http://usj.sagepub.com/Urban Studies

http://usj.sagepub.com/content/35/7/1019The online version of this article can be found at:

DOI: 10.1080/0042098984466

1998 35: 1019Urban StudGregory K. Ingram

Patterns of Metropolitan Development: What Have We Learned?

Published by:

http://www.sagepublications.com

On behalf of:

Urban Studies Journal Foundation

can be found at:Urban StudiesAdditional services and information for

http://usj.sagepub.com/cgi/alertsEmail Alerts:

http://usj.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://usj.sagepub.com/content/35/7/1019.refs.htmlCitations:

What is This?

- Jun 1, 1998Version of Record >>

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

Urban Studies, Vol. 35, No. 7, 1019±1035, 1998

Patterns of Metropolitan Development: What HaveWe Learned?

Gregory K. Ingram

[Paper received in ®nal form, December 1997]

Summary. Urban develop ment pattern s in both industrial and develop ing countries with

market-orien ted economies show strong regularities consisten t with basic urban location theory.

Large metropolitan areas are converging to similarly decentralised structures with multiple

sub-cen tres, decen tralised manufacturing and more centralised service employm ent. Decentralis-

ation is increasin g the relian ce on road-based urban transport for both passengers and freigh t.

Land markets are strong determ inants of decentralisat ion , and the development pattern s of cities

without land markets differ greatly from cities with even poorly function ing land markets.

Demand pattern s in urban housing are similar across cities, but supply-side im pediments vary

widely, resulting in a wide range of the ratio of housing prices to income. The ef®ciency of public

infrastru cture provision also varies widely across cities and across sectors within cities. Large

metropolitan areas in low-income countries will continue to grow as these countries urbanise.

1. Introduction

Although much of our knowledge aboutmetropolitan development is still imperfect,in the past 35 years a great deal of theoreticaland empirical work has been carried out oncities and metropolitan areas in both devel-oped and developing countries with market-oriented economies. This work has produceda set of empirical ®ndings with remarkablystrong regularities across countries and cities.Moreover, many of these empirical regulari-ties are quite consistent with urban locationtheory and tend to indicate the broad appli-cability of our basic theory to market-basedcities. This paper attempts to summarisemany of these empirical regularities aboutpatterns of metropolitan development. Theseregularities offer insights about the develop-

ment and growth pressures that exist in manycities and indicate what directions futuredevelopment is likely to take. It would betempting to argue that all of the empiricalregularities discovered are consistent withtheory, have normative content, or re¯ectunderlying outcomes that are ef®cient. Inmany cases this may be true, but care mustbe taken in drawing such conclusions becausesome of these stylised facts may be based ontechnological or demographic factors as muchas they are on theory or market outcomes.

2. Urbanisation and Economic Develop-

ment

Across countries, there is a strong relation-ship between the level of economic develop-

Gregory K. Ingram is with the Research Advisory Staff, The World Bank, Room N-7-043, 1818 H Street, NW, Washington, DC 20433,USA. Fax: 202-522-0304. E-mail:gingram@worldbank.org. The author thanks Ezra Bennathan, Douglas Keare, Kyu Sik Lee andSlobodan Mitric for useful comments on this paper. He also thanks for their discussion, comments and suggestions the participantsat the TRED Conference, particularly Alex Anas and Anita Summers.

0042-0980/98/071019-17 $7.00 Ó1998 The Editors of Urban Studies at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1020

Figure 1. Urbanisati on increases with incom e.Source: World Bank data.

ment and the degree of urbanisation. Figure 1shows the relation between GDP per capita(in 1985 purchasing power parity dollars)and the percentage of the popula tion living inurban areas for 101 countries in both 1960and 1985. In Figure 1, a line segment con-nects each country’s position in 1960 with itsposition in 1985. For every country, the per-centage urban in 1985 exceeded the percent-age urban in 1960. Figure 1 shows that thepercentage of the population living in urbanareas increases rapidly with GDP per capitaat low-income levels (below $2000) and thenmuch more slowly with income growth atmiddle- and high-income levels. The mostrapid growth in urban populations thus oc-curs as countries move from low-income tomiddle-income levels. Moreover, the in-crease in urbanisation with income, as illus-trated in Figure 1, applies across countries ata single point in time and over time to anindividual country experiencing economicgrowth.

Urbanisation levels rise with income asresources are moved from the primarily ruralagricultural sector to the more urbanised in-dustry and service sectors. The attractivenessof urban areas for the location of industryand services stems from scale economies inproduction (ef®cient plant sizes are large),lower transport costs (reduced by clusteringactivities together), the modest use of land byindustry and services as an input to pro-duction (allowing high densities), externali-ties among ®rms (sharing of information),linkages across ®rms (providing intermediateinputs to each other) and potential agglomer-ation economies (because large clusters ofactivities use specialised inputs moreef®ciently). (For a variety of views on thecauses of urbanisation, see Henderson, 1985;and Krugman, 1995.)

If this relationship between developmentand urbanisation continues to hold, globallevels of urbanisation will increase markedly(World Bank, 1990; United Nations, 1993).

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1021

Table 1. Average density intercep ts and gradient s, selected North and Latin American cities

1970 1960 1950

D b D b D b

Large cities ( . 2.5 million)North American (six) 16 000 0.11 18 000 0.12 24 000 0.17Latin American (four) 26 000 0.12 30 000 0.16 35 000 0.20

Small cities ( , 2.5 million)North American (six) 4 400 0.12 3 900 0.14 3 300 0.15Latin American (four) 20 500 0.26 16 000 0.31 10 000 0.32

Note: Densities are persons per sq km.Source: Calculated from Ingram and Carroll (1981).

In 1995, low-income countries (GNP per cap-ita less than $765 in 1995) had 56 per cent ofthe world’s population (World Bank, 1997,p. 36), and low-income countries as a grouphave been experiencing high rates of econ-omic growth. From 1980 to 1990, GNP percapita in low-income countries increased 4.1per cent per year, compared to a global aver-age of 1.5 per cent (World Bank, 1997,pp. 36 and 132). Projections indicate that in2010, of the 59 cities in the world withpopulations over 5 million, 47 will be indeveloping countries (Berghall, 1995, p. 12).

China and India contain nearly two-thirdsof the population in all low-income countries,and both already have large cities and aboveaverage rates of economic growth. If theurban populat ion of China or India doubles,will their large cities also double in size?Evidence indicates that existing large cities inlarge developing countries grow at roughlythe same, or a bit lower, rate as the overallurban populat ion. In addition to country size,the type of government also in¯uences thegrowth of large cities. In countries with uni-tary governments, large cities have tended togrow faster than the urbanisation level, whilelarge-city growth has been slower in coun-tries with federal governments (Mills andBecker, 1986, p. 59). This suggests that largeChinese and Indian cities are likely to growless than in propor tion to their countries’overall urban populat ion, but they will growin size.

