measuring urban form - alliance for community treesactrees.org/files/research/japasong.pdf ·...

Measuring Urban Form

Is Portland Winning the War on Sprawl?

Yan Song and Gerrit-Jan Knaap

Many articles on urban sprawl start with the disclaimer that sprawl isill defined, but we know it when we see it. In this article, we presentseveral measures of urban form and use these to evaluate development

patterns and trends of single-family residential neighborhoods in the Portland,Oregon, metropolitan area. In doing so, we seek not only to increase precision inthe debate on urban sprawl, but to evaluate whether the Portland metropolitanarea—internationally known for its leadership in growth management—is win-ning its war on urban sprawl.

Research Context

Portland is Oregon’s largest metropolitan area and has an international repu-tation for its growth management policies. According to Benfield et al. (),for example, “Portland has become a sort of living laboratory for efficient urbanplanning and living. The results are benefiting both the environment and theregion’s economy” (p. ).

Portland is perhaps best known for its urban growth boundary (UGB). Firstestablished in and expanded little since then, the boundary encompasses

cities, parts of three counties, and approximately . million people. Under therequirements of Oregon’s land use statutes, all land outside the UGB—withexceptions—is designated for resource use and prohibited from urban develop-ment. All land both inside and outside the UGB must be planned by the appro-priate city or county. Zoning must correspond with plans. This comprehensiveform of land use planning and management has been both celebrated and ma-ligned (Abbott, ). Proponents argue that Portland’s UGB has successfullyserved to contain urban sprawl, minimize public service costs, and protect naturalresources and open space (Nelson, ). Opponents argue that the UGB has ledto housing price inflation and helped to stifle urban growth (Staley et al., ).

Portland is also known for its light-rail transit system, established on the eastside of the metropolitan area in . The westside extension was placed in opera-tion in . To increase ridership and accommodate growth within the UGB, anumber of policies were adopted to facilitate transit-oriented development, in-

Although many have written abouturban sprawl, few have sought tomeasure it. In this article, we presentseveral quantitative measures of urbanform and compute these for neighbor-hoods of varying age in WashingtonCounty, the western portion of thePortland, Oregon, metropolitan area.Our results suggest () neighborhoods inWashington County have increased insingle-family dwelling unit density sincethe s; () internal street connectivityand pedestrian access to commercialareas and bus stops have improved sincethe early s; () external connectivitycontinues to decline; and () the mixingof land uses remains limited. We con-clude that while several measures appearto be improving, Portland’s war onsprawl is not yet won.

Yan Song is an assistant professor at theDepartment of City and RegionalPlanning at the University of NorthCarolina at Chapel Hill. She received herPh.D. in urban and regional planningfrom the University of Illinois at Ur-bana-Champaign. Gerrit-Jan Knaap isdirector of the National Center for SmartGrowth Research and Education at theUniversity of Maryland. His researchinterests include the economics andpolitics of land use planning, the efficacyof economic development instruments,and the impacts of environmental policy.

Journal of the American Planning Association,

Vol. , No. , Spring .

© American Planning Association, Chicago, IL.

cluding transit area overlay zones with minimum densityrequirements and several public/private partnerships estab-lished to encourage high-density housing and employmentgrowth around station areas. As with the UGB, the light-rail system has been controversial. Proponents claim thatthe system has been an effective vehicle for creating a lessauto-dependent urban development pattern ( Friendsof Oregon, ). Opponents claim that it has survivedonly because of substantial subsidies by the federal govern-ment (O’Toole, ).

The Portland area is also distinctive in at least oneother regard: It has the only directly elected regional gov-ernment in the United States. The roles of Metro (for-merly the Metropolitan Service District) include manage-ment of the urban growth boundary, regional land useplanning, transportation planning, and data management.Although Metro lacks authority to zone and impose sub-division regulations, it can require local governments torevise their plans and regulations if it finds that they do notserve regional goals (Metro, a). In , Metro beganwork on the Growth Concept designed to shapemetropolitan development for a -year period. To accom-modate the approximately , new residents and, additional jobs projected for the area withoutencouraging urban sprawl, the Concept encourages rede-velopment within the UGB, especially in designated urbancenters and transit corridors (Metro, ). Followingprinciples of New Urbanism, the goals of the Conceptinclude the transformation of the metropolitan area intoa multinucleated urban form, the development of a multi-modal transportation system, and the designation ofmixed-use regional and town centers (Calthorpe, ;Katz, ; Metro, ). The implementation vehicle, theUrban Growth Management Functional Plan, was adoptedin November . In it Metro set binding targets andperformance measures, such as designating small-lot sub-divisions and establishing minimum housing densities, forits subordinate cities and counties. Metro also requiredcities and counties to change their comprehensive plans toassure compliance with this plan (Metro, ).

Our study focuses on Washington County in thewestern portion of the metropolitan area, the most rapidlygrowing of the three counties within Metro’s jurisdiction.The study area contains the cities of Beaverton, Hillsboro,Tigard, Sherwood, Tualatin, King City, Cornelius, ForestGrove, and Durham (see Figure ). In response to theFunctional Plan, local governments in Washington Countyhave revised their comprehensive plans in a variety of ways(City Council of Beaverton, ; City Council of Hills-

boro, a). Some, for example, increased planned den-sities in their comprehensive plans to conform with thefunctional plan. To promote adequate traffic circulationand full street connections, the subdivision regulations ofBeaverton and Washington County state that street inter-sections should be provided at intervals of no less thaneight per mile (excluding barriers); no cul-de-sac streetslonger than feet should be designed; no more than

dwelling units should be provided on a closed-end streetsystem except when topography and barriers prevent streetextension; access to adjacent developments should beprovided through street, pedestrian, and bicycle networks;street connections between neighborhoods should beencouraged; block lengths between local and collectorsstreets should not exceed feet; and the maximumperimeter of the blocks formed by local and collectorstreets should not exceed feet (City Council ofBeaverton, ; Washington County, ). Local landuse plans and regulations also provide standards for streetdesign, sidewalk width, and shape of blocks. To increasepedestrian/bicycle accessibility, for example, local plansencourage short, directed public rights-of-way to connectresidential uses with nearby commercial services, schools,parks, transit stops, and other public facilities. Finally,local land use plans and zoning ordinances encouragemixed-use developments, which include combinationsof residential uses, offices, retail and service business, andpublic and institutional uses (City Council of Beaverton,; City Council of Hillsboro, b). For the unincor-porated areas of Washington County, the CommunityPlan Code includes transit-oriented districts that encouragea mixture of land uses and includes urban design prin-ciples, standards, and guidelines (Washington County,).

