1# RECORD 8c46154fbaf164447e17b77d9874009c161a116a2# TITLE: The effect of light rail transit service on nearby property values: Quasi-experimental evidence from Seattle3# YEAR: 2018.0 VENUE: nan DOI: 10.5198/jtlu.2018.10964# AUTHORS: Michael R. Ransom5# PAGES: 18 CANDIDATE TABLE PAGES (1-based): [7, 10, 12, 15]6# ABSTRACT: This paper examined the effect of the construction of light rail transit stations on surrounding residential property values in Seattle, Washington. It studied sales of homes in the areas around the seven stations that serve primarily residential areas in the Rainier Valley, using a difference-in-differences regression technique to obtain plausibly causal estimates of the effect of new rail service. For these seven stations, the estimated impact of light rail service was positive for only one station and negative for two stations. Estimated impacts for the other stations were small and statistically insignificant. These results suggest that light rail service did not provide value to the neighborhoods in the Rainier Valley of Seattle. I speculate that the transport service provided by light rail was not a significant improvement relative to the bus lines that serviced the area before light rail was built.789===== PAGE 1 (score 10, numeric-paren 0, table-words 0, var-words 10) =====1011T J T L U http://jtlu.org12V. 11 N. 1 [2018] pp. 387–40413Abstract: This paper examined the effect of the construction of 14light rail transit stations on surrounding residential property values 15in Seattle, Washington. It studied sales of homes in the areas around 16the seven stations that serve primarily residential areas in the Rainier 17Valley, using a difference-in-differences regression technique to obtain 18plausibly causal estimates of the effect of new rail service. For these 19seven stations, the estimated impact of light rail service was positive 20for only one station and negative for two stations. Estimated impacts 21for the other stations were small and statistically insignificant. These 22results suggest that light rail service did not provide value to the neigh23borhoods in the Rainier Valley of Seattle. I speculate that the transport 24service provided by light rail was not a significant improvement rela25tive to the bus lines that serviced the area before light rail was built.261 27Introduction28This paper analyzes the impact of new light rail service on the price of residential property in Seattle, 29Washington in the United States. Light rail has been popular with local transit agencies in North Amer30ica. Since 1990, more than 20 cities in the United States have built new light rail lines or expanded ex31isting lines. Planners generally believe that these systems will improve transit access, reduce dependency 32on automobiles, and that stations will provide a focal point for high density development. However, 33the systems are very expensive to build and to operate. For example, the construction cost of the system 34studied here, Seattle’s Central Link “starter,” was $2.7 billion (Lindblom, 2016). Furthermore, in the 352014 fiscal year, operating costs for Seattle’s light rail exceeded fares by about $45.5 million (Federal 36Transit Authority, 2016), and fares covered only about 26 percent of operating costs.37Understanding the impact of new light rail service on housing prices is important for at least two 38reasons. First, it provides market based evidence of whether light rail provides improved transportation 39access compared to transit systems in place before light rail was built. If residents place value on the 40new transport option, this value should be capitalized into house prices where that access is available. 