Retail Rent with Respect to Distance from Light Rail Transit Stations in Dallas and Denver

47 A growing body of recent research is challenging the assumptions underlying the half-mile-circle 48 in planning for development around transit stations. In this article we review this literature and 49 extend it to include retail land uses. We estimate the rent premium conferred on retail properties 50 in metropolitan Dallas and metropolitan Denver, both of which have extensive light rail transit 51 systems. We find that consistent with half-mile-circle assumptions, retail rent premiums extend 52 only to about 0.30 mile from transit stations with half the premium dissipating after a few 53 hundred feet and three quarters within the first 0.10 mile. We offer implications for planners and 54 public officials. 55 56 Introduction 57 Much has been written about the association between rail transit and residential property values 58 but less has been written about the association with respect to other property values and nearly 59 none about the association with respect to retail property values. We help close this gap. We 60 begin with a literature review. This is followed by our research design, study area and data which 61 we apply to metropolitan Dallas and Denver. After presenting our results we offer implications 62 for planning and public policy. 63 64 Literature 65 The basic theory of urban economics (1, 2) can be summed up as follows: as a location becomes 66 more central to economic activity in a region, demand for such location increases and through 67 the market bidding process land becomes more valuable and development becomes more intense. 68 Central business district (CBD) location is an obvious example of this. Assuming Von Thünen’s 69 unfettered plain (3), land values and development intensity tend to fall at a declining rate from the 70 CBD. But areas outside CBDs can also enjoy accessibility advantages. This occurs when 71 transportation investments confer more efficient accessibility to non-CBD nodes than elsewhere 72 outside the CBD. The result should be more intensive development around those. Where those 73 investments are highways, congestion often follows thereby undermining efficiencies (4, 5). 74 Transit, as an “uncongestible” transportation option (6) can restore accessibility efficiencies, 75 leading to yet more intensive development. 76 77 But do all types of urban land uses react similarly transit station proximity? In an 78 important meta-analysis of studies through the middle 2000s, Debrezion, Pels and 79 Rietveld (7) identified variations between land uses. Like Bartholomew and Ewing (8), 80 they note that most studies of transit-station effects on property values address residential 81 property values and most of them address single family values – presumably because 82 data available for those properties are more plentiful than for other land uses. There were 83 about an equal number of studies on attached residential and office properties, but very 84 few for other land uses. We will not review the details of their findings except to observe 85 that, generally, the literature on industrial, hospitality (principally hotels), and retail 86 property value with respect to transit station proximity is small. 87 88 Generally, most prior studies have assume perhaps as an article of faith that urban land uses 89 will cluster mostly within the first one-quarter mile and a few out to about one-half mile. 90 Emerging research is challenging the half-mile-circle mantra. For instance, Petheram et al. (9) 91 found that apartments capitalized light rail transit station proximity to about 1.25 miles in Salt 92 Lake County, Utah. Ko and Cao (10) found office and industrial rent premiums with respect to 93 distance from the Hiawatha light rail transit stations in Hennepin County, Minnesota to extent 94 0.9 mile. For metropolitan Dallas and Denver, however, we (11) find the office rent premium to 95 extend about two miles from light rail transit stations though three-quarters of the premium 96 dissipates at about two-thirds mile. 97 98 We find only one relevant study estimating the rent premium on the association between 99 rail transit proximity and retail properties. Cervero and Duncan (12, 13) find that retail 100 land use value increases substantially within 200 feet of light and commuter rail transit 101 stations, perhaps 167 percent higher than distances beyond 200 feet in San Diego 102 County, California. Our study contributes to knowledge about whether and the extent to 103 which there is an association between retail land uses and, in particular, light rail transit 104 station proximity. We apply our inquiry to metropolitan Dallas and Denver. 