DO SUBURBS HARM CITIES? A REVIEW OF THE EVIDENCE
Brett W. Hawkins, Department of Political Science, University of Wisconsin-Milwaukee
and Roberta Derlin, Department of Educational Management and Development, New Mexico State University
Many people claim, as if it were established fact, that suburbs exploit cities and cause their decline. But that viewpoint has never been free from scientific challenge nor convincingly backed by evidence. Ever since the 1950's, when social scientists first began to write about the issue, they cautioned readers not to rule out the possibility that suburbs generated more benefits than costs for cities. They stressed that suburbanites created property values wherever they worked, shopped, and played, and that without systematic benefit-cost analysis, higher city expenditures alone did not demonstrate exploitation by suburbs (Margolis, 1957; Fitch, 1957; Brazer, 1959; Davies, 1965; Davies, 1969). Even so, since the 1960's, city politicians have used the assertion that suburbs hurt cities as a political tool to seek redistribution of suburban resources. But political usefulness is not scientific validity. The issue should be decided not by opinion or expediency, but by the weight of the evidence, which forms the subject of our essay: what does the most relevant research tell us about the effects of suburbs on cities?
Two points should be emphasized at the beginning. First, the fact that spending by cities grows in part to serve suburban commuters does not prove that suburbs exploit cities, unless the costs outweigh the benefits. Writers sometimes recite the costs (which are real) without acknowledging evidence that suburbs benefit cities. Some ways that suburbanites benefit cities are by (a) contributing to city sales and income taxes where these exist, property taxes where suburbanites own city property, and user fees; (b) investing money and skills in city businesses; (c) buying city goods and services; (d) paying taxes to overlapping county and special district governments that provide a disproportionate share of their services to city residents; and (e) providing suburban services used by city residents.
Second, the reader should be aware of a particular version of the claim that suburbs harm cities, which declares that cities and suburbs are "interdependent" but only in one direction: the suburbs need a prosperous city. Invariably, these unidirectional declarations about interdependence are accompanied by demands that the federal government and states make suburbs pay for the burdens they allegedly impose on cities. However, if suburban-generated benefits exceed or equal costs, the burdens are already paid for. In that case, interdependence is mutual: cities benefit from healthy suburbs just as suburbs benefit from healthy cities.
Publications estimating both benefits and burdens are quite rare, but any study that does not consider both is inconclusive. We have found fourteen such studies, including two co-written by one of the authors. Some of these studies focus on commuters; some relate to the size and growth of suburbs; and some compare city and suburban taxes paid to counties and special districts with the services provided to cities and suburbs by those overlapping jurisdictions. All such studies depend on assumptions to estimate benefits and costs.
We discuss the benefit-cost literature chronologically and under two headings: the city impact of all suburbanites (i.e., suburban growth), and the city impact of suburban commuters (a classification suggested by Greene, Neenan, and Scott 1974). Cost-benefit studies are numbered one to fourteen. (Lettered studies do not estimate both burdens and benefits, and accordingly are not included in the summary section. We discuss three such studies because they raise issues important to the analysis of costs and benefits.)
Taken as a whole, studies that compare benefits and burdens--and these are the ones we deem most relevant--show either that suburbs benefit cities or that they do not harm them.
(1) Using data from the 1950's on the thirty-six largest metropolitan areas, Margolis found that indicators of city conditions reflecting a beneficial suburban influence (total retail sales, for example) were of a higher magnitude than indicators of a suburban burden. For that reason, he characterized the argument that suburbs exploited central cities as "not well established" (1961: pp. 257-259).
(2) In a pioneering study of whether cities subsidized suburbs through the actions of overlapping governments, Banovetz estimated city and suburban benefits from county expenditures relative to county taxes paid by city residents and suburbanites. His data were from the Twin Cities. Regarding welfare, Banovetz discovered evidence of suburban subsidization of cities through the counties. Regarding a combination of several less costly county services, he found city subsidization of suburbs. The amounts of subsidization were small in both cases, and he concluded that neither city nor suburb was subsidizing the other to any appreciable extent. He called charges of subsidy "ill-founded" (1965).
