d - ‘Q.. ..--~-‘-. I. MUNICIPAL EXPENDITURES AND URBAN STRUCTURE: I ' - PLANNING IMPLICATIONS FOR ' MICHIGAN’S CENTRAL CITIES Thesis for the'Degree of M. U. P. MICHIGAN STATE UNIVERSITY . PAUL R. STUHMER 1977 -.~. (.- P ~- ; 0'17 'v .-. . , .- . A l.~~..p..,.r~,la .4 gin“ 0'... map- «w ~‘I"I‘~“M w - 4 .0,-.- ..... l’t' ‘.. .... .. vawva GUI. 5" ~ ‘. .vofln-l-Orro— Dfitvfll'fi "1 ..., ’. I. - a ’- ‘.' r,—. . .- - ‘ . nu. ‘5 P--..wdl' ' —-.r n \r - (r- ~-.-v - x v H"- - .o ‘0... ‘ u . ' . 4—9-1- m“; w .. . . -Qkfl‘fv ”In 1 5' gym IIIIIIIIIIIIIIIIIIIIIIIIIIIII 3 1293 10649 1339 w‘W-V . ) . 'l Vfia‘ ’ R '1 ‘- ’ . 4" I I' L Q 3‘ v T” . 1- \ ‘5‘”:me 9“ "3.. APR 1 6“87‘"‘-’" ‘3 '7 K2. 1 5 fin” .' . A a“ ' ‘1‘}“EE? fl - - \ L) ,. 2 \J (j H.146 L/O ABSTRACT MUNICIPAL EXPENDITURES AND URBAN STRUCTURE: PLANNING IMPLICATIONS FOR MICHIGAN'S CENTRAL CITIES BY Paul R. Stuhmer Problems of urban decline and financial crisis that were once thought to occur primarily in large urban centers have now spread to many of Michigan's medium-sized cities. A necessary step in finding solutions to these problems is integrating studies of municipal expenditures into the urban planning process. This will introduce new areas of concern to planners and increase the feasibility and relevance of many planning activities. One facet of relating municipal finance to planning is establishing Michigan's central cities as a "client group" that would benefit from advocacy planning. This study seeks to show that: 1. Michigan's central cities have sufficient distin- guishing characteristics to allow them to be considered as a class. Paul R. Stuhmer 2. The characteristics which distinguish the central cities have a significant effect on the structure of municipal costs. 3. Differences in municipal costs can be explained by a limited number of structural and socio-economic variables which cannot be changed over the short run. If these hypotheses are correct, then the central cities will require programs which would benefit central cities as differentiated from other units of government, and which would provide compensation over the short term while longer term correctives are designed. The methodology used to investigate the factors explaining expenditure variables was to create two sets of multiple regression equations for a number of specific costs. Observations of cost and explanatory variable were taken from a sample of 202 local governments in southern Michigan. The sample was divided by category of local government and category differences for both structural and cost variables were calculated. The second set of regressions was used to identify explanations of cost variations within a subsample consisting of cities of over 10,000 population within the sample. The findings of the study indicate that the central cities do in fact have sufficient differentiating character- istics to consider them a distinct and separate group for finance and planning purposes. Central city means for Paul R. Stuhmer variables were compared to those for urban townships, suburban cities, suburban townships, and rural local govern- ments. Only two variables failed to show statistically significant differences. 1 Several factors were identified which have signifi- cant impacts on expenditure variations between local governments. The most important is the density of commercial and industrial value. Second in importance was the age of housing stock, which was felt to be the best available surrogate for age and condition of urban infrastructure. Percent poverty and employment characteristics had a marginal impact. Significant differences between categories were especially striking in total local tax rates and tax burden related to income. Recommendations include the establishment of a statewide business property tax to replace local property taxes on businesses, and the investigation of alternatives for "life-cycle" financing for cities. Research recommenda- tions include investigation of patterns of county expendi- tures, improvements in reporting of municipal expenditures and research into cost patterns in declining urban areas. MUNICIPAL EXPENDITURES AND URBAN STRUCTURE: PLANNING IMPLICATIONS FOR MICHIGAN'S CENTRAL CITIES BY \.& Paul Rf Stuhmer A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF URBAN PLANNING School of Urban Planning and Landscape Architecture 1977 ACKNOWLEDGMENTS Many individuals have contributed time and expertise to produce this thesis. Foremost among these I wish to thank Dr. Roger Hamin, my advisor, for setting the direction of the research effort and his advice, encouragement, and criticism along the way. Dr. C. Robert Muth sponsored the research and helped to define many of the issues of municipal and education finance. Mary Guile added both technical expertise and a great deal of effort in tracking down information and form- ulating the study. Special thanks to Jan Leland for her time and skill in editing and proofreading this manuscript. Thank you one and all. ii TABLE OF CONTENTS LIST OF TABLES. . . . . . . I. INTRODUCTION . . . . . Background . . . Study Design and Scope . Research Objectives . . Research Methodology . . Study Limitations . . II. RELATED LITERATURE AND RESEARCH Early Studies of Municipal Expenditures . . . Urban-Suburban Relations and Metropolitan Areas . . Other Recent Directions in Expenditure Studies. . III. METHODOLOGY. . . . . . Sample Description. . . Local Government Sample Multiple Regression Analysis of Expenditures . . . Group Means and Analysis of Variance . . . Description of Variables. Independent Variables . Cost Variables . . . Group Mean Variables . IV. ANALYSIS OF DATA . . . . Whole Sample Regression Analysis Cities Regression Analysis iii 20 28 38 45 46 51 55 59 61 62 67 70 70 87 Selected Expenditures By Category. . . Groups Means and Analysis of Variance . . . . . . . . . . V. CONCLUSIONS AND RECOMMENDATIONS . . . . Findings . . . . . . . . . . . Policy Implications and Recommendations . . . . . . . . Recommendations For Further Research . . . . . . . . . . APPENDICES O O O 0 O O O O O O O 0 Appendix A. Means, Standard Deviation and Number of Valid Cases For Regression Variables B. Correlation Coefficients for Full Sample Regression . . . . . . . C. Simple Correlations for Cities Only Variables . . . . . . . . D. List of Data Sources. . . . . . . E. Sample of Municipalities . . . . . REFERENCES 0 O O O O O O O O O 0 0 iv Page 96 101 116 117 121 128 132 132 133 134 135 136 140 Table 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. LIST OF TABLES Sample size and breakdown. . Sample classification . Revised sample classification . Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression Regression summary for full sample. . . for dependent for dependent for dependent for dependent for dependent for dependent for dependent for dependent for dependent for dependent for dependent variable variable variable variable variable variable variable variable variable variable variable summary for cities . for dependent for dependent for dependent for dependent variable variable variable variable TOTEXP. GENAD . GOVT . DEBT . BLDGS . HUD. . HEALTH. POLICE. FIRE . SEWER . ROADS . GENAD . POLICE. FIRE . SEWER . Page 51 53 54 70 73 75 77 78 79 80 81 82 83 85 86 88 90 91 93 94 Table Page 21. Regression for dependent variable ROADS. . . . 95 22. Breakdown for GOVT expenditures . . . . . . 98 23. Breakdown for GENAD expenditures . . . . . . 99 24. Breakdown of FIRE expenditures . . . . . . 100 25. Breakdown for POLICE expenditures. . . . . . 100 26. Category mean populations . . . . . . . . 102 27. Population density by category. . . . . . . 103 28. Residential value density by category . . . . 103 29. Industrial value density by category. . . . . 104 30. Industrial as a percent of total value . . . . 105 31. Percent housing built before 1940 by category. 1 106 32. Overcrowded housing by category . . . . . . 106 33. Percent employed as professional, technical and managerial by category . . . . . . . 108 34. Percent employed in manufacturing by category. . 108 35. Percent of families below the poverty level . . 109 36. Percent of workers with incomes below $5,000 per year . . . . . . . . . . . . . 110 37. Per capita income by category . . . . . . . 111 38. Full per capita residential value. . . . . . 111 39. Per capita assessed value by category . . . . 112 40. Local property tax millage . . . . . . . . 113 41. Local government charges by category. . . . . 114 vi I . INTRODUCTION This study is a test of several hypotheses concern- ing the structure of local government types and the effect of community structure on local government expenditures. Chapter one presents the background relating the study of expenditures to planning theory and practice, a description of the study design and scope and study limitations. Chapter two describes relevant studies and research. Chapters three and four present the study methodology and data, and Chapter five summarizes findings and relates these to recom- mendations for policy and further research. Chapter one provides the context for the expenditure study. The background relates the study of expenditures to fiscal planning and fiscal impacts of planning decisions. The sc0pe and design provide the specific context in which the study was undertaken. The next sections summarize the specific objectives, techniques and limitations of the study. Background Among the most prominent and widely publicized trends of the late 19608 and 19708 has been the decline of the nation's central cities. In problems ranging from financial crisis to physical decay and massive outmigration, the impact of this decline upon the urban planning profession has been profound. Planners as well as most city officials have been unprepared and unable to c0pe with the demographic and structural changes that have accompanied changing social conditions and preferences. Concern for these problems has typically been directed to the largest urban centers, where civil disorders and other graphic conditions have brought wide public attention. More recently, many of the same symptoms of decline have become evident in small and medium-sized cities that were once thought to be immune to such problems. Although this situation has been exacerbated in Michigan as a result of declines in auto and related industries, it is obvious that the decline of cities reflects a more basic pattern that is being replicated in manyareas across the country. As outmigration to the suburbs has spread from the largest metropolises to smaller cities and regional centers, it has taken the leading edge of society with it, leaving many of the poor and less fortunate to deal with the complex prob- lems of urban management. The consistent movement away from central cities seems to be qualitatively different from individual city declines due to changes in industrial location and production. Thus the problem may merit special attention, as well as having different solutions or alternatives. Attempts by the planning profession to deal with this crisis have not met with any substantial degree of success. Part of the problem seems to stem from a bais lack of knowledge of the processes by which economic growth and revitalization can be encouraged. A specific aspect of this lack is that little has been done in the State of Michigan to analyze the root causes of specific increases in munici- pal expenditures. This research is of direct importance to the ability of planners and city officials to estimate the feasibility, costs and requirements for changes in public services. The determination of the causes for expenditure trends may enable planners to increase their lead time in developing and implementing more effective programs to deal with problems of rising costs and deteriorating services: . . .Planning for public service requirements in smaller cities is an essential, but often neglected part of the process of encouraging economic growth. Many smaller cities will serve as growth centers for multi-county deveIOpment regions and these cities are likely to experience large increases in population. At the same time, other cities may continue their decline in p0pula- tion. It is, however, essential for both of these groups of cities to maintain adequate levels of govern- mental services. In the growth centers, inadequate levels of government may retard the region's growth by discouraging new industry or new residents from locating in the area. In declining cities, inadequate service levels may accelerate the population outflow, and produce unnecessary hardships for the peOple who remain (Stinson, 1970, p. 134). Finance has long remained on the periphery of urban planning, appearing perhaps most frequently in relation to capital improvements planning. But the pattern of municipal expenditures has a direct relationship to many facets of urban planning. The requirement for public services and the ability to support them will directly affect the ability to carry out such programs as neighborhood restoration and housing rehabilitation. In the past, many of these programs have failed because of the city's inability to provide the necessary level of support in related services. Planners for such activities must have an adequate knowledge of the likely changes in expenditures that result from demographic and structural changes. These will have both budgetary and policy implications in the city and its planning activities: Current public financial planning relies heavily on the extrapolation of existing trends. This technique, while adequate for most marginal increases in population, may prove misleading both for rapidly growing and declining cities. In these cities, economies and diseconomies of scale, fixed capacity levels and changes in both the quantity and quality of public services can no longer be ignored. In addition, some revenue sources, such as the property tax, may not increase at the same rate as expen- diture needs, forcing increased reliance on other sources of revenue. Consequently, projections of future service requirements and revenue estimates may need to be based on something more than the previous year's trend (Stinson, 1970, p. 135). Many planning activities now require the submission of detailed plans covering both program cost and eligibility for specific programs. This type of plan in many programs is a condition for state or federal aid. The creation of detailed information of this type requires a background in both physical and financial planning. Agencies without this capability will be at a disadvantage in applying for such programs. The analysis of expenditures may also shed some light on planning--related issues, particularly those related to planning for public services. It has long been held that increasing the scale of operations of many municipal services will lower per capita costs. However, some recent evidence indicates that this is not always the case. Some large cities may in fact experience diseconomies of scale (Muller, 1975, p. 95). It is not the purpose here to debate this issue, but to illustrate areas where expenditure analysis is essential to developing sound planning strategies. A growing use of the financial aspect of planning is for what is termed fiscal impact analysis. The basic purpose of fiscal impact analysis is to determine what costs will be created by a proposed activity and whether revenues generated by the activity will cover projected costs. The most common use to date is to predict the impact of land development projects, and its use for that purpose is increasing rapidly (Muller, 1975, p. 1). Fiscal impact analysis is a new but rapidly growing planning activity. Its growth has taken place at about the same time and is roughly analogous to, the growth in environ- mental impact analysis. The number of fiscal impact studies during 1973 and 1974 was greater than all such studies during the previous five years (Muller, 1975, p. 1). Additionally, at least two states, Florida and Vermont, have legislation requiring fiscal impact studies for some classes or projects. As the scale of projects or proposed activities changes, the tools of impact analysis also change. Three types of studies normally undertaken are: development project studies, community-wide studies and regional impact studies. Project or subcommunity level (termed microlevel) analysis generally relies on data developed at the local level. Come munity and larger studies (termed macrolevel) tend to rely on aggregated local state and national data. The