annotations of studiesonlinepubs.trb.org/onlinepubs/tcrp/tcrp_rpt_39-d.pdf · the costs of...

The Costs of Sprawl–Revisited Annotations of Studies

Rutgers • Brookings • Parsons Brinckerhoff • ECONorthwest 133 TRANSIT COOPERATIVE RESEARCH PROGRAM(TCRP) H-10

SECTION

ANNOTATIONS OF STUDIES

This section contains detailed

information on specific source materialsof the literature on sprawl. Theinformation is presented according to fivebasic categories (and their subcategories)of sprawl's impacts. A listing of worksincluded here by author follows thereferences section of this monograph.

I. Public and Private Capital andOperating Costs1. Alternative Development Analyses2. Fiscal Impacts, Exactions, and

Impact Fees3. The Effects of Growth Controls on

Housing Costs4. Urban Form and Sprawl

II. Transportation and Travel Costs1. Changes in Automobile Travel2. The Effects of Density on Travel

Choice3. Unlimited Outward Extension

4. Spatial Segregation of Uses (LandUse Mix and Urban Design)

5. Dispersed Employment6. The Costs of Travel

III. Land/Natural Habitat Preservation1. Land Preservation and Community

Cohesion2. Land/Habitat Preservation:

Empirical Studies

IV. Quality of Life1. Popular Literature2. Indicators, Report Cards, and

Benchmarks3. Economics Literature4. Sociology Literature5. Psychology Literature

V. Social Issues1. The Growth of Cities and

Metropolitan Areas2. Urban Decline3. Urban Renewal

III

The Costs of Sprawl–Revisited Annotations of Studies


The citations that have been annotatedhere are far from an exhaustive coverageof the individual topic. Yet, within thefive basic categories of sprawl's impacts,there is a representation of some of thefield's most important current andhistorical literature. Approximately 25percent of the references have beenannotated. Each of the annotated works isreviewed in a similar manner. The firstportion of the analysis places the work inan overall context. A second componentinvolves a discussion of why the work isimportant to the sprawl literature. A thirdcomponent discusses methodology and

data sources employed, if appropriate. Afinal component discusses results andconclusions of the work.

The annotations have been chosen toinclude both positive and negativepositions on sprawl as well as a balancebetween descriptive and empirical works.

The annotations included here contributeto the judgments on strengths andweaknesses of the literature contained inthe previous chapter. Sprawl has a richand diverse literature, as will be seen inthis compilation of annotations.

The Costs of Sprawl–Revisited Capital and Operating Costs


CHAPTER

Annotations of Studies

PUBLIC/PRIVATE CAPITALAND OPERATING COSTS

This chapter annotates studies that relateto sprawl and the alleged public/privatecapital and operating costs that itimposes. Public capital and operatingcosts are costs associated with residentialand nonresidential development: roads,utilities, public buildings and the costs ofproviding day-to-day police, fire, generalgovernment, recreational, and educationalservices. Private capital and operatingcosts are the expenses incurred inoccupying residential and nonresidentialproperties—in other words, the costsimposed on housing and commercialdevelopment related to type, location, anddensity of development. The chapter isdivided into four parts:

Alternative Development AnalysesFiscal Impacts, Exactions, and Impact

FeesThe Effects of Growth Controls on

Housing CostsUrban Form and Sprawl

The Alternative Development Analysessection includes the works of Robert W.Burchell (1997b, 1992a, and others),James Frank (1989), James Duncan

(1989), RERC (1974), and Paul Tischler(1994), among others. Fiscal Impacts,Exactions, and Impact Fees includes, inaddition to the above, the works of theAmerican Farmland Trust (1986, 1992)and DuPage County, Illinois (1989, 1991).The Effects of Growth on Housing Costsliterature includes the works of Katz andRosen (1987), George Parsons (1992), andSeymour Schwartz et al. (1981 and 1989).Finally, the section on Urban Form andSprawl includes the articles of Helen Ladd(1991) and Richard Peiser (1989).

ALTERNATIVE DEVELOPMENTANALYSES

Burchell, Robert W. 1997b. SouthCarolina Infrastructure Study:Projection of Statewide InfrastructureCosts, 1995-2015. New Brunswick, NJ:Center for Urban Policy Research,Rutgers University.

This study involves a 20-year projectionof infrastructure need in the State of SouthCarolina. It encompasses all public andquasi-public infrastructure required in thestate including developmental (roads,bridges, water/sewer), educational,

9



commercial, public safety, public health,recreational/cultural, and environmental.Twenty-eight individual categories ofinfrastructure are contained in the aboveseven groupings listed above. Findings forthe state are as follows:• $56.7 billion in required infrastructure

costs from 1995-2015• $16.7 billion in potential infrastructure

savings due to technology, differingmeans of provision, and costs ofsprawl savings. Much of the abovesavings to come from technology anddiffering means of provision, asopposed to costs of sprawl savings

• Three-quarters of the remaining $40billion ($2 billion per year for 20years), or an average of $1.5 billionannually, could be raised via 10percent infrastructure set-asides in allstate, county, municipal, and schooldistrict general fund budgets andintergovernmental transfer revenues

• The remaining $500 million annuallymust be raised from a variety ofrevenue sources, including propertytax, sales tax, the tolling of roads,development impact fees, water/sewercharges and the like. A c2/ /gallongasoline tax increase would raise only$56 million in revenues annually forinfrastructure purposes.

Burchell, Robert W. 1992a. ImpactAssessment of the New Jersey InterimState Development and RedevelopmentPlan. Report II: Research Findings.Trenton: New Jersey Office of StatePlanning.

This study is an analysis of the effects ofimplementing a managed growth strategyin the state of New Jersey in the form of aState Plan. Two alternative futures aremodeled—one with development as usual(TREND), and one with developmentaccording to the proposed State

Development and Redevelopment Plan(PLAN).

The study showed that the state could save$1.4 billion in infrastructure funding over20 years for roads, utilities, and schools, ifit followed the PLAN versus the TRENDscheme. This savings occurred mainlythrough more intensive use of existinginfrastructure, as opposed to buildingadditional infrastructure. The PLANapproach directed new growth to whereexcess capacity existed, rather than tovirgin territories. The PLAN scheme wasalso more compact than the TRENDscheme, therefore requiring less distanceto be covered when linking developmentsby local and county roads. In addition,concentrating larger numbers of people inmore compact areas provides foreconomies of scale, such as larger waterand sewer treatment facilities, with lowercosts per individual user. The Rutgersstudy, using the same increases inpopulation (520,000 persons in 20 years),jobs, and households, for both TRENDand PLAN, found that PLAN would save:• $699 million in roads—a 24 percent

savings• $561 million in water and sewer

costs—a 7.6 percent savings• $173 million in school capital costs—

a 3.3 percent savings• $1.4 billion, overall, or just under 10

percent of all development-related(roads, water/sewer, public buildings)infrastructure expenditures.



Burchell, Robert W. et al.1. The Economic Impacts of Trendversus Vision Growth in theLexington Metropolitan Area.Report Prepared for BluegrassTomorrow, Lexington, KY. 1995b(November).2. Impact Assessment of DELEPCCMP versus STATUS QUO onTwelve Municipalities in the DELEPRegion. Report Prepared for by theGovernment Committee of theDelaware Estuary Program.Philadelphia, PA. August 15, 1995.3. Fiscal Impacts of AlternativeLand Development Patterns inMichigan: The Costs of CurrentDevelopment Versus CompactGrowth. Southeast MichiganRegional Council of Governments.1997a.

These three studies extend and broadenthe application of Burchell's New Jerseymodeling of development alternatives(sprawl versus compact growth) todifferent geographic settings. A broaderbase enables refinement and testing of themodels under different taxing structures,differing means of providing and fundinginfrastructure, and differing geographiclevels of investigation.

Each of the studies—Lexington,Kentucky, the Delaware Estuary, and theState of Michigan—looked at the land,infrastructure, housing, and fiscal costs ofsprawl versus compact development.Compact development was differentlydefined in each study; sprawl which wasequated with historical development,varied only marginally from place toplace. There was much more consistencyin the definition of trend development orsprawl across these studies than there wasin the specific alternatives to sprawl.Findings from the studies are included inthe following table:

Duncan, James E. et al. 1989. TheSearch for Efficient Urban GrowthPatterns. Tallahassee: Department ofCommunity Affairs.

This analysis encompasses detailed casestudies of the actual costs (and revenues)incurred by several completed residentialand nonresidential projects throughoutFlorida. The projects chosen arerepresentative of five differentdevelopment patterns ranging from"scattered" to "compact." Although theFlorida study did not intend such ananalysis, it is possible to group the fivedevelopment patterns into two aggregatedevelopment profiles, "trend" and"managed/planned" growth. The term"trend" includes the Florida developmentpatterns labeled "scattered," "linear," and"satellite"; the term "managed/planned"refers to the Florida patterns of"contiguous" and "compact" growth. Withthis grouping, the relative capital costs fortrend versus managed or planned growthcan be determined from the base Floridacase study information. The data showthat the total public capital costs for adetached unit built under trend conditionsin Florida approached



$16,000; under planned development thecapital cost was about $11,000 per unit, orroughly 70 percent of the cost of "trend"(see table below).

ECONorthwest. 1994. Evaluation of NoGrowth and Slow Growth Policies for thePortland Region. Portland, OR:Metropolitan Portland Government.

As part of its Region 2040 planningprogram, Metro (Portland, Oregon'smetropolitan area's regional government)evaluated alternative futures for the regionand the policies available to achieve thosefutures. This report examined whether theregion could or should adopt policies toreduce and slow the anticipated growth inthe region.

The study identifies both reactive andproactive policies which would slowgrowth. Reactive policies woulddiscourage growth through amunicipality's economic developmentactivities, its investment in infrastructureand public facilities, and its programs forenvironmental amenities and socialservices. Proactive government policieswould discourage growth by limiting thesupply and use of buildable land (viazoning and planning regulations),charging more for public facilities, andimposing increased environmental anddesign standards on existing and newdevelopments.

The authors concluded that whilemunicipally initiated reactive or proactivepolicies would slow growth in the shortrun at the local level, a viable strategy forthe Portland metropolitan area must beimplemented at a regional scale, so thataggregate growth is decreased and notsimply redistributed within themetropolitan area.

Finally, the study pointed out thatalthough not always considered, slowgrowth policies could have negativeeffects on the local economy. Theseinclude decreased economic opportunitiesand income growth, and increasedhousing costs.

Fodor, Eben V. 1997. "The Real Cost ofGrowth in Oregon" Population andEnvironment 18-4.

Oregon municipalities levy systemdevelopment charges to offsetdevelopment costs. Within Oregon, for atypical three-bedroom home, thesedevelopment charges can range from$1,000 to $6,500 per unit. Fodor,however, conservatively estimates that theactual public cost is closer to $24,500 perunit. He provides a breakdown as shownin the table below.

Fodor concludes that Oregon (like otherstates) is only recovering a fraction of thepublic infrastructure costs through systemdevelopment charges and otherdevelopment fees. Rather, he postulatesthat communities are subsidizing growthby keeping housing prices artificially low.Implementing growth managementstrategies, while providing thesesubsidies, works at cross purposes. Fodorsuggests that communities should pursuealternatives, such as the public acquisitionof land to prevent development, as aviable, cost-saving policy for growthmanagement.



Frank, James E. 1989. The Costs ofAlternative Development Patterns: AReview of the Literature. Washington,DC: Urban Land Institute.

This study reviews the national literatureconducted over roughly four decadesconcerning development costs. Frankorders the findings of the various reportsand expresses them in equivalent dollarterms (1987 dollars). He concludes fromthe national literature that multiple factorsaffect development costs includingdensity, contiguity of development,distance to central public facilities (e.g.,sewage and water plants), as well as othercharacteristics, such as municipalimprovement standards. In brief, capitalcosts are highest in situations of low-density sprawl, and for developmentlocated a considerable distance fromcentral facilities. By contrast, costs can bedramatically reduced in higher-densitydevelopment that is centrally andcontiguously located. As described byFrank:

When all capital costs are totaled ... the totalcost for low-density ... sprawl ... is slightlymore than $35,000 per dwelling unit.Further, if that development is located 10miles from the sewage treatment plant, thecentral water source, the receiving body ofwater, and the major concentration ofemployment, almost $15,000 per dwelling

unit is added to the cost, for a total of$48,000 per dwelling unit ...

The cost can be reduced to less than$18,000 ... by choosing a central location,using a mix of housing types in whichsingle-family units constitute 30 percent ofthe total and apartments 70 percent, and byplanning contiguous development instead ofleapfrogging. (Frank 1989, 39)

To the extent that planned or managedgrowth fosters the more efficient patternsdescribed above—centrally located,contiguous development that includesunits at somewhat higher density—it canachieve infrastructure savings relative totraditional development.

Peiser, Richard B. 1984. "Does It Pay toPlan Suburban Growth?" Journal ofthe American Planning Association 50:419-433.

The aim of this study is to provide acomprehensive method for comparing onedevelopment pattern to another. Toachieve this aim, Peiser conducts a quasi-controlled experiment comparing theplanned development of a 7500-acre tractin southwest Houston with thehypothetical "unplanned" development ofthe same tract. Peiser patterns unplanneddevelopment after development thatoccurred to the north of Houston in anarea called Champions. The authorevaluates the capital costs associated withland development and transportation foreach of the two development alternatives.Other social costs are also examined in aqualitative analysis. Unlike previousstudies, Peiser includes non-residentialland uses in his analysis.



Peiser makes four assumptions in hiscomparison. First, total density and totalacreage of each case is assumed to beequal. Second, he assumes that eachcommunity derives the same level of totalbenefits. The relative advantage of onecommunity over the other is determinedby the differences in costs associated witheach type of development. Third, thestudy focuses upon the differences incosts of the overall community design, notthe costs associated with differences inhousing, building types, interior streets orutilities for residential subdivisions.Finally, travel costs are accounted for toand from the edge of the developmentsite.

Peiser finds that planned developmentproduces higher net benefits thanunplanned development for the three costcomponents investigated: landdevelopment, transportation, and socialissues. Overall, transportation costsprovide the greatest net benefit. However,the magnitude of the difference is small,only accounting for one to three percent oftotal costs. Further, Peiser acknowledgesthe obstacles to planning large-scaledevelopments. He highlights severalconstraints associated with suchdevelopment including the cost andavailability of financing, the labyrinthinepermitting process, and the difficulties ofmanaging large-scale projects.

Real Estate Research Corporation(RERC). 1974. The Costs of Sprawl:Environmental and Economic Costs ofAlternative Residential DevelopmentPatterns at the Urban Fringe: (VolumeI: Detailed Cost Analysis; Volume II:Literature Review and Bibliography).Washington, DC: U.S. GovernmentPrinting Office.

