Exploring the synergistic economic benefit of enhancing neighbourhood bikeability and public transit accessibility based on real estate sale transactions

Abstract

Chinese

Planners and policy makers are increasingly promoting biking and public transit as viable means of transportation. The integration of bicycling and transit has been acknowledged as a strategy to increase the mode share of bicycling and the efficiency of public transit by solving the first- and last-mile problem. However, the economic outcomes of jointly promoting neighbourhood bikeability and transit accessibility are still poorly understood. This study aims to assess the property value impact of neighbourhood bikeability, transit accessibility, and their synergistic effect by analysing the single-family and condominium property sale transactions during 2010–2012 in Austin, Texas, USA. Our Cliff-Ord spatial hedonic modelling approach, which is also known as the general spatial model (or SAC), controls for the spatial dependent effects in the sale price and the error terms simultaneously. In order to quantify neighbourhood bikeability and transit accessibility, we use Bike Score and Transit Score as publicly available indices. We have assessed how residents’ willingness to pay (WTP) for bikeability and transit accessibility depend on various socio-demographic and built-environment factors, and whether the WTP is influenced by the bicycle-transit synergy. The results from this research show that jointly enhancing bikeability and transit accessibility can generate positive synergistic effects on property values. The effects would behoove policy makers to pursue the coordination of bicycle master plans with regional transit plans and to consider strategies of spatially-joint bicycle and transit investment.

Keywords

bikeability Bike Score economic benefit spatial hedonic models Transit Score

Introduction

Bicycling and public transit use have attracted increased attention among North American planners and policy makers, given the climbing rates of cycling and transit travel in the US and Canada in recent years (APTA, 2009; Statistics Canada, 2009). Efforts to coordinate bicycling and transit use have been reflected in the proliferation of bike sharing programs in cities such as Austin, Minneapolis and New York as well as public campaigns to encourage ‘bike and ride’ trips (DeMaio, 2009; Pucher et al., 2010). The environmental, health and congestion relief benefits associated with the two modes have been documented in some previous literature; strategic coordination of them has the potential to magnify these benefits by generating more bicycling and public transit use (Doolittle and Porter, 1994; Martens, 2004, 2007; Schneider, 2005; USDOT, 1998). Strategies to jointly develop bicycle and transit infrastructure will likely increase the catchment area of transit stations, promoting increased transit ridership and bicycle access and egress trips to and from stations (Krizek and Stonebraker, 2010; Pucher and Buehler, 2009). The integration of the two modes may help solve the first- and last-mile problem by improving access to transit stations, especially if transit stations and vehicles are adequately equipped with bicycle parking, bicycle racks on transit vehicles and bicycle sharing systems (Krizek and Stonebraker, 2010; Schneider, 2005; Wang and Liu, 2013). Additionally, bicycle-transit integration would provide the longer term benefits of promoting increased non-automobile transport options and achieving environmental, public health and social equity goals in North American cities (Bachand-Marleau et al., 2011; Krizek and Stonebraker, 2010; Wang and Liu, 2013).

While the multitude of benefits have been recognised by planners, the link between bicycling and transit, especially in terms of how one mode affects the other, is not well understood (Krizek and Stonebraker, 2010; Singleton and Clifton, 2014). The limited body of literature exploring the linkage between bicycling and transit has focused mainly on commuter trips using household surveys. In terms of economic benefits such as property values, only the proximity to transit stations has been adequately examined (Bowes and Ihlanfeldt, 2001; Hess and Almeida, 2007; Munoz-Raskin, 2010; Zhong and Li, 2016). The impact of bicycle infrastructure on property values has been largely limited to studies on separated bicycle paths and trails (Krizek, 2006; Parent and Vom Hofe, 2013). While residential location choice is influenced by a wide array of factors, there will likely be a greater demand for neighbourhoods that jointly promote bicycle and transit infrastructure, translating into increased premiums attributed to the synergistic impacts of bike-transit use. These types of neighbourhoods may be particularly attractive for young professionals who are increasingly shifting away from car use and embracing sustainable and active transportation modes (McDonald, 2015).

This study aims to assess the property value impact of neighbourhood bikeability, transit accessibility and their synergistic effect by analysing the single-family and condominium property sale transactions during 2010–2012 in Austin, Texas, USA. We employ the Cliff-Ord type spatial hedonic models (also known as the general spatial model, or SAC) in order to control for the spatial dependence effects and obtain unbiased estimates. Bike Score (Front Seat Management LLC, 2015a) and Transit Score (Front Seat Management LLC, 2015b) are used as indices to measure neighbourhood bikeability and transit accessibility, respectively. We have assessed how residents’ WTP for bikeability and transit accessibility depends on various socio-demographic and built-environment factors, and whether there is a synergistic effect of jointly promoting bicycling and transit. This study is a novel contribution to the literature by quantifying the property value premiums of bike-friendly and transit-orientated areas within the context of a diverse urban region. From a policy perspective, the results of this study have significant implications for identifying which areas should be targeted for bicycle and transit projects.

In the next section, we provide a review of literature relevant to the economic benefits of bicycling and transit use. We introduce our data and analytical methods in the following section, and then the interpret results. In the final section, we discuss the policy implications of our findings and provide concluding remarks.

Literature review

While there has been growing interest among planners during the past decade in jointing developing cycling and transit, relevant empirical studies still remain sparse, especially for North American cities. A key reason is the lack of reliable data about bicycle-to-transit facilities and their users, such as bike parking and the characteristics of bicyclists who ride transit (Schneider, 2005). Many US transit agencies still do not collect detailed data for this subset of travellers, which remain small compared to other transportation modes. Therefore, much of the existing literature to date has relied mostly on descriptive data and focuses on case studies from Western Europe where bicycling and transit are more prevalent (Martens, 2004, 2007; Pucher and Buehler, 2009, 2012).

However, the disconnect between transit and bicycling has been recently addressed through a handful of studies that focus on multimodal integration and the potential for bicycle-transit synergies (Singleton and Clifton, 2014). There have also been at least two research reports describing current best practices in integrating bicycle and transit (Doolittle and Porter, 1994; Schneider, 2005). While each of the two modes has been studied as discrete trips, few studies have explored how these two modes can be linked into a single multimodal trip. Additionally, few studies have examined how bicycling can influence transit use, and vice versa; nor have these studies explored the long-term socio-economic impacts of integrating cycling and transit. A more recent study by Welch et al. (2015) modelled the network access to bike and public transit facilities in Portland (Oregon, USA). They found that housing values may be positively impacted by increased proximity to bike lanes/multi-use paths, but can significantly benefit from increased proximity to light rail/street car stations.

