Urban Resurgence and the Consumer City - Glaeser

This article was downloaded by:[University of Reading]On: 7 October 2007Access Details: [subscription number 769893539]Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Urban StudiesPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t713449163

Urban resurgence and the consumer cityEdward L. Glaeser a; Joshua D. Gottlieb ba Department of Economics, Harvard University and NBER, Littauer Center,Cambridge, MA, USAb Taubman Center for State and Local Government, Harvard University, Cambridge,MA, USA

Online Publication Date: 01 July 2006To cite this Article: Glaeser, Edward L. and Gottlieb, Joshua D. (2006) 'Urbanresurgence and the consumer city', Urban Studies, 43:8, 1275 - 1299To link to this article: DOI: 10.1080/00420980600775683URL: http://dx.doi.org/10.1080/00420980600775683

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expresslyforbidden.

The publisher does not give any warranty express or implied or make any representation that the contents will becomplete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should beindependently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with orarising out of the use of this material.

http://www.informaworld.com/smpp/title~content=t713449163

http://dx.doi.org/10.1080/00420980600775683

http://www.informaworld.com/terms-and-conditions-of-access.pdf

Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

Urban Resurgence and the Consumer City

Edward L. Glaeser and Joshua D. Gottlieb

[Paper first received, April 2005; in final form, February 2006]

Summary. Cities make it easier for humans to interact, and one of the main advantages of dense,urban areas is that they facilitate social interactions. This paper provides evidence for the USsuggesting that the resurgence of big cities in the 1990s is due, in part, to the increased demandfor these interactions and due to the reduction in big city crime, which had made it difficult forurban residents to enjoy these social amenities. However, while density is correlated withconsumer amenities, we show that it is not correlated with social capital and that there is noevidence that sprawl has hurt civic engagement.

1. Introduction

Over the past 50 years, the automobile hasdramatically changed the world’s urban land-scape. Traditional dense walking cities havebeen replaced by cities designed entirelyaround the automobile. These new cityscapeslook completely unlike earlier ones becausethey are so much less dense—a result of theform of transport they exploit. The cardemands a physically large city becausedriving and parking require vast amounts ofspace and the car’s nature permits low-density living. As the car supplanted previoustechnologies, traditional cities fell into a spiralof decline and seemed doomed to increasingobsolescence.1

With this backdrop, the remarkable resur-gence of a number of big cities in the 1990sis quite remarkable. In the US, New York,Chicago and Boston have all experiencedremarkable decades relative to their recentpast. London has also had a turnaround sincethe 1970s. There are two main explanations

for this turnaround. First, over the past 20years, there has been a remarkable increasein the importance of knowledge in theeconomy and the biggest, densest citiesappear to have a comparative advantage infacilitating the flow of knowledge. Secondly,over the past 20 years, the desire of consumersto live in these cities has increased enor-mously as a result of changes in style ofgovernment, improvements in law enforce-ment technology and rising incomes thathave raised demand for high-end urbanamenities.

In part 2 of this paper, we review the basicfacts of urban resurgence in the US. Since1980, urban housing prices have risen, some-times dramatically. Urban incomes havedone moderately well and population levelsstabilised after the dismal 1970s. There isclear evidence for a widespread resurgence,but population levels are still rising faster inthe car-oriented Sunbelt than in traditionalurban areas. Moreover, the evidence suggeststhat, when we adjust for costs of living, real

Urban Studies, Vol. 43, No. 8, 1275–1299, July 2006

Edward L. Glaeser is in the Department of Economics, Harvard University and NBER, Littauer Center, 1875 Cambridge Street, Cam-bridge, MA 02138, USA. Fax: 617 495 7730. E-mail: [email protected]. Joshua D. Gottlieb is in the Taubman Center for Stateand Local Government, Harvard University, 79 John F. Kennedy St, Cambridge, MA 02138, USA. Fax: 617 495 7730. E-mail:[email protected]. The Taubman Center for State and Local Government provided helpful funding. The authors would liketo thank Jesse Shapiro, Andrei Shleifer, the Editors and two anonymous referees for valuable comments. The authors also thankAlbert Saiz, Jesse Shapiro and Jacob Vigdor for providing data and Josh Samuelson for excellent research assistance.

0042-0980 Print=1360-063X Online=06=081275–25 # 2006 The Editors of Urban Studies

DOI: 10.1080=00420980600775683

Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

wages have been falling in dense urban areas.Standard economic theory tells us that thismeans that urban resurgence is not primarilythe result of rising urban productivity.Instead, falling relative real wages are betterseen as evidence for an increased desire ofpeople to live in urban areas. Big cities arehaving a renaissance as places of consump-tion, not production.

In part 3, we turn to the social aspects ofdense urban centres. More specifically, welook at whether the urban edge in facilitatingsocial contact has played a role in the renais-sance of many of the largest cities. Urbanproximity facilitates both positive and nega-tive social interactions. Positive social inter-actions include the fun of meeting people orgoing out to concerts, museums, and otheractivities more readily available in urbanareas, and the city’s important role as a mar-riage market. Negative social interactionsinclude crime, the spread of disease throughclose physical proximity, and congestion.Urban success depends, in part, on the positivesocial interactions outweighing the negativeinteractions.

Since 1980, crime rates have fallen dra-matically. This decline can potentiallyexplain as much as one-third of risinghousing prices in some cities (Schwartzet al., 2003). The effect of declining crimerates on population levels is likely to bemuch smaller. We provide evidence on thegreater tendency of people in cities to go torestaurants, concerts, museums and bars. Allof these activities rise with income and mostrise with education. Hence, one plausiblehypothesis is that urban resurgence can beunderstood as reflecting the rising demandfor urban amenities caused by rising incomeand education levels nation-wide.2

In the final section of this paper, we turn tothe question of whether car-based living, orsprawl, as it is sometimes called, is destroyingthe social fabric of the US. Putnam (2000,p. 215) argues that “the residents of largemetropolitan areas incur a ‘sprawl civicpenalty’ of roughly 20 per cent on mostmeasures of community involvement”. Infact, the data do not show any negative link

between social capital and conventional con-cepts of sprawl. To be sure, there is nodoubt that people who live in towns withless than 50 000 residents or rural areas aremore likely to join social groups than thosewho live in bigger cities or their suburbs.But comparing small towns with metropolitanareas sheds no light on anything resemblingthe conventional concept of sprawl—low-density or car-based living within metro-politan areas. Big metropolitan areas are notsprawl and small towns are not the oppositeof sprawl. In fact, it is small towns that havethe lower population densities and greaterautomobile ownership usually associatedwith sprawl.3 Consequently, Putnam’sfinding that social capital is higher in smalltowns does not mean that there is a ‘sprawlcivic penalty’.

If anything, civic participation appears todecline in dense communities and rise in thesuburbs. Putnam suggests that sprawl is anenemy because of longer commute times,but low-density, car-based living is associatedwith shorter—not longer—commutes.Putnam’s language appears to give supportto the view of some opponents of sprawl thatsuburbs are bad for social connection, butPutnam describes the evidence correctly

Even taking into account the educationaland social backgrounds of those who havemoved there, the suburbs have faintlyhigher levels of trust and civic engagementthan their respective central cities, whichshould have produced growth, not decay,in social capital over the last generation(Putnam, 1996).

As this quotation suggests, there is no evi-dence suggesting that discouraging medium-density, car-based living will improve socialcapital and there is some evidence to suggestthe contrary.

2. Understanding the Resurgent City

We begin with an overview of facts about theresurgent city. We first recapitulate two essen-tial elements of spatial economics. First,migration across cities ensures that cities

1276 EDWARD L. GLAESER AND JOSHUA D. GOTTLIEB

Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

that are more attractive along one dimension(such as having higher wages) are less attrac-tive along some other dimension (such ashaving worse weather). This principle liesbeyond the hedonic literature (for example,Rosen, 1979) that estimates the value ofurban amenities. Secondly, mobility of firmsensures that cities that provide one kind of pro-ducer advantage (like high productivity) willhave other production disamenities (likeworkers who demand high wages). Thisframework tells us simultaneously that highnominal wages imply high levels of pro-ductivity and that high real wages (wages con-trolling for cost of living) imply consumerdisamenities. If the high nominal wages werenot accompanied by high productivity, thenthe firms would leave. If the high real wageswere not accompanied by offsetting disame-nities, then workers would flock to the city.

A large body of empirical work, such asRoback (1982) and Gyourko and Tracy(1989), has shown the value of this frame-work. Glaeser et al. (2001) use this insightto create an index of attractiveness andindeed find that the places with low wagesrelative to housing prices are generally all inCalifornia. The places with high real wagesare generally in the colder or less pleasantparts of the country (like Alaska).

This framework suggests the three commonmetrics for studying urban economics: popu-lation growth, income growth, and housingprice growth. Nominal income growthimplies rising productivity. Housing pricegrowth implies a greater willingness of consu-mers to pay to receive a place’s bundle ofwages and amenities. Population growth cap-tures the quantity side of popular demand tolive in a city. No measure on its own impliessuccess but, taken together, the combinationof rising population, income and jobs suggestsurban health.4 Moreover, the three measurescan together tell us something about thenature of success. High wages accompaniedby stagnant housing prices and rising popu-lations suggest an increase in productivity.Housing prices that rise faster than nominalwages suggest that consumption amenitiesare increasing in the city.