(3) Kee, using data from 1953 to 1962 on the nation's twenty-two largest metropolitan areas, found that suburban growth did not correlate with increases in overall or non-education city spending (1967). In another study using some of the same data, Kee correlated nonresident work force with city property values, incomes, value added by manufacturing, and retail sales per capita (1962 data). He found a significant positive correlation between nonresident work force and city retail sales (.676) but "virtually no statistical relationship" with city income, property value, or value added (1968).
Given the fact that Kee's 1967 and 1968 studies use similar data from the same time period--examining possible burdens in the 1967 study, and possible benefits in the 1968 study--we treat them as one study of the effect of suburban growth, concluding that together they indicate no suburban harm to cities. The 1967 study shows no harm, and the 1968 study shows one benefit out of four possibilities. This we judge conservatively to indicate no benefit.
(A) Campbell and Sacks discovered a significant reservation about the exploitation thesis. Like earlier researchers (Brazer 1959), they found that the greater the share of the total metropolitan population living outside the city, the greater the city's spending per person. But they also found that the greater the suburban share of population (for thirty-six metropolitan areas), the greater the suburban expenditure per person. They interpreted this finding as revealing a city-suburban relationship "much more complex" than the exploitation thesis suggested (1967). However, Campbell and Sacks did not compare suburban-generated costs with benefits.
(4) In 1972 Kasarda published an extraordinary study of the impact of suburban growth on cities. His data on 168 metropolitan areas provided a basis for comparing suburban-generated costs and benefits. He did not make the comparison, but we do.
His data showed that suburban size (1960 and 1970) and growth (from 1950 to 1960 and from 1960 to 1970) correlated with the size and growth of city government expenditures (which he interpreted as signifying a suburban-generated cost). Suburban size and growth also correlated with the size and growth of sales and receipts in the city for retail business, wholesale business, and services (which he did not interpret as signifying a suburban-generated benefit). Government expenditures are of course paid for by taxpayers; but business receipts are not. Significantly, in Kasarda's data, the average correlation of suburbanization with city business health was greater than its average correlation with city government spending; all correlations were positive. But Kasarda acknowledged only the evidence of cost, ignoring the spur to city business while observing that city residents are partly reimbursed for suburban-generated costs "through employment and sales taxes" (p. 1123). He called for the consolidation of city and suburban governments into a single area-wide government in order to facilitate the redistribution of suburban incomes to the city, a recommendation that seems to contradict his own evidence of the benefits of suburbanization.
(5) Hawkins and Hendrick compared the share of property taxes that the suburbs and the city contributed to Milwaukee County with the share of services that each area received from eight 1991 county programs (corrections, social services, museum, zoo, parks, handicapped transportation, mental health, and capital spending). Most of the county's property tax revenue came from the suburbs (with about 35 percent of the county's population); but, for all eight programs, most of the service benefits went to city residents. The authors concluded that the role of the county was to redistribute income from the more affluent suburbs to the city (1994).
(6) Hawkins and Hendrick have also done a study of taxes paid to and benefits received from metropolitan special districts (Milwaukee area sewerage and post-secondary technical education districts). For both districts, suburban taxpayers contributed most of the property taxes, and city residents received most of the 1993 service benefits (1997).
We should note that Greene, Neenan, and Scott (1974) criticized the Twin Cities study in a way that also applies to the Milwaukee studies. All three studies assumed that government expenditures generate benefits for direct program recipients only; the studies did not attempt to estimate indirect benefits attributable to external effects or to class differences in willingness to finance the benefits. However, Greene, Neenan, and Scott themselves noted that the direct benefits approach is the more traditional one and may be employed with weaker assumptions. It is for that reason that many other publications concerning the distribution of services by local governments have looked at direct user benefits only. They avoid the estimation of indirect and willingness-to-pay benefits because of the extraordinary difficulty of doing so without making "strong assumptions" that predetermine results (see for example Boyle and Jacobs, 1982). In addition, the Twin Cities and Milwaukee studies included services that one might assume generate spillover benefits for city residents, and for which they are thus willing to pay the smaller share they do pay (for the greater direct benefits they also receive)--suburban parks, suburban technical college campuses, zoos, and capital expenditures, for example.
