UBC Faculty Research and Publications

Not in my neighbour's back yard? : Laneway homes and neighbours' property values Davidoff, Tom; Pavlov, Andrey; Somerville, Tsur 2019-03-08

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
52383-Davidoff_Tom_et_al_Not_my_neighbour's.pdf [ 2.16MB ]
Metadata
JSON: 52383-1.0376805.json
JSON-LD: 52383-1.0376805-ld.json
RDF/XML (Pretty): 52383-1.0376805-rdf.xml
RDF/JSON: 52383-1.0376805-rdf.json
Turtle: 52383-1.0376805-turtle.txt
N-Triples: 52383-1.0376805-rdf-ntriples.txt
Original Record: 52383-1.0376805-source.json
Full Text
52383-1.0376805-fulltext.txt
Citation
52383-1.0376805.ris

Full Text

Not in my neighbour’s back yard? Lanewayhomes and neighbours’ property valuesTom DavidoffA,* Andrey PavlovB Tsur SomervilleA8 March 2019AUniversity of British ColumbiaBSimon Fraser UniversityAbstractSingle family zoning is often defended by homeowners as necessary to preserveneighbourhood amenities, but blamed by economists for making housing less af-fordable by constraining supply. In 2009, the City of Vancouver (Canada) rezoned95% of single family areas to allow small “laneway homes” behind main residentialstructures. The goal was to increase rental housing supply with minimal disrup-tion to rezoned neighbourhoods. We exploit this regulatory change to estimate themagnitude of externalities from density. We find small and statistically insignificantaverage spillovers from the addition of laneway housing on the values of adjacentproperties. Laneway homes’ impact is greater and statistically different from zeroon more expensive properties.Key words: Externalities (D62), Land use regulations (R52), Housing Demand(R21)1 We gratefully acknowledge BC Assessment’s assistance in providing transaction and property characteristicsdata for this paper and a SSHRC grant for funding. We have also benefited from the comments of seminarparticipants at the Chinese University of Hong Kong, National University of Singapore, Tel Aviv University,UBC, the University of Toronto, and the University of Washington and discussants at the AREUEA, AsRES,and UEA conferences. Corresponding author: Thomas Davidoff; thomas.davidoff@sauder.ubc.ca; Sauder Schoolof Business, University of British Columbia, 2053 Main Mall, Vancouver, BC V6J 3S1; (604) 822-8325/1 IntroductionThe high cost of housing in “Superstar Cities” such as London, New York, SanFrancisco, Sydney, Toronto, and Vancouver has risen to prominence as a matter ofboth academic study and policy. Economists have pointed to zoning constraintson supply as a primary cause of high house prices, recommending increases inthe allowable density of housing as a mechanism to address the affordability chal-lenges in these cities. 2 Beyond the issue of affordability for locals, recent work (Moretti and Hsieh (2017), Glaeser and Gyourko (2018)) suggests lower nationaleconomic growth from a misallocation of human capital as households are lesswilling to move from low productivity low house price cities to high productivitybut expensive supply-constrained cities.In supply-constrained cities with very high land values like Vancouver, raising al-lowable density would raise property values for properties ripe for redevelopment. 3Yet, local officials often cite the opposition of incumbent homeowners as the causeof their own reluctance to allow more density in single family neighbourhoods. Lo-cal resident opposition to policies for increasing residential densities that wouldprovide them with clear financial gain suggests that incumbent residents attacha considerable value to maintaining the status quo character of their neighbour-hoods. This is consistent with anecdotal evidence and signalling in public hearings2 A long literature links supply restrictions to higher prices. This includes: Pollakowskiand Wachter (1990), Glaeser and Gyourko (2003), Quigley and Raphael (2004), Glaeser,Gyourko, and Saks (2006), Ball, Allmendinger, and Hughes (2009), Gyourko and Molloy(2015), and Hilber and Vermeulen (2014). Some recent work raises issues with econo-metric identification and these effects in a system of open cities; see Aura and Davidoff(2008), Anenberg and Kung (2018), and Davidoff (2014)3 Even with the City of Vancouver taking a substantial share of the increase in valuefrom rezoning, changing a property’s allowed use from single family detached to town-house (an approximate change in the ratio of building floor area to lot size of 0.65 to1.2) results in a 20-25% price premium, with higher premiums for greater increases inthe allowed density. In our case, the allowable floor area ratio increase is much smaller,and owners are not allowed to stratify the property.2on redevelopment applications for zoning changes and variances where local oppo-sition is ubiquitous. The motivations for local homeowners’ opposition has beenthe subject of work in both economics, planning and urban studies, which pointto a variety of correlates with this opposition. 4Some planners have embraced built forms that are a compromise between addingsupply and preserving low density. This “gentle densification” or “missing mid-dle” housing represents a built form between the extremes of low-density singlefamily detached units and high-rise apartments that has been promoted as lessdisruptive to local neighbourhoods. 5 Yet, even modest density infill housing canstir up dramatic opposition. 6 A relevant question is whether this approach bettermanages the tradeoff between the general benefits of increased supply and thepresumed local negative externalities from density that motivate neighbourhoodopposition to rezoning for higher density.In this paper we estimate the external effects on neighbouring property values asa result of a increase in density at the lowest end of the missing middle contin-uum. Accessory dwelling units (ADUs) are smaller stand-alone structures addedto existing single family properties, which do not have separate title. Since 2010,regulatory changes in cities such as Portland, OR, Seattle, WA, Toronto, ON andWashington, DC have encouraged or legalized this built form, and in 2017 Cali-4 See Hamilton (1975), Fischel (2001), and Ortalo-Magne and Prat (2011), who considermarket-level pricing and its role in zoning. Among planning and urban studies: Simsand Bosetti (2016) Monkkonen and Manville (2018)5 See the following: http://missingmiddlehousing.com/, Gee, M., Globe andMail, Aug 18, 2018 https://www.theglobeandmail.com/news/toronto/toronto-has-lots-of-houses-and-towers-but-not-enough-inbetween/article36030909/,https://www.planning.nsw.gov.au/policy-and-legislation/housing/medium-density-housing/6 Dougherty, C. New York Times, Dec. 1, 2017https://www.nytimes.com/2017/12/01/business/economy/single-family-home.htmlprovides a detailed example of just this type of opposition.3fornia effectively legalized ADUs statewide. 7 We exploit a land use policy changeby the municipal government in Vancouver, BC Canada that allowed propertyowners in all single family residential zones to add a ADUs as a single additionalinfill unit (In Vancouver, and throughout this paper, ADUs are referred to as“laneway” units). The policy, first effective July 2009, allowed property ownerswhose lots and existing units met certain size, dimension, and house layout criteriato construct a separate infill unit behind the main house adjacent to the lane oralley running at the rear of the property; hence the term ”laneway”. This appliedto 95 percent of single family zoning areas and was extended to the remainingareas in 2013.Certain features of Vancouver’s laneway program make it attractive for policyanalysis. First, the City of Vancouver has remarkably high housing prices, oftenwinning prizes for its low ratio of income to price from sources such as Zoocasa,and yet has retained single family zoning in the majority of its residential zones.Second, Vancouver has more ADUs in place than any other city in North America:from July 2009 to 2017, approximately 2,913 laneway units were built. Third,whereas zoning changes are typically endogenous to local market conditions, thesmall change in allowable density embodied in the laneway program was appliedquite broadly to all single family zones.Both the blanket introduction of the laneway policy and the way laneways havebeen added help us avoid the endogeneity problem of studying externalities of highdensity development on lower density neighbours. If variation in zoning reflectschanges in market demand across neighbourhoods, then we can not attributechanging property values to the effects of rezoning alone. The attendant bias7 For a discussion see https://https://accessorydwellings.org or Kirk, M. Jan 16,2018 https://www.citylab.com/design/2018/01/the-granny-flats-are-coming/550388/and see Cohen, J. Jan. 14, 2018 https://nextcity.org/daily/entry/california-adu-applications-skyrocket-after-regulatory-reform for the response in California4to estimates of externality effects can be positive because higher density occurswhere land values are highest, or negative because local governments rezone inless well-off, less politically powerful areas. The City of Vancouver’s applicationof the laneway policy to all single family areas avoidsthese endogeneity concernsrelated to location.We estimate the externality of laneway construction on the values of neighbouringsingle family properties at different levels of granularity. Most broadly, we estimatethe effect on a unit’s transaction price of the number of laneway units within a100 meter radius. We conduct a second, narrower test by looking at the effectof an immediate neighbour building a laneway on one’s own property value. Ourcleanest estimation of this effect uses only newly built neighbours, comparingwhen the new build has a laneway and when it does not. Using new homes avoidsthe difficulty in controlling for unobservable aspects of a neighbourhood or theneighbouring houses’s structure that are correlated with an existing householdsdecision to incur the significant costs of replacing their garage with a lanewayhome.Our estimates’ validity depend critically on the random assignment of lanewayconstruction, both by neighbour and neighbourhood. We conduct several tests ofthe assumption that whether a newly built single family property has a lanewayunit or not is random. We generally fail to reject zero relationship between mea-sures of immediate neighbourhood quality and placement of laneway homes, de-spite differences in laneway construction across Vancouver more broadly.In general, we find that the effects of laneways on home values are small andlocalized. The presence of a laneway within 100m of a house is associated