We start with urban population growthbetween 1950 and today. Table 1 shows popu-lation growth by decade for the 10 cities thathad the most people as of 1950. We havefocused on cities (which are defined by theirpolitical boundaries) rather than metropolitanareas because cities correspond better to thetraditional dense cores of urban areas. Thesecities no longer dominate America’s urbanlandscape but, in the middle of the 20thcentury, these places characterised big citiesin the US. Except for Los Angeles, they areall in the Northeast or Midwest and locatedon major waterways. Moreover, generallythese cities grew large before mass auto-mobile transport. Eight of these cities werealso among the 10 largest cities in the US in1910 at the dawn of the automobile era.

The post-war period was generally not kind tothese places. Eight of the 10 cities today haveless than 80 per cent of their 1950 populations.Detroit, Cleveland and St Louis have lostmore than 47 per cent of the 1950 population.As Table 1 shows, the 1970s were the worstdecade for all of these cities; every city had itslowest population growth during that period. Itis true that US population growth slowedduring this decade, but it was still a respectable11.5 per cent and US population growth wasslower still in the 1980s, when these cities didmuch better. Moreover, population growth inthe 1970s declined 2 percentage points relativeto population growth in the 1960s, but theunweighted average of the growth rates ofthe cities declined from 24.6 per cent to214.1 per cent between the 1960s and 1970s.The 1970s were certainly the great period ofpopulation decline for America’s older cities.

But perhaps these data overstate the truesignificance of the 1970s. Urban decline ismediated by the durability of the housingstock (Glaeser and Gyourko, 2005). Citiesoften decline much more slowly than theireconomic situation might predict becausehousing remains and people want to live inthe houses. This theory tells us that, evenif a city’s decline is steady during periods ofdeclining household size, the populations ofdeclining cities will fall much more quicklythan during periods of steady household size.

URBAN RESURGENCE AND THE CONSUMER CITY 1277

Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

American household sizes have declinedsteadily since 1940, when there were 3.5people in a home, but the decline in householdsize in the 1970s was considerable. Duringthis decade, the average household size fellfrom 2.96 persons to 2.56 persons. Some ofthe extreme falls in population during thisperiod surely reflect declining householdsizes during that decade.

But there is evidence that confirms that thisis not the only thing that is special about the1970s. First, declining household size might

mean that cities that are losing populationshould lose more population, but it wouldnot mean that growing cities would startdeclining—yet New York City gained popu-lation in both the 1960s and 1980s but lostpopulation in the 1970s. Boston, likewise,started to gain population again in the 1980s.Moreover, the data on per capita incomegrowth and housing price growth alsosuggest that the 1970s were special.

In Table 2, we present housing price growthand per capita income growth for the 10

Table 1. Growth in top 10 cities by 1950 population

PopulationPercentage population growth

PopulationCity name 1950 1950s 1960s 1970s 1980s 1990s 2000

New York City, NY 7 891 957 21.4 1.5 210.4 3.5 9.4 8 008 278Chicago, IL 3 620 962 21.9 25.2 210.7 27.4 4.0 2 896 016Philadelphia, PA 2 071 605 23.3 22.7 213.4 26.1 24.3 1 517 550Los Angeles, CA 1 970 358 25.8 13.6 5.4 17.5 6.0 3 694 820Detroit, MI 1 849 568 29.7 29.5 220.4 214.6 27.5 951 270Baltimore, MD 949 708 21.1 23.5 213.1 26.5 211.5 651 154Cleveland, OH 914 808 24.2 214.3 223.6 211.9 25.4 478 403St Louis, MO 856 796 212.5 217.0 227.2 212.4 212.2 348 189Washington, DC 802 178 24.8 21.0 215.6 24.9 25.7 572 059Boston, MA 801 444 213.0 28.1 212.2 2.0 2.6 589 141

United States 150 697 361 19.0 13.3 11.5 9.8 13.2 281 421 906

Note: All data are from the US Bureau of the Census.

Table 2. Growth in top 10 cities by 1970 Population

Percentage populationgrowth

Percentage growth inincome per capita

relative to US growth

Percentage growth inowner-occupied housevalues relative to US

growth

City name 1970s 1980s 1990s 1970s 1980s 1990s 1970s 1980s 1990s

New York City, NY 210.4 3.5 9.4 226.0 20.6 212.6 235.4 120.8 232.0Chicago, IL 210.7 27.4 4.0 222.4 24.2 2.7 228.5 20.6 13.8Los Angeles, CA 5.4 17.5 6.0 217.8 0.1 220.6 41.6 56.8 249.2Philadelphia, PA 213.4 26.1 24.3 224.0 4.8 213.4 228.6 26.7 224.7Detroit, MI 220.4 214.6 27.5 227.3 226.5 2.3 272.8 229.9 79.0Houston, TX 29.4 6.4 15.1 6.2 219.8 29.8 30.6 232.3 211.8Baltimore, MD 213.1 26.5 211.5 222.1 7.6 29.4 6.1 14.2 220.2Dallas, TX 7.1 11.3 18.1 27.5 22.1 213.8 26.3 7.0 230.3Washington, DC 215.6 24.9 25.7 27.4 12.0 21.0 26.8 5.0 219.8Cleveland, OH 223.6 211.9 25.4 222.0 220.9 1.1 249.8 221.6 20.4

Notes: All data are from the US Bureau of the Census. Growth in income per capita and owner-occupied house values are

expressed as a difference from national growth.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

largest cities in the US as of 1970, in the1970s, 1980s and 1990s. We have normalisedincome growth and housing price growth bysubtracting the average growth of incomeand housing prices in the country as awhole. While this is a slightly differentsample of cities, it remains true that, for allof them except Houston, population growthwas lowest during the 1970s, despite the factthat US population grew by more in the1970s than in the 1980s.

The fourth, fifth and sixth columns showthe difference in growth in income per capitabetween these cities and the US as a whole.Again, for every city except Houston,growth in income was lowest during the1970s. Each of these cities, except Houston,had a significant decline in per capitaincome over the decade of the 1970s relativeto other Americans. This decline reflectsboth the fact that the economies of thesecities were weakening and also the fact thatwealthier people were fleeing these oldercities and often being replaced by poorerresidents.

The final three columns show the change inself-assessed housing values for owner-occupied housing. Again, we have normalisedby subtracting the growth in house values forthe country as a whole. For four of the top fivecities (Los Angeles is the exception), the1970s saw huge drops in housing values rela-tive to the US as a whole: New York, Chicago,Philadelphia and Detroit all had plummetinghousing prices relative to the US. Thechanges in housing values were somewhatmore mixed in the smaller cities. Houstonboomed in the 1970s. Washington, DC, andBaltimore also had rising housing pricevalues. Dallas had only a very modestdecline in housing values, but Cleveland’shousing market was a disaster.

Table 2 confirms that along every dimen-sion the 1970s were a remarkably bleakperiod for major American cities. Populationsdeclined. Incomes fell relative to the countryas a whole and housing prices in someplaces plummeted. During this period,people fled the older cities of the Northeastand Midwest for suburbs and for sun. Crime

soared and the future of the city was verymuch in doubt. The dark period of the 1970sexemplifies the basic downward spiral thathad struck so many of America’s largestcities after the Second World War. ThisAmerican decline was mirrored in the UK asLondon faced significant problems duringthe 1970s. The continental cities were differ-ent; Paris did not have the disappointing1970s that New York and London did.

American cities suffered during the 1970sbecause of two great secular changes: themove to sun and the move to sprawl. Duringthe 20th century, changes in transport technol-ogy led to massive changes both within urbanareas as people fled the central cities forsuburbs and across regions as people fledcolder places for warmer climes. Both ofthese changes should ultimately be seen asthe result of changing transport technologies.

The move to sprawl is ultimately almostentirely the result of the rise of the automobileand its remarkable advantages as a means ofmoving about. Cars are particularly importantfor cities because they require a vastly differ-ent urban form from what walking citiesrequire. Public transport suggests high den-sities and centralisation; cars require low den-sities and decentralisation. High urbandensities were required in a pre-car erabecause non-car methods of transport ulti-mately require foot traffic to get to and fromthe public transport stops. Centralisation isalso common because it is natural for thecity to cluster around public transport depotslike Grand Central Station or the downtownport of New York. Land use regulation alsocontributed to these patterns (see, forexample, Walker, 1977).

In walking cities, where we still find highdensities and much centralisation, people areunable to take advantage of the benefits con-ferred by an automobile. Cars use massivequantities of land and centralisation in aworld of automobiles only leads to conges-tion. Figure 1, which shows the relationshipbetween commute times and density acrossAmerican cities, makes it clear that auto-mobiles have not been able to eliminate thelong commutes found in dense areas but are


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

associated with much shorter commutes inless dense areas.