Despite clear evidence that regional and local govern-ments in the Portland metropolitan area have made signifi-cant investments in plans, regulatory structures, and publicfacilities to mitigate urban sprawl, the extent to whichdevelopment patterns in the Portland metropolitan areahave changed remains in dispute. In this article, we evalu-ate development patterns and trends in the Portland met-ropolitan area by computing several measures of urbanform and examining how these have changed over time.Our intent is not to conduct a policy analysis of the im-pacts of a particular plan or regulations. We hold thatgrowth management instruments in Portland are toonumerous, too mutually interactive, and too difficult todate stamp to isolate the impacts of any one instrument.Instead we offer a general examination of whether Port-

Song and Knaap: Measuring Urban Form

land’s suite of policy instruments implemented over anextended period have had a measurable influence on vari-ous elements of urban form.

Quantitative Research on UrbanSprawl

Longstanding attempts to quantify urban sprawl focuson the growth of suburbs relative to central cities (Chinitz,). These studies show that suburbs have grown—andcontinue to grow—more rapidly than the central citiesthey surround. Such studies provide information about thelocation of population growth but little specifically abouturban form. Another longstanding approach to the studyof sprawl focuses on density. Such work demonstrates thaturban population density gradients have fallen over time,and that the trend is both global and centuries old (Mills,). Several recent studies compare growth in urbanpopulations with growth in urbanized land areas (Fultonet al., ; Sierra Club, ) in attempts to identify“who sprawls the most.” But these also reveal only thegross density of new development.

In very recent articles, sprawl has received more de-tailed quantitative analysis. Wassmer () examineddevelopment patterns and trends in Sacramento, Califor-nia, and a small number of comparable cities. As was donein previous studies, Wassmer computed the share of met-ropolitan population that lived in the central city, thecentral county, and the urbanized area at selected points intime. He also computed and compared similar shares foremployment, retail sales, farmland, poverty rates, incomelevels, employment rates, and commuting times. He con-cluded that Sacramento exhibited a “comparably highdegree of centralization as measured by population andemployment densities at its center,” and “displayed a highlevel of negative metropolitan outcomes generally associ-ated with a high degree of urban sprawl” (Wassmer, ,p. ). Like previous studies, however, Wassmer’s researchreveals little about urban form.

Galster et al. () conducted perhaps the mostdetailed analysis of sprawl to date. They identified eightdimensions of sprawl: density, continuity, concentration,clustering, centrality, nuclearity, mixed uses, and proxim-ity. Each dimension reflects spatial relationships amongunits of analysis. These units are defined by assigning

Journal of the American Planning Association, Spring , Vol. , No.

Figure . Portland metropolitan area and study area.Source: Metro (b)

housing units from census blocks into -mile or /-mile grids. Then, using GIS, they computed each measurefor metropolitan areas (the details of the computationmethods are too complex to describe here). Finally, theynormalized all the measures and, based on equal weightsfor each normalized measure, computed overall sprawlrankings. The raw scores they reported for six of the di-mensions are presented in Table .

The results by Galster et al. () offer new andinteresting information about urban development patternsin the United States. Consistent with conventional wis-dom, Atlanta, Detroit, Miami, and Denver ranked highon the sprawl index while New York, Philadelphia, Boston,and Chicago ranked low. Contrary to conventional wis-dom, Los Angeles ranked fairly low on most measures ofsprawl. What’s more, the sprawl rankings of most citiesvaried a great deal among measures. Houston, for example,ranked the highest in clustering and lowest in nuclearity.

The measures developed by Galster et al. () pro-vide considerable information about urban form beyonddensity and suburban population growth. Specifically, theyreveal differences in spatial relationships among -milegrids within metropolitan areas and thus reveal quantita-tive insights about urban form. Yet their measures offerlittle of use to urban residents or policymakers. Shouldpublic officials in Houston, for example, be concernedor pleased that it ranked high in clustering and low innuclearity? If so, how should they respond?

Detailed and policy-relevant measures of urban formhave been developed by Eliot Allen and his colleagues atCriterion. Allen’s measures, part of a planning supportsystem called INDEX, include over measures computedat a variety of geographic scales (Allen, ). When fedinto a companion forecast model, these yield forecasts ofvehicle miles traveled, ambient air emissions, and jobs/housing balance. Both the original measures and the fore-casts they produce can then be used to evaluate alternativedevelopment scenarios, formulate plans, and monitor planimplementation.

Allen’s indicators offer a number of advantages overprevious measures. Like all good indicators, they are welldefined, relatively easy to compute (when GIS data areavailable), and easily interpreted. Perhaps most impor-tantly, however, they are policy relevant. Whereas grossmeasures of density, nuclearity, and centrality provide in-teresting information about metropolitan form, measuresof transportation options, residential proximity to retailand industrial uses, and accessibility of parks, shops, andtransit is of direct concern to citizens and policymakers.This is why INDEX serves well to evaluate alternativedevelopment proposals and land use plans.