41Another reason to study the impact on property values is that many proponents of rail systems have 42proposed that new light rail systems might be financed with “value capture” schemes, whereby the tran43The effect of light rail transit service on nearby property values: 44Quasi-experimental evidence from Seattle45Michael R Ransom46Brigham Young University47ransom@byu.edu48Article history:49Received: November 7, 201650Received in revised form: August 5117, 201752Accepted: December 28, 201753Available online: May 16, 201854Copyright 2018 Michael R Ransom55http://dx.doi.org/10.5198/jtlu.2018.109656ISSN: 1938-7849 | Licensed under the Creative Commons Attribution – Noncommercial License 4.0 57The Journal of Transport and Land Use is the official journal of the World Society for Transport and Land Use (WSTLUR) 58and is published and sponsored by the University of Minnesota Center for Transportation Studies. This paper is also 59published with additional sponsorship from WSTLUR.606162===== PAGE 7 (score 110, numeric-paren 75, table-words 2, var-words 16) =====636439365The effect of light rail transit service on nearby property values66Table 1 presents summary statistics for the neighborhoods that are analyzed in this paper. Houses 67in most of these neighborhoods are generally comparable in terms of sale prices, size, and number of 68bathrooms. However, homes in the neighborhood around the Mount Baker stations sell for significantly 69more than those in other neighborhoods, and they are larger. Generally, houses near stations that are 70further from downtown Seattle are newer, have larger lots, and are less expensive.71Around 20 percent of sales took place after the opening of the light rail line, as indicated by the 72variable, “After.” The mean of the variable “Close” indicates the fraction of houses in the neighborhood 73that are within the 800-meter catchment basin of the corresponding rail station. In the Mt. Baker and 74Beacon Hill neighborhoods, a large fraction of houses (roughly half) is close to the station. On the other 75hand, in the SeaTac and Tukwila areas, only around 5 percent and 10 percent of houses in the neigh76borhoods fall within the respective catchment areas. This difference has important implications for the 77analysis. One common approach in this type of analysis is to compare all the “close” houses to all the 78“distant” houses, combining all stations. In the Seattle case, the comparison would be disproportionately 79Table 1: Summary statistics80Variable81Sea Tac 82Airport83Tukwila 84Int’l Blvd85Rainier 86Beach87Othello88Columbia 89City90Mount 91Baker92Beacon93Hill94Logarithm of 95Real Relative 96Sale Price9712.07598(0.305)9912.002100(0.293)10112.059102(0.338)10312.180104(0.401)10512.213106(0.379)10712.564108(0.470)10912.200110(0.315)111Sale Price 112223,408113(112,914)114206,744115(99,384)116224,515117(119,702)118268,564119(165,497)120270,963121(159,799)122394,869123(239,306)124261,237125(139,049)126Number 127Bathrooms1281.948129(0.85)1301.893131(0.878)1321.897133(0.872)1341.994135(0.945)1361.876137(0.896)1382.139139(0.949)1401.689141(0.818)142Size (ft2)1431,677144(641)1451,670146(669)1471,669148(642)1491,756150(717)1511,652152(636)1532,000154(761)1551,582156(594)157Age15839.077159(18.559)16043.262161(24.532)16248.325163(27.608)16450.391165(30.835)16657.325167(31.505)16865.668169(30.686)17067.305171(28.934)172Waterfront Lot1730.159174(0.366)1750.003176(0.055)1770.021178(0.144)1790.024180(0.154)1810.019182(0.137)1830.000184--1850.000186--187Outstanding 188View1890.070190(0.255)1910.058192(0.233)1930.161194(0.367)1950.142196(0.349)1970.152198(0.359)1990.301200(0.459)2010.120202(0.326)203Lot Size (acres)2040.241205(0.13)2060.253207(0.16)2080.174209(0.13)2100.146211(0.07)2120.125213(0.06)2140.114215(0.05)2160.106217(0.04)218Size Garage (ft2)219277.80 220(256.64)221221.11 222(251.83)223182.80 224(220.77)225168.15 226(211.12)227137.64 228(192.23)229121.01 230(175.96)23195.49 232(155.57)233Detached 234Garage?2350.161236(0.367)2370.250238(0.433)2390.205240(0.403)2410.193242(0.395)2430.234244(0.424)2450.289246(0.453)2470.286248(0.452)249Close (<1/2 250mile)2510.042 252(0.201)2530.103 254(0.304)2550.177 256(0.382)2570.293 258(0.455)2590.211 260(0.408)2610.416 262(0.493)2630.554 264(0.497)265After 20092660.179 267(0.383)2680.170 269(0.375)2700.179 271(0.383)2720.210 273(0.407)2740.205 275(0.404)2760.229 277(0.420)2780.214 279(0.411)280After×Close2810.007 282(0.082)2830.018 284(0.134)2850.029 286(0.169)2870.067 288(0.251)2890.047 290(0.212)2910.088 292(0.283)2930.108 294(0.311)295296297===== PAGE 8 (score 14, numeric-paren 0, table-words 1, var-words 12) =====298299394300JOURNAL OF TRANSPORT AND LAND USE 11.1301between the “close” houses in the Beacon Hill and Mt. Baker neighborhoods with the “distant” houses 302in the Tukwila and Sea Tac neighborhoods. A comparison of the houses between those two groups is 303likely to provide a biased estimate of the effect of light rail on housing prices. Thus, I adopt the approach 304of analyzing each neighborhood separately. This approach is also consistent with the idea that the value 305of a new transit option in a neighborhood is likely based on the location of that neighborhood relative 306to employment and commercial centers, existing commuting routes and patterns that naturally differ 307across neighborhoods, as well as the characteristics of the individuals who choose to live there.3084 309The hedonic pricing model and difference-in-differences estimation310The hedonic approach is based on an economic theory developed by Rosen (1974). The basic idea, as 311applied to housing markets, is that housing prices in a neighborhood reflect a general equilibrium in 312the supply and demand of housing services. In the hedonic model, each of the characteristics of a house 313provides some benefit, and thus each of these components has a price. Some of these characteristics rep314resent elements of the house itself, such as the size, age, or style of the house. Others represent locational 315characteristics, such as pleasant views, good neighborhood schools, or convenient access to downtown 316labor markets. To the extent that a new light rail line provides faster or more comfortable connections to 317areas of the city that residents want to visit, this should be reflected in housing prices. Those who most 318value the transit services provided by light rail will be willing to bid more for houses that are close to 319light rail stations. On the other hand, a rail line may also detract from housing values by creating noise, 320local traffic, or increasing neighborhood crime. In Seattle, light rail replaces or supplements bus services 321that were already available in these neighborhoods. It is possible that light rail service is inferior to exist322ing bus services, as explained in Richmond (2001), who discusses how reconfiguring the transit system 323around a rail line may impose costs on riders. 324 325The vast majority of the work that studies the impact of transit services on housing prices relies 326on the standard hedonic framework. Most of the work utilizes cross-sectional comparisons, which are 327likely to lead to bias because of the omission of important variables. The concern in the Seattle case 328focuses on the characteristics of the station neighborhoods. To the extent that the locations of the sta329tions are different in ways that cannot be measured with available data on housing characteristics, simple 330cross-sectional comparison of prices will suffer from omitted variable bias. For example, the ability to 331economically acquire right-of-way for the rail line dictated its exact location. Were the areas thus chosen 332for stations less desirable that other locations in the neighborhoods that the line serves? 333One way to deal with the problem of omitted variables is to use a quasi-experimental approach, 334such as difference-in-differences estimation. In this case, difference-in-differences compares the changes 335in housing prices in the areas close to stations (within the catchment area) with other houses in the same 336neighborhoods that are further away. This is analogous to an experiment—it compares how initiation of 337rail service affects the price of houses that are well-served by rail (the “experimental group”) with houses 338in the same neighborhoods that are not well-served (the “control group”). This implicitly controls for 339unobserved characteristics, since the analysis focusses only on changes in the relative prices. In this pa340per, the analysis follows closely the methods in Pope and Pope (2015) which are discussed in detail in 341Parmeter and Pope (2012).342I implement the above ideas in a simple regression model that describes the logarithm of the real 343relative sale price, as follows:344ln(pi* )= α0+α1 Closei+α2 Afteri+α3 Close×Afteri+β’ Xi+εi 345 346 347(1)348Where pi* represents