3. The Distribution of Population within

Cities

Any analysis of the physical development ofa metropolitan area requires an understandingof how the residential population is dis-tributed within the area and how this distri-bution changes over time. A great deal isknown about this both for individual citiesover time and across cities at the same pointin time (Ingram and Carroll, 1981; Y. J. Lee,1985; Mills and Tan, 1980; Zhang, 1991). Inindustr ial countries, large cities (over 2.5million inhabitants) have higher densitiesthan small cities and both large and smallcities tend to be decentralised–with popu-lation densities that decline slowly as dis-tance from the city centre increases. Indeveloping countries, the popula tion densi-ties of large cities also decline slowly asdistance from the centre increases, whereasthe popula tion densities of small cities dropoff rapidly. In terms of overall density pat-terns, large cities in industr ial and developingcountries are quite similar with high densitiesand ¯at density gradients. These patterns areillustrated by Table 1 which shows for se-lected North American and Latin Americancities average intercepts (D) and gradients (b)from the standard density gradient equation,

D(x) 5 D exp ( 2 bx).

Over time, a universal ®nding is that metro-politan populations have become more de-

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1022

centralised (population density gradientsbecome ¯atter)–due to the effects of in-creases in income (promoting housing con-sumption) and improvements in transportperformance (higher speeds and lower costsrelative to incomes) (Meyer and Meyer,1987). Population growth in large cities usu-ally does not increase the popula tion densityof high-density areas, but promotes densi®-cation of less-developed areas and expansionat the urban fringe. In particular, populationdensities in the most central zones fre-quently decline as households are displacedby the expansion of other activities. This isa very robust ®nding in both industr ial anddeveloping countries and has been observedin cities as diverse as Bangkok, Bogota,Mexico City, Shanghai and Tokyo.1 In theUS, from one-third to one-half of large cen-tral cities have lost population over the past25 years (Downs, 1994, ch. 5; Meyer andGomez-Ibanez, 1981, ch. 2).

The preference for large-lot, single-familyhousing exhibited in the US does not seemto be ubiquitous. It is not readily observedin Europe, and the few analyses of this issuebased on household data from developingcountries have not found evidence of a pref-erence for large lots (Dowall and Treffeisen,1991; Ingram, 1984). Although gross resi-dential densities towards the periphery oflarge metropolitan areas in developing andindustr ial countries are similar, the net resi-dential densities in developing countriesare typically higher than those in the US(Mohan, 1994).

Development towards the periphery isdriven by lower land prices and lower devel-opment costs (Meyer and Gomez-Ibanez,1981). It is less costly to build on vacantland than to redevelop encumbered siteswhich requires the expenditure of resourcesto destroy existing physical assets and theloss of the assets as well. This is economi-cally feasible when transformation will pro-duce large increases in densities or the shiftof a parcel from residential to commercial orindustr ial use, but it is rare. Even in the USwhere redevelopment is thought to be en-demic, only about half of 1 per cent of

existing dwelling units are demolished eachyear (US Census of Housing, various years).Peripheral development is also permitted bythe wide availability of motorised modes ofpassenger transport in the cities of bothindustr ial and developing countries. Theshift from walking to motorbuses travellingon streets, the most common transit modein developing countries, typically triplestravel speeds from 5 km per hour to 15±20 km per hour (World Bank, 1986, p. 53).The shift from motorbus to motorcar (bothon common rights of way) typically onlydoubles speeds– the average door-to-doorspeed for work trips by car in the US in1980 was 38 km per hour (Downs, 1992,p. 11).

In addition to being more decentralised,the popula tion distribution in large cities ismore variegated than in small cities. Largecities in both industr ial and developingcountries usually have an original centre orcentral business district (CBD), but they alsohave a number of sub-centres which com-bine to form a polycentric development pat-tern (Dowall and Treffeisen, 1991). Smallcities, especially in developing countries, aremore likely to have a single, well-de®nedcentre (Ingram and Carroll, 1981). In ad-dition, households often sort themselvesamong locations in cities in particular ways.For example, larger households typicallyprefer larger dwelling units. Since housingprices and rents are lower at the periphery ofcities than at the centre, large households areoften more decentralised than small house-holds. In Bogota, Colombia, for example,the average household size at the centre in1978 was roughly two persons and house-hold size increased regularly with distancefrom the centre, reaching ®ve persons at theperiphery (Mohan, 1994). This pattern isconsistent with urban location theory. How-ever, the relationship between household in-come and distance from the centre does nothave a consistent pattern in cities in devel-oping countries, although high-incomehouseholds are clearly decentralising inmany large cities in developing countries(Ingram and Carroll, 1981).

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1023

4. The Distribution of Employm ent within

Cities

Because of data unavailability, fewer studieshave examined the spatial distribution of em-ployment than the spatial distribution of

population in cities. Nevertheless, a highlyregular set of ®ndings have emerged. Studieswithin cities over time indicate that there is a

marked tendency for employment to decen-tralise– the propor tion of jobs in the centrefalls over time and most new growth in

employment is located out of the centre oflarge cities (Meyer and Gomez-Ibanez,1981). Suburbs in the US contain more thanhalf of all urban jobs and are the site for

three-quarters of the new of®ce space(Downs, 1992; Diamond and Noonan, 1996).Analysis of US data indicates that industry is

attracted by freeways and special facilitiessuch as airports, but not by central locations(Shukla and Waddell, 1991). In developing

countries, urban industr ial employment ex-hibits strong patterns of decentralisation (K.S. Lee, 1989; Lee and Choe, 1989; Y. J. Lee,1985; Hamer, 1985a ). In Shanghai, for exam-

ple, decentralisation stems from the develop-ment of both specialised satellite industr ialtowns and rural industry. A third of Shang-

hai’s industr ial workers were rural in 1991(Ning and Yan, 1995). In Bogota, employ-ment decentralisation seems to be driven

mainly by market forces, whereas in SaoPaulo and Seoul, government policies alsoencouraged it (K. S. Lee, 1989). The ®rststudy of employment location in Africa doc-

uments a strong pattern of decentralisation ofmanufacturing employment in the Johannes-burg metropolitan area (Rogerson and Roger-

son, 1996).At the same time, employment is typically

more centralised in an urban area than is the

population. That is, if a cordon line is drawnat an arbitrary distance from the city centre,the speci®ed area will contain a larger pro-portion of all urban jobs than of all urban

population (Hamilton, 1982). This meansthat the typical commuter in an urban areacommutes from a residence more distant

from the centre to a workplace less distant

from the centre. However, employm ent is notheavily concentrated in the central businessdistricts of large cities (Meyer and Gomez-Ibanez, 1981). In the US it is rare to havemore than 8 per cent of a metropoli tan area’sjobs located in the central business district.New York City and Washington, DC, top thelist with 14 and 12 per cent respectively;Philadelphia is more typical with 7 per cent(US Department of Transportation, 1975). Indeveloping countries, the central businessdistricts in large cities are likely to havebetween 10 and 20 per cent of all metropoli-tan employment, but the percentages are fall-ing rapidly as most job growth is locatedoutside of the central business district (K. S.Lee, 1989; Lee and Choe, 1989; Y. J. Lee,1985).