105 106 Research Design, Study Area, Model and Variables 107 We extend work of others including Ko and Cao by evaluating the retail rent premium associated 108 with light rail transit station proximity in metropolitan Dallas and Denver. We chose those 109 systems for four reasons. First, they are among the oldest LRT systems in the US. The Dallas 110 Area Rapid Transit (DART) system began LRT service in 1996 while metropolitan Denver’s 111 Regional Transportation District began operating its FasTracks LRT in 1994. Only Portland’s 112 (1986), Sacramento’s (1987) and San Diego’s (1981) LRT systems are older. 113 114 Second, unlike Portland, Sacramento and San Diego, DART and FasTracks serve metropolitan 115 areas that are largely sprawling metropolises undeterred by terrain (the Rocky Mountains are 116 tens of miles away from downtown Denver) and policy (neither explicitly contains urban 117 development). 118 119 Third, they are among the nation’s largest LRT systems. In 2012, DART had 60 stations and 120 nearly 100,000 daily passengers while FasTracks had 46 stations and nearly 90,000 daily 121 passengers. 122 123 Fourth, their sheer size allow for sufficient data on office rents to undertake hedonic analysis (as 124 we discuss below). Indeed, our study area includes the central counties of Dallas and Denver as 125 well as Arapahoe and Jefferson counties in Colorado. It is thus the largest study area of any study 126 of its kind. 127 128 We employ the following hedonic model in our analysis: 129 130 Ri = f(Bi, Si, Ci, Li) 131 132 where: 133 134 R is the market rent per square foot for property i; 135 136 B is the set of building attributes of property i; 137 138 S is the set of socioeconomic characteristics of the vicinity of property i; 139 140 C is a composite measure of urban form of the vicinity of property i; and 141 142 L is a set of location attributes of property i. 143 144 Our dependent variable, R or rent per square foot, and independent variables comprising B, 145 building attributes, come from CoStar, with permission. Through proprietary access during fall 146 2012, we were able to collect an inventory of all retail structures within the study area including 147 their address, square feet, occupied and vacant space to derive the vacancy rate, stories, effective 148 age (by the later of the construction or renovation year), and weighted average contract rent per 149 square foot though we do not have lease terms for individual tenants. These variables include: 150 151 Socioeconomic data, S, come from either the 2010 census (for percent census tract population 152 that is not White non-Hispanic) or the 2012 5-year American Community Survey (for census 153 tract median household income). 154 155 C is a unique variable which measures urban form from most sprawled/diffused/disconnected to 156 most compact/integrated/connected at the level of the census tract. This index places urban 157 sprawl at one end of a continuous scale and compact development at the other. The original 158 index was developed in 2002 for metropolitan areas and counties (14, 15). In a recent study, the 159 compactness indices were refined and updated to 2010 for metropolitan areas, urbanized areas, 160 counties and census tracts and all are posted on a National Institutes of Health website (16). For 161 census tract indices, Ewing and Hamidi used the same methodology and the same type of 162 variables as in larger area analyses. They extracted principal components from multiple 163 correlated variables using principal component analysis and transformed the first principal 164 component to an index with the mean of 100 and a standard deviation of 25. Because the number 165 of component variables is greater for street accessibility than land-use mix, and greater for land166 use mix than development density, the resulting index gives more weight to street accessibility 167 than mix, and to mix than density. This is not unintentional, since the built environment-travel 168 literature suggests that density is the least important of the three D variable types (17). Given that 169 retail land uses that depend especially on accessibility this is an appropriate composite variable 170 to include. 171 172 Finally, L, the set of location variables, measures the distance of the centroid of each parcel to 173 the center of central business district of Dallas or Denver, the nearest entrance onto a limited 174 access highway and its quadratic term, and distance to the nearest LRT station and its quadratic 175 term. Distances are measured in miles. 176 177 Although the CoStar retail building database is the most comprehensive available from any 178 source, only about a quarter of the retail properties include rent. The reason is that most firms 179 either own the buildings they use and do not rent space to other tenants, or tenants have long180 term exclusive tenancy agreements with property owners. Nonetheless, with more than 700 retail 181 properties comprising more than 36 million square feet, we believe our analysis will reveal 182 central tendencies helping to clarify whether and the extent to which LRT station proximity 183 1 http://gis.cancer.gov/tools/urban-sprawl Accessed July 28, 2014. confers rent premiums on retail property. 184 185 Results 186 Table 1 reports results of lin