Density Concentration Clustering Centrality Nuclearity Proximity

Atlanta . . . . . .

Boston . . . . . .

Chicago . . . . . .

Dallas . . . . . .

Denver . . . . . .

Detroit . . . . . .

Houston . . . . . .

Los Angeles . . . . . .

Miami . . . . . .

New York . . . . . .

Philadelphia . . . . . .

San Francisco . . . . . .

Mean . . . . . .

Standard deviation . . . . . .

Table . Indicators of urban sprawl (Source: Galster et al., ).

Quantitative Measures of Urban Form

To begin our evaluation of development patterns inmetropolitan Portland, we present several measures ofurban form. Many are similar to those developed by Allen.To facilitate meaningful comparison, however, we usethese attributes to measure the urban form of neighbor-hoods. A definition of each measure and how it was com-puted is provided below. Later we return to a discussionof neighborhoods.

Street Design and Circulation SystemsAccording to critics of sprawl, contemporary suburban

developments contain too many winding streets and cul-de-sacs, creating blocks that are too big and thus lackconnectivity. According to this point of view, better con-nectivity leads to more walking and biking, fewer vehiclemiles traveled, better air quality, and a greater sense ofcommunity among residents (Benfield et al., ). Ourmeasures of connectivity involve the number of nodes andintersections, the distance between points of access intothe neighborhood, the number and lengths of blocks, andthe lengths of cul-de-sacs. Four measure connectivitywithin a neighborhood; one measures connectivity be-tween neighborhoods:

• Int_Connectivity—number of street intersectionsdivided by sum of the number of intersections and thenumber of cul-de-sacs; the higher the ratio, the greaterthe internal connectivity.

• Blocks_Peri—median perimeter of blocks; the smallerthe perimeter, the greater the internal connectivity.

• Blocks—number of blocks divided by number of hous-ing units; the fewer the blocks the greater the internalconnectivity.

• Length_Cul-De-Sac—median length of cul-de-sacs; theshorter the cul-de-sacs, the greater the internal connec-tivity.

• Ext_Connectivity—median distance between Ingress/Egress (access) points in feet; the shorter the distance,the greater the external connectivity.

DensityAccording to critics of sprawl, contemporary urban

development is dominated by single-family dwelling units(SFDUs) on large lots. They believe this low-density de-velopment increases automobile dependence, consumesfarmland, and raises the cost of public infrastructure(American Planning Association [APA], ). We offer

three measures of single-family development density: SFDUlot size, density, and floor space. (Note that we excludedmultifamily units in computing density measures becausethe data were not available. We speculate that the inclusionof multifamily units would largely affect total residentialdwelling-unit density, since multifamily residential devel-opment in higher densities is an important growth man-agement strategy in the area.)

• Lot_Size—median lot size of SFDUs in the neighbor-hood; the smaller the lot size, the higher the density.

• SFDU_Density—single-family dwelling units dividedby the residential area of the neighborhood; the higherthe ratio, the higher the density.

• Floor_Space—median floor space of SFDUs in theneighborhood; the smaller the floor space, the higherthe density.

Land Use MixAccording to critics of sprawl, contemporary urban

developments are homogeneous and lack a mix of landuses. They believe that greater mixing of uses facilitateswalking and biking, lowers vehicle miles traveled, improvesair quality, and enhances urban aesthetics (APA, ).We offer two measures of land use mix: One measures theactual mix of nonresidential land uses in the neighbor-hood; the other measures the mix of zoned nonresidentialland uses.

• Mix_Actual—acres of commercial, industrial, andpublic land uses in the neighborhood divided by thenumber of housing units; the higher the ratio, thegreater the land use mix.

• Mix_Zoned—acres of land zoned for central commer-cial, general commercial, neighborhood commercial,office commercial, industrial, and mixed land uses inthe neighborhood divided by the number of housingunits; the higher the ratio, the greater the mix.

AccessibilityAccording to critics of sprawl, contemporary urban

development is characterized by too much separationbetween uses, creating travel distances that are too long(APA, ). Here we offer three measures of accessibility:distance to commercial uses, distance to a bus stop, anddistance to a public park. Each is measured as the mediandistance from the centroid of every single-family parcel inthe neighborhood to the centroid of the nearest commer-cial use, bus stop, or public park.


• Com_Dis—median distance to the nearest commercialuse; the shorter the distance, the greater the accessibil-ity.

• Bus_Dis—median distance to the nearest bus stop; theshorter the distance, the greater the accessibility.

• Park_Dis—median distance to the nearest park; theshorter the distance, the greater the accessibility.

Pedestrian AccessAccording to critics of sprawl, contemporary develop-

ment patterns create great distances between destinationsand thus discourage walking. They believe that pedestrianaccess encourages residents to walk, lowers vehicle milestraveled, and improves human health (Frank & Engelke,). Pedestrian access is widely accepted as /-milenetwork distance (Duany & Plater-Zyberk, ). Thuswe measure pedestrian access as the percentage of single-family homes that are within /-mile network distance(i.e., walking distance) of all existing commercial uses orbus stops.

• Ped_Com—percentage of SFDUs within / mile of allexisting commercial uses; the higher the percentage,the greater the pedestrian access.

• Ped_Transit—percentage of SFDUs within / mile ofall existing bus stops; the higher the percentage, thegreater the pedestrian access.

Definition of Neighborhoods

The neighborhood has long been regarded as the basic building block of urban form. What constitutes aneighborhood is disputed. Lacking clear theoreticalguidance, we explored three data-driven alternatives:census tracts, block groups, and subblock groups. Weobserve that boundaries of block groups and subblockgroups, in most cases, coincide with wide and busy arteri-als that divide an area into neighborhoods with distinctlydifferent physical and social characteristics. To chooseamong these, we computed several measures of urbanform at these three levels of census geography. That is,we calculated measures of urban form for WashingtonCounty’s census tracts, block groups, and

subblock groups (subblock groups are formed by furtherpartitioning block groups by major arterials). Then wedetermined the age of the neighborhood defined at eachlevel by the median “year built” attribute of all single-family units it contained.

