoning and density affect housing prices through supply constraints at the regional level, as such analyses require a comparison between multiple markets. Fourth, urban planners commonly advocate for an increase in development density to promote environmental sustainability (Moos, 2017). The central theme of this article, however, is the impact of zoning and density on housing prices. Evaluating potential environmental benefits associated with higher development density is outside of the scope of this study. Finally yet importantly, this study reveals only the correlations between zoning, density and single-family home prices due to the cross-sectional nature of our data. The statistical associations between these variables do not imply causation. The literature suggests that zoning could influence home prices through both supply constraint and amenity effects. Also because of the cross-sectional nature of our data, we are not able to disentangle the two effects in our analysis.

Despite its limited scope, this study sheds light on the dynamic relationship between zoning, development density and single-family housing prices in rising real estate markets. Portland’s residential zoning codes allow a wide range of densities in each zoning category which provides us with an opportunity to observe how home developers behave when operating in a flexible density zoning system. We find that most of the new single-family homes were built upon relatively small lots, confirming that residential developers will build higher-density single-family homes when they can. On the other hand, we also observe considerable heterogeneity in development density within each zoning category, which suggests that many other factors are at work in addition to zoning in determining development densities.

A prominent trend in Portland’s housing market over the past few decades is that single-family homes are getting bigger. The quick expansion of home size contrasts with the dwindling household size and stagnant middle-class income in Portland and more broadly in the nation. There could be several reasons for the bigger-home trend. First, as discussed earlier, low-density zoning constrains lot size to be larger than a given amount and pushes developers to build larger homes to recoup the land cost. Developers are even more incentivised to build larger single-family homes in cities where land is expensive and low-density zoning is pervasive. Second, banks tightened lending standards in and after the 2008 housing crisis, making it more difficult for low- and moderate-income families to access home loans. Developers thus focused on the high-end markets and build larger homes. Third, in addition to satisfying basic shelter needs, housing is a positional good that conveys social standing. The high housing spending of top-earners has inspired conspicuous housing consumption (Greenwood and Holt, 2010). No matter what the reasons are, the bigger-home trend reduces housing choices for medium- and low-income households and leads them to spend a higher proportion of their income on housing expenditures (Dong, 2018).

On the other hand, single-family homes are being built on smaller lots despite their growing floor areas. Because developers tend to build smaller homes on smaller lots, the reduction of lot size helps to contain the quick rise of home size and cost in Portland. However, the association between lot size and home price is a lot weaker than that between home size and price, particularly for new single-family homes. Everything else being equal, a 10% decrease in lot size is associated with a 1.6% decrease in new single-family home sales prices and a 2.6% decrease in existing single-family home sales prices. In contrast, a 10% increase in floor area is associated with a 5.5% increase in new single-family home sales prices and a 5.0% increase in existing single-family home sales prices. Therefore, decreasing lot sizes have not been able to fully counter the effect of growing floor areas on housing prices.

Density zoning could directly influence property values by shaping the expectations of redevelopment capacities and future neighbourhood change. We find a negative association between higher-density zoning and the sales prices of existing single-family homes; this does not apply to the sales prices of new single-family homes. Existing single-family homes in higher-density zones tend to sell at lower prices, ceteris paribus, suggesting that the concern of future neighbourhood change prevails over the opportunity of redevelopment.

Single-family homebuyers in Portland demonstrate a strong preference for homes that are close to the city centre. This proximity preference is consistent with a previous finding that the prices of homes that are closer to the city centre were more resilient in the 2008 housing crisis in the Portland region (Dong, 2015). Moreover, homebuyers in Portland show a willingness to pay a premium for living in denser neighbourhoods. This willingness could be good news for new urbanist advocates who associate compact cities and neighbourhoods with many environmental and health benefits. It also could be bad news as these environmentally friendly neighbourhoods command a price premium that will make them less affordable. While new urbanist design is expected to promote socioeconomic diversity, homebuyers in Portland show a preference for neighbourhoods with higher proportions of the White population and higher incomes. Liberating low-density zoning may still help alleviate the upward pressure on housing prices by increasing housing supply at the regional level, but its effects at the dwelling-unit and neighbourhood levels may be weaker than expected, particularly in places where people are more likely to accept new urbanist design.

Footnotes

Acknowledgements

This study is supported by a research grant awarded by the Gazarian Real Estate Center in the Craig School of Business at the California State University, Fresno.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Hongwei Dong

References

Aurand

(2010) Density, housing types and mixed land use: Smart tools for affordable housing? Urban Studies 47(5): 1015–1036.

Chakraborty

Knaap

Nguyen

Shin

(2010) The effects of high-density zoning on multifamily housing construction in the suburbs of six US metropolitan Areas. Urban Studies 47(2): 437–451.

Cunningham

(2007) Growth controls, real options, and land development. The Review of Economics and Statistics 89(2): 343–358.