One might suspect that higher use of publictransport might make up for the ineffective-ness of the automobile in extremely densetraditional cities. On average, however, itturns out that public transport is muchslower than the automobile. The averagecommute by car is 23 minutes in the US andthe average commute using public trans-port is 47 minutes. This should not sur-prise us. Glaeser et al. (2000) document the15–18-minute fixed cost associated withpublic transport. For each form of publictransit used, a rider must get to a pickuppoint, wait for the bus or train, and thenwalk from the drop-off point to the final desti-nation. Although clearly not a causal relation-ship, Figure 2 shows the significant positiveassociation between commute times andpublic transport usage across American cities.

Since cars provide an enormous benefit infacilitating movement outside dense urbanareas, we should not be surprised that the auto-mobile has reorganised urban life and led to anexodus from urban areas. On the other hand,owning an automobile is expensive, can bedangerous, and requires learning how todrive. It is also associated with a different,low-density lifestyle which some peopleprefer and others dislike. The real wage datasuggest that, three decades ago, living atlower densities was an amenity for the bulkof people but that, today, preferences seemto have reversed for at least some citieswhose amenities have increased.

The move to sun is also best seen as a resultof the tremendous decrease in the price ofmoving goods over the 20th century. Theurban agglomerations that once lined thegreat lakes and the northern rivers of the UShad the great advantages that they wereclose to basic resources (like coal or lumber)and that they could access cheap water-borne transport. As the real costs of movinggoods declined by more than 90 per centover the 20th century (Glaeser and Kohlhase,2004), the need to agglomerate around riversor coal mines disappeared. Eventually, produ-cer cities located in places to increase pro-ductivity were replaced by consumer citieslocated in places which had amenities thatconsumers wanted, like warmth. Of course,the attractiveness of warm places alsoincreased with the advent of air conditioningand the eradication of malaria, hookworm,and other diseases in the American South.

This initial technology-driven exodus fromcold, dense cities was complemented by thetendency of poorer people to live inthe older, more centralised cities. Cars savetime, but cost money, so it should not surpriseus that poorer people with less money and alower opportunity cost of time decided tostay in the traditional urban areas (Glaeseret al., 2000). Moreover, in declining cities,housing prices decline to offset decreases inwages. This will make those cities particularlyattractive for those workers who have mar-ginal attachments to the labour market, butwho still need housing (such as those who

Figure 1. Commute times and densities across USmetropolitan areas.

Figure 2. Commute times and public transportuse.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

live off government support), and for thatreason we would expect urban decline to beaccompanied by an influx of poor people(Glaeser and Gyourko, 2005).

The importance of these factors is wellillustrated by the counter-examples of Franceand some other European cities. In continentalEurope, extremely high petrol taxes (oftenmore than 300 per cent) made cars far lessattractive to middle-income workers. More-over, smaller size and climatic homogeneitymeant that there was no compelling factorpushing towards regional realignment. As aresult, the traditional European cities neversuffered the tremendous difficulties of the1970s because they did not face the sameexodus to sun and sprawl.

Somewhat surprisingly, as Tables 1 and 2show, the post-war spiral of many ofAmerica’s largest cities had already levelledoff by the 1980s. For example, whileNew York lost 10 per cent of its populationduring the 1970s, in the 1980s it actuallygained population. The turnaround inhousing prices is even more dramatic. While6 out of 10 big cities lost housing value rela-tive to the US as a whole in the 1970s, 6 outof the same 10 cities had relative increasesin housing prices over the 1980s. Incomesrose with housing prices and, again, many ofthe same cities that had lost income (relativeto the US as a whole) in the 1970s gainedincome in the 1980s.

2.1 Consumption vs Production and theResurgent City

Why did the resurgence of big cities in the1980s and 1990s occur? To make somesense of this change, we must turn to basicelements of urban economics. Cities are theabsence of physical space between peopleand firms. The comparative advantage ofcities is reducing the costs of interactionthrough proximity. Much of economicgeography focuses on the role that reducingtransport costs play in attracting firms andincreasing their productivity (for example,Fujita, Krugman and Venables, 1999). Butthe role that cities play facilitating interactions

in non-market contexts may be just asimportant.

Cities are not just about production. AsTerry Nichols Clark (2004) has written, theyare also ‘entertainment machines’. Just asthe elimination of transport costs betweenfirms improves productivity, eliminatingtransport costs between people can radicallyalter social life. No one who has experiencedthe buzz of London or New York can doubtthat big, dense cities have a feeling all theirown and that this impacts on every part ofurban life. Of course, the impacts of proximitycan be both negative and positive. The resur-gence of big cities in the 1990s is connectedboth to a reduction in the negative social inter-actions and to an increase in positive socialinteraction.

Cities are about both consumption and pro-duction and, in a sense, the first question iswhether the urban renaissance after 1980had more to do with improving urban pro-ductivity or increases in the quality of life indense, urban areas. One approach to this ques-tion is to use the basic spatial economics fra-mework discussed above and to look atchanges in income and costs of living overtime. If dense US cities were becomingmore productive, relative to the US as awhole, then wages in those cities should berising relative to the US as a whole. If densecities were becoming more pleasant andattractive places to live, then real wages—i.e. wages adjusted for local costs ofliving—should be declining in those citiesrelative to the US as a whole. Following thebasic framework, if urban resurgence hasbeen based on consumer amenities, thenwages controlling for local costs of livingshould have fallen in these big cities. Ifurban resurgence has been based on risingurban productivity, then wages divided bylocal cost of living indices might increaseand nominal wages should certainly increase.

One piece of evidence on the relativeimportance of production and consumptionis given by Table 2. In this table, the growthin housing values between 1980 and 2000was much higher than the growth in percapita income for the big cities that did well


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

like New York and Chicago. In the 1990s,none of the big cities had incomes that rosemore than 3 per cent more quickly than percapita income in the US as a whole. Decadalchanges in housing values are more volatileand New York actually lost housing valuebetween 1990 and 2000 relative to the US,but, when we combine the 1980s and 1990s,New York and Chicago had sizeable increasesin housing values and corresponding declinesin the cost of living adjusted wage.

In Figures 3 and 4, we show the relation-ship between the logarithm of real wages inthe city and the logarithm of city populationin 1970 and 2000. The real wage variable isbased on the Census Individual Public UseMicro Sample. We have used these data tofirst run wage regressions where we correctfor age and schooling and then use thecity-fixed effect estimated by this wageequation. In this context, it is appropriate tocontrol for human capital variables, but notvariables that reflect local demand, so wedo not control for local industry structure.In principle, we must be able to explainwage patterns both from the firm’s andfrom the worker’s perspective. For firms to

pay higher wages, they must be more pro-ductive, so measures that predict productivityshould also predict higher wages. But forworkers to receive lower real wages in anarea, they must be receiving some otheramenity. This regression is oriented towardsunderstanding amenity levels by looking atthe real wage, not towards understandingproductivity.5

Figure 3 shows the relationship between thelogarithm of population density in 1970 andreal wages across 26 large metropolitan areas.There is a strong positive relationship. Asmetropolitan area population rises by 1 logpoint (approximately 100 per cent), realwages rise by 0.042 log points (approximately4 per cent). The same relationship holds in1980. By 1990, there is no relationshipbetween city size and real wages and, in2000, the relationship is negative. Figure 4shows the connection between real wages andcity population across the same 26 metropoli-tan areas and shows that, as metropolitan areapopulation rises by 1 log point, real wagesdecline by 0.070 log points. This negativerelationship also holds when we examine amore complete set of cities (local price data

Figure 3. Log of real wages and city size, 1970.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

are unfortunately not available for this largerset of cities in 1970 or 1980).

The natural interpretation of these results isthat, in the 1970s, workers needed to be com-pensated for living in big cities. These citiessuffered from crime and other disamenities.In 2000, crime rates had fallen and risingincomes led people to value urban social ame-nities more strongly. As a result, real wages arenow lower in big cities than in smaller areas.

Figures 5 and 6 show how real wagesdeclined. This decline does not come from

declining nominal wages, but rather fromrising prices in big, dense cities. Figure 5shows the relationship between the changein the logarithm of median house valuebetween 1980 and 2000 across metropolitanareas within the US (Anchorage is excluded).The raw correlation is 40 per cent and asdensity rises by 1 log point (approximately100 per cent), housing price growth was0.08 log points higher (approximately 8 percent). People’s willingness to pay to live inthese denser cities has risen dramatically as

Figure 4. Log of real wages and city size, 2000.

Figure 5. Housing value changes and urbandensity, 1980–2000.

Figure 6. Income growth and population density,1990–2000.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

the crime rates in these areas have fallen andas these places have become more attractiveplaces in which to live. Figure 6 shows therelationship between changes in per capitaincome and the logarithm of density acrossmetropolitan areas within the US. Therelationship is strongly negative, as wewould expect if cities have become moreattractive with time and people have becomemore willing to live there.

In Table 3, we show the elasticity of percapita income, median housing values andmedian rents with respect to city populationfor cities in 1970, 1980, 1990 and 2000. Thefirst column shows that the elasticity ofincome with respect to city population risesinsignificantly over the time-period from0.085 to 0.095. The second column showsthe elasticity of housing values with respectto city population which soars from 0.09 to0.19 in 1990 and 0.15 in 2000. The thirdcolumn shows a dramatic jump from 0.05 to0.11 in the elasticity of median rent withrespect to metropolitan area populationbetween 1980 and 1990. The connectionbetween housing prices or rents and city popu-lation has become much stronger over time.The connection between per capita incomeand population is essentially flat.