witha 0.5% lower transaction price for that house. The estimated effects are largerwhen the laneway is an immediate neighbour: the transaction price of a house is2.8% lower on average when a newly built adjacent house has a laneway than if5the newly built unit was constructed without a laneway. However, in that smallersample, this latter result is not statistically different from zero.The average effects mask a systematic pattern across houses. We allow for varyingcoefficients by testing whether the effect of laneway homes on property values aresensitive to the predicted value of subject properties. A natural conjecture is thatbetter-off households in more expensive housing have some combination of a lowermarginal value of the additional income flow from a laneway in-fill, a larger distastefor sharing their property with another household, greater aversion to renters oradded neighbourhood congestion, and a greater opportunity cost to losing somegarage space that is a result of adding a laneway. Consistently we find that thenegative spillover effect increases in absolute magnitude with predicted subjectproperty price. Effects are close to zero and statistically not different from zerofrom zero for housing below the median and the estimated negative spillover isstatistically different from zero for the highest valued properties (above the 75thpercentile).In the next section we describe Vancouver’s laneway policy. Section 3 reviews theliterature on density externalities and reiterates the contribution of this study.Section 4 details our data sources and identification strategy. Results are describedin Section 5. Section 6 concludes with comments on aggregate welfare effects andpolicy implications.2 Laneway PolicyOur analysis focuses on the construction of laneway infill units (accessory dwellingunits) in the backyards of single family detached properties in the Vancouver,BC. 8 The laneway policy was introduced by the City of Vancouver in July 20098 The City of Vancouver is the central city (2016 population 631,486) of the 23 mu-nicipalities in the Vancouver Census Metropolitan Area (CMA), Canada’s third largest6with the objective to further “Council priorities on Affordable Housing and Sus-tainability.” Both affordability and sustainability were understood to be achievedin part through increased density. As expressed by the City, laneway housinghad the advantage of increasing residential density through a mechanism thatresulted in “little or no visible change in existing neighbourhoods” contributing“to the rental housing mix and housing choice by providing small rental units inestablished neighbourhoods.” 9 Zoning rules were changed in July 2009, allowinglaneways in all RS1 and RS5 single family zones, which represented 95 percentof properties with single family zoning. The remaining single-family and a smallnumber of non-single-family zones were added in July 2013. Laneway units werenot allowed on every single family zoned lot: the presence of a lane, lot area anddimensions, and main structure footprint affect whether construction is possible(these constraints are detailed in the Appendix). Approximately 16 percent of lotswould not qualify, and a further 9 percent of properties could not add a lanewaybecause the existing structure does not allow a sufficient set back from a laneway.For the latter, if the existing structure was torn down, the lot itself could accom-modate a laneway with a newly built, differently placed main structure. Lanewayhomes cannot be sold separately from the main property as they do not have adistinct property title.Laneway units do not substantively change building massing in detached singlefamily neighbourhoods because they effectively substitute for two-car garages. Alaneway footprint is no larger than that of a two-car garage, but at one and ahalf stories they are taller with larger structure massing than the two-car garageCMA (2016 population of 2,463,431). Among major US and Canadian cities, not metroareas, its population density of 14,226 persons per sq. mi. was third highest to NewYork City (27,016 persons per sq. mi. in 2010) and the City of San Francisco (a 2010density of 18,679 persons per sq. mi.).9 All quotes from June 9, 2009 City of Vancouver Policy Report, Urban Structure#08-2000-20, https://council.vancouver.ca/20090721/documents/phea6.PDF.7they typically replace. The net addition to structure density is limited: lanewayhomes are either one- or two-bedroom units, with a median size of 663 sq. ft.(61.6 sq. m.), with the range from the 10th to 90th percentile of 489 to 865 sq.ft. (45.4 to 80.4 sq. m.), compared to 2,275 sq. ft. (211 sq. m.) for the mean newsingle family structure. A laneway does not limit the allowed size of the mainstructure, though it can affect its positioning on the lot because of the requiredminimum distance between the main unit and the laneway. The opportunity costsof a laneway unit are mainly in the loss of garage space: for the typical 33 footwide lot, a laneway unit designs limit the property to up to a a single car garageand additional carport space. Wider lots can accommodate a two car garage aspart of the laneway design.Laneways are not inexpensive and do not provide the lowest cost rental space.To address concerns about lost amenity, the city implemented strict rules onlaneway design, spacing and layout. As a result they are fairly high quality, withconstruction costs for the median size unit is roughly $C 350 per square footor $US 260, which is at the high end of single family construction costs. 10 Thebenefits for renters come from potential increases in the aggregate stock of rentalunits and the addition of stand-alone rental units in neighbourhoods where theywere not previously available.Between the citywide rezoning to allow for infills in July 2009 and the end of2017, 2,913 laneway units were built in the City of Vancouver. This representsapproximately 3.9 percent of the total single family housing stock of 74,835 single10 https://www.vancouverhome.builders/how-much-cost-to-build-house/. The higherper square foot cost comes reflects first the amortization of the fixed costs for a kitchenand bathroom over a smaller unit size. Second, the cost per square foot is lower when thelaneway is built along with a new main structure because for a new build the marginalcost of the laneway is only the cost above that of building a two car garage and theplanning, permitting, landscaping, and engineering costs are a fixed cost required forthe main build.8family detached structures, of which 66,893 were in area zoned for the lanewayinfill units after 2013. Laneway homes represent 10.4 percent of all single, attached,and multi-family units built in the city over this period. In Figure 1 we show thedistribution of these laneway infill units, differentiating between those built aspart of the construction of a new house (built 2010 or later) and those addedto existing structures. The single family zones (RS1/RS5) where the lanewayswere first permitted are shown in the dark hue, with the slightly lighter hue thesecond July 2013 rezone. Over two thirds of laneways were built as part of anew construction, which in all cases required the teardown of an older, existingsingle family unit. In total, for zones that allowed laneways, 1,984 of 6,334 singlefamily properties (31.3 percent) built after 2009 (new homes) had a laneway, while920 of 60,306 single family properties (1.5 percent) built prior to 2010 (existinghomes) added a laneway. For units built after 2013 the laneway share among newbuilds is over 40 percent. While both types are distributed throughout the city, theconcentration is higher is on the eastern side of the city, where 72% of lanewayswere built.The distribution of laneways correlates broadly with price dispersion in the city.Figure 2 shows the distribution of residuals from a hedonic house price regressionwith lot and structure characteristics and time dummies for year month 2012-2017, so no controls for location (the data and controls are detailed below). Theresiduals, which in part reflect geographic variation, are higher in magnitude onthe city’s west side, where the incidence of laneways is lower. This reflects thedifference in mean transaction prices between the two sides of the city of $C2.81M and $C 1.07M west vs. east. In the analysis we control for these differencesusing neighbourhood fixed effects, for the areas outlined in Figure 2.9Fig. 1. Location of all laneways.10Fig. 2. Hedonic Price Residuals - No Geographic Controls.113 LiteratureIn the absence of externalities, binding caps on residential density imposed byzoning cause deadweight loss because the market is prevented from adding spacefor which people would pay more than construction cost. However, density affectswelfare through a variety of external mechanisms. Urban models of size and struc-ture commonly include negative externalities of residential density as congestionthat limits the aggregate positive effect of density enhancing agglomeration exter-nalities (Ahlfeldt, Redding, Sturm, and Wolf (2015), Ahlfeldt and Pietrostefani(2017), and Brinkman (2016)). In these models the effects of density are capturedat an aggregate, often metropolitan area, level. Zoning may also be necessary forefficiency in the provision of local public goods: Hamilton (1975) demonstrateshow zoning that forces properties to be built at specific values addresses the prob-lem of fiscal free riding when public goods are funded by ad valorem propertytaxes. Limits on residential size or use may also deliver benefits by preservinglocal amenity, which Fischel (2001) characterizes as opposition by local residentsto density that is in the aggregate welfare increasing. Consistent with these ben-efits to regulation, McMillen and McDonald (2002), who show that introductionof zoning in Chicago in 1923 increased residential property values.Local opposition to density is the subject of an extensive literature. Whittemoreand BenDor (2018) provide a summary of the work in planning on this opposi-tion. The planning work highlights a wide a variety of contributing factors to theopposition, much of which is tied to neighbourhood composition and change. Inthe economics literature, with the exception of Turner, Haughwout, and van derKlaauw (2014), recent efforts such as Glaeser and Gyourko (2018) have focusedon the aggregate welfare gains from increased supply, rather than identifying thenature of negative externalities that development can impose on nearby residents.The interactions among land use regulation, welfare, and housing supply suggest a12modelling treatment of house prices that