Among the most often cited studies in thehistory of planning literature, this three-

volume series is the final report of anearly effort to isolate the variable ofdensity from those of structural age,obsolescent layout, and low-incomepopulation, in order to measure some ofthe most important and costlyconsequences of urban form. Sponsoredjointly by the U.S. Council onEnvironmental Quality, the Department ofHousing and Urban Development, and theEnvironmental Protection Agency, theproject was carried out by the Real EstateResearch Corporation. Its centralconclusion, that low density developmentis extremely costly on energy,environment, and fiscal grounds, has beengenerally accepted as both intuitivelycorrect and accurately determined.

The basic study method was to makedetailed estimates of the costs associatedwith six hypothetical new communities—each containing 10,000 dwelling units,each housing an "average" urban fringepopulation mix, and each constructed in a"typical" environmental setting.

The report's basic conclusions are:

1) The high-density planned communitywould be optimal with reference to allfour key indicators examined: energycost, environmental impact, capitalcost, and operating cost. The low-density sprawl community would beleast desirable with reference to allfour.

2) The high-density planned communitywould require 44 percent less energythan the low-density sprawlcommunity.

3) The high-density planned communitywould generate 45 percent less airpollution.

4) The high-density planned communitywould require a capital investment 44



percent less than the low-densitysprawl community; the largestproportionate savings would be inroad and utility construction, but thelargest absolute saving would be inthe cost of residential constructionitself.

5) The operating cost of communityservices would be about 11 percentlower in the high-density plannedcommunity than in the low-densitysprawl community

A classic in its field, The Costs of Sprawlhas not failed to attract criticism. Perhapsthe most glaring limitation is that itsenergy, pollution, and capital costcomparisons all require correction. Theauthors assumed different space standardsfor the different types of dwelling units.The savings in capital and operating costscalculated in the report are mainly afunction of the difference in size, notwhere these units are located or thedensity of their development.Furthermore, the energy savings attributedin the report to high-density developmentappear significantly overstated; and sincethe estimates of air pollution reductionwas made a direct function of energysavings, these estimates must be deflatedto a similar degree.

Despite these qualifications, The Costs ofSprawl merits the close attention of thoseinterested in an analysis of urban form.Although it is important to recognize thefragility of the main conclusions of TheCosts of Sprawl, it is equally appropriateto recognize this report as a landmarkfrom which most research on theconsequences of urban form has branched.

Souza, Paul. 1995. New Capital Costs ofSprawl, Martin County, Florida.Gainesville, Fl: University of Florida.

This is a study of the capital requirementsof growth in a developing county in SouthFlorida. Souza assumes that MartinCounty's population will grow at a similarrate as the rest of South Florida. He thencalculates the density- and distance-related costs of providing certain publicinfrastructure. Costs associated withproviding roads, a potable water supply,and sanitary sewers within three differenthousing densities (3, 5, and 10 units peracre) are assessed using Martin County-specific unit costs and developmentassumptions. Souza then computes costsassociated with connecting three sites atvarying distances from the urban servicearea.

Souza finds that the provision ofinfrastructure within the lowest densitydevelopment is over 100 percent morecostly than the provision of infrastructurewithin the highest density development.He also finds that distance costs can notbe separated from density costs. Asdistance from the center increases, densitydecreases.

The provision of infrastructure to thelowest density housing pattern situated onthe site farthest from the urban servicearea is 181 percent more costly than theprovision of infrastructure to the highestdensity housing pattern situated on a sitewithin the urban service area. Low-densitysprawl is significantly more costly thanhigh-density non-sprawl. Roads comprisethe largest proportion of both density- anddistance- related costs.



Tischler & Associates. 1994. MarginalCost Analysis of Growth Alternatives—King County, Washington. Bethesda,MD: Tischler & Associates, Inc.

This paper was prepared for the KingCounty, Washington Growth ManagementPlanning Council (GMPC) to evaluateoptions for future development. Theanalysis consists of three basic land-usescenarios, with and without high capacitytransit between planned urban centers,yielding five alternatives for study.

The study identifies the costs of newdevelopment for each alternative over atwenty-year projection period, estimatingcosts and revenues associated with growthfor roads, transit, water and sewer utilities,and government administration.

The study draws the followingconclusions:

• The alternatives with moredevelopment in the urban centers andcities are more fiscally beneficialwhen roads, utilities, and general fundactivities are considered.

• The scenarios using eight centersgenerate higher net revenues for thegeneral fund and utility districts thanthose with fourteen centers, becausethe eight-center scenarios assumefewer new households in theunincorporated counties inside theurban growth area (UGA) than do thefourteen-center scenarios.

• The fourteen-center scenarios indicatenet road costs which are 11 percentbelow those of the eight-centerscenarios. Road costs (especiallyrights-of-ways) are higher for theeight-center scenarios because thesealternatives include major increases innew households and jobs, which entail

both the construction of new roadsand purchases of rights-of-ways in thematuring urban centers.

Windsor, Duane. 1979. "A Critique ofThe Costs of Sprawl." Journal of theAmerican Planning Association 45, 2:(July) 279-292.

This article is a critical review of theRERC study, The Costs of Sprawl. Thefirst part summarizes the findings fromthe RERC study. The author thencriticizes RERC for not disentanglingdensity from other factors. Windsorcriticizes RERC for using different sizeunits for different densities, i.e., smallerunits for higher densities. Because smallerunits have lower floor areas, they are lesscostly to build. This is a major reason whylarger low-density, single-family units areconsidered more costly to build and topublicly service than higher-density units.Total floor area is 44 percent lower inhigh-rise developments than single-familyneighborhoods. Differences in housingcosts and public capital costs largelyparallel these floor area differences.

Windsor recomputes the RERC analysisassuming all housing units are 1,200square feet. This computation greatlyreduces the housing and capital costadvantages of high density, though theadvantages still exist. However, thisapproach is equally unrealistic. In reality,higher-density units are indeed smaller, onaverage, so RERC is not entirely wrong.Windsor concludes that the only way toavoid the problem is to calculate resultsfor both methods on a per square footbasis and compare them.

Another criticism Windsor levels at theRERC study is that it assumes structurecosts are highest for single-family homes,



lower for high-rise construction, andlowest for walk-up units. Windsor, incontrast, believes that high-rise unitsshould have the highest costs per squarefoot. He also takes issue with RERC'sassumption that developers have tocontribute more land to the public sectorunder single-family development thanunder high-rise development.

RERC assumes that clustering in higher-density patterns results in savings ofvacant land. But, Windsor argues that ifthe model assumed that a given amount ofland is developed at different densities,the total population accommodated on theland could vary. Some ability to accountfor saved land when residents have moreland immediately around them thanpermitted by existing zoning should havebeen developed.

The author also criticizes RERC for: "theunderlying assumption that costminimization is an appropriate principlefor planning and development ... Costminimization is not a planning principleunless benefits are constant." Since RERCignores the benefit side, its conclusions ofreduced costs ignore the reduced benefits.However, the author thinks it is notnecessary to measure benefits. He claimsthe prevalence of low-density settlementsreflects the benefits to consumers:"Consumers choose to live at highdensities only where land costs are veryhigh, as in central cities." Where landcosts are lower, as in suburbs, they preferlow density environments.

The author claims that suburbs resist highdensity on several grounds, not just costs."Voters are opposed to rapid populationgrowth, the possible characteristics of newresidents, the fiscal implications, and theloss of suburban amenities like openspace, semi-rural ambiance, etc.Exclusionary land-use controls are

intended, in part, to force low-densitydevelopment; they function as a form ofgrowth management." (291)

By ignoring these nuances, says Windsor,RERC "does not properly evaluate therelative economic efficiency of alternativedevelopment patterns." (291)

York, Marie L. 1989. EncouragingCompact Development in Florida. FortLauderdale, FL: Florida AtlanticUniversity–Florida InternationalUniversity Joint Center forEnvironmental and Urban Problems.

York analyzes three areas: growthmanagement programs from around thecountry (Maine, Massachusetts, NewJersey, Oregon, Vermont); innovativedevelopment strategies by region; andproblems with redevelopment. Theanalysis of the statewide programsconsists of descriptive summaries of statepolicies designed to address increasinggrowth management problems, as well assummaries of the situations that promptedthe development of these policies. Eachstate's section concludes with a discussionof how well the policies and legislationhave worked in achieving the state's goals.Some of the difficulties, noted by York,revolve around definitions of rural versusurban uses; areas that are exempted fromthe legislation for one reason or another;property taxes as a disincentive tocompact growth; opposition fromdevelopers and municipalities; and themethods of determining urban growthboundaries.

The innovative development strategiessection of this document examines the useof urban growth boundaries (UGBs)—aproactive growth management tool tocontain, control, direct, or phase growth



close to existing urban development.Basic to this strategy is the delineation ofperimeters around urban developmentareas, within which urban densities arenormally encouraged, and outside whichurban uses and densities are discouraged.Also discussed are transfers ofdevelopment rights, point systems, andrevenue sharing as means of growthmanagement. All four of the subsectionsconclude with the experiences ofcommunities in several states that haveimplemented these strategies.

The section on problems of in-fill anddevelopment provides overviews of issuesand approaches, followed by discussionsof programs and projects in Florida andother states.

A secondary analysis of several statestudies is undertaken to quantify theimpact of UGBs on land prices. However,York admits that the data are neithercomplete nor consistent enough to drawfirm conclusions concerning the land priceimpacts of urban growth boundaries.

The report concludes with separate sets ofrecommendations for encouragingcompact development, redevelopment,and in-fill development. Therecommendations address land use, fiscal,and infrastructure issues.

FISCAL IMPACTS, EXACTIONS,AND IMPACT FEES

Altshuler, Alan. 1977. "Review of TheCosts of Sprawl," Journal of theAmerican Planning Association 43, 2:207-209.

This short article reviews the RERCreport and was published about two yearsafter the report. The review is so clear andso well done that it changed forever the

way the Costs of Sprawl report wasviewed. For purposes of simplicity,Altshuler focuses on the two extremecases analyzed by RERC—high-densitymultifamily housing and low-densitysingle-family housing. He begins bysummarizing the major findings of theRERC study. He then asks threequestions: (1) Have the results of thetheoretical analysis been calibrated againstactual community experiences? (2) Doesthe report itself fully support theconclusions stated in its summary? and (3)Are the reported advantages of high-density over low-density developmentclearly differentiated as to reason? Hisanswer to all three questions is "No"!

One key issue Altshuler raises is whetherdensity per se affects the demand forcommunity services. RERC explicitlyassumes that it does not, but Altshulerchallenges that assumption. Low-densityareas often have no sidewalks; they haveabove-ground utility lines, and infrequentstreet lights, when measured against high-density areas. The demand for publicsafety services is also likely to be lower inlow-density development. Further, RERCdoes not include any estimates of masstransit spending in its study, though suchspending would surely be higher in high-density communities that rely more onmass transit. Therefore, Altshulermaintains high-density settlements arelikely to have higher community servicecosts than low-density ones, which wouldoffset many of the savings projected forthe latter types of settlements.

According to RERC, the savings forhigh-density living versus low-densityliving are only $238 per year in operatingcosts, for a density rise from 3.5 to 19.0units per acre plus more intensiveplanning. Four-fifths of the savings areattributed to density alone. Altshulerbelieves, given the omissions mentioned



above, that this small amount wouldvanish if the analysis were done correctly.

He also makes the point, which Windsorlater picks up on, that dwelling units in thehigh-density settlement are 34 percentsmaller than those in the low-densitysettlement, and this accounts for a largepart of their differences in capital andenergy costs. Five-sixths of the heatingand air conditioning savings in high-density development is attributable tosmaller housing unit sizes.

RERC further assumes that averageannual travel per household in high-density developments is about 9,891 milesversus 19,673 miles per household in low-density developments. Generally, onlylocal trips vary by density, but RERCfallaciously attributes cost savings to theentire travel mileage. According toAltshuler, correcting this error eliminatesfour-fifths of the claimed savings in autoenergy consumption. With properanalysis, the total energy savings of high-density versus low-density development isabout 3 percent, of which only 1 percent isattributable to density alone.

Another issue Altshuler addresses iswhether higher residential density leads tohigher density in other types of land uses.He thinks not: "The case that low-densityliving is a highly expensive luxuryremains to be made" (209).

Nonetheless, the author commends RERCfor having putting forth a systematicanalysis that can serve as a starting pointfor other studies.

Altshuler, Alan A., and Jose A. Gomez-Ibanez. 1993. Regulation for Revenue:The Political Economy of Land UseExactions. Washington, DC: BrookingsInstitution.

The central issues and themes of this textrelate to government-mandated exactionspaid by real estate developers. Exactionsmay be in-kind or involve monetaryoutlays. The legal theory underlyingdevelopment exactions is thatgovernments, having reasonablydetermined that certain public needs are"attributable" to new development, mayrequire that their costs be "internalized" aspart of the development process. A keypremise of the argument for exactions isthat land development is a major cause ofescalating local infrastructure demandsand costs.

This study looks at the costs of growth inbuilt-up communities. Alternativeestimates of revenues and expenditures forthe city of San Francisco are discussed, asare approaches for allocating publicexpenditures for growth among countybusinesses and residents in MontgomeryCounty, Maryland.

This text also critiques the Real EstateResearch Corporation's The Costs ofSprawl study. As Altshuler found in his1977 study, the principal problem with theRERC study is the meaning associatedwith the cost differences. Altshuler andGomez-Ibanez argue that the degree ofvariation between the quality of housingunits from one community to another doesnot allow costs to be fairly compared. Ifconditions can not be replicated in futurestudies, community cost impacts willcontinue to be difficult to compare.

Overall, the chapter on fiscal impactanalysis reflects at least one of the authors'inexperience with this technique.



The authors rely too heavily on secondaryanalyses and critiques. Furthermore, thebook does not give enough recognition tothe role (both constructive anddestructive) of fiscal impact analysis toeither counter or abet development(sprawl) during subdivision or site planreview.

American Farmland Trust. 1986.Density-Related Public Costs.Washington, DC: AFT.

This study tests the hypothesis that, inrural areas, public costs for newresidential development exceed the publicrevenues associated with this type ofdevelopment. Using Loudoun County,Virginia data, the study attempts todevelop a methodology to estimate: (1)the net public costs (public costs minuspublic revenues) of new residentialdevelopment; and (2) how these costsvary with different development densities.

The study's methodology entails a four-step process. First, major categories ofpublic expenditures (education, health andwelfare, and safety) and public revenues(property taxes, state funds, and otherlocal taxes) are identified, based on thecounty's annual budget. Second, thedemographic profile of a 1,000-householdnew development is determined, based onsurrounding demographics, to consist of3,260 residents (including 940 school-agechildren). Third, four developmentscenarios are projected at different densitylevels, varying from 0.2 dwelling units(d.u.) per acre (rural, low-density) to 4.5d.u. per acre (suburban, high-density),while retaining the demographic profile ofthe development. Finally, fiscal impactanalysis models are run to determinepublic costs and revenues associated witheach density scenario.