Figures , , and illustrate the relationships betweeninternal connectivity and median year built measured atthe census tract, block group, and subblock group level,respectively. As demonstrated by observation and statisticaltests, “neighborhoods” defined at the census tract level(Figure ) provide less information about changes in urbanform over time than neighborhoods defined at the blockgroup level (Figure ). The former have less variation inconnectivity, due to aggregation, and thus demonstratesmaller changes over time. Neighborhoods measured at thesubblock group level (Figure ), however, reveal little moreinformation than those measured at the block group level.Thus we chose to use block groups as neighborhoods anddivided the study area into neighborhoods for theremaining analysis.

Measures of Urban Form for TwoDifferent Neighborhoods

To test our measures of urban form we first selectedtwo distinctly different neighborhoods from the inthe study area (see Figure ). The first is the Orenco Sta-tion neighborhood, located next to an Intel plant and alight rail line. This development has gained a great dealof notoriety as an exemplar of a successful New Urbanistdevelopment (National Association of Homebuilders[NAH], ). The second is the Forest Glen neighbor-hood, located in southern Beaverton. In Forest Glen,measures of urban form lie near the median value of allneighborhoods in Washington County, and thus it repre-sents a “typical” neighborhood in our study area. As shownin Table , Orenco Station and Forest Glen differ signifi-cantly in most measures of urban form. Orenco Stationhas better internal street connectivity; more mixing of landuses within the neighborhood; better pedestrian access toparks, commercial areas, and bus stops; but lower externalconnectivity than Forest Glen.

This simple comparison of two neighborhoods doesnot, of course, provide conclusive evidence that Orenco isa better place to live than Forest Glen. It does, however,demonstrate that differences in urban form can be meas-ured and that such measures appear to capture differencespertinent to the current debate on urban sprawl. Theextent to which any characteristic of a neighborhood isdesirable, and the extent to which one desirable character-istic should be traded off against others, remain matters ofpreference we do not resolve here.


Comparing the Urban Form ofNeighborhoods Built at DifferentTimes

To examine trends in urban form in our study area,we present plots of data and the results of regressions thathave measures of urban form as the dependent variableand the age of the neighborhood as the only independentvariable. Though statistical analysis is necessary to iden-tify statistically significant relationships, the plots revealintuitive information that statistical analyses alone oftenobscure, especially for new, unfamiliar quantitative meas-ures. For economy of exposition, however, we representplots of data for only one measure of each type of urbanform indicator. To illustrate where a New Urbanist anda traditional neighborhood lie in the distribution of eachmeasure, we indicate values for the Orenco Station andForest Glen neighborhoods with a star and triangle, respec-tively. The coefficient on age in the regression analysis tellsus if the indicator measure is increasing or decreasing overtime. We also tested to see whether the relationship be-tween the indicator measure and age changed to a signifi-cant degree at any point over the study period. The resultsof the regression analyses are presented in Table .

Table lists the measures of urban form and theintercept and slope coefficients (and corresponding t-statistics) of the following equations:


Figure . Internal connectivity by age of neighborhood at census tractlevel.

Figure . Internal connectivity by age of neighborhood at block grouplevel.

Figure . Internal connectivity by age of neighborhood at subblockgroup level.

Median year built

.

.

.

.

.

.

Internal connectivity of census tracts

Median year built

.

.

.

.

.

.

Internal connectivity of subblock groups

Median year built

.

.

.

.

.

.

.

Internal connectivity of block groups


Figure . Forest Glen (upper) and Orenco Station (lower) residential neighborhoods.Source: Metro (b)

Urban form measure Forest Glen Orenco Station

Internal connectivity (intersections / [intersections + cul-de-sacs]) . .

External connectivity (median distance between access points in feet) ,

Median block size (median perimeter in feet) ,

Number of blocks per SFDU . .Median length of cul-de-sacs (feet)

Actual nonresidential area per SFDU (sq. ft.) ,

Zoned nonresidential area per SFDU (sq. ft.) ,

Median distance to nearest commercial use (feet) ,

Median distance to nearest park (feet) ,

Median distance to nearest bus stop (feet) ,

Percentage of SFDUs within / mile of all existing commercial uses . .

Percentage of SFDUs within / mile of all existing bus stops . .

Table . Urban form measures for Forest Glen and Orenco Station neighborhoods.


Data measurement periodst test

1940–2000 1940–1989 1990–2000 for Year of equality trend

α β α β

–X α β

–X of means change

Int_Connectivity . −. . −. . −. . . −.

(.)*** (−.)*** (.)*** (−.)*** (−.)*** (.)***

Blocks_Peri −. . −. . . −. . −.*

(−.) (.) (−.)*** (.)*** (.)** (−.)**

Blocks . −. . −. . −. . . −.

(.)*** (−.)*** (.)*** (−.)*** (−.)** (.)**

Length_Cul-De-Sac . −.(.)*** (−.)***

Ext_Connectivity −. .(−.)*** (.)***

Lot_Size −.(.)*** (−.)***

SFDU_Density −. . −. . . − . . .***

(−.)*** (.)*** (−.)* (.)** (−.)*** (.)***

Floor_Space −. .(−.)*** (.)***

Mix_Actual . −.(.) (−.)

Mix_Zoned −.(.) (−.)

Com_Dis −. . − . . −. . .***

(−.)*** (.)*** (−.)*** (.)*** (.)*** (−.)***

Bus_Dis −. .(−.)*** (.)***

Park_Dis . −.(.) (−.)