Many more hedonic studies have examined the economic impact of transit access compared to neighbourhood bikeability. The literature is replete with studies examining the impact of the proximity to bus and rail transit stations on property values (Bartholomew and Ewing, 2011; Bowes and Ihlanfeldt, 2001; Cervero and Duncan, 2002a, 2002b; Duncan, 2011; Hess and Almeida, 2007; Munoz-Raskin, 2010; Pan et al., 2014; Rodríguez and Targa, 2004). These studies have generally showed a positive effect on property values, with rail stations showing a higher positive impact than bus stations.

The research examining bikeable communities has largely emphasised the environmental and public health aspects, while studies on the economic benefits have been somewhat limited in scope. The literature on the economic benefits of bicycling has explored the effect of proximity to off-street bike paths/trails and on-street bike lanes on property values using the hedonic pricing model (Krizek, 2006). A review of the literature shows mixed findings, with some studies associating proximity to multi-purpose trails with positive home values, such as Parent and Vom Hofe (2013), and others such as Krizek (2006) showing that homes adjacent to bicycle facilities, including both off-street bicycle trails and on-street bike lanes, are subject to a negative impact on home values particularly in suburban locations. However, these differences may be due to different analytical approaches as well as the type of bicycle facility: for example, the effects of spatial dependence were accounted for by Parent and Vom Hofe (2013) but not by Krizek (2006); Parent and Vom Hofe exclusively examined the impacts of off-street greenways and paths while Krizek compared both off-street and on-street facilities. Additionally, Nicholls and Crompton (2005) and Asabere and Huffman (2009) found positive impacts on property values for homes located near trails and greenbelts. We believe the reason for the scarcity of bicycle studies compared to transit studies is due to the lack of comprehensive bicycle data available.

In this study, we use the hedonic pricing model to measure the synergistic economic benefit of jointly promoting bicycle and transit use, which has not been examined in the literature. Our data and methodology allows us to address this important gap through a comprehensive analysis of sociodemographic and land use characteristics that influence bicycle and transit use in urban and suburban areas.

Methodology

Our primary motivation for this study is to examine how the recent trends in bicycle and transit infrastructure investments, as well as the increased emphasis on bike share and bike-and-ride programs, have been translated into market demand for neighbourhoods with excellent bicycle and public transit accessibility. These neighbourhoods, which generate societal benefits such as reduced household transportation costs and improved public health outcomes, are in limited supply in the US. Hypothetically, their bicycle and public transit accessibility could be capitalised into higher sale prices for homes, which could generate additional revenue from property taxes to finance bicycle and transit improvements.

Hence, we designed this study to answer the following three research questions:

Does increased neighbourhood bikeability benefit property values?

Does increased transit accessibility benefit property values?

Is there any synergistic effect of bicycling and transit on property values? We define the synergistic effect as an interaction term between bikeability and transit accessibility. Therefore this question can be broken down to two sub-questions: a) does bikeability have a larger impact on property values when transit is enhanced? b) does transit accessibility have a larger impact on property values when bikeability is enhanced?

We analysed both single-family and condominium housing types because previous research (Duncan, 2008; Li et al., 2015a) has shown that condominium residents may have different sociodemographic and economic characteristics and thus value environmental amenities differently from single-family housing residents.

Data

Housing and neighbourhood data

Our final housing sample consists of 3495 condominium transactions and 12,149 single-family transactions from January 2010 to November 2012 in Austin, Texas. The physical characteristics of housing, such as lot size, age and number of bedrooms and bathrooms, were extracted from the Multiple Listing Service (MLS) data obtained with permission from the Austin Board of Realtors. The original MLS data included 5062 condominium transactions and 26,107 single-family transactions; we truncated the top and bottom one percentile of the condominium/single-family sample to exclude extremely high or low sale prices, in addition to missing and mistyped values in any variable considered in the model (as listed in Tables 1 and 2). Additionally, as our spatial modelling work considers the interdependent relations among properties within the same neighbourhood, we excluded properties located in Census Block Groups where there were fewer than four sale transactions.

Table 1.

Summary statistics of the condominium housing sample.

Variable definition and unit	Mean	Std. dev.	Minimum	Maximum
Home sale price, US$1000	203.7541	127.1606	33.4000	830.0000
Structural characteristics
Living area, m²	103.3125	41.4288	32.0505	335.0903
Lot area, m²	585.5760	3708.5100	4.0469	58,719.8800
House age at the year of sale, year	21.2255	15.5619	0.0100	90
Number of days on market	114.2913	94.4857	1	1149
Number of bedrooms	1.8784	0.6883	1	5
Number of full bathrooms	1.6092	0.5534	1	4
Number of half bathrooms	0.2844	0.4550	0	2
Binary: 1 = has 2 or more stories	0.3505	0.4772	0	1
Binary: 1 = has 1 or more garage spaces	0.2844	0.4512	0	1
Binary: 1 = has pool	0.4847	0.4998	0	1
Binary: 1 = has view	0.4558	0.4981	0	1
Binary: 1 = located on the waterfront	0.0418	0.2001	0	1
Binary: 1 = large/medium tree on parcel	0.5296	0.4992	0	1
Neighbourhood variables
Network distance to nearest major road intersection, metre	865.6023	597.9949	2.5564	2983.5270
Direct distance to nearest lake, metre	599.6000	464.6736	4.3512	2299.5090
School quality score in the past year, 0–100	73.0953	10.6572	28	96
Binary: 1 = highway within 400 m	0.8192	0.3849	0	1
Binary: 1 = railroad within 400 m	0.1625	0.3690	0	1
Bike Score, 0–100	65.4836	21.9698	10	97
Transit Score,0–100	47.6724	15.7281	9	79
Covariates interacting with Bike Score and Transit Score
Total length of sidewalks within 1.6 km, km	146.9981	38.5548	40.5906	217.5880
Average speed limit within 1.6 km, km/hour	56.0457	4.4496	45.0001	67.2549
Percent of households with no car, 0–100	10.0323	9.1039	0.0000	52.8395
Percent of African Americans, 0–100	5.5162	5.9319	0.1149	46.0318
Percent of Asians, 0–100	6.4494	6.4719	0.1493	27.8404
Percent of non-White Hispanics, 0–100	22.9703	17.3382	4.3678	87.1594
Percent of other non-Whites, 0–100	2.3925	0.7340	0.5825	4.5293
Percent of under 18 years, 0–100	13.2463	9.2390	0.0929	41.3559
Percent of 65 years or older, 0–100	7.4849	6.1322	0.0000	33.8100
Per capita income, US$1000	40.4616	24.69782	1.736	108.099
Percent of college degree holders, 0–100	79.0232	17.6848	18.4009	100
Percent of population in poverty, 0–100	15.6139	19.3308	0	79.3770
Number of population per square km, 1000	2.4859	1.9404	0.0648	8.9500
Number of jobs per square km, 1000	2.5957	5.3005	0.0160	52.6010

Note: The sample includes 3495 condominium sale transactions during 2010–2012 in Austin, Texas.

Table 2.

Summary statistics of the single-family housing sample.