analysis of municipal expenditures across com- munities is used in fiscal impact analysis for several purposes. It serves to establish a range of expenditure levels for particular government services. When different community types are adequately represented, it may be used to develop a trend in expenditures across community types. This trend may then be used to predict likely changes in expenditure levels as community characteristics change. Planners have also traditionally not been involved with the issue of local tax levels. But as public expendi- tures and tax levels have risen, planning activities have become more directly linked to the financial condition of the city. Recent Michigan legislation, notably the Industrial Facilities Tax Act of 1974 and the Downtown Development Authority Act of 1975, directly relate development activities to the city's financial condition and taxing powers. If the powers authorized in these and similar acts are to be used effectively, it is imperative that urban planners understand both the nature of municipal expenditures and the implica- tions of the resulting tax burden. Planners must be much more knowledgeable about the workings and incidence of local taxes, especially the property tax. As institutions engaged in the revitalization of cities turn more toward economic development incentives and financial tools, they are increasingly turning to other experts rather than urban planners to formulate development plans. Urban planners who are unable to understand and work with the city's finances may be left out of the most critical decisions concerning the city's future. If this is to be avoided, fiscal strategies must become part and parcel of the urban planner's tools. Study Design and Scope One of the public services most dramatically affected by the increasing cost of urban government has been public education. Many educational planners have recognized that support for other governmental services directly affects financing for local education. The specific effect of the local government's tax level on local school district revenue has been termed ”municipal overburden." This term was originally used in New York to explain expenditure variations between school districts (Polley, 1961). Arguments over the existence of municipal overburden were instrumental in the development of municipal expendi- ture analysis. Initially, the attempt was to delineate types of communities where overburden was thought to occur. Com- pensation was then recommended communities on the basis of community type, or specific characteristics such as density or total population. Many of the same arguments are being offered today for more general programs of aid to the cities. In 1974 the Ford Foundation funded a study to develop similar information for cities in the state of Michigan. Although much of the study was concerned with purely educa- tional activities, an important segment dealt with differences in specific governmental costs between central cities, sub- urban and rural governments. Of specific interest was the idea of applying the "municipal overburden" theory to Michigan cities. In order to do this it was necessary to compare the expenditure patterns for a large number of Michigan's local governments to determine what expenditure levels were and the most important factors in determining them. The original intent of this thesis was to examine the effect and importance of conflicting political bounda- ries, especially those between school districts and associ- ated minor civil divisions. This involved methods of ascertaining boundaries and allocating tax base and costs between minor civil divisions. However, as the study progressed, it became evident that this was not suitable. Most planning tasks do not involve the calculation of this type of information. The problem, as originally stated, did not deal with the creation of boundaries, or with methods of changing them. The problem is at best a limited one involving the allocation of costs and benefits of special purpose districts. Allocation methodology for these has long been established with little controversy through the use of tax code districts and special assessment districts, and the use of tax code districts did not prove suitable for expenditure analysis between minor civil divisions. At the same it was noted that there is a lack of original research in Michigan relating the cost structure and determinants of municipal finance to urban planning functions. As the research for the Ford Foundation study proceeded along these lines, it became evident that analysis of the substance of municipal expenditures would prove more beneficial, and analysis of the substance would provide a greater opportunity to develop and evaluate techniques for analysis for such functions as the preparation of fiscal impact studies. This thesis, then, is an out-growth of that research. Research Objectives The thesis examines three major hypotheses covering both the divisions of local governments into categories and the effects of differences between those categories. The hypotheses are: 10 1. Michigan's central cities are a distinct class of Local Government. That is, they have characteris- tics which distinguish them from other local governments. 2. The characteristics which distinguish the central cities have a significant effect on the structure of municipal costs. 3. The differences in municipal costs can be explained by a limited number of structural and socio-economic factors which cannot be changed over the short run. If these statements are true, there are two specific implications for planning in the central cities that will be examined: 1. Central cities should be a viable ”client group” for advocate planners. If central cities are a class, then planners may advocate programs and activities which would benefit central cities as distinguished from other units of local government. 2. Since the factors affecting municipal costs cannot be changed over the short run, planners must seek compensation for them in the short run while plan- ing longer term corrective actions. Research Methodology The basic methodology of the study integrates three basic types of information for a cross-section sample of Michigan minor civil divisions. The sample consists of 222 11 minor civil divisions in southern Michigan, and is divided for purposes of analysis into the categories of central city, urban township, suburban city, suburban township, rural municipality, and rural township. A more complete descrip- tion of the sample is included under the description of the study's methodolOgy. A major objective in the sample construction was the idea that the comparisons should not include just cities, but those governments in the same area which are nonurban. Conventionally, this is done by the Census divisions of Standard Metropolitan Statistical Areas (SMSA's) into urban- ized and nonurbanized sectors. This sample design makes possible the analysis of differences within as well as between those sectors. A first set of questions concerns expenditures for local public services. The study will attempt to outline major variables that affect costs. This will center around such phenomena as the distribution of income, industrializa- tion, city age, and density. The methodology for this involves the creation of multiple regression equations using current Operating costs for specific municipal functions as dependent variables. The primary reason that current Operating costs are used and not capital expenditures is that current operating costs more accurately reflect the local government's financial situation than the capital expenditures for a given year. A characteristic of capital expenditures is that they are 12 "lumpy" and occur at uneven intervals. There is no assur- ance that large differences would occur randomly, and the results would distort the financial picture. Current operating expenditures are also typically a much larger portion of the government's budget than are capital expendi- tures. Because a regression which includes all types of municipalities will contain variations inherent to or caused by the type of government itself, two sets of regressions are presented. The first covers the entire sample, and deals with variations between as well as within categories. The second set of equations includes a subsample of those cities of over 25,000 population. This is presented as an alternative to the more conventional approach of using a ”dummy" variable to incorporate all variation within a single set of equations. A second set of questions deals with the extent of differences between the categories of local governments. These differences are expressed mainly through measures of the populations of each local government. This section also includes some measures of the difference in selected per capita expenditures by category. The purpose of this is to illustrate the magnitude of differences. Statistical measures may demonstrate the statistical "significance" of differences between groups, but the importance of those 13 differences is a value judgement. Thus it is necessary to go back to the basic data to determine whether or not impor— tant differences do exist. The method by which these data are presented is that of group mean data, with analysis of variance used where appropriate. This technique allows for examination of the differences between groups compared to the differences within groups, and will be explained more fully in the methodology section. The third section deals with the ability of local governments to pay for services. One portion of this deals with a unified picture of income and tax burden for each of the categories. The purpose is to assess the tax effort made by each category, and to indicate whether or not there are significant differences in tax effort by category. A significant portion of this lies in separating out the residential portion of the local property tax to determine where the burden of local taxes falls, and what implications this has for urban policy. The concluding section will draw implications for policy and administration from the data, including recommen- dations for further research and improvements in data collection. This section will also, where appropriate, indicate measures which have already been taken in both legislation and administration. 14 Study Limitations The design of any study imposes upon some inherent limitations, and data availability and usability often impose further restrictions. Some of these affect the validity of finding while others limit the extent to which inferences may be made from the findings. Because of the desire to provide uniform information regarding state influences on the sample governments, trends and inference from this data set may be applied only to a very limited extent to local governments outside Michigan. The purpose behind this restriction is to provide both a uniformity and a depth is data which is not possible if a multistate sample is used. Many authors have noted the critical weakness of inconsistent data definitions inherent in the multi-state sample, and many have acknowledged the power that the state has in setting the conditions for local finance: . . .The states hold power over these important ground rules: they specify the internal workings of urban tax systems; they set conditions that determine cost and use of land; they have a major influence over the internal content of basic services like education, transportation, health care, public safety, and justice (Campbell, 1970, p. 129). States also set the accounting rules and in many cases specify the level of government and structure within which a function is carried out. Thus the best place for analysis is where the entire sample is located within one consistent set of operating conditions and structure. 15 Time is another factor which also affects the con- sistency and validity of data. Accounting methods and categories, areas and even the structure and functions of local government change over time. Because of this, most time-series studies are limited to one variable, population. The cross-section method was selected for this study primar- ily because within the study, the underlying conditions remain constant. This applies especially to such data as cost categories and methods of counting costs. The cross- section, which is a set of observations at a single point in time also "holds constant” such factors as changes in technology. The analysis of municipal expenditures in this study has been restricted to current operating expenditures. It was felt that, for a cross-section analysis, this would provide the most valid indicator of relative expenditure levels. Capital flows would certainly be more relevant for many planning purposes. However, even with a state these are very difficult to examine in a uniform fashion. For example, many communities may form semi-autonomous govern- ment corporations to carry debts for local projects. This debt, while it may be the financing for a local government function, would not be reported with the local government's debt servicing. Thus that community's debt as reported would be lower than that for a community which financed the same project directly through the local government. This 16 process does not occur randomly, as some classes of local governments must use this more regularly to avoid debt ceilings. There also remains to be developed an acceptable method of reporting and equalizing for differing loan terms, periods and interest rates. Other limits are inherent in the regression technique. Of prime importance is that the correlations exhibited by the regression equations do not by themselves establish any causality. This is of special significance when variables such as intergovernmental revenues are discussed. Some studies have found significant correlations between expenditures and intergovernmental revenues, imply- ing that there is a cause and effect relationship. While there is no firm evidence, it seems likely that the relation- ship may proceed either way. Government aid may be a result of increased expenditures which generate pressure on politicians to seek aid. For this reason, some variables of this type have been excluded. Relationships among independent variables may also distort the findings. The tendency here is for correlations to be lower and standard errors larger than would be the case if there were no interrelationships. Distortion may also occur where there is significant nonreporting or misreporting among particular cost variables. These limits confirm that the regression is an imperfect tool for approximating reality. The strength of the regression in the context of this study is that it 17 indicates which variables may be the most important and the degree to which variation may be explained. The actual value of the regression coefficients is subject to change and may be considered of lesser importance. The number of variables which may be used with confidence in a regression is limited, and is obviously much less than the number of factors in the real world which affect costs. In many cases, no data exist for these variables and they cannot be included in such a study. The degree to which many variables are still unexplained indi- cates that there is much yet to be discovered in the field of expenditure analysis. A severe limitation even within the confines of a single state is the lack of uniform accounting procedures. Analysis of the 1972 Census of Governments data indicate that in many cases local governments have not reported expenditures in categories where they almost certainly occurred. Such nonreporting is most common with very small governments. A related problem is the lack of uniform definitions for accounting categories. This makes it virtually impos- sible to make cross-government comparisons of many specific expenditure categories. For this reason, the number of specific expenditure categories considered in this study is quite small. It is limited to those categories which can be considered common to most if not all local governments. 