The flat connection between nominalincome and city population over time cer-tainly suggests that urban labour productivityis not rising relative to other areas, but thereis other evidence that corroborates thatclaim. The decentralisation of employmentwithin metropolitan areas continued unabatedduring the 1990s as citizens chose to live inthe suburbs and as central-city land becamemore expensive (Glaeser and Kahn, 2001).The increase in reverse commuting wherepeople live in the city but work in thesuburbs further supports the notion thatthe urban renaissance is driven by consump-tion, not production. Together with the evi-dence on housing prices, this evidencestrongly suggests that people began activelyto choose to live in cities in the 1980s and1990s rather than just work there.

This evidence pushes us towards a con-sumption-based theory of urban renaissance.While it is true that big cities remain moreproductive than smaller places, this has notchanged over time. The remarkable changehas been in the desire of people to live incities, not in the wages paid in these places.Figure 7 shows the time-series of a surveyquestion administered by DDB Needham(1998) asking people whether they wouldprefer to live in a big city or a small town.The basic pattern is striking. Around 1985,survey respondents became much more enthu-siastic about big cities than they had been pre-viously. This survey evidence confirms whatthe data on real wages have already suggested.Big cities became much more attractive placesin which to live after 1980. In section 3, we tryto understand why that occurred, but first weput this resurgence in perspective.

2.2 Resurgent Cities and Sprawl

In the previous section, we have argued thatthe 20 years since 1980 have been muchbetter for America’s biggest cities than the20 years before 1980. While this is surelytrue, it should not blind us to the fact thatthe general trend to sun and sprawl has contin-ued relatively unabated over the past 20 years.While some cities have improved relative to

Table 3. Changes in income and housing priceelasticity with population

Incomeelasticity

Housing valueelasticity

Rentelasticity

1970 0.083 0.092 0.067[0.007] [0.012] [0.010]

1980 0.077 0.088 0.050[0.008] [0.014] [0.007]

1990 0.095 0.194 0.108[0.008] [0.019] [0.009]

2000 0.095 0.15 0.103[0.008] [0.016] [0.009]

Notes: Income data are per capita income from the Bureau of

Economic Analysis. Housing value is the median housing

value reported in the US Census for that year. Rent is the

median rent reported in the US Census for that year. All

regressions report the coefficient where the logarithm of

the dependent variable is regressed on the logarithm of

metropolitan area. All estimates are based on a common

set of 318 metropolitan areas.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

the catastrophic 1970s, the fundamental fact isthat the places oriented around the car havecontinued to gain relative to places orientedaround public transport, that suburbs andedge cities have continued to grow relativeto traditional downtowns and that density con-tinues to predict decline.

First, Table 1 itself shows that, even in the1990s, every one of the 10 cities includedlost population relative to the US as a whole,which grew by 13 per cent in the 1990s. Inthe 1990s, there was a weak negative corre-lation between population and growth and astrong 230 per cent correlation betweendensity and growth (Glaeser and Shapiro,2003). Across metropolitan areas, the corre-lation between density and growth is 220per cent. Despite the claims of some newurbanists about how demand for high-densityliving is on the rise, more dense areas contin-ued to lose population relative to denser areasin the 1990s.

Likewise, public transport continues topredict urban decline. Across cities, the corre-lation between the percentage of workers whodrive to work and urban growth is also 30 percent. In a regression with controls for initialpopulation, median age, land area and initial

income, cities or metropolitan areas wheremore than 5 per cent of the population takepublic transport grew at least 5 per centmore slowly than cities where less than 5 percent of the population take public transport.Figure 8 shows the robust correlationbetween cars per capita in 1990 and popu-lation growth between 1990 and 2000 acrosscities with more than 100 000 inhabitants.

Just as low-density, car-based living con-tinues disproportionately to attract residents,the Sunbelt continues to grow at the expenseof colder climes. Across cities, the correlation

Figure 7. Time series on liking cities, 1975–95.

Figure 8. Population growth and cars per capita,1990–2000.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

between mean January temperature andmetropolitan area growth is 42 per cent. Thecorrelation between this variable and citygrowth is 35 per cent. The key point of thesefacts is that, while New York, Boston andChicago have seen a renaissance, the bigstory in American society continues to bethat of car-based living in the warmer climatesof the country. The good news from the 1980sand 1990s is that this rise does not need tomean complete disaster for the older cities.

3. Consumption and Urban Resurgence

In this section, we attempt to understand whythe desire to live in cities like New York,Chicago and London surged in the 1980sand 1990s. While it is certainly true that theincreasing value of information in theeconomy has helped the economies of bigdense places, it is also true that rents have gen-erally risen faster than wages in these areasand that wages adjusted for cost of livinghave fallen in these places relative to othersmaller cities. Hence, we focus on thechanges in consumer amenities in theseplaces and the changes in demand for the con-sumer amenities that are provided primarily inthe biggest places. We begin with a brief his-torical overview of cities as centres for con-sumption—rather than production. We thenturn to understanding how cities changed ascentres for consumption after the SecondWorld War.

3.1 Consumption in Cities

Historically, the most important limit on citygrowth was an urban disamenity—disease.Until the 20th century, cities were killing-fields. Just as physical proximity speeds theflow of products or ideas, it speeds the flowof germs and bacteria. The lack of health inurban areas should be understood as a nega-tive social interaction. In a world withoutmassive public infrastructure for water andpipes, clean water was almost impossible toget in dense areas. Together, these factorsreduced life expectancy by 5 years or morefor people living in larger cities before 1900.

The public provision of clean water issurely the single biggest improvement in thequality of urban life in history. Over the past200 years, cities have increasingly madeinvestments in aqueducts and sewagesystems that have completely changed thehealth costs of living close together. Someof these improvements are the result of tech-nology, but many owe as much to improve-ments in the quality of urban governmentand the ability to borrow (Glaeser, 2004;Cutler and Miller, 2006). All told, thesuccess of cities in the late 19th and early20th centuries owes a great deal to thereduction of urban health hazards and, today,life expectancies are often longer in urbanareas.

Offsetting the negative effect of urbandisease, cities traditionally offered a muchricher social life than rural areas. Beforeradio and television, proximity to otherhuman beings provided the best form of enter-tainment and this was most available in bigcities. Novels, like Dreiser’s Sister Carrie,that discussed the move from rural to urbanliving in 19th century America, emphasisethe dreariness of low-density living and therelative charms of moving to a great city.Probably the main social advantage of citiesin this time-period was proximity to otherpeople, but there were other advantagesassociated with city life such as live theatre,restaurants and other commercial forms ofentertainment. Cities specialised in thesethings both because these industries involvefixed costs, which can be spread over a largerconsumer-base in cities, and because denseurban areas reduce the travel costs for consu-mers and producers.

Over the first half of the 20th century, theurban social advantage abated because ofentertainment technologies, including radio,movies, television and the car, that improvedthe social opportunities available in low-density living. The radio and televisionmeant that consumers no longer had to travelto entertainment venues to hear music orexperience theatre. Movies allowed perfor-mers to stay in Hollywood and still entertainmillions. The car meant that people found it


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

easier to travel to restaurants despite living inlow-density areas. In a sense, all of these inno-vations can be seen as revolutions in transportor reductions in the cost of delivering servicesover distance. As always, reductions in trans-port costs tend to reduce the advantages thatcome from people living close together.Even if there had not been a collapse inurban safety, these technologies would haveincreased the consumption amenities in low-density places relative to dense cities.

The other important change in big cities inthe post-war period was the rise in crime after1960. Figure 9 shows the long-run patterns ofcrime in New York City. This figure focuseson homicide because this crime statistic isthe best measured and the least likely tomove because of changes in reporting technol-ogy. There have been three great periods ofrising homicide rates in New York City.Between 1820 and 1870, the homicide rateincreased by 0.15 per 100 000 per year andessentially tripled over that 50-year period.This was the era when New York transformeditself from a town of 100 000 to a city of1 000 000. Between 1890 and 1935, the homi-cide rate per 100 000 rose by 0.08 per year onaverage which effectively doubled the crime

rate. Some of this rise might be associatedfrom the new immigration of the 1890s;some of the rise might be associated with pro-hibition. Crime rates fell between 1935 and1960, possibly due to the end of prohibitionand improvements in law enforcement.

As the graph shows, though, the most spec-tacular increase in crime occurred between1960 and 1980. Over 20 years, the murderrate in New York City exploded, rising by1.1 per 100 000 per year. Homicides essen-tially quadrupled over a 12-year period. Asthe numbers and figure make clear, this riseis of a completely different order of magnitudefrom any other comparable period inthe history of New York City. In New York,the 1970s were a far more deadly decadethan any time before or since. While crimerose everywhere in the US, the increase wasconcentrated in the big, dense cities and, by1970, there was a powerful positive relation-ship between crime and city population(Glaeser and Sacerdote, 1999).

We should not be surprised that there ismore crime in big cities. Proximity facilitatescrime as much as any productive activity—after all, transport costs are particularlyimportant when you are carrying illegal

Figure 9. Homicides in New York City, 1800–2000. Source: Monkonnen (2001).