the real relative price of the house that sold in transaction i. Close is an indi349350351===== PAGE 10 (score 218, numeric-paren 169, table-words 6, var-words 16) =====352353396354JOURNAL OF TRANSPORT AND LAND USE 11.1355Table 2 presents estimates of the latter comparison. That is, it compares sales that took place after 356December 2009 relative to those that took place before January 2009. This table reports an estimate of 357the proximate effect of light rail service without considering the possible forward-looking responses to 358the announcement of the system by buyers and sellers. The hedonic variables included are: number of 359bathrooms, size of the house and its square (in square feet), age of the house and its square (in years), 360acreage of lot and its square, size of garage (in square feet), if the garage is detached, whether the lot is 361located on a waterfront, and whether the lot has outstanding views. 362The coefficient on Close indicates the relative value of parcels close to stations before 2009. Thus, 363except for the Beacon Hill catchment area, houses close to light rail stations sold at a discount compared 364Table 2: Difference-in-Differences Regression Estimates of Effect of Opening of Light Rail Differences compare years before 365opening (1992-2008) with those after (2010-2016) (Dependent variable is logarithm of relative real sale price)366Variable367Beacon Hill368Mount369Baker370Columbia371City372Othello373Ranier374Beach375Tukwila/376Intnl Blvd377SeaTac378Airport379Close (within ½ 380mile of station)3810.0384**382(0.0136)383-0.1298***384(0.0148)385-0.1048***386(0.0096)387-0.0828***388(0.0095)389-0.0031390(0.0085)391-0.0106392(0.0096)393-0.0222394(0.0154)395After (sale after 3962009)3970.2031***398(0.0206)3990.1263***400(0.0175)4010.1645***402(0.0098)4030.1199***404(0.0111)405-0.0614***406(0.0085)407-0.1539***408(0.0077)409-0.1051***410(0.0082)411Close x After412-0.0256413(0.0285)414-0.0069415(0.0284)4160.1019***417(0.0203)4180.0932***419(0.0196)420-0.0116421(0.0203)422-0.0060423(0.0226)424-0.1347***425(0.0367)426Baths4270.0285**428(0.0107)4290.0565***430(0.0091)4310.0204**432(0.0064)4330.0347***434(0.0067)4350.0241***436(0.00554370.0185***438(0.0053)4390.0192***440(0.0057)441Size (ft2)4420.00035***443(0.00004)4440.00048***445(0.00002)4460.00030***447(0.00002)4480.00031***449(0.00002)4500.000431***451(0.000020)4520.00030***453(0.00002)4540.00031***455(0.00002)456Size2457-5.23e-08***458(1.15e-08)459-3.44e-08***460(3.17e-09)4612.03e-08***462(5.53e-09)463-2.15e-08***464(2.77e-09)465-5.10e-08***466(4.76e-09)467-2.37e-08***468(3.13e-09)469-1.94e-08***470(3.68e-09)471Age472-0.00593***473(0.000890)474-0.00213**475(0.000821)476-0.00509***477(0.000474)478-0.00484***479(0.000497)480-0.004017***481(0.000390)482-0.006067***483(0.000376)484-0.006522***485(0.000495)486Age24875.00e-05***488(7.49e-06)4893.40e-05***490(7.43e-06)4914.25e-05***492(4.07e-06)4933.81e-05***494(4.36e-06)4953.56e-05***496(3.48e-06)4973.86e-05***498(3.79e-06)4994.83e-05***500(6.92e-06)501Waterfront502--5030.6723***504(0.0712)5050.2902***506(0.0253)5070.2269***508(0.0234)5090.4724***510(0.0201)5110.0372512(0.0498)5130.0584***514(0.0077)515View5160.0509**517(0.0181)5180.1599***519(0.0142)5200.2325***521(0.0103)5220.2869***523(0.0124)5240.2198***525(0.0087)5260.0611***527(0.0119)5280.1484***529(0.0112)530Lot (acres)5310.0509**532(0.0181)5331.3753***534(0.1827)5350.8103***536(0.0983)5370.8986***538(0.0824)539-0.0818*540(0.0355)5410.2000***542(0.0267)5430.5928***544(0.0494)545Lot25463.0950547(1.8734)548-0.8996***549(0.1645)550-0.2619**551(0.0904)552-0.1283*553(0.0597)5540.0322**555(0.0099)556-0.0064557(0.0114)558-0.2006***559(0.0319)560Garage (sf)5610.00033***562(0.00005)5630.00022***564(0.000045)5650.00021***566(0.00003)5670.00020***568(0.00002)5690.000193***570(0.000018)5710.000171***572(0.000014)5730.000172***574(0.000014)575Detached 576Garage5770.0775***578(0.0140)5790.1413***580(0.0149)5810.0701***582(0.0087)5830.0420***584(0.0102)5850.0784***586(0.0079)5870.0480***588(0.0069)5890.0730***590(0.0085)591Constant59211.7027***593(0.0537)59411.3914***595(0.0366)59611.6696***597(0.0278)59811.5490***599(0.0244)60011.4713***601(0.0236)60211.6384***603(0.0195)60411.5527***605(0.0210)606R26070.386080.696090.526100.666110.576120.636130.70614N6151,8516162,8716176,0156184,5066195,9216204,6736213,817622Notes: * p<0.05; ** p<0.01; *** p<0.001623624625===== PAGE 11 (score 14, numeric-paren 0, table-words 3, var-words 8) =====626627397628The effect of light rail transit service on nearby property values629houses further away. In some cases, these estimated differences are large and statistically significant. For 630example, in the Mt. Baker neighborhood, houses close the eventual location of the rail station sold for 63113 percent less than houses further away, on average. 