18 Finally, this study does not deal with many other facets of local, state and federal regulations which affect costs and services. In most cases these do not seem suscep- tible to the type of analysis conducted here, but require an approach incorporating the political decision making pro— cess as well as the dollars-and-cents results. II. RELATED LITERATURE AND RESEARCH The analysis of municipal expenditures is not a new activity; many studies have attempted to relate expenditure levels to various demographic, structural and financial variables. The vast majority of those studies have taken place since the Bureau of the Census began consistent collec- tion of municipal finance data. The 1950 designation of Standard MetrOpolitan Areas and the expanded information ' provided for those areas also facilitated the growth of such studies. This section does not attempt to cover all expendi- ture studies, but only those which have some significance either in establishing techniques or are of special interest to planners. These include some earlier studies which developed the regression approach and several later studies which give the analysis of expenditures a planning focus. The purpose of the search of related research on municipal expenditures was to establish basic techniques used for expenditure analysis and to examine the extent to which this has been related to planning practices. To 19 20 accomplish this, literature in planning, economics, and finance was reviewed, and the relevant studies divided into three categories. These include: (1) Early studies which established the basic techniques for expenditure analysis. These were used not only for methods of analysis, but to determine which variables might prove most useful in explain- ing expenditure variation; (2) A second category of studies addresses more directly issues of metropolitan fiscal relations and urban-suburban differences. (3) The last category deals with recent direction and special interests in expenditure studies. This yields some measure of progress being made and the flexibility of techniques. It was origi- nally thought that this area might have greater applicability to planning issues. Early Studies of Municipal Expenditures Among the earliest and most widely known analyses is Brazer's 1959 study (Brazer, 1959). This study involved the analysis of expenditures of a nationwide sample of 462 cities of over 25,000 population. Data for the study were drawn from the 1940 and 1950 Census and 1952 Census of Governments. Subsamples for analysis were composed of the 40 largest cities, 35 cities in California, 30 in Massachusetts and 32 in Ohio. Brazer's major focus was to correlate expenditure levels to independent variables describing the city's structure. The basic methodology was to construct regression 21 equations for each of eight expenditure categories using six ”independent“ variables. Following is a list of the vari- ables used: EXPENDITURE VARIABLES INDEPENDENT VARIABLES total operating expenses p0pulation police density common functions population growth fire median family income highway intergovernmental revenue recreation percent employed in general control manufacturing and services sanitation The same variables were also used for the subsample analysis. Although some previous attempts to use this format had been made, this was the first successful study using a number of specific expenditure categories in a multiple regression format. One of Brazer's major findings, though not reported quantitatively, was that state and regional location account for a great deal of the variation in expenditures. This pointed the way toward examination of expenditures within individual states and regions. However, the individual state analyses presented by Brazer did not seem to add signi- ficantly to the explanation of expenditure variation, nor did the coefficients of the 40 largest cities differ markedly from the whole sample. In developing hypotheses of urban interaction, Brazer relied heavily on the earlier work of Amos Hawley. This was one of the earliest studies which found a positive correlation 22 between central city expenditures and suburban growth. The specific hypothesis was that the expenditures (for the cen- tral city) are positively correlated with the percent of population living in the suburban portion of the metropolitan area. Brazer's findings added some support to this. Brazer attempted to shed some light on the question of optimum city size by including population as one variable. His findings was that, when other sources of variation were taken into account, size alone does not seem to significantly alter the spending pattern. Whether or not this would remain across all city sizes was not clear. Several weaknesses showed themselves in this early attempt. The selection of variables left many possible areas of causality uncovered. In part, this was undoubtedly due to lack of data for such variables as property valuation and age of structures. While Brazer's study included median income as a variable, it had no measure of those families or individuals living in poverty. Likewise, the distinction between manufacturing and other types of cities applied to the labor force, not the location of facilities. Brazer concluded that the class of city was also a major factor influencing expenditures. Class in this con- text referred to the city's structure, i.e., whether the city was a central city of a metrOpolitan area, an industrial or residential suburb, an independent city or a major resort area. However, little quantitative evidence was offered to 23 support that conclusion or to indicate the specific manner in which class would influence expenditure levels. The problem of correlation versus causation in regression studies is illustrated by the findings on inter- governmental revenue. Brazer highlighted his findings on intergovernmental revenue as this was the only variable found to be significant with all expenditure categories. He felt that, given the broad nature of the sample and the strength of the relationship, there would be a cause-effect relation- ship. However, there are many possible explanations, among them that city governments are most likely to seek aid for those expenditure categories which are highest or increasing most rapidly. This illustrates the danger involved in using this form of analysis without a strong underlying theory. A study by Adams (1963), utilizing a national sample but with a different construction, yielded somewhat similar results. Adams constructed a sample of 478 county areas by aggregating the expenditures of local governments within those counties. This gave a single set of expenditures for each county unit. Adams divided his variables into five categories: socioeconomic, physical, income, population, and political. This was a significant increase in the number of variables of previous studies, since each category had several vari- ables. Variables under socioeconomic included density, percent urbanized and per capita hotel, motel, and camping 24 receipts. Physical included region (of the United States), age of housing and percent multiple-unit housing. Population included percent foreign-born and non-white. Political included percent (of pOpulation) recent immigrants to the local area and the number of political units. This last is a variable unique to this type of sample construction. Those results which differed from Brazer's indicated that political fragmentation affected the per capital expen- ditures significantly, as did density and urbanization. The finding on political fragmentation was not unexpected or important in itself, but pointed the way towards a more systematic analysis of the costs and benefits of metropolitan government. One critical aspect of Adams' study which may have influenced the findings lies in the sample construction. The independent as well as cost variables were pooled into county areas, cancelling out much of the urban-suburban vari- ation. In effect higher city costs and lower suburban costs (where they existed) were diluted. Where ”urbanization“ variables such as density and percent urbanization were used, the results would be weighted by very large metropolitan areas. Thus one would not expect this type of study to show the same kind of results as one which included city and sub- urban governments as distinct observations. A different problem with this study and several others lies with variable choice. As many observers have noted, those areas with high expenditure levels also seemed 25 to have a fairly large percentage of non-white population. This has led many researchers to include non-white as a separate variable, and in some cases to achieve statistically significant result, as Adams did for Some categories. The real problem lies with the theory behind this correlation. The general hypothesis is that these peOple require more services because they are poor or some similar condition. But if this is so, then proverty or some related measure would be much better. The use of variables such as non-white or percent minority, unless accompanied by very specific explanations, creates two problems: The first problem is that it puts the researcher in a position of "blaming the victim". Many researchers move quickly from the correlation to the inference that the minorities actually cause increased expenditures. The second problem is statistical in nature, but may also damage the credibility of the research. Poverty and race are often highly correlated, for well-known historical reasons. Where these two are used together the result is likely to be highly confounded, obscuring the relationship with each of the variables and possibly reducing the appar- ent or calculated significance of both. The reason behind this is that multiple regression is based on the assumption that the independent variables are not related, or at least the relationship is not known. With race and income there is a direct and usually significant relationship. It is therefore unwise to use the two together without extreme 26 safeguards and equally unwise to use a variable such as race unless there is some very good reason for including it. A major focus in and since Brazer's study has been the separation and identification of "supply" and ”demand" components of cost variation. Adams devotes a major portion of his study to developing just such a formulation, with unconvincing results. The idea of the supply and demand functions comes from the conception of public finance as an analog to the private market, with a notion of "supply” and ”demand" setting the level of public expenditures: . . .The annual cost associated with the operation of local government is argued to be functionally related to various demand factors as well as technologically based supply side factors. To the extent that either are not specified; by default, they will be found with- in the error term of the determinants equation (Beaton, 1974, p. 251). This search for the supply and demand functions is expressed somewhat differently in a review of expenditure studies (Weicher, 1970, p. 379). Weicher notes four relationships which are analyzed by most studies: the possi- ble existence of economies of scale, effects of political .fragmentation, the effects of intergovernmental aid on local expenditures, and the response of expenditures to changes in fiscal capacity. He indicated that the studies he examined dealt inadequately with "tastes" (demand) and with service conditions (supply). The result, in his opinion, is that these studies fail to account for much variation, especially with regards to highway and sanitation 27 expenditures. Several later studies have continued the search for supply and demand functions. However, warnings have been sounded about the appli- cability of the supply and demand analogy. In particular, the analogy between the supply theory of the firm and municipal expenditures has been severely criticized (Bahl, 1968, p. 24). The use of the firm's supply function in a municipal expenditure context would require that the local government produce its own population. This type of reason- ing has led some researchers away from more productive forms of analysis. During this early period of expenditure analysis, most studies have been directly or indirectly concerned with the problems of growth (Campbell and Sacks, 1967). This is reflected in the attempts to ascertain the dynamic character of hypothesized economies of scale. Netzer (1969) has focused on this quality, especially the developmental aspects of suburbanizing areas. His cen- tral findings have been that smaller municipalities have higher per capital expenditures as a result of high "start-up” costs in both physical develOpment and human services (Beaton, 1974, p. 251). However, this seems to address more the problems of a high rate of growth than size or structure. Netzer may also have surveyed smaller commun- ities (much less than 25,000) than those who have not found economies of scale. 28 Urban-Suburban Relations and MetropOlitan Areas A major result of the early expenditure studies is that greater direction shows up in later attempts. Two notable trends are those toward the fiscal aspects of metropolitan governments and the effects of suburban growth. Bahl (1969) presents one of the most complete analyses with regards to variation in metropolitan areas. Bahl concentrated on developing what he termed a "behavioral theory” of municipal expenditures, ignoring for the most part the traditional supply and demand theory. The technique involved both the creation of static regression equations for a national sample of 198 SMSA's and the exploration of the dynamic properties of selected variables. The results of Bahl's study are among the most con- vincing and relevant of expenditure studies, in good part because they are oriented toward policy and practice rather than theory. In assessing the implications of his results. Bahl addresses planners and politicians rather than econo- mists. In detailing urban-suburban differences, the study goes to some length to point out the fiscal implications of long-range land use policies: . . .The planner must anticipate the general effects of the metropolitan land-use plan on the central city fisc. A plan that does not discourage residential migration to suburbs is almost certain to sap the fiscal resources of the city, but as demonstrated, is not as certain to reduce expenditure requirements (1969, p. 131). 29 The dynamic qualities of growth and decline explored by Bahl address several issues. These include the changing character of urban-suburban differences, the effects of differing rates of change and the question of whether there is a ”life-cycle" pattern to urban structure. Bahl made considerable progress in defining the issues and detailing which variables might be useful in further research. As in most expenditure studies, less progress was made toward high- lighting those variables which account for the greatest portion of variation. A problem in accounting for variation seems to lie in approaches which attempt to maximize the percent of variance accounted for (R2) rather than isolating single important variables. Bahl's equations included 18 indepen- dent variables. Even with the relatively large sample size, this many variables tends to "use up" available variation, obscuring some relationships. Where this happens, other measures must be used to indicate which variables have a singularly important effect for the researcher. Unlike some researchers or theorists, the planner or practitioner is not seeking to account for "variation." The goal is to find out which variables should be manipulated, and in what fashion, to achieve desired impacts. Such research should also demonstrate what side-effects or externatities might be expected when certain variables change. Despite some methodological problems, Bahl points out some trends of real importance to planners. As noted, 30 population decline due to residential outmigration will not lower service costs, even when variations due to other factors (such as increased poverty) are accounted for. Population size was significant only for total operating expenditures. In this instance the coefficient was not large, more or less sustaining Brazer's argument that econo- mies of scale are limited at best. The 1950-1960 rate of growth was significant only for highway expenditures, indicating a very weak relationship between growth (of established cities) and costs. On the other hand, pOpulation density was signifi- cant for all but highway expenditures. Regression coeffi- cients (the B-value or "slope") tended to be fairly large, although in most cases the R2 values of the whole equations were low. This means that although the relationship was both significant and important, there remained a great deal of variation to account for. In very few cases (notably fire and police expenditures) were R2 values above .50. One very noticeable trait of Bahl's regressions is that in most of the equations, five or six variables account for most of the variation. This leads to the conclusion that careful construction and consideration, not the number of variables, will determine the success of equations. An example of this is the use of the fringe popula- tion ratio. Both planning theory and other studies have indicated that a high fringe (suburban) population imposes costs on the central city. Bahl included this as an 31 independent variable, yet the results are obscured. Bahl admitted that this is because of the intercorrelations with other, less important variables (1969, p. 80). Removing a number of variables from consideration could have clarified some outstanding relationships. A later study focused more closely on the urban- suburban population question (Kasarda, 1974). Kasarda began with a much more specific question than most studies. His was an attempt to verify and expand the "urban-suburban exploitation thesis" (Neenan, 1970). Briefly, the urban-suburban exploitation thesis states that suburban areas "exploit" the central city by characteristically using more public services than they apply for in taxes, charges and transfers. Neenan measured this effect for the Detroit SMSA and found that suburban residents experienced a net gain from Detroit city services. This gain ranged from approximately $1.75 to $12.50 per capita annually. The gain results both from free public services provided by Detroit to nonresidents and subsidizing some regional services. Kasarda expanded the concept to deal with services rather than net fiscal gain. He created an analysis of 158 SMSA's considering four functions: retail trade, wholesale trade, business and repair service, and public service. This construction is unusual in that the private sector services are not normally thought of as "costs." However, they are very important in establishing the role of suburbs. 