Variable definition and unit	Mean	Std. dev.	Minimum	Maximum
Home sale price, US$1000	266.0795	177.5109	63	1480
Structural characteristics
Living area, m²	164.1259	63.0057	33.1653	611.8394
Lot area, m²	868.9548	671.7190	161.8742	21,084.1200
House age at the year of sale, year	37.0553	22.0558	0.01	126
Number of days on market	96.6638	79.2003	1	1090
Number of bedrooms	3.1960	0.6994	1	11
Number of full bathrooms	1.9848	0.5969	1	6
Number of half bathrooms	0.2588	0.4566	0	9
Binary: 1 = has 2 or more stories	0.3032	0.4596	0	1
Binary: 1 = has 1 or more garage spaces	0.6213	0.4851	0	1
Binary: 1 = has pool	0.0561	0.2302	0	1
Binary: 1 = has view	0.1701	0.3757	0	1
Binary: 1 = located on the waterfront	0.0125	0.1112	0	1
Binary: 1 = large/medium tree on parcel	0.7829	0.4123	0	1
Neighbourhood variables
Network distance to nearest major road intersection, metre	1353.6700	683.7679	8.3796	4876.7780
Direct distance to nearest lake, metre	765.9550	527.4095	0.0100	2734.5620
School quality score in the past year, 0–100	69.0893	13.6921	28	96
Binary: 1 = highway within 400 m	0.6168	0.4862	0	1
Binary: 1 = railroad within 400 m	0.1536	0.3606	0	1
Bike Score, 0–100	49.2962	18.7480	2	98
Transit Score,0–100	35.9598	11.5292	6	78
Covariates interacting with Bike Score and Transit Score
Total length of sidewalks within 1.6 km, km	128.9428	32.7964	31.5596	220.2860
Average speed limit within 1.6 km, km/hour	53.8114	4.5745	43.7482	71.6800
Percent of households with no car, 0–100	5.8221	6.9728	0.0000	52.8395
Percent of African Americans, 0–100	6.6791	9.1535	0	62.7986
Percent of Asians, 0–100	3.8280	4.6708	0	27.6934
Percent of non-White Hispanics, 0–100	29.1343	20.1420	1.5649	89.2997
Percent of other non-Whites, 0–100	2.1969	0.7661	0.2899	4.1514
Percent of under 18 years, 0–100	20.8605	6.6096	1.0863	46.2216
Percent of 65 years or older, 0–100	9.4163	5.5418	0.3621	33.8698
Per capita income, US$1000	33.7784	16.9404	6.6320	131.4130
Percent of college degree holders, 0–100	71.1099	20.0209	4.5365	100.0000
Percent of population in poverty, 0–100	6.5872	7.3867	0.0000	69.3527
Number of population per square km, 1000	1.8098	0.8393	0.0056	8.9500
Number of jobs per square km, 1000	0.7355	0.8876	0.0045	19.8267

Note: The sample includes 12,149 single-family home sale transactions during 2010–2012 in Austin, Texas.

We obtained and merged datasets from numerous sources to capture neighbourhood characteristics. Geographic Information Systems (GIS) shapefiles¹ of road and railway networks were provided by ESRI Inc. and the city of Austin provided several shapefiles, including the water body polygons and sidewalk networks, as well as geocoded datasets of speed limit information, collision statistics and crime rates.² Our school performance data were extracted from the Texas Assessment of Knowledge and Skills database provided by the Texas Department of Education. Various socio-demographic variables, including race/ethnicity, age, income, education and poverty, were generated based on the 2007–2011 five-year estimates of the US Census Bureau American Community Survey, aggregated at the 2010 Census Block Group level; the data for population and employment densities were provided by the Capital Area Metropolitan Planning Organization, aggregated at the 2008 Traffic Analysis Zone level.

Measurement of neighbourhood bikeability and transit accessibility

We collected the Bike Score and Transit Score data from walkscore.com, as composite indices of neighbourhood bikeability and transit accessibility for each individual property transaction. Our hedonic models included the interaction terms of the two indices to quantify the synergistic effect of bicycling and transit on property values.

The Bike Score for a location is a number from 0 to 100 based on four equally weighted measures – bike lane scores, hill scores, destinations & connectivity and bicycle mode share – each of which is also normalised to a score from 0 to 100 (Front Seat Management LLC, 2015a). In addition to consultation with experts,³ the walkscore.com researchers created an online forum to gather input from the general public about what are the most important factors to include in a Bike Score calculation. The votes received for the four measures are among the highest supported factors.⁴

The bike lane score for a location was calculated by applying a distance decay function to each segment of bike lane within 1000 m from the location; off-street lanes’ value was twice that of on-street lanes. The hill score was calculated as the steepest grade within a 200 m radius from the location, based on the National Elevation Data from the United States Geological Survey; a grade of 10% to 2% was normalised to a score between 0 and 100. Destinations & connectivity was measured as the Street Smart Walk Score, whose algorithm considers the routing distances to the common destinations (e.g. grocery stores, parks and restaurants/bars) as well as the street connectivity and the average block length (Li et al., 2015a). The bicycle mode share was calculated as the bicycling commuting mode share in the US Census data. Depending on their Bike Scores, neighbourhoods are classified as Biker’s Paradise (90 ≤ Bike Score ≤ 100) where daily errands can be accomplished by biking, Very Bikeable (70 ≤ Bike Score ≤ 89) where biking is convenient for most trips, Bikeable (50 ≤ Bike Score ≤ 69) where some bike infrastructure is available, or Somewhat Bikeable (0 ≤ Bike Score ≤ 49) where bike infrastructure is minimal (Front Seat Management LLC, 2015a). Recently, Bike Score has been used by several researchers as a proxy for bikeability (Braun and Malizia, 2015; D’Angelo et al., 2015; Malizia and Song, 2016; Winters et al., 2015).

We use Transit Score developed by Front Seat Management LLC (2015b) to measure how well a location is served by transit. Similar to Bike Score, a Transit Score is also a normalised score from 0 to 100 based on the distance to the nearest stop of nearby routes, the frequency per week of the route and the type of route. A raw Transit Score for a location is the sum of the values of nearby routes, where a value is determined by the frequency per week weighted by mode (1 for bus, 2 for rail transit and 1.5 for ferry/cable car/other) and then adjusted based on a distance (to the nearest stop on the route) decay function; the routes where their nearest stop is beyond 1.6 km of the location generally have little effect on the score. To normalise the scores, researchers at Front Seat Management LLC (2015b) utilised the average raw score of the centre of five cities (San Francisco, Chicago, Boston, Portland and Washington, DC) as the benchmark full score of 100. Front Seat Management LLC (2015b) classifies neighbourhoods as Rider’s Paradise (90 ≤ Transit Score ≤ 100) where public transportation is ‘world-class’, Excellent Transit (70 ≤ Transit Score ≤ 89) where transit is convenient for most trips, Good Transit (50 ≤ Transit Score ≤ 69) where many nearby public transit options are available, Some Transit (25 ≤ Transit Score ≤ 49) where there a few nearby transit options, and Minimal Transit (0 ≤ Transit Score ≤ 24) where it is possible to get on a bus. Recent researchers have utilised Transit Score to measure transit accessibility (Beiler et al., 2015; Duncan et al., 2012; Nedwick and Burnett, 2015) and to predict travel behaviour (Hirsch et al., 2013).

