Urban decline and residential preference: The effect of vacant lots on housing premiums

Abstract

Vacant land is a ubiquitous urban phenomenon. The existence of vacant land in a neighborhood can either lower or heighten nearby housing values, depending on its relative development potential. However, this condition has rarely been examined longitudinally, nor has it been examined thoroughly across different socioeconomic conditions. This research examines the impact of vacant lots on housing premiums using 2006–2015 single-family home sale transactions in the City of Minneapolis, Minnesota. The study area was divided into low-, middle-, and high-income levels. The results show that vacant lands have negative impacts on nearby single-family houses and these impacts differ by income level per neighborhood. The study sheds light on how planners and researchers should conceive vacant lands differently in various surroundings and conditions.

Keywords

Vacant land hedonic modeling urban decline depopulation socioeconomic conditions

Introduction

The ratio of vacant land (VL) to urban land in U.S. is increasing (Newman et al., 2016). Around 1/6 of urban land in U.S. cities is deemed vacant, on average, an area roughly the size of the Netherlands (Branas et al., 2011). With the global urban population projected to increase by more than three billion people between 2010 and 2050 (Buhaug and Urdal, 2013), cities have a vested interest redeveloping or managing VL. The questions of how to deal with VL and how its existence impacts cities and neighborhoods are ones that urban scholars have struggled to answer for more than half a century (Brophy and Vey, 2002). VL consists of both underutilized or unused properties, as well as lots with abandoned structures (Bowman and Pagano, 2004; Pagano and Bowman, 2000). Smaller proportions of VL can be an indicator of healthy economic growth in that they allow flexible space for cities to grow; excess amounts are typically considered symptoms of urban decline in that high VL to territorial size ratios are indicative of economic downturns, depopulation, and structural crisis (Accordino and Johnson, 2000).

Recent studies have shown that housing abandonment has surpassed deindustrialization as the primary driver of VL (Gu et al., 2019). Housing market issues (e.g. the 2008 recession) can result in intensifications of structural abandonment and vacancy rates. While primary larger-scaled factors contributing to VL increases include a lack of reinvestment into declining areas and deindustrialization (Newman et al., 2018), on a smaller scale, physical conditions such as the size, shape, and location of individual VLs (Lee et al., 2018) have also been shown to result in VL increases. A recent survey to large U.S. cities by Newman et al. (2016) showed that most VLs were reported to be small (70.7%), odd shaped (39.7%), and disconnected/in the wrong location (41.4%), making them difficult to regenerate. This condition presupposes that the spatial characteristics of VLs can be a significant reason for increases or decreases in a city’s VL supplies, yet this has not been examined in much of the current literature. Perhaps a reason for this lack of research is due to a lack of detailed spatial information in VL inventories. Only two-thirds of all cities (63%) reported having a VL inventory and only 13% had an inventory of abandoned structures (Newman et al., 2016). Simultaneously, longitudinal studies on VL are also limited due to this circumstance.

It has been shown that as VLs remain vacant for longer periods of time, they have a greater influence on converting adjacent active lots to inactive (Greenstein and Sungu-Eryilmaz, 2004). As duration of vacancy increases, VLs can begin to cluster, resulting in increased threats to long-term structural crisis and urban blight (Newman and Kim, 2017). Yet, we still know little in regards to the impacts that VLs can have on surrounding properties. The current increase in the amount of VL in U.S. cities, coupled with the lack of neighborhood-scaled research and longitudinal analyses of the vacancy condition, presents a dire need to better understand these situations. To help fill these gaps, this research assesses how the different conditions associated with VL impact the value of nearby single-family housing, examining this phenomenon across different types of neighborhoods with differing socioeconomic conditions. To assess this, a hedonic model is applied to longitudinal VL data in Minneapolis, Minnesota, USA.

Literature review

Causes and effects of VL

Decades of suburbanization and disinvestment have left many American cities with an overabundance of VL (Schilling and Logan, 2008). VLs can create significant problems for neighborhoods, with research showing VLs linked to reductions in property values and increased crime (Branas et al., 2011). The negative externalities accompanying VLs can further increase vacancy rates by encouraging remaining homeowners to leave declining neighborhoods, making VL a causal factor of itself; this process is known as the spillover effect. As VL accrues within an area, it can cluster, resulting in blight, decreased property values, and decreased tax base (Setterfield, 1997). This process can then create a circle of declining property values that drives further vacancy (Accordino and Johnson, 2000).

The presence of long-term VL disconnects communities, amplifies crime, and decreases quality of life (Kivell, 2002). Excessive amounts of long-term VLs can lead to widespread urban decline (Greenstein and Sungu-Eryilmaz, 2004; Zhang et al., 2015). Factors widely discussed as contributing to increases in VL and its increased duration include population loss (Rieniets, 2009), policy enforcement (Németh and Langhorst, 2014), local economy decline (Ryan, 2012), suburbanization (Audirac, 2007), and each parcel’s individual spatial characteristics (Henry et al., 2001). Neighborhoods with low social status are typically associated with longer-term vacancies (Immergluck, 2016). Relatedly, minority neighborhoods experiencing a high capacity of foreclosures tend to show longer-term vacancies than non-minority neighborhoods (Silverman et al., 2013). Ryan’s (2012) findings also suggest that neighborhoods with high socioeconomic status are less likely to confront problems with high VL amounts.

Hedonic modeling and VL: An overview

The location and characteristics of local-scaled VL can be a significant factor in determining the market price of a property (Park et al., 2016). In hedonic studies, various attributes of a property are employed to estimate the implicit prices of the characteristics of the property (Can, 1990; Rosen, 1974). Consumers’ preferences are revealed by deriving monetary values from the characteristics that differentiate each property among other closely related properties. Hedonic modeling has been found to be the most reliable method of establishing the implicit value of non-market attributes associated with residential properties (Des Rosiers et al., 1996). Thus, employing hedonic modeling in an attempt to examine impact of different attributes of VL can provide more reliable and explicit results that go beyond anecdotal evidence that much of the previous literature has found.