32 Their inclusion is designed to indicate the degree to which the central city acts as a service center for suburban areas. This sets the stage for the relationship between suburban population and central city costs. The technique used by Kasarda is called path analy- sis. This is somewhat different from regression in that it separates the direct and indirect effects of each indepen- dent variable on the dependent. This is acheived by examin- ing the intercorrelations between independents and comparing these with the direct effects. From these direct effects it is possible to determine which independent causes the greatest change in the dependent. Unlike the regression, the end result is not an equation which calculates the estimated amount of change in the dependent for each change in the independent variables. Thus path analysis is better suited to isolating important variables but not for esti- mating impacts. When expenditures are controlled for city age, per capita income and percent non-white, Kasarda finds that ". . .The size of the suburban population, rather than the size or composition of the central city population is the most important determinant of central city expenditures for public services (1974, p. 145). When commuters replace suburban population in the relationship, the results are essentially the same. The relationship between commuters and suburban population is almost perfectly linnear, adding 105 commuters for each 1,000 suburban population. This relationship seems to 33 indicate that suburbs do not become more "independent" of the central city as they grow larger. On the average, there were 132 commuters for each 1,000 central city residents (1960 data). One factor which does change is that highway expendi- tures were not significantly correlated with suburban population. When commuters were substituted for population this relationship was positive and highly significant. This indicates the value of the analysis in isolating direct from indirect effects. This is especially important for this type of urban analysis, where many variables are highly correlated with each other and have both direct and indirect effects. Kasarda's analysis does serve to strengthen the "exploitation" argument. The relationship described is limited in that the data cover only the ten-year period through 1960. This particular study should be extended through 1970 data to see whether the trends indicated still hold. This would be of particular interest for the private sector variables. The study covers the period when residen- tal outmigration had been substantial, but before massive commercial and industrial relocation had taken place. Neenan's study, using later data, indicated that the fiscal relationship still held. But the pattern of use may be changing toward a trend of the central city providing more public and less private sector service. This should produce drastically different fiscal implications. 34 The preceding studies have shown that the format of lumped independent variables used indiscriminantly against cost variables will not account for the greatest part of cost variation. Even within states, these commonly leave the majority of variation unaccounted for. As many authors have realized, politics plays an exceedingly important role in the allocation of public funds. While many political con- sideration are not yet subject to quantitative analysis, attempts have been made to include government structure and organization. This deeply touches many aspects of planning, as it is virtually an article of faith that efficiency in public services can be improved by the restructuring of local government functions. This has been applied to repeated proposals for metropolitan government and regional planning as well as tax-base sharing and multiple-unit service arrangements. It is possible to include some measures of political structure in a quantitative format. Adams' political frag- mentation variable was one simple example. More sophisti- cated measures would account for structural differences in functional categories within the same political unit. If successful, this might significantly increase our explana- tion of variation by allowing us to specify different changes or variables for each function in a systematic fashion. Beaton (1974) attempts to create just such a system for expenditure analysis. He begins by outlining four 35 general structural conditions: city size, growth, decline, and "fiscally unique functions performed by the different agencies” (1974, p. 250). There are two differing opinions on the effects of city size on per capita costs. One position, that taken by the Advisory Commission on Intergovernmental Relations (ACIR) is that for most cities there are not significant economies or diseconomies of scale. Diseconomies of scale seem to be present only for cities over 250,000 (ACIR, 1968, p. 52). This general positions is also supported by Gabler (1971). The ACIR reviewed its position in 1976 and found no reason to alter it (ACIR, 1976, p. 58). Beaton, on the other hand, finds that when costs are disaggregated it is possible to discern size effects. This is because when city size increases, interactions among the population tend to increase exponentially. Therefore those functions relating to human interaction would find their cost increasing faster than the population. Functions which have a production function similar to the private firm's may experience economies of scale. These services would have decreasing costs per capita, at least up to some limit. As a result of these differences between functions, Beaton aggregates cost into two functional categories. These functions were earlier delineated by Hirsch (1959) on the basis of long term response to increasing municipal size. The two categories used by Beaton are: (l) horizon- tally integrated functions, in which a single government 36 unit controls all units performing the same functions, e.g. police department; and (2) circularly integrated functions in which the same unit provides a number of complementary services. Thus, in contrast to most other studies, Beaton has only two cost categories. Circular costs include: general government operating costs, general and financial administration and planning. Horizontal costs include public safety, public works and recreation. Beaton takes a second major step in disaggregating cities by what he terms "environmental parameters" (1974, p. 256). The total sample of 567 New Jersey municipalities is broken down into 16 subsamples. This is roughly equiva- lent to a scheme of city classification. The sample is broken into four size groups: (1) 1,000-4,000, (2) 4,000- 15,000, (3) 15,000-25,000, and (4) 25,000-383,000. Each size category is divided into growing or declining categor- ies, and each of those is divided into "rich" and "poor" cities. Rich cities are those with a property value of over $8,500 per capita, poor are those below $8,500. Regression equations are then set up for each subsample and Beta coefficients used to rank the results. With variation already divided into the subsets, the coefficients appear very impressive: in most cases at least 90 percent of the variation is accounted for. However, some very serious flaws present themselves. Because of the sample division, many subsamples are very small; one sub- sample contained no observations and was deleted. There 37 were only 46 total declining cities, divided into eight subsamples. Therefore many of the results are statistically invalid. They are not necessarily false, but cannot be relied on even within the context of the sample. A more serious problem is that the empirical research approach seems inconsistent with Beaton's theoreti- cal approach. The division of the sample by both environ- mental structure and cost category is designed to allow a choice of variables to be used in accord with theory. For example, Beaton theorizes that horizontal functions are more likely to be explained by socioeconomic variables than circular functions. Since Beaton uses all 16 of his indepen- dent variables for every equation, this effect cannot be discerned. The 16 independent variables cover the full spectrum normally used for expenditure regressions, includ- ing income, employment, aid, density, crime rate, age and debt. To follow Beaton's theory, variables should have been grouped, those with high intercorrelations omitted or used in separate equations and the success of explanation for each equation noted. Thus although Beaton's theory makes some significant advances over more traditional approaches, the empirical test falls short of proof. It would appear at this point that the basic techniques are adequate if more sOphistication is used in: (1) developing samples and subsample categories; and (2) making regression equations and other statistical techniques accord with theory. More needs to be done to 38 refine functional categories such that all important cost variables can be categorized. As data improve this will increasingly need to include such functions as housing, which do not easily fit the horizontal-circular dichotomy. From the planner's point of view, the idea of structurally different types of services offers a chance to realize some improvements in services and more accurate anticipation of trends affecting services. The theoretical improvement is that, for an individual city it is probably too much to ask to formulate a unique set of variables affecting each individual cost function. On the other hand, specifying one set of variables for service functions in general will not be specific enough for planning and budget- ing purposes. A refined set of functional categories would allow the isolation of organizational traits which can link the independent variables to specific functions for projec- tions. Other Recent Directions in Expenditure Studies More recent studies have given some focus to declining metropolitan areas. Stinson (1970) analyzed both the expenditure and revenue patterns for cities following population shifts. His sample consisted of United State cities outside SMSA's with populations between 25,000 and 50,000. Stinson analyzed sources of revenue, increases in revenue, changes in expenditures by function and allocations (percent of budget) by function. 39 Stinson did not use any form of multivariate techniques, but pointed out the simple relationships between population change and expenditure and revenue changes. He did this by dividing the sample into two parts: cities reporting a population decline during 1950-1960 and cities with a growth rate between 33 and 100 percent during the same period. Controls for other sources of variation were not included in the statistical analysis. Stinsons basic intention in this was to give a broad picture of expendi- tures, revenues and debt structure without an overburdening methodology. The results of expenditure comparisons were not definitive for specific categories, but indicated no sub- stantial shift in the allocation of expenditures between functions. One significant difference was the higher proportion of capital expenditures in growing cities. This is not an unexpected difference, but might indicate that declining cities were not making substantial replacements of infrastructure. The other notable difference was that declining cities had a higher prOportion of short-term versus long-term debt. Although Stinson's research produced few specific results, it marked an early attempt to assess the planning and finance implications for declining cities. Stinson recognized the short-comings of such a broad study and simple technique and urged the development of more sophis- ticated techniques for revenue and expenditure projections. 40 His concerns for smaller cities also foreshadowed the spread of outmigration and its attending symptoms of decline from larger metropolitan areas to a more general pattern. Another recent study (Liebert, 1974) addresses the question of relating municipal expenditures to community decision-making processes and structures. Liebert analyzed a number of comparative studies on municipal structure to determine whether the effects of different power arrange- ments were reflected in expenditure patterns. One of the first variables examined by Liebert was the functional scope of local government. In many ways, this more specifically states the objections to multi-state samples: different local governments are responsible for different functions. Most/expenditure analyses are organized by local government level: not by function. But the same / local government levels/in different states do not have the same responsibilities; what is assigned to the municipal level in one state may be found at the county level in another. Much of the higher average expenditures in New England are explainable in this fashion (Liebert, 1974, p. 767). This may be cdmplicated for some functions by the addition of multi-county and multi-state regional agencies and districts. Liebert's conclusion from this is that the "general municipal expenditure level" is actually a two-dimension variable incorporating both the functional inclusiveness and performance level (1974, p. 770). To study performance 41 levels, inclusiveness must be controlled. In the planning context, this might be reversed; performance could be con- trolled for to study the comparative effects of functional inclusiveness. Two hypotheses or proposals that emerge are that it is possible to create measures for both inclusiveness and decentrization and that these will have an impact on per capita municipal expenditures. Liebert cites a study (1974, p. 777) in which an index of decentralization was positively correlated with per capita expenditures. The tentative explanation for this is that as more people enter the decision-making process, more demands will be made for special programs. Alternative explanations are possible, particularly that as more peOple participate in the decision-making process, they will be more willing to pay (taxes) for public services. But whatever the explanations, the relevant point is that measures can be devised to test decision- making alternatives that relate directly to the organization of planning functions. The measures discussed by Liebert directly relate expenditure levels to organization type and structure. Planners attempting to decide where routine decision-making should be carried out and the appropriate governmental levels for specific functions may look directly to this type of empirical information to