Analytical methods

Hedonic pricing method

We employ the hedonic pricing method (Rosen, 1974) as our main analytical framework. Our starting model is the following:

P = β_{0} + S \cdot β_{S} + N \cdot β_{N} + X \cdot β_{X} + ε

(1)

where:

P is a vector of logarithm-transformed property values;

S and N are matrices of the housing structural and neighbourhood characteristics (see Tables 1 and 2 for specific variables) respectively; all continuous variables in S and N are also logarithm-transformed according to suggestions from the Box-Cox transformation tests.

$β_{S}$ and $β_{N}$ are vectors of coefficients associated with S and N; $β_{0}$ is a vector of the constant terms; and $ε$ is a vector of the error terms.

X is a matrix of bikeability (or transit accessibility) variable and the interaction terms associated with it, including the interaction term between bikeability and transit accessibility and the interaction terms between it and various environmental and socio-demographic variables. Each variable x in X was normalised as $\tilde{x} = \frac{x - \bar{x}}{\bar{x}}$ , where $\bar{x}$ is the sample mean of x; such a normalisation process was practiced by previous researchers (Anderson and West, 2006; Li et al., 2015b; Saphores and Li, 2012) to facilitate the interpretation of the coefficients, $β_{X}$ .

According to Rosen (1974), the price of a product could be explained by implicit prices of its various characteristics only under some strict assumptions, including continuum of products, market equilibrium, symmetric information between buyers and sellers and perfect competition. Our sample sizes for both the condominium and the single-family analysis are large enough so that we could generally assume the continuum of products. The market equilibrium could also be assumed as our period of analysis is less than two years during which there were no major market shocks; the symmetric information between buyers and sellers too could be assumed because of the professional services from realtors and inspectors and the information available from the internet. Bajari and Benkard (2005) argued that the perfect competition assumption is unnecessary because a hedonic pricing function could also produce valid estimates of implicit prices of products’ characteristics based on the demand side of a market which is not perfectly competitive.

To systematically investigate how the premiums for bikeability and transit accessibility depend on their synergistic effect, we considered the following four sets of models:

Model A1: $X \cdot β_{X}$ is defined as Bike Score and its interaction terms; the sample transactions are condominiums;

Model A2: $X \cdot β_{X}$ is defined as Bike Score and its interaction terms; the sample transactions are single-family houses;

Model B1: $X \cdot β_{X}$ is defined as Transit Score and its interaction terms; the sample transactions are condominiums;

Model B2: $X \cdot β_{X}$ is defined as Transit Score and its interaction terms; the sample transactions are single-family houses.

Spatial econometric modelling

After estimating Equation 1 with the Ordinary Least Squares (OLS) estimator for Models A1, A2, B1 and B2, we used the Moran’s I test suite developed by LeSage (2010) to examine the OLS residuals for evidences of spatial dependence, which is the inter-dependent relationship between sample observations located spatially close to each other. Previous researchers (Anselin, 1988; LeSage, 2010; Saphores and Li, 2012) provided both theoretical explanations and empirical evidence for the threat of spatial dependence to unbiasedness and consistency of estimates. In this study, the Moran’s I statistic was highly significant for all four sets of models. Then we performed the Lagrange Multiplier tests and their robust forms on these models with the software package developed by Lacombe (2013); we confirmed that the spatial dependence effects existed both in the property value variables and in the residuals. Following the theoretical guidelines (Anselin, 1988; Arraiz et al., 2010; Cliff and Ord, 1981) and published case studies (Li et al., 2015b; Saphores and Li, 2012), we adopted the spatial Cliff-Ord type model (also known as the general spatial model, or SAC) as follows:

{\begin{matrix} P = β_{0} + λ \cdot W \cdot P + S \cdot β_{S} + N \cdot β_{N} + X \cdot β_{X} + ε \\ ε = ρ \cdot W \cdot ε + e \end{matrix}

(2)

where:

W is a spatial weight matrix, which was constructed to illuminate the spatially dependent relation between sample observations. In our study, the matrix was set up so that the sale prices of neighbouring properties located in the same census block group had the same weight of influence on the value of a property;

e is a vector of adjusted residuals by excluding the spatial dependence effects;

all other notations in this equation are the same as Equation 1.

Other empirical considerations

There are several empirical econometric issues to consider in order to generate unbiased estimates with the hedonic pricing methods. As no consensus is achieved on the optimal functional form of the hedonic models, we applied the Box-Cox transformation test to a base model with key structural and neighbourhood characteristics to guide our decision; the lambda parameters were −0.005 and 0.209 for the condominium and single-family samples respectively. We decided to choose the log-log functional form for both samples. In order to mitigate the risk of multicollinearity among our explanatory variables, we excluded the bicycle and pedestrian rates from the explanatory variables; the Variance Inflation Factors (VIF) for all our models are under the commonly used threshold of 10 (Kutner et al., 2004). We compared results with and without these variables and found no major difference. Our spatial regression results were generated with the maximum likelihood estimator; these results were very similar to those estimated by the generalised spatial two-stage least-squares estimator with the heteroskedastic option; therefore, heteroscedasticity has little impact on our results.

Results

We estimated the spatial hedonic models with the SPPACK software package developed by Drukker et al. (2011). The spatial lag coefficients are highly significant for all models of both housing types except for the spatial lag coefficient for price in Model A1; these provide further evidence about the potential threat of spatial dependence. Our results were generated with the maximum likelihood estimator. For brevity, we do not interpret the estimates on the various housing structural and neighbourhood characteristics, the market-related variables including the time on market and the monthly dummy variables or on the constitutive terms used to generate the interaction terms. Generally speaking, these variables were associated with expected signs.

Premiums for bikeability and transit accessibility

The direct effects of Bike Score and Transit Score on property values are the linear operation of the coefficients presented in Table 3.

Table 3.

Main estimation results based on spatial regressions.