(7) and (8) Greene, Neenan, and Scott's 1974 study in the Washington, D.C. area is surely the most comprehensive study ever done of the costs and benefits of public services provided by central city and suburban jurisdictions. It employed two different approaches. One estimated direct costs and direct benefits to users, comparable to the approach used in the Twin Cities and Milwaukee studies. This method is the more traditional one and can be employed using weaker assumptions that are (Greene et al. say "appear to be" [p. 25]) more objective. The findings arrived at by this method showed the District enjoying strong net benefits from both the Maryland and Virginia suburbs: "The District ?exploits' the suburbs" (p. 146). (In the summary, we refer to this study as "Greene, Neenan, and Scott method #1.")
Greene and his associates, however, preferred another method on grounds that suburbanites gained indirectly from the redistributive expenditures of city governments and, according to opinion surveys done at the time, were more willing (as higher-income persons) to pay for poverty programs than were lower-income residents of cities. When estimates of this willingness to pay are used, and treated as measures of the true cost of service, the taxes from suburbanites flowing to the District are "swamped by the much larger net flow of benefits from the District into the suburbs." The conclusion of their analysis is that "there exists considerable ?fiscal exploitation' of the city by the suburbs in the Washington metropolitan area" (p. 148). (In the summary we refer to this study as "Greene, Neenan, and Scott method #2.")
While Greene and his associates consider the direct benefits approach to be of limited validity, we consider the "welfare basis" approach to be limited. In addition to the points made previously, the welfare basis approach assumes that significant indirect benefits flow only in one direction, thus guaranteeing a pro-city result. Also, we believe that the approach is time-bound and does not apply to the 1990's as it did to the 1970's. It is true that in the 1960's opinion surveys often showed higher income people supporting redistributive public policies; now, however, the persistent criticism of such programs, backed by research in some cases, suggests that better educated, higher income people may be indisposed to support them. Their support is all the more unlikely if it is true, as an award-winning book contends, that redistributive policies (which are by definition financed by the affluent while serving the non-affluent) give higher-than- average taxpayers an incentive to leave cities that emphasize such policies, and therefore run counter to the interest that all cities have in their economic well-being (Peterson, 1981).
(B) Still another perspective on the city influence of all suburbanites is provided by Nathan and Adams. In 1976 and 1989, they examined the effects of city-suburban differences, which they termed "disparities." Both studies identified twenty-two central cities (of fifty-five studied) that were undergoing relatively high social and economic hardship in 1970, measured by six social and economic variables. They then divided the twenty-two cities into nine that manifested higher city-suburban socioeconomic disparity and thirteen that manifested lower disparity. (They measured disparity with a ratio of the six social and economic characteristics of cities to the same six characteristics applied to corresponding suburbs.) From 1970 to 1980, socioeconomic conditions in the nine "high hardship/high disparity" cities improved by 4 percent, while conditions in the thirteen "high hardship/low disparity" cities improved by 15 percent.
Nathan and Adams interpreted these results as showing that city-suburban disparity harms cities. But we believe that an alternative interpretation is equally consistent with their data. According to this data, high hardship cities encountering both higher and lower disparity improved during the 1970's. Even extreme city-suburban disparities were associated with improvement in cities. Therefore disparity--a concept that reflects suburban economic health-- either did not hurt, or actually helped cities. It is true that more initial disparity was associated with less subsequent improvement, and so disparity may have limited, but did not arrest, city improvement. However, such a claim would need to be supported by evidence that the nine cities showing less improvement did not have higher hardship initially.
The key question is whether the nine cities showed less improvement because they were worse off than their suburbs, or because they were worse off than the other high hardship cities. If it is the latter, we are left with the unedifying conclusion that socioeconomic improvement in cities experiencing disparity with their suburbs is principally determined by worse city conditions in the first place. In studying the city effect of city-suburban disparity, it is necessary to examine the effect of the disparity net of the city's initial socioeconomic conditions, since either could devitalize the city in a subsequent time period. Nathan and Adams did not do this. Furthermore, even as they argue that only burdens are associated with city-suburban disparity, their analysis shows the opposite: i.e., that only benefits are associated with it. (But because their studies do not expressly estimate benefits and burdens, they will not be included in our summary section.)