determine fiscal 42 consequences. This would be particularly true if per capita expenditures are reasonable measures of public preference. Other studies have organized expenditure research around variables such as voting. Davies and Haine (1966) suggest that voter interest groups add significantly to the explanation of per capita expenditures. It is noteworthy that social welfare concerns do not even enter into this model, which might be described as a test of pure "interest group” theory. The results were not overwhelming, but indicate there are certainly elements of truth to the theory. On a somewhat more conservative note, Sharkansky (1967) indicates his preference for previous expenditures as the primary determinant of current (and future) expendi- tures. This incremental approach is certainly valid for some functions. However, it should be noted that Sharkan- sky's study was developed before the popularization of PPBS approaches to public budgeting. It is likely that past expenditures will play a decreasing role in future alloca- tions. At a broader level, Dusansky and Nordell (1975) put forth the possibility of constructing a general model dealing with all aspects of the urban-suburban fiscal situation. However, their model is more closely derived from conventional economic theory and does not rely on decision-making alternatives to impact finances. What good are these studies? A survey of the progress in expenditure analysis reveals that this branch of 43 analysis has progressed beyond the "background information" stage and is now an active ingredient in both policy and planning. One of the most important users of these studies is the ACIR. They provide not only recommendations and models for federal legislation, but develop state guidelines and assist other federal agencies in drawing up regulations and administrative policy. ACIR has taken an interest in the organization, ‘ extent and effectiveness of urban and regional planning (ACIR, 1976). Finance is both a measure and means of success in planning. Expenditure studies which point towards more effective levels of government and which relate local expenditures to public decision-making processes enable ACIR to make recommendations for the proper focus of federal legislation and appropriations. State and regional agencies may also find important uses for expenditure studies. An example is the Michigan State Housing and Development Authority's study of Michigan cities over 10,000 population. This may aid in establishing criteria for statewide programs and determining trends in expenditures (and taxes) affecting the housing stock. Perhaps the most important contribution expenditure research can make is the inclusion of future finances in the planning process. Linking physical planning to the city budget is difficult and can only be successful if planners understand the critical relationship between finance and the city structure: 44 ” . .The urban finance problem, perhaps the most com- plex of all urban problems, presents an especially pressing need for the coordination of fiscal and physical planning” (Bahl, 1969, p. 131). III. METHODOLOGY The methodology for develOping an integrated analy- sis of municipal expenditures and characteristics consists of several discrete steps. The first is the method of classifying or grouping local governments. The second is the method of analyzing cost data with structural and other community variables. The last is the analysis of a group differences in both cost and community variables. The first section of this chapter describes the creation and structure of the study sample. This includes the sample's classification system. The second section presents the regression technique used to analyze expendi- ture levels. The third section describes the analysis of variance that is used in conjunction with the classification system and which identifies differences between community types. The last section contains a description and listing of variables used in the study. Data sources are contained in a separate appendix. An integral part of the analysis of expenditures and other variables is the attempt to classify or categorize local governments. The basis or rationale for this is that 45 46 if different classes of local government have unique expen- diture characteristics, these may be measurable and stable enough to constitute the basis for specific government programs, either at local or higher levels. Some criteria must be used to define which are rational dividing points between types of governments, both for purposes of analysis and as a basis for policy decisions. The regression equations for expenditure, on the other hand, operate independently of the classification system. Therefore any weaknesses in the classification scheme will not affect this portion of the analysis. The relationships thus established may be used to test the classification scheme. This may be of special value where legislation or administration are required to distinguish in some manner between categories of local governments. Thus, what appears to be a separation of techniques in the study is actually the method by which each part is tested and reinforced. Sample Description The classification system used in this study is designed around the idea of urban--suburban—-rural differ- ences, taking into account the basic differences between forms of government. Many studies have used the concept of separating urban and nonurban communities, but few take the specific form of government into account. The more usual practice, especially when using census data, is to separate 47 the sample into: Central city of an SMSA (urban) noncentral city area of the SMSA (suburban) and non-SMSA (rural). The lack of differentiation in government types has been noted in political studies (Campbell, 1970, p. 174), but the point is equally valid for expenditure studies where the political or governmental structure must be taken into account. In Michigan, the revenue limitations inherent in township government are especially important in assessing expenditure patterns for basic services. The selection of school districts for the Finance Study determined the selection of municipalities, but not the number of municipalities in each category. One problem resulting from that approach was that the category of rural municipalities was severely underrepresented. Thus, few conclusions can be drawn about communities in this category. The selection of school districts was designed to serve essentially the same purposes as the sample of munici- palities. That is, it was designed to illustrate expendi- ture differences and community characteristics of urban, suburban, and rural communities in the same area. Therefore, the methodology for selecting the school districts will be presented first, then the municipal sample selection. In selecting the sample for the Finance Study, the intent was to achieve a sample of districts which would represent each of several types of districts, based on geographic location, structure and selected financial 48 criteria. The sample is not intended as a random sample or representative of the state as a whole. Rather, the purpose was to construct a sample that would be useful for examining specific issues relating to differences between urban, sub- urban, and rural districts in lower Michigan. The sample was designed to be representative of this area. In order to emphasize the structure and problems of mediumesized cities and their surroundings, the selection of districts centered around the 13 districts of the Middle Cities Education Association, 11 of these Cities are also the central cities of SMSA's; two, Benton Harbor and Battle Creek, are regional centers outside of SMSA's. All 13 of these districts are included in the sample, such that the sample consists of the central city district of each area with a sampling of its surrounding suburban and rural districts. Originally it was decided that the sample would include districts of five categories (of district type) as classified by the Michigan Department of Education. This classification consists of the categories of: metropolitan core, independent city, independent town, urban fringe (suburban) and rural. The system is essentially based on the district's relationship to the nearest central city and the size of the included municipality. However, it was determined that there was an insufficient number of independent cities and towns (categories two and three) to be useful. Also, it was felt that the department of 49 Education's limit of ten miles is to the nearest central city for consideration as an urban fringe district was too res— trictive, since normal commuting activity extended well beyond that range. To deal with this problem, category two and three districts were reclassified into categories one and four on the basis of characteristics of their constituent municipalities. The reclassification resulted in one category two district (Benton Harbor) being moved to category one and one (St Joseph) being moved to category four. All category three districts were moved to category four. If Benton Harbor is considered on the basis of its historic role as a regional center in Southwestern Michigan, the reclassifi— cation agrees more closely with Census definitions of urban and suburban areas, while retaining greater differentiation than would be possible if the categories were based solely on Census criteria. After preliminary analyses had been conducted, it was determined that the wide variation within the urban fringe (category four) group necessitated the creation of a fourth category. Rather than revert to the Department of Education's independent town or city classification, a category of suburban city was created. This category included all category four districts located with or pri- marily within the boundaries of incorporated municipalities. This change was made, however, after the selection and 50 classification of the municipal sample. But the change did bring the school district sample closer to the underlying philosophy and structure of the municipal sample. One original criterion for sChool districts was that they would be as nearly as possible coterminous with munici- pal or minor civil division boundaries. This was drOpped since few if any of the districts actually met the criterion. This meant that there was an imperfect, though close match- ing of municipal and school district boundaries. In practi- cal terms it was partially responsible for the high number of suburban townships in the municipal sample. As preliminary research had indicated a significant difference in millage rates and tax base between urban, suburban, and rural districts, a set of districts exhibiting both the extremes and central tendencies was desired. Therefore, the sample includes for each central city group: the central city, the low millage-low valuation (SEV) district, the high-millage-low SEV district, and the high millage district. In adding the central tendency, the district nearest the mean values was chosen, unless that district exhibited unusual coterminality problems. Where that occurred, the next district nearest the mean was chosen. Table one indicates the sample size and breakdown by category. 51 Table l.--Sample size and breakdown. Category No. of Districts Percent of Percent of Category Sample I 13 100 19 II 36 30 53 III 19 32 28 Sample total: 68 districts Local Government Sample The sample of local governments consists of those local governments which are included within the boundaries of the sample school districts. For the purpose of the Ford Foundation Finance Study, the object was to include in the sample jurisdictions having the same characteristics as the school district sample, and where relevant to merge school district and local government data. In creating the sample, an attempt was made to create a sample size that would prove workable, yet large enough to yield meaningful data both for trends and sub- samples. While it was possible to set subsample size in the school district sample, the same was not possible with the municipal sample, resulting in the rural municipalities being underrepresented. It was not felt that this was a serious loss, since the total number of communities in this category is very small. The size of categories becomes more critical when dealing with finance data. This is due 52 not to the nature of the information itself, but to the sit- uation where, unlike Census data, there may be a substantial amount of missing data. Thus the number of observations used for any one calculation may be smaller than the sample total. The sample was originally constructed using all cities, villages and townships which fell inside the sample school district boundaries as mapped by the Michigan Depart- ment of Education. Townships with less then ten percent of their area within the districts were excluded. The results were then rechecked by reference to a similar map from the Municipal Advisory Council of Michigan. The exclusion of townships with less than ten percent of their area within the school district sample served to reduce the categories of suburban township and rural township, which are still large in relation to the total sample. I A final check consisted of adding portions of the state equalized valuation allocated to school districts from the constituent local governments. This total for each school district was compared to the actual valuation reported by the district. Although some minor discrepancies were found, the system proved essentially sound for most purposes. This matching process was carried out to insure the comparability of the two systems. It does not carry any implications for analysis within the municipal sample beyond its use in the classification system. 53 For purposes of analysis, the 222 local governmen- tal units contained in the sample were classified by both category and type of government. In drawing up the original sample, villages were included as separate units in the categories of suburban cities and rural municipalities. Local governments were assigned to categories based on the category of the school district in which the majority of the area of the local government was located. The original classification of the sample is given in table two. Table 2.--Sample classification. Category Classification Number 1 Central City 13 2 Urban Township 18 3 Suburban Municipality 32 4 Suburban Township 77 5 Rural Municipality 20 6 Rural Township 62 Total 222 When data collection began, it soon became apparent the villages should have to be dropped as separate units of government. This is because both the 1970 Census and the state tax records do not use the village as a separate reported category, except for millage rates, population and total housing count. For other data, villages are included 54 as parts of the townships they are located in. This resulted in the revised sample given in table three. Table 3.