The results reveal a public revenueshortfall for residential development at alldensities. However, the lowest-densitydevelopment results in a shortfall threetimes larger than the highest densitydevelopment ($2,200 per d.u. versus $700per d.u.). To cover the shortfall, thecounty would have to apply somecombination of reduced public services,higher taxes, or increased commercialzoning. In addition to imposing higherpublic costs, the lower-densitydevelopments also remove larger amountsof agricultural land—a result counter toLoudoun County's goal to preserve itsagricultural economy.

American Farmland Trust. 1992a. DoesFarmland Protection Pay? The Cost ofCommunity Services in ThreeMassachusetts Towns. Washington, DC:AFT.

This report summarizes the AmericanFarmland Trust's findings from studies ofthree Pioneer Valley, Connecticut towns(Agawan, Deerfield, and Gill). The reportaddresses the five basic steps undertakenby the studies: 1) discussions with localsponsors to define land-use categories(residential, industrial, commercial, andfarm/forest/open land); 2) the collection ofdata for each town; 3) the review of publicrevenues, allocated by land use; 4) thereview of public expenditures, allocatedby land use; 5) data analysis andcalculation of revenue-to-cost ratios.

This study is part of a series of studies onthe costs and benefits of unimproved landas it relates to a community's fiscal well-being. The bulk of support for theseanalyses comes mostly from groupsdesiring to preserve agricultural lands inperpetuity as opposed to general academicinquiry into this area.



It is important to summarize the results ofthese studies before talking aboutmethodologies. In general, they concludethat:

• Residential development does not payits own way.

• Nonresidential development does payits own way but is a magnet forresidential development.

• Open space or agricultural lands havehigher revenue-to-cost ratios than bothresidential and non-residentialdevelopment.

Several aspects of the American FarmlandTrust studies cause concern. For instance,these studies are not termed "fiscalimpact" but rather Cost of CommunityService (COCS) studies. They are notapproached in a standard cost/revenueframework, yet they proffer standardfiscal impact conclusions.

Further, no one is viewed as being athome "tending" the farm. The farmcontains no residents or workers. Thecosts/revenues for residents are deflectedto other land uses—predominantlyresidential. No costs are assigned toagricultural workers, and no highwaycosts, garbage costs, traffic costs, orhealth/social service costs are assigned tothe farm. Nor are municipal legal orelection costs factored in.

• If a reasonable cost-assignmentmethod could not be found, a"default" procedure was used, whichassigned costs by the distribution ofrevenues or by the value of property.Given the low agriculturalassessments, predictably large sharesof local costs were assigned toresidential and nonresidential uses;very small amounts were assigned toagricultural uses. The study'sconclusions were blunt: The results ofthe study show that the residential

category is being supported by theagricultural and commercial/industrialcategories. The residential sector isdemanding more in services than it iscontributing in revenues.

• This study provides a fiscal argumentfor the protection of farmland andopen space.

Despite the lopsided findings of the study,the fact remains that fiscal impactrelationships between agricultural andother land uses are not well-documented.Most of the cost and revenue calculationprocedures developed since The FiscalImpact Handbook (Burchell and Listokin1978), ignore open space and agricultureas either a significant cost or revenue. Allcosts are assigned to the residential andnonresidential sectors, and all forthcomingrevenues come exclusively fromdeveloped properties whether or not theyhave inclusive open space. Most studiesassume that open space or unimprovedlands have neither a significant negativenor positive cost/revenue impact, which isprobably an accurate assumption. Neitheragricultural nor open space lands costvery much or provide much in localrevenue.

Buchanan, Shepard C. and Bruce A.Weber. 1982. "Growth and ResidentialProperty Taxes: A Model forEstimating Direct and IndirectPopulation Impacts." Land Economics58, 3 (August): 325-337.

Buchanan and Weber's analysis is anattempt to determine the extent to whichpopulation growth affects single-familyresidential property taxes, and how theseeffects are transmitted. To answer thesequestions, the authors examine both taxrates and assessed values of properties inOregon, and the possible influence ofincreased population on each of thesevariables. Among other procedures, they



study intermediate variables, such as ageof housing stock, personal income, andpopulation density.

The authors find that up until 1979,increases in new single-family homesapparently both directly increased averagehomeowner assessments and indirectlyincreased tax rates on all properties. In1979, however, the Oregon legislatureenacted a tax relief program under whichthe average rates of increase in assessedvaluation of both residential property andall other property on a statewide basiswere limited to a maximum 5 percent peryear. After this change was enacted,assessments and tax rates slowed in theirrate of increase. The costs of servicingnew properties could no longer be"exported" to old properties beyond afixed percentage per year.

The authors suggest that similarlegislation be enacted in other stateswhere there is a quest for homeownerproperty tax relief. They assert that themodel they developed for localgovernments in Oregon is generalizable toother states with similar tax systems. Inaddition, they believe that the model couldbe adapted for use in analyzing the impactof population growth on the tax bills ofowners of all types of property.

Dougharty, Laurence, Sandra Tapella,and Gerald Sumner. 1975. MunicipalService Pricing Impacts on FiscalPosition. Santa Monica, CA.: RAND.

This report analyzes the impact ofalternative municipal service pricingpolicies on urban structure and on acommunity's financial position. A pricingpolicy is a method of allocating the cost ofa service to one sector or another of thecommunity (e.g., new residents, existingresidents, the entire community). In thisreport, primary attention is paid to pricing

policies for the capital infrastructurerequired to service new development.

The pricing structures estimated in thereport are developed using actual datafrom San Jose and Gilroy, California. SanJose is a fairly large city that has alreadyexperienced fairly rapid growth; Gilroy isa city that could potentially undergoexplosive growth. Both are believed bythe authors to be prototypical of a numberof American cities.

This analysis finds that pricing policiesallocated to the largest base of payees hasthe least overall negative effect on theeconomy of the community.

Downing, Paul, ed. 1977. Local ServicesPricing and Their Effects on UrbanSpatial Structure. Vancouver:University of British Columbia Press.

The central claim of this book is that localgovernments are increasingly turning toalternative sources to raise revenues.Many now impose user charges.

A user charge is an explicit price on theconsumption of a public service. Inaddition to being a source of revenue, usercharges are also a direct measure oftaxpayers' willingness to pay for theservices provided by local government. Auser charge makes the payment explicitand directly associated with the publicservice that is being delivered. It gives thetaxpayer a better understanding of whatthe choices are, enabling a moreintelligent decision regarding the array ofgoods provided by local government. Inaddition, the user charge performs thefunction of price. It rations demand,according to who values the service mosthighly.

The first section of this book presents therationale for developing new sources of



revenue for local government. It identifiesthe possible scope of user charges as ameans of filling this demand, and presentsa detailed examination of how a usercharge is designed and calculated. Thesecond section presents an analysis ofspatial variation in the costs of providingpublic services. The third section isconcerned with institutional systems andtheir effects on the financing and supplyof services. The fourth section argues thatwhile location may not affect costs,political bodies are willing to account forcost differences in the way they assignuser charges. The final section presentsthe author's conclusions about the findingsand relates their implications for futurepolicy.

DuPage County DevelopmentDepartment. 1989, 1991. Impacts ofDevelopment on DuPage CountyProperty Taxes. DuPage County, IL:DuPage County Regional PlanningCommission.

Five years ago, a brouhaha emerged in theChicago area involving the cost ofnonresidential uses. Debate in DuPageCounty centered on whether or notcommercial uses paid for themselves. Anumber of experts subsequently gatheredto evaluate the findings of the DuPageCounty Planning Commission's study,Impacts of Development on DuPageCounty Property Taxes.

A regression analysis by DuPage Countyinferred a strong relationship betweennonresidential development and propertytax increases. Although preservationistsleaped to defend these findings, otherspointed to the weaknesses of the study.The most convincing of the critiques tooka position in the middle—pointing out thatsome evidence backed up the findings ofDuPage County, but the evidence was notnearly as strong as had been presented inthe report.

The DuPage County report is not a classicfiscal impact analysis, but rather aregression equation in logarithmic form.The dependent variable is total propertytaxes levied; independent variablesinclude change in nonresidential firms,change in ratio of nonresidential-to-residential equalized valuation, medianresidential property tax levy, and the ratioof taxes to the total municipal equalizedassessed valuation.

Some critics of the analysis believed bothsides of the regression equation formed anidentity, whose intercorrelation preventedsolution. Some thought the researchdesign should undergo significantalteration; others thought both dependentand independent variables should berecast. DuPage County, however,continued to defend both the analysis andits results.

In reality, the analysis must be put into afiscal impact frame wherein all costs canbe compared to all revenues. In addition,the quality and quantity of services, therelative levels of tax and nontax revenue,the presence of deficient or excess servicecapacities, and the effects of other landuses in similar situations should beviewed. A number of studies with similarconclusions about nonresidential growth'simpact on property taxes have beendocumented. Most results can be traced tothe nonaccountability of elected/appointedofficials, which in turn led to significantservice increases for primarilynonresidential properties.

The DuPage study points out thatnonresidential development and itsassociated surplus fiscal revenues couldimprove service quality and quantity in acommunity. However, it may alsoincrease local expenditures. Withoutknowing the type and quantity of publicservices produced before and after thenonresidential development is put in



place, no judgment can be made aboutnonresidential development and future taxrates.

Logan, John, and Mark Schneider.1981. "Suburban MunicipalExpenditures: The Effects of BusinessActivity, Functional Responsibility, andRegional Context." Policy StudiesJournal 9: 1039-50.

The authors set out to explain thevariations in the level of suburbangovernment spending. They firstsummarize existing explanatory models,which stress either local stratification anddiscrimination, the structure of localdecision making, ecological position, orpublic choice. These models suggestvarying hypotheses about which suburbsspend more and which less. Logan andSchneider then evaluate each of thealternative models and propose majordirections for further research.

The authors conclude that each model hasits strengths and certain hypotheses fromeach is supported by the data. Certainvariables stand out, however, as havingparticularly strong explanatory power inall models. The strongest, in terms ofdetermining suburban expenditures, iseconomic function. Regardless of anyother differences among communities,suburbs with strong employment basesspend more than those with weakemployment bases.

Of nearly equal importance is the set ofservice responsibilities that a suburbanmunicipality assumes. This finding maytake historical inquiry to fully explain.

Finally, differences in SMSA structure areshown to explain part of the expenditurevariation, although no simple reason isgiven for this influence. Instead, theauthors call for additional research to be

directed towards explaining what kinds ofeconomic, historical, political, or socialsituations cause the differences insuburban expenditures detected in thisstudy.

THE EFFECTS OF GROWTHCONTROLS ON HOUSING COSTS

Katz, Lawrence, and Kenneth Rosen.1987. "Interjurisdictional Effects ofGrowth Controls on Housing Prices."Journal of Law and Economics 30(April): 149-160.

In this article, Katz and Rosen argue thatthe widespread proliferation of land useand environmental regulations, primarilyimposed by local governments, forces thehome-building industry to work within amuch more complex and often morecostly regulatory framework.

Local governments have used a widevariety of procedures to control residentialdevelopment, and these controls havebecome increasingly complex andinnovative over time. In manymunicipalities, traditional land-usecontrols have been augmented byenvironmental and fiscal impactprocedures, urban growth managementsystems, utility connection moratoria,multiple permit systems, overall growthlimitations, or a combination of thesemeasures.

Katz and Rosen examine the effects oflocal land use regulations on house pricesin the San Francisco Bay Area. They findthat land use regulations appear to havehad a substantial effect on housing prices.The authors conclude that the widespreaduse of controls in communities limitsavailable housing supply. The spread ofthese regulatory techniques tometropolitan areas outside Californiacould have substantial negative effects on



the affordability of housing in theselocations.

Parsons, George. 1992. "The Effects ofCoastal Land Use Restrictions onHousing Prices." Journal ofEnvironmental Economics andManagement (February): 25-27.

In this paper, the author estimates theeffect on housing prices of land-userestrictions for property abutting theChesapeake Bay. He examines the changein housing prices before and after theintroduction of restrictions in AnneArundel County, the most populated ofthe 16 counties that were affected. Theprice changes are compared to the changein housing prices in several inlandlocations, not affected by the restrictions,over the same monitoring period.

Parsons finds that the coastal land-userestrictions appear to have caused aconsiderable increase in housing prices.To make this assertion, he assumes that,absent the controls, the change in housingprices in coastal areas would have beenthe same as it was for inland locations.Except for land use restrictions, the authorbelieves, both locations share the samegeneral market conditions. This is asignificant assumption, considering theeffect of "water frontage" on housingprice.

Parsons also believes that the land userestrictions create winners and losers inthe local housing market. The winners arethe current owners of housing in thecommunity. The losers include currentowners of undeveloped and restrictedland, renters, and future purchasers ofhousing in the coastal community.

Parsons closes with an admonition that thelarge transfer of wealth from futureresidents to current residents through

housing price increases, and the absenceof future residents in the political processin which the restrictions were establishedraise suspicion about the fairness andefficacy of these regulations.

Propst, Luther, and Mary Schmid.1993. The Fiscal and Economic Impactsof Local Conservation and CommunityDevelopment Measures: A Review ofLiterature. Tucson, AZ: SonoranInstitute.

The authors undertook an effort to reviewabout 200 articles and other publications,with a special emphasis on those dealingwith actual experiences in realcommunities. This literature reviewincludes economic studies of the impactsof zoning and other land use regulation,and the economic impacts of acommunity's appearance, architecture, andnatural environment.

Both a subject-related analysis andannotations of the more important studiesappear in this review.

The author's general conclusion is thatenvironmentally sensitive land useplanning need not have a detrimentaleffect on real estate values, economicvitality, or the local tax base. Rather, theopposite is often true.

The authors close on the optimistic notethat the lessons contained in the studiesmay help Greater Yellowstone (nearBozeman, Montana) communitiessuccessfully manage rapid growth andchange as they choose their own futures.



Schwartz, Seymour I., David E.Hansen, and Richard Green. 1981."Suburban Growth Controls and thePrice of New Housing." Journal ofEnvironmental Economics andManagement 8 (December): 303-320.

In this paper, the authors study the effectsof suburban growth control programsupon the price of new housing. Theprograms that limit growth employ avariety of devices, including phasedzoning, reduced development densities,and increased development charges. Someprograms take an even more directapproach, by setting restrictions on thenumber of housing units permitted, or byimposing population or housing unit caps.The authors then analyze the effects ofperhaps the most direct form of control:the housing quota of Petaluma, California.

To estimate price effects, the authorscompare price changes of new single-family housing in Petaluma between 1969and 1977 (after the quota was enacted) tothe price changes in two nearbycommunities. The authors limit theanalysis to new housing because itprovides a relatively consistent basis forevaluation.

The study's results suggest that Petaluma'sgrowth control program was responsiblefor an increase in housing prices. Becauseof the complexity of the issue, however, atotally unambiguous finding is notpossible.

The authors are more confident whendescribing the effect of growth controls onthe quality of construction. Here, they findevidence that higher construction qualityand prices are attributable to growthcontrols. The authors close by calling foradditional research to analyze the effectsof growth controls on housing price.