allows for different types of effects. BothStrange (1992) and Turner et al. (2014) disaggregate the different ways zoning re-strictions can effect property values and welfare. The former presents a theoreticalmodel that categorizes density effects both within and across neighbourhoods. InStrange’s model, the direct effects are the presumed negative spillovers on nearbyproperties and the indirect effects occur as increases in density in one area inducewholesale zoning changes elsewhere, resulting in positive aggregate supply effects.Turner et al. (2014) describe the effect of density changes as own-lot, external,and supply effects. The first is the potential positive price effect from increasesin the allowed development density on one’s own lot. The second is the negativeexternality higher density on a lot may impose on neighbouring properties. Thethird is the aggregate effect of increases to housing supply that is the source of thepositive welfare effects. Notably, Turner et al. (2014) find that both the own-lotand external effects of density are positive. In this paper we focus on identifyingthe existence and magnitude of spillovers, Strange’s first and Turner, Haughwout,and van der Klaauw’s second effects. The nature of our data and identificationstrategy generate a clean estimate of this effect for modest increases in densityThere is a long empirical literature that attempts to estimate the effects of densityon neighbouring property values. Early work by Sagalyn and Sternlieb (1973), andStull (1975) found that multi-family properties in single family neighbourhoodslower the values of the nearby single family units. Others such as Crecine, Davis,and Jackson (1967), Rueter (1973), Maser, Riker, and Rosett (1977), and Markand Goldberg (1986) do not find that proximity, as measured by a sample of singlefamily properties a given distance from the location of the density under study,has an effect on their prices.A problem with these empirical studies is the non-random allocation of density.First, as noted by Strange (1992), higher density at a particular location is likely to13affect the possible redevelopment option value of nearby units, essentially positiveredevelopment spillovers (Turner et al. (2014) own effect), because these units arenow more likely to be re-zoned for higher density. Working in the other direction,it may be that areas that are conducive to higher density have other aspects thataffect their prices, either positively or negatively depending on one’s model of thepolitical economy of zoning. If rezoning to higher density follows the market, as perWallace (1998), then higher density occurs in areas with high land prices. On theother hand if density is allocated to areas with less well organized local opposition,then density is likely to occur in lower value areas. Given these conditions, it isunlikely that single family units near the density are a random sample.A further identification challenge arises if residents of dense structures differ mea-surably from existing residents, then the observed price effects may not the den-sity per se, but a negative association with the occupants. Diamond and McQuade(2016), find that the construction of subsidized housing for low-income residents,where these residences are at a higher density than the surrounding neighbour-hood, lowers the prices of nearby lower density single family structures in high-priced neighbourhoods, but the effect does not occur in lower priced areas whereresidents of the existing and subsidized properties are more similar. Given thehigh quality of laneway homes, this is less of a challenge in our context, but couldbe a channel through which we find larger adverse effects among more expensivehomes.We estimate the negative spillover associated with greater household density, inthis case an increase of one household per lot in density. Unlike Turner et al.(2014), we study realized changes in density, rather than differences in municipallevel land use regulation intensity. As well, we do so with substantially greaterprecision and a higher degree of proximity then in their work, though this comesat the expense of the greater geographic breadth they study. In their paper, iden-14tification results from variation in the distance to a neighbouring municipalitywhere the regulatory regime is different. They study the effect of proximity to amunicipal boundary on land prices, which can be sensitive to unobserved factorsthat affect the parcels development option value. Their approach does not distin-guish whether there are actual differences across the border in density or builtform, as they measure regulatory environment using municipality level measures.Our contribution lies in a clear test of an actual increase in density, adding anadditional neighbouring housing unit, while both controlling for the standard setof house, lot, and neighbourhood characteristics or fixed effects and using a treat-ment effect that is homogenous throughout the sample (a newly built neighbourwith or without the additional laneway infill unit).4 Estimation and Identification4.1 Option Grant and Announcement EffectsAround the time of the July, 2009, enactment of the laneway program, singlefamily property values were affected in several ways. First, homes eligible forlaneway construction were endowed with a potentially valuable option. Second,the potential for new laneway homes as rental supply should have affected pricesand rents everywhere through the downward slope of demand for space. Third, thepresent discounted value of external effects of possible laneway option exercise bynear and distant neighbours should have been realized. Fourth, the hedonic valueof two-car garages may have started to rise to the extent that market participantsforesaw their replacement with laneway homes.We have explored regression specifications that use changes in prices of differentlyzoned and built homes around the time of the first and subsequent laneway an-nouncements, but do not believe that we can convincingly estimate any of these15four effects in isolation or collectively. A generic problem is that we do not knowthe timing of changes in beliefs about the probability of the laneway program andsubsequent revisions; the program was discussed publicly for some time beforeenactment. It is tempting to consider RS-zoned homes that are eligible versusnot within some neighbourhood boundary to isolate the option grant effect. How-ever, because eligibility is something we must estimate and because our estimatesare spatially correlated, the anticipated externalities of neighbouring propertiescannot plausibly be held constant across laneway-eligible and ineligible properties.4.2 Laneway Option Exercise EffectsA natural conjecture is that among types of homes where it is common to buildersconstructing new homes, both with and without a laneway, adding a lanewayshould add roughly as much value as the typical incremental cost of constructionplus the appropriate return for the builder. If the value added were much greater,then particularly among new homes built speculatively, it would be a mistaketo fail to add a laneway. If the value added were far below plausible incrementalcosts, it would be a mistake for builders to add these to new homes. There is moreroom for market value added to be less than the marginal cost among existinghomes, where incumbent owners may have idiosyncratic preferences for lanewayhomes versus privacy and larger garages.To estimate the impact of the option exercise, we compare sales prices amonghomes that are eligible to build a laneway home that sell after the 2009 optiongrant, with and without a laneway home. We regress:pi,j,t =β0 + β1Xi + β2Li,t + µj,t + γt + i,t. (1)16In (1) pi,t is the log transaction price p for house i, in neighbourhood j, and at timeperiod (month-year) t. Xi are characteristics of the property, µj,t is a dummy forthe neighbourhood-year combination, 11 and γt is the effect of sale in year-montht.β2 measures the marginal contribution of a laneway to value. We restrict thisestimation to properties eligible for a laneway house at that time. This includesall RS-1 and RS-5 zoned properties up through 2013, and then adds propertiesin all remaining RS zoning categories after 2013. We require for identification ofβ2 as the causal effect of a laneway on own value that i,t is independent of theindicator for the presence of a laneway, Li,t.4.3 Effect on Adjacent PropertyBuilding a laneway house represents a modest increase in structure density and sitecoverage, and also an increase in the number of residents and vehicles, potentiallyincreasing neighbourhood congestion. To estimate the existence and magnitudeof this potential externality we test for the effect of a laneway on the transactionprices of properties within a fixed radius. As the effect is likely to decay withdistance, we additionally test for effect on the transaction price of a property thathas a neighbouring property with a laneway unit.The first more general neighbourhood effect has the following specification:pi,t =β0 + β1Xi + β2Li,t + β3NLi,t + β4NPi,t + µj,t + γt + i,t (2)11 Examples of the 23 neighbourhoods for which we have controls µ are: False Creek,Kitsilano, and Dunbar.17In (2), in addition to the variables defined above in (1), we add NLi,t, the numberof laneways within 100 meters of transacting property i in period t, and NPi,t,the number of residential properties within 100 meters of transacting property iin period t.Our more refined test replaces the laneway count in the neighbourhood NLi,t witha dummy variable NNLk,t for neighbour k that is adjacent to property i that takeson the value of one of one if neighbour k has a laneway and 0 otherwise:pi,t =β0 + β1Xi + β2NNLk,t + β3Yk,t + µj,t + γt + i,t (3)The specification also includes the neighbour’s property characteristics Yk,t, whichwe limit to age, floor and lot area. The parameter β2 estimates the mean effectsof a neighbouring laneway above and beyond the neighbour’s total floor and lotarea, relative to a neighbour without a laneway.4.4 Random Placement of LanewaysFor our estimates of the externalities from density to be unbiased there must bea random allocation of laneways conditional on observable property characteris-tics, neighborhood-year dummies, and citywide year-month dummies. Explicitly,whether a unit has a laneway must be independent of neighbourhood, own unit,and neighbouring unit characteristics that affect property value and are not oth-erwise controlled for in the regression specification. Figures 1 and 2 above showthat at a citywide level this does not hold: laneways are not randomly