(C) Using data on the twenty-two largest metropolitan areas, Kee examined changes from 1953 to 1962, concluding that, while there was some evidence for the suburban exploitation thesis, it was not clearly supported. The data demonstrated that growth of the nonresident work force in cities (commuters) correlated positively with total city spending per capita for five services (police, fire, sanitation, recreation, and general control). Important as they are, these five services are only a fraction of all services provided in cities. Curiously, Kee did not compute the effect of commuters on total city expenditures, education expenditures, or general non-education expenditures, even though these were part of his data set. More significantly, Kee's study did not single out those expenditures in cities that were made by city governments (and supported by city taxpayers); instead, he combined city government expenditures with those of the overlying units of local government. But of course the programs of counties and special districts are supported by suburban taxpayers as well as city ones. Therefore, his measure of "city" spending on services commingles suburban-generated costs with suburban-generated benefits. Finally, Kee made no estimate of commuter-generated benefits for cities. Accordingly, the 1967 Kee study (as it applies to commuters) is not included in the summary.
(9) Neenan investigated commuter-generated benefits and costs and found evidence that in 1966 Detroit was exploited by all six of its suburbs that he studied (1970). The net subsidy from Detroit ranged from $1.73 per capita to $12.58 per capita, with a median of $6.78. However, because the average level of local spending per capita in 1966 was about $200.00, Bradford and Oates have characterized Neenan's evidence as having "minor quantitative significance" (1974: p. 48).
(10) One follow-up study of Detroit and the same six suburbs criticized Neenan's "very strong" assumption that per capita benefits from government programs increased with increasing income. The authors proposed three alternative assumptions which they thought as reasonable as Neenan's, and showed that Neenan's conclusions did not hold for two of them. (One alternative assumption was that benefits were independent of, rather than proportional to, income.) The authors called the Neenan study too sensitive to its assumptions to generate any more than the most cautious conclusions (Auld and Cook, 1972).
(11) Another follow-up study to Neenan's Detroit study carried out an alternative test of suburban exploitation. Exploitation of city residents exists, said Ramsey, if the following sum is negative: city revenue collections generated by suburban residents less the cost of city services used by suburban residents plus the cost of services provided by the suburbs and used by city residents less the suburbs' revenue collections generated by central city residents. The results of using this approach "drastically" reduced the per capita subsidy that Neenan found the six suburbs receive from Detroit (to about six cents per Detroit resident). This analysis also showed that three of the six suburbs were exploited by Detroit (Ramsey, 1972).
(12) In a 1970 Ph.D. dissertation at Columbia University, Book estimated the net fiscal impact of commuters on New York City. He concluded that commuters should be paying millions of dollars more for the benefits they receive, but the only benefit he considered was the city's nonresident earnings tax. He made no estimates of suburban contributions to sales or property taxes, or to property values and income (cited in Greene, Neenan, and Scott, 1974: pp. 18-19).
(13) With 1960 data, Vincent estimated commuter-generated benefits (compared with costs) in sixty-five metropolitan areas, and concluded that cities were benefited by commuting suburban workers and shoppers (1971). His study has been criticized, however, because it does not consider the likelihood that the productivity of suburbanites is positively related to the proximity of a city (Greece, Neenan, and Scott, 1974: p. 20). If this criticism is valid, we can infer that today, the productivity of city residents is higher than it would otherwise be because of their proximity to the suburbs, where most contemporary business takes place. Therefore, this suburban-generated benefit would have to be included in contemporary studies of commuter effects.
(14) In 1972 Smith studied the exploitation question in the San Francisco area. He concluded that commuters contribute more (in fact millions more) than it costs the city to service them (1972). But he too was criticized--in his case for assuming that city expenditures are divided in equal proportions among industrial, commercial, and residential land uses, and for assuming that the value of a local government service to its user is equal to its cost (Greene, Neenan, and Scott, 1974: pp. 20-21).(1)
For purposes of this summary--with two exceptions--we simply adopt the conclusions of the authors of the benefit-cost studies reviewed above (thus bypassing the question of whose assumptions and measures are superior) when classifying them as showing a net suburban benefit for cities, a net suburban burden, or approximately equal benefits and burdens (i.e., no harm). The exceptions we make are the Kasarda study, which in our opinion demonstrates the opposite of what the author concludes, and the 1967 and 1968 Kee studies using very similar data. As explained above, we have chosen to combine the two Kee studies so that they add up to a benefit- cost study.