Schwartz, Seymour I., Peter M. Zorn,and David E. Hansen. 1989. "ResearchDesign Issues and Pitfalls in GrowthControl Studies." Land Economics, 62,(August): 223-233.

In this article, the authors analyze theexperiences of Davis, California, acommunity that attempted to mitigate theeffects of growth controls on the price ofhousing. The authors seek to determinewhether Davis was successful in reducingthe expected increase in the per-unit priceof housing services due to growth control,and also to determine the extent to whichDavis was successful in limiting theexclusionary impact of growth controls onlower income households.

The authors discover that growth controlsincreased per-unit housing prices due tothe reduction in supply, but that price-mitigating programs are also effective.Furthermore, the incentives created byprice-mitigation lead developers to buildsmaller, lower quality units.

Surprisingly, the study's results also showthat growth controls increase the salesprice of older housing, but the per-unitprice declines, implying an increase in thequality of old houses sold. This isexplained by the fact that the decrease inthe quality of new housing may haveencouraged households desiring higherquality (larger) housing to turn to theolder housing market, resulting in anincrease in the demand for higher qualityolder homes.

The authors conclude that price-mitigatingmeasures are only partially successful inreducing the price effects of growthcontrols, since they mostly shift theimpacts of growth controls from the newto the old housing market.



URBAN FORM AND SPRAWL

Black, Thomas J. 1996. "The Economicsof Sprawl," Urban Land 55, 3.

For most businesses, the decentralizationof activity and improvements incommunications and transportation createa choice of location options. In addition,globalization trends (particularly in theproduction of some goods) are placingmany U.S. businesses under increasingcompetitive pressure. According to Black,low-density, dispersed urban settlementpatterns result from these powerfuleconomic forces that continue to be drivenprimarily by changing transportation costsand production requirements.

The major benefit to businesses andresidents in metropolitan areas of thedispersed settlements is reduced landrents, writes Black. Increased locationoptions and lower transportation costsmean highly competitive land marketsand, consequently, lower land prices,giving U.S. firms a significant advantageover their foreign competition. Thistranslates into more competitive pricingby U.S. firms, a lower cost of living forworkers, and higher returns on financialcapital and labor.

It is obvious that high-density urban areasmake sense only for economic activitiesthat can justify hefty rents. As analternative, extended transportationsystems and lower-cost trucktransportation also have enabled industrialand warehouse facilities to spread out toless dense locations to save money onbuilding and land costs.

Black points out that another major issuesurrounding current development patternsis the extent to which general taxpayers,especially those in built-out communities,subsidize new incremental, low-densitydevelopment. Examples include subsidies

in the form of income tax mortgage,interest, and property tax deductions;federal highway expenditures in excess ofuser taxes and fees; and state and localgovernment subsidies for infrastructureexpansion to support new development.

Views on these issues differ, and Blackends this article by calling for a realisticanalysis of the forces at work to avoidfaulty conclusions about the economics ofsprawl.

Ladd, Helen, and William Wheaton.1991. "Causes and Consequences of theChanging Urban Form." RegionalScience and Urban Economics (21): 157-162.

The papers compiled in this work wereinitially presented at a conference on theCauses and Consequences of theChanging Urban Form, October 1990, rununder the auspices of the Lincoln Instituteof Land Policy. The goal of theconference was to bring togetherempirical, theoretical, and policy-orientedeconomists to improve the understandingof the nature, causes, and implications ofpolycentrism in metropolitan areas.

Recently, older, traditionally monocentriccities, such as Boston and Chicago, havedeveloped significant suburbansubcenters. Other newer cities, such asPhoenix, Dallas, and Los Angeles areperceived as lacking in any sense of"centrality."

Two of the papers in this collectionprovide insights on the nature andfunction of employment subcenters anddecentralization; two develop and testempirical models that specifically includeelements of polycentrism. One paperprovides a new theory of subcenterformation in the context of a dynamic



model; and two more examine variousimplications of changing urban form forlabor markets.

McKee, David L. and Gerald H. Smith.1972. "Environmental Diseconomies inSuburban Expansion." The AmericanJournal of Economics and Sociology 31,2: 181-88.

In this paper, the authors set out to isolateand analyze the components of suburbansprawl. Causes of sprawl are discussed,and an attempt is made to specify theeconomic effects sprawl has on urbanareas, people, and the economy. Theauthors intention is to develop suitablepolicy recommendations for dealing withsprawl.

The authors conclude that sprawl appearsto be the result of strong market forces,and that solutions to the problem may bebeyond the abilities of public policymakers to solve. They call for moregovernment intervention, especially at thelocal level, and for action on the part ofregional planning agencies. To date, theseinstitutions have not been able to dealadequately with the problem, because ofoverlapping jurisdictions and otherpolitical considerations. The authorsfurther call for greater cooperation andsharing of resources at all levels ofgovernment to find solutions to suburbansprawl.

Mills, David E. 1981. "GrowthSpeculation and Sprawl in aMonocentric City." Journal of UrbanEconomics 10: 201-226.

Mills presents an economic theory ofsprawl in a growing, monocentric city. Heposits that where decision makers haveperfect knowledge, leapfrog developmentand discontinuous land-rent functions may

occur and be efficient in both an ex-postand ex-ante sense. Where the extent offuture growth is uncertain, decisionmakers become speculators, and thespatial pattern of development is morecomplicated. Ex-post inefficiencygenerally occurs.

In the context of Mill's formalmonocentric-city model, three land-usepatterns qualify as examples of sprawl.Leapfrog development occurs when a vonThunen ring of undeveloped landseparates rings of developed land. Thisform of sprawl involves radicaldiscontinuity. Scattered development, thesecond form of sprawl, occurs when thereare annuli with both developed(homogeneously) and undeveloped land inthem. Mixed development occurs whenthere are annuli with more than onedeveloped use. Scattered development andmixed development forms of sprawlinvolve circumferential discontinuity.

Mills provides theoretical explanations foreach form of sprawl. Leapfrogdevelopment can be explained byintertemporal planning on the part ofdecision makers who anticipate futuregrowth with certainty. The essential ideahere is similar to the notion put forth byOhls and Pines (1975), that is, that landinside of the urban fringe is sometimeswithheld from early development andpreserved for more remunerative futureoptions. Theoretical explanations forscattered and mixed development formsindicate that decision makers are uncertainabout future growth and make speculativedecisions.

Several criticisms of sprawl are cited andaddressed with evidence generated fromthe monocentric city model constructedfor this analysis.



Ohls, James C., and David Pines. 1975."Discontinuous Urban Developmentand Economic Efficiency." LandEconomics 3 (August): 224-234.

Many observers argue that discontinuousdevelopment (wherein land that is closerto urban centers is skipped over in favorof land further away) is inefficient forseveral reasons. First, this developmentpattern fails to make use of the mostaccessible land. Second, the expense ofproviding public services, such as roadsand sewage systems, to new developmentis high. In contradiction, Ohls and Pinesargue that discontinuous developmentmay be desirable and efficient in certaincases. For instance, the development ofretail and commercial services near theurban fringe must often wait for thematuration of critical scale. In rapidlyexpanding urban areas, contexts arise inwhich it may be efficient to skip over landfor a period of time in order to reserve itfor commercial uses after market scaleincreases.

This strategy has been implemented insome planned communities. In Columbia,Maryland, for example, the planners ofthis "new town" explicitly reserved vacantland in residential areas for thedevelopment of shopping clusters in thefuture—after increased residentialdensities make such shopping enterpriseseconomically feasible.

Peiser, Richard B. 1989. "Density andUrban Sprawl." Land Economics 65, 3(August): 193-204.

In this article, Peiser argues that, contraryto accepted thinking, if a free urban landmarket were allowed to function it wouldinherently promote higher-densitydevelopment. He offers theoreticalarguments and empirical evidence tosupport this thesis.

Peiser argues that uniformly low-densityurban development is inefficient, becauseit increases transportation costs, consumesexcessive amounts of land, and adds to thecost of providing and operating publicutilities and public services. Furthermore,he claims that the data show that overtime discontinuous development patternsactually promote higher density. So publicpolicies aimed at preventingdiscontinuous development may bemisguided. They may lead to developmentpatterns in which densities might be lowerthan they ordinarily would be withoutsuch a policy.

Three case studies (Dallas, TX;Montgomery County, MD; and FairfaxCounty, VA) are presented in which lotsizes are examined over time along majorarterial roadways. Higher densities (i.e.smaller lot sizes) are found in later in-filldevelopment.

Peiser concludes that polices thatencourage sequential development shouldbe avoided. Instead, he argues that acompetitive land market will achievehigher density through discontinuousdevelopment followed by later in-filldevelopment.

The Costs of Sprawl–Revisited Transportation and Travel Costs


CHAPTER

Annotations of Studies

TRANSPORTATION ANDTRAVEL COSTS

Transportation and travel costs as theyrelate to sprawl involve mode of travel,pattern of residential development anddevelopment access, density of residentialdevelopment, and location/type of non-residential development. The specifictopics that group the annotations of thissection reflect the above topical concerns.They are:

Changes in Automobile TravelThe Effects of Density on Travel ChoicesUnlimited Outward ExtensionSpatial Segregation of UsesDispersed EmploymentThe Costs of Travel

Changes in Automobile Travel includesthe works of Gordon and Richardson(1989, 1991) and Alan Pisarski (1992).The Effects of Density on Travel Choicesincludes the works of Anthony Downs(1992), Robert Cervero (1991b, 1996),Frank and Pivo (1994), John Holtzclaw(1990, 1994), and Newman andKenworthy (1989). Unlimited OutwardExtension includes the works of JudyDavis (1993) and Parsons Brinckerhoff(1997). Spatial Segregation of Usesincludes the works of Robert Cervero(1989, 1995, 1996), Reid Ewing (1994),

S. Handy (1992, 1995), Robert Kitamuraet al. (1994), and Parsons Brinckerhoff(1996c). Dispersed Employment includesthe works of Robert Cervero et al. (1996,1997). The Costs of Travel includes theworks of Apogee Research (1994), M. E.Hanson (1992), J. J. MacKenzie et al.(1992), Parsons Brinckerhoff (1996a), andMichael Voorhees (1992).

CHANGES IN AUTOMOBILETRAVEL

Gordon, Peter, Harry W. Richardson,and Myung-Jin Jun. 1991. "TheCommuting Paradox: Evidence fromthe Top Twenty." Journal of theAmerican Planning Association 57, 4:416-20.

This comparison of mean commutingtimes of residents of core counties in the20 largest U.S. metropolitan areas showsthat average trips times declined orremained the same between 1980 and1985, even as population increased inmost areas. The authors hypothesize thatconstant or declining trip times were theresult of commuters changing residencesor jobs so that their origins and

10



destinations were closer to each other orso they could travel faster on lesscongested routes. However, the AmericanHousing Survey database they use doesnot contain the information needed toconfirm or disprove this hypothesis.

Levinson, D. M., and A. Kumar. 1994."The Rational Locator: Why TravelTimes Have Remained Stable." Journalof the American Planning Association60: 319-332.

The authors compare travel diary datafrom the Washington, DC metropolitanarea for the years 1968 through 1988.They conclude that greater dispersion ofactivities has helped keep travel timesconstant. During this 20-year period, themetropolitan area became more dispersed;population grew by 30 percent,employment grew by 85 percent, and thenumber of daily motorized trips perperson increased from 2.3 to 2.8. Yet,differences of means test show that formost modes and purposes, average timesfor home-to-work and work-to-home werethe same at the beginning and end of theperiod. The authors conclude that the"locators"—households and firms—actedrationally and relocated to keep traveltimes constant.

Pisarski, Alan E. 1992. New Perspectivesin Commuting. Washington, DC: U.S.Department of Transportation. July.

Pisarski identifies trends in commutingusing data from the 1990 Census and the1990 Nationwide Personal TransportationStudy. He shows that the proportion of alltrips that were for work purposes declinedslightly, from 20.4 percent in 1983 to 20.1percent in 1990. Although the miles oftravel for work increased, average traveltimes for work trips increased by only 40seconds. This result is partly due to a 35

percent increase in the number of worktrips made in single-occupant vehicles,usually the fastest mode of travel. Theadditional 22 million people who drovealone exceeded the number of workersadded to the labor force. Meanwhile, theabsolute number of people using transitremained at about 6 million, but the shareof users declined due to populationgrowth.

Richardson, Harry W., and PeterGordon. 1989. "Counting NonworkTrips: The Missing Link inTransportation, Land Use, and UrbanPolicy." Urban Land (September): 6-12.

One recent transportation phenomenon inthe United States has been the growth innon-work travel, both during peak andoff-peak hours. Using data from theNationwide Personal TransportationStudies for 1977 and 1983, the authorsfind that the number of non-work tripsincreased three to four times faster thanwork trips during that time frame in allsizes of SMSAs. Non-work travel evenincreased faster than work travel duringthe peak periods. Richardson and Gordoncontend that suburbanization, especially inthe largest metropolitan areas, was aprincipal cause of the increase in non-work travel, although they acknowledgethat demographic and workforce changeswere probably also involved.Suburbanization of businesses means thatsuburbanites have more close-by shoppingand recreational opportunities and,therefore, may make more trips to satisfyimmediate needs rather than wait untilthey have a list of needs. The study,however, does not demonstrate either thatshopping and recreational opportunitieshave increased in suburbia or thathouseholds take more trips because ofsuch an increase. Nor does the study ruleout the effects of other factors such as



rising incomes, greater participation ofwomen in the workforce, and changes inleisure activities on non-work travelchoices.

Rossetti, M. A., and B. S. Eversole.1993. Journey to Work Trends in theUnited States and Its Major MetropolitanAreas, 1960-1990. Washington, DC:U.S. Department of Transportation.

The authors compare mean commutingtimes in 1980 and 1990 for the 39metropolitan areas with populations inexcess of one million in 1990. They findthat commuting times increased in 35 ofthe metropolitan areas, and that theincreases ranged from 0.47 percent inPhiladelphia to 13.69 percent in SanDiego. All four of the metropolitan areaswith commuting time increases of morethan 10 percent are Sunbelt cities: LosAngeles, San Diego, Sacramento, andOrlando. The only cities wherecommuting time declined are New York (-7.70 percent), Pittsburgh (-1.05 percent),New Orleans (-0.57 percent), and SaltLake City (-1.92 percent).

THE EFFECTS OF DENSITY ONTRAVEL CHOICES

Simulations

Downs, Anthony. 1992. Stuck in Traffic:Coping with Peak Hour Congestion.Washington, DC: BrookingsInstitution; and Cambridge, MA:Lincoln Institute of Land Policy.

Downs develops a hypothetical urban areamodel to test the extent to which changesin the location and density of developmentwould change average commutingdistances. The basic model uses values forthe proportion of jobs in the CBD, centralcity, suburbs, and exurbs, and commuting

distances similar to the averages for largemetropolitan areas. Different densities arecreated by varying the size of the suburbsand exurbs (and adjusting the proportionof population and jobs in each area asneeded to match the size). The studyshows that the density of growth at theurban fringe has a significant impact oncommuting distances; a move from verylow to medium densities has the greatestimpact. Increasing exurban densities from886 persons per square mile to 2,800reduces commuting distances by 8percent. An increase from 886 persons to4,363 persons per square mile decreasescommuting trip lengths by 14 percent.Beyond that, large increases in densityshorten trips by only a small amount.