It is important to note that when evaluating potential biases in census-based hedonic valuation methods, Shultz and King (2001) found that VL best contributed to hedonic models at the track level or lower. Their results showed that smaller-scaled VL data created higher accuracies in models. More importantly, it also showed that, among all land-use variables, VL had a greater relative effect on housing values than did industrial or commercial land uses. Only a small amount of research has been conducted using hedonic modeling to assess the effects of VLs on urban areas, primarily due to a lack of both spatial and transactional data.

Hedonic modeling: Impacts of VL in high development potential areas

Most hedonic-based VL studies have been conducted in areas of high demand for development/redevelopment, viewing VL as a supply for future development rather than a threat to urban decline. These studies were conducted both as longitudinal (severely limited) and as single instance case studies (more abundant); most found that VL to be a significant uncertainty in regards to land pricing (Newman et al., 2017). It is generally assumed that the increased price uncertainty due to the presence of VL can delay the timing of new development (when in demand) and, therefore, can raise surrounding land prices; it has also been found that higher development potential VL is generally developed quicker than lower quality VL (Lopez and Arnott, 2018). Using parcel-scaled characteristics and real property transactions for Seattle, Washington, Cunningham (2006) tested this theory, finding evidence in support. Results showed that a one-standard-deviation increase in uncertainty lowered the likelihood of development by 11% and raised land prices by 1.6%. In a similar study, it was found that when VL is present and when controlling for a development ceiling and zoning restrictions, housing prices increase, resulting in developmental spillover (Pollakowski and Wachter, 1990).

Other, more historical-based studies, have also found that VL can catapult land prices when development potential increases in adjacent lots. For example, a study on the price of VL in New York City’s central business district in the 19th century reported that VL located within 2–3 miles of New York’s City Hall (used as a proxy for the Central Business District) increased more than fourfold in price between 1860 and 1870; VL prices were, on average, also more than six times higher between 1870 and 1900 than before this time period (Atack and Margo, 1998).

Non-U.S.-based research has shown similar results from hedonic models in areas of high development potential. Baba and Hino (2019) described the built environment factors leading to abandonment in Japan using a logistic regression model. The research found that the further away a property owner is from a vacant house, the greater the likelihood that that vacant house will become abandoned, long-term. Similarly, Bourne (2019) combined housing economic data to gain understanding of VL uses in England and Wales also using a logistic regression. The analysis found that VL tended to be found in the most and least affordable areas and that the probability of VL being more expensive than a regular home increases as affordability decreases and tourism increases. Further, when estimating the value of VL, the research found that an empty homes tax of 1% would generate the equivalent to 11% of the current council tax.

When analyzing 35 large- and medium-sized cities between 2004 and 2014, both high amounts of VL and high housing prices were shown to coexist in China’s real estate market (Zhang et al., 2015). An empirical panel analysis of VL and housing prices proved a significant correlation between the two variables. While this study did not separate VL by type, it was determined that many owners of VLs mothballed properties, waiting on the market to increase further before profiting on their turnover. These findings may not be typical of those found in depopulating cities or in areas with a low development potential, but the research does show the important role played by VL for adjusting real estate market prices, suggesting that it can be a primary driver of fluctuations in real estate prices.

Hedonic modeling: Impacts of VL in low development potential areas

As noted, development potential is not typically high in areas with heightened amounts of VL, especially within depopulating cities. Structural abandonment is the primary type of VL in many depopulating cities, with a majority of abandonment as residential land uses (Díaz et al., 2011). The ability to counteract abandonment can be dependent upon a city’s ability to afford the cost of razing unoccupied structures. Gedal and Ellen’s (2018) study on teardown (removal of abandoned structures) and VL sales analyzed approximately 3800 abandoned structure teardowns and 4900 VL sales occurring in New York City between 2003 and 2009, finding that a majority of VLs were disproportionately located in distressed (defined as those with low socioeconomic conditions) neighborhoods and were valued less highly than teardown parcels, even in similar neighborhoods. In some cities, regeneration of VLs has been addressed through the implementation of new public housing (Immergluck, 2016). Tach and Emory (2017) found that the redevelopment of VLs into public housing had significant direct and indirect spillover effects on neighborhood racial and economic composition; VLs in gentrifying neighborhoods were also found to be more likely to be filled by higher-income residents than vacant units in lower-income areas. Relatedly, Hollander et al. (2019) found that severely depopulating neighborhoods in Baltimore exhibited significant spatial clustering of VLs.

Some hedonic models have examined the effects of converting VLs into non-developmental land uses (e.g. green infrastructure) on property values. Housing proximity to urban green space has been continuously found to contribute to increased home sale prices (Kong et al., 2007; Nassauer and Raskin, 2014; Netusil et al., 2014; Noh and Rogers, 2016). Further, research using hedonic pricing models has continually reported that converting abandoned railways into greenways positively impacted sales prices of nearby single-family houses (Noh, 2019). Because of these and related findings, urban greening has been used as a management strategy to reduce the negative influence of VLs. VL is actually considered an open space in a fair amount of urban VL inventories. When classified as such, in many studies, VL was found to be the least valued type of open space and was shown to have a higher negative impact on land values than industrial land uses (Ready and Abdalla, 2005). A spatial difference-in-differences analysis of the economic impact of greening urban VL showed a significant impact on increases in housing prices near formerly VLs in Philadelphia, PA; properties surrounding greened VL had a greater increase in value than properties surrounding non-greened VLs (Heckert and Mennis, 2012).