There are also similar job location patternswithin cities by type of industry. Manufactur-ing employment is more decentralised thanservice employment (Y. J. Lee, 1985). Firmshave literally changed locations over time,and the annual mobility rates of manufactur-ing ®rms in developing and industr ial coun-tries are similar at around 3±5 per cent peryear (K. S. Lee, 1989). Printing is the onlycentralised manufacturing activity in both in-dustrial and developing countries. Moreover,there is a tendency for large manufacturingplants to be more decentralised than smallplants, and for areas close to the centre tospecialise in the location of new, small man-ufacturing enterprises in `incubator areas’(World Bank, 1990; K. S. Lee, 1989). Themovement of manufacturing ®rms is oftenstimulated by the need for more space, betterinfrastructure services and improved freighttransport by truck (Hamer, 1985a; K. S. Lee,1989). The relocation of manufacturing ac-tivity also reduces the demand for freighttransport in central areas which can reducecentral traf®c congestion.

As manufacturing jobs move out of thecentre, they are replaced by service sectorjobs. The evolution of service sector jobs inthe centre is less regular, but in many largecities in developing countries retail activitiesremain concentrated in the centre for sometime. Eventually, retail activity disperses and

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1024

is replaced by employm ent in ®nance, lawand other activities which are less oriented tohouseholds but require good communicationand face-to-face contact. In the US, retailemployment is now widely decentralised(Diamond and Noonan, 1996), whereas suchemployment is still centralised in many de-veloping countries where large retail estab-lishments are still located centrally (Y. J.Lee, 1985). The strong decentralisation ofretail establishments may not occur until autoownership reaches fairly high levels.

5. Location Patterns and Transport

There are many reasons why people taketrips in urban areas, but trips to work andschool are major components of travel. Indeveloping countries, the journey to worktypically accounts for 40±50 per cent of ur-ban trips, and trips to school account foranother 20±35 per cent (Mohan, 1994).Moreover, travel problems are most acuteduring peak hours, and work trips alone mayaccount for three-quarters of peak travel.Accordingly, the work-trip travel pattern incities in developing countries is a key deter-minant of transport demand and the overallneed for added transport capacity. Industr ialcountries have experienced much growth innon-work travel, so that their work trips maynow be less than a third of all trips and halfor less of peak-hour trips (Meyer andGomez-Ibanez, 1981; Small, 1992). How-ever, for both industr ial and developingcountries, the patterns of populat ion andemployment decentralisation, summarisedabove, obvious ly have profound implicationsfor transport because they are important de-terminants of the work-trip travel pattern.

Decentralisation of both jobs and resi-dences spreads work-trip travel ¯ows over abroader area. If all jobs were in the centre,there would be high traf®c ¯ows on radialroutes into the centre and high corridor vol-umes because the transport system wouldhave many origins for work trips but a con-centrated destination. As employment dis-perses from the centre, the transport system

has to serve many additional destinations.This reduces traf®c volum es between originsand the central destination, raises volum es toother destinations, and lowers radial corridorvolumes. These changes make serving trans-port demand with transit systems more costlybecause transit costs are higher and/or transitservice levels are lower at lower levels ofcorridor passenger ¯ows (Meyer et al.,1965). The increasing costs and decreasingservice levels of transit that accompanydecentralisation lead more travellers inmiddle- and high-income countries to useprivate autos, which further lowers transitpassenger volum e and further degrades tran-sit performance.

Decentralisation of jobs and residencestypically reduces transit performance but im-proves auto performance by reducing aver-age work-trip lengths and lowering traf®cvolumes on radial corridors. Distributingtraf®c ¯ows more widely across the transportnetwork and reducing radial corridor vol-umes may reduce congestion. In addition, thedecentralisation of employm ent and resi-dences has the potential for reducing thetravel distances of commuters (Downs,1992). Employm ent decentralisation, in par-ticular, can be envisaged as moving joblocations closer to residential locations andimproving the jobs±housing balance. The po-tential for this is illustrated by average traveltimes from residence to workplace, such asthose shown in Table 2, which almost alwaysindicate that work-trip travel times arelongest for trips from suburban residences tocentral business district (CBD) workplaces,and shortest from suburban residences tosuburban workplaces.

Table 2. Average minutes of one-way commutingby location of residence and workplac e, US, 1980

Workplace

Residence CBD Rest of city Suburb

City 24.9 20.0 26.4Suburb 35.1 27.2 18.8

Source: Downs, (1992, p. 20).

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1025

Regression analyses of average commut-ing times across 82 metropolitan areas in theUS demonstrate that greater centralisation ofemployment (measured as the propor tion ofmetropolitan employment in the largest city)increases average work trip commute times(Gordon et al., 1989). An analysis of Bogota,Colombia, indicated that the average distancefrom home to work remained constant whilethe city’s population grew by 40 per centbecause of the decentralisation of employ-ment (Pineda, 1982). A comparison of Lon-don and Paris, inquiring why residents of thetwo cities travel nearly identical amountseven though London has 20 per cent morepeople and is much more spread out thanParis, concluded that the greater dispersionof London’s population and employment isthe key reason (Mogridge, 1986).

The changed traf®c volum es that occur asdecentralisation proceeds, noted above, mayrequire that a wider range of transit modesand service levels be made available. Largecities in developing countries often utilise avariety of on-street transit vehicles rangingfrom group taxis, jitneys and vans; throughfull size buses; to articulated buses (WorldBank, 1986; Kain, 1991). The optimal ve-hicle size is a function of passenger routevolumes and desired headways (time elapsedbetween vehicles) (Walters, 1979). In periph-eral areas and on the routes with low passen-ger volumes, it is often economical to usesmaller vehicles. This is often observed indeveloping countries, whereas industr ialcountries typically have less variety in thesizes and types of transit vehicles. In the US,this is largely due to regulations put in placeto protect transit franchises that were grantedto private ®rms (Meyer et al., 1965). Morerecently, smaller vehicles are used in the USto serve suburban areas and especially thehandicapped through dial-a-ride programmes(Meyer and Gomez-Ibanez, 1981).

The typical direction of causation in his-torical studies of metropolitan developmenthas been to view transport as a determinantof land use. Large transit systems built in thelate 19th and early 20th centuries had amajor impact on their cities’ development

patterns, improving access to the centralbusiness district and promoting relativelyhigh-density development along well-de®nedtransit corridors.3 However, now that jobgrowth in the central business district is low(or negative), particular care must be takenwhen analysing transport investments thatare strongly oriented to serve the centralbusiness district. Transport projects, such assubways and freeways, that increase trans-port capacity to the central business districtare often very expensive because of theirseparate rights of way and high costs ofconstruction– running to US$100 millionper kilometer for an installed subway systemand close to that for an urban limited-accesshighw ay (Kain, 1991; World Bank, 1986,p. 52).