distributedin all single family zones, but are concentrated more intensively in Vancouver’sless expensive eastern side. Failure to absorb this gross difference would virtuallyguarantee a negative estimated bias in our external effect.18To absorb these coarse geographic differences, all regressions have on the righthand side the neighbourhood-year fixed effect µj,t for twenty-three neighbour-hoods in the relevant single family zones as defined by BC Assessment. The crit-ical question issue is whether within neighbourhoods the allocation is random,conditional on observable characteristics of the transacting unit and on time. Totest the validity of this assumption we conduct three tests for the robustness ofour analysis to concerns that laneways are not randomly placed within broaderneighbourhoods.First, we test whether a new property has a laneway or not is independent ofthe unit’s predicted value. That is: are laneways systematically built with betteror worse new builds, conditional on observables and geography? To do so weregress predicted value, which itself is conditional on property characteristics andthe geographic and time fixed effects, on whether the unit has a laneway or not.Economically significant estimated coefficient on predicted value would be groundsfor concern that whether a neighbouring new build has a laneway or not reflectsthe reference property’s value.Li,t = f(Pˆi,t, µj,t, γt) (4)As above in (1), Li,t takes on the value of 1 if a transacting property has alaneway and 0 otherwise. This is a function, depending on the specification (linearprobability or probit), of the predicted price Pˆi,t the neighbourhood-year fixedeffects µj,t , and citywide year-month fixed effects γt. Pˆi,t is the predicted valuefrom a first stage hedonic estimate that iexcludes Li,t as a right hand side variablein the first stage: 1212 For all fitted values of price, time is controlled to be Jan. 2015 so that predicted value19Pˆit = β0 + β1Xi + µj,t + γt + i,t (5)Our second robustness test is related to the first, but addresses the value of the newbuild’s neighbour, which is the reference property in (3). Are laneways built nextto better or worse properties? Here we estimate among homes that are adjacentto new houses built after laneways are allowed whether there is any statisticaldifference in their prices before the policy introduction among those that get anadjacent new home with a laneway and those that get an adjacent new homewithout a laneway after the policy change. Is there a pre-policy value differencewhich is related to an unobserved factor rather than the laneway presence in theex-post regressions for (3)? Among all properties with a new neighbouring homebuilt after 2009, and for all t < T = July, 2009 we regress log transaction pricepi,t against the specification from (3), except the neighbour has a laneway dummythat is time invariant, taking on the value of 1 if at any point post-time T theneighbour has a laneway and 0 otherwise:pi,t =β0 + β1Xi + β2NNLk + β3Yk,t + µj,t + γt + i,t (6)Again where β2, the estimated coefficient on NNLk that is economically or sta-tistically different from zero would lead to concern about the orthogonality of thelaneway dummy variable in specifications (1), (2) and (3).Our final robustness test is whether the probability that a laneway is built is afunction of sub-neighbourhood local-area characteristics. Specifically, are lanewaysis time invariant. Of course the coefficients µ and γ will differ between (4) and (5).20more likely to be built in higher or lower value areas within our 23 broader neigh-bourhoods. Here we determine time t > T = Dec, 2009 share of new homes withlaneways SJ within more localized census tracts and postal code groups, a finerlevel of geography J than the neighbourhoods j we use more generally.Share of New with Laneway ≡ SJ = Number of New with LanewayNumber of all New(7)We then include this share of laneway houses variable in a standard hedonicregression on all transactions prior to 2009. , For t < T = Jan 2009.pi,t = β0 + β1Xi + β2SJ + µj,t + γt + i,t, (8)Collectively these tests address whether the location of a laneway is clearly non-random, relative to a property’s own quality (test (6)), nonrandom relative to thequality of the laneway’s neighbour (test (7)), and nonrandom relative to the qual-ity of the neighbourhood (test (8)). In all cases, an estimated coefficient that is notstatistically different from zero is consistent with the assumption of randomness.4.5 DataOur data is the universe of properties in the City of Vancouver and all transac-tions of single family properties from 2005 through 2017. These data come fromBC Assessment, the Province’s property assessment authority. We use transac-tions from different time periods for different tests, so for simplicity limit ourdescription here to the primary tests of externalities. We combine the transactiondata from Vancouver for single family houses with property roll data that provides21a rich set of property characteristics. By combining the transactions set with theuniverse of all properties we are able to identify the characteristics of immedi-ately adjacent properties, even if the neighbour does not transact. We derive areduced sample of 20,930 transactions of single family units from a full sample of31,534 single family transactions from 2012-2017. The full sample is windsorizedat the 0.5th and 99.5th percentiles on price, floor area, and lot size. We limit ourtransactions to those in areas with single family zoning in 2017 and where eachincluded transacting property has a single family unit adjacent on either side. Weuse transactions from 23 of the city’s 30 neighbourhoods that had lots zoned forsingle family detached units that permitted a laneway infill. 13An important distinction we make is between laneways built as part of new con-struction and those added to an existing unit. The marginal costs of a lanewayadded to a new build has a lower marginal cost than that replacing an existinggarage where the main structure remains whole. The counts of transactions showthat 20.6% of our new build transactions have a laneway versus 1.6% of existingstructures. In Figures 3 and 4 we show the distribution of transactions for newand existing homes respectively, in each case distinguishing between those withand without a laneway unit.13 Of the seven excluded neighbourhoods, five are in the high density downtown core,where the single family units are unusual heritage preservation outliers. The other twohave large numbers of single family units, but are currently zoned for other types ofresidential uses such as duplex, triples, quadplex and for townhouses.22Fig. 3. Location of all new build transactions.23Fig. 4. Location of all existing build transactions.24Laneways built behind existing units present potential problems in identification.First, the large difference in the incidence of a laneway between new and existingunits suggests a different dynamic in the decision to build one. The motivationsmay be different. Builders with whom we have spoken indicate that based clientproperty owners who add a laneway to an existing unit are motivated by reasonsother than the present value of the gained rental income. Typically, they plan onhaving a family member live in the unit or move into it themselves as they age, al-lowing themselves to downsize without relocating. Both the cost and the personalconsumption benefit for adding a laneway to an existing unit may thus indicateof a long intended tenure, which might be correlated with unusual unobservedinvestment in the property or neighbourhood quality.In Table 1 we provide select summary characteristics for the total sample andselect subsets. For both new and existing units we provide separate summaries forunits with laneways and the total for each group. All are for properties that trans-act over the period 2012-2017. Newer properties have larger floor areas, highermean prices, and are more likely to be on the more expensive west side of thecirty. However, mean lot size are the same across both new and existing prop-erties. New properties with laneways have lower average mean prices by 28%,smaller floor area by 7%, and smaller lot area by 10%. This difference in partreflects laneway homes’ higher incidence on the lower value east side of the city,which also has fewer large lots. Existing units with a laneway are slightly newerthan existing structures without, but otherwise identical in observable propertycharacteristics, though not location, and hence have a slightly lower mean price.We find that 21% of transactions of new homes have laneway units, as opposedto only 2% of transactions of existing homes.25Meanmean sd N New homes Existing homesVariable All homes All Has Laneway All Has LanewayAge 29.84 20.3 20930 1.79 1.24 34.89 30.37( -196 ) ( -189 ) ( 26 ) ( 0.51 )Floor area (000 sq ft) 1.69 0.71 20930 2.17 2.03 1.61 1.58( 44 ) ( 21 ) ( -11 ) ( -3.14 )Has laneway home 0.04 0.21 20930 0.21 1 0.02 1( 24 ) ( 661 ) ( -12 ) ( 661 )No. laneway homes within 100m. 1.03 1.31 20920 0.96 1.57 1.04 1.26( -3.01 ) ( 9.06 ) ( 0.75 ) ( 2.73 )Lot size (000 sq ft) 5 1.79 20930 5 4.5 5 4.98( 0 ) ( -11.58 ) ( 0 ) ( -0.27 )New house 0.15 0.36 20930 1 1 0 0( 342 ) ( 342 ) ( -60 ) ( -60 )Price ($000) 1766 1215 20930 2447 1794 1643 1490( 24.85 ) ( 0.67 ) ( -10.33 ) ( -4.95 )No. residential units within 100m. 45.97 17.46 20920 46 49.2 45.96 45.93( 0.09 ) ( 4.12 ) ( -0.06 ) ( -0.03 )No. single-family within 100m. 41.83 11.36 20920 42.23 44.46 41.76 41.57( 1.89 ) ( 6.47 ) ( -0.6 ) ( -0.42 )West side 0.32 0.47 20930 0.46 0.2 0.29 0.23( 14.85 ) ( -7.53 ) ( -6.4 ) ( -3.56 )Table 1Summary statistics by type of single family home. t statistics for type differences from all homes in parentheses26In our immediate neighbour specification, identification comes from the effect onthe transacting reference unit of whether a neighbour has a laneway or not. Wetest this for all neighbours, and then in a more focused sample using transactionsthat have a newly built neighbour, again comparing the effect on the referenceproperty’s price if the new neighbour includes an additional infill laneway unit. Allof these new houses result from demolition of an existing older single family homesthat is replaced by the newer, larger single family house. Teardowns have beenused to extract estimates of urban land values in developed areas and to valueredevelopment options. 