Research objectives

As shown, evidence from several VL-based studies shows negative impacts on nearby properties, especially in declining cities (i.e. vacant lot clustering, long-term structure crisis and urban blight, negative spillover effects). VLs in declining cities possess a potential threat to cities, both economically and environmentally. However, there is no clear understanding on how different conditions of VLs impact neighborhoods with different income levels. The findings from these questions can lead to policy implications that can specifically aim to remediate declining or blighted neighborhoods considering the condition of the surrounding VLs. To explore these possibilities, this research asks: (1) How do the different conditions of VLs (i.e. size, distance, density, and distribution) impact the value of single-family houses? and (2) How do impacts differ between different neighborhoods of low-, middle-, and high-income? We focus on single-family houses for several reasons: (1) there are more number of transactions than other uses, (2) the preference for single-family houses well depicts how VLs are conceived by residents, and (3) the transaction data include both the structure and the lot and therefore provides an additional direction in interpreting the results.

Data and methods

Study area

In this study, we examine the City of Minneapolis, Minnesota, USA, to examine the impact of VL on single-family housing values. Previously a depopulating city, Minneapolis has a population of 411,452 with a recent population growth of 10.4% since 2010 (United States Census Bureau, 2011, 2019) and includes 11 communities with approximately 30,000 residents in each (City of Minneapolis, 2014). Within the communities, 84 neighborhoods are officially designated. As of 2017, there were 183,344 housing units with nearly 60% single-family household and 42.9% as single-family-detached (United States Census Bureau, 2019). The boundary of the study area was determined based on the jurisdiction of the city, which includes 366 Census Block Groups. Note that we used Census Block Groups as an operational neighborhood boundary to identify neighborhood-level attributes including socioeconomic characteristics and VL information. The official 84 neighborhoods were not specific enough to differentiate the characteristics around each residential sale transaction.

During the last decade, Minneapolis put intense efforts into examining land uses, providing an inventory of VL, and offering ways to revitalize VLs and abandoned structures. Annual VL information is collected by Hennepin County for taxation purposes. The year-end vacancy information, which refers to a property with no other official use, is publicly shared yearly since 2005 (Metropolitan Council of Minneapolis – St Paul, 2016). As of 2015, Minneapolis identified 5544 vacant properties (see Figure 1) and initiated the Vacant Housing Recycling Program to purchase, remodel, and resell VLs. The program evolved into the Minneapolis Homes Program to expand the scope, aiming to increase affordable housing opportunities and population diversity (City of Minneapolis, 2019).

Figure 1.

Study area and neighborhoods by income level.

Empirical data and data collection

In this research, the dependent variable is the sales prices of single-family houses in Minneapolis. The boundary includes 42,478 sales transactions in a 10-year period spanning from 2006 to 2015, annually. The residential sales transaction, structure data, and spatial data were gathered from the open Minneapolis web portal.¹ In the portal, two independent datasets, sales history data and assessor’s data, are linked. Key variables include structure area, number of bedrooms and bathrooms, construction material, exterior material, heating system, fire place, actual transaction amount, and date of sale. Other data include the location of airports, highways, cemeteries, industrial sites, major retails, parks, schools, golf courses, water bodies, and the downtown area. Locations of VLs were matched from the VL data from the Open Minneapolis web portal for each year.

Through preliminary analyses, we treated the original dataset and excluded all extremes. First, sales transaction prices from 2006 to 2015 were adjusted to 2015 market prices using the Housing Price Index from the Federal Housing Financing Agency. From the initial 42,478 single-family sales transactions, we excluded the top and bottom 1 percentile of the sales which did not reflect the ordinary market value (Li et al., 2015; Noh, 2019; Woo and Lee, 2016). As a result, 850 sales records with the prices lower than $73,693 and higher than $2,804,306 were removed from the analysis. Further, there were 3460 missing or mistyped values that could not be restored; 1370 for structural variables, 189 for neighborhood variables, and 1901 for sales records were excluded. After the preliminary regression, 260 cases with extreme residuals, caused by an unordinary number of bed and bath rooms or size of lots, were confirmed and removed. As a result, the final sample of the single-family housing sales transactions included 37,908 cases.

To analyze heterogeneity in the impact of the VLs in neighborhoods with different income levels, we conducted separate hedonic regressions for the sales in low-, middle-, and high-income neighborhoods. Based on the 2010 census data, block groups with median income less than 80% of the median household income of the study area were classified as low-income, between 80 and 120% as median income, and higher than 120% as high-income neighborhoods.² Among the 37,908 sales transactions, 33.1, 37.8, and 29.1% of the transactions belonged to low-, middle-, and high-income neighborhoods, respectively. Figure 1 shows the VLs and the distribution of neighborhoods by income level. Low-income neighborhoods are concentrated in the central and north-western areas while high-income neighborhoods are located on the south side of the city. Middle-income neighborhoods are in between low- and high-income areas with a concentration in the north-eastern portion of the city.

In this study, we include distance from the nearest VL (at parcel scale), size of the nearest VL, dummy for sale on VLs, distribution score, Getis-Ord Gi*³ of the block group, number of VLs in the block group, and average area of a VL in the block group. Including these variables allows us to understand the impact of the VLs to the residents in a more comprehensive manner (i.e. under what circumstances the residents favor or disfavor the VLs and how they are different among neighborhoods with different income levels).

For the housing attributes, size of house, number of rooms and bathrooms, construction material, exterior material, heating system, fire place, and age of the house at the time of sale are also introduced into the analysis.

For the locational variables, we include distance to amenities and dis-amenities such as parks, schools, airport, neighborhood commercial nodes,⁴ industrial site, golf courses, water bodies, central business districts, shopping malls, cemetery, and major highways using data provided by Open Minneapolis web portal. For neighborhood variables, income, unemployment, race/ethnicity, vacancy rate, and education level data at the block group level from the 2010 census are introduced to account for neighborhood differences. Lastly, time of the sale, sale year and sale quarter, was accounted in the regression.