--Revised sample classification. Category Classification Number 1 Central City 13 2 Urban Township 18 3 Suburban Municipality 28 4 Suburban Township 77 5 Rural Municipality 4 6 Rural Township 62 Total 202 As noted previously, the selection of school dis- tricts was not a random one. However, within categories the selection of local governments was essentially random since there is no necessary correlation between school districts and local governments for other than population and area. The financial criteria that were used in select- ing the school district sample have no direct counterpart in local government although, as mentioned, the relationship between school district and local government finance was a subject of investigation for the Finance Study. One problem in the sample, especially with regard to central and suburban city classification, arises when approaching the Detroit Metropolitan Area. Many suburban 55 cities in that area are larger than some central cities in the sample. For example, the smallest central city, Benton Harbor, has a population near 15,000. Several other central cities have pOpulations less than 50,000, yet are undeniably regional centers. By contrast, Ypsilanti is twice Benton Harbor's size and Southfield, the largest suburban city in the sample, exceeds 70,000. This difference in size obviously affects some class differences. This is a major reason why the expenditure analysis must Operate indepen- dently of the classification system. Multi 1e Re ression Analysis of Expenditures The heart of the analysis of municipal expenditures is the multiple linear regression technique. What this technique attempts to do is to quantify the expression, "all other things being equal." That is, it "holds still" cer- tain factors which constantly change in real life while it tests the relationship between each "independent" variable and the "dependent” variable. In a sense, the theory behind multiple regression is one of implied causation. In simple bivariate regres- sion (correlation), the coefficient merely states the degree to which the two variables are related, or vary together. Multiple regression goes a step further and allows the researcher to specify which variables are ”independent”, i.e. created by factors outside the system, and which variable is "dependent" or affected by the 56 "independent" variables. Note that in this construction, causation is implied, while the strength of the relation— ship is the test hypothesis. The regression operates by testing the simple relationship of each independent variable with the depen- dent, with variation created by the relationships to the other variables excluded. This carries with it an important assumption. For each variation to be correctly measured, it is assumed that the independent variables are not in any functional way related to each other. Where this assumption is violated, variation ”moves" from one variable to another, causing larger error coefficients and indicating a weaker relationship than may actually be the case. In practice, few independent variables are totally unrelated, so that most regressions suffer to some extent from this problem, which is termed multicollinearity. The relevant question is to what degree this impairs the equation. This question cannot be answered by statistics alone, but requires a working knowledge of the variables and theories being tested. In a normal multiple regression, a problem.may occur due to the other in which variables are introduced into the equation. The first variables entered may "use up“ the variation in the dependent variable. The result of this will be that later variables may show up as less important than they actually are. To deal with this pro- blem, this study uses a variation of multiple regression 57 known as stepwise regression. In this variation, the inde- pendent variables are separately tested against the depen- dent. The independent variable with the strongest relation- ship is entered first, the second strongest the second, and so on. The strength of the relationship is measured by the degree of correlation and the statistical significance of the correlation. This stepwise procedures is intended to give the most accurate picture of the relationship between a dependent and several independent variables. It represents the best measure of the linear relationship between vari- ables, although it is entirely possible for the relationship to be much stronger in some non-linear function. Linear regression is most commonly used in the social sciences because in many cases it represents the "state of the art" in theory. Most theories put forth some direct relationship between two or more variables, without specifying the precise.limits of the relationship or the specific rates of change. As social and planning theories improve, non-linear equations may be formed to test those theories. For example, such an equation could eventually be develOped to specify limits and reverses to economies of scale when theories are advanced to explain those changes and sufficiently sensitive data are collected. The linear regression, with all its imperfections, remains the best tool at present. 58 The specific form of regression used for this study is the Statistical Package for the Social Sciences (SPSS) version 6.0 stepwise regression. This is a "packaged" com- puter program consisting of a basic fixed format with a number of options. One of these, termed pairwise deletion, allows variables with some missing observations. This was sometimes necessary where some local governments did not report specific expenditure items. Its use is indicated by the number of valid cases associated with each variable. In using data from the smallest as well as the largest communities, some trends are established which extend beyond the urban network. These trends will not be as sensitive to differences within the category of cities. Therefore this study uses two sets of regression equations. These two sets are not designed to answer the same questions, but are complementary in nature. They seek to determine whether trends which account for the differences encountered in the whole sample also account for differences between cities. I The use of a subsample of cities over 10,000 also allows the use of another variable. The average salary for common municipal functions is collected for units of govern- ment over 10,000 and can be tested as a determinant of expenditure levels. To do this, however, one cannot simply use dollar expenditure amounts. This is because with the same tasks and employees, higher salaries would result in higher expenditures. This would in effect use the same 59 variation on both sides of the equation. To get around this, the “city" subsample uses per capita employees for each function rather than dollar amounts. This should indicate changes in gross service levels, rather than just the result of increased cost per employee. It should be noted that, like the larger sample, the "city” subsample does not differentiate between categor- ies. Both central cities and suburban cities are included. This eliminates complications arising from the classifica- tion of cities. For example, several of the suburban cities are larger than the smallest of the central cities. Thus this sample is more likely to indicate differences due to factors inherent in the urban structure and population itself. Group Means and Analysis of Variance Not all data describing cities may be entered into regression equations. In some cases, such as with measures of income, variables are too closely related to be used in the same equations. In other cases, the variables within the equation give an inadequate picture of the differences which exist between and within categories. The group mean breakdown is one device for portray- ing differences between categories. The form used here displays the group mean, standard deviation and the same information for the sample as a whole. The standard devi- ation gives a measure of dispersion compared to the mean. 60 This gives the researcher some indication of how homogeneous the group is with respect to that variable. For example, the mean per capita income of the central city group is approximately $3500, with a standard deviation of $526. This suggests a fairly homogeneous group with respect to income. The mean for urban townships is $5,086, but the standard deviation is $2,013, indicating a much wider dis- persion in this group. In the aggregate, such information may help answer the question of how accurate and useful the classification system is. In the example of income, the group standard deviations are below the standard deviation for the whole sample with the exception of urban townships and suburban cities. This suggest a possible need to further refine those categories. A further test of the homogeneity of groups is the analysis of variance (ANOVA). ANOVA techniques are based on the concept that any result may be achieved by chance, and the probability of that occurring is measurable. The ANOVA measures the dispersion (variance) both within and between groups and compares the two. The test hypothesis is that there is no difference between the groups, and that individual differences are random. The results are given in the form.of the probability of the groups being the same. For example, a probability of .05 means that the results could be achieved purely by chance 5 percent of the time. Five percent is a commonly accepted level for statistical significance. To say that a result is 61 statistically significant means that the result is not likely to be achieved by chance, not that there is anything necessarily important about the results. Confusing statis- tical significance with importance is a very common error in analysis. Where it is used in this study, the ANOVA merely indicates group differences as they are constructed. This is a test of the classification system. There are other possibilities; for example, variance of the groups may be "pooled" to test for urban-suburban differences. This would require a separate underlying theory stating which categor- ies should be combined and is outside the scope of this study. If a specific type of difference is suspected (which is not the case here), a ”one-way" test of variance may be made to determine not whether there is simply a difference, but whether one category is always larger than another. A test theory or hypothesis of this sort might be that suburbs have higher per capita incomes than central cities. In this case we do not need to test whether or not suburban incomes are lower than central cities, hence the test is "one-way." Description of Variables This section presents a list of variables used in the analysis, together with a preliminary hypothesis of the suspected effect of the variable. First listed are the 62 independent and dependent variables included in the regres- sions, then those variables used only for breakdowns and analysis of variance. Independent Variables 1. Value density--Two different measures of value density are used in this study. The first of these, density of business value, is a measure of the amount of industrial and commercial assessed valu- ation per square mile. This is used in the whole sample regression. The second measure is the total assessed valuation per square mile. The inclusion of value and value density is a relatively recent innovation in cross-section studies, most because until very recently no reliable information has been available on valuation for minor ciVil divisions. These two measures act as indicators of activity density, a concept designed to indicate the amount of activity in an area requiring municipal services. Population density is a similar measure, but may understate the services required for a downtown area which has a great deal of activity but a low per- manent population. One part of the hypothesis is that businesses may require specialized services not needed in residential areas. Population density--As in the case of activity density, pOpulation density is a measure of service 63 needs. In largely undifferentiated areas and over large regions, population density should be a reasonably good measure of activity and required services. However, it does not accurately reflect structural differences between communities, such as sparse population due to lack of settlement and sparse population due to the presence of large industries. Aged and deteriorated housing and infrastructure-- Unfortunately, there has never been develOped a comprehensive and reliable measure of a community's age and condition. The best currently available measure is the age of housing stock. The inference is that the age of the housing stock will be closely related to the age of roads, utilities and similar facilities, and fairly closely related to business growth._ The specific measure used in this study is the percent of housing in the community constructed prior to 1940. The hypothesis linking age to municipal expenditures is that cities will incur higher maintenance and operating costs as a result of aging structures. Also, many buildings and facilities are nearing the end of their useful life and will have to be recon- structed or replaced at much higher cost than new facilities constructed in lower density areas. 64 Crowded housing-~This variable uses the Bureau of Census definition of living in crowded conditions as those dwelling units with more than one person per room. The specific meaSure used is the percent of dwelling units in a community with more than one person per room. Crowded housing is seen as a measure of deterioration requiring higher mainte- nance and operating costs, and for local governments a measure of higher density which may create the need for more social services. Families and persons with low incomes--Two slightly different measures of low income were used. In the full sample regressions, a broader measure of all workers is used; this is the percent of employed persons earning less than $5,000 per year. In the subsample of cities, a more restricted measure to poverty-oriented services, is the percent of fami- lies below the federally-designated poverty level in 1970. Note that since welfare payments of any form are not included in the expenditure categories, this is not intended to measure direct costs due to poverty. Percent of professional, technical and managerial workers in the community labor force--Since the size of these occupational groups is a strong indicator of income level and ability to pay, there is an interest in discovering whether any higher 65 service costs are linked to this variable. In this context, the proportion of professional and techni- cal workers is not used as an indicator of city type or structure, as has been frequently done. It is used rather as a social class indicator. It should be noted that this variable lists workers by place of residence, not place of work; there is often considerable divergence between the two. Salary levels for common municipal functions--Where similar service levels are maintained, differences in salary levels should contribute significantly to cost differences. Data on salary levels for common functions was available only for cities over 10,000; it is therefore used only in the city subsample regressions. There are two common hypotheses offered for salary differentials, each with considerably different implications. The first is that higher salaries are required for some cities to compensate for living costs and conditions. This would indi- cate that many central cities would have to pay higher salaries. The second hypothesis is that higher salaries are a community or collective luxury item and are offered by communities rich enough to afford them. This should indicate a high correla- tion with cities with high valuations (per capita) and high per capita incomes. 66 There is no strong theory at present which expresses the direction and magnitude of net salary effects on expen- ditures. However, the widespread interest in this variable prompted its inclusion. Cost Variables The cost or dependent variables selected are intended to cover the basic services common to most, if not all, local governments. They do not include such services as parks and recreation, libraries or certain specialized activities. Although it would certainly be desirable to include these within the study, consistent data are not available for enough of the study sample to permit their inclusion. In particular, parks and recreation expenditures would be a desirable variable, and it is often included in expenditure analyses. However, as such studies (including this one) do not include county expenditures which often provide services for outlying areas, the data would be badly distorted. In an attempt to examine the problem particular to cities, three variables are included in the category of services not common to all local governments. This includes sewers, health care and housing and urban develOpment. The category of sanitation was the least useful, with a total of six reported cases out of the sample total of 202. A likely explanation is that for most local governments this was included under the category of general operating expenses. 