Ewing, Reid. 1997b. Transportation andLand Use Innovations. Chicago, IL:Planners Press, American PlanningAssociation.

The author's goal in this book is toprovide suggestions for improvingmobility by reducing congestion andautomobile dependence. He definesmobility as "the ability [of individuals] toengage in desired activities at moderatecosts to themselves and to society."

Ewing cites two key implementationstrategies for solving mobility problems.First, according to the author, the length,mode, and frequency of trips of householdtravel is affected by residentialaccessibility—the accessibility from aperson's home to their destination; and bydestination accessibility—the accessibilityfrom one destination location to another.Better accessibility can be achieved,according to Ewing, through better land-use planning at the regional andcommunity levels.

The second strategy espoused by theauthor is travel demand management



(TDM). TDM attempts to reduce thenumber of automobiles on the roads atpeak travel times. This strategy promotessuch techniques as carpooling, staggeredwork hours, compressed work weeks, andtelecommuting. The success of TDMprograms hinges on employers'willingness to implement them and toprovide incentives to their employees toutilize them.

The author also examines how cities cancreate conditions favorable to transit,pedestrian, and bicycle use. He notes thatto accomplish this goal, these modes mustbecome more flexible and feasible. Again,Ewing concludes that land usedevelopment patterns must become moresupportive of transit and alternative modesof travel.

In conclusion, Ewing calls for a paradigmshift in land use and transportationplanning. He believes there "should beless emphasis on how fast vehicles moveand more emphasis on how well people'stravel needs are met."

Metro. 1994. Region 2040Recommended Alternative Decision Kit.Portland, OR.

This analysis of alternative urban forms ofgrowth for the Portland, Oregonmetropolitan area shows that moreconcentrated development, in conjunctionwith expansion of transit service, reducesvehicle miles of travel and use of theautomobile. This study uses one of themost advanced travel demand models inthe United States to simulatetransportation outcomes. It determinesthat under continued current developmentpatterns, the urban area would have toexpand by more than half of its currentsize over the next 50 years.

The study also tests three differentscenarios that concentrate variousamounts of growth in transit corridors,centers, and in neighboring cities. In the"Growing Out" scenario, a larger share ofsingle-family housing is built than theregion has at present, with more than one-fourth of future growth placed outside thecurrent urban growth boundary. The"Growing Up" scenario keeps all futuregrowth inside the urban growth boundaryby increasing densities and building alarger share of multifamily housing. The"Neighboring Cities" scenario movesabout one-third of the expected growth toother cities within commuting distance ofthe urban area. Not surprisingly, thehighly concentrated development of the"Growing Up" scenario produces thehighest transit use (6 percent of all trips)and the greatest reduction in VMT overbase case levels (16.7 percent). The moredispersed patterns, while consuming moreland, have lower levels of congestion.

Despite the results of this study, theability to change travel behavior islimited, because much of the capitalinfrastructure that will serve the builtenvironment for the next 50 years isalready in place. Some of the study's otherproposed changes in the way the regionsdevelop may also not be feasible toundertake for political and economicreasons.

Mobility for the 21st Century TaskForce. 1996b. Strategic Goals for the21st Century. Washington, DC:American Public Transit Association.

In 1995, the American Public TransitAssociation formed the Task Force onMobility for the 21st Century (M21). TheM21 Task Force believes that theproblems caused by urban sprawl arerapidly worsening. The Task Force



concluded that over the next fifty years,with continued existing developmentpatterns, the nation will "slip into adownward spiral of economic,environmental, and social decline."

In response, the M21 Task Force engagedin a year-long strategic planning processto develop a plan for an alternative future.As a result, the Task Force devised fourplausible scenarios for the future,developed a preferred vision for the year2050, and adopted six goals andrecommendations to make the vision areality.

The four scenarios of how the Task Forcebelieves urban development patterns mayevolve over the next fifty years follow.

Boundless Sprawl—Continued growthand unchecked urban sprawl; U.S.maintains economic growth, but centralcities decline and urban problems worsen.

Dying Cities—Continued growth andunchecked urban sprawl feed economicand social decline, causing a downwardspiral. Central cities are faced withincreased poverty, crime, and otherproblems.

Community-oriented Growth—Growthcontinues, but in the form of infill andmixed-use, pedestrian-scale communitiescentered around transit stations.

Reinventing the City—A new urbanpattern emerges following the tenets ofsustainable development. All developmentoccurs within an urban growth boundary(surrounded by greenbelts), and everylocation can be reached easily by transit.

Based on the analysis of these plausiblescenarios, the M21 Task Force developeda vision of their preferred future. Thisfuture, based on sustainable community

development, benefits the economy,environment, social equity, communitylife, and individual quality of life.

In this vision, both central cities andsuburbs thrive. Although people continueto live in suburbs, transit-orienteddevelopments (TODs) have replaced low-density suburbs as the preferredneighborhood design. In addition, theseneighborhoods allow easy pedestrian andbicycle movement. Also, TODs offer awide choice of housing type, densities,and costs.

TODs offer numerous additional benefits.They require less new infrastructure andutilize existing infrastructure andmaximum capacity, ease trafficcongestion, save commuting time, andreduce pollution.

The Task Force acknowledges that thevision is a long stretch but believes that itis achievable if the following six strategicgoals and recommendations are adopted.

1. Build on the principles of ISTEA.2. Invest in innovative sustainable

technologies.3. Create desirable land-use and

development patterns.4. Strengthen regional and metropolitan

planning and decision making.5. Shift toward true cost pricing.6. Provide creative leadership initiatives.

Empirical Studies

Cervero, Robert. 1991b. "Land Use andTravel at Suburban Activity Centers."Transportation Quarterly 45, 4: 479-491.

This article analyzes the effects of density,land-use mix, and parking characteristicson commuting behavior in suburbanactivity centers. The study uses data from



83 randomly selected buildings in sixsuburban activity centers, collected as partof a project called Travel Characteristicsof Large-Scale Suburban Activity Centers,for the National Cooperative HighwayResearch Program.

The strongest relationship evidenced inthe study was between density (measuredas the height of each building) and transituse. Having retail operations in thebuilding had only modest effects on modechoice; primarily it increased transit andwalking mode shares. Parking supply hadless effect on mode choice, probablybecause most of the office buildings hadgenerous supplies of parking spaces.

Using buildings as the unit of analysis inthis study poses some problems. Forexample, the study fails to consider othercenter characteristics that may playimportant roles in determining commutingbehavior, such as distances betweenbuildings and opportunities to shop andconduct personal business at otherlocations within the center.

Cervero, Robert, and Kara Kockelman.1996. "Travel Demand and the 3Ds:Density, Diversity, and Design."Transportation Research Digest 2, 3:199-219.

The authors claim that a host of urbandesign philosophies—new urbanism,transit-oriented development, traditionaltown planning—have gained popularity inrecent years as ways of shaping traveldemand. All share three commontransportation objectives: (1) reduce thenumber of motorized trips; (2) of trips thatare produced, increase the share that arenon-motorized; and (3) of the motorizedtrips that are produced, reduce traveldistances and increase vehicle occupancylevels. An expected outcome of weaningpeople from their cars, proponents hope,

will be a lessening of the negativeconsequences of an automobile-orientedsociety—namely, air pollution, fossil fuelconsumption, and class and socialsegregation.

Cervero and Kockelman describe hownew urbanists, neo-traditionalists, andother reform-minded designers argue forchanging three dimensions, or the 3Ds, ofthe built environment—density, diversity,and design—to achieve these objectives.While the effects of density on traveldemand have been acknowledged, theeffects of diversity and design have just aslong been ignored. This paper examinesthe connection between the 3Ds of thebuilt environment and travel demand. Ittries to sort through the relative influencesof these three dimensions after controllingfor other variables, such as travelers'demographic characteristics. It does thisby applying the technique of factoranalysis to gauge the relative influence ofeach dimension as well as their collectiveimpacts.

The research findings of this paper lendsome degree of credibility to the claims ofnew urbanists and others that compact,mixed-use, pedestrian-friendly designscan reduce vehicle trips, vehicle milestraveled (VMT) per capita, and motorizedtravel. The research suggests that theeffects of the Bay Area's builtenvironment on travel demand weremodest to moderate at best. Densitiesexerted the strongest influence onpersonal business trips. Additionally,residential neighborhoods that werespatially accessible to commercialactivities, reflected by an accessibilityindex variable, tended to averageappreciably less VMT per household.Diversity also had a modest impact ontravel demand, although where it wassignificant, its influences was somewhatstronger than that of density. Having retailactivities within neighborhoods was



most closely associated with mode choicefor work trips. Further, the dimension ofwalking quality was generally moderatelyassociated with travel demand. Finally,several specific design elements of thebuilt environment seemed to beparticularly relevant to non-work trip-making. Notably, neighborhoods withhigh shares of four-way intersections, as aproxy for grid-iron street patterns, andlimited on-street parking abuttingcommercial establishments, tended toaverage less single-occupant vehiculartravel for non-work purposes.

The researchers believe that higherdensities, diverse land uses, andpedestrian-friendly designs must co-existto a certain degree if meaningfultransportation benefits are to accrue.Having nice sidewalks, attractivelandscaping, and other pedestrianamenities in a low-density, residential-only neighborhood is unlikely to promptmany residents to walk to shops andstores. However, the synergy of the 3Ds incombination is likely to yield moreappreciable impacts.

Dunphy, R. T., and K. M. Fisher. 1994.Transportation, Congestion and Density:New Insights. Paper presented at the73rd Annual Meeting of theTransportation Research Board,Washington, DC: January.

Using data on urbanized areas fromHighway Statistics, 1990, the authorsinvestigate the relationships (usinggraphs) between density and vehicle milesof travel and travel use. They find somecorrelation between urbanized areapopulation density and transit use (26percent), but little correlation betweenvehicle miles of travel and density (8percent). However, using data from the1990 Nationwide Personal TransportationSurvey, the authors find that people in

denser areas make nearly the samenumber of daily trips as people at lowerdensities, but they drive less. At mostdensities the average number of persontrips per day is just below 4.0; only at30,000 persons/square mile or more dotrip numbers dip to 3.4 trips per day.However, the average number of trips bycar drops from about 3.5 at densitiesbelow 30,000 persons/square mile to 1.9at 30,000 persons/square mile. Peopleliving at higher densities drive lessbecause both transit and walking/bikingare more viable options. The authors,however, find a strong correlationbetween density and life cycle stage. Theycontend that demographics may be moreof a contributing factor to differences intravel behavior than density.

This study is descriptive, suggestingrelationships that need further analysiswith multivariate techniques to sort outthe relative effects of householdcharacteristics versus land-use density.The data analyzed in this study are alsoaggregate, comparing whole regionsrather than specific places within regionswhere people live and work.

Dzurik, Andrew. 1993. "TransportationCosts of Urban Sprawl: A Review of theLiterature." State Transportation PolicyInitiative. Center for UrbanTransportation Research. November.

This article cites nine studies that dealwith transportation costs and sprawl,including the classic RERC The Costs ofSprawl study (1974a), reviews byAltshuler (1977) and Windsor (1979), andseveral others. The first part of the articlecites a 1965 study—"The Nature andEconomics of Urban Sprawl" by Harveyand Clark—that defined threecharacteristics of sprawl, low-density,ribbon, and leapfrog development.



Automobile use was also viewed as thecatalyst for urban sprawl. John Kain(1967) argued, however, that any savingsfrom developing high-density areas maybe offset by higher construction costs perunit.

RERC's report is cited extensively. Itestimated that a low-density sprawlcommunity would require more than sixtimes the amount of minor streets than aplanned high-density community. Onlyroad length costs were considered asdirect costs in the analysis oftransportation variations among thesecommunities. However, two indirect costswere also considered: travel time and airpollution. The RERC report assumedtwice as much VMT in a low-densitycommunity, which accounted for a largedifference in such costs.

Dzurik's article then cites Altshuler'scriticisms of the RERC report. Altshuler'sbook The Urban Transportation System(1979) argues that the American publichas strong preferences for autotransportation and low-densitysettlements. Therefore, Americans willrefuse to live in densities high enough tobring about any changes in the problemsassociated with sprawl, which he believeshave been exaggerated anyway. Bowlercompleted a study in 1977 that showedthat "user-operated transportation"accounted for about one-seventh ofconsumer spending, a proportion thatstayed roughly constant from 1950through 1973. He argues that suburbanliving results in higher use of energy andland resources for transportation thanhigher-density living.

When the urban environment is modeledas polycentric, however, the percentage ofsuburban dwellers who increase theirtravel distances for journeys to work nolonger continues to rise, since work places

also become decentralized. Yet manymodels assume that rising commutingcosts are a major transportation cost ofsuburbanization.

Gordon also argues that because worktrips are declining as a percentage of alltrips, the relative importance ofaccessibility to workplaces as a motive forchoosing places to work and places to liveis falling. Gordon and Richardson arguethat decentralization is an antidote totraffic congestion because it scatters bothorigin and destination points and makessuburb-to-suburb trips shorter than anyother types.

BART failed to replace the auto as thepreferred means of commuting in the BayArea, in spite of its enormous cost. Wherelight rail systems have been created, citieshave experienced small gains in publictransit ridership over pure bus systems,but they have also incurred major costincreases. Light rail tends to replace bustravel more than auto travel.

Dzurik reviews the argument overcompact development. One advantagecompact development is supposed to haveover sprawl is that it uses the excesscapacity in existing infrastructures, ratherthan create a need to build newinfrastructures. This was a major source ofthe economies found in the New Jerseysprawl studies. But such savings do notalways materialize.

Dzurik discusses how much subsidy fromlocal governments goes into highways andmass transit. In Milwaukee, Wisconsin,for example, he points out, the localburden of highway costs equals 59 percentof the local property tax levy. Becauseuser fees do not pay the entire cost of autotravel, more sprawl occurs than wouldotherwise take place.



Dzurik's study claims that thetransportation costs associated with urbansprawl have not been studied in theappropriate quantitative terms. Therefore,most questions about this issue are stillunanswered. His article cites unfavorableviews of sprawl's transportation costs in12 articles and studies, with one-sentencesummaries of their major complaints.

Dzurik's article contains almost nooriginal quantitative analysis. Numerouscited studies offer contradictory evidence,and the author does not critically analyzewhy these differences exist. Further, heoffers few better approaches to researchthe areas of transportation and urbansprawl.

Frank, L. D., and Gary Pivo. 1994. TheRelationship Between Land Use andTravel Behavior in the Puget SoundRegion. WARD 351.1. Olympia, WA:Washington State Department ofTransportation.