If transport is a determinant of land-usedevelopment, what impact will the construc-tion of a large transit system have on thedevelopment pattern of an existing metro-politan area? Will it increase residential den-sities or centralise employment? A review ofexperience with new subways in Montreal,San Francisco, Toronto and Washington,DC, found very modest effects on metropoli-tan development patterns, with some evi-dence of development around stations inToronto and Washington, and some evidenceof CBD development in Montreal and SanFrancisco (Meyer and Gomez-Ibanez, 1981).More recent analyses of experience with newsubway systems in the US show that costshave been well above, and ridership levelswell below, forecasts or projections madewhen the projects were reviewed and ap-proved (Pickrell, 1989). This has also beenthe experience of many rail rapid transit sys-tems in developing countries (Fouracre et al.,1990). The construction costs of metros indeveloping countries are so high that theycrowd out many other investments and caneven have consequences for macroeconomicmanagement. Most systems have operatingde®cits that severely strain local budgets , asin Pusan and Mexico City (World Bank,1996; Kain, 1991). Perhaps the strongest®nancial performance from a recently con-structed subway is in Hong Kong, where

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1026

fare-box revenues cover operating costs andcontribute to capital costs (Fouracre et al.,1990). Incorrect cost and travel forecasts arealso common in new toll-road projects, suchas those in Mexico, and revenue shortfallshave had severe ®nancial consequences forinvestors (World Bank, 1994).

As noted above, the pattern of decentralis-ation in older metropoli tan areas and the lowdensities of newer auto-oriented cities makeit more costly to serve urban travellers bytransit. Moreover, superimposing transit sys-tems on existing cities has had minimal ef-fects on land-use patterns. These two factshave led many analysts to argue that land-usecontrols should be used as an instrument toaffect urban travel demand. This view isoften embodied in reports that recommendmore and better land-use planning as a sol-ution to urban transport and environmentalproblems (Diamond and Noonan, 1996).Sometimes the proposal is explicit, as in ananalysis of transport policy for Canberra andAsian cities.

Unlike º most cities worldw ide, all ofCanberra’s land is in public ownership.This leasehold land tenure system makespossible a much more interventionist styleof government in which the approachshould be that of the benign and prudentlandlord managing a large and complexestate in stewardship for the bene®t ofpresent and future generations (Black,1992, p. 8).

The notion of using land-use controls as aninstrument to affect transport demand alsostems from cross-sectional studies that indi-cate that low-density cities are more auto-oriented than high-density cities (Newmanand Kenworthy, 1991), and from the debateabout wasteful commuting in the US (Hamil-ton, 1982; Small and Song, 1992).

The most noteworthy attempt to use landuse to affect transport has been the develop-ment of planned communities that contain arough balance of both residences and jobs.The hypothesis has been that workers wouldprefer to live and work in the same com-munity, minimising their commuting travel.

Results have not ful®lled expectations. Stud-ies of British New Towns in 1990 foundthem to be only 5.4 per cent more self-contained than comparable towns (Bae andRichardson, 1993, p. 6). Analysis of plannedcommunities in the US found their residents’commuting patterns to be no different fromthose in unplanned communities (Downs,1992, p. 104). These results should not besurprising. Residential location theory pre-dicts that commuters will commute down therent gradient, trading off transport costs togain location rent savings; it does not predictthat commuters minimise travel time. Thefact that residents of British New Townscommute out (mainly by train) from thesetowns to their jobs in nearby large cities, andthat workers in those towns drive away fromtheir jobs to homes in the surroundingcountryside is what residential locationtheory predicts.

The commuting patterns in planned com-munities indicate that balancing the spatialdistribution of jobs and workers within com-munities, or focusing only on residential den-sities or on employment densities, may havelittle effect on travel patterns that are drivenby trade-offs between travel costs and loca-tion rents. Directly controlling household lo-cation choices is not very popular where ithas been tried. Some local governments re-quire municipal workers to live within themunicipality, a requirement that often makesthe workers worse-off. In addition, the focuson residential densities may omit other as-pects of residential choice that strongly affecthousehold welfare. For example, the analysisby Newman and Kenworthy (1991) relatesgasoline use in metropolitan areas to theiraverage densities, and ®nds that low-densityareas consume more gasoline. However, thehigh-density metropolitan areas with lowgasoline use also typically have high housingprices and low consumption of residential¯oor space per person. High gasoline con-sumption is associated with high housingspace consumption (see Figure 2).

Using land-use controls as an instrumentto affect transport demand in developingcountries is not likely to succeed. Many of

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1027

Figure 2. Gasoline consum ption and resident ial ¯oor area per person in selected cities. Note: Low gasolineuse is associate d with resident ial crow ding: 89 per cent of the variatio n in per capita gasoline use isassociate d with the square of ¯oor area per person. Sources: gasoline use– Newman and Kenworthy

(1991); ¯oor area– World Bank (1996, p. 61).

these countries have elaborate systems ofzoning and land-use controls on the books,but very little enforcement capacity (Hayashiet al., 1992; Miyamoto and Udomsri, 1992).In many cities in developing countries, up tohalf of the dwellings that are constructedeach year are illegal (World Bank, 1993).Roughly 60 per cent of the houses now exist-ing in Bogota were origina lly illegal (Mohan,1994; Hamer, 1985b). Land-use changeshave also been analysed as a means of im-proving air quality in urban areas. The con-clusion: changing spatial structure is aninef®cient way to improve urban air quality ,and raising densities may increase exposureto high pollutant concentrations (Bae andRichardson, 1993).

6. Land Markets

Many of the empirical generalisations orstylised facts summarised above come fromcities that have land markets which functionmore or less well. The outcomes obtainedresult from the actions of decentralised de-cision-makers in a market setting as well as

the in¯uence of other factors such as techni-cal change and the durability of structures.Location theory shows that the decliningdensity gradients are systematically related tounderlying land-rent gradients. It predictsthat the land-rent gradients will be less steepthan the density gradients because capital issubstituted for land in the production ofhousing (Mills, 1972, pp. 79±84). The strongempirical regularities found with populationdensities suggest that there are similar regu-larities in land-rent patterns. However, land-rent data are rare. The relatively fewempirical studies dealing directly with landrents tend to obtain results consistent withtheory (for example, Ingram, 1982; Mills andSong, 1977; Mohan and Villamizar, 1982).

An indirect test of the relationship be-tween populat ion densities and land rents isillustrated by `the exception that proves therule’. At least one city, Moscow, developedover a period of 70 years with no underlyingland market, and its population density pat-tern differs dramatically from that in othercities (Bertaud and Renaud, 1997). Figure 3,comparing density patterns in Moscow and

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1028

Figure 3. Residentia l densities in Moscow and Paris.Source: Bertaud and Renaud (1997, p. 141).

Paris, shows that Moscow has a densitygradient that actually increases from the cen-tre. The reduction in population density inMoscow from 4±8 km from the centre re¯ectsa ring of industr ial land use in that area. Thelack of a land market to allocate land orpromote its redevelopment, and the presenceof administrative procedures that promotedthe hoarding of inputs, produced a startlingindustr ial land-use pattern. Not only is indus-try located close to the centre, industr ial landuse occupies 31.5 per cent of the total built-up land in Moscow. This compares with 5±6per cent in most large cities. This land wasused for plants and to `warehouse’ industr ialinputs and outputs, and it seems that indus-trial land was being warehoused as well. Therelative regularity of populat ion density pat-terns in other cities–which have land mar-kets of widely varying ef®ciency– suggeststhat even poorly functioning land marketshave a strong effect on metropolitan spatialstructure relative to no land market at all.