67 For variables used in the first regression set, the "cost” is the budgeted expenditure for that item as reported in the 1972 Census of Governments. This includes all current operating expenditures for that item. In the second set of regressions a derivative of cost is created by using the number of personnel per capita assigned to each function. While this has some advantages over straight budget expendi- tures, it also has some limitations. It excludes expendi- tures for specialized equipment and would be understated where a local government used contract personnel or a cost-sharing arrangement with another local government. GroupMean Variables In addition to those variables used as explanatory variables in the regression analysis, some variables have been used for the group mean breakdown where it was felt that they would shed some light on the structure of the sample. No breakdown is included for salaries, since this was available for only a portion of the sample. Variables added for the group means analysis are listed below: 1. Per capita income 2. Per capita state equalized valuation 3. Per capita residential valuation 4. percent industrial valuation 5. density of industrial valuation 6. density of residential valuation 7. population density 68 8. percent employed in manufacturing 9. local tax millage 10. ratio of total tax bill to community income 11. ratio of per capita income to per capita local tax 12. per capita local government charges For convenience, each regression variable, both independent and cost, has been assigned a short name which is used in the analysis section. The short names and full variable names are listed below: List of Independent Variable Names VDEN - total assessed valuation per square mile BUSVAL - business and industrial valuation per square mile POPDEN - population density OLDHOUS - percent of housing built before 1940 CROWD - percent of overcrowded housing LOWINC - percent of workers earning less than $5,000 per year POV - percent of families below the poverty level PROF - percent of professional, technical and managerial workers SALARY - average salary for common municipal functions List of Cost Variable Names POLICE - police and public safety FIRE - fire protection SEWER - sewer system operations and maintenance ROADS GENAD BLDGS DEBT HUD HEALTH 69 local roads and highway—related operations and maintenance general control and financial administration maintenance and operation Of government buildings local government debt service general (not allocated to specific categories) housing and urban renewal health services other than hospitals IV. ANALYSIS OF DATA Whole Sample Regression Analysis One of the factors complicating the analysis of expenditures is that independent variables are likely to have different effects on each cost variable. The same factors most important in explaining highway expenditures will not necessarily suffice for police. However, there may be some variables that will stand out for all costs. Table four presents a simplified summary of all regressions using the full sample. Table 4.--Regression summary for full sample Cost Variable Important Independents R2 TOTEXP BUSVAL .62 GENAD BUSVAL .10 GOVT BUSVAL, OLDHOUS .13 DEBT LOWINC .03 BLDGS BUSVAL .14 HUD LOWINC, BUSVAL .39 HEALTH BUSVAL, OLDHOUS .39 POLICE BUSVAL, OLDHOUS .61 FIRE BUSVAL, OLDHOUS .55 SEWER BUSVAL, OLDHOUS, PROF .40 ROADS BUSVAL, OLDHOUS, .17 70. 71 The summary of regressions serves to indicate the importance of BUSVAL for all costs except DEBT and HUD. The R2 for DEBT (.03) indicates that no variable was impor- tant, while HUD expenditures were reported for only 15 municipalities. Since this did not include variation for most of the sample, the structure of differences for HUD is likely to be different. The presence of BUSVAL is most important for TOTEXP, FIRE, and POLICE. In these cases a majority of variation is accounted for by the equations. In all cases BUSVAL has a large and positive value. The strength of the relationship between BUSVAL and most cost variables indicates clearly the large impact of the physical structure itself on the public cost of servic- ing the community. It should be noted that the actual measure is density of commercial value, not total value. Thus smaller communities with dense commercial areas (such as suburban malls) also seem to experience this. On the other hand, commercial value for townships would be diluted through the much larger area. It is important to note that BUSVAL explains by far the greatest variation in TOTEXP. This is an indication of the magnitude of its effects not only on basic services, but on all local government expenditures. Although this does not provide the precise key to the causes of expendi- ture variation it suggests that commercial and industrial value density is the key to understanding many expenditures. 72. The second variable of real importance is that of aged housing. The sample includes fairly large newer com- munities as well as older ones. The importance of OLDHOUS is that it distinguishes expenditures due to age from those due to factors such as poverty and presence of industry. A limiting factor is that OLDHOUS is to some extent inter- correlated with other explanatory variables. Without the presence of BUSVAL, for example, OLDHOUS should have a slightly greater effect. However, the simple correlation between OLDHOUS and BUSVAL is less than .04, so this effect should not be great. The correlation between OLDHOUS and LOWINC is quite high (.53); this may explain why the effect of LOWINC appears very weak when theory seems to indicate a stronger link. Part of this difference lies in the fact that, unlike most studies, this sample includes a number of rural areas. These contain a fair proportion of low-income population, which also has low service expenditures. If rural areas were excluded the effect of LOWINC should be greater. Although OLDHOUS is not highly correlated with BUSVAL, it seems to be highly associated with central cities. The correlation between OLDHOUS and sanitation expenditures is .81. Sanitation as a separate municipal expense was reported by only six large cities and was there- fore not used in the regression analysis. The high corre- lation between OLDHOUS and central cities may explain the effect of OLDHOUS on some variables not directly related 73 to structure, such as HEALTH. This is also an expenditure associated with central cities, as counties are likely to assume that function outside of urban areas. It is interesting to note that OLDHOUS is third in importance for TOTEXP and is significant for GENAD. This finding indicates that OLDHOUS really has a more specific relationship to cost than does BUSVAL. Table 5.--Regression for dependent variable TOTEXP F Variable B Std Error B Significance Beta -4 -6 Busval .l4x10 .9148x10 249.8 .7289 .000 Prof .1656 .0536 9.536 .1719 .002 Oldhous .0181 .0292 11.28 .1825 .001 Lowinc -.1681 .8406 3.950 -.1092 .048 Crowd .0962 .0548 3.074 .0803 .081 (Constant) -l6.68 21.35 .6105 .436 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .7676 .5893 .7676 286.98 .000 Prof .7760 .6022 .2710 150.65 .000 Oldhous .7845 .6154 .0747 105.64 0 Lowinc .7886 .6220 -.0177 81.04 .000 Crowd .7923 .6278 .0759 66.13 .000 The equation for TOTEXP presents one of the stron- gest relationships in the study. The final equation accounts for 62 percent of the variation, with BUSVAL 74 accounting for almost 59 percent. Of the variables, only CROWD is not significant at the 95 percent confidence level. The level of total expenditures is in a sense the starting point from which individual functions are disaggre- gated. The analysis of TOTEXP serves to indicate which factors exert the greatest influence on the whole budget. It is therefore much more general than the effects of speci- fic functions. The implications are wider but the causes not as specific. It is interesting to note that the sign of LOWINC is negative. A common theory is that city expenditures are higher because of increased services to the poor, but this may not be the case. Alternatives possible are that services are diverted from other areas to the poor, or that the influence of the poor on revenues (and thus indirectly on expenditures) is greater than the direct pOsitive influence on expenditures. One test of this is the influence of LOWINC on other costs. A uniformly negative sign would indicate the revenue effect; varied signs would suggest a diversion effect. Expenditures for general control and financial admin- istration (GENAD) are a more specific cost associated with administering the city. Here most of the variation remains unexplained. Of the independent variables, only BUSVAL has a statistically significant relationship. 75 Table 6.--Regression for dependent variable GENAD F variable B Std Error B Significance Beta Busval .7900x10'6 .2014x10'6 15.37 .2857 .000 Iowinc .0260 .0186 1.955 .1214 .164 Prof .0095 .0118 .6586 .0713 .418 Crowd -.0031 .0120 .0681 -.0188 .794 Oldhous .0014 .0064 .0520 .0195 .820 (Constant) 6.651 4.703 2.000 .159 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .3103 .0963 .3103 20.46 .000 Lowinc .3240 .1050 .1194 11.20 .000 Prof .3295 .1085 .0690 7.715 .000 Crowd .3301 .1090 .0035 5.780 .000 Oldhous .3305 .1092 .0613 4.611 .001 The lack of significance of any other than BUSVAL may reflect several alternative explanations. (hue is that local governments may have a great deal of difficulty in distinguishing costs that belong in this category. Many small governments may lump part of this with unallocated COBtB (GOVT). GENAD seems to indicate factors which would generate the most ”paperwork” for local governments, given variations in government type and responsibilities. One inference is that cities with more programs (of any type) would experi- ence higher total administrative costs, hence higher GENAD expenditures. In this instance, BUSVAL would be acting as 76 a surrogate for larger city size. The fairly high positive correlation between TOTEXP and GENAD (.42) lends some sup- port to this hypothesis. One statistical explanation for the poor performance of all variables for GENAD lies in the strong relationships between the independent variables. If, for example, OLDHOUS and LOWINC are closely related in terms of their own effect on GENAD and their effects on GENAD work in Opposite direc- tions, they would to some extent cancel out, leaving a minimum of explanation. This theoretical issue differs from the purely statistical problem of multicollinearity. A final possible explanation is that GENAD costs may reflect salary differences. Salary does not seem to have a linear relationship to city wealth or structure. The relationship is a more complex one where both rich and poor cities have high salaries. Since salary would be the major component of GENAD costs, this may be the most important factor. Neither expenditures for DEBT nor GOVT show a signi- ficant relationship to the independent variables, although in GOVT the variable BUSVAL is significant and explains some variation. The respective coefficients (R2) of .03 and .13 indicate that none of the variables contribute a great deal. The fact that the variation of DEBT is unexplained by major factors in city structure may indicate that 77 Table 7.--Regression for dependent variable GOVT F variable B Std Error B Significance Beta Busval .8885x10-6 .2317xio'6 . 14.70 .2947 .000 Oldhous .0196 .0074 7.031 .2402 .009 Lowinc -.0152 .0214 .5058 -.0652 .478 Prof .0060 .0135 .1950 .0409 .659 Crowd .002 .0138 .0234 .0116 .878 (Constant) 1.617 5.410 .0893 .765 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .3074 .0945 .3074 17.54 .000 Oldhous .3596 .1293 .1978 12.40 .000 Lowinc .3654 .1335 .0707 8.529 .000 Prof .3667 .1344 .0192 6.408 .000 Crowd .3668 .1345 .0196 5.101 .000 suburbs require, and are getting, higher per capita capital investment. That suburbs require this is not unexpected (Netzer, 1969). The negative relationships of LOWINC to DEBT and GOVT may indicate the situation where most programs dealing directly with the poor are handled at the county or state level with minimal effort from local governments. It was for that reason that welfare-related costs were excluded from the analysis. The operation and maintenance of government build- ings (BLDGS) is the last "general” cost considered. All others are more specific functions. Again, the degree of explanation is quite low and BUSVAL is the only significant 78 Table 8.--Regression for dependent variable DEBT F variable 8 Std Error B Significance Beta Lowinc -.0209 .0197 . 1.125 -.1302 -6 -6 .291 Busval .2306x10 .2132x10 1.170 .1113 .282 Crowd -.0098 .0127 .5883 -.0784 .445 Prof -.0093 .0125 .5594 -.0929 .456 Oldhous -.0045 .0068 .4444 -.0808 .507 (Constant) 15.09 4.977 9.194 .003 Summary Table variable Multiple R R Square Simple R Overall F Significance Lowinc .1327 .0176 -.l327 1.90 .171 Busval .1572 .0247 .0725 1.33 .269 Crowd .1667 .0278 -.0719 .99 .400 Prof .1756 .0308 .0487 .81 .516 Oldhous .1872 .0350 .1060 .74 .594 factor. The correlation between BLDGS and GOVT is quite high (.47), although perhaps not as high as would be expected. It is noteworthy that the correlation with LOWINC is again negative, although the standard error is rather high. This again leads to the suggestion that direct services to the poor do not create large increases in local government expenditures. The relationship between BLDGS and BUSVAL may be explained by the higher cost of maintaining structures in higher density downtown areas. These same conditions are much less likely to be found in the lower-density town- ships. With the possible exception of highways, townships 79 Table 9.--Regression for dependent variable BLDGS F variable B Std Error B Significance Beta Busval .1093x10'6 .2858x10-6. 14.63 .3413 .000 Crowd -.0013 .0017 .6395 -.0703 .425 Lowinc -.0031 .0026 1.412 -.1264 .237 Oldhous .0012 .0009 1.350 .1430 .176 Prof .0004 .0016 .0879 .0319 .767 (Constant) 1.045 .6674 2.454 .120 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .3376 .1140 .3376 16.085 .000 Crowd, .3506 .1229 -.0710 8.690 .000 bowinc .3561 .1268 -.0477 5.956 .001 Oldhous .3735 .1395 .0709 4.944 .001 Prof .3743 .1401 .1144 3.943 .002 should also have fewer functions, hence a much smaller physical plant. Higher maintenance cost would be expected for both rich and poor cities. This could explain the non-significant relationship with factors such as LOWINC. However, one might have expected a stronger relationship with OLDHOUS, which should be a surrogate for older city structures. Expenditures for HEALTH and HUD are purely urban variables; only 27 cities reported HEALTH expenditures and only 15 HUD. The purpose of their inclusion is to examine those urban variables in the same sample and format as more basic expenditures. This should provide additional 80 Table 10.--Regression for dependent variable HUD F Variable B Std Error B Significance Beta Lowinc .0910 .0538 . 2.855 .5528 -6 -6 .125 Busval .5093x10 .5826x10 .7641 .2398 .405 Prof -.0133 .0341 .1519 -.1291 .706 Oldhous -.00757 .0186 .1657 -.l3l6 .693 Crowd .0065 .0349 .0353 .0512 .855 (Constant) 1.813 13.58 .0178 .897 Summary Table Variable Multiple R R Square Simple R Overall F Significance Lowinc .5726 .3278 .5726 6.341 .026 Busval .6101 .3722 .2588 3.558 .061 Prof .6164 .3800 -.2927 2.247 .140 Oldhous .6267 .3928 .2338 1.617 .245 Crowd .6286 .3951 .1860 1.176 .391 information both on the functioning of the structural variables and the urban portion of the sample. In both cases a substantial portion of the varia- tion (39 percent) is explained by the independent variables. In both cases most of the variation is accounted for by one‘ variable; BUSVAL in the case of HEALTH, LOWINC in the case of HUD. Of the variables, only BUSVAL has a statistically significant relationship. Given the small number of observations, historical developments should have more influence than any of the structural variables on HUD. Expenditure levels would differ, for example because of changes in federal funding 81 Table ll.--Regression for dependent variable HEALTH F variable B Std Error B Significance Beta -6 -6 Busval .1488x10 .4493x10 . 10.96 .5559 O 003 Oldhous .0017 .0013 1.789 .2476 O 194 Prof .0015 .0024 .4035 -.1201 .532 (Constant) -.6145 .8632 .0050 .944 variable Partial Tolerance Significance Lowinc .020 .631 .0083 Crowd .019 .910 .00843 Summary_Table variable Multiple R R Square Simple R Overall F Significance Busval .5401 .2917 .5401 10.29 .004 Oldhous .6197 .3841 .3238 7.484 .003 Prof .6282 .3947 -.1188 4.999 .008 formulas and the timing of development. HUD reflects only the city's share of housing and urban development expendi- tures. It is difficult to tell in this case whether BUSVAL is acting as a surrogate for “central city" or whether the factor itself is influencing the level of HUD; in either case the effect is not large or statistically significant. In the case of HEALTH, BUSVAL cannot be regarded as the direct causal factor. Rather it seems to represent those areas which developed local health care systems before the counties and state stepped in. The operation of public hospitals is not included in HEALTH, but the two 82 would to some degree be related so that older areas with local public hospitals should also have higher HEALTH expen- ditures. These functions illustrate that, even within uniform state requirements, great variation exists in certain expen- ditures. This variation cannot be accounted for by variables dealing only with structure or city ”type." The explanation probably would not be improved by employment or other simi- lar statistical measures. As indicated by Liebert(1974) and others, political and historical factors also play a large role in present expenditure levels. Table 12.