	Model A (Bike Score)				Model B (Transit Score)
	Model A1 (Condos)		Model A2 (S.F.)		Model B1 (Condos)		Model B2 (S.F.)
Variable names	Coefficients	Rob.SE	Coefficients	Rob.SE	Coefficients	Rob.SE	Coefficients	Rob.SE
Bikeability and Transit Access
Normalised Bike Score	0.2999***	0.0495	0.0279	0.0161
Normalised Transit Score					0.3946***	0.0609	0.0990***	0.0245
Norm. Bike Score X Norm. Transit Score	0.8122***	0.1525	0.0647	0.0459	0.9885***	0.1472	0.1035**	0.0457
Interaction terms: Normalised Bike/Transit
Score with normalised variables
Total length of sidewalks within 1.6 km, km	−0.3408	0.2093	0.3206***	0.0642	−1.1120***	0.2155	0.3375***	0.0745
Average speed limit within 1.6 km, km/hour	1.8258***	0.4715	0.3132**	0.135	−0.3155	0.5278	0.2313	0.1624
Percent of households with no car, 0–100	−0.2066***	0.0638	0.0386***	0.0148	−0.5589***	0.0871	0.0949***	0.0224
Percent of African Americans, 0–100	0.5517***	0.0482	0.0136	0.0142	0.7511***	0.0801	−0.0833***	0.0261
Percent of Asians, 0–100	0.1436	0.0737	−0.0312***	0.0118	0.0092	0.0921	0.0475**	0.0202
Percent of non-White Hispanics, 0–100	−0.6656***	0.1446	0.0589	0.0532	−0.3616**	0.1774	0.1785**	0.0833
Percent of other non-Whites, 0–100	−0.1265	0.1578	−0.0909	0.0518	0.3913	0.2334	0.0864	0.0847
Percent of under 18 years, 0–100	0.0816	0.1072	0.1782***	0.0683	0.2151	0.1338	0.0802	0.0993
Percent of 65 years or older, 0–100	0.0781	0.0725	−0.0346	0.0301	0.3340***	0.0918	−0.0325	0.0457
Per capita income, US$1000	0.0481	0.1155	−0.0119	0.0371	0.1398	0.1463	−0.0769	0.0567
Percent of college degree holders, 0–100	−0.7741	0.4466	0.5605***	0.1311	−1.0464	0.581	0.8663***	0.1795
Percent of population in poverty, 0–100	−0.1766**	0.0754	0.0252	0.0174	0.0679	0.0957	−0.0582**	0.024
Number of population per square km, 1000	0.1833**	0.0726	0.0748***	0.0252	0.3963***	0.0709	0.051	0.0314
Number of jobs per square km, 1000	0.0098	0.0087	−0.0071	0.0069	0.0168***	0.006	−0.0181**	0.0076
Number of observations	3495.00		12,149		3495		12,149
Spatial Lag Coefficient for Price	−0.0965	0.0802	−1.4256***	0.1221	−0.2388***	0.0849	−1.3785***	0.1243
Spatial Lag Coefficient for Error Term	0.8759***	0.0149	0.9449***	0.0038	0.8928***	0.0113	0.9439***	0.0040
Akaike Information Criterion (AIC)	−2370.86		−5583.57		−2346.12		−5624.34
Bayesian Information Criterion (BIC)	−1859.65		−4968.95		−1841.08		−5017.13

Note: ** p < 0.05, *** p < 0.01; The adjusted R-squares for the spatial models are not available; for the OLS model counterparts, they are 0.85 (A1), 0.80 (A2), 0.84 (A3) and 0.79 (A4); AIC and BIC are commonly-used indices comparing performances of different statistical models (Kuha, 2004).

Table 3, Model A presents coefficients associated with preferences on bikeability. The coefficients on the normalised Bike Score are 0.2999 (statistically significant) for the condominium sample and 0.0279 (0.05 < p < 0.1) for the single-family sample, respectively; these results show that, keeping all structural and neighbourhood characteristics constant and all other variables at their sample averages (we name such a condition as Condition α for the rest of this article), a one percent increase in Bike Score would increase condominium property values by 0.2999% and single-family property values by 0.0279%.

The premium for bikeability depends on various environmental and socio-demographic factors. Keeping Condition α, having a higher-than-sample-average speed limit within 1.6 km would significantly enhance the premium for bikeability; a higher Bike Score, which generally reflects enhanced safety factors, would become more desirable when higher traffic speed poses potential threat to the safety of cyclists. In areas where the population density is higher than the sample average, bikeability becomes significantly more valuable; higher population density may be associated with higher overall cost for automobile travel, therefore making biking more appealing. In areas where the proportion of young population is higher than the sample average, bikeability is significantly more desirable according to the single-family sample; this age factor seems to work in the same direction for the condominium sample, for which the effect is non-significant. We also found inconsistent preferences towards bikeability in the condominium and single-family residents, depending on some factors; for example, having a higher-than-sample-average proportion of carless households (or college degree holders) would not significantly influence the premium for bikeability according to the condominium sample but would significantly increase the premium according to the single-family sample.

Table 3, Model B presents coefficients relevant to the premium for transit accessibility. When keeping Condition α, a one percent increase in Transit Score would significantly increase condominium property values by 0.3946% and single-family property values by 0.0990%. Similarly, the preferences for transit accessibility also depend on various environmental and socio-demographic factors, most of which have different directions of influence between the condominium and the single-family samples. Transit accessibility would be valued by single-family residents significantly more if the neighbourhood sidewalk density, proportion of carless households, percentage of non-White Hispanic population or percentage of college degree holders was higher than their corresponding sample average; for the condominium residents, all these factors significantly decrease the valuation of transit. On the other hand, having a higher-than-sample-average job density significantly increases the premium for transit for the condominium residents, but significantly decreases the premium for the single-family residents.

The synergistic effect of bicycling and transit on property values

The coefficients on the interaction terms between Bike Score and Transit Score in Models A and B reveal the direct synergistic effects on property values. According to Model A, having a Transit Score higher than the sample average (47.67 for the condominium sample and 35.96 for the single-family sample) would enhance the premium for Bike Score. When the Transit Score doubles from the sample average, the premium for Bike Score will have a more than threefold increase for both housing markets: from the benchmark of 0.2999% (under Condition α) to 1.1121% for the condominium market; and from 0.0279% (under Condition α) to 0.0926% for the single-family market. On the other hand, Model B suggests that having a Bike Score higher than their sample average (65.48 for the condominium sample and 49.30 for the single-family sample) would enhance the premium for Transit Score for both housing samples. When the Bike Score is double the sample average, the premium for Transit Score will increase from 0.3946% (under Condition α) to 1.3831% for the condominium sample, which is also an increase exceeding threefold; for the single-family sample, the premium for Transit Score will more than double, from 0.0990% (under Condition α) to 0.2025%. These results have revealed strong preferences towards the joint development of bicycling and transit. Therefore, the efforts to design, plan and implement bicycling and transit systems in synthesis are rewarded economically, reflected in benefits to property values.

We can gain more insights into the revealed preferences for jointly developing and bicycling and transit from the total effects⁵ generated from our spatial hedonic models. Instead of being a point estimate, the total effects for Bike Score and Transit Score may vary depending on individual property records; therefore, they can be best presented as a distribution.