632The coefficient on After indicates the relative growth rate of houses in the analysis neighborhoods, 633compared to houses in the Seattle metro area generally. (Recall that the dependent variable is the loga634rithm of the sale price of a house deflated by the metro-wide Case-Schiller price index.) These estimates 635indicate that relative to prices in all of Seattle, housing prices have increased more rapidly in the analysis 636neighborhoods that are closer to downtown Seattle—Othello, Columbia City, Mt. Baker, and Beacon 637Hill. Generally, the increase has been larger for neighborhoods that are closer to downtown. For neigh638borhoods that are most distant from downtown, the relative prices have fallen.639The crucial coefficient for the analysis is the interaction variable—Close×After. This shows how 640housing prices changed with the opening of the rail system, compared to the changes of prices for houses 641in the same neighborhood that are more distant from the stations. For all but one of the neighborhoods, 642the estimate of the effect of light rail is negative. For Sea Tac and Othello, these changes are also statisti643cally significantly different from zero. However, for the Columbia City neighborhood, the estimated 644impact of the initiation of rail service is positive, quite large, and statistically significant. 645646647===== PAGE 12 (score 207, numeric-paren 168, table-words 5, var-words 8) =====648649398650JOURNAL OF TRANSPORT AND LAND USE 11.1651Table 3 reports estimates using the earlier date window, comparing sales before January 2004 with 652those that took place after December 2009. The estimates from this analysis are very similar, both quali653tatively and quantitatively, to the previous analysis reported in Table 2. In this case, the sample sizes are 654somewhat smaller, since the analysis excludes sales that took place between January 2004 and December 6552009. 656As a test for sensitivity, I have also examined slightly different time periods. For example, I have 657estimated the model comparing the period between 2003 and 2007 with the period from 2011 through 6582015, as well as comparing the period from 1999-2003 with the period from 2011-2015. Results are 659qualitatively very similar. Thus, the results to not appear to be sensitive to the particular definitions of 660“before” and “after.” 661Table 3: Difference-in-Differences Regression Estimates Difference compares years before construction (1992-2003) with 662those after opening (2010-2016) (Dependent variable is logarithm of relative real sale price)663Variable664Beacon665Mount666Baker667Columbia668City669Othello670Ranier671Beach672Tukwila/673Intnl Blvd674SeaTac675Airport676Close6770.0394*678(0.0159)679-0.1449***680(0.0174)681-0.1234***682(0.0113)683-0.0907***684(0.0119)685-0.0098686(0.0104)687-0.0050688(0.0119)689-0.0206690(0.0182)691After6920.2474***693(0.0217)6940.1514***695(0.0187)6960.2074***697(0.0103)6980.1696***699(0.0120)700-0.0273**701(0.0092)702-0.1298***703(0.0084)704-0.0985***705(0.0092)706Close x After707-0.0217708(0.0297)7090.0042710(0.0297)7110.1183***712(0.0213)713-0.0883***714(0.0211)715-0.0079716(0.0215)717-0.0101718(0.0243)719-0.1330***720(0.0384)721Baths7220.0282*723(0.0121)7240.0578***725(0.0101)7260.0154*727(0.0072)7280.0272***729(0.0078)7300.0198**731(0.0065)7320.0088733(0.0063)7340.0193**735(0.0067)736Size (ft2)7370.000352***738(0.000050)7390.000483***740(0.000024)7410.000322***742(0.000026)7430.000307***744(0.000017)7450.000447***746(0.000023)7470.000323***748(0.000017)7490.000297***750(0.000021)751Size2752-5.52e-08***753(1.33e-08)754-3.58e-08***755(3.73e-09)756-2.60e-08***757(6.23e-09)758-2.09e-08***759(3.03e-09)760-5.38e-08***761(5.60e-09)762-2.92e-08***763(3.68e-09)764-1.34e-08**765(4.59e-09)766Age767-0.005313***768(0.001009)769-0.001945*770(0.000932)771-0.004767***772(0.000535)773-0.004454***774(0.000592)775-0.004380***776(0.000465)777-0.006907***778(0.000462)779-0.006782***780(0.000571)781Age27824.16e-05***783(8.55e-06)7843.22e-05***785(8.50e-06)7863.55e-05***787(4.62e-06)7883.02e-05***789(5.18e-06)7903.43e-05***791(4.17e-06)7924.13e-05***793(4.66e-06)7944.74e-05***795(7.97e-06)796Waterfront797-- 7980.6472***799(0.0771)8000.2623***801(0.0283)8020.2040***803(0.0276)8040.4644***805(0.0227)8060.0394807(0.0584)8080.0471***809(0.0091)810View8110.0611**812(0.0200)8130.1537***814(0.0162)8150.2521***816(0.0115)8170.2952***818(0.0145)8190.2219***820(0.0102)8210.0620***822(0.0144)8230.1446***824(0.0130)825Lot (acres)8260.5880827(0.6628)8281.4153***829(0.2061)8300.8670***831(0.1091)8320.9429***833(0.0931)834-0.0913*835(0.0417)8360.2292***837(0.0313)8380.5780***839(0.0567)840Lot28411.4301842(2.4902)843-0.9184***844(0.1765)845-0.2835**846(0.0937)847-0.1501*848(0.0636)8490.0345**850(0.0108)851-0.0061852(0.0123)853-0.2115***854(0.0349)855Garage (sf)8560.000354***857(0.000052)8580.000235***859(0.000051)8600.000214***861(0.000028)8620.000223***863(0.000028)8640.000200***865(0.000021)8660.000165***867(0.000017)8680.000191***869(0.000016)870Detached 871Garage8720.0829***873(0.0156)8740.1489***875(0.0168)8760.0660***877(0.0099)8780.0415***879(0.0120)8800.0823***881(0.0092)8820.0452***883(0.0082)8840.0729***885(0.0100)886Constant88711.6382***888(0.0636)88911.3575***890(0.0420)89111.6148***892(0.0315)89311.5228***894(0.0282)89511.4539***896(0.0278)89711.6385***898(0.0233)89911.5705***900(0.0249)901R29020.429030.709040.559050.679060.579070.619080.70909N9101,4519112,2649124,6979133,3569144,4549153,4849162,905917Notes: * p<0.05; ** p<0.01; *** p<0.001918919920===== PAGE 13 (score 6, numeric-paren 0, table-words 0, var-words 6) =====921922399923The effect of light rail transit service on nearby property values9246 925Residual analysis926An important underlying assumption for the analysis is that light rail stations are not sited in areas for 927which there may have been preexisting trends in housing prices. Although I do not expect this to be the 928case, it is possible to examine the question graphically. To do this, I estimate for each of the neighbor929hoods a hedonic price regression model but leave out the indicators for the timing of the light rail, and 930the location of the house relative to the rail station. Thus, the residuals from these regressions “contain” 931the effects that are emphasized in the analysis. I then average these residuals for each year, separately for 932the “close” houses and the “not close” houses.933Figure 4: Residual analysis by station neighborhoods.934-.4935-.29360937.2938.4939Residual940199094119959422000943200594420109452015946Year947< 0.5 mi9480.5-2.0 mi949Beacon Hill950-.3951-.2952-.19530954.1955.2956Residual957199095819959592000960200596120109622015963Year964< 0.5 mi9650.5-2.0 mi966Mount Baker967-.4968-.29690970.2971.4972Residual973199097419959752000976200597720109782015979Year980< 0.5 mi9810.5-2.0 mi982Columbia City983-.3984-.2985-.19860987.1988.2989Residual990199099119959922000993200599420109952015996Year997< 0.5 mi9980.5-2.0 mi999Othello1000-.11001-.05100201003.051004.11005Residual1006199010071995100820001009200510102010101120151012Year1013< 0.5 mi10140.5-2.0 mi1015Rainier Beach1016-.11017-.05101801019.051020.11021Residual1022199010231995102420001025200510262010102720151028Year1029< 0.5 mi10300.5-2.0 mi1031Rainier Beach1032-.41033-.31034-.21035-.1103601037.11038residual1039199010401995104120001042200510432010104420151045year1046< 0.5 mi10470.5-2.0 mi1048SeaTac Airport104910501051===== PAGE 15 (score 229, numeric-paren 179, table-words 9, var-words 9) =====105210534011054The effect of light rail transit service on nearby property values1055Table 4: Impact of Light Rail Opening on Characteristics of Houses Sold1056Variables1057Beacon1058Hill1059Mount1060Baker1061Columbia1062City1063Othello1064Rainier1065Beach1066Tukwila/1067Intnl