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

goods. Moreover, high levels of populationoften make crime-solving more difficultbecause, in a large city, there are inevitablymore suspects for a given crime. Perhapsmost importantly, cities appear to be relatedto social breakdowns of all sorts, in partbecause it is easier to escape community sanc-tions (see Glaeser and Sacerdote, 1999, for adiscussion). As a result, there has been a long-standing connection between urban size andcrime and, during much of the 20th century,the breakdown in urban law and orderserved to deter the resurgence of the largestcities.

The rise of crime in New York between1960 and 1975 is mirrored in many of thegreat cities of the developed world. Thisperiod combined economic distress, as manu-facturing fled the older urban areas, withincreasing social freedom and an increasingunwillingness to allow the police brutalitythat had been such a part of crime preventionbefore 1960. It has been difficult to establishthe causes of the great crime explosion ofthe 1960s and 1970s. The largest single corre-late of crime rates across cities in those yearswas the share of households with single-parent heads, which suggests the importanceof breakdowns in traditional social structures.However, arrest rates also declined over thisperiod and economic distress also increasedin many of the larger American cities. Further-more, an increasingly youthful population canexplain some of the increase as well (aboutone-fifth of the rise over the period).

While it is difficult to establish the impactthat crime had on urban distress in the1970s, Berry-Cullen and Levitt (1999) showthat an increase of 1 reported crime percapita reduces city population by 1 per cent.Between 1960 and 1980, reported crimes percapita in New York rose from 0.02 to 0.1.This increase is spectacular, but the Berry-Cullen and Levitt coefficient estimatesuggests that it would only predict a popu-lation decline of 0.08 per cent. Given thatNew York’s population fell by almost 10 percent in the 1970s, this estimate suggests thatthe increase in crime can only explain amodest amount of the decline in population.

While the rise in crime cannot explain thefall in New York’s population, it can poten-tially explain the decline in housing values.Schwartz et al. (2003) estimate an elasticityof property value with respect to the violentcrime rate of 0.15. Thaler (1978) estimates alower elasticity of 0.05. Violent crime ratesincreased almost tenfold from 0.0024 to0.021 between 1960 and 1980. Even thelower Thaler estimate would predict a 40 percent drop in housing prices, which roughlycoincides with the decline observed inNew York relative to the country as a whole.Given this coefficient estimate, the drop inNew York City property values can beexplained by the increase in crime. In otherlarge cities, with similar increases in crime,property value declines can also be attributedto the rise in crime.

3.2 Urban Amenities since 1980

In section 2, we argued that the evidencesuggests that urban resurgence in the US hadmore to do with rising consumer amenitiesthan with rising productivity. The strongestevidence for this was the declining connectionbetween real wage and city population. A sec-ondary piece of evidence was the opinionpolls which suggest an increased desire tolive in big cities.

This change can conceivably come fromtwo sources. First, city amenities may indeedhave changed. For example, crime rates haveplummeted since 1980 in many large cities.Some cities have invested more in quality oflife. Secondly, the willingness to pay forurban amenities relative to the amenities ofsuburbs may have increased. For example, itmay be that the cities have always had moremuseums, but that the value placed onproximity to museums rises with income andeducation. In this case, the increased attrac-tiveness of the city comes from a greater will-ingness to pay for a constant bundle of urbanamenities.

We will not be able perfectly to separate outthese two explanations, but we will be able toput together some evidence on both of thesehypotheses. We begin with changes in the


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

crime rate. As Figure 9 shows, the periodsince 1980 and particularly since 1990 hasseen a remarkable decrease in the homiciderate in New York. Within the US as awhole, violent crime victimisation hasdeclined by 50 per cent since 1994. Almostall the decline has been centred in the largeAmerican cities (which, after all, had muchmore crime to begin with) and, as a result,the difference in crime rates between citiesand smaller areas is much smaller than itever has been.

Measuring crime other than homicides isproblematic because reported crimes arepotentially subject to changes in reporting ten-dencies and there is substantial mismeasure-ment in victimisation surveys. Keeping thisin mind, in Figure 10, we show the trends invictimisation from both robbery and burglaryin New York and Chicago. These figuresshow 3-year moving averages of the victi-misation rates from the National Crime Victi-misation Survey. In both cities, there has beena substantial decline in self-reported victimi-sation from both types of crime. Comparingthe peak with the trough, the burglary victimi-sation rate declined from 5 per cent in 1981 in

New York to 1 per cent in 1998. There are alsosignificant drops in robbery shown inFigure 10.

How much of the resurgence of New Yorkand other cities can be attributed to thereduction in crime? Since 1980, the crimerate in New York has fallen from 0.1 to0.03. Using the Berry-Cullen and Levitt esti-mate, this would predict a 0.07 per centincrease in population over this time-period.Just as crime cannot explain New York’sdecline in the 1970s, the fall in crime ratescannot explain the rise in population overthe later time-periods. Similar results occurfor other cities. The Berry-Cullen and Levittestimate, which is currently by far the bestavailable, suggests that changes in populationare unlikely to be driven by the reduction incrime.

There is a better chance that the decrease incrime can explain rising property values.Schwartz et al. (2003) estimate that approxi-mately one-third of New York’s housingprice appreciation after 1988 can be explainedthrough declining violent crime rates. After1980, there was a 50 per cent drop in theviolent crime rate in New York City. Using

Figure 10. Crime victimisation in New York and Chicago, 1980–2000.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

the Thaler estimate, this would predict a 2.5per cent rise in housing values. The largerSchwartz et al. (2003) estimates suggest thatthis might explain a housing price increaseof up to 10 per cent, which is far less thanthe actual price increase.

Overall, we are led then to the view thatcrime rates are important, but that decliningcrime can only explain a modest amount ofthe increased demand for New York Cityand other big cities. Even though crime ratesfell dramatically, the generally accepted esti-mates of the relationship between eitherprices or population and crime suggest thatdeclining crime is only part of the story. Wetherefore now turn to other reasons for theincreasing attractiveness of dense places asconsumer cities.

While cities have always had negativesocial aspects—not all proximity is good—they have also had some positive socialfeatures. Certainly, the most obvious socialadvantage is the reduced cost of face-to-facecontact with other human beings. The powerof this advantage is most obvious in the ten-dency of many cities to attract single peopleand to act as a marriage market. By facilitating

matching, cities make it easier to connect witha mate, just as cities make it easier to connectwith the right employer. More generally, big,dense urban areas support connections of allkinds and this is part of what makes them fun.

Beyond reducing interaction costs, urbandensity also supports forms of consumptionsuch as museums, restaurants, bars, movietheatres and concert halls. Most of these facili-ties have some form of fixed costs and higherdensities make it possible to generate enoughcustomers to cover the fixed costs. Figure 11shows the relationship between the numberof museums and metropolitan area populationin 1990. As we all know, there are moremuseums in big places. Furthermore, sincespecialisation and the division of labour arelimited by the extent of the market, bigcities allow there to be more types ofmuseums, restaurants, bars and so forth.

To investigate whether cities actually facili-tate different forms of entertainment, Table 4uses the DDB Needham Life Style Survey(1998) and looks at the connection betweencentral-city residence and different forms ofleisure activities. The DDB Needham surveypresents categorical responses about how

Figure 11. Museums and metropolitan area population, 1990.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

Table 4. Social activities and cities

(1) (2) (3) (4) (5) (6) (7)Visited an art

gallery ormuseum

(n ¼ 80 260)

Went to a baror tavern

(n ¼ 67 802)

Went out todinner at arestaurant

(n ¼ 76 991)

Went to themovies

(n ¼ 79 849)

Went to a popor rockconcert

(n ¼ 80 326)

Went to aclassicalconcert

(n ¼ 80 224)

Entertainedpeople in my

home(n ¼ 79 919)

City resident 0.1891 0.051 0.1071 0.202 0.1379 0.1479 20.0606[0.0112]�� [0.0125]�� [0.0112]�� [0.0111]�� [0.0113]�� [0.0112]�� [0.0112]��

Suburb resident 0.0808 20.0049 0.0636 0.1149 0.0749 0.0531 20.0062[0.0100]�� [0.0112] [0.0101]�� [0.0099]�� [0.0101]�� [0.0101]�� [0.0100]

City x after 1990 20.0064 20.0057 20.0726 0.0497 20.0008 0.0093 0.076[0.0164] [0.0173] [0.0167]�� [0.0163]�� [0.0166] [0.0165] [0.0164]��

Suburb x after 1990 20.0041 0.0366 20.0132 0.0577 0.0092 0.0136 0.0435[0.0144] [0.0153]�� [0.0147] [0.0143]�� [0.0146] [0.0145] [0.0144]��

Survey year 20.0055 0.0022 20.0023 20.0065 0.0023 20.009 20.0307[0.0008]�� [0.0011]�� [0.0008]�� [0.0008]�� [0.0008]�� [0.0008]�� [0.0008]��

College graduate 0.3581 20.0589 0.1339 0.1579 0.0569 0.3434 0.0103[0.0086]�� [0.0091]�� [0.0086]�� [0.0085]�� [0.0086]�� [0.0086]�� [0.0085]