14 Our work is different in that we do not attempt to modelthe redevelopment choice, but instead use these properties for both control andtreatment groups. We prefer the newly built neighbour sample because the controlgroup (newly built neighbours without laneway infills) is much more likely to besimilar to the treatment group (newly built neighbours with laneway infills) thanfor older houses in general with and without an added laneway: for neighbouringunits the mean floor space to lot area ratio for new homes is 0.44 with a standarddeviation of 0.054, while for existing homes the mean is 0.33 with a standarddeviation of .103. 15The sample size when we limit transactions to those with a newly built neighbour(defined as less than five years old) is much smaller. Of these, further droppingcorner units or those with a property other than a single family detached house14 See Rosenthal and Helsley (1994), Helms (2003), Dye and McMillen (2007), Clapp andSalavei (2010), Clapp, Bardos, and Wong (2012), and McMillen and OSullivan (2013)for research on teardowns15 An additional consideration is that among older homes that have not yet built alaneway or been rebuilt, construction of a laneway in the near future may be perceivedas an option that is likely to be exercised in the not too distant (and not too heavilydiscounted) future. Thus the impact of exercising (versus retaining) that option mayhave a small impact on neighbours. By contrast, building a new two-car garage with anew home involves a significant reduction in the likelihood of building a laneway homein the near future. Old garages may also be more unappealing to neighbours than newlybuilt garages.27adjacent, results in a sample of 2,329 transactions, 558 of which have a new buildneighbour with a laneway and 1,771 have a new build neighbour without. Weshow the distribution of these transactions in Figure 5. As with the larger sampleof all neighbouring units, our incidence of having a new build neighbour with alaneway is higher on the east side.28Fig. 5. Location of transactions that are adjacent to a new house.Table 2 reports the summary statistics separately for transacting properties with anewly built neighbour that includes a laneway house and those with a newly builtneighbour that does not have a laneway infill unit. In general the transactingunits with an adjacent newly built neighbouring house with a laneway are lessexpensive, smaller, and on smaller lots, than transacting properties with a newlybuilt adjacent single family house without a laneway. Much of this reflects thewest-east split in price and laneway incidence in the city noted earlier, which alsoextends to difference in mean lot and property size.29New-built neighbour with a laneway(1) (2) (3) (4) (5)VARIABLES N mean sd min maxprice 558 1,635,479 936,243 195,250 5,500,000Finished area 000sf 558 1.619 0.603 0.600 4.041Lot size 000sf 558 4.812 1.592 2.750 27.080Age - renovation adjusted 558 27.685 22.809 0.000 90.000Property has a Laneway unit 558 0.165 0.371 0.000 1.000West-side 558 0.237 0.425 0.000 1.000No of laneways within 100m, excl. own 558 2.581 1.619 0.000 9.000No of 1-fam within 100m 558 42.876 10.680 8.000 67.000No of all resid units within 100m 558 47.138 15.423 8.000 140.000New-built neighbour without a laneway(1) (2) (3) (4) (5)VARIABLES N mean sd min maxprice 1,771. 2,205,315 1,359,172 124,000 9,790,476Finished area 000sf 1,771 1.913 0.736 0.654 6.209Lot size 000sf 1,771 5.223 1.878 2.011 26.277Age - renovation adjusted 1,771 23.591 21.218 0.000 102.000Property has a Laneway unit 1,771 0.040 0.196 0.000 1.000West-side 1,771 0.511 0.500 0.000 1.000No of laneways within 100m, excl. own 1,770 0.844 1.136 0.000 8.000No of 1-fam within 100m 1,770 41.167 11.147 4.000 69.000No of all resid units within 100m 1,770 44.695 18.756 5.000 469.000Table 2Summary statistics The table reports the summary statistics for transactions with a newly built neighbourwith and without a laneway.5 Results5.1 Impact of Own Laneway HouseOur first test is the estimation of equation (1) which considers the own propertyvalue impact of building a laneway house. With perfect foresight and full informa-tion this should not exceed the exercise cost of the laneway option, the marginalconstruction cost. Moreover, if we see similar new homes with laneway homes insignificant numbers in the same neighbourhood, then for identical builders the30increase in sale price of new homes associated with presence of a laneway homeshould not be less than the marginal cost of providing a laneway unit over atwo-car garage. The results reported in Table 3 splits the sample by new vs. old(existing) houses, and then for new houses between the east and west sides of thecity, and above and below median predicted price. Houses built within 5 years ofthe transaction date are defined as new.Building a laneway house is associated with higher property value for a new houseof roughly 11.3 percent, per specification (1). At the sample mean this is $C259,300, which is within the plausible range of cost to build a laneway homeinstead of a two car garage. Adding a laneway to an existing house does notappear to increase its resale value. However, as we note above existing homes aremore likely to have unobserved characteristics that are correlated with lanewayexercise, complicating the interpretation of this coefficient. This may be more truefor units where a laneway is added, and then the home is sold soon afterwards.Adding a laneway to an existing home may also indicate a low redevelopmentoption value, as it would likely not make sense to build a laneway home and thensell the property if the property was less ripe for teardown and redevelopment.The results are consistent with the difference in laneway uptake between theeast and west sides of Vancouver. The figures and summary tables above showa notably higher incidence of laneway uptake on the east side. In regressions (2)and (3) the marginal contribution of a laneway to property transaction price isfour times as large for new East Side properties as new West Side, where onaverage house prices are twice as high. Using predicted price instead of geographyto segment the city, we see similar results: large positive and statistically differentfrom zero effect on price among new houses below the median in predicted price,but no statistically different than zero effect on transaction prices for new houses31(1) (2) (3) (4) (5) (6)VARIABLES New House Old House New: East New: West New: Low Price New: High PriceProperty has a Laneway unit 0.113*** -0.007 0.154*** 0.037 0.197*** 0.029(0.036) (0.035) (0.041) (0.092) (0.053) (0.058)Dummy, = 1 if laneway suitable 0.025 -0.001 0.035 0.045 0.092** 0.020(0.021) (0.011) (0.028) (0.035) (0.041) (0.026)Lot size 000sf 0.159*** 0.135*** 0.120** 0.139** -0.360 0.141***(0.038) (0.008) (0.059) (0.050) (0.254) (0.041)Lot size squared -0.007*** -0.003*** -0.005 -0.006 0.058* -0.006**(0.002) (0.000) (0.004) (0.004) (0.034) (0.003)Finished area 000sf -0.317*** 0.062** -0.331*** -0.194 0.246 -0.035(0.095) (0.026) (0.106) (0.181) (0.348) (0.122)Finished area squared 0.095*** -0.002 0.084*** 0.084** -0.102 0.045*(0.021) (0.004) (0.021) (0.035) (0.110) (0.024)Number of bedrooms -0.010 -0.007* -0.005 -0.015** 0.010 -0.014**(0.006) (0.004) (0.010) (0.006) (0.011) (0.006)Number of bathrooms (full+half) 0.026*** 0.006 0.026*** 0.028*** -0.008 0.032***(0.007) (0.005) (0.009) (0.010) (0.016) (0.007)Number of stories 0.037 0.034*** 0.078** 0.006 0.072 0.032(0.032) (0.012) (0.035) (0.058) (0.063) (0.041)Dummy, =1 if has full basement 0.007 0.033*** -0.001 0.007 -0.017 0.004(0.016) (0.010) (0.029) (0.017) (0.043) (0.018)Age - renovation adjusted 0.214*** -0.015*** 0.237*** 0.195*** 0.205*** 0.212***(0.015) (0.001) (0.018) (0.023) (0.025) (0.018)Age Squared -0.038*** 0.000*** -0.041*** -0.036*** -0.037*** -0.038***(0.003) (0.000) (0.003) (0.004) (0.005) (0.003)Had major renovation -0.004 0.040*** -0.028 0.018 -0.055 0.012(0.015) (0.008) (0.020) (0.021) (0.035) (0.016)Dummy, =1 if has single car garage 0.130*** 0.021** 0.148*** 0.074 0.153*** 0.158**(0.034) (0.010) (0.036) (0.108) (0.040) (0.060)Dummy, =1 if has multi-car garage 0.235*** 0.000 0.245*** 0.179** 0.264*** 0.220***(0.035) (0.010) (0.039) (0.082) (0.047) (0.053)Observations 3,195 17,735 1,715 1,480 781 2,414R-squared 0.812 0.677 0.789 0.648 0.608 0.782Neighborhood/time effects + controls Yes Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 3Effect of own laneway home. The table reports the impact of having a laneway unit as estimated by Equation(1). We report the estimates for a new house (five years or less of age) and an existing house (over five years old).We further break the sample down by East/West halves of the city and by price. The positive effect of lanewayhouses on price is statistically different from zero only for lower-priced properties and for new properties on theeast side. All regressions have controls for both neighbourhood-year and city wide year-month fixed effects32with above median predicted prices. 165.2 Impact of Nearby Laneway HousesOur primary objective in this paper is to measure the spillover from density.Our first approach to do so uses equation (2), where we add to the specificationused above in Table 3 the number of laneway houses within a 100-meter ringaround each observation and the number of single-family properties within thesame radius as a control for potential number of laneways and local single familydensity. 