The authors use data from the 1989Transportation Panel Survey for thecentral Puget Sound region, along withhousehold characteristics from the 1990Census, employment data from the stateemployment agency, and land-use datafrom the county assessor to identify thefactors that affect travel behavior. Theyfind that density, mix, and jobs/housingbalance are all related to travel behavior,with employment density andjobs/housing balance having the strongestrelationships. At higher densities, trips areshorter but take more time. More trips aremade using alternatives to the single-occupant vehicle. As land-use mixincreases, trip distances, times, and auto-mode shares decrease. As jobs andhousing become more balanced, tripdistances and travel times go down. Therelationships between density and modesplit are not linear. The authors identify

thresholds at which there is a substantialincrease in transit use. These thresholdsare 50-75 employees and 9-13 persons pergross acre for work trips, and 75employees and 18 persons per gross acrefor shopping trips. The use of carpooling,however, seems unrelated to urbandensities or other land-use attributes. Thestudy controls for householdcharacteristics, such as income andvehicle availability.

Gordon, Peter, A. Kumar, and HarryW. Richardson. 1989. "The Influence ofMetropolitan Spatial Structure onCommuting Time." Journal of UrbanEconomics, 26: 138-151.

The authors combine data on residentialand employment densities (residents orworkers per acre of land zoned for thatpurpose) for 82 SMSAs from twelvestates (from the U.S. Geological SurveyLANDSAT file) with census data toidentify factors that influence commutingtimes by auto and transit. Their researchfinds that lower residential densities areassociated with shorter commuting timesboth by car and by transit. For auto trips,concentration of industrial employmentleads to shorter travel times, whereasconcentration of commercial employmentincreases trip times. The clustering ofmanufacturing produces economies indriving, but the clustering of commercialactivities (such as in the CBD) producescongestion that reduces times. Othervariables (land area, income, economicstructure) have the expected positive ornegative influences, and the equations arefairly robust, explaining 61 to 87 percentof the variability in mean travel times. Asa result, the authors conclude thatpolycentric or dispersed spatial structuresreduce commuting times.

The authors' use of SMSAs as the unit ofanalysis, however, raises questions aboutwhat density means. No SMSA has



uniform density throughout. Perhapslower regional density is a proxy for ageof development, city size, or some otherfactor that influences transit use.

Holtzclaw, J. 1990. Explaining UrbanDensity and Transit Impacts on AutoUse. Paper presented to the State ofCalifornia Energy ResourcesConservation and DevelopmentCommission by Natural ResourcesDefense Council and the Sierra Club.April 19.

Holtzclaw compares the annual vehiclemiles of travel in five communities withvarious densities in the San Francisco BayArea to test whether higher residentialdensities combined with better transitservice and neighborhood shopping resultin less driving. The study finds thatdoubling residential density reducesannual vehicle miles by 20 to 30 percent.Better transit access also reduces vehicletravel.

Holtzclaw's study, however, is a cross-sectional one, that only demonstratescorrelation between density and vehiclemiles of travel. It does not show, forexample, that increasing density in aparticular neighborhood would reducevehicle miles of travel. Neither does thestudy control for income levels or othercharacteristics of households thatinfluence vehicle miles of travel.

Holtzclaw, J. 1994. Using ResidentialPatterns and Transit to Decrease AutoDependence and Costs. San Francisco,CA: Natural Resources DefenseCouncil.

Holtzclaw uses smog check odometerreadings for 28 communities in SanFrancisco, Los Angeles, San Diego, andSacramento—all with at least 20,000residents—to evaluate the relationship

between density and land use. The studyfinds that neighborhood density isnegatively related to both automobileownership rates and vehicle miles oftravel, controlling for household incomeand size. When household densitiesdouble, vehicle miles of travel decline by16 percent, controlling for such factors astransit service intensities and vehicleownership. Better access to transit alsoreduces vehicle miles of travel. Shoppingopportunities and the pedestrianenvironment, on the other hand, are notstatistically significant in explaining travelbehavior.

Although, income is controlled in thisstudy, residents could still vary by numberof children, number of workers, or othercharacteristics that influence travelbehavior.

While, this cross-sectional analysis showsa relationship between density andautomotive use in existing communities, itdoes not demonstrate that if low-densitycommunities became denser fewer tripswould be made by automobile.

Newman, Peter W. G., and Jeffrey R.Kenworthy. 1989a. Cities andAutomobile Dependence: AnInternational Sourcebook. Brookfield,VT: Gower Publishing.

Newman and Kenworthy assemble a setof data on the transportation and land-usecharacteristics of ten large U.S. cities, fiveAustralian, twelve Western Europe, threeAsian, one Canadian, and one Russiancity for the period 1950 to 1980. Usinggasoline consumption per capita as theprimary measure of automobiledependence (other measures such astransit mode share are highly correlatedwith this measure), they identify therelationship between automobiledependence and urban density. Lowdensities are associated with high



automobile dependence, and highdensities with less dependence on theautomobile. This relationship holds forregions as a whole, for inner areas (pre-World War II parts of the cities), and forouter areas. As a result, the authorsconclude that more compact cities wouldreduce automobile use.

Reviewers, however, have questioned thevalidity of using gasoline consumption asthe measure of automobile dependence,noting that many factors, such as gasprices and fleet characteristics, influencegasoline consumption. Newman andKenworthy's analysis of automobiledependence also fails to make full use ofthe data collected, employing only asingle variable—urban density—toexplain automobile use, when otherfactors are clearly involved. As a result,the role of density may be overstated.

Parsons Brinckerhoff Quade andDouglas. 1996b. "Commuter and LightRail Corridors: The Land UseConnection." In Transit and UrbanForm, Vol. 1. Washington, DC: TransitCooperative Research Program,Transportation Research Board.October.

This study updates the work of Pushkarevand Zupan (1982) by analyzing the effectsof residential densities and CBDemployment levels and densities on lightrail and commuter rail boardings. The dataare from eleven cities in the United Stateswith a total of nineteen light rail lines andsix cities with a total of fortysevencommuter rail lines. Boardings and transitservice characteristic data were providedby transit agencies. Employment andpopulation characteristics are from the1990 Census. The data are used todevelop models of light rail and commuterrail boardings and costs. The empiricalresults are then used to estimate boardings

and costs for hypothetical light rail andcommuter rail corridors.

The study finds that residential densitieshave a significant influence on light railboardings. A 10 percent increase inresidential density within two miles ofstations increases station area boardingsby 5.9 percent, holding constant otherfactors affecting ridership, such asincome. Residential densities matter lessfor commuter rail boardings. Commuterrail is a high fare mode of travel, andmany of the high-income riders comefrom low-density suburban areas somedistance from the city center.

Both the size and density of the CBDinfluence light rail ridership. A 10 percentincrease in CBD employment densityraises light rail boardings per station byabout 4.0 percent, holding constant thenumber of CBD employees, theresidential density of stations, and otherfactors affecting ridership. For commuterrail, a 10 percent increase in CBDemployment densities increases stationboardings outside the CBD by 7.1 percent.

The study concludes that light rail is mostcost-effective and efficient in the citieswith larger CBDs and denser corridors.Commuter rail works best with denseCBDs. Other factors within the control oftransit agencies, such as the availability offeeder bus service and park-and-ride lots,also influence ridership and costs.

Pushkarev, B., and J. M. Zupan. 1977.Public Transportation and Land UsePolicy. Bloomington, IN: IndianaUniversity Press.

The authors estimate the effects ofpopulation density on transit use byemploying areawide population densitiesand transit use data from 105 urbanized



areas. They show that population densityexplained 55 percent of the variation intransit use in 1960, and 66 percent in1970.

The authors also estimate the effects ofresidential density, downtown floor space,and the presence or absence of rail transitfor 27 urbanized areas. Using these factorsincreases the explanatory power of theequations, but the new variables are stillless significant than residential density inexplaining transit use. Pushkarev andZupan attribute this result to greatervariability in office floor space than inresidential densities among the areasstudied.

UNLIMITED OUTWARD EXTENSION

Davis, Judy. 1993. "The Commuting ofExurban Residents." Urban Geography,14, 1: 7-29.

A comparison of the commuting times ofworkers who bought homes in the suburbsand those who bought homes in the exurbsof Portland, Oregon, shows that theaverage exurban home buyer has acommuting trip six to seven minuteslonger than his counterpart in suburbia,controlling for occupation, income, andother household and job characteristics.The data is from a survey conducted bythe author of about 750 households thatbought and occupied homes in 1987.

Although some exurban households havecommutes similar to those of suburbanhouseholds, the average exurbaniteappears to trade off longer travel times formore space, a rural environment, lowerhousing prices, a better place to raisechildren, or some combination of thesefactors. However, exurban residents seemto sort themselves out so that those wholive close to the urban area have central

city and suburban jobs, whereas thosewho live farthest out most likely work inexurban towns.

Parsons Brinckerhoff Quade &Douglas. 1997. "TCRP Project H-13A—Draft Report: Consequences ofthe Interstate Highway System forTransit." Washington, DC: TransitCooperative Research Program,Transportation Research Board.

This report describes the intended andunintended consequences of thedevelopment of urban interstate highwaysfor transit by examining four metropolitanareas in the United States, four selectedcities in Germany, and one city in Canada.The report includes profiles of thesecommunities, their transportation systems,and the positive and negative impacts oftheir transportation choices.

The authors gathered information frompublished articles, official plans, andinterviews with officials and otherknowledgeable people in eachcommunity. They also visited each site.Their goal was to understand the historyof the development of the interstatehighway system within the urbanized areaand the interactions between high-speedlimited-access roads and changes in thetransit system and land-use patterns.

The case studies were selected to test twomajor hypotheses identified from theliterature review: that the interstatehighway program biased transportationinvestments in favor of high-speedlimited-access highways that madeautomobile travel much more attractivethan transit use; and that interstatehighways facilitated the suburbanizationof households and firms, producing apattern of development that is difficult forpublic transit to serve.



The authors present evidence from thesecase studies that confirms that thedevelopment of the interstate highwaysystem adversely affected public transit.The data show declines in transitridership, increasing difficulty inmaintaining transit service levels, and thedecentralization and dispersion ofhouseholds and jobs in case study regionswith the highest use of interstatehighways. Yet the authors correctly pointout that transit was in decline well beforethe interstate system was operational, andother factors supported development inoutlying areas, such as low property taxrates, inexpensive land, and the growth ofcompeting local governments.

The authors couch their main findingswithin other significant influences in thedecline of transit:

1) The magnitude and certainty of publicfunding has influenced modalinvestment choices. These choices, inturn, have affected regional travel.

2) Those cities whose citizens haveshown a continuous commitment to,and investment in, high-quality transitservice have strong urban centers andhigh transit use.

3) Strong, well-respected institutionsbuild and operate transit systems inthe regions where the impacts ofhighways on transit are low.

4) Active, well-organized citizen groupsmitigated the impacts of highways bysuccessfully opposing certain highwaydesigns as well as highwayconstruction itself.

5) The integration of transportation andland-use policies, plans, and projectshas mitigated the impacts ofautomobile infrastructure.

6) In cities where highways' adverseimpacts are fewest, public policiessupport the use of alternative modesof transportation.

7) In general, city centers with fewerfreeways have experienced lessadverse impacts from automobiletravel.

SPATIAL SEGREGATION OF USES(LAND-USE MIX AND URBANDESIGN)

Suburbs (Employment and ResidentialAreas)

1000 Friends of Oregon. 1996. Makingthe Land Use, Transportation, AirQuality Connection (LUTRAQ): Analysisof Alternatives. Vol. 5. Portland, OR.May.

This study compares auto-oriented versustransit-oriented alternatives of land useand transportation patterns in suburbanWashington County, in the Portland,Oregon metropolitan area. Eachalternative utilizes the same land area andhas the same overall density.

In the auto-oriented alternatives, mostnew multifamily housing and jobs are atthe urban fringe. The "no build" variationincludes few transportationimprovements, whereas the "highways"variation includes a bypass freeway andother highway improvements. With thetransit-oriented alternatives, most newmultifamily housing and jobs locate onvacant lands near transit routes. Thisalternative also takes into account transitinvestments, retrofitting of pedestrianimprovements, selected highwayimprovements, and a demandmanagement program that includesparking charges for work trips. Theregion's travel demand model, which wasenhanced to increase its sensitivity todensity and design, is



used to simulate the transportationoutcomes in 2010 of each of thealternatives.

The study finds that the package oftransit-oriented development andtransportation improvements that focus onnon-automotive modes generates thefollowing effects within the study area:

• Reduces auto ownership rates by 5percent from auto-oriented levels

• Reduces single-occupant auto use forwork trips to 58 percent compared to76 percent in auto-oriented alternatives

• More than doubles the share of worktrips made by transit over auto-orientedalternatives (18.2 percent versus 8.8percent)

• Reduces daily vehicle trips perhousehold from 7.5 to 7.2 trips

• Reduces the delay over "no build"levels by more vehicle hours thanhighway building alternative (53percent reduction versus 43 percentreduction)

• Reduces peak period vehicle hours oftravel at three times the rate that the"highway" building alternative does(15.7 percent reduction versus 5.6percent)

• Reduces daily vehicle miles of travelby 6.4 percent, whereas the "highway"building alternative increases them by1.6 percent.

One caveat, however, is important to note.The study area encompasses the fastestgrowing part of the Portland metropolitanarea. The impacts would likely be less iftransit-oriented land uses andtransportation improvements were builtthroughout the metropolitan area, sincethe remainder of the region has lessgrowth to focus toward transit-orienteddevelopments.

Activity Centers

Cambridge Systematics. 1994. TheEffects of Land Use and Travel DemandStrategies on Commuting Behavior.Washington, DC: U.S. Department ofTransportation, Federal HighwayAdministration.

This study tests the influence ofemployment site design characteristics oncommuting mode choice at suburban worksites in the Los Angeles area. The researchinvolved on-site data collection of specificurban design and land-use attributes toensure a careful calibration of theindependent variables. The results indicatethat the presence of land-use mix andcertain urban design features, such asshade trees and sidewalks, in coordinationwith demand management programs, areresponsible for increasing the percentageof work trips made by transit by three tofour percentage points. An attractiveurban environment proved to be the onlyfactor that influenced mode choice in theabsence of a travel demand program. Inother words, mixed uses and access toservices within the employment centerwere not strong enough incentives, bythemselves, to generate more commutingby transit. This study did not control forfactors such as the level of transit serviceto the site, however.

Cervero, Robert. 1989. America'sSuburban Activity Centers: The LandUse–Transportation Link. Boston:Unwin-Hyman.

Cervero compares the commutingcharacteristics of workers in 57 suburbanemployment centers. These centers allhave at least one million square feet ofoffice space, 2,000 or more workers, andare at least five miles distant from theCBD. He uses cluster analysis to identifysix types of centers—office park, office



center, large mixed-use center, moderatemixed-use center, sub-city, and largecorridor. Cervero then uses analysis ofvariance techniques to determine whetherthe center types differ in commutingcharacteristics. He concludes thatlocations of higher densities and greaterland use mix do result in more commutingby transit, ridesharing, and walking.Ridesharing is greatest in the centers withhigher densities, whereas walking isgreatest in centers with significant retailactivity and nearby multifamily housing.These denser, more mixed centers alsohave slower speeds of travel because ofgreater congestion within the centers. Thisstudy did not control for transitavailability and the quality of thepedestrian environment, however.