Like other prices in an economy, landprices perform two roles: they have an al-locative function and a distributive function.In their allocative role, land prices indicatethe value of land to producers and signal howland should be used. In their distributive role,

land rents and increases in land values pro-duce returns to land-owners. When landprices are high, they indicate that the landshould be developed intensively and/or beoccupied by an activity that highly values thesite. In residential use, for example, highland values signal that land be developed athigh residential densities. At some sites, thevalue of land in non-residential use mayexceed its value in residential use, and onlynon-residential uses can afford the site. Ob-jections to the income to land-owners pro-duced by rising land prices have led somecountries to intervene directly in the landmarket in ways that have proved to be coun-ter-productive, with India’s Urban LandCeiling Act a noteworthy example (WorldBank, 1993, p. 29). Rather than interveningin the land market directly, it is possible togain the advantages of the allocative role ofland prices and moderate the impact of thedistributive role of land prices by taxingincomes earned from land rent or land sale.

In developing countries, the observation isoften made that land rents are `too high’, buta diagnosis of pathology requires a theory ofhealth. How high should land rents be? Rela-tive land rents in urban areas vary by loca-tion. High rents are observed at locations that

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1029

Figure 4. Patterns of spending on housing.Source: World Bank (1993, p. 75).

are easily accessible and have high concen-trations of economic activity, and land rentsand densities are closely related, as notedearlier. Often the centre of an urban area isthe most accessible location and has thehighest land rents and densities. This com-mon sense (and theoretically sound) ap-proach explains relative land rents in a city,but what explains the total value of land rentin a city or its average level? Empirical workon this topic is not extensive, but there issuggestive evidence that land rents at thenational level absorb a roughly constantshare of output,3 and that the total value ofland rents in a city varies directly with thevalue of the economic product produced inthe city (Ingram, 1982). Economic growth ina metropoli tan area will therefore raise ag-gregate land rents and values.

7. Housing, Residential Location and

Labour Markets

Urban housing and residential location havebeen the subject of much theorising and em-pirical work. Both theory and empirics indi-cate that household residential locations aresystematically determined relative to thehousehold’s workplaces. The direction ofcausation is not clear. We do not know if

households choose their workplace and theirresidence simultaneously or in a particularorder (Waddell, 1993). However, there arestrong regularities. As noted in section 5, thedistance from home to work is usuallygreater for workers employed at the citycentre than for those employed elsewhere inthe metropolitan area, and the average dis-tance from home to work is shorter, thefurther from the centre is the workplace. Inaddition, workers tend to live and work in thesame radial corridor of the city (Meyer et al.,1965; Mohan, 1994). These patterns, whichexist in both industr ial and developing coun-tries, provide some guidance for planners(and developers) who are trying to locatenew residential developments or industr ialparks in expanding metropolitan areas.

Much empirical work has been done onhousing demand and the expenditures thathouseholds make on housing in urban hous-ing markets. A major and very robust resultin both industrial and developing countries isthat the propor tion of income spent on hous-ing by households within a particular city ishigher for low-income households than forhigh-income households (Mayo et al., 1986;Malpezzi and Mayo, 1987). This is illus-trated by Figure 4, which also shows thatthe city-wide average share of income spent

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1030

on housing rises with overall average cityincome levels.

In the jargon of economics, the demandfor housing is inelastic with respect to in-come at a ®xed point in time, but may beelastic over time as incomes grow. It is note-worthy that the average city-wide share ofincome spent on housing peaks for middle-income, developing countries and is lower inindustr ial countries. This pattern is similar tothat displayed by the share of GDP investedin housing (Burns and Grebler, 1976; Annezand Wheaton, 1984; World Bank, 1993,p. 102). In addition, housing expenditures in-crease somewhat with family size becauselarger families buy larger units than smallerfamilies (Malpezzi and Mayo, 1987). Theseempirical regularities allow us to predict howmuch different households would be willingto pay for housing or what the distributionalimpact would be of a housing or propertytax, but it is less useful for predicting theincome level of households living in dwell-ing units with particular rents. In particular,we often ®nd relatively high-income house-holds living in housing units in areas whererents are low. This means that low-rentneighbourhoods are often not necessarily ef-fective location proxies for low-incomehouseholds (Ingram, 1984).

Relative to the work done on housing de-mand, there are few analyses of housingsupply. The work that has been done indi-cates that housing construction is typically avery competitive industry that uses simpletechnolog ies and has few barriers to entry(Mohan, 1994), although distortions in theland market can concentrate market power ina few ®rms (World Bank, 1993). In manycases, households can help to construct theirown units and many do (Hamer, 1985b) .Industry studies indicate that the largest con-struction ®rms have a small propor tion of themarket (Ingram, 1982). This is not alwaystrue of the construction materials industrywhich can be non-competitive, particularlyfor materials such as cement and plumbing®xtures in developing countries (Berghall,1995, p. 64; World Bank, 1993, p. 139).Analyses of the cost of building a standard

housing unit across developing countries in-dicate that construction costs vary less thanincomes. For example, construction costsonly double across countries whose per cap-ita income levels differ by a factor of ®ve(World Bank, 1993, p. 80).

Impediments to ef®cient housing supplyinvolve much more than constructioninef®ciencies. In many developing countries,infrastructure provision is carried out by thepublic sector and is not responsive to demand(World Bank, 1994, pp. 30±31). Servicedland ready for development often commandsa scarcity premium well above the cost ofproviding infrastructure (Green et al., 1994;World Bank, 1993, p. 81). Unreasonablyhigh construction standards and restrictiveland-use and zoning regulations can raisehousing costs dramatically (Angel and Mayo,1996). Regulatory hurdles and procedure cantake vast amounts of time–sub-division andtitling of lots were estimated to take between®ve and seven years in Malaysia in the mid1980s (World Bank, 1993, p. 85). Acrossvarying regulatory regimes, average ratios ofhousing prices to household incomes varyfrom a low of 2.5 in Bangkok to a high of5±7 in Seoul or Kuala Lumpur (World Bank,1993, p. 85).

It has long been evident to urban analyststhat an urban labour market can be viewed asthe dual of an urban housing market. Hous-ing analysts often take the location of theworkplace as ®xed and study residential lo-cation and housing choice in terms of thetravel cost-location rent trade-off. A similarapproach can be taken to urban labour mar-kets: assume that residential location is ®xedand study employm ent location and jobchoice in terms of trade-offs between travelcost and wage differences across workplaces.After all, individuals change their jobsroughly twice as often as they change theirresidence (Simpson, 1992, ch. 2).

Very little empirical work has been doneusing this approach to urban labour markets.This is partly because the urban wage gradi-ent is much ¯atter than the land-price gradi-ent, and partly because of the heterogeneityof both workers and jobs (Moses, 1962).