Ped_Com . −. . −. . −. . . −.**

(.)*** (−.)*** (.)*** (−.)*** (−.)* (.)*

Ped_Transit . −. . −. . −. . . −.***

(.)*** (−.)*** (.)*** (−.)*** (−.)** (−.)**

* p < . ** p < . *** p < .

α = intercept; β = slope;–X = average

Table . Regression results for all urban form measures.

Y = α + βt + εand

Y = α + βt + α + βt + εCoefficients for the second equation are presented only

if Chow tests demonstrated that α and, α and β and β

differ significantly. That is, Table presents the resultsof two sets of coefficients if tests reveal that the trend inurban form changed over time. For the measures that havebeen identified with a trend change, Table also presentsthe mean values of the measures in two periods and theresults of t tests for the equality of means. The final col-umn of Table lists the year in which the trend changed—if indeed it did change.

ConnectivityFigure illustrates the relationship between internal

connectivity and the age of the neighborhood. From thisfigure, it is clear that neighborhoods developed in thes and s had higher internal connectivity thanthose developed more recently. Further, the results revealthat the internal connectivity fell fairly consistently as theneighborhoods aged from to . Around ,however, the trend appears to change, such that neighbor-hoods developed in had internal connectivity meas-ures about equal to those developed in the s. Further-more, Chow tests confirm that the trend in connectivitychanged significantly in . As shown in Table , inter-nal connectivity fell by approximately . per year until; after , internal connectivity rose by approxi-mately . per year.

As also shown in Table , similar changes in trendsoccurred in the number of blocks and the length of blockperimeters. Until the early s, the number of blocks inneighborhoods was falling while the perimeters of blockswere rising. Beginning in the early s, however, thenumber of blocks began to rise while the perimeters ofblocks began to fall. Somewhat surprisingly, the lengthsof cul-de-sacs fell throughout the postwar period. Thetrends in cul-de-sacs notwithstanding, these results suggestthat the street network in neighborhoods built during thes began to exhibit a pattern less characteristic of urbansprawl. The differences in means, however, in all the meas-ures of internal connectivity before and after , are notstatistically significant.

The index of external connectivity, however, exhibits adifferent pattern. As shown in Figure , external connectiv-ity appears to have fallen over the entire postwar period.Indeed regression analysis reveals that the measure of ex-ternal connectivity fell by about . feet per year. That is,

the distance between access points into the neighborhoodincreased by approximately feet per year. Again, it isnoteworthy that even the Orenco Station neighborhoodranks low in external connectivity.

Development DensityFigure illustrates the relationship between single-

family lot size and neighborhood age. The figure suggeststhat lot sizes have fallen fairly consistently over time froman average of , square feet in to , squarefeet in . Figure also shows that while the rate ofdecline does not appear to accelerate, the number of ex-tremely low-density neighborhoods diminished after .As shown in Figure , the SFDU density increased from anaverage of dwelling units per acre in to an averageof in . The rate of increase in SFDU density accel-erated after . This, however, may reflect a decline inthe area of the neighborhood not developed for single-family use. Not surprisingly, the floor space per SFDUhas risen consistently over time, from an average of ,

square feet in to an average of , square feet in, as shown in Figure . But some neighborhoodscontinue to have homes with less than , square feet offloor space. These results suggest that SFDUs have in-creased in size and density over the entire postwar period,indicating the presence of smaller lots and larger houses.The rate of increase in density of single-family units in aneighborhood accelerated in the early s, but this may


Figure . Internal street connectivity by age of neighborhood.

.

.

.

.

.

.

.

Internalconnectivity

Median year builtForest Glen Orenco Station

reflect higher proportions of land developed for single-family use and the elimination of extremely low-densitydevelopments. Nevertheless, the mean density in theperiod after was significantly higher than the meandensity of development before .

Land Use MixTo explore relationships between land use mix and age

of neighborhood, we computed the number of square feetin each neighborhood not used or zoned for residentialpurposes per SFDU. Actual nonresidential land uses ofneighborhoods built at different times are presented inFigure . As shown, the range is large, with no apparentsystematic relationship, which is confirmed in Table .


Figure . External connectivity by age of neighborhood.

Median distance betweenaccess points (feet)

Figure . SFDU density by age of neighborhood.

SFDU density

Figure . Median SFDU floorspace by age of neighborhood.

Median SFDU floorspace (sq. ft.)

Figure . Single-family lot size by age of neighborhood.

Median single-family lot size (sq. ft.)





The same is true for zoned land uses. These results suggestthere has been little change in the zoned or actual mix ofland uses in these neighborhoods over time. The fact thatsingle-family neighborhoods generally lack mixed uses isperhaps a vivid reflection of NIMBYism.

AccessibilityTo explore associations between accessibility and age

of neighborhood, we computed the median value of thedistance from every SFDU to several potential destinationsin each neighborhood, including commercial land uses,public parks, and public transit. As shown in Figure ,distance to the nearest commercial land use has risen untilfairly recently. As shown in Table , distances to commer-cial uses indeed rose until and then began to fall.Once again, however, the recent fall appears to reflect theelimination of outliers more than a fall in the average.Indeed, the average distance to commercial uses remainshigher after than before. As also shown in Table ,distance to the nearest bus stop is greater in newer neigh-borhoods, while distance to public parks appears unrelatedto neighborhood age. Neither trend changed over thestudy period.

Pedestrian AccessTo assess pedestrian access, we calculated the percent-

age of SFDUs in each neighborhood located within /mile (walking distance) of all existing commercial uses andbus stops. As shown in Figure , the percentage of SFDUswithin / mile of commercial uses falls over time. Theresults in Table , however, suggest that this relationshipchanged in . After , the percentage of dwellingswithin walking distance to commercial uses rose. Figure suggests, however, that this result reflects the influence of afew highly accessible neighborhoods, including OrencoStation. Access to bus stops exhibits a similar trend, thoughthe change occurred years later. However, the meanpercentage of homes with access to both commercial usesand bus stops was lower after than before.