Cervero, Robert. 1996. "Jobs-HousingBalance Revisited." Journal of theAmerican Planning Association 62, 4:492-511.

Using data from the 1980 and 1990Censuses, Cervero compares the jobs-housing balance of the 23 largest cities inthe San Francisco Bay Area. His evidenceshows that the jobs-housing balancegenerally improved during the decade,particularly as jobs increased in formerlyhousing-rich areas. However, housing didnot grow significantly in job-rich areas,largely because zoning and growthcontrols prevented housing growth.Fifteen of the communities studiedshowed small increases in the ratio ofinternal commuting to externalcommuting. Nonetheless, about twice asmany people commuted in and out of theaverage community as commuted withinit. Thus, he concludes that despite lesssegregation of uses (measured at a grosscity-wide scale), many people continue tocommute considerable distances in partbecause of mismatches between the jobs

available in their community and the typeof housing found there.

Among other things, this descriptive studydemonstrates that the transportationconsequences of spatial segregation ofuses need more careful consideration thanjust a look at the numbers of residencesand jobs. The mismatches between theincomes of employees and housing pricesand between new jobs and housingavailability also must be considered.

Neighborhoods

Cervero, Robert, and R. Gorham. 1995."Commuting in Transit VersusAutomobile Neighborhoods." Journalof the American Planning Association 612: 210-225.

This study compares work trip modeshares and trip generation rates betweenmatched pairs of transit-oriented and auto-oriented neighborhoods. Seven of thepairs are located in the San Francisco BayArea, and six in the Los Angeles area.Transit-oriented neighborhoods aredefined as those built around streetcars orrail stations prior to 1945, which have agrid street pattern. Auto-orientedneighborhoods are those built after 1945,with little orientation to transit, and withmore curving streets and cul-de-sacs.

The neighborhood pairs in the study hadsimilar incomes and, as far as possible,similar levels of transit service. Six of theseven San Francisco pairs showed theexpected results of lower auto ownershipand more use of transit and walking forwork trips. (In one pair with a largeuniversity in the transit neighborhood, thetransit neighborhood had less incidence ofdriving alone but walking oftensubstituted for transit.) The difference inthe share of drive-alone rates betweenneighborhood pairs ranged from 2.0 to



17.5 percent of trips. The results weremore mixed in Los Angeles than in SanFrancisco, however.

The authors conclude that neighborhooddesign matters little in the Los Angelesarea because of the overwhelmingdominance of the automobile in thisregion. The results may also be muddledbecause transit service levels were lessclosely matched in the Los Angeles pairsthan they were in the San Francisco pairs.

Ewing, Reid, P. Haliyur, and G. W.Page. 1994. "Getting Around aTraditional City, a Suburban PUD, andEverything In-Between."Transportation Research Record 1466:53-62.

In this study, the authors compare sixcommunities in Palm Beach County,Florida, on the basis of work accessibility,neighborhood shopping opportunities, andpedestrian accessibility. They find littleevidence that accessibility to retail affectsmode choice or vehicle hours of travel perperson. The shortest shopping andrecreational trips occurred in a classic1970s planned-unit development (i.e., asuburban auto-oriented community)because ample stores and recreationalfacilities could be found within thecommunity. This result suggests that themix of uses is as important as the layoutof streets and other design features indetermining travel behavior.

Handy, S. 1992. "Regional VersusLocal Accessibility: Neo-TraditionalDevelopment and Its Implications forNon-Work Travel." Built Environment18, 4: 253-267.

Handy compares the shopping trip travelmodes of residents of traditional andsuburban neighborhoods in the San

Francisco Bay Area. She finds thatresidents of traditional neighborhoods,where shopping opportunities are locatednearby, make 2.75 to 5.5 times as manyshopping trips by walking as residents ofmore auto-oriented neighborhoods.Residents of both types of neighborhoodsmake about the same number of auto tripsto regional shopping malls, suggestingthat neighborhood shopping trips maysupplement rather than replace longertrips.

Handy, S. 1995. Understanding the LinkBetween Urban Form and TravelBehavior. Paper presented at the 74thAnnual Meeting of the TransportationResearch Board, Washington, DC.January.

In this study, Handy makes detailedcomparisons of non-work trips in foursuburban neighborhoods in the SanFrancisco Bay Area. A "traditional" and a"typical" suburban neighborhood areidentified in Silicon Valley, where thereare good transit connections to the rest ofthe region. Another pair is selected inSanta Rosa, on the fringe of themetropolitan area. The data used stemfrom original surveys. An analysis ofvariance shows that differences in travelbehavior do occur because of urban form,controlling for household type (i.e.,number of adults and number ofworkers). People make more shoppingtrips on foot in the "traditional"neighborhoods where the downtowns areconnected to residential neighborhoodsand offer services to those residents. It isnot clear whether these trips replace autotrips, or merely supplement them,however. What is clear, is that peoplevalue choices and on average visit morethan one grocery store and more than oneregional mall in a month, if the choicesare available. Having choice adds to



travel since trips are made to places moredistant from home.

Kitamura, R., P. L. Mokhtarian, and L.Laidet. 1994. A Micro-Analysis of LandUse and Travel in Five Neighborhoodsin the San Francisco Bay Area. Instituteof Transportation Studies, University ofCalifornia, Davis. November.

The authors studied the travel behavior ofseveral hundred families in five SanFrancisco Bay Area neighborhoods. Theareas were selected because they hadsimilar median incomes. But some hadhigh density, some low, and they varied inmix of use and access to rail transit.Three-day travel diaries were collected,and site surveys were made to identifyurban design characteristics. Modelsestimated individual travel behavior and,therefore, controlled for individualcharacteristics such as income,occupation, education, and vehicleownership. Differences in travel wereexplained both by individualcharacteristics and by land-use measures,especially residential density, publictransit accessibility, and the presence ofsidewalks. Density was most important inexplaining the share of non-motorizedtrips. Access to transit influenced thenumber of non-motorized trips and theshare of transit trips. The mix of uses wasnot a very powerful indicator of travelbehavior, but a dummy variable for place(combining all the land use attributes) wassignificant.

Overall, however, the models developedin this study had limited explanatorypower; they were able to explain onlyabout 15 percent of the variability in thenumber or share of trips by variousmodes.

Parsons Brinckerhoff Quade andDouglas. 1996c. "Influence of Land UseMix and Neighborhood Design onTransit Demand." Unpublished reportfor TCRP H-1 Project. Washington,DC: Transit Cooperative ResearchProgram, Transportation ResearchBoard. March.

In this report, separate studies examine theeffects of neighborhood land-use mix andurban design on the demand for transitand other alternatives to the automobile.

The first study uses Annual HousingSurvey data for 1985 for 11 largemetropolitan areas to compare modechoices for work trips of residents in areaswith and without easy access to a "cornerstore" or other commercial activities. Asecond study of the greater Chicago areauses transit and land-use data to identifythe factors that influence individual transittrips. The third study compares the modechoices for work and non-work trips in"traditional" and "suburban"neighborhoods in the San Francisco BayArea, using original survey data. All ofthe studies use multi-linear regressiontechniques to control for income and otherhousehold characteristics.

Overall, the studies show that the typesand mix of land uses do influence thedemand for transit, as well as the use ofnon-motorized modes. People who live inmixed-used neighborhoods have a lowerprobability of commuting by car (3 to 4percentage points), a slightly higherprobability of using transit (1 to 2percentage points), and a much higherprobability of walking or bicycling (10 to15 percentage points) for work trips. Inthe Chicago area, a 10 percent increase inresidential density is associated with an 11percent increase in the number of trips bytransit. Residents of "traditional"neighborhoods in San Francisco are more



likely to use non-automotive modes fornon-work trips than residents of"suburban" neighborhoods. Theneighborhood comparison study, however,did not find statistically significantdifferences in mode choice for work tripsbetween the two types of neighborhoods.

Moreover, all these studies found itdifficult to sort out the effects of land-usemix and urban design, because thesecharacteristics are so strongly correlatedwith density. When density is included inan equation, mix and design variablesgenerally explain little about mode choice.Each of the studies controlled forresidential characteristics such as incomeand auto ownership. Because the studiesare cross-sectional, however, they showonly correlation between land-usecharacteristics and mode choice, notcausality.

DISPERSED EMPLOYMENT

Cervero, Robert; Timothy Rood; andBruce Appleyard. 1997. "JobAccessibility as a PerformanceIndicator: An Analysis of Trends andtheir Social Policy Implications in theSan Francisco Bay Area." Institute forUrban Regional Development,University of California at Berkeley.

The authors claim that "accessibility," asan indicator of opportunities to reachdestinations efficiently, has gainedincreasing attention as a complement totransportation planning's more traditionalmobility-based measures of performance,like "average delays" and "levels ofservice." They maintain that evaluatingtransportation performance in terms ofaccessibility allows a more balancedapproach to transportation analysis andproblem-solving.

Increasing accessibility by bringing urbanactivities closer together through morecompact development and the inter-mix ofland uses, as well as by promotingteletravel, can substitute for physicalmovements. Although not a replacementfor mobility-based planning, accessibilitymeasures help gauge progress towardmeeting other regional objectives likesustainability and social equality.

The authors use census transportationplanning data to study trends in jobaccessibility between 1980 and 1990, withthe San Francisco Bay Area serving as acase context. The objectives of theanalysis are multifold: 1) The work seeksto advance the use of accessibilityindicators as inputs to long-rangetransportation planning and monitoring; 2)The work aims to enrich how jobaccessibility is measured by introducingan "occupational match" refinement; 3)The authors employ empirical measures ofjob accessibility to address the spatialmismatch question; and 4) The work callsfor more formally institutionalizing andexpanding the use of accessibilityindicators for evaluating and monitoringlong-term transportation systemperformance as well as progress towardachieving broader social welfareobjectives.

The research showed that the Bay Area'slargely market-driven patterns of regionalemployment growth failed to improve jobaccessibility among residents of theregion's poorest inner-city neighborhoods.Minority neighborhoods in the inner EastBay and parts of downtown San Franciscoaveraged the worst occupationalmismatches in terms of proximity toavailable jobs throughout the 1980s.Controlling for occupationally matchedaccessibility, educational levels, andvehicle availability, Bay Areaneighborhoods with high shares ofAfrican Americans still had



disproportionately high unemploymentrates in 1990.

Cervero, Robert and Kang-Li Wu.1996. "Subcentering and Commuting:Evidence from the San Francisco BayArea, 1980-1990." Paper presented atthe 1996 TRED Conference onTransportation and Land Use.Cambridge, MA: Lincoln Institute.October.

This paper examines the growth ofdispersed subcenters in the San FranciscoBay Area and the effects of this growth oncommuting. Cervero identifies 22employment centers with 7 or moreworkers per gross acre and 9,500 or moreemployees in 1990. Downtown SanFrancisco is the largest and most denselypopulated subcenter. Other centers are inSilicon Valley and the East Bay core area(Oakland, Berkeley, and Emeryville); 16more are located further out in suburbs.Two of the subcenters did not exist in1980. Employment in these subcentersgrew on average by 23.6 percent annuallyin the 1990s, increasing the regional shareof employment in centers from 47.5percent to 48.2 percent.

The growth of these subcenters hasproduced an increase in vehicle miles oftravel (VMT) for commuting trips. Onaverage, one-way VMT increased from7.1 to 8.7 miles during the 1980s—a 23percent increase, with the largest increasesto be found in suburban centers. Thisincrease in vehicle miles of travel islinked to both longer distances and togreater use of single-occupant vehicles. Ofthese, longer distances between home andwork had more influence on VMT, sinceoutside of downtown San Francisco andthe eastern Bay Area, the vast majority ofcommuters used cars in both 1980 and1990. Cervero estimates that more thanfour-fifths of the growth in VMT is due to

longer distances between home and work.Longer distances were especiallyimportant in increasing VMT in the moreperipheral centers.

While at least one of their previous studiessuggested that job decentralizationshortened commutes, this result has beenexplained mainly in terms of recordedtravel times, and typically measured at theaggregate, metropolitan-wide level. Thisstudy sought to refine the analysis ofspatial implications on commuting bydisaggregating data among employmentcenters, measuring highway and transitnetwork distances, and examiningcommuting behavior during the 1980-1990 window of rapid suburbanemployment growth. When combiningrefined commute distance measures withdata on shifts in modal distributions andoccupancy levels, the finding is thatemployment decentralization is associatedwith substantial increases in commuteVMT per employee. Cervero attributesthese longer distances both to regionalgrowth and to mismatches in the job andhousing markets that necessitate longcommutes.

THE COSTS OF TRAVEL

Apogee Research, Inc. 1994. The Costsof Transportation: Final Report.Conservation Law Foundation. March.

This report reviews the literature on thecosts of transportation and estimates theper-mile costs of several modes forBoston, Massachusetts, and Portland,Maine. The study divides costs into threetypes: user costs, governmental costs, andsocietal costs. Extensive data werecollected for the case study regions, in aneffort to accurately reflect the cost oftravel in these specific places. Somecosts—land loss, water pollution, solidand hazardous waste pollution, and social



isolation—could not be quantified and arenot included in the analysis.

The report estimates costs for variousmodes, in different kinds of environments.For example, it estimates that a peak-period trip in a dense part of Boston usinga single-occupant vehicle (SOV) on anexpressway costs $1.05 per mile. Of the$1.05, $0.88 are user costs (including$0.24 for travel time), $0.05 aregovernmental costs not paid by the user,and $0.12 can be regarded as societalcosts. In the off-peak period, the same tripcosts $0.89 per mile, with $0.73attributable to user costs ($0.10 for traveltime), $0.05 to governmental costs, and$0.11 to societal costs. In a low-densitysetting the peak and off-peak SOV tripsboth cost $0.71. For the SOV mode, usercosts, including travel time, vary the mostamong different settings.

By contrast, a high-occupancy vehicle(HOV) expressway trip in high-densityBoston at peak hours costs $0.58 per mile,a commuter rail trip $0.58, a rail transittrip $1.04, a bus trip $1.09, a bicycle trip$0.73, and a walking trip $2.56. Therelatively higher cost of rail transit, bus,and walking trips is primarily attributableto the added travel time. Costs in thesmaller city of Portland are generallylower for all modes and densities.

The authors believe that transportationdoes influence sprawl, and this impactshould be considered a societal cost of thetransportation system. They do not,however, measure this cost, since studieshave neither identified the full range ofthe costs of sprawl nor the proportion ofthese costs that are due to thetransportation system.

This report documents the ways that travelcosts differ with the physical environmentand the modes available. As far as

possible, costs are based on actual data forthe locations studied, although measuresof societal costs are generally taken fromnational studies.