17 We report our results below in Table 4. In Table 5 we use separatefixed effects for one, two, and three or more laneways within 100m. In both tableswe include in the specification, but exclude reporting, lot and structure hedoniccharacteristics.16 The predicted price, here and throughout the paper, is based on a model that includesthe same covariates as the specification presented in Table 3 plus neighbourhood andtime effects. For the purpose of predicting the price for splitting the samples in this table,and in all subsequent tables, we use the actual physical characteristics and location ofeach property, but fix the time to a specific point (January 1, 2015). This is designedto identify the properties that are high value because of characteristics or location, notbecause of time of sale.17 We have run the regressions with the total number of residential units in the 100mring and gotten identical qualitative results and extremely similar point estimates. Wehave also used a 250m ring and obtained the same pattern of results.33All laneways and 1-family counts, 100-meter ring(1) (2) (3) (4) (5) (6) (7) (8)VARIABLES Full sample High pp Very high pp w10 w25 w50 w75 w90No of laneways within 100m, excl. own -0.006** -0.013** -0.022*** -0.001 0.001 -0.001 -0.005** -0.012***(0.003) (0.005) (0.008) (0.003) (0.002) (0.002) (0.002) (0.003)No of 1-fam within 100m 0.001 0.001 0.001 0.000* 0.001*** 0.001** 0.000 -0.001(0.001) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) (0.001)Observations 20,920 9,782 4,451 20,920 20,920 20,920 20,920 20,920R-squared 0.700 0.553 0.440 0.520 0.473 0.576 0.764 0.833neighbourhood/time effects + controls Yes Yes Yes Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 4Effect of nearby laneway homesThe table reports the estimation of Equation (2) for all transactions using the laneway counts and single-family counts. “High pp”and “Very high pp” are based on sub-samples of properties whose predicted price is above the median and above the 75th percentile, respectively. The locally-weightedleast squares regressions, (4) through (8), use weights that are inversely proportional to the absolute distance measured in predicted price to the listed percentile ofthe predicted property price distribution. The negative effects of both laneways and overall density are concentrated in the higher-valued properties. We detect noeffect in the average or below average-priced properties. All regressions have controls for property and lot characteristics, neighbourhood-year fixed effects, and citywide year-month fixed effects.34All laneways and 1-family counts, 100-meter ring(1) (2) (3)VARIABLES Full sample High pp Very high ppOnly 1 laneway house within 100 m -0.024*** -0.041*** -0.044**(0.008) (0.011) (0.018)Two laneway houses within 100 m -0.034*** -0.043*** -0.065***(0.010) (0.014) (0.021)Three or more laneway houses within 100 m -0.021* -0.031 -0.044*(0.012) (0.021) (0.026)No of 1-fam within 100m 0.001 0.001 0.002(0.001) (0.001) (0.001)Observations 20,920 9,782 4,451R-squared 0.700 0.553 0.441neighbourhood/time effects + controls Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 5Effect of nearby laneway homesThe table reports the estimation of Equation (2) for all transactions usingthe laneway counts and single-family counts similarly to Table 4, except the counts are included as categoricalvariables. “High pp” and “Very high pp” are based on sub-samples of properties whose predicted price is abovethe median and above the 75th percentile, respectively. Overall, it is the first and second laneways that areassociated with lower property values, but in the sample of higher value homes three or more can also havea statistically different than zero negative effect. Point estimates of the negative externality are higher for thesample restricted to higher-valued properties, based on predicted value. All regressions have controls for totalsingle family units within the 100m ring, property and lot characteristics, neighbourhood-year fixed effects, andcity wide year-month fixed effects.In Table 4 the number of laneway houses within a 100-meter radius is associ-ated with a 0.4 percent lower price per laneway, though this point estimate is notstatistically different from zero. The effect of nearby laneway houses on a givenproperty’s price is larger when we limit the sample to houses with higher predictedprices by a factor of 4 to 5 times and the point estimates are statistically differentthan zero. We further present the results of locally-weighted least-squares regres-sion that uses the inverse of the absolute difference in predicted price as the weightsin regressions (4) through (8). In regression (4) the weights are higher for prop-erty’s with a predicted price closer to the lowest 10 percentile of predicted price. Inregression (8) the weights are highest for property’s with a predicted price closerto the 90th percentile of predicted price. The impact of nearby laneway houses is35very close to zero or insignificant for all quantiles except for the highest decile. Ofnote, the effects of overall residential density are similar.In Table 5 we use fixed effects for different numbers of laneways rather than thecontinuous measure used in Table 4. 18 In the full sample, one laneway within 100mis associated with a 2.4% lower price, a second adds a further 3.64% decline, andthree or more a further 2.1% decline. The effects for the higher predicted valuehomes sub-sample are up to twice the magnitude of those for the overall sample.These findings suggest that while on average a laneway in a neighbourhood hasa negative impact on an affected property’s price, this effect is sensitive to aneighbourhood wealth and driven by effects among the most expensive homes,with below the median value homes unaffected by the increase in neighbourhooddensity.Since counts of nearby homes in 100-meter rings and unobservable neighbourhoodcharacteristics are spatially correlated, the standard errors reported in Table 4could be biased. However, we find that when we account for spatial autocorrela-tion using the methods developed by Conley (1999), there is no increase in ourestimated standard errors. In fact, in some specifications accounting for the spatialcorrelation of the errors increases the precision of our estimates. 195.3 Impact of a Neighbouring Laneway HouseTo further investigate the spillover from density we limit the analysis to the ef-fect if one of the two immediate neighbours of the transacting property has alaneway. The sample is limited to transacting properties with an adjacent single18 45.8% of transactions have no laneway within 100m, 27.6% have one, 14.5% have two,6.5% have three, 3.0% have fur, and 2.6% have more than four.19 We find this in a setting where OLS and HAC standard errors are computed withneighbourhood dummies and year-month dummies, but not the interactions includedin the main specifications. Those interactions would make the computation of spatiallycorrected standard errors infeasible.36family unit on each side, which eliminates corner lots and lots bordering non-single family land uses. Using the specification from Equation (3), we estimatethe effect on a property’s transaction price when one of the two immediate neigh-bours has a laneway house and present these in Table 6. As above, we present theresults for sub-samples of high and very high predicted price as well as the resultsfrom weighted least-squares estimation for various quantiles of the predicted pricedistribution.Table 6 presents the results for all neighbours, whether a newly-constructed prop-erty or an existing property. With the exception of the weighted regression centredon the highest-value properties, the impact of a neighbouring laneway house isvery small and the point estimates are not statistically different from zero. Thepattern is representative of that in Table 4, where the effect of a nearby lanewayis more negative for higher value properties.A potential problem with the results shown in Table 6 is bias because a lanewayneighbour may be non-random. As described above, most laneway houses are builtas part of new construction. Therefore, the impact of a neighbouring laneway ispotentially confounded by the impact of a newly built neighbouring property. Ifnewly built neighbours replace older rundown housing they may generate positivespillovers biasing a possible negative pure laneway effect. In addition, we remainconcerned about unobservables in the decision of an existing homeowner to adda laneway.To address this issue, we repeat the neighbouring laneway house estimation forthe sub-sample of newly constructed neighbours only. We show results for thusre-estimation of Equation (3) when the sample is limited to newly built neigh-bours only in Table 7. The identification comes from the difference between thetransaction price of a house that has a newly built neighbour that has a lanewayand the transaction price of a house that has a newly built neighbour that does37not have a laneway. For the purposes of this table, we eliminate the rare cases oftwo newly constructed neighbours. While the impact of a neighbouring lanewayhouse is not significant for the full sample, the results in Table 7 display a clearand more consistent pattern of the negative effect of an adjacent laneway as partof a new build compared to neighbouring new builds without a laneway on thetransaction prices of the highest priced properties than we see in Table 6’s analysisof all properties. As with other specifications the negative spillovers are small inmagnitude and not statistically different from zero for all but the highest pricedhomes, whether we segment or weight the sample.38Single-family neighbourhoods, all neighbours(1) (2) (3) (4) (5) (6) (7) (8)VARIABLES Full sample High pp Very high pp w10 w25 w50 w75 w90A nieghbor has a laneway -0.003 0.002 -0.009 0.027 0.003 -0.006 0.002 -0.046***(0.019) (0.029) (0.045) (0.019) (0.011) (0.015) (0.016) (0.014)Observations 11,149 5,493 2,820 11,149 11,149 11,149 11,149 11,149R-squared 0.723 0.575 0.439 0.577 0.509 0.589 0.770 0.844neighbourhood/time effects + controls Yes Yes Yes Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 6Effect of immediate neighbor’s laneway home.The table reports the estimation of Equation (2) for all transactions. “High pp” and “Very high pp” are basedon sub-samples of properties whose predicted price is above the median and above the 75th percentile, respectively. The locally-weighted least squares regressions,(4) through (8), use weights that are inversely proportional to the absolute distance measured in predicted price. Having a neighbour with a laneway has a small andinsignificant impact on property value in almost all sub-samples. The only exception is that properties at the 90th percentile in terms of predicted price and onlywithin single-family neighbourhoods seem to experience a significant negative price effect from a neighbouring laneway. All regressions have controls for property andlot characteristics, neighbourhood-year fixed effects, and city wide year-month fixed effects.39Single-family neighbourhoods, new neighbours only(1) (2) (3) (4) (5) (6) (7) (8)VARIABLES Full sample High pp Very high pp w10 w25 w50 w75 w90New neighbour has laneway -0.028 -0.036 -0.089** -0.023 -0.011 -0.027 -0.013 -0.069***(0.026) (0.029) (0.038) (0.024) (0.020) (0.018) (0.018) (0.020)Observations 1,330 878 488 1,330 1,330 1,330 1,330 1,330R-squared 0.775 0.674 0.563 0.760 0.670 0.679 0.770 0.840neighbourhood/time effects + controls Yes Yes Yes Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 7Effect of immediate neighbor’s laneway home (new neighbors only). The table reports the estimation of Equation (2) for transactions where there is anewly built (past 5 years) neighbouring single family property. “High pp” and “Very high pp” are based on sub-samples of properties whose predicted price is abovethe median and above the 75th percentile, respectively. The locally-weighted least squares regressions, (4) through (8), use weights