Study Limitations

For several reasons the research results must be inter-preted with some caution. First, because of the lack ofimportant data on multifamily developments, such as thenumber of units in a multifamily building and the year itwas built, we were not able to include multifamily units inour urban form measures. The exclusion of multifamily

developments affects our results in several ways. Our streetnetwork design and circulation measures are based onstreet networks only; therefore, the exclusion of multifam-ily housing would not have any impact on these measures.Since the land use mix measures treat all residential areasthe same, the omission of multifamily units would also notaffect these results. For accessibility measures, however, the


Figure . Actual nonresidential area per SFDU by age of neighborhood.

Median nonresidential area (sq. ft.)

Figure . Median distance to nearest commercial use by age ofneighborhood.

Median distance to nearestcommercial use (feet)



exclusion of multifamily units could affect our results. Weobserved that multifamily units are located closer to com-mercial uses, parks, and bus stops than single-family units.

For our neighborhoods, the median distances frommultifamily parcels to the nearest commercial areas, parks,and bus stops are , , and feet, respectively, com-pared to ,, , and , feet for single-family units.For measures of density, our omission of multifamily unitsalso has potentially significant impacts. If we measureddensity in all residential units, instead of only single-familyunits, we suspect that we might find an even more rapidlyrising trend over time. For pedestrian access measures, theinclusion of multifamily units might also show a morepromising trend. We observed that more multifamilyparcels are located at shorter distances to commercial usesand bus stops than singe-family units. In our study area,% of multifamily parcels are within / mile of commer-cial uses, compared to only % of single-family units;% of multifamily parcels are within / mile of busstops, compared to only % of single-family units. Wepropose to do more research to include multifamily unitsin measuring urban form when relevant data becomeavailable.

Second, it is possible that some of our measures—suchas commercial land uses, density, and proximity to busstops—reflect redevelopments or changes in bus routes.As our measure of neighborhood age is based on the me-dian age of single-family homes in the neighborhood, it is

possible that, for example, a neighborhood first developedin the s but partially redeveloped for commercial usein recent years could cause us to overestimate the mixtureof land uses in the s.

Third, some of our results may reflect the influenceof outliers on both ends. For example, some of the in-creases in density may reflect the elimination of extremelylow-density neighborhoods, and some of the increases inpedestrian accessibility may reflect the influence of a fewrecently developed New Urbanist neighborhoods. Theplots of data reveal that the median values of several meas-ures have not changed a great deal. Even so, this does notdiminish the statistically significant changes in trends. Infact, it is quite likely that the absorption of land within theUGB has made extremely low-density development infeasi-ble. And it is quite likely that several recent new urbanistdevelopments have not changed the form of the typicalsuburban development but have altered average county-wide trends.

Summary & Conclusions

In this article, we presented several measures of urbanform and computed them for neighborhoods in Washing-ton County, Oregon. We first tested these measures in twoselect neighborhoods and found that they appear to meas-ure meaningful differences between a typical neighborhood


Figure . Pedestrian access to commercial use by age of neighborhood.

.

.

.

.

.

.

−.

Percentage of SFDUs within 1/4 mile of commercial uses

Figure . Pedestrian access to bus stops by age of neighborhood.

.

.

.

.

.

.

.

−.

Percentage of SFDUs within 1/4 mile of bus stops



and a highly touted New Urbanist neighborhood. We thenexamined whether our measures of urban form variedsystematically with the age of neighborhoods.

We found systematic changes over time in most meas-ures of urban form, excluding measures of land use mixand distance to parks. Further, we found that many ofthese measures improved in the early s. Specifically,we found increases in measures of connectivity, pedestrianaccess, and density—especially after . The resultssuggest that neighborhoods in Washington County arebecoming better internally connected, more pedestrianfriendly, and denser but remain relatively homogeneousin land uses. Taken at face value, the results suggest thateither market forces or the many growth managementpolicies in the Portland metropolitan area are causing sig-nificant changes in urban form. As mentioned before, weemphasize that our intent in this article was not to conducta policy analysis of the impacts of a particular plan or regu-lations. Because of the complex nature of growth manage-ment strategies, we do not attempt to isolate the impacts ofany one instrument nor to answer the question of whichspecific instruments of the region’s growth managementprogram are connected to the identified urban formchanges noted here. Nevertheless, we speculate that re-cently developed neighborhoods have relatively higherinternal street connectivity and better pedestrian access,and this might be attributed in part to street connectivitypolicies adopted in by Metro as part of the UrbanGrowth Management Functional Plan (Daisa et al., ).External connectivity is decreasing, and this might be dueto the Access Management Program (AMP) in state andcounty road departments. The AMP has limited localstreet connections to the arterial network over the past years. Another factor is that since the late s, mostsingle-family developments have been sited on parcelsabutting existing development. In many cases, these abut-ting developments lacked future street connections. TheUGB in the Portland metropolitan area might in partcause the density of single-family development to increaseover time. As Abbott () pointed out,

Data . . . suggest that the UGB is continuing to work.In , the Portland area was building new housing ata density of five dwelling units per acre. By and, the density of new development averaged eightdwellings per acre, actually exceeding the Plantarget. (p. )

These results, which characterize a trend in urban

form toward denser, more internally connected, and morepedestrian friendly yet relatively homogenous neighbor-hoods with poor external connectivity are rather encourag-ing for the neighborhoods, but less so for the region. Atthe regional level, we speculate, better land use mixing andregional connectivity is constrained by economies of scalein commercial uses and transportation infrastructure. Theresults are consistent with the concerns expressed by previ-ous studies. According to Bookout (a, b), Fulton(), Lehrer and Milgrom (), Furuseth () andSouthworth and Balaji (), the traditional neighbor-hoods that New Urbanists hope to replicate are character-ized by compactness, small scale, and diversity of buildings,but the economic and lifestyle demands of urban and sub-urban life seem to require certain facilities on a massivescale. In sum, Portland is winning the war on urban sprawlat the neighborhood scale, or at least appears to have wonsome important battles; but progress remains elusive at theregional scale.