Hedonic price model

Hedonic price modeling is based on a multiple regression that infers implicit prices of amenities and dis-amenities associated with a property by analyzing variation in housing prices while controlling for structural and neighborhood characteristics (Redfearn, 2009). It allows the ability to decompose values for attributes of properties in different dwellings in multiple places (Malpezzi, 2002). To examine the impact of VLs on housing premiums, the hedonic model employs the following attributes: (1) variables that represent the condition of each VL, (2) structural characteristics of existing buildings, (3) locational characteristics (i.e. distances to amenities and dis-amenities), and (4) neighborhood characteristics (i.e. sociodemographic characteristics such as race, education, age, etc.). The hedonic framework in this study is as follows

{{\log (price)}_{i} = α V_{i} + β S}_{i} + δ L_{i} + γ N_{i} + θ T_{i t} + ε

where V _i represents a vacancy related variable, S_i a structural variable, L_i a locational variable, N_i a neighborhood variable, and T_it a time variable;

α, β, δ, γ, and θ

are associated parameter vectors; and

ε

is a normally distributed random error associated with uncertainty in the measurement of variables, and unexplained variables (see Table 1).

Table 1.

Descriptive statistics.

Variable	Mean	Std. dev.	Min	Max
Log of sales value	12.44	0.50	11.21	14.91
Number of vacant lands in BG	8.01	12.80	0.00	115.00
Average area of vacant lands in BG (m²)	1129.75	2312.76	0.00	26,353.23
Getis-Ord Gi* z score of BG	0.15	0.81	−1.95	8.05
Distance from the nearest vacant land (1 km)	0.18	0.13	0.00	0.88
Size of the nearest vacant land (m²)	2871.03	16580.98	0.69	341,632.70
1 = sale data on vacant land	0.00	0.06	0.00	1.00
Area (m²)	5942.61	1887.46	0.00	45,320.00
Number of bathrooms	1.91	0.92	0.00	9.00
Number of bedrooms	3.07	0.90	0.00	11.00
Number of fireplaces	0.59	0.74	0.00	7.00
Building material_woodframe (dummy)	1.00	0.06	0.00	1.00
Exterior building material_stucco (dummy)	0.46	0.50	0.00	1.00
Exterior building material_metal (dummy)	0.26	0.44	0.00	1.00
Exterior building material_wood (dummy)	0.17	0.37	0.00	1.00
Age of house	58.86	22.60	0.00	114.00
Age of house_sq	5282.01	26428.24	0.00	12996
Heating (dummy)	0.26	0.44	0.00	1.00
Distance to airport (km)	4.80	2.60	0.01	10.87
Distance to highway (km)	0.91	0.57	0.01	2.89
Distance to cemetery (km)	1.64	0.99	0.02	4.89
Distance to commercial node (km)	0.56	0.44	0.00	2.92
Distance to industry (km)	1.32	0.92	0.00	4.28
Distance to major retail (km)	2.38	1.12	0.13	6.24
Distance to park (km)	0.29	0.18	0.00	1.21
Distance to school (km)	0.62	0.41	0.01	2.63
Distance to golf course (km)	1.84	0.87	0.00	4.67
Distance to water bodies (km)	0.70	0.47	0.02	2.84
Distance to downtown (km)	5.30	1.96	0.31	9.46
Percent of non-white	22.84	20.85	0.00	96.21
Percent of Hispanic	3.25	6.60	0.00	45.12
Percent of renter	26.51	22.87	0.00	100.00
Percent of population with college degree	52.78	20.83	0.00	91.50
Percent of population under poverty	10.25	11.02	0.00	84.13
Percent of population unemployed	5.75	5.32	0.00	34.30
Median income of household ($1000)	77.32	33.60	0.00	218.15
Percent of vacant housing units	5.45	6.39	0.00	39.05
Sale on first quarter (dummy)	0.17	0.37	0.00	1.00
Sale on second quarter (dummy)	0.33	0.47	0.00	1.00
Sale on third quarter (dummy)	0.30	0.46	0.00	1.00
Sale on year 2006 (dummy)	0.16	0.37	0.00	1.00
Sale on year 2007 (dummy)	0.11	0.32	0.00	1.00
Sale on year 2008 (dummy)	0.08	0.27	0.00	1.00
Sale on year 2009 (dummy)	0.09	0.28	0.00	1.00
Sale on year 2010 (dummy)	0.08	0.26	0.00	1.00
Sale on year 2011 (dummy)	0.06	0.25	0.00	1.00
Sale on year 2012 (dummy)	0.09	0.29	0.00	1.00
Sale on year 2013 (dummy)	0.10	0.30	0.00	1.00
Sale on year 2014 (dummy)	0.11	0.31	0.00	1.00

BG: block group.

Table 2.

Regression results: All sales transaction.

ln_sale_adjusted	Coef.	Std. err.	P > t
Number of VLs in BG	−0.00105	0.000139	0
Average area of VLs in BG (1000 m²)	−0.00195	0.000647	0.003
Getis-Ord Gi* z score of BG	−0.01438	0.001724	0
Distance from the nearest VL (1 km)	0.075231	0.011492	0
Size of the nearest VL (1000 m²)	0.000148	0.0000794	0.062
1 = sale data on VL	0.043667	0.020917	0.037
Structural variables Neighborhood variables Time variables	All complies with the previous literature
R²	0.7571

BG: block group; VL: vacant land.

In this study, we did not employ spatial hedonic modeling because the dataset consists of sales transactions over a 10-year period. This weakens the assumption of spatial autocorrelation for sales transactions that happened several years apart. It also eliminates repeated sales transactions in the dataset when we employ spatial models since each property can only hold one sales transaction data in the matrix. When we performed spatial models in the test runs, the results were similar to those without the treatment. For these reasons, we decided that hedonic models with different neighborhood levels were more appropriate for this study.