DeCorla-Souza, Patrick; Rathi, AjayK.; and Caldwell. 1992. "NationwideInvestment Requirements for NewUrban Highway Capacity UnderAlternative Scenarios," TransportationResearch Record 1359: 57-67.

The paper provides cost estimates formaintaining the nation's urban highwaycapacity at 1985 levels of service throughthe year 2005. Besides providing thecapital costs as a function of variouslevels of travel growth projections, theauthors add a policy dimension to theirestimates. Both land-use andtransportation systems management policyconstraints are included. While significantcosts will be required despite policylevels, the paper demonstrates thatappropriate policies could contain thecapital investments required.

The cost models were run under thecondition of no additional managementimposed, then under moderate and highmanagement conditions. The policyconstraints incorporated were complex,containing several thrusts each in theland-use and traffic management areas.

Land use policies involved restrictingdevelopment to where existing capacitywas available, incentives for high densitydevelopments at commute trip terminusareas, mixed-use developments in thesuburbs, and traditional neighborhooddevelopments. On the transportationmanagement side, policies includedencouraging alternative commutingmodes, work rescheduling programs,discouraging solo commuting, andincreasing traffic control provisions.



The paper describes the process used inthe analysis. There are two phases to theprocedure: first to calculate the additionallane-miles required, then to calculate theequivalent dollars needed. The analysisyielded lane-mile increases ranging from33 percent (low growth) to 49 percent(high growth) under baseline policyconditions. For the high managementcondition, the range was from 22 percentto 34 percent. At the low-growthcondition which the authors thought wasthe more likely, policy management cutthe required increase by a third. It shouldfurther be noted that with the elevatedpolicy management, the high-growthcondition almost equaled that of lowgrowth and policy status quo condition.The capital investment needs varied from1.2 trillion dollars under the high-growth,baseline management conditions down to375 billion dollars under the low-growth,high management scenario.

The authors conclude that significantincreases in highway funding (up to apossible 1.2 trillion dollars) will berequired to maintain the 1985 levels ofservices. With the imposition of only amoderate increase in management policyand under the assumption of a lowunderlying travel growth rate, the requiredinvestment can be halved.

Delucchi, M. A. 1996. The AnnualizedSocial Cost of Motor-Vehicle Use in theU.S., 1990-1991: Summary of Theory,Data, Methods, and Results. Davis, CA:Institute of Transportation Studies.August.

In this 20-volume study, Delucchi and hiscolleagues estimate the total social cost ofautomobile use in the United States for1991. The study shows that many costfunctions are non-linear and dependentupon location. Therefore, the study'sestimates cannot be divided by total

automobile mileage or some othermeasure of use to produce an accurateaverage price to use in other studies oranalyses, although the methods may beapplied in other studies.

Delucchi divides costs into six categories:

1) personal non-monetary costs, such astravel time;

2) motor vehicle goods and servicespriced in the private sector, such asvehicle ownership, maintenance, anduse costs;

3) motor vehicle goods and servicesbundled with other goods and servicesin the private sector, such asemployer- or business-providedparking;

4) publicly provided motor vehicle goodsand services, such as roads;

5) monetary externalities of motorvehicleuse, such as accident costs not paid bythe responsible party; and

6) nonmonetary externalities of motorvehicle use, such as air pollution andglobal warming.

This report estimates that the total socialcost of motor vehicle use is between $1.88trillion and $2.839 trillion per year. Ofthese costs, 38 to 50 percent of the costsare for private-sector goods and services;21 to 22 percent of the costs are forpersonal non-monetary purposes; 13 to 21percent are for non-monetary externalities;4 to 5 percent are for monetaryexternalities; 4 to 8 percent are forbundled private-sector costs; and about 7percent are for public infrastructure andservices. Delucchi also estimates thatpayments by motor vehicle users total$109 billion to $173 billion dollars a year,which is less than the $125 to $207 billionestimate of the amount spent on publicinfrastructure and services. He argues,however, that it is not necessary for userpayments to match governmentexpenditures for efficient



provision of resources. The differencebetween taxes paid by users and theprovision of public goods and servicesrelated to motor-vehicle use must bejudged on other grounds.

Delucchi does not include urban sprawl asa cost of automobile use. He argues thatsprawl is a result of locational decisions,not motor vehicle use. Althoughtransportation systems and costs mayinfluence location decisions, he contendsthat the costs of different patterns ofdevelopment are not directly a result ofthe use of motor vehicles. Furthermore, heclaims the proper corrective action is tocharge correctly for infrastructureprovisions and other aspects of urbanform, not to change automobile prices.

Ewing, Reid. 1997. "Is Los Angeles-Style Sprawl Desirable?" and Gordon,Peter and Harry W. Richardson. 1997a."Are Compact Cities a DesirablePlanning Goal?" Journal of theAmerican Planning Association 63, 1(winter): 107-126.

These two articles published in theAmerican Planning Association Journaldebate the sprawl issue. The first articleby Reid Ewing paints sprawl asundesirable. It defines sprawl not assuburbanization per se but rather as awasteful form of outward development.Sprawl is characterized by: 1) leapfrog orscattered development; 2) commercialstrip development; or 3) large expanses oflow-density or single-use development.Ewing points out two indicators thattypify sprawl—suburban environmentsthat are difficult to access and those thatlack functional open space. Locations thatare difficult to access are those far fromthe core and from each other; locationsthat lack functional open space are definedas those where open space is totally

private and cannot be used to linkneighborhoods, buffer incompatible uses,or provide space for social interaction,recreation or civic functions.

According to Ewing, sprawl is reinforcedby consumer preference, technologicalinnovation, public transportationsubsidies, and the "more than shelter"concept of the housing market.

Costs of sprawl include increased: 1)vehicle miles traveled; 2) energyconsumed; 3) public/privateinfrastructure; 4) depletion of developableand fragile lands; and 5) psychic andsocial stress. The cures for sprawl includemore government oversight in the form ofstate and regional planning and morecompact, mixed-use cluster development.

The second article by Peter Gordon andHarry W. Richardson attempts to attack"compact cities" as an alternative tospread-out metropolitan development—or sprawl. Gordon and Richardson definecompact cities as those with 1) highdensities at a macro or metropolitanlevel; 2) even higher densities at a microor neighborhood/community level; or 3)even higher densities at a downtown orcentral city level. Gordon and Richardsonreject compact cities because: 1) peoplelike low, rather than high-density living;2) there is no real chance that at either anational or global level there will be ashortage of land; 3) there is currently anenergy glut, and therefore no need toalter residential preferences to conservefuel; 4) the automobile is the mostefficient and preferred way to accessresidential neighborhoods, and ideallysuited to spread development; 5) suburbsare not congested and by their locationhave contributed to less inner-city andinner-suburb congestion; 6) inner city(compact) and suburban (spread) workand shopping trips are compatible; 7)



agglomeration economies, once theprovince of cities, have now moved tosuburbs; and 8) central locations have nomarket and continue to decline, and theirrescue is a wasteful misallocation ofpublic funds. Given these reasons, no casecan be made for compactness as adescription of desired urban form.

Which article is right? The answer isprobably both. People appear to 1) preferthe accroutrements of suburban living butdislike strip commercial development; 2)want to distance themselves from urbanproblems, but worry about energy andland consumption; 3) like theirautomobiles but see merit in transit; 4) seea growing sophistication in suburbs butacknowledge a need for safe andfunctioning cities; and 5) travel andfunction in an environment that is lessthan efficient and less than beautiful, butvery, very comfortable.

Hanson, M. E. 1992. "AutomobileSubsidies and Land Use: Estimates andPolicy Responses." Journal of theAmerican Planning Association 58, 1:60-71.

This paper estimates the subsidies ofautomobile use in Madison, Wisconsin, amedium-sized city, in 1983. Hanson usesdata on highway costs and taxes in thecity and determines that direct subsidiesfor highway infrastructure, maintenance,and policing were equivalent to $0.024per passenger-mile or $105 per person in1983. Indirect subsidies for air and waterpollution, petroleum prices, land-useopportunity costs, and personal injurywere estimated from national data, andare, therefore, less precise than thehighway data. Nonetheless, he calculatesthat indirect subsidies were equal to$0.034 per passenger-mile or $257 perperson. In this estimation of costs, thelargest subsidies were for personal injury

(36 percent), highways (23 percent), andair pollution (15 percent).

Hanson contends that subsidization of theautomobile produces more dispersedpatterns of development than would occurotherwise. Furthermore, he claims thatsprawled development limitstransportation options by making theautomobile the only viable source oftravel.

Litman, Todd. 1995. TransportationCost Analysis: Techniques, Estimatesand Implications. Victoria, BC: VictoriaTransport Policy Institute. February.

Based on a review of existing studies,Litman estimates the cost per mile for anumber of different modes oftransportation: average car, fuel-efficientcar, electric car, van, rideshare passenger,diesel bus, electric bus/trolley,motorcycle, bicycle, walking, andtelecommuting. The report includes costestimates for 20 different factors thataffect travel choice, ranging from thecosts of operating a vehicle to the cost oflack of transportation options.

Litman estimates that for urban travelduring peak periods, a mile of travel byautomobile costs $1.33. Of this amount,$0.16 is attributable to variable vehiclecosts, $0.25 to fixed vehicle costs, $0.31to user time and risk, and $0.61 toexternal or societal costs, such aspollution and land use impacts. The samemile of travel in an urban area during theoff-peak hours costs $1.06, with $0.14attributable to various vehicle costs, $0.25to fixed vehicle costs, $0.33 to user timeand risk, and $0.34 to external or socialcosts.

Litman does not separate out governmentalcosts of travel. Those costs paid by users,such as roads built with gasoline



taxes, are considered user costs; thosepaid through general taxes, such aspolicing, are lumped in external costs. Thelargest external costs in Litman's schemeare for air pollution, accident costs notpaid by the user, the opportunity costs ofland currently used for roads, and externalcosts of energy consumption such as taxsubsidies, energy security, andenvironmental damage.

Litman contends that land-use costs are alegitimate cost of automobile use becauseauto use encourages sprawl. It requireslarge amounts of land for transportationfacilities and makes development of theurban fringe much easier. The effectsinclude loss of prime farmland andwetlands, aesthetic degradation, loss ofcommunity, and higher transportationcosts. Indeed, Litman estimates thatlanduse effects cost about 7 cents permile, compared to 33 to 35 cents per milefor owning and operating the vehicle and17 to 23 cents for travel time.

This study provides relative measures ofthe various costs of using the automobileversus other modes of travel; thecalculations are based on estimates madeby others. The data used rarely cover thefull range of modes for which the authorestimates costs. Thus, the figures in histables are often simplified. The authorattempts to monetize all costs despite thelack of hard data on many costs. Thenumbers are average estimates and do notconsider location-specific factors such asdifferences in costs for urban and ruralroad building or congestion. The types ofoutcomes that the author counts as land-use impacts of transportation are generallycounted elsewhere in an analysis of thecosts of the sprawl and should not becounted again.

MacKenzie, J. J., R. C. Dower, and D.D. T. Chen. 1992. The Going Rate: WhatIt Really Costs to Drive. Washington,DC: World Resources Institute.

This report estimates the amount spent onautomobile subsidization in the UnitedStates; it defines subsidies as costs notpaid directly by the user. According to thestudy, road users pay only about 60percent of the $53.3 billion annualgovernmental costs of building andmaintaining roads. They pay only 25percent of the $91.0 billion police, fire,and other municipal costs associated withautomobile use. Free employer-providedparking accounts for the largest portion ofthe subsidy. The authors estimate that 85percent of the $100 billion cost ofemployee parking is not paid by the user.The report also estimates that users payvirtually none of the air pollution costs(estimated to be $37 billion), securitycosts for maintaining a reliable supply ofoil ($25 billion), petroleum subsidy ($0.3billion), or noise pollution costs ($9billion). About 15 percent of accidentcosts, or $55 billion worth, are alsoestimated to be paid by someone otherthan the responsible party. The authorswere unable to estimate some costs,however, such as the opportunity costs ofland devoted to roads.

Estimates are based on data from previousstudies. Estimates of externality costs aremore speculative than other costs.

Parsons Brinckerhoff Quade andDouglas. 1996a. Cost of Travel inBoulder. City of Boulder, CO. July 15.

This study employs the methods of theApogee study; local data for governmentalcosts; and national data for societal costs;and local and national data for user coststo estimate the total cost of typical



trips by various modes within the builtenvironment of Boulder. The study isbased on actual travel times to and fromspecific locations.

The authors estimate that the cost ofcommuting to Denver (25.5 miles) is$24.61 by single-occupancy vehicle(SOV) and $15.79 by transit. The SOVtrips breaks down as follow: $19.40 foruser costs (mostly travel time); $1.16 forgovernmental costs; and $4.04 for societalcosts. The transit trip includes $10.68 foruser costs; $4.70 for governmental costs(mostly for transit provision), and $0.41for societal costs. Although, in this case,transit is a cheaper trip, for a multi-purpose shopping trip of 9.75 miles withinthe city of Boulder, an SOV trip costsmuch less than a transit trip, $11.66 versus$29.17. Transit is more expensive becauseof the time involved and because of therelatively high governmental expenses foroff-peak transit travel. For a short 2-miletrip to downtown Boulder, more optionsare considered. An SOV trip costs $4.02,a transit trip $3.43, a bike trip $1.74, and apedestrian trip $5.59. The relatively highcosts of pedestrian travel is due to thelonger time needed to complete the trip.

This study shows that travel costs varywith the environment and by type oftravel. Transit costs less for longcommutes; walking and bicycling areviable alternatives for short trips in acompact city; the car is best for linkedtrips.

Voorhees, M. T. 1992. The True Costs ofthe Automobile to Society. City ofBoulder, CO. January.

This study estimates the total annual costof automobile use in the United States in1990. The author divides costs into twomain categories: 1) the direct expenses ofautomobile ownership and use, including

the cost of highways; and 2) externalcosts, including direct monetary costs, foremergency medical care; lost economicgain due to air pollution and otherexternalities; and the opportunity costs ofusing land for roads and parking. Relyingon data from other studies, he estimatesthat in 1990, the total cost of automobileuse in the U.S. was $1.152 trillion. Thelargest costs were direct expenditures forautomobile ownership and use ($440billion, or 38 percent); land-useopportunity costs ($246 billion, or 21percent); congestion costs ($146 billion,or 13 percent); air pollution costs ($100billion, or 9 percent); and highway costs($80 billion, or 7 percent). Voorhees alsoargues that the automobile has two majorland-use impacts; it consumes largeamounts of land for roads and parking andit encourages sprawl. He does not try toestimate the costs of sprawl, however,because he lacks data and because thesecosts are already calculated in the amountof fuel consumed and other costs of usingthe automobile that result from a moredispersed pattern of development.

His external cost estimates are quitesubjective and would easily be changedby making different assumptions. The costestimate for land opportunity costs isrelatively large compared to estimates inother studies.

annotations of studiesonlinepubs.trb.org/onlinepubs/tcrp/tcrp_rpt_39-d.pdf · the costs of...

Documents