Of the fourteen benefit-cost studies reviewed, seven report a net suburban benefit for cities; four report benefits approximately equal to costs; and three report a net suburban cost.(2)
If construed conservatively, this pattern supports the conclusion that suburban effects on cities through growth and commuters have generated benefits that at least reimburse the cities for their costs. Construed liberally, the pattern suggests that, far from damaging cities, suburbs improve them.
What are the broader implications of this pattern of evidence? One implication is that suburbanization is not the cause of city decline.(3) If suburbanization is not responsible, what is? Research does not yet yield a complete answer; but one alternative cause is consistent with the evidence (although counter to the orthodoxy of suburban causation): economic and technological changes have caused both the growth of suburbs and the decline of cities, in much the same way that earlier changes caused both city growth and rural decline. Suburban growth correlates with city decline not because it causes it, but because both are caused by a third factor. The first correlation is therefore spurious.
Another implication has to do with the way we view commuters. After decades of complaint about suburban commuters, the American metropolis is experiencing more and more commutation in the other direction. Since suburbs are the principal location of business activity, that is hardly surprising. But if the suburban exploitation thesis was ever valid, these new commuters are by its logic exploiting the suburbs. It is only a matter of time, however, before writers with a pro-city agenda suggest that this commutation, unlike the early one, is a net benefit for those on the receiving end.
In addition, if city-suburban interdependence is a valid concept, and city prosperity benefits suburbs, the pattern of evidence reviewed here has to be read as implying that suburban prosperity also benefits cities. In other words, the evidence implies a concept of interdependence that faces both ways, is not tied to a political agenda, and recognizes that cities are better off when their suburbs are well of, just as suburbs are better off when their cities are well off.
As for policy implications, the evidence implies that suburbs are not a problem for cities but part of the solution. Therefore, federal, state, and local policies should not be based on a claim that city problems are caused by suburbanites who contribute nothing to the well-being of cities. Besides being inaccurate, this assertion invariably leads to demands that resources be taken away from suburbs and given to cities. Such policies are counterproductive because suburban distress hurts cities, and many suburbs are already distressed.
(1)1. Kasarda's 1968 study of 168 metropolitan areas also investigated the effect of commuters on cities, but it did not consider benefits.
(2)2. The seven reporting a net benefit are Margolis; Kasarda (which we reinterpret); Hawkins and Hendrick 1991; Hawkins and Hendrick 1997; Greene, Neenan, and Scott method #1; Vincent; and Smith.
The four reporting roughly equal benefits and costs are Banovetz; Kee (combining his 1967 and 1968 studies of growth); Auld and Cook; and Ramsey.
The three reporting a net cost are Neenan; Book; and Greene, Neenan, and Scott method #2.
3. Suburbs have been growing faster than cities since the 1920's. Cities have been the locus of a smaller share of population and business since the 1960's. They have been declining in absolute numbers since the 1970's, while individual cities have been declining for much longer. Cities would have declined still more, both relatively and absolutely, if they had not annexed large amounts of suburban territory during the 1960's. Ninety-eight percent of the growth of the nation's central cities during the 1960's was due to annexation (Marando, 1976).
The contemporary metropolis is no longer dominated economically by a single center. It is economically, as well as politically, polycentric.
(3)4. A 1992 study prepared for the National League of Cities (NLC), a lobbying organization with individual cities and municipal leagues as members, concludes that city-suburban disparity hurts cities in a subsequent time period. It also hurts suburbs and whole metropolitan areas. This, the study suggests, demonstrates interdependence (Ledebur and Barnes, 1992).
For a critique of the NLC study which argues for a reversal of that study's causal direction-- for example, city job decline causes disparity rather than the other way around--see Hill, Wolman, and Ford 1995. When we first read the NLC study, that was our reaction as well.
Auld, D.A.L. and Gail C.A. Cook. 1972. "Suburban-Central City Exploitation Thesis." National Tax Journal 25: 595-597.
Banovetz, James M. 1965. Governmental Cost Burdens and Service Benefits in the Twin Cities Metropolitan Area. Minneapolis: University of Minnesota.
Book, Samuel H. 1970. "Costs of Commuters to the Central City as a Basis for Commuter Taxation. Ph.D. dissertation, Columbia University.