that are inversely proportionalto the absolute distance measured in predicted price. Having a neighbour with a laneway house has a very marginal and insignificant effect for low and medium-priced properties. However, it has a large and significant negative effect for higher-priced properties. All regressions have controls for property and lot characteristics,neighbourhood-year fixed effects, and city wide year-month fixed effects.405.4 Testing for Randomness in Laneway “Assignment”Our results in Table 7 depend on the assumption that laneway homes are con-structed at random among all newly built houses that are eligible for a lanewayhome, contingent on controls. If laneway houses are systematically built as partof a better or worse quality new build, or adjacent to better or worse qualityneighbours, or in superior or inferior areas, or in a way associated with onob-served features that affect the price of neighbouring homes, then the parameteron neighbouring laneway house may potentially be capturing the effect of oneof these aspects rather than the impact of a laneway. Such a possibility wouldundermine the identification assumptions of Equations (2) and (3) that “Numberof Laneways” NL and “New Neighbour has Laneway” NNL is uncorrelated withthe error term and generate biased results.Our first test is whether laneway units are built behind better or worse qualityhomes within neighbourhoods, per Equation (4). We use the fitted price of a newbuild property as an explanatory variable in a regression of whether the propertyhas a laneway or not. First, we estimate a predicted price for all new propertiesas in (5). The predicted price is then used in a probit model to estimate the prob-ability of having a laneway home. If predicted price has no explanatory power inthe likelihood a new build has a laneway, then our approach is consistent with therandomness assumption relative to own house value. The results in Table 8 no sta-tistically different from zero or consistent pattern of relationship between a unit’spredicted price and whether it has a laneway, contingent on unit characteristicsand neighbourhood.Our second test for random assignment evaluates whether houses where newlybuilt neighbouring units had laneways are systematically different from single fam-ily properties where the newly built neighbouring units did not include a laneway.To test this condition we take the sample of properties with new neighbours 2012-41(1) (2) (3)Has laneway Has laneway Has lanewayVARIABLES LABELS Full sample High predicted price Very high predicted pricelnPhat Predicted log-Price 0.085 -0.815 0.342(0.263) (0.529) (1.509)Observations 2,526 1,107 179neighbourhood/time effects Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 8Robustness test. The table reports the probability a property has a laneway as a function of its predictedtransaction price. The predicted price is based on a model with all covariates we use in Table 3 plus neighbourhoodand time effects, but benchmarked to Jan. 2015. Models “High pp” and “Very high pp” are based on sub-samplesof properties whose predicted price is above the median and above the 75th percentile, respectively. In all caseswe consider, the predicted price of a property does not predict the probability of a laneway house.2017 and use their transaction prices from 2005-2008, but include as a dummyvariable whether they will have a new build with laneway in the future 2012-2017period: their price prior to the announcement of the laneway policy as a functionof the ex-post realization of their new build neighbour having a laneway or not.The results of these regressions are shown below in Table 9. Globally, propertiesthat will in the future have a new build neighbour with a laneway do.not havetransactions prices prior to the announcement of the laneway policy that are sys-tematically different in ways that would bias or previous results. As well, unlikeabove, the estimated coefficients in the w10 and w90 locally weighted regressions(specifications (4) and (8)) are essentially identical. There is also no clear patternby own predicted price as we have also seen in all other regression tables above.The point estimates, though none are statistically different from zero, are for themost part greater than zero, which would suggest that any bias in the Table 7 andothers would be towards finding a positive rather than negative effect of laneways.42(1) (2) (3) (4) (5) (6) (7) (8)VARIABLES Full Sample High PP Very high PP w10 w25 w50 w75 w90New neighbor has laneway 0.019 0.004 0.117 0.036 0.045* 0.042* 0.005 0.032(0.030) (0.077) (0.177) (0.025) (0.024) (0.021) (0.023) (0.027)Observations 983 492 248 983 983 983 983 983R-squared 0.796 0.762 0.847 0.807 0.711 0.692 0.782 0.876neighbourhood/time effects Yes Yes Yes Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 9Random assignment test. The table reports the effect of having a future new build neighbour with a laneway versus without on the price of a property prior tothe rezoning that allowed laneways. Regressions include all lot and property characteristics plus neighbourhood and time effects. Models “High pp” and “Very highpp” are based on sub-samples of properties whose predicted price is above the median and above the 75th percentile. Properties that will get a laneway post 2009 donot have lower transaction prices 2005-2008 than those that will not among properties that will have new build neighbour post 2009.43Finally, using data for the 2012 - 2017 period, we calculate the share of propertieswith laneway houses out of all newly built properties by census tract and by 5-digit postal code. 20 We then use this geographic concentration measure as anexplanatory variable in estimating (8). If laneway houses are more likely to bebuilt in a superior or inferior neighbourhoods, then the post-2011 concentrationvariable would affect pre-2009 prices. We focus on the post-2011 concentrationsbecause most laneways in our sample were built after 2011, reflecting the timeto take advantage of the newly permitted use (proposal development, approval,construction, and sale). The results of this third test are shown in Table 10. Wefind that the ex-post concentration of laneway houses is small and not significantin explaining pre-2009 transaction prices for any of the definitions or sub-sampleswe use. The sign switches between positive and negative between specifications,just as one might expect if the assignment is random. That nonmonotonicity ofthe relationship between fitted property value and coefficient on laneway presenceprovides some comfort that the rising negative coefficient on neighbouring lanewayhomes with property value reflects a causal external effect.20 This is the first five characters in the Canadian alpha numeric X0X 0X0 postal codeand is approximately five full blocks in a single family area. The BC Assessment neigh-bourhood definitions we use for geographic fixed effects have approximately 3 censustracts and 60 5-digit postal codes for each of the larger fixed effect neighbourhoods.44(1) (2) (3) (4) (5) (6)5-digit code 5-digit codeVARIABLES Tract Tract 5-digit code 5-digit code High PP Very high PPShare of new w laneway out of all new, CT -0.021 0.037(0.063) (0.056)Share of new w laneway out of all new, PC 0.031 0.016 0.035 -0.011(0.034) (0.020) (0.032) (0.049)Observations 8,627 8,602 8,598 8,574 4,438 2,154R-squared 0.498 0.611 0.499 0.612 0.477 0.318neighbourhood/time effects Yes Yesneighbourhood/time effects + controls Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 10Random assignment test. This table verifies the random assignment of laneway houses among all newly built properties. We first compute the share of new houseswith laneways out of all new houses built in the 2012 - 2017 period by census tract and by 5-digit postal code. We then include these shares in hedonic regressions forall single family transactions prior to 2010. If laneways are systematically built on superior or inferior properties, the share of laneway houses would have a significantimpact on pre-2010 transaction prices. Instead, the share of new houses with laneways out of all post 2010 new houses has no explanatory power in any of the modelspecifications we consider. We interpret this result as evidence that laneway houses are randomly assigned among all new houses.45(1) (2) (3) (4) (5)VARIABLES w10 w25 w50 w75 w90Share of new w laneway out of all new, PC 0.048 0.033 0.015 -0.014 -0.027(0.045) (0.032) (0.011) (0.014) (0.017)Observations 8,574 8,574 8,574 8,574 8,574R-squared 0.500 0.331 0.317 0.443 0.709neighbourhood/time effects + controls Yes Yes Yes Yes YesRobust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Table 11Random assignment test. This table verifies the random assignment of laneway houses among all newly builtproperties in a manner similar to Table 10 except using weighted regression. We first compute the share of newhouses with laneways out of all new houses built in the 2012 - 2017 period by 5-digit postal code. We then includethese shares in hedonic regressions for all single family transactions prior to 2010. If laneways are systematicallybuilt on superior or inferior properties, the share of laneway houses would have a significant impact on pre-2010 transaction prices. Instead, the share of new houses with laneways out of all post 2010 new houses hasno explanatory power in any of the model specifications we consider. We interpret this result as evidence thatlaneway houses are randomly assigned among all new houses.6 ConclusionWe explore the impact of exercising the right to build a laneway home on one’sown property and one’s immediate neighbour’s property values. Exercising theright in a newly built home generally means reducing the size of the garage andthe presence of a rental unit. Consistent with builder optimization conditions,adding a laneway adds value to a property on average an amount approximateto reasonable incremental costs. However, the positive benefit is smaller and maybe below costs for higher priced properties and neighbourhoods, where lanewayhomes are less frequently found. The value of a laneway to a buyer does declinewith property value.We view our clearest contribution as the estimate of the externality associatedwith the presence of a laneway unit in a newly built home on neighbours. Acrossall properties an adjacent or nearby laneway has little meaningful effect on a prop-erty’s value. However, for higher priced properties, the negative effects of nearby46laneways becomes more pronounced and statistically different from zero amongthe most valuable properties. We find evidence consistent with a