Results

As the aim of the study is to examine the impact of VLs, we focus on the interpretation of six key variables representing the condition of each VL. All structural, locational, and neighborhood variables correspond to previous literature.⁵ We confirm that there is no multicollinearity among explaining variable with the results of correlation and variance inflation factor.⁶

Hedonic modeling for all properties

In general, VLs show a negative impact on nearby single-family houses. As shown in Table 2, the number of VLs is negatively associated with the value of single-family homes at the block group level. When there is one more VL in the block group in the year a property was sold, single-family houses experience a 0.105% decrease in their value, holding other factors constant. Housing premium is also penalized when a VL is nearby. The coefficient for the distance from the nearest VL presents a positive value. This suggests that the farther the nearest VL from a sales property, the more the value of the property. When the nearest VL is located 100 meters closer to a single-family home, the housing value decreases by 0.75%.

In addition, the size of VLs is negatively associated with housing premiums. When average area of VL in the block group increases by 1000 square meters, the housing premium decreases by 0.195%. The result is also related to the hot and cold spot analysis from the Getis-Ord Gi*. The negative coefficient for the z score of the Getis-Ord Gi* indicates that when larger VLs are physically concentrated in a block group, it negatively impacts housing premiums. This finding suggests that larger VLs are generally considered dis-amenities in the neighborhood. However, when we interpret the coefficient for the size of the nearest VLs from each sale, the results indicate otherwise. The size of the nearest VL from each of the sold single-family houses is positively associated with the value of the properties. When the size of the nearest VL increases by 1000 square meters, the value of the property increases by 0.015%. Lastly, VLs had higher premium of 4.37% than occupied lands.

Spatial heterogeneity in the effects

Most of the structural, locational, and neighborhood attributes have similar coefficient values in each model.⁷ Before we examine the coefficients of variables related to the VLs, understanding the descriptive statistics for VL in each income level neighborhood provides a meaningful background for interpreting the related coefficients. Area of VL does not show a significant difference among each neighborhood type: 10.68% for low-income, 11.9% for middle-income, and 9.04% for high-income. In total, 10.24% of all parcels are considered as VL. However, average size and number of VLs show significant differences among the neighborhoods. There are a greater number of VLs in low-income neighborhoods, with a smaller average size. VLs in high-income neighborhoods are about three times larger than VLs in low-income neighborhoods. The average size of VLs in the middle-income neighborhoods is twice the size of the ones in the low-income neighborhoods. In terms of number of VLs, the order is opposite, low-income neighborhoods hold the most and high-income the least.

The regression models by different income levels present conflicting coefficients. Table 3 presents the regression results for low-, middle-, and high-income neighborhoods. In general, the model for the low-income neighborhoods has all the coefficients except for one, size of the nearest VL, showing to be statistically significant. Middle- and high-income models each have three coefficients (out of six) that are statistically significant. For middle-income neighborhoods, average area and distribution of VLs, and distance from the nearest VL were statistically significant. On the other hand, for high-income neighborhoods, coefficients of the variables (other than the three previously mentioned), number of VLs, size of the nearest VL, and sale on VL were statistically significant.

Table 3.

Regression results: Neighborhoods by income level.

	Low-income	Middle-income	High-income
ln_sale_adjusted	Coef.	Coef.	Coef.
Number of VLs in BG	−0.00171***	0.000337	0.00364***
Average area of VLs in BG (1000 m²)	−0.0147***	−0.00426***	0.000809
Getiz-Ord I* z score of BG	0.0203***	−0.0318***	Omitted
Distance from the nearest VL (1 km)	0.0924***	0.0842***	−0.0173
Size of the nearest VL (1000 m²)	−0.0000849	−0.000113	0.00156*
1 = sale data on VL	0.149***	0.00164	−0.140***
Structural variablesNeighborhood variablesTime variables	All complies with the previous literature
R²	0.6479	0.6663	0.7388

BG: block group; VL: vacant land.

***p < 0.01; *p < 0.10.

Table 4.

Vacant lands by neighborhoods.

Income level	Total area (m²)	VL areas (m²)	VL/Total (%)	No. of VLs	Average size of VL (m²)	No. of block group
Low-income(below −20%)	31,830,000(21.9%)	3,400,041(22.8%)	10.68	2340(42.2%)	1453	105(28.7%)
Middle-income (between −20 and 20%)	42,836,772(29.4%)	5,098,326(34.2%)	11.90	1783(32.2%)	2859	105(28.7%)
High-income (20% and over)	70,988,898(48.7%)	6,418,922(43.0%)	9.04	1421(25.6%)	4517	156(42.6%)
Total	145,655,670	14,917,289	10.24	5544	2690	366

VL: vacant land.

The number of VLs shows different impacts between low- and high-income neighborhoods (see Table 4). A greater number of VLs in each block group impacts housing premiums negatively for the low-income but positively for high-income neighborhoods. One additional VL within a block group decreases single-family housing premiums by 0.17% in low-income neighborhoods while it increases premiums by 0.36% in the high-income neighborhoods. To help interpret this conflicting result, we also examined the impact of the size of the nearest VL. For the high-income neighborhoods, when the size of nearest VL increases by 1000 square meters, there is 0.16% increase in the property value. For other neighborhoods, the coefficients are statistically insignificant. The results indicate that, in high-income neighborhoods, VLs near the properties are considered as a positive condition. A greater number of VLs with a larger size adds value to housing premiums in high-income neighborhoods. This is probably due to the fact that in high-income neighborhoods, a majority of the VLs are either greenfields or VLs in the subdivisions waiting to be occupied. VLs in high-income neighborhoods are typically larger in size and located in areas with high development potential. Undeveloped areas near affluent neighborhoods possess high development potential. Inversely, VLs in low-income neighborhoods are typically considered to be abandoned or deserted, rather than undeveloped. The coefficients for the sale on a VL support this premise. When a VL is on sale, it holds a 14.9% premium in low-income neighborhoods when holding other factors constant. In low-income neighborhoods, often times, the buildings are aged and may require demolition and maintenance before building a new structure. However, in high-income neighborhoods, a sale on a VL decreases the value by 14.0%. Undeveloped VLs require many improvements on areas such as the front and back yards, entry roads, and fencing, which are not accounted for in the hedonic model.