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1031

Recent work on urban labour markets basedon search theory has yielded testable predic-tions– such as that more skilled workers willsearch more broadly across the urban labourmarket and therefore have longer commutes(Simpson, 1992). The predictions of thesesearch models have reasonable empiricalsupport, but additional work is needed togain insights about what can be done at themetropolitan level to improve urban labourmarket outcomes.

8. Infrastructure and Basic Services

Cities in industr ial and developing countriesvary greatly with respect to the ef®ciency oftheir infrastructure provision (World Bank,1994). Infrastructure investment, mainte-nance funds and the services themselveshave been provided by the public sector inmost countries until recently. This is nowchanging in both industr ial and developingcountries, and private-sector involvement ininfrastructure provision is increasing. LatinAmerican countries are privatising infrastruc-ture services such as telecommunications,electric power and transport services, whilein east Asia private investment is ®nancing agreater share of infrastructure investmentneeds through a variety of concessions andcontracts (Ingram and Kessides, 1995). Casestudies indicate that privatising ®rms andusing private contractors for infrastructureconstruction, operation or maintenance isless costly and more ef®cient than havingsuch work done by public employees (Galalet al., 1994; Gyam® et al., 1992; Heseltineand Silcock, 1990; Newbery and Pollitt,1996).

The ®nancial arrangements to pay for in-frastructure vary from user fees to generaltax revenues. Infrastructure agencies whichutilise user fees and which can alter the feesto cover the costs of service have fewer®nancial problems, are more likely to extendservice and are more ef®cient than agenciesthat have to rely on general tax revenues forsupport (World Bank, 1994). In some cases,revenue from user fees covers investmentcosts as well as operating costs.

Some infrastructure agencies have had suc-cess using betterment fees to pay for infra-structure investments. Experience indicatesthat households are willing to pay such feesonly for infrastructure investments that di-rectly bene®t them– such as connecting theirhouse to a water main, paving the sidewalk infront of their house, or installing local streetlights. Households are not willing to paybetterment fees for infrastructure investmentsthat are near them, but which have manybene®ciaries– such as the improvement ofnearby arterial streets (Mohan, 1994; WorldBank, 1988).

Particular attention must be paid to theinfrastructure needs of industry in order toincrease economic productivity. Infrastructureservices such as electricity and transport areimportant intermediate inputs to enterprises,and the quality and reliability of infrastructureservices can be an important determinant of a®rm’s location (K. S. Lee, 1989; Lee andAnas, 1992). The ef®ciency of infrastructureservice provision varies across countries butis not related to per capita GDP, and withina country the ef®ciency of service productionin one infrastructure sector tells us virtuallynothing about the ef®ciency of service pro-duction in another (World Bank, 1994). Thesetwo ®ndings suggest that the organisation andincentives within infrastructure sectors areprimary determinants of the quality of infra-structure services that are produced.

Recent macro-level studies indicate thatinfrastructure services contribute signi®-cantly to economic growth in the US(Aschauer, 1989; Munnell, 1992), and theresults from other countries are mixed (Can-ning and Fay, 1993; Ford and Poret, 1991).The returns estimated in some of these studiesare often surprisingly (and unbelievably)large, but they are very sensitive to speci®-cation and the level of aggregation of the data(Holtz-Eakin, 1992; Gramlich, 1994). There isstill substantial disagreement among analystsabout why this is the case. The results fromthese macro-level studies are not solid enoughto provide guidance for designing infrastruc-ture investment policies or programmes. Thebest estimate of the return of an infrastructure

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1032

investment in a metropolitan area continuesto be a comprehensive bene®t±cost analysisat the investment project level.

9. Potential Sources of Regularities

The ®ndings summarised here are basedmainly on studies that have been carried outin market-based or mixed economies. Thereare many common behavioural patterns ex-hibited by households and ®rms in urbanareas across industr ial and developing coun-tries. In many cases, even the behaviouralparameter estimates–such as the elasticitiesin housing demand or mode choice equa-tions– are very similar across countries withvery different income levels. These similari-ties are often so striking that they demand anexplanation. One obvious explanation is thatacross countries households are fundamen-tally similar; in economic terms, they havesimilar utility functions.

A second possible explanation for the em-pirical regularities across urban areas is em-bodied in the analysis underlying purchasingpower parity estimates of GDP across coun-tries which focuses on the determinants ofthe prices of tradable versus non-tradablegoods (Kravis et al., 1978). Most urban ser-vices are non-tradables. The prices of trad-ables are set internationally, whereas theprices of non-tradables are mainly a functionof domestic income levels. This implies thatthe relative prices of many urban services,and the ratio of the prices of many urbanservices to local incomes, will not varygreatly across countries, particularly relativeto tradable goods. Given the underlyingsimilarities of households (and of householdutility functions) across countries, the simi-larities of urban service±relative price ratiosand price±income ratios are likely to producesimilar behaviour in economic explanationsinvolving urban goods and services.

10. Conclusion

The development patterns of cities in devel-oping and industr ial countries with market-

based economies exhibit similar patterns ofdecentralisation of both population and em-ployment, with the largest metropolitan areasconverging to similarly decentralised struc-tures with multiple sub-centres, highly de-centralised manufacturing employm ent andemerging specialisation of the central busi-ness district in service employm ent. Cities indeveloping countries typically have some-what higher population densities than thosein industr ial countries, but the differenceshave been narrowing over time in the largestmetropolitan areas.

Decentralisation of population and em-ployment increases reliance on road-basedtransport for both passengers and freight.Industr ial countries have experienced de-creases in transit use as auto ownership lev-els have risen. Many developing countriesshow early signs of a similar pattern, al-though their transit-ridership levels are stillhigh and their transit systems often offer arich mix of options in terms of vehicle sizesand levels of service.

Land markets are strong determinants ofdecentralisation, and cities without landmarkets exhibit quite different developmentpatterns from cities with even poorly func-tioning land markets. In market-based cities,land rents are closely related to developmentdensities, although empirical work on landrents and values is relatively rare because ofa lack of data. Analyses of urban housingmarkets indicate that demand patterns arevery similar across cities in developing andindustr ial countries, but that supply-side im-pediments vary widely– resulting in a widerange of ratios of housing prices to incomes.Similarly, the ef®ciency of public-sector in-frastructure provision varies widely acrosscities, and across sectors within cities.

The coming decades will see an increasein global urbanisation with most of the in-crease taking place in low-income countrieswhich contained in 1995 nearly 60 per centof the world’s population. Many of thesecountries already have large metropolitanareas whose populat ions will continue togrow as urbanisation increases. An improvedunderstanding of metropolitan development

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1033

continues to be critical to urban policy-making, particularly in low-income coun-tries.

Notes

1. Results for Bangkok and Mexico are fromthe author’ s own analysis. Other sources areBogota (Mohan, 1994); Shanghai (Ning andYan, 1995); and Tokyo (Zhang, 1991).