Blvd1068SeaTac1069Airport1070Panel A: Size (ft2)1071Close107246.571073-492.6***107415.761075-144.0***1076-1.8561077-21.101078-511.7***1079(40.64)1080(50.99)1081(30.98)1082(32.10)1083(31.52)1084(39.78)1085(44.40)1086After108749.181088165.2***1089120.2***1090198.2***1091101.2***1092128.4***109379.87***1094(53.28)1095(54.23)1096(23.36)1097(39.96)1098(23.82)1099(30.67)1100(29.08)1101Close x After1102105.61103-6.2111104-2.1051105-72.6411069.026110719.141108213.61109(72.24)1110(73.89)1111(45.06)1112(55.37)1113(61.70)1114(89.35)1115(137.6)1116Constant11171,535***11182,315***11191,636***11201,788***11211,659***11221,652***11231,687***1124(30.28)1125(38.88)1126(14.16)1127(20.22)1128(13.61)1129(14.52)1130(14.87)1131Observations11321,84611332,86811346,01011354,49911365,90311374,66011383,8151139R-squared11400.00911410.06811420.00611430.01811440.00411450.00511460.0251147Panel B: Lot Size (Acres)1148Close1149-0.00540**1150-0.0132***1151-0.001611152-0.0171***1153-0.0352***11540.01101155-0.0277***1156(0.00233)1157(0.00277)1158(0.00291)1159(0.00295)1160(0.00354)1161(0.00956)1162(0.00441)1163After1164-0.00817**11650.0036411660.0011511670.0117**11680.0041511690.0059911700.006841171(0.00340)1172(0.00437)1173(0.00248)1174(0.00558)1175(0.00505)1176(0.00665)1177(0.00674)1178Close x After11790.0105**11800.004911181-0.0164***1182-0.0151**11830.0021211840.0097711850.01181186(0.00432)1187(0.00812)1188(0.00464)1189(0.00666)1190(0.00723)1191(0.0205)1192(0.0149)1193Constant11940.110***11950.126***11960.126***11970.153***11980.179***11990.251***12000.242***1201(0.00188)1202(0.00229)1203(0.000976)1204(0.00217)1205(0.00247)1206(0.00349)1207(0.00291)1208Observations12091,84612102,86812116,01012124,49912135,90312144,66012153,8151216R-squared12170.00712180.00912190.00412200.01212210.01112220.00112230.0021224Panel C: Age1225Close12268.644***12276.734***1228-3.293**1229-6.230***1230-7.467***12310.593123214.34***1233(1.948)1234(1.627)1235(1.533)1236(1.479)1237(1.398)1238(1.465)1239(1.085)1240After1241-4.23512425.476***12436.416***12449.457***124511.53***12468.939***124711.37***1248(3.250)1249(1.880)1250(1.194)1251(1.358)1252(1.004)1253(1.131)1254(0.855)1255Close x After125612.26***12572.8631258-14.78***1259-5.728**12601.59512613.6181262-7.1041263(3.856)1264(2.758)1265(2.880)1266(2.612)1267(2.753)1268(2.940)1269(5.166)1270Constant127162.08***127260.96***127357.20***127450.46***127547.40***127641.59***127736.47***1278(1.704)1279(1.175)1280(0.645)1281(0.769)1282(0.555)1283(0.524)1284(0.406)1285Observations12861,84612872,86812886,01012894,49912905,90312914,66012923,8151293R-squared12940.04612950.02112960.01512970.02312980.03712990.02113000.0731301Panel D: Number of Bathrooms1302Close1303-0.152***1304-0.607***1305-0.02771306-0.060513070.135***13080.01171309-0.636***1310(0.0546)1311(0.0575)1312(0.0404)1313(0.0439)1314(0.0471)1315(0.0559)1316(0.0619)1317After13180.293***13190.252***13200.214***13210.242***13220.138***13230.192***13240.119***1325(0.0769)1326(0.0577)1327(0.0329)1328(0.0454)1329(0.0313)1330(0.0396)1331(0.0375)1332Close x After1333-0.007591334-0.073013350.216***13360.03961337-0.06761338-0.077113390.1781340(0.104)1341(0.0827)1342(0.0681)1343(0.0799)1344(0.0891)1345(0.119)1346(0.167)1347Constant13481.712***13492.466***13501.840***13511.978***13521.858***13531.862***13541.954***1355(0.0440)1356(0.0427)1357(0.0195)1358(0.0243)1359(0.0177)1360(0.0190)1361(0.0204)1362Observations13631,84613642,86813656,01013664,49913675,90313684,66013693,8151370R-squared13710.03113720.08813730.01613740.01213750.00613760.00613770.0241378Notes: Standard errors in parenthesis. Standard errors are heteroskeasticity robust and clustered to account for multiple sales 1379of homes in the sample. *** p<0.01, ** p<0.05, * p<0.1138013811382===== PAGE 16 (score 13, numeric-paren 0, table-words 0, var-words 13) =====138313844021385JOURNAL OF TRANSPORT AND LAND USE 11.11386Panel B reports the estimates for lot size. In this case, the experimental effects estimates are small 1387and not statistically significant for most of the stations. However, for Beacon Hill the estimate is positive 1388and significant, and for Columbia City and Othello, it is negative and significant. The differences range 1389from about 