High school dropout 20.1658 0.0104 20.1834 20.1152 20.0074 20.1217 20.1125[0.0114]�� [0.0125] [0.0115]�� [0.0113]�� [0.0115] [0.0114]�� [0.0114]��

Age 0.0101 20.018 0.008 20.0288 20.0285 0.0066 20.0164[0.0014]�� [0.0016]�� [0.0015]�� [0.0014]�� [0.0015]�� [0.0014]�� [0.0014]��

Age squared 20.0001 0.0001 20.0001 0.0002 0.0002 0 0.0002[0.0000]�� [0.0000]�� [0.0000]�� [0.0000]�� [0.0000]�� [0.0000] [0.0000]��

Female 0.095 20.3808 20.0411 0.0119 20.0329 0.068 0.1103[0.0071]�� [0.0076]�� [0.0071]�� [0.0070] [0.0071]�� [0.0071]�� [0.0070]��

Log income 0.0747 0.0593 0.225 0.114 0.0564 0.046 0.0725[0.0050]�� [0.0059]�� [0.0051]�� [0.0050]�� [0.0051]�� [0.0051]�� [0.0050]��

Income data missing 0.0515 0.0069 0.0253 0.0364 0.032 0.0741 0.0334[0.0194]�� [0.0195] [0.0197] [0.0192] [0.0196] [0.0194]�� [0.0193]

Constant 9.687 23.7707 2.0709 12.7154 24.1534 16.9775 60.3858[1.6537]�� [2.1055] [1.6697] [1.6384]�� [1.6711]�� [1.6579]�� [1.6488]��

T-test for residence 0.0000�� 0.0000�� 0.0000�� 0.0000�� 0.0000�� 0.0000�� 0.0000��

T-test for interactions 0.8926 0.0163�� 0.0005�� 0.6295 0.5570 0.7995 0.0544

Notes: Data are from DDB Needham Life Style Survey. Standard errors are in brackets.�� indicates significance at 5 per cent and �� indicates significance at 1 per cent. The last two lines

report p-values for t-tests of the null hypotheses that: the city coefficient equals the suburbs coefficient; and, the coefficient on city x after 1990 equals the coefficient on suburb x after 1990.

Controls for Black, Asian, other race, missing race, marital status and missing marital status are included in the regression but coefficients are not reported. Data are from 1975–98 except

that question (2) was only asked from 1979 to 1998.

UR

BA

NR

ES

UR

GE

NC

EA

ND

TH

EC

ON

SU

ME

RC

ITY

12

91

Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

often these activities occur—never, between 1and 4 times per year, etc. Thus, the coeffi-cients in the table are not easy to interpret.Nevertheless, the table gives a qualitativesense about how big city living enablescertain forms of entertainment.

We split the population into three groups:those who live centrally in a city with morethan 50 000 residents, those who live in thesuburbs of those cities, and those who live incities with less than 50 000 people or outsidecities altogether. We also interact these vari-ables with a dummy variable that takes on avalue of 1 if the year is after 1990. This inter-action is meant to capture the possibility thatcities are getting better at delivering theseactivities over time.

We have also controlled for marital status,education, age, year (the survey data includeresults from 1975 to 1998), race, sex, the log-arithm of income, and dummy variables forwhether the income variable, marital statusvariable, and race variable are missing. TheDDB Needham survey does not reportincome per se, but rather a range of differentvalues for income. We have replaced thisrange with its mid-point and deflated so thatall income is in 2000 dollars. Since manyrespondents refused to give their incomelevels, we assign those respondents theaverage income in the sample and include adummy variable for missing income.

The first regression looks at the determi-nants of going to an art museum. Living in acentral city positively predicts having visitedan art museum. Living outside a cityaltogether strongly predicts not attending amuseum. There is no significant interactionbetween place of residence and time. Thepopulation averages tell us that 49 per centof central-city residents have attended amuseum in the past year, 45 per cent of subur-banites have attended a museum in the pastyear, and 33 per cent of non-city residentshave attended a museum in the past year. Ina probit formulation (not reported) with thesame controls, central-city residence raisesthe probability of attending a museum by 5per cent and living outside a city altogetherlowers the probability of attending a

museum by 4.5 per cent. Of course, thesecoefficients represent both treatment andselection. Significantly, museum attendancerises with both education and income. Asmuseum attendance appears to be a normalor even a luxury good, economic theorywould tell us that the demand for museumsshould rise as incomes rise in a country.

The second regression looks at the fre-quency with which people attend bars.Central-city residence increases the frequencyof this activity. Living in a suburb has noimpact on the probability of this activity. Inthe regression, this coefficient is not signifi-cant, but in a probit (not reported) where theprobability of having been to a bar at all inthis past year is regressed on residence, non-city residents are 3.3 per cent less likely tohave gone out drinking and this difference isstatistically significant. There is no significantinteraction between this central-city status andthe dummy variable for being after 1990.However, the impact of suburbs on going toa bar or tavern does appear to have risenwith time. The demand for this activity riseswith income.

The third regression looks at having goneout to a restaurant for dinner. Central-city resi-dents are more likely to have been out to a res-taurant than non-city residents, but this effectbecomes muted after 1990. Suburbanites arealso more likely to have gone out to dinnerthan non-city residents. Again, the incomeelasticity is positive.

In the fourth regression, going out to themovies is regressed on city residence. This isamong the strongest coefficients in the table.Furthermore, the central-city effect gets stron-ger over time. In a probit formulation, central-city residents are 3 per cent more likely tohave attended more than 5 movies in thepast year and non-city residents are 6 percent less likely to have attended more than 5movies. This is also a leisure activity withpositive income elasticity.

In the fifth regression, we examine havinggone to a pop or rock concert. Again,central-city residence strongly predicts morefrequent concert-going and non-city residencepredicts not going to a concert. In a probit


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

formulation, central-city residence increasesthe probability of having been to a concertby 1.4 per cent and non-city residencereduces this probability by 4.8 per cent.These changes might seem small except thatonly 15 per cent of the sample attended aconcert at all.

The sixth regression examines going to aclassical concert. Again, there is a strongcentral-city effect and there is little interactionbetween this effect and time. The seventhregression looks at entertaining guests athome. Here, there is a negative effect ofliving in a central city, but this effect disap-pears after 1990.

Taken together, these coefficients suggestthat city residence supports almost all social,leisure activities. Going to concerts, restau-rants, museums and the movies are morecommon in central cities. However, the timeeffects are generally weak so there is little evi-dence that this effect has changed over time.Thus, urban resurgence cannot be explainedby changes in the accessibility of theseactivities.

However, another explanation is that thevalue placed on urban amenities has risenover time, especially by people who are richenough to pay for big city residence. Foreach activity, income increases the tendencyto engage in the social activities in Table 4.In every case, except for going to a bar, edu-cation increases the tendency to engage inthese social activities. Hence, one reasonableview is that urban amenities have gotmore valuable over time as people havebecome richer or more educated. It is impossi-ble to go from a table like Table 4 to willing-ness to pay to be in New York. Hence, thismust remain a conjecture, but it seems to usto be the leading candidate for explainingthe revitalisation of downtown areas.

3.3 Immigrant Cities

No discussion of urban resurgence would becomplete without recognising the enormousimportance that immigration played inbuoying big city populations in the 1990s. In2000, more than 35 per cent of the population

of New York City was foreign-born. US- bornpopulation growth in places like New Yorkand Chicago continued to be negative in the1990s; immigrants provided the increase inpopulation. Across metropolitan areas, thecorrelation between log of total populationand the foreign-born share of the populationwas 35 per cent in 1970 and 48 per cent in2000. Thus, the rise in immigrant populationsis not just the result of rising levels of immi-gration into the US, but also the result ofimmigrants’ increasing attraction to big cities.

Historically, the link between immigrationand big cities had at least three roots. Mostsimply, cities like New York and Bostonwere ports that served as gateways intoAmerica. In the 19th century, high transportcosts made it attractive for immigrants tostay where they landed. Over the 20thcentury, these transport costs have becomeless important, but other factors continued toattract immigrants to big urban areas. Wecan separate these factors, as well, into con-sumption and production. First, big citieshad a number of amenities, including existingconcentrations of immigrants, making themnatural places for recent immigrants to live.The fact that big cities do not require residentsto own multiple cars continues to be a potentforce attracting poorer immigrants to denseareas. Secondly, big cities have economiesthat do a better job of incorporating theforeign-born, especially when those foreign-born lack good English skills.

The influx of immigrant populations tolarge cities in the 1990s shows the continuingimportance of these factors. Cutler et al.(2005) show that recent immigrants havebeen particularly attracted to high-densityareas with public transport. While richernative populations have fled to car-basedliving in the exurbs, poorer immigrantsremain drawn to the less expensive areas oftraditional cities that allow them to avoidbuying and maintaining cars.

4. Cities, Sprawl and Social Capital

The social advantages of big cities that wehave stressed might lead some people to


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

think that cities have a real advantage in theformation of social capital—i.e. socialgroups and networks. Social capital, by itsessence, involves externalities and if a socialtrend, like sprawl, is indeed destroying com-munity in America, then it makes sense tothink about policies that stemmed that socialtrend. In this section, we investigatewhether, as Robert Putnam (2000) suggests,

suburbanisation appears to reduce socialcapital. After all, we have been arguing thatcities facilitate various forms of social con-nection, so it might well be reasonable tobelieve that the move to car-based living hasgenerally reduced our interactions.