2. A classic study of the impact of transpor t onurban developm ent analyses Boston(Warner, 1970).

3. In the US, land rents absorbed 7.7 per cent ofnational incom e in 1850 and 6.4 per cent in1956 (Mills, 1972, p. 49).

References

ANGEL, S. and MAYO, S. (1996) Enabling policiesand their effects on housing sector perform -ance: a global comparison . Paper presented atHabitat II Conferen ce, Istanbul , Turkey.

ANNEZ, P. and WHEATON, W. (1984) Econom icdevelopm ent and the housing sector: a cross-national model, Econom ic Developm ent andCultura l Change , 32, pp. 749±766.

ASCHAUE R, D. A. (1989) Is public expendit ureproducti ve?, Journal of Monetary Econom ics,23, pp. 177±200.

BAE, C.-H. C. and RICHARDSON, H. W. (1993)Automobiles, the environm ent and metropolit anspatial structure . Working Paper, Lincoln Insti-tute of Land Policy, Cambridge, MA.

BERGHALL, P. E. (1995) Habitat II and the urbaneconom y: a review of recent developm ents andliterature , UN University /World Institute forDevelopm ent Econom ics Research, Helsinki.

BERTAUD, A. and RENAUD, B. (1997) Socialistcities without land markets, Journal of UrbanEconom ics, 41, pp. 137±151.

BLACK, J. A. (1992) Policy measures for land useand transpor t demand managem ent and theirimplicatio ns in managing grow ing Asianmetropol ises, Regional Developm ent Dialogue ,13(3), pp. 3±26.

BURNS, L. and GREBLER, L. (1976) Resource allo-cation to housing investm ent: a comparativeinternati onal study, Econom ic Developm entand Cultural Change , 25, pp. 95±121.

CANNING , D. and FAY, M. (1993) The effect oftransport ation networks on econom ic grow th .Working Paper, Colum bia University , NewYork.

DIAMOND, H. L. and NOONAN, P. F. (1996) LandUse in America , Lincoln Institute of Land Pol-icy. Washington, DC: Island Press.

DOW ALL, D. E. and TREFFEISEN, P. A. (1991) Spa-tial transformation in cities of the develop ingworld, Regional Science and Urban Econom -ics, 21, pp. 201±224.

DOW NS, A. (1992) Stuck in Traf®c. Washington,DC: Brookings Instituti on and Lincoln Instituteof Land Policy.

DOW NS, A. (1994) New Visions for Metropolit anAmerica . Washingto n, DC: Brooking s Insti-tution and Lincoln Institute of Land Policy.

FORD, R. and PORET, P. (1991) Infrastructure andprivate sector product ivity, OECD Econom icStudies , 17, pp. 63±89.

FOURACRE, P., ALLPORT, R. and THOMSON, J. M.(1990) The perform ance and impact of railmass transit in developi ng countrie s, ResearchReport 278, Transpor t Research Laborato ry,Crow thorne, UK.

GALAL, A., JONES, L., TANDON, P. and VOGEL-

SANG, I. (1994) Welfare Conseque nces of Sell-ing Public Enterprise s. New York: OxfordUniversity Press.

GORDON, P., KUMAR, A. and RICHARDSON, H. W.(1989) The in¯uence of metropoli tan spatialstructur e on commuting time, Journal of UrbanEconom ics, 26, pp. 138±151.

GRAMLICH, E. M. (1994) Infrastructure invest-ment: a review essay, Journal of Econom icLiterature , 32, pp. 1176±1196.

GREEN, R. K., MALPEZZI, S. and VANDELL, K.(1994) Urban regulatio ns and the price of landand housing in Korea, Journal of Housing Eco-nomics, 4, pp. 330±356.

GYAMFI, P., GUTIERREZ, L. and YEPES, G. (1992)Infrastru cture maintenan ce in LAC: the costs ofneglect and options for improvem ent, 3 vols.Report 17, Regional Studies Program , WorldBank, Washingto n, DC.

HAMER, A. (1985a) Decentrali zed urban develop-ment and industria l location behavio r in SaÄ oPaulo . Staff Working Paper 732, World Bank,Washington, DC.

HAMER, A. (1985b ) Bogota’s unregula ted sub-division s: the myth and reality of increm entalhousing construct ion . Staff Working Paper 734,World Bank, Washington, DC.

HAMILTON, B. (1982) Wasteful commuting, Jour-nal of Political Economy, 90, pp. 1035±1053.

HAYASHI, Y., WEGENER, M., DOI, K. and SUPARAT,R. (1992) An internati onal comparative studyon land use transport plannin g policies as con-trol measures of urban environm ent. Paper pre-sented to the Sixth World Conferen ce onTransport Research , Lyon, June, (Mimeo-graph).

HENDERSON, J. V. (1985) Urban Developm ent:Theory, Fact and Illusion . New York: OxfordUniversity Press.

HESELTINE , P. M. and SILCOCK, D. T. (1990) Theeffects of bus deregula tion on costs, Journal of

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

GREGORY K. INGRAM1034

Transport Econom ics and Policy , 24, pp. 239±254.

HOLTZ-EAKIN, D. (1992) Public-sector capital andthe producti vity puzzle . Working Paper 4122,National Bureau of Econom ic Research, Cam-bridge, MA.

INGRAM, G. K. (1982) Land in perspect ive: its rolein the structur e of cities, in: M. CULLEN and S.WOOLERY (Eds) World Congress on Land Pol-icy, pp. 103±118. Lexington , MA: LexingtonBooks.

INGRAM, G. K. (1984) Housing demand in thedevelopi ng metropolis . Staff Working PaperNo. 663, World Bank, Washingto n, DC.

INGRAM, G. K. and CARROLL, A. (1981) The spa-tial structure of Latin American cities, Journalof Urban Econom ics, 9, pp. 257±273.

INGRAM, G. K. and KESSIDE S, C. (1995) The®nancing of infrastr ucture in develop ing coun-tries, in: Developm ent Issues , Presentati ons tothe 50th Meeting of the Developm ent Com-mittee, World Bank, Washingto n, DC.

KAIN, J. F. (1991) A critical assessm ent of publictransport investm ents in Latin America . Inter-American Developm ent Bank, Washingto n,DC.

KRAVIS, I. B., HESTON, A. and SUMMERS, R. (1978)Internat ional Comparisons of Real Productand Purchasin g Power. Baltim ore, MD: JohnsHopkins University Press.

KRUGMAN, P. (1995) Urban concentr ation: therole of increasin g returns and transpor t costs,in: Proceeding s of the 1994 World Bank An-nual Conferen ce on Developm ent Econom ics,pp. 241±263. Washington, DC: World Bank.

LEE, K. S. (1989) The Location of Jobs in aDevelopin g Metropolis . New York: OxfordUniversity Press.

LEE, K. S. and ANAS, A. (1992) Costs of de®cientinfrastru cture: the case of Nigerian manufactu r-ing, Urban Studies , 29, pp. 1071±1092.

LEE, K. S. and CHOE, S. C. (1989) Changing loca-tion patterns of industri es and urban decentra l-ization policies in Korea, in: J. KWON (Ed.)Korean Econom ic Developm ent, pp. 429±436.New York: Greenw ood Press.