The concept of abandoned versus undeveloped land can be extended to the concept of developable lands when analyzing results in low- and middle-income neighborhoods. In both neighborhoods, average area of VL and proximity to the nearest VL are negatively associated with the value of single-family houses. When the average area of VLs in the block group increases by 1000 square meters, the value decreases by 1.47 and 0.43% in low- and middle-income neighborhoods, respectively. Likewise, being located 1 kilometer closer to the nearest VL decreases the value of a property by 9.24 and 8.42% in low- and middle-income neighborhoods, respectively. These results correspond to what is commonly accepted in the previous literature (Immergluck, 2016; Ryan, 2012; Silverman et al., 2013). Interestingly, in low-income neighborhoods, clustering of larger VLs positively impacts the value of the properties in the block group, whereas in middle-income neighborhoods, clustering of VL decreases housing premiums. In low-income neighborhoods, where the average size of VLs is relatively smaller than other neighborhoods, clustering of VLs may draw more development interest than other portions of the neighborhoods.

Discussion

This research determined how the different conditions of VLs impact the value of single-family houses and how these impacts differ between neighborhoods of different income levels. Overall, the existence of a VL tended to decrease neighborhood housing values when not separated by income level. Increases in both the number and amount of VLs showed to decrease housing premiums. This is consistent with a majority of the literature examining formerly or currently depopulating cities, such as Minneapolis. When considering the large amount of VLs in Minneapolis (5544), it is important to note that many of these are abandoned residential structures, which tend to negatively impact their surroundings more so than simply unused parcels with no buildings upon them.

Interestingly, our findings show that, when not separated by income levels, being in a closer proximity to a VL does not necessary result in a decrease in nearby individual housing values. This indicates that the mere presence of VL may be more important that what is located around it. The abundance of isolated, small, singular, and non-clustered VLs in Minneapolis could have played a large role in this outcome.

When analyzing the impact on neighborhoods by income level, the impact varies for each condition based on each individual VL. Consistent with much of the existing literature, lower-income neighborhoods tended to have more negative impacts from VL. In lower-income neighborhoods, when more VL exists, property values tend to decrease and larger-sized VLs tended to result in lower housing prices. When VLs cluster in lower-income neighborhoods, however, housing premiums tend to increase. This is a unique finding as it is incongruent with much of the literature in lower-income neighborhoods. This finding indicates a possible increase in the rate of gentrification in lower-income communities in Minneapolis. Large clusters of VLs are more easily developed in such neighborhoods. In most low-income neighborhoods, VLs tend to be relatively small, oddly shaped, and located at less preferred locations around poorly maintained streetscapes and infrastructure. In these conditions, VL types are typically abandoned structures or too small to possess much development potential. For this reason, larger VLs which are spatially clustered are more appreciated for their revitalization potential.

Inversely, this research finds that being closer to a VL increases housing price values in high-income neighborhoods, staying consistent with literature related to VL with high development potential. A greater number of and larger VLs actually added value to the neighborhood. In high-income neighborhoods, VLs are usually greenfields or subdivided lots waiting to be built out. This type of land possesses a potential for development which positively impacts nearby residential areas.

Only middle-income neighborhoods agreed with what much of the literature found in regards to the impact of the existence of high and low development potential VLs. Middle-income neighborhoods corresponded to what was expected for VLs, negatively impacting in terms of their size, clustering, and proximity. Also, when VLs cluster in middle-income neighborhoods, prices decrease significantly.

Conclusions

As noted, there have been relatively few studies examining the impact of VL in regards to the socioeconomic condition of neighboring residential properties. The study sheds light on how planners and researchers should conceive VLs differently in various surroundings and conditions. From the different effects of VLs on the residents’ preference in each income level related neighborhood, local governments can utilize the result to plan customized policies for each neighborhood when dealing with VLs. Lower-income neighborhoods should be targeted first when developing regeneration plans for cities, due to the high negative impacts created by VLs and their threat to eventual blight due to the spillover effect. Hierarchical approaches to regeneration should occur, beginning with clustered VLs, then addressing larger lots, then concluding with developing ways to repurpose small and disconnected unused parcels. A blend of permanent, more developmental based functions in large and clustered VLs and temporary and green infrastructure-based uses on smaller, unkempt parcels should be utilized to maximize regeneration potential.

While we have devised the methods as succinctly as possible to develop these findings, there are still a few minor limitations to this study. A cautious interpretation is encouraged with our results as the vacancy data are aggregated by the municipality under investigation for all properties with no use. Hennepin County collects the vacancy information for taxation purposes and this dataset includes all VLs including open land, abandoned properties, land held by public institutes, and small-sized vacancies that are difficult to be developed. The inability to differentiate VLs by type limits the interpretation of results at the block group level in regards the location and socioeconomic status around each vacant property by the assumption that all VLs are the same. Also, the information on vacancies, such as average VL sizes and vacancy clustering, was aggregated into the block group level. Obviously, aggregating data into a singular unit can limit the results in that a sample located at the further end of the block group boundary represents data at the same magnitude as a sample adjacent to an actual sales property. In this sense, the block group level was the smallest and most appropriate unit level, given the dataset. Shultz and King (2001) have also confirmed that land use related variables are best aggregated at block group level in hedonic studies.