Putnam’s Bowling Alone claims that thesuburbanisation of the American populationis one of the prime causes of declining

Table 5. Civic engagement and cities

(3)(1)

Attendedchurch or

other placeof worship

(n ¼ 79686)

(2)Worked

on acommunity

project(n ¼ 80069)

Wrote aletter toan editor

of a magazineor newspaper(n ¼ 45037)

(4)Contacteda publicofficial

(n ¼ 6396)

(5)Are you aregistered

voter?(n ¼ 7232)

City resident 20.1516 20.1236 0.0427 20.1631 20.0568[0.0111]�� [0.0112]�� [0.0204]�� [0.0340]�� [0.0325]

Suburb resident 20.1401 20.1068 20.0003 20.0917 20.0589[0.0099]�� [0.0100]�� [0.0188] [0.0294]�� [0.0303]

City resident x 0.0178 0.0496 20.0063after 1990 [0.0163] [0.0164]�� [0.0245]

Suburb x after 1990 20.0113 0.0273 0.0179[0.0143] [0.0144] [0.0224]

Survey year 20.0127 20.0112 20.0003 0.4363 0.1543[0.0008]�� [0.0008]�� [0.0032] [0.2542] [0.1195]

College graduate 0.1847 0.2242 0.0858 0.2075 0.1554[0.0085]�� [0.0085]�� [0.0114]�� [0.0311]�� [0.0264]��

High school dropout 20.1974 20.1442 20.0192 20.1035 20.4221[0.0113]�� [0.0114]�� [0.0169] [0.0380]�� [0.0442]��

Age 0.0176 0.0141 0.0055 0.0306 0.0269[0.0014]�� [0.0014]�� [0.0020]�� [0.0055]�� [0.0047]��

Age squared 20.0001 20.0001 0 20.0003 20.0001[0.0000]�� [0.0000]�� [0.0000] [0.0001]�� [0.0000]��

Female 0.2624 0.0808 20.0121 20.1404 0.084[0.0070]�� [0.0070]�� [0.0096] [0.0252]�� [0.0229]��

Log income 20.0086 0.0385 20.0043 0.0475 0.161[0.0050] [0.0050]�� [0.0075] [0.0259] [0.0182]��

Income data missing 20.0131 0.0155 0.0218 20.0109 0.0622[0.0192] [0.0193] [0.0227] [0.1041] [0.0664]

Constant 24.3441 21.4723 0.4044 2865.5804 2310.1723[1.6346]�� [1.6508]�� [6.4533] [503.7789] [238.3129]

T-test for residence 0.2687 0.1095 0.0207�� 0.0235�� 0.9374T-test for

interactions0.0820 0.1860 0.2916

Notes: Data are from DDB Needham Life Style Survey. Standard errors are in brackets. �� indicates significance at 5 per cent

and �� indicates significance at 1 per cent. The last two lines report p-values for t-tests of the null hypotheses that: the city

coefficient equals the suburbs coefficient; and, the coefficient on city x after 1990 equals the coefficient on suburb x after 1990.

Controls for Black, Asian, other race, missing race, marital status and missing marital status are included in the regression but

coefficients are not reported. Data for questions (1)–(3) are from 1975 to 1998, question (4) was only asked in 1981/82 and

question (5) was only asked in 1994/95.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

social capital. He particularly emphasises longcommutes as a distraction from civic engage-ment. His primary evidence for this claim isthat people who live in small towns aremore engaged in civic associations. This factis certainly true, but it is hard to see how theprevalence of social capital in small townsprovides evidence on sprawl and suburbanisa-tion. If sprawl is low-density, suburban living,then central cities and high densities are itsopposite—not small towns. If sprawl deterscivic engagement then, by definition, weshould expect to see more civic engagementin high-density places and less civic engage-ment in the suburbs.

In Table 5, we look at the determinants ofcivic engagement. We use the same DDBNeedham survey data and our methodologyis the same as in Table 4. We look at fivedifferent types of social engagement: attend-ing church, working on a communityproject, writing a letter to a magazine or news-paper, contacting a public official, and being aregistered voter. For four of these five vari-ables, central-city residence predicts lesssocial engagement than residing in a non-city area such as Putnam’s small towns.Only in the case of writing a letter to themedia does central-city residence predictmore civic engagement. Living in thesuburbs also appears to have negativeeffects, but if anything these effects are lessnegative than those for centre cities. Statisticaltests confirm this finding. The final row ofTable 5 reports the p-value for a t-test forthe equality of the coefficients on city resi-dence and suburban residence. Only two ofthe outcomes (writing a letter and contactinga public official) have a significant difference(p , 0.05) and one of these (contacting apublic official) happens more in the suburbs.The fact that there is no positive correlationbetween central-city residence and civicengagement should make us immediatelysceptical about the view that sprawl isdestroying social capital.

Since sprawl is ultimately about density, wenow look at whether civic engagement isgreater in high-density metropolitan areas.Figure 12 shows the relationship between

metropolitan area density and working oncommunity projects. The regression line isfitted weighting the observations by theinverse of the square root of the number ofobservations of the dependent variablewithin the metropolitan area. This weightingis necessary to correct for the larger standarderrors of the estimates of community engage-ment in smaller areas. The basic relationshipis weakly negative. Denser, less sprawlingplaces are less likely to have residents whowork on community projects. Figure 13repeats this using volunteering as the depen-dent variable. Again the relationship isnegative.

Table 6 shows these relationships in aregression form. Again, all of theseregressions have been estimated weightingby the inverse of the square root of thenumber of observations of the dependent vari-able. We have also controlled for the share ofadults with college degrees, the median age inthe city, the median January temperature andper capita income. In all of the first threeregressions, density decreases civic engage-ment. Volunteering, working on communityprojects and going to club meetings alldecline with population density. In the fourthregression, we find that attending churchrises with density. The fifth regression showsresults for another variable which is occasion-ally linked to social capital: believing thatpeople are honest. Again, denser placesappear to have less, rather than more, socialcapital.

Figure 12. Community activism and urbandensity across metropolitan areas.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

It is hard to look at the figures and the tableand think that dense living will increasesocial capital or that the rise of sprawl hashurt civic engagement. On the contrary, amore natural interpretation of the results isthat whatever caused the decline in civicengagement (the television, perhaps) hasbeen partially offset by sprawl. YetPutnam’s basic logic may seem strong: livingin the suburbs means long commutes andthese commutes should crowd out otheractivities like civic connection.

The problem with Putnam’s story is thatsprawl is not responsible for longer commutesand that, in fact, commutes are actuallyshorter in low-density metropolitan areas.Figure 1 shows the relationship betweencommute times and density level acrossmetropolitan areas. A 1 log point increase indensity is associated with a 2.4 minuteincrease in average commute times. Lowerdensity, sprawling areas are not associatedwith longer commutes; rather, they are associ-ated with shorter commutes. Thus, it shouldnot surprise us that these places have moresocial capital. The longer commutes in denseplaces discussed in section 2 should detersocial engagement.

While high-density living is certainly a plusfor many forms of social connection, civicassociations do not seem to thrive in higherdensity areas. Within metropolitan areas,there is no connection between central-cityresidence and most forms of social capital.Across metropolitan areas, density is associ-ated with less, not more, social capital—perhaps in part because density is associatedwith longer commutes. Sprawl may havenegative consequences along other

Table 6. Social capital and density across metropolitan areas

Volunteering(n ¼ 270)

Workingon a

communityproject

(n ¼ 271)

Gone to aclub meeting

(n ¼ 271)

Attendanceat church(n ¼ 271)

Believeothers are

honest(n ¼ 271)

Share of adults with 0.0095 0.0093 0.0006 0.0074 0.0049college degrees [0.0038] [0.0021] [0.0036] [0.0077] [0.0029]

Log of people per 20.0869 20.0316 20.0074 0.0940 20.0531square mile [0.0174] [0.0094] [0.0165] [0.0351] [0.0129]

Median age 0.0020 0.0046 0.0121 0.0226 0.0154[0.0053] [0.0029] [20.0050] [0.0106] [0.0039]

Mean January temperature 20.0024 20.0003 20.0023 20.0170 20.0032[0.0008] [0.0004] [0.0008] [0.0016] [0.0006]

Per capita income, 1989 0.0122 20.0062 0.0026 20.1208 0.0056[0.0096] [0.0052] [0.0091] [0.0193] [0.0071]

Constant 2.4791 1.5001 2.0602 4.9571 3.5132[0.1834] [0.0993] [0.1739] [0.3694] [0.1361]

R 2 0.1539 0.1362 0.0649 0.4786 0.1833

Notes: Data are from the DDB Needham Life Style Survey, 1975–98 and from the 1990 US Census. Observations are

weighted by the inverse of the square root of the number of DDB Needham respondents within each metropolitan area. Stan-

dard errors are in brackets.

Figure 13. Volunteering and urban density acrossmetropolitan areas.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

dimensions, but it cannot be credited withkilling social capital.