LEE, Y. J. (1985) The spatial structure of themetropolit an regions of Brazil . Staff WorkingPaper 722, World Bank, Washingto n, DC.

MALPEZZI, S. and MAYO, S. (1987) The demandfor housing in develop ing countrie s, Econom icDevelopm ent and Cultural Change , 35,pp. 687±721.

MAYO, S., MALPEZZI, S. and GROSS, D. (1986)Shelter strategie s for the urban poor in develop -ing countrie s, World Bank Research Observer ,1, pp. 183±203.

MEYER, J. R. and GOMEZ-IBANEZ, J. A. (1981)Autos, Transit, and Cities. Cambridge, MA:Harvard University Press.

MEYER, J. R. and MEYER, L. K. (1987) Econom icdevelopm ent cities, and the urban transport a-tion problem . Discussion Paper 258, HarvardInstitute for Internat ional Developm ent, Cam-bridge, MA.

MEYER, J. R., KAIN, J. F. and WOHL, M. (1965)The Urban Transporta tion Problem . Cam-bridge, MA: Harvard University Press.

MILLS, E. S. (1972) Urban Econom ics. Glenview ,II: Scott-Foreman.

MILLS, E. S. and BECKER, C. M. (1986) Studies inIndian Urban Developm ent . New York: OxfordUniversity Press.

MILLS, E. S. and SONG, B. N. (1977) Korea’surbaniza tion and urban problem s 1945±1975 .Working Paper 7701, Korea Developm ent In-stitute.

MILLS, E. S. and TAN, J. P. (1980) A comparisonof urban populat ion density function s in devel-oped and developi ng countrie s, Urban Studies ,17, pp. 313±321.

MIYAMOTO, K. and UDOMSRI, R. (1992) Presentsituation s and issues of transport plannin g andplan implem entation in developi ng metropo-lises from the viewpoint of integrati on withland-use policies . Paper presented to the SixthWorld Conferen ce on Transport Research ,Lyon, (mimeograph ).

MOGRIDGE, M. J. H. (1986) If London is morespread out than Paris, why don’t Londoner stravel more than Parisians?, Transporta tion , 13,pp. 85±104.

MOHAN, R. (1994) Understan ding the DevelopingMetropolis : Lessons from the City Study ofBogota and Cali, Colombia . New York: OxfordUniversity Press.

MOHAN, R. and VILLAMIZAR, R. (1982) The evol-ution of land values in the context of rapidurban growth: a case study of Bogota and Cali,Colombia, in: M. CULLEN and S. WOOLERY

(Eds) World Congress on Land Policy ,pp. 217±252. Lexington , MA: LexingtonBooks.

MOSES, L. (1962) Towards a theory of intra-urban wage differen tials and their in¯uenceon travel patterns, Papers and Proceedin gsof the Regional Science Associatio n, 9, pp. 53±63.

MUNNELL, A. H. (1992) Infrastr ucture investm entand econom ic grow th, Journal of Econom icPerspective s, 6(4), pp. 189±198.

NEWBERY, D. M. and POLLITT, M. G. (1996) Therestructu ring and privatisa tion of the CEGB–was it worth it? DAE Working Paper 9607,Departm ent of Applied Econom ics, CambridgeUniversity .

NEWMAN, P. W. G. and KENW ORTHY, J. R. (1991)Cities and Automobile Dependen ce: An Inter-nationa l Sourcebook. Aldershot : AveburyTechnical Publishing.

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from

PATTERNS OF METROPOLITAN DEVELOPMENT 1035

NING, Y. and YAN, Z. (1995) The changin g indus-trial and spatial structur e in Shanghai, UrbanGeograph y, 16, pp. 577±594.

PICKRELL, D. H. (1989) Urban rail transitprojects: forecast versus actual ridership andcosts , Transport System s Center, US Depart-ment of Transport ation, Cambridge, MA.

PINED A, J. F. (1982) Residentia l location decision sof multiple worker househo lds in Bogota,Colombia, Revista Camara de Comercio deBogota , 46, pp. 163±187.

ROGERSON, C. M. and ROGERSON, J. M. (1996)Manufactu ring location in the developi ngmetropolis : the case of Greater Johannesburg .Washingto n, DC: World Bank, (mimeograph ).

SHUKLA, V. and WADDE LL, P. (1991) Firm loca-tion and land use in discrete urban space,Regional Science and Urban Econom ics, 21,pp. 225±253.

SIMPSON, W. (1992) Urban Structure and theLabour Market . Oxford: Clarendon Press.

SMALL, K. (1992) Urban Transporta tion Econom-ics. Philadelphia, PA: Harwood Academ ic Pub-lishers.

SMALL, K. and SONG, S. (1992) Wasteful commut-ing: a resoluti on, Journal of Political Econom y,100, pp. 888±898.

UNITE D NATIO NS (1992) Building Materials forHousing , Report of the Executive Director HS/C/14/7, Commission on Human Settlem ents,Nairobi, Kenya.

UNITE D NATION S (1993) World Urbanizat ionProspects, 1992 Revision . New York: UN.

US CENSUS OF HOUSING (variou s years) Compo-nents of Invento ry Change: U.S. and Regions .Washingto n, DC: US Bureau of the Census.

US DEPARTMENT OF TRANSPORTATIO N (1975) Ur-ban Data Book, Report No. DOT-TSC-OST-

75±45. Washington, DC: US Departm ent ofTransport ation.

WADD ELL, P. (1993) Exogenou s workplace choicein resident ial location models: is the assum p-tion valid?, Geograph ical Analysis , 25, pp. 65±84.

WALTERS, A. A. (1979) Cost and scale of busservices . Staff Working Paper No. 325, WorldBank, Washingto n, DC.

WARNER, S. B., JR (1970) Streetcar Suburbs: TheProcess of Growth in Boston 1870±1900 . NewYork: Atheneum .

WHIT E, M. (1994) Housing and the journey towork in U.S. cities, in: Y. NOGU CHI and J.POTERBA (Eds) Housing Markets in the UnitedStates and Japan , pp. 133±159. Chicago: Uni-versity of Chicago Press.

WORLD BANK (1986) Urban Transport: A WorldBank Policy Study . Washington, DC: WorldBank.

WORLD BANK (1988) World Developm ent Report1988 . New York: Oxford University Press.

WORLD BANK (1990) Urban Policy and Econom icDevelopm ent: An Agenda for the 1990s . Wash-ington, DC: World Bank.

WORLD BANK (1993) Housing: Enabling Marketsto Work . Washington, DC: World Bank.

WORLD BANK (1994) World Developm ent Report1994: Infrastru cture for Developm ent. NewYork: Oxford University Press.

WORLD BANK (1996) Sustainable Transport: Pri-orities for Policy Reform . Washingto n, DC:World Bank.

WORLD BANK (1997) World Developm ent Indica-tors. New York: Oxford University Press.

ZHANG, X. P. (1991) Metropoli tan spatial structureand its determ inants: a case study of Tokyo,Urban Studies , 28, pp. 87±104.

at Massachusetts Institute of Technology on December 31, 2013usj.sagepub.comDownloaded from