Supplemental Material

sj-pdf-1-epb-10.1177_2399808320933906 - Supplemental material for Urban decline and residential preference: The effect of vacant lots on housing premiums

Supplemental material, sj-pdf-1-epb-10.1177_2399808320933906 for Urban decline and residential preference: The effect of vacant lots on housing premiums by Youngre Noh, Galen Newman and Ryun Jung Lee in Environment and Planning B: Urban Analytics and City Science

Footnotes

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

Youngre Noh

Notes

Youngre Noh is an Assistant Professor at Texas A&M University. He is also a member of American Institute of Certified Planners. His work aims to quantify the social equity, environmental, and housing implications of physical planning by using large data sources that includes longitudinal historic records. His publications include analyzing longitudinal impact of converting oil fields to nature preserves on nearby housing values, comparing different methodological measures for spillover effects of physical planning, and examining causality of the spillover effects.

Galen Newman is an Associate Professor in Landscape Architecture and Urban Planning and Director of the Center for Housing and Urban Development at Texas A&M University. He also serves as Coordinator of the Bachelor of Landscape Architecture Program Texas A&M University as well as Vice President of Research and Creative Scholarship for the Council of Educators in Landscape Architecture. He teaches undergraduate and graduate landscape architecture studios, landscape history classes, and advanced visualization courses. His research interests include urban regeneration, land use science, spatial analytics, community flood resilience, and community/urban scaled design.

Ryun Jung Lee is a lecturer in the Department of Landscape Architecture and Urban Planning at Texas A&M University. She has a PhD in Urban and Regional Sciences. Her research interests include land use planning, spatial analytics, and urban design, with a focus on urban vacant land and marginalized communities. She teaches courses in landscape history, planning theory, and design studios.

References

Accordino

Johnson

(2000) Addressing the vacant and abandoned property problem. Journal of Urban Affairs 22(3): 301–315.

Atack

Margo

(1998) “Location, location, location!” the price gradient for vacant urban land: New York, 1835 to 1900. The Journal of Real Estate Finance and Economics 16(2): 151–172.

Audirac

(2007) Urban shrinkage amid fast metropolitan growth (two faces of contemporary urbanism). Available at: http://www.coss.fsu.edu/durp/sites/coss.fsu.edu. durp/files/Audirac2009.pdf (accessed 25 September 2009).

Baba

Hino

(2019) Factors and tendencies of housing abandonment: An analysis of a survey of vacant houses in Kawaguchi City, Saitama. Japan Architectural Review 2(3): 367–375.

Bourne

(2019) Empty homes: Mapping the extent and value of low-use domestic property in England and Wales. Palgrave Communications 5(1): 1–14.

Bowman

Pagano

(2004) Terra Incognita: Vacant Land and Urban Strategies. Washington, DC: Georgetown University Press.

Branas

Cheney

MacDonald

, et al. (2011) A difference-in-differences analysis of health, safety, and greening vacant urban space. American Journal of Epidemiology 174(11): 1296–1306.

Brophy

Vey

(2002) Seizing City Assets: Ten Steps to Urban Land Reform. Washington, DC: Brookings Institution Center on Urban and Metropolitan Policy.

Buhaug

Urdal

(2013) An urbanization bomb? Population growth and social disorder in cities. Global Environmental Change 23(1): 1–10.

10.

Can

(1990) The measurement of neighborhood dynamics in urban house prices. Economic Geography 66(3): 254–272.

11.

City of Minneapolis (2014) Council wards. Available at: http://www.ci.minneapolis.mn.us/council/maps/index.htm (accessed 13 April 2019).

12.

City of Minneapolis (2019) MinneapolisHomes. Available at: http://www.minneapolismn.gov/cped/housing/MinneapolisHomesREHAB.

13.

Cunningham

(2006) House price uncertainty, timing of development, and vacant land prices: Evidence for real options in Seattle. Journal of Urban Economics 59(1): 1–31.

14.

Des Rosiers

Lagana

Thériault

, et al. (1996) Shopping centres and house values: An empirical investigation. Journal of Property Valuation and Investment 14(4): 41–62.

15.

Díaz

Nahuelhual

Echeverría

, et al. (2011) Drivers of land abandonment in Southern Chile and implications for landscape planning. Landscape and Urban Planning 99(3–4): 207–217.

16.

Gedal

Ellen

(2018) Valuing urban land: Comparing the use of teardown and vacant land sales. Regional Science and Urban Economics 70: 190–203.

17.

Getis

Ord

(2010) The analysis of spatial association by use of distance statistics. In: Anselin L., Rey S (eds) Perspectives on Spatial Data Analysis. Berlin, Heidelberg: Springer, pp.127–145.

18.

Greenstein

Sungu-Eryilmaz

(eds) (2004) Recycling the City: The Use and Reuse of Urban Land. Cambridge, MA: Lincoln Institute of Land Policy.

19.

Newman

Kim

, et al. (2019) Neighborhood decline and sustainable urbanism: Mitigating housing abandonment in shrinking cities. Land Use Policy 83(1): 505–5011.

20.

Heckert

Mennis

(2012) The economic impact of greening urban vacant land: A spatial difference-in-differences analysis. Environment and Planning A: Economy and Space 44(12): 3010–3027.

21.

Henry

Schmitt

Piguet

(2001) Spatial econometric models for simultaneous systems: Application to rural community growth in France. International Regional Science Review 24(2): 171–193.

22.

Hollander

Johnson

Drew

, et al. (2019) Changing urban form in a shrinking city. Environment and Planning B: Urban Analytics and City Science 46(5): 963–991.

23.

Immergluck

(2016) Examining changes in long‐term neighborhood housing vacancy during the 2011 to 2014 US national recovery. Journal of Urban Affairs 38(5): 607–622.