AcknowledgementsWe express our appreciation for the support from the Lincoln Instituteof Land Policy and the U.S. Department of Housing and UrbanDevelopment. We alone are responsible for any remaining errors.

Notes. Because of the lack of important data on multifamily residential units,such as the number of units in a multifamily building and the year built,we were not able to include multifamily units in our urban form meas-ures. This exclusion has some effects on our measures, as we discuss nearthe end of the article.. For a historical review of land use policies in Oregon, see Abbott().. For more on the UGB, see Knaap () and Abbott ().. In addition to employing neotraditional architectural design princi-ples, New Urbanism also refers to larger scale planning issues (Furuseth,; Katz, ). Attempting to redefine urban landscapes by address-ing criticisms of suburban development and urban sprawl, regional-scaleNew Urbanism includes the following major elements: () high-densitydevelopment; () proximity of residences, jobs, and retail destinations topublic transit facilities to encourage pedestrian-oriented development;() mixed land uses, with retail and employment locations close to resi-dential areas; and () grid street networks, which are not divided intothe arterial collector/local road classification system found in most sub-urban areas (Ryan & McNally, ; Southworth, ).. For example, in a widely circulated email message, Andrés Duanystates:

I have visited Portland five times. All of them were invitationsto public speaking and so in all cases I was . . . shown the manywonderful places that make it a great, livable city. I agreed thatPortland was great and livable and praised the city accordingly. . . .


On the fifth trip, however, before my engagement, I “escaped” myhosts and went out to visit the famous urban boundary on myown. What did I find, to my surprise? That as soon as one left theprewar urbanism (to which all my prior visits had been confined)the sectors all the way to the urban boundary were chock full ofthe usual sprawl that one finds in any American city, no betterthan in Miami. So the outcome wasn’t that different after all; inPortland most of the prewar urbanism is excellent and most ofthe postwar version is junk. But can’t we say that, despite that,an urban boundary is an achievement in itself? That depends onwhere you draw it, and from that point of view, Portland’s wasnot a great achievement. It was drawn years out; which is to sayslack enough to include years of future growth. (Charles, ,p. )

. Criterion Planners/Engineers Inc., Portland, Oregon. More informa-tion is available on-line at http://www.crit.com/index/index.html.. All the calculations were computed using ArcInfo and ArcView, withdata from Metro’s Regional Land Information System (RLIS).. For more on connectivity, see Allen () and Southworth ().. Parks are defined as in Metro’s RLIS as public parks, greenways, andneighborhood parks.. For more on neighborhood definition, see Cervero and Gorham(), Crane and Crepeau (), and Olson ().. The median “year built” attribute for all single-family units in aneighborhood is, of course, an imperfect measure of neighborhood age.A neighborhood with half its housing built in the s and half built could have the same median year built attribute as one with all itshouses built in the s, though the former is certainly much older.Such limitations are important to remember in the interpretation of theresults, but we found less variation in the year built attribute withinneighborhoods than across neighborhoods; and we know of no betterway to estimate a neighborhood’s relative age with the available data.. We do not argue that census blocks are the best aerial unit formeasuring urban form. We only claim that it is a convenient unit thatillustrates the effects we seek to capture.. All urban form measures in this project are calculated by employing RLIS data rather than data over time. Therefore, a more accuratestatement would be that we are measuring associations between urbanform measures and the median age of single-family dwellings in neigh-borhoods. However, we use the term trend in urban form in this sectionto keep the interpretation concise. Based on our methods, to state that“trends in urban form measures changed” is a valid inference.. We use age as the only independent variable because in most caseswe have no theoretical basis for including other variables and because weare primarily interested in changes over time.. Other plots of data are available from the authors.. We thank an anonymous referee for this explanation of why theexternal connectivity has been decreasing.

References Friends of Oregon. (). The LUTRAC alternative. RetrievedMay , , from http://www.friends.org/resources/lut_vol.htmlAbbott, C. (). Planning a sustainable city: The promise andperformance of Portland’s urban growth boundary. In G. D. Squires(Ed.), Urban sprawl: Causes, consequences and policy responses (pp. –). Washington, DC: Urban Institute Press.