Davis

Heathcote

(2007) The price and quantity of residential land in the United States. Journal of Monetary Economics 54: 2595–2620.

Dong

(2015) Were home prices in new urbanist neighborhoods more resilient in the recent housing downturn? Journal of the Planning Education and Research 35(1): 5–18.

Dong

(2018) The impact of income inequality on rental affordability: An empirical study in large American metropolitan areas. Urban Studies 55(10): 2106–2122.

Fischel

(2001) The Homevoter Hypothesis: How Home Values Influence Local Government Taxation, School Finance, and Land-Use Politics. Cambridge, MA: Harvard University Press.

FRED, Federal Reserve Bank of St. Louis (2017) S&P Dow Jones Indices LLC, S&P/Case-Shil ler OR-Portland Home Price Index© (POXRSA). Available at: https://fred.stlouisfed.org/series/POXRSA (accessed 17 February 2017).

Glaeser

Gyourko

(2003) The impact of building restrictions on housing affordability. Economic Policy Review 9(2): 21–39.

10.

Glaeser

Gyourko

(2018) The economic implications of housing supply. Journal of Economic Perspectives 32(1): 3–30.

11.

Glaeser

Gyourko

Saks

(2005a) Why have housing prices gone up? The American Economic Review 95(2): 329–333.

12.

Glaeser

Gyourko

Saks

(2005b) Why is Manhattan so expensive? Regulation and the rise of housing prices. Journal of Law and Economics 48(2): 331–369.

13.

Gordon

Richardson

(1997) Are compact cities a desirable planning goal? Journal of the American Planning Association 63(1): 95–106.

14.

Green

Malpezzi

Mayo

(2005) Metropolitan-specific estimates of the price elasticity of supply of housing, and their sources, The American Economic Review 95(2): 334–339.

15.

Greenwood

Holt

RPF

(2010) Growth, inequality and negative trickle down. Journal of Economic Issues 44(2): 403–410.

16.

Gyourko

Saiz

Summers

(2008) A new measure of the local regulatory environment for housing markets: The Wharton residential land use regulatory index. Urban Studies 45: 693–729.

17.

Herbert

McCue

Sanchez-Moyano

(2013) Is Homeownership Still an Effective Means of Building Wealth for Low-income and Minority Households? (Was it Ever?). Boston, MA: Harvard University, Joint Center for Housing Studies.

18.

Ihlanfeldt

(2007) The effect of land use regulation on housing and land prices. Journal of Urban Economics 61: 420–435.

19.

Knaap

Meck

Moore

, et al. (2007) Do we know regulatory barriers when we see them? An exploration using zoning and development indicators. Housing Policy Debate 18(4): 711–749.

20.

Lindblad

Quercia

(2015) Why is homeownership associated with nonfinancial benefits? A path analysis of competing mechanisms. Housing Policy Debate 25(2): 263–288.

21.

McDonald

Siegel

(1986) The value of waiting to invest. The Quarterly Journal of Economics 101(4): 707–728.

22.

Manson

Schroeder

Van Riper

Ruggles

(2018) IPUMS National Historical Geographic Information System: Version 13.0 [Database]. Minneapolis: University of Minnesota.

23.

Moos

(2017) The velocity of density: Can we build more sustainable cities fast enough? Sustainability 9(2326): 1–10.

24.

Oregon Department of Education (n.d.) Report cards for school year 2012–2013. Available at: http://www.ode.state.or.us/data/ReportCard/Media.aspx (accessed 25 July 2018).

25.

Paulsen

(2013) Commentary: “Zoning restrictiveness and housing foreclosures: Exploring a new link in the subprime mortgage crisis”. Housing Policy Debate 23(2): 463–466.

26.

Pogodzinski

Sass

(1991) Measuring the effects of municipal zoning regulations: A survey. Urban Studies 28(4): 597–621.

27.

Portland Metro (n.d.) Regional Land-use Information system (RLIS). Available at: http://rlisdiscovery.oregonmetro.gov/ (accessed 25 July 2018).

28.

Saiz

(2010) The geographic determinants of housing supply. The Quarterly Journal of Economics 125(3): 1253–1296.

29.

Schuetz

(2008) Guarding the town walls: Mechanisms and motives for restricting multifamily housing in Massachusetts. Real Estate Economics 36(3): 555–586.

30.

Streiner

(2005) Finding our way: An introduction to path analysis. Canadian Journal of Psychiatry 50(2): 115–122.

31.

Talen

Knaap

(2003) Legalizing smart growth: An empirical study of land use regulation in Illinois. Journal of Planning Education and Research 22: 345–359.

32.

US Census Bureau (n.d.) Characteristics of new housing. Available at: https://www.census.gov/construction/chars/sold.html (accessed 11 April 2017).

33.

Zabel

Dalton

(2011) The impact of minimum lot size regulations on house prices in Eastern Massachusetts. Regional Science and Urban Economics 41: 571–583.