Figure 1, Panel A illustrates the distribution of the individual elasticities of property values with respect to Bike Score, over the space of Transit Score.⁶ The mean and median of the elasticities for the whole condominium sample are 0.3208 and 0.1480 respectively. As shown in Panel A1, the average elasticity for Bike Score is 0.1083 for condominium properties without good transit (Transit Score < 50) nearby; the average elasticity increases to 0.3506 for those served by Good Transit, and 1.2382 for those served by Excellent Transit. These averages are significantly different from each other. A similar pattern is observed for the single-family sample, as shown in Panel A2. The mean and median of the elasticities for the whole single-family sample are 0.0647 and 0.0210 respectively. The average elasticities are 0.0313, 0.3004 and 0.4821 respectively for the above three Transit Score levels sorted in ascending order. Again, these averages are significantly different between any pair from the three subsamples.

Figure 1.

Elasticities of price with respect to bikeability and transit accessibility.

Figure 1, Panel B illustrates the distribution of the individual elasticities of property values with respect to Transit Score, over the space of Bike Score. The mean and median of the Transit Score elasticities for the whole condominium sample are 0.3348 and 0.2109 respectively. As shown in Panel B1, the average Transit Score elasticity for condominium properties in Somewhat Bikeable neighbourhoods is 0.4148, which, surprisingly, is not significantly different from the Bikeable subsample; the average elasticities for the Very Bikeable and Biker’s Paradise subsamples are 0.2209 and 0.3356 respectively. On the other hand, the mean and median of Transit Score elasticities for the whole single-family sample are 0.1204 and 0.1344 respectively. As shown in Panel B2, a clear upward pattern is observed: the average Transit Score elasticities are 0.0039, 0.1533, 0.4141 and 0.5498 for the above four subsamples ordered ascendingly by bikeability levels.

We have converted the total effects introduced into dollar values (per individual property unit), as presented in Tables 4 and 5. On average, a one percent increase in bikeability can benefit an individual condominium (or single-family) property value by only US$67.59 (or US$173.28 for single-family) if good transit is not available nearby, compared to US$1030.77 (or US$1000.92 for single-family) if good transit is available and US$3900.46 (or US$2054.23 for single-family) if excellent transit is available. Similarly, on average a one percent increase in transit accessibility can benefit an individual condominium (or single-family) property value by US$509.18 (or US$88.86 for single-family) if minimal bike infrastructure is available, compared to US$1329.92 (or US$2080.32 for single-family) if the neighbourhood bikeability level allows daily errands to be accomplished by biking. We notice that even though the coefficients on Bike Score and Transit Score are higher for condominiums than for single-family houses (as shown in Table 3), there is no discernable pattern when comparing the average premiums in dollar values between the two property types.

Table 4.

Effect of a one percent increase in bikeability on property values.

Neighbourhood transit accessibility level	Condominiums		Single-family houses
Neighbourhood transit accessibility level	Actual dollar values	Percentage of average property value	Actual dollar values	Percentage of average property value
Transit Score < 50	US$67.59	0.1083%	US$173.28	0.0313%
50 ⩽ Transit Score ⩽ 69	US$1030.77	0.3506%	US$1000.92	0.3004%
70 ⩽ Transit Score ⩽ 89	US$3900.46	1.2382%	US$2054.23	0.4821%
90 ⩽ Transit Score ⩽ 100	N/A	N/A	N/A	N/A

Note: The dollar values in this table represent the average total effects per individual property. These values were calculated based on the average elasticities and property values for each subsample depending on different transit accessibility levels.

Table 5.

Effect of a one percent increase in transit accessibility on property values.

Neighbourhood bikeability level	Condominiums		Single-family houses
Neighbourhood bikeability level	Actual dollar values	Percentage of average property value	Actual dollar values	Percentage of average property value
Bike Score < 50	US$509.18	0.4148%	US$88.86	0.0039%
50 ⩽ Bike Score ⩽ 69	US$374.03	0.3911%	US$573.02	0.1513%
70 ⩽ Bike Score ⩽ 89	US$457.45	0.2209%	US$1565.60	0.4141%
90 ⩽ Bike Score ⩽ 100	US$1329.92	0.3356%	US$2080.32	0.5498%

Discussions and conclusions

The results from this research show that high-quality bicycle and transit investments have the potential to increase property values for both condominium and single-family housing markets. In neighbourhoods with good transit service or better (a Transit Score of 50 or higher), investing in bicycle infrastructure would yield a much greater payoff in terms of property values of both housing types compared to neighbourhoods that are not well-served by transit. However, the condominium market and the single-family market responded differently in terms of the impact of transit accessibility on property values: for condos, improving transit accessibility in neighbourhoods with the highest level of bikeability (a Bike Score of 90 or higher) yields a much higher increase in property values compared to neighbourhoods with lower levels of bikeability; for single-family homes, the threshold for Bike Score tends to be as low as 50. Despite the difference in thresholds, the results consistently show that the greatest appreciation in property values occurs in neighbourhoods with an extensive bicycle and transit network. This suggests that both condominium and single-family residents place a high value on bike friendly, transit-rich neighbourhoods. Due to the scarcity of relevant literature and our unique approach to measuring the revealed preferences on bicycle and transit networks, we are unable to directly compare our results with previous studies. However, our findings generally agree with several previous researchers who found that close proximity to bicycle infrastructure (Asabere and Huffman, 2009) and transit stations (Cervero and Duncan, 2002a; Pan et al., 2014) results in increases in property values. Our results are also consistent with the recent Portland (Oregon, USA) study by Welch et al. (2015) showing that increased accessibility to either bicycle facilities or transit stations may benefit housing values. The revealed preferences on bicycle and transit networks, as shown in our research, are also consistent with previous studies (Krizek and Stonebraker, 2010; Schneider, 2005; Wang and Liu, 2013) which discussed the positive effects of bicycle infrastructure near transit stations for solving the first- and last-mile problem.

Overall, our results show that jointly enhancing bikeability and transit accessibility can generate positive synergistic effects on property values. The effects would behoove policy makers to pursue the coordination of bicycle master plans with regional transit plans and consider strategies of spatially-joint bicycle and transit investment. Such plans and strategies are not only for economic benefits in terms of property values and tax revenues which could be used to make further improvements to bicycle and transit systems but also to promote increased public health, transportation options and social equity. While our results suggest that targeting bicycle investments in transit-rich neighbourhoods would generate higher property values, in some cases it may be necessary to invest in areas with lower transit accessibility to provide more non-automobile options or to achieve greater transportation equity.

Given the fiscal constraints facing municipal governments, strategic investment in bike/transit infrastructure is not only necessary but prudent to maximise the potential for increasing active travel. Previous studies have shown that bicycles and transit are complementary modes that could benefit from each other, and cities that jointly develop both modes enjoy higher levels of walking, bicycling and transit use. Improving the bicycle network around transit stations would reduce the burden of transfer between bicycle and transit modes, alleviating the first- and last-mile problem. Improving bicycle access to transit stations would greatly expand the catchment area of these stations which would lead to increased cycling and transit ridership. Our study has shown that residents value these benefits and place a high premium on neighbourhoods that have extensive bicycle and transit accessibility. Many people who value a high-quality bicycle network also value a high-quality transit network as it provides multimodal transportation options. Hence, investments that are made to integrate bicycle and transit, creating environmental and congestion relief benefits, may be partially recovered from increased property tax revenues.