Boyle, John and David Jacobs. 1982. "The Intracity Distribution of Services: A Multivariate Analysis." American Political Science Review 76: 371-379.
Bradford, David F. and Wallace Oates. 1974. "Suburban Exploitation of Central Cities and Governmental Structure." Pp. 43-90 in Harold M. Hochman and George E. Peterson eds., Redistribution Through Public Choice. New York: Columbia University Press.
Brazer, Harvey E. 1959. City Expenditures in the United States. New York: National Bureau of Economic Research.
Campbell, Alan K. and Seymour Sacks. 1967. Metropolitan America: Fiscal Patterns and Governmental Patterns. New York: Free Press.
Davies, David. 1965. "Financing Urban Functions and Services." Pp. 119-153 in Robinson O. Everett and Richard H. Leach, Urban Problems and Prospects. Dobbs Ferry, New York: Oceana Publications.
Davies, David. 1969. "Fiscal Effort, A Comment." National Tax Journal 22: 423.
Fitch, Lyle C. 1957. "Metropolitan Financial Problems." The Annals of the American Academy 314: 66-73.
Greene, Kenneth V., William B. Neenan, and Claudia D. Scott. 1974. Fiscal Interactions in a Metropolitan Area. Lexington, Massachusetts: Lexington Books.
Hawkins, Brett W. and Rebecca M. Hendrick. 1994. "Do County Governments Reinforce City- Suburban Inequalities? A Study of City and Suburban Service Allocations." Social Science Quarterly 75: 755-771.
Hawkins, Brett W. and Rebecca M. Hendrick. 1997. "Do Metropolitan Special Districts Reinforce Sociospatial Inequalities? A Study of Sewerage and Technical Education in Milwaukee County." Forthcoming in Publius 127: No. 1.
Hill, Edward W., Harold L. Wolman, and Coit Cook Ford III. 1995. "Can Suburbs Survive Without Their Central Cities? Examining the Suburban Dependence Hypothesis." Urban Affairs Review 31: 147-174.
Kasarda, John D. 1972. "The Impact of Suburban Population Growth on Central City Service Functions." American Journal of Sociology 77: 1111-1124.
Kee, Woo Sik. 1967. "Suburban Population Growth and Its Implications for Core City Finance." Land Economics 63: 202-211.
Kee, Woo Sik. 1968. "City-Suburban Differentials in Local Government Fiscal Effort." National Tax Journal 21: 183-189.
Ledebur, Larry C. and William R. Barnes. 1992. Metropolitan Disparities and Economic Growth. Washington: National League of Cities.
Marando, Vincent. 1976. "The Politics of Metropolitan Reform." Pp. 24-49 in Alan K. Campbell and Roy W. Bahl, State and Local Government: The Political Economy of Reform. Free Press: New York.
Margolis, Julius. 1957. "Municipal Fiscal Structure in a Metropolitan Region." Journal of Political Economy 65: 225-236.
Margolis, Julius. 1961. "Metropolitan Finance Problems: Territories, Functions, and Growth." Pp. 229-270 in National Bureau of Economic Research, Public Finances, Needs, Sources, and Utilization. Princeton, New Jersey: Princeton University Press.
Nathan, Richard P. and Charles F. Adams. 1976. "Understanding Central City Hardship." Political Science Quarterly 91: 47-61.
Nathan, Richard P. and Charles F. Adams. 1989. "Four Perspectives on Urban Hardship." Political Science Quarterly 104: 483-508.
Neenan, William. 1970. "The Suburban-Central City Exploitation Thesis: One City's Tale." National Tax Journal 23: 117-139.
Peterson, Paul E. 1981. City Limits. Chicago: University of Chicago.
Ramsey, David D. 1972. "Suburban-Central City Exploitation Thesis: Comment." National Tax Journal 25: 599-604.
Smith, R.F. 1972. "Are Nonresidents Contributing Their Share to Core City Revenues?" Land Economics XLVIII: 240-247.
Vincent, Philip E. 1971. "The Fiscal Impact of Commuters." Pp. 41-143 in Werner Z. Hirsch et al., Fiscal Pressures on the Central City: The Impact of Commuters, Nonwhites, and Overlapping Governments. New York: Praeger.