Allen, E. (). INDEX: Software for community indicators. In R. K.Brail & R. E. Klosterman (Eds.), Planning supporting systems: Integratinggeographic information systems, models, and visualization tools (pp. –). Redlands, CA: ESRI Press.American Planning Association. (). The principles of smart develop-ment (PAS report #). Chicago: Author.Benfield, F. K., Raimi, M. D., & Chen, D. D. (). Once there weregreenfields: How urban sprawl is undermining America’s environment,economy and social fabric. New York: Natural Resources DefenseCouncil.Bookout, L. W. (a, February). Neotraditional planning, cars,pedestrians and transit. Urban Land, , –.Bookout, L. W. (b, April). Neotraditional town planning: Buckingconventional codes and standards. Urban Land, , –.Calthorpe, P. (). New urbanism: Not perfect, but improving.Retrieved March , , from http://www.pbs.org/newshour/newurbanism/calthorpe.htmlCervero, R., & Gorham, R. (). Commuting in transit versusautomobile neighborhoods. Journal of the American Planning Association,, –.Charles, J. A. (). Lessons from the Portland experience. In J. S.Shaw & R. D. Utt (Eds.), A guide to smart growth: Shattering myths,providing solutions (pp. –). Washington, DC: The HeritageFoundation and the Political Economy Research Center.Chinitz, B. (). City and suburbs. Englewood Cliffs, NJ: PrenticeHall.City Council of Beaverton. (). Comprehensive plan for the Cityof Beaverton. Retrieved November , , from http://www.ci.beaverton.or.us/departments/CDD/ComprehensivePlan/maincomp.htmCity Council of Hillsboro. (a). Hillsboro comprehensive plan.Retrieved November , , from http://www.ci.hillsboro.or.us/Planning_Department/COMP_PLAN/COMPPLANTitle.pdfCity Council of Hillsboro. (b). Hillsboro subdivision ordinance.Retrieved November , , from http://www.ci.hillsboro.or.us/Planning_Department/SUB_ORD/Sub_Ord_Title.pdfCrane, R., & Crepeau, R. (). Does neighborhood design influencetravel? A behaviorial analysis of travel diary and GIS data. TransportationResearch D, (), –.Daisa, J. M., Kloster, T., & Ledbetter, R. (). Does increased streetconnectivity improve the operation of regional streets? Case studies from thePortland metro regional street design study. Retrieved November , ,from http://www.fehrandpeers.com/publications/papers/street_connectivity.pdfDuany, A., & Plater-Zyberk, E. (, May). The second coming of theAmerican small town. Plan Canada, –.Frank, L., & Engelke, P. (). The built environment and humanactivity patterns: Exploring the impacts of urban form on human health.Journal of Planning Literature, (), –.Fulton, W. (). The new urbanism: Hope or hype for Americancommunities? Cambridge, MA: Lincoln Institute of Land Policy.Fulton, W., Pendall, R., Nguyen, M., & Harrison, A. (). Whosprawls most? How growth patterns differ across the U.S. Washington, DC:Brookings Institute. Retrieved March , , from http://www.brook.edu/dybdocroot/es/urban/publications/fulton.pdfFuruseth, O. J. (). Neotraditional planning: A new strategy forbuilding neighborhoods? Land Use Policy, (), –.


Galster, G., Hanson, R., Ratcliffe, M. R., Wolman, H., Coleman, S.,& Freihage, J. (). Wrestling sprawl to the ground: Defining andmeasuring an elusive concept. Housing Policy Debate, (), –.Katz, P. (). The new urbanism: Toward an architecture of commu-nity. New York: McGraw-Hill.Knaap, G. (, June). The urban growth boundary in metropolitanPortland, Oregon: Research, rhetoric, and reality. Paper presented at theworkshop on urban growth management policies in the United States,Japan, and Korea, Seoul.Lehrer, U. A., & Milgrom, R. (, June). New (sub)urbanism:Countersprawl or repackaging the product. CNS, (), –.Metro. (). The Metro growth concept. Portland, OR: Author.Metro. (). regional framework plan and functional plan.Portland, OR: Author.Metro. (a). Metro at a glance. Portland, OR: Author.Metro. (b). RLIS lite [CD-ROM]. Portland, OR: Author.Mills, E. (). Urban economics (nd ed.). Glenwood, IL: ScottForesman.National Association of Homebuilders. (). Smart growth examples.Retrieved April , , from http://www.nahb.com/smartexamples/menu.htmNelson, A. C. (). Oregon’s urban growth boundary policy as alandmark planning tool. In C. Abbott, D. Howe, & S. Adler (Eds.),Planning the Oregon way: A twenty-year evaluation (pp. –). Corvallis,OR: OSU Press.Olson, D. P. (). What is neighborhood? Retrieved May , ,from http://www.kcneighbornet.org/O’Toole, R. (). Ten transit myths (Reason Public Policy Study#). Retrieved May , , from http://www.rppi.org/ps.htmlRyan, S., & McNally, M. G. (). Accessibility of neotraditionalneighborhoods: A review of design concepts, policies, and recentliterature. Transportation Research A, A(), –.Sierra Club. (). Sprawl: The dark side of the American dream.Retrieved May , , from http://www.sierraclub.org/sprawl/report/Southworth, M. (). Walkable suburbs? An evaluation of neotradi-tional communities at the urban edge. Journal of American PlanningAssociation, , –.Southworth, M., & Balaji, P. (). The suburban public realm II:Eurourbanism, new urbanism and the implications for urban design inthe American metropolis. Journal of Urban Design, (), –.Staley, S. R., Edgens, J. G., & Mildner, G. S. (). A line in the land:Urban-growth boundaries, smart growth, and housing affordability.Retrieved May , , from http://www.rppi.org/urban/ps.htmlWashington County, OR. (). Community development code.Retrieved May , , from http://www.co.washington.or.us/deptmts/lut/plan/comdev.htmWassmer, R. (). Urban sprawl in a U.S. metropolitan area: Ways tomeasure and a comparison of the Sacramento area to similar metropolitanareas in California and the U.S. (Lincoln Institute of Land PolicyWorking Paper). Retrieved May , , from http://www.lincolninst.edu/workpap/wpap.html


A CLASSIC IN THE MAKING...

“The Bulldozer in the Countryside is solid environmentalhistory, telling a remarkably broad story of politicaleconomy, culture, and physical environments on a

national scale...Rome writes gracefully, with a sense of drama that makes the book hard to put down.”

–Journal of American History

The Bulldozer in the Countryside is the first scholarlyhistory of efforts to reduce the environmental costs ofsuburban development in the United States. The bookoffers a new account of the mass migration to thesuburbs and the rise of the environmental movement.

The Bulldozer in the CountrysideSuburban Sprawl and the Rise of American Environmentalism

By Adam Rome$58.00: Hardback: 0-521-80059-5 $21.00: Paperback: 0-521-80490-6

www.cambridge.org

WINNER OF THEFREDERICK JACKSON

TURNER PRIZE OFTHE OAH

measuring urban form - alliance for community treesactrees.org/files/research/japasong.pdf ·...

Documents