It is important to note that while property value benefits from public investments in bicycle and transit infrastructure can help generate some additional property tax revenue to fund bicycle and transit services, they may not fully cover capital and operating costs. Therefore, cities are increasingly employing innovative strategies, such as tax increment financing and special assessment districts, to encourage private landholders to contribute to the finance of these services. Known as land value capture, these are methods of funding infrastructure improvements by recapturing some of the increased property value attributable to public infrastructure investment (Mathur and Smith, 2013; Medda, 2012; Smith and Gihring, 2006). These strategies have been used in several cities such as New York City and Chicago that have made recent transit improvements and expansions to their bicycle networks; these could also be used in Austin to supplement increased property tax revenue from bicycle/transit investments.

While our findings support policy measures to promote the investment of both bicycle and transit facilities, it is worth mentioning that the appreciation of property values also raises possible concerns about gentrification from a planning equity perspective. While bike share programs have been growing in popularity throughout the US and bicycle infrastructure in many US cities is rapidly expanding, there has been some neighbourhood opposition to bike lanes in cities such as Washington, DC and New York (Stein, 2011). Proposals for bike lanes in minority neighbourhoods in particular have often met opposition due to fears that investments in bicycle infrastructure may draw affluent young professionals into the community and displace existing residents; for example, such opposition was documented by Lubitow and Miller (2013) in their case study of North Williams Avenue, Portland (Oregon, USA). However, based on our empirical evidence, it is difficult to assess how bicycle and transit facilities impact housing affordability given that the housing market is influenced by many external factors. Nonetheless, this is an important concern that should be considered by planners and policy makers.

This study is subject to the following limitations. First, our measurement of neighbourhood bikeability and transit accessibility does not directly capture some specific infrastructure elements relevant to bicycle and transit trip makings, such as bicycle sharing, parking and on-board racks, even though they may positively facilitate the integration of the two modes; our measurement of transit accessibility does not directly incorporate how well various amenities can be reached via transit in terms of time or distance. Therefore, with the bikeability and transit interaction terms, we can only measure the general synergistic economic benefit; our analysis is unable to consider different types of trip activities (Krizek and Stonebraker, 2011): 1) bike-and-ride (cycling to the transit station, parking the bike and boarding the transit vehicle); 2) bike, ride-with-bike and bike (cycling to the transit station and bringing the bike aboard the transit vehicle); 3) ride-and-bike (cycling from the transit station by utilising a bike share facility at or near the transit station); or a combination of bike-and-ride and ride-and-bike (options 1 and 3). Second, results from this Austin-based case study may not be generalisable to other cities where the geographic, climatic and socio-demographic characteristics are different. Future studies may consider applying our analytical framework to other cities to gain insights on the revealed preferences on bicycle-transit integration; in addition, contingent valuation/ranking surveys may be conducted to assess the stated preferences on bicycle-transit integration; results from these surveys can be compared with the revealed preferences as documented in this study and shed light on how the specific infrastructure elements are valued by residents.

Footnotes

Acknowledgements

During data collection and processing, we received assistance from numerous colleagues, particularly Mr Han Park, Dr Ayoung Woo, Dr Tim Lomax, Dr Chanam Lee and Dr Jun-Hyun Kim; we are grateful for their assistance. The authors also thank the Austin Board of Realtors for providing the Multiple Listing Service data and for providing the Bike Score and Transit Score data.

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

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Anderson

S-T

West

S-E

(2006) Open space, residential property values, and spatial context. Regional Science and Urban Economics36(6): 773–789.

Anselin

(1988) Spatial Econometrics: Methods and Models. Dordrecht: Kluwer Academic Publishers.

APTA (2009) Transit Statistics. Available at: http://www.apta.com/research/stats/ (accessed 1 July 2015).

Arraiz

Drukker

Kelejian

et al . (2010) A spatial Cliff-Ord-type model with heteroskedastic innovations: Small and large sample results. Journal of Regional Science50(2): 592–614.

Asabere

Huffman

(2009) The relative impacts of trails and greenbelts on home price. The Journal of Real Estate Finance and Economics38(4): 408–419.

Bachand-Marleau

Larsen

El-Geneidy

(2011) Much-anticipated marriage of cycling and transit: How will it work?Transportation Research Record: Journal of the Transportation Research Board 2247: 109–117.

Bajari

Benkard

(2005) Demand estimation with heterogeneous consumers and unobserved product characteristics: A hedonic approach. Journal of Political Economy113(6): 1239–1275.

Bartholomew

Ewing

(2011) Hedonic price effects of pedestrian-and transit-oriented development. Journal of Planning Literature26(1): 18–34.

Beiler

Seketa

Swiatek

(2015) Investigating transit and pedestrian accessibility for age-restricted communities using spatial analysis. International Journal of Sustainable Transportation 10: 881–893.

10.

Bowes

Ihlanfeldt

(2001) Identifying the impacts of rail transit stations on residential property values. Journal of Urban Economics50(1): 1–25.

11.

Braun

Malizia

(2015) Downtown vibrancy influences public health and safety outcomes in urban counties. Journal of Transport & Health2(4): 540–548.

12.

Cervero

Duncan

(2002a) Benefits of proximity to rail on housing markets: Experiences in Santa Clara County. Journal of Public Transportation5(1): 1–18.

13.

Cervero

Duncan

(2002b) Transit’s value-added effects: Light and commuter rail services and commercial land values. Transportation Research Record: Journal of the Transportation Research Board1805: 8–15.

14.

Cliff

Ord

(1981) Spatial Processes: Models & Applications. London: Pion.

15.

D’Angelo

Evenson

Rose

et al . (2015) Examination of community and consumer nutrition, tobacco and physical activity environments at food and tobacco retail stores in three diverse North Carolina communities. Preventive Medicine Reports2: 730–736.

16.

DeMaio

(2009) Bike-sharing: History, impacts, models of provision, and future. Journal of Public Transportation12(4): 41–56.

17.

Doolittle

Jr Porter

(1994) Integration of Bicycles and Transit. Washington, DC: Transportation Research Board Transit Cooperative Research Program Synthesis 4.

18.

Drukker

Peng

Prucha

et al . (2011) SPPACK: Stata Module for Cross-Section Spatial-Autoregressive Models. Available at: http://ideas.repec.org/c/boc/bocode/s457245.html (accessed 1 July 2015).

19.

Duncan

Aldstadt

Whalen

et al . (2012) Validation of Walk Scores and Transit Scores for estimating neighborhood walkability and transit availability: A small-area analysis. GeoJournal78(2): 407–416.

20.