As a final note, it is worth remembering thatdense cities are also places where racial segre-gation is higher. Figure 14 shows the relation-ship between metropolitan area density andracial dissimilarity in 2000. While it is some-times thought that sprawl increases racial seg-regation, there is little evidence that this istrue. Indeed, Figure 14 suggests that suburba-nisation and integration are more likely tomove together.

5. Conclusion

The success and failure of big cities dependsin large part on the urban edge in consump-tion, not production. Urban decline in thepost-war period was caused in large part bychanges in technology that made big citiesless effective at catering to consumers’ prefer-ences. The automobile meant that commuteswould be faster in car-oriented cities and sub-urban employment cities. Improvements intransport technology meant that people nolonger needed to live near waterways, butcould instead live in sunny, dry climates. Asa result, people fled from traditional citiesbuilt around older forms of transport andmoved to warm centres of sprawl. Thiseffect was somewhat strengthened by risingcrime between 1960 and 1980, which canexplain falling housing prices (but notfalling population) in big urban areas.

Since 1980, cities have rebounded, but thisrebound shows up mainly in the form ofhigher housing prices. In 1970, real wagesrose with urban populations. In 2000, realwages weakly declined with city size. Thenatural interpretation of this is that thesecities have become much more attractiveplaces in which to live. Some part of thisincreased attractiveness may be the result ofdeclining crime rates. Another portion of thisincrease seems due to rising incomes and edu-cation levels which increase demand for urbanamenities like museums, restaurants andconcerts.

But these urban advantages do not meanthat big cities today are centres of socialcapital or political engagement. People inthe suburbs are as likely to volunteer or beregistered voters as people in central cities.Across metropolitan areas, density is associ-ated with less—not more—social capital.Hence, it makes little sense to think that fight-ing sprawl is a good means of encouragingcivic engagement.

Notes

1. Transport technologies are not the only forceshaping urban form; land use regulation andproduction technologies are also important.Still, the available evidence strongly supportsthe primacy of the automobile in the dispersalof population and employment (Glaeser andKahn, 2004).

2. This hypothesis requires the income elasticityof demand for local amenities to be positive,which is shown by DiPasquale and Kahn(1999).

3. The correlation between city population andcar ownership per capita in the 2000 UScensus was 20.33.

4. While employment, or jobs, in a city mightseem like an attractive alternative measure,the correlation between this measure andpopulation is 98 per cent in the US in 2000and the correlation of change in employmentand change of population in the 1990s is 90per cent. These correlations are sufficientlyhigh that it is hard to examine employmentseparately from population.

5. If industries and occupations are seen asreflecting workplace amenities rather thanfirm productivity, then controlling for themmakes more sense. Still, the basic patterns

Figure 14. Density and segregation acrossmetropolitan areas.


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

of wages and urban size do not change whenwe add these controls (Glaeser and Mare,2001).

References

BERRY-CULLEN, J. and LEVITT, S. (1999) Crime,urban blight and the consequences for cities,Review of Economics and Statistics, 81(2), pp.159–169.

CLARK, T. N. (2004) The City as an EntertainmentMachine. Oxford: Elsevier.

CUTLER, D. and MILLER, G. (2006) Water,water everywhere: water supply and municipalgovernance in America’s cities, in: E.GLAESER and C. GOLDIN (Eds) Corruption andReforms: Lessons from America’s History, ch.5. Chicago, IL: University of Chicago Press.

CUTLER, D., GLAESER, E. and VIGDOR, J. (2005) Isthe melting pot still hot? Explaining the resur-gence of immigrant segregation. WorkingPaper No. 2071, Harvard Institute of EconomicResearch.

DIPASQUALE, D. and KAHN, M. (1999) Measuringneighborhood investments: an examination ofcommunity choice, Real Estate Economics,27(3), pp. 389–424.

FUJITA, M., KRUGMAN, P. and VENABLES, A. J.(1999) The Spatial Economy: Cities, Regions,and International Trade. Cambridge, MA: MITPress.

GLAESER, E. (2004) Public ownership in the Amer-ican city, in: A. E. SCHWARTZ (Ed.) Essays inHonor of Dick Netzer, pp. 130–162. Northamp-ton, MA: Edward Elgar.

GLAESER, E. and SHAPIRO, J. M. (2003) Urbangrowth in the 1990s: is city living back?,Journal of Regional Science, 43(1), pp. 139–165.

GLAESER, E. and GYOURKO, J. (2005) Urbandecline and durable housing, Journal of PoliticalEconomy, 113(2), pp. 345–375.

GLAESER, E. and KAHN, M. (2001) Decentralizedemployment and the transformation of theAmerican city, Brookings-Wharton Papers onUrban Affairs, 2, pp. 1–47.

GLAESER, E. and KAHN, M. (2004) Sprawl andurban growth, in: J. V. HENDERSON and J. F.THISSE (Eds) Handbook of Urban and RegionalEconomics Vol. 4, ch. 56. Oxford: OxfordUniversity Press.

GLAESER, E. and KOHLHASE, J. (2004) Cities,regions and the decline of transport costs,Papers in Regional Science, 83(1), pp. 197–228.

GLAESER, E. and MARE, D. (2001) Cities andskills, Journal of Labor Economics, 19(2), pp.316–342.

GLAESER, E. and SACERDOTE, B. (1999) Why isthere more crime in cities?, Journal of PoliticalEconomy, 107, pp. S225–S258.

GLAESER, E., KAHN, M. and RAPPAPORT, J. (2000)Why do the poor live in cities? Working PaperNo. 7636, NBER.

GLAESER, E., KOLKO, J. and SAIZ, A. (2001) Con-sumer city, Journal of Economic Geography, 1,pp. 27–50.

GYOURKO, J. and TRACY, J. (1989) Local publicsector rent-seeking and its impact on local landvalues, Regional Science and Urban Economics,19(3), pp. 493–516.

MONKONNEN, E. (2001) Homicides in New YorkCity, 1797–1999. Study No. 3226, Inter-Univer-sity Consortium for Political and SocialResearch, Ann Arbor, MI.

PUTNAM, R. (1996) The strange disappearance ofcivic America, The American Prospect, 7(24),pp. 204–215.

PUTNAM, R. (2000) Bowling Alone: The Collapseand Revival of American Community.New York: Simon and Schuster.

ROBACK, J. (1982) Wages, rents, and the quality oflife, Journal of Political Economy, 90(6), pp.1257–1278.

ROSEN, S. (1979) Wage-based indices of urbanquality of life, in: P. MIESZKOWSKI andM. STRASZHEIN (Eds) Current Issues in UrbanEconomics, pp. 74–104. Baltimore: JohnsHopkins University Press.

SCHWARTZ, A., SUSIN, S. and VOICU, I. (2003) Hasfalling crime driven New York’s real estateboom?, Housing Policy Debate, 14(1), pp.101–135.

THALER, R. (1978) A note on the value of crimecontrol: evidence from the property market,Journal of Urban Economics, 5(1), pp. 137–145.

WALKER, R. (1977) The suburban solution: capit-alism and the construction of urban space in theUnited States. Unpublished dissertation, JohnsHopkins University.

Appendix

DDB Needham Life Style Data

The data on social capital come from the DDBNeedham Life Style data, compiled by DDBWorldwide (Chicago) from surveys administeredfrom 1975 to 1998. These are the data used byPutnam (2000) and come from his website(http://www.bowlingalone.com/data.php3).They are extensively documented in Putnam(2000, Appendix I).

The DDB Needham survey presents respon-dents’ answers to a variety of questions about thefrequency of their participation in social activities.Most queries give respondents categorical optionsto quantify the frequency of their participation inthe activity in question over the past year. Inorder to run regressions with the frequency of anactivity as the dependent variable, we recode


Dow

nloa

ded

By:

[Uni

vers

ity o

f Rea

ding

] At:

16:2

2 7

Oct

ober

200

7

these categorical responses into individualnumbers as follows:

a. None (0)

b. 1–4 times (2.5)

c. 5–8 times (6.5)

d. 9–11 times (10)

e. 12–24 times (18)

f. 25–51 times (38)

g. 52þ times (52)

Income data are similarly coded at the mid-point ofthe reported range with answers ‘under $x’ codedslightly below $x and answers ‘$yþ ’ codedslightly above $y. Missing income data areassigned the sample mean so that they can beincluded in the regressions and a dummy variableequal to 1 for respondents who do not report theirincome is included in order to account for anydifference in the dependent variable associatedwith differences between these and otherrespondents.

US Census Data

Cross-city and cross-metropolitan area data arefrom the decennial census data obtained and pub-lished by the US Bureau of the Census. Whenresults are reported by metropolitan area, 1993metropolitan area definitions are used.

Crime Data

New York City homicide data are from the Inter-University Consortium for Political and SocialResearch study: Homicides in New York City,1797–1999 (Monkonnen, 2001).

Burglary and robbery victimisation data inNew York and Chicago come from the Bureau ofJustice Statistics’ Crime and victimisation in thethree largest metropolitan areas, 1980–98, com-piled from the National Crime VictimizationSurvey and available at http://www.ojp.usdoj.-gov/bjs/abstract/cv3lma98.htm.


Documents

Urban Resurgence and the Consumer City - Glaeser