monotonicallymore negative impact, continuing to increase in percentage and absolute termsas we limit or weight the sample towards higher predicted value properties. Thispattern is consistent across all out tests.These findings are consistent with the notion that occupants of lower-priced neigh-bourhoods and properties have a stronger preference for additional rent or a sepa-rate living space for a family member a laneway house can provide. By comparison,the relatively better-off residents would appear to place a greater value on largergarage space and less shared space relatively to rental income, so a laneway doesnot add value to their properties.The policy implications of our results are mixed because of the negative effect ofhigher density on higher priced homes. The addition to the rental stock and thebroadening of the location and unit type choices for renters is an unambiguouspositive effect on renters’ welfare, a clear concern for many affordability-challengedcities. The pattern of our results suggests that laneways in neighbourhoods wherehouses are below the 75th percentile in house value does not impose negativeexternalities. As the cost imposed on nearby homeowners increases and the directbenefit to a property owner with added rental density declines as the value ofa property increases., we are not clear on the point where Pareto gains end northe aggregate effect of allowing laneways in all neighbourhoods and balancing thenegative externalities in higher end areas against the broader gains. At a minimumour results drawing on the Vancouver context indicate that there are gains tobe had from increasing density modestly in many single family neighbourhoods,a process that would help address affordability challenges in high house priceand rent cities. The fact that laneway homes add little value to own value andsubtract value from neighbouring properties among the most desirable homes need47not imply that densifying these neighbourhoods is unhelpful. Natural alternativeapproaches in expensive neighbourhoods with very high land value are to allowthe sale of infill units, or even permitting significant increases in density thatwould change the entire housing stock.487 Appendix7.1 Laneway GuidelinesDetails on rules and regulations can be found at http://vancouver.ca/home-property-development/building-your-laneway-house.aspx. We summarize the more detailedcity guiidelines below as follows:(1) The property needs to back on a lane or another street. Properties that haveno lane or street separating them from the property behind are not eligible tobuild a laneway house. This restriction applies even for corner lots, which intheory have the necessary access for fire and other services. For this reason,we identify all properties for which the lot polygon border is not within 4meters of a laneway as ineligible.(2) The requirements for cite coverage and access imply that properties witheither of these characteristics are not eligible for a laneway house:• Lot is less than 110 ft deep or narrower than 25 ft• Lot is both less than 33 ft wide and less than 122 ft deep(3) The total site coverage of the main house and the laneway cannot exceed 40%of the property area. This restriction is particularly binding for propertiesknown as “Vancouver Special” because of their size and location within thelot. To identify these properties, we apply the following filters:• Property built between 1963 and 1986, and• One story with full basement, and• Floor area exceeds 1500 sq. ft., and• Floor area to lot size greater than .5, and• Lot size is less than 9000 sq. ft.,• But allow laneways for large lots which exceed 148 x 36 feet.The above criteria for selecting properties eligible for laneway homes appears to49be accurate. All laneway homes observed in our data are located on propertiesthat satisfy the above criteria. There are no laneway homes on properties that ourmechanism labels as ineligible. However, it is conceivable that properties labeledas eligible in our analysis are in fact not eligible. This could potentially bias downany impact of own laneway home eligibility that we find. As well, applicants mayappeal to the Director of Planning for a variance.50ReferencesGabriel M. Ahlfeldt and Elizabetta Pietrostefani. The economic effects of density:A synthesis. Discussion Paper 210, ESERC, 2017.Gabriel M. Ahlfeldt, Stephen J. Redding, Daniel M. Sturm, and Nikolaus Wolf.The economics of density: Evidence from the berlin wall. Econometrica, 83(6):2127–2189, 2015. doi: 10.3982/ECTA10876. URL https://onlinelibrary.wiley.com/doi/abs/10.3982/ECTA10876.Elliot Anenberg and Edward Kung. Can More Housing Supply Solve the Afford-ability Crisis? Evidence from a Neighborhood Choice Model. Finance and Eco-nomics Discussion Series 2018-035, Board of Governors of the Federal ReserveSystem (U.S.), May 2018. URL https://ideas.repec.org/p/fip/fedgfe/2018-35.html.Saku Aura and Thomas Davidoff. Supply constraints and housing prices. Eco-nomics Letters, 99(2):275–277, 2008.Michael Ball, Phil Allmendinger, and Cathy Hughes. Housing supply and planningdelay in the south of england. Journal of European Real Estate Research, 2(2):15169, 2009.Jeffrey C. Brinkman. Congestion, agglomeration, and the structure of cities. Jour-nal of Urban Economics, 94(C):13–31, 2016.John M. Clapp and Katsiaryna Salavei. Hedonic pricing with redevelopmentoptions: A new approach to estimating depreciation effects. Journal of UrbanEconomics, 67(3):362–377, 2010.John M. Clapp, Katsiaryna Salavei Bardos, and S.K. Wong. Empirical estimationof the option premium for residential redevelopment. Regional Science andUrban Economics, 42(1-2):240–256, 2012.T.G. Conley. Gmm estimation with cross sectional dependence. Journalof Econometrics, 92(1):1 – 45, 1999. ISSN 0304-4076. doi: https://doi.org/10.1016/S0304-4076(98)00084-0. URL http://www.sciencedirect.com/science/article/pii/S0304407698000840.John P. Crecine, Otto A. Davis, and John E. Jackson. Urban property markets:Some empirical results and their implication for municipal zoning. Journal ofLaw and Economics, 10(1):79–99, 1967.Thomas Davidoff. Supply constraints are not valid instrumental variables forhome prices because they are correlated with many demand factors. Workingpaper, Sauder School of Business, University of British Columbia, 2014.Rebecca. Diamond and Timothy McQuade. Who wants affordable housing intheir backyard? an equilibrium analysis of low-income property development.Working Paper 22204, NBER, 2016.Richard F. Dye and Daniel P. McMillen. Teardowns and land values in the chicagometropolitan area. Journal of Urban Economics, 61(1):45–64, 2007.William Fischel. The Homevoter Hypothesis. Harvard University Press, Boston,512001.Edward Glaeser, Joseph Gyourko, and Raven Saks. Urban growth and housingsupply. Journal of Economic Geography, 6:71–89, 2006.Edward L. Glaeser and Joseph Gyourko. The impact of building restrictions onhousing affordability. FRBNY Economic Policy Review, 9:21–39, 2003.Edward L. Glaeser and Joseph Gyourko. The economic implications of housingsupply. Journal of Economic Perspectives, 32(1):3–30, 2018.Joseph Gyourko and Raven Molloy. Regulation and housing supply. Handbook ofRegional and Urban Economics, 5:1289–1337, 2015.Bruce W. Hamilton. Zoning and property taxation in a system of local govern-ments. Urban Studies, 12(2):205–211, 1975.Andrew Helms. Understanding gentrification: An empirical aaalysis of the de-terminants of urban housing renovation. Journal of Urban Economics, 54(3):474–498, 2003.Christian A. L. Hilber and Wouter Vermeulen. The impact of supply constraintson house prices in england. The Economic Journal, 126(591):n/a–n/a, 2014.ISSN 1468-0297. doi: 10.1111/ecoj.12213. URL http://dx.doi.org/10.1111/ecoj.12213.Steven M Maser, William H Riker, and Richard N Rosett. The effects of zoningand externalities on the price of land: An empirical analysis of monroe county,new york. Journal of Law and Economics, 20(1):111–32, 1977.Daniel P. McMillen and John F. McDonald. Land values in a newly zoned city.Review of Economics and Statistics, 84(1):62–72, 2002.Daniel P. McMillen and Arthur. OSullivan. Option value and the price of teardownproperties. Journal of Urban Economics, 74(1):107–129, 2013.Paavo. Monkkonen and Michael Manville. Opposition to development or opposi-tion to developers? survey evidence from los angeles county on attitudes towardsnew housing. Working paper, UCLA, 2018.Enrico Moretti and Chang-Tai Hsieh. Housing constraints and spatial misalloca-tion chang-tai hsieh. Working paper, University of Chicago and NBER, 2017.Francois Ortalo-Magne and Andrea Prat. On the political economy of urbangrowth: Homeownership versus affordability. CEPR Discussion Papers 8243,C.E.P.R. Discussion Papers, February 2011. URL http://ideas.repec.org/p/cpr/ceprdp/8243.html.Henry O. Pollakowski and Susan Wachter. The effects of land-use constraints onhousing prices. Land Economics, 66(3):315–324, 1990.John M. Quigley and Steven Raphael. Is housing unaffordable? Why isn’t it moreaffordable? Journal of Economic Perspectives, 17(1):191–214, 2004.Stuart S. Rosenthal and Robert W. Helsley. Redevelopment and the urban landprice gradient. Journal of Urban Economics, 35(2):182–200, 1994.Frederick H. Rueter. Externalities in urban property markets: An empirical testof the zoning ordinance of pittsburgh. The Journal of Law & Economics, 1652(2):313–349, 1973.Lynne B. Sagalyn and George Sternlieb. Zoning and housing costs: The impactof land-use controls on housing price. The ANNALS of the American Academyof Political and Social Science, 409(1):223–224, 1973.Sam Sims and Nicolas Bosetti. Stopped: Why people oppose residential devel-opment in their backyard. Technical report, Centre for London, 2016. URLhttps://www.centreforlondon.org/publication/nimby-opposition/.William C. Strange. Overlapping neighborhoods and housing externalities. Jour-nal of Urban Economics, 32(1):17–39, 1992.William J. Stull. Community environment, zoning, and the market value of singlefamily homes. Journal of Law and Economics, 18(2):535–557, 1975.Mathhew A. Turner, Andrew Haughwout, and Wilber van der Klaauw. Land useregulation and welfare. Econometrica, 82(4):1341–1403, 2014.Nancy E. Wallace. The market effects of zoning undeveloped land parcels: Doeszoning follow the market? Journal of Urban Economics, 23(3):307–326, 1998.Andrew H. Whittemore and Todd K. BenDor. Reassessing nimby: The demo-graphics, politics, and geography of opposition to high-density residential infill.Journal of Urban Affairs, Forthcoming:1–20, 2018.53

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.52383.1-0376805/manifest

Comment

Related Items