1/60th of an acre to about 1/100th of an acre, but in these neighborhoods the lots are typi1390cally quite small—around one-eighth of an acre for the Columbia City sample.1391Panel C reports estimates for age. Again, the estimates for Beacon Hill are positive and statistically 1392significant and for Columbia City and Othello they are negative and significant. Thus, the houses that 1393are selling near Columbia City and Othello are disproportionately newer suggesting a relative influx of 1394new housing around the Columbia City and Othello stations. The positive age effect for Beacon Hill 1395suggests a relative influx of new housing in parts of the neighborhood that are not close to the station—1396that the catchment area for the rail station is more stagnant with respect to new construction.1397Panel D shows estimates for the number of bathrooms. The only statistically significant experimen1398tal effect is for Columbia City, and it is rather large at about .22. 1399Taken together, these estimates indicate that there have been rather significant changes in the com1400position of housing near the Columbia City station. It is possible that at least some of the large positive 1401estimated impact of rail access in this neighborhood is due to significant changes in the composition of 1402the housing stock in that neighborhood. For most of the other stations, it appears that there has been 1403little or no difference between the catchment areas and more distant parts of the neighborhood in terms 1404of the composition of the housing stock. Clearly, some caution is warranted in the interpretation of the 1405results for the Columbia City station. 14068 1407Discussion1408Although many studies report beneficial impacts of light rail on property values, the mostly null (or 1409negative) findings of this study are not unique. For example, Kim and Lahr (2014) study the impact 1410of the Hudson-Bergen light rail line in Northern New Jersey using a repeat sales method that is similar 1411to the differences-in-differences method used here. They find different effects for different stations, 1412arguing that those neighborhoods that experienced the largest housing price appreciation were in areas 1413that experienced much improved transit service. Housing near stations that were already well-served by 1414public transit options experienced little or no benefit. Chatman, Tulach, and Kim (2012) find neutral 1415to negative impact on housing prices of the River Line in the New Jersey suburbs of Philadelphia using 1416a repeat-sales method. Also, using a model of longitudinal changes in housing prices, Papon, Nguyen-1417Luong, and Boucq (2015) found no beneficial impact of a new light rail line in Paris. Many of the cross-1418sectional studies discussed by Higgins and Konaroglu (2016) also find negative proximity impacts of rail 1419transit stations, but as explained earlier, it is difficult to interpret these results as causal as there may be 1420reasons for light rail stations to be built in areas with depressed housing values. 1421Taking the estimates of the present study at face value, it suggests that the transportation benefit 1422of the new light rail line in Seattle is small, or that non-transportation externalities, such as noise and 1423higher auto congestion near stations, outweigh the benefits. While the public perception of light rail is 1424that it is superior to systems based on buses, this is not necessarily the case, as Richmond (2005) argues 1425in his analysis of the development of Los Angeles’ light rail network. In particular, redesigning the bus 1426transportation network to accommodate the fixed route of a rail line may significantly reduce the utility 1427of public transit for many users (Richmond, 2001). 1428There are limitations to using this study as a basis for evaluating the total benefit of the new light 1429rail service in Seattle. The study makes no attempt to estimate the benefits to users that do not live in the 1430residential neighborhoods of the Rainier Valley. Due to limitations in the data, this study is unable to 1431evaluate the impact of light rail service on apartment dwellers. Commercial properties in downtown ar1432