Duncan

(2008) Comparing rail transit capitalization benefits for single-family and condominium units in San Diego, California. Transportation Research Record: Journal of the Transportation Research Board 2067: 120–130.

21.

Duncan

(2011) The impact of transit-oriented development on housing prices in San Diego, CA. Urban Studies48(1): 101–127.

22.

ESRI Inc. (2015) ArcGIS Online Help – Shapefiles. Available at: https://doc.arcgis.com/en/arcgis-online/reference/shapefiles.htm (accessed 1 July 2015).

23.

Front Seat Management LLC (2015a) Bike Score^TM. Available at: https://www.walkscore.com/bike-score-methodology.shtml (accessed 1 July 2015).

24.

Front Seat Management LLC (2015b) Transit Score^®. Available at: https://www.walkscore.com/transit-score-methodology.shtml (accessed 1 July 2015).

25.

Hess

Almeida

(2007) Impact of proximity to light rail rapid transit on station-area property values in Buffalo, New York. Urban Studies44(5–6): 1041–1068.

26.

Hirsch

Moore

Evenson

et al . (2013) Walk Score® and Transit Score® and walking in the multi-ethnic study of atherosclerosis. American Journal of Preventive Medicine45(2): 158–166.

27.

Krizek

Stonebraker

(2011) Assessing options to enhance bicycle and transit integration. Transportation Research Record: Journal of the Transportation Research Board 2217: 162–167.

28.

Krizek

(2006) Two approaches to valuing some of bicycle facilities’ presumed benefits: Propose a session for the 2007 national planning conference in the city of brotherly love. Journal of the American Planning Association72(3): 309–320.

29.

Krizek

Stonebraker

(2010) Bicycling and transit: A marriage unrealized. Transportation Research Record: Journal of the Transportation Research Board2144: 161–167.

30.

Kuha

(2004) AIC and BIC comparisons of assumptions and performance. Sociological Methods and Research33: 188–229.

31.

Kutner

Nachtsheim

Neter

(2004) Applied Linear Regression Models. 4th edn.McGraw-Hill Education.

32.

Lacombe

(2013) Donald J. Lacombe’s MATLAB Code for Spatial Econometrics Page. Available at: http://community.wvu.edu/~djl041/matlab.html (accessed 1 July 2015).

33.

LeSage

(2010) Econometrics Toolbox. Available at: http://www.spatial-econometrics.com/ (accessed 1 July 2015).

34.

Joh

Lee

et al . (2015a) 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.

35.

Saphores

J-DM

Gillespie

(2015b) A comparison of the economic benefits of urban green spaces estimated with NDVI and with high-resolution land cover data. Landscape and Urban Planning133: 105–117.

36.

Lubitow

Miller

(2013) Contesting sustainability: Bikes, race, and politics in Portlandia. Environmental Justice6(4): 121–126.

37.

Malizia

Song

(2016) Does downtown office property perform better in live–work–play centers?Journal of Urbanism: International Research on Placemaking and Urban Sustainability 9(4): 372–387

38.

Martens

(2004) The bicycle as a feedering mode: Experiences from three European countries. Transportation Research Part D: Transport and Environment9(4): 281–294.

39.

Martens

(2007) Promoting bike-and-ride: The Dutch experience. Transportation Research Part A: Policy and Practice41(4): 326–338.

40.

Mathur

Smith

(2013) Land value capture to fund public transportation infrastructure: Examination of joint development projects’ revenue yield and stability. Transport Policy30: 327–335.

41.

McDonald

(2015) Are millennials really the “go-nowhere” generation?Journal of the American Planning Association81(2): 90–103.

42.

Medda

(2012) Land value capture finance for transport accessibility: A review. Journal of Transport Geography25: 154–161.

43.

Munoz-Raskin

(2010) Walking accessibility to bus rapid transit: Does it affect property values? The case of Bogotá, Colombia. Transport Policy17(2): 72–84.

44.

Nedwick

Burnett

(2015) How can the LIHTC program most effectively be used to provide affordable rental housing near transit?Cityscape17(2): 113.

45.

Nicholls

Crompton

(2005) The impact of greenways on property values: Evidence from Austin, Texas. Journal of Leisure Research37(3): 321–341.

46.

Pan

Zhang

et al . (2014) Effects of rail transit on residential property values: Comparison study on the rail transit lines in Houston, Texas, and Shanghai, China. Transportation Research Record: Journal of the Transportation Research Board2453: 118–127.

47.

Parent

Vom Hofe

(2013) Understanding the impact of trails on residential property values in the presence of spatial dependence. The Annals of Regional Science51(2): 355–375.

48.

Pucher

Buehler

(2009) Integrating bicycling and public transport in North America. Journal of Public Transportation12(3): 79–104.

49.

Pucher

Buehler

(2012) City Cycling. Cambridge, MA: MIT Press.

50.

Pucher

Dill

Handy

(2010) Infrastructure, programs, and policies to increase bicycling: An international review. Preventive Medicine50: S106–S125.

51.

Rodríguez

Targa

(2004) Value of accessibility to Bogotá’s bus rapid transit system. Transport Reviews24(5): 587–610.

52.

Rosen

(1974) Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy82: 34–55.

53.

Saphores

J-D

(2012) Estimating the value of urban green areas: A hedonic pricing analysis of the single family housing market in Los Angeles, CA. Landscape and Urban Planning104(3–4): 373–387.

54.

Schneider

(2005) Integration of Bicycles and Transit. Washington, DC: Transportation Research Board Transit Cooperative Research Program Synthesis 62.

55.

Singleton

Clifton

(2014) Exploring synergy in bicycle and transit use: Empirical evidence at two scales. Transportation Research Record: Journal of the Transportation Research Board2417: 92–102.

56.

Smith

Gihring

(2006) Financing transit systems through value capture. American Journal of Economics and Sociology65(3): 751–786.

57.

Statistics Canada (2009) Employed Labour Force by Means of Transportation. Available at: http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/labr88c-eng.htm (accessed 1 July 2015).

58.

Stein

(2011) Bike lanes and gentrification: New York City’s shades of green. Progressive Planning: The Magazine of Planners Network 188: 34–37.

59.

USDOT (1998) Bicycles and Transit: A Partnership that Works. Washington D.C.: US Department of Transportation, Federal Transit Administration.

60.

Wang

Liu

(2013) Bicycle-transit integration in the United States, 2001–2009. Journal of Public Transportation16(3): 95–119.

61.

Welch

Gehrke

Wang

(2015) A hedonic spatial panel approach to estimating the impact of network access to bike and public transit facilities on housing prices. Transportation Research Board 94th Annual Meeting. Washington D.C., January 2015.

62.

Winters

Sims-Gould

Franke

et al . (2015) “I grew up on a bike”: Cycling and older adults. Journal of Transport & Health2(1): 58–67.

63.

Zhong

(2016) Rail transit investment and property values: An old tale retold. Transport Policy 51(October 2016): 33–48.