24.

Kivell

(2002) Land and the City: Patterns and Processes of Urban Change. London: Routledge.

25.

Kong

Yin

Nakagoshi

(2007) Using GIS and landscape metrics in the hedonic price modeling of the amenity value of urban green space: A case study in Jinan City, China. Landscape and Urban Planning 79(3–4): 240–252.

26.

Lee

Newman

Park

(2018) A comparison of vacancy dynamics between growing and shrinking cities using the land transformation model. Sustainability 10(5): 1513.

27.

Joh K, Lee C, et al. (2015) Assessing benefits of neighborhood walkability to single-family property values: A spatial hedonic study in Austin, Texas. Journal of Planning Education and Research 35(4): 471–488.

28.

Lopez JCG and Arnott RJ (2019) Is higher-quality land developed earlier? Environment and Planning B: Urban Analytics and City Science. First published online: February 15, 2019. https://doi.org/10.1177/2399808319828728

29.

Malpezzi

(2002) Hedonic pricing models: A selective and applied review. In: Gibb K., and O’Sullivan A. (eds) Housing Economics and Public Policy, pp.67–89. Oxford: Blackwell. https://doi.org/10.1002/9780470690680.ch5

30.

Metropolitan Council of Minneapolis – St Paul (2016) MetroGIS. Available at: https://www.metrogis.org/.

31.

Nassauer

Raskin

(2014) Urban vacancy and land use legacies: A frontier for urban ecological research, design, and planning. Landscape and Urban Planning 125: 245–253.

32.

Németh

Langhorst

(2014) Rethinking urban transformation: Temporary uses for vacant land. Cities 40: 143–150.

33.

Netusil

, et al. (2014) Valuing green infrastructure in Portland, Oregon. Landscape and Urban Planning 124: 14–21.

34.

Newman

Brody

Smith

(2017) Repurposing vacant land through landscape connectivity. Landscape Journal 36: 37–57.

35.

Newman

Hollander

Lee

, et al. (2018) Smarter shrinkage: A neighborhood-scaled rightsizing strategy based on land use dynamics. Journal of Geovisualization and Spatial Analysis 2(2): 11.

36.

Newman

Kim

(2017) Urban shrapnel: Spatial distribution of non-productive space in a growing city. Landscape Research 42(7): 699–715.

37.

Newman

Bowman

AOM

Jung Lee

, et al. (2016) A current inventory of vacant urban land in America. Journal of Urban Design 21(3): 302–319.

38.

Noh

(2019) Does converting abandoned railways to greenways impact neighboring housing prices? Landscape and Urban Planning 183: 157–166.

39.

Noh

Rogers

(2016) Recovery of the value of natural amenities: From oil fields to nature preserve. Land Use Policy 57: 11–22.

40.

Pagano

Bowman

AOM

(2000) Vacant Land in Cities: An Urban Resource. Washington, DC: Brookings Institution, Center on Urban and Metropolitan Policy, pp.1–9.

41.

Park

Huang

Newman

(2016) A statistical meta-analysis of the design components of new urbanism on housing prices. Journal of Planning Literature 31(4): 435–451.

42.

Pollakowski

Wachter

(1990) The effects of land-use constraints on housing prices. Land Economics 66(3): 315–324.

43.

Ready

Abdalla

(2005) The amenity and disamenity impacts of agriculture: Estimates from a hedonic pricing model. American Journal of Agricultural Economics 87(2): 314–326.

44.

Redfearn

(2009) How informative are average effects? Hedonic regression and amenity capitalization in complex urban housing markets. Regional Science and Urban Economics 39(3): 297–306.

45.

Rieniets

(2009) Shrinking cities: Causes and effects of urban population losses in the twentieth century. Nature and Culture 4(3): 231–254.

46.

Rosen

(1974) Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy 82(1): 34–55.

47.

Ryan

(2012) Design After Decline: How America Rebuilds Shrinking Cities. Philadelphia: University of Pennsylvania Press.

48.

Schilling

Logan

(2008) Greening the rust belt: A green infrastructure model for right sizing America’s shrinking cities. Journal of the American Planning Association 74(4): 451–466.

49.

Setterfield

(1997) Abandoned buildings: Models for legislative & enforcement reform. Hartford, CT: Trinity College, Trinity Center for Neighborhoods, Research Project 23, p.5.

50.

Shultz

King

(2001) The use of census data for hedonic price estimates of open-space amenities and land use. The Journal of Real Estate Finance and Economics 22(2–3): 239–252.

51.

Silverman

Yin

Patterson

(2013) Dawn of the dead city: An exploratory analysis of vacant addresses in Buffalo, NY 2008–2010. Journal of Urban Affairs 35(2): 131–152.

52.

Tach

Emory

(2017) Public housing redevelopment, neighborhood change, and the restructuring of urban inequality. American Journal of Sociology 123(3): 686–739.

53.

United States Census Bureau (2011) 2006–2010 American Community Survey. Available at: http://factfinder2.census.gov.

54.

United States Census Bureau (2019) 2013–2017 American Community Survey. Available at: http://factfinder2.census.gov.

55.

Van Zandt

Mhatre

(2009) Growing pains: Perpetuating inequality through the production of low-income housing in the Dallas/Fort worth metroplex. Urban Geography 30(5): 490–513.

56.

Woo

Lee

(2016) Illuminating the impacts of brownfield redevelopments on neighboring housing prices: Case of Cuyahoga county, Ohio in the US. Environment and Planning A: Economy and Space 48(6): 1107–1132.

57.

Zhang

Yuan

(2015) Panel data of China’s real estate prices and real estate vacancy – Based on data analysis of 35 large and medium size cities. In: 2015 International Conference on Management Science and Innovative Education, November, pp. 297–301. Paris: Atlantis Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.10 MB