--Regression for dependent variable POLICE F variable B Std Error B Significance Beta -5 -6 Busval .1571x10 .1332x10 139.1 .6743 0 Oldhous .020 .00425 23.15 .3236 .000 Prof .027 .0078 12.00 .2388 .001 Lowinc -.0209 .0123 2.87 -.1155 .092 Crowd .0107 .0079 1.819 .0764 .180 (Constant) -5.823 3.110 3.506 .063 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .7305 .5337 .7305 155.68 0 Oldhous .7463 .5569 .1795 84.86 .000 Prof .7767 .6033 .2649 67.92 0 Bovine .7814 .6107 .0143 52.16 .000 Crowd .7848 .6160 .0557 42.34 0 83 Table 13.--Regression for dependent variable FIRE F variable B Std Error B Significance Beta -6 -7 Busval .9188x10 .8946x10 . 105.4 .6177 0 Oldhous .0136 .0028 22.95 .3393 .000 Prof .0134 .0051 6.918 .1864 .010 Iowinc .0075 .0083 .8181 .0652 .367 (Constant) -4.619 1.971 5.488 .021 Summary Table variable Multiple R R Square Simple R Overall F Significance Busval .6747 .4553 .6747 114.5 0 Oldhous .7329 .5371 .3109 78.91 .000 Prof .7464 .5571 .1273 56.62 0 Lowinc .7482 .5598 .2108 42.61 .000 POLICE and FIRE are perhaps the most ubiquitous of local government functions. Yet even here some local governments in the sample reported no expenditures in these categories. However, some very good trends can be seen in the equations for these functions. It is not surprising that BUSVAL figures prominently in both expenditure equations. In FIRE, BUSVAL accounts for 45 percent of the variation and 53 percent in the case of POLICE. The large Beta values in both cases indicate that BUSVAL has a very strong effect on both costs. Both BUSVAL and OLDHOUS directly describe conditions that lead to 84 demands for more service. As would be expected, the age of structures exerts a stronger influence on FIRE than on POLICE. Again we see the interesting relationship of poverty- oriented variables. In both equations, neither CROWD nor LOWINC is statistically significant. It is part of the conventional wisdom that low income areas breed crime and require greater expenditures. But this is complicated where poor people cannot effectively demand more services, and suburban areas where the fear of crime may generate higher expenditures than crime itself. LOWINC has a negative sign that suggests this. Further evidence is given by the positive and signi- ficant relationship with PROF. As may be seen from the group means, PROF is a fairly good proxy for suburban areas. The more dense suburban areas also seem to have higher POLICE and FIRE expenditures (density being measured by value, not population). Since both POLICE and FIRE are labor-intensive ser- vices, salary should be an important factor. A test of the "cities only” regressions will be to determine if this is the case. This argument is one directly affecting larger cities such as New York, where high wage levels are alleged to be a major factor in the fiscal crisis. It would be interesting to see whether this trend extends to smaller cities. 85 Table 14.--Regression for dependent variable SEWER F Variable B Std Error B Significance Beta Busval .5118x10—5 .7788x10-6 43.18 .5050 .000 Prof .1362 .0456 8.897 .2768 .004 Oldhous .0830 .0248 11.14 .3023 .001 Lowinc .0794 .0720 1.217 -.1011 .272 Crowd .0345 .0467 .5455 .0563 .462 (Constant) -29.26 18.18 2.590 .110 Summary Table Variable Multiple R R Square Simple R Overall F Significance Busval .5691 .3239 .5691 55.58 .000 Prof .5915 .3498 .2759 30.94 .000 Oldhous .6330 .4007 .1413 25.40 .000 Lowinc .6375 .4065 -.0183 19.34 .000 Crowd .6398 .4093 .0162 15.52 .000 SEWER expenditures reveal several conflicting trends. The most important single variable is BUSVAL, indicating greater per capita costs in denser areas. This tends to refute arguments for economies of scale. This is not to say that economies of scale do not exist, but rather that they are overpowered by such factors as increased pollution from commercial and industrial users. The positive correlation with OLDHOUS may mean higher operating and maintenance costs associated with older equipment, or simply that many larger cities tend to 86 be older. However, many rural localities with very low or no sewer expenditures also tend to be old. This would tend to weaken the relationship. PROF also shows up as a fairly strong positive relationship with SEWER. This suggests higher costs in sub- urban areas, an effect well described in theory and case studies. However, this trend is not nearly as strong as BUSVAL, indicating perhaps that smaller suburban cities can operate this service most effectively. Table 15.--Regression for dependent variable ROADS F variable B Std Error B Significance Beta -6 -6 Busval .5962x10 .260x10 22.36 .3450 .000 Prof .0114 .0073 2.397 .1364 .123 Oldhous .0084 .0040 4.381 .1799 .038 Lowinc -.0119 .0116 1.047 -.0890 .307 Crowd -.0055 .0075 .5324 -.0528 .467 (Constant) 2.893 2.943 .9662 .327 Summary Table Variable Multiple R R Square Simple R Overall F Significance Busval .3688 .1360 .3688 28.03 .000 Prof .3838 .1473 .1803 15.28 .000 Oldhous .4065 .1652 .0803 11.61 .000 Lowinc .4133 .1708 -.0367 9.01 .000 Crowd .4164 .1733 -.0714 7.29 .000 Weicher (1970) complained that highway expenditures were poorly explained in relation to other expenditure 87 variables. Here again this is the case. The implication from the equation is that there is some tendency (BUSVAL) for the denser areas to have higher road maintenance costs. However, the cause of this is not so clear. Cities in Michigan tend to take responsibility for most roads within their borders, even fairly major ones. In outlying juris- dictions these same roads are handled by county or state highway departments. The variable ROADS may describe who pays, not necessarily what it costs. Conventional theory suggests that suburban areas experience higher highway cost as a result of commuting. The positive sign of PROF offers some support for this, but the effect is not very large or statistically significant. Since the cities also have to deal with the commuters they will also bear a high cost for highways. It seems likely that communities with a high value for PROF also have a high percentage of commuters. These governments would have higher highways costs than communities with fewer commuters. Cities Regression Analysis The regression analysis for cities is a more restricted analysis designed to show qualitative differences among the cities. Certain trends and the importance of some variables in the whole sample may have been due to the inclusion of townships. To the extent possible, this is tested by a regression which excludes smaller communities and townships. 88 By using cost in terms of personnel, this analysis also can test the effect of salaries on cost. This is only possible with a greatly restricted set of variables. For example, TOTEXP cannot be meaningfully translated into per- sonnel without controlling for type of government. This is beyond the scope of this study. A summary of regression results is given in Table 16. Table l6.--Regression summary for cities Cost variable Important Independents R2 Genad Oldhous, Vden .31 Police Pov, Salary, Oldhous .85 Fire Oldhous, Vden .79 Sewers Oldhous, Pov .63 Roads Oldhous, Salary, Pov .35 The regression summary indicates important differ- ences from the full sample. Changes in coefficients are to be expected for several reasons. The outstanding effect is that among cities, value density is no longer the dominant factor. In most cases, age replaces it. If a city's fiscal health may be measured in terms of rising cost, then age may be a very useful indicator. Note especially that OLDHOUS is the most important variable for all structure-related costs (FIRE, ROADS, and SEWERS) as well as GENAD. This is quite different from the results of the full sample regressions and more in line with recent observations of cities. 89 POV also plays a more important role than LOWINC did in the full sample results. This may be because the differ- ence in POV is relatively greater than full sample differ- ences in LOWINC. It is also possible that POV is a more sensitive measure of those conditions which create increased service burdens. The question of the effects of SALARY is left unanswered. SALARY was significant enough to be entered into only two equations. In all cases SALARY was entered in the original equation. Where it had no measurable effect the program dropped the variable and it does not appear in the final equation. SALARY had a significant relationship only to POLICE expenditures. In ROADS, the relationship is positive but not significant as a result of the large standard error. *The relationship is positive, meaning that high expenditure governments tended to have high salaries. This does not indicate the cause of either variation, however, although other factors in the POLICE equation tend to support the “combat pay” hypothesis. This hypothesis is briefly that central cities must pay higher salaries than their suburban counterparts to compensate for poor working and living conditions. Population density was also entered into this set of regressions as a comparison to value density. In general, value density proved the better explainer. In the case of FIRE, for example, VDEN increased total explanation (R2) 90 from .73 to .79 when it was substituted for POPDEN. Only in the case of POLICE did POPDEN prove more significant. Neither density figure added to the equation; POPDEN was slightly more significant but interestingly, the sign is negative; this indicates more police per capita in less dense cities, although the quantitative effect appears negligible. Table l7.--Regression for dependent variable GENAD F variable B Std Error B Significance Beta Oldhous .9910x1o'5 .3710x10"5 7.1321 .4716 _10 _11 .014 Vden .1712x10 .999lx10 2.9371 .3026 _3 _3 .101 (Constant) .3563x10 .2698x10 1.7441 variable Partial T01erance Significance Pov .0165 .3632 .0057 .940 Summary Table variable Multiple R R Square Simple R Overall F Significance Oldhous .4713 .2222 .4713 6.570 .017 .5602 .3138 .3022 5.030 .016 Only the variables OLDHOUS and VDEN had a suffic- iently significant relationship with GENAD to enable inclusion in the final equation. Of these, only OLDHOUS is significant at the 95 percent confidence level. These two together account for less than a third of the total varia- tion. The clear indication is that more specific variables are needed to determine factors affecting GENAD. The 91 variables used point in the direction of older, denser cities incurring higher GENAD costs, but the links are very indirect. This may reflect the administrative costs of such programs as urban renewal. This type of program, requiring state and federal coordination (i.e., higher administrative cost) would not be as prevalent in newer suburban cities. Table 18.--Regression for dependent variable POLICE 8 Variable B Std Error B Significance Beta Pov .8709x10"4 .2965x10'4 8.6230 .4094 _5 _6 .008 Salary .1667x10 .2623x10 40.413 .5806 _4 _5 .000 Oldhous .1014x10 .4595x10 4.8725 .3073 _7 _7 .039 Popden -.4300x10 .3736x10 1.3248 -.1051 _3 _3 .263 (Constant) -.4445x10 .2462x10 3.2570 .086 Summary Table variable Multiple R R Square Simple R Overall F Significance Pov .7109 .5055 .7109 23.511 .000 Salary .8999 .8099 .6346 46.874 .000 Oldhous .9212 .8486 .7061 39.254 0 Popden .9263 .8580 .1768 30.227 .000 POLICE provides the best case of the relationship between SALARY and cost variables, and the best support for the ”combat pay" argument. POV has a slightly greater addition to the R2 but the higher Beta value for SALARY indicates a stronger effect. POV, OLDHOUS, and SALARY all have positive correlations with POLICE. POPDEN is negative but not significant. It should be noted that high 92 population density is not (in this sample) synonymous with larger or older cities. The simple correlation of POPDEN with size (pOpulation) is .15, that with OLDHOUS is .06. Even the relationship of POPDEN to VDEN is .80, indicating that while very closely related even these two are somewhat different. The high correlation between POV and OLDHOUS (.80) undoubtedly cause some problems in the equation. The corre- lation between LOWINC and OLDHOUS in the full sample is considerably lower (.53), suggesting that when the rural poor are taken out, poverty does mean the older central cities. But this association does not seem to show up strongly in other variables. The relationship of POV to POLICE cost clearly associates poverty with the perception of crime in cities. (Since POV is more uniformly measured than crime, the rela- tionship might prove to be strong enough and consistent enough to use in an aid formula similar to that employed by the federal government for Community Development allocations. The equation for FIRE provides an excellent example of the close relationship between a cost variable and the suspected structural causes of cost variation. Both OLDHOUS and VDEN are positive and significant; OLDHOUS has by far the greater effect. SALARY did not have a sufficiently high correlation to remain in the equation. POV is positive, but not significant, probably because of its high correlation to OLDHOUS. 93 Table l9.--Regression for dependent variable FIRE F Variable B Std Error B Significance Beta -4 -5 Oldhous .2644x10 .5348x10 ~ 24.422 .8095 _10 -11 .000 Vden .2262xlO .8816x10 6.5853 .2572 _4 _4 .018 Pov .1105x10 .3460x10 .1019 .0525 _3 _3 .753 (Constant) .3675x10 .2383x10 2.376 .138 Summarerable variable Multiple R R Square Simple R Overall F Significance 01dhous .8509 .7240 .8509 60.355 0 Vden .8903 .7927 .2612 42.076 .000 Pov .8909 .7937 .7176 26.939 0 The effect of aged structure could affect FIRE costs in several ways. 01d structures and lack of fireproof materials would make fires more likely. Lack of sprinkler systems and other automatic firefighting systems would tend to make fires worse when they occur. Density would also tend to make fires more danger- ous and costly to fight. But the greatest cost increase could well come from the specialized equipment and training needed to fight commercial and industrial fires. With POLICE, this would be analogous to special techniques and equipment needed to patrol large deserted commercial and industrial areas at night. The effect of business density was more clearly shown in the full sample, where the range of density was greater. 94 Table 20.-Regression for dependent variable SEWER F variable 8 Std Error B Significance Beta Oldhous .6175x10'5 .3147x10-5' 3.8488 .4321 _4 _4 .064 Pov .3573x10 .2011x10 3.1570 .3881 _6 _6 .091 Salary .1556x10 .2132x10 .53258 .1251 _11 _11 .474 Vden .1538x10 .6609x10 .054192 .0399 _3 _3 .818 (Constant) -.3126x10 .1690x10 3.7729 .066 Summary Table variable Multiple R R Square Simple R Overall F Significance Oldhous .7570 .5731 .7570 30.886 .000 Pov .7958 .6333 .7496 19.001 .000 Salary .8096 .6555 .2574 13.323 .000 Vden .8102 .6565 .1532 9.556 .000 Value density may not work as well for among-city variation, since older areas often experience declining assessments. This produces an apparent decline in value density while physical density remains high. Explanation of SEWER in the city sample is consider- ably improved over the full sample. As with FIRE, OLDHOUS explains by far the greatest portion of variation. All variables have high standard errors; none is significant at the 95 percent confidence level and only OLDHOUS and POV at the 90 percent. SALARY and VDEN are both well below accept- able levels of significance. With Beta values of .43 and .38, OLDHOUS and Pov have a strong impact on SEWER costs. 95 The impact of OLDHOUS tends to confirm the associa- tion between aged housing and aged or obsolescent infras- tructure. The effect seems to be consistent throughout a number of variables when considered only among cities. The effect of VDEN seems to be subject to the same limitation given under FIRE. Although declining assessments would affect only part of the city, this could be enough to lower the aggregate density. This would distort the corre- lation and so its impact in the equation. Table 21.--Regression for dependent variable ROADS F Variable B Std Error B Significance Beta -5 Oldhous .5573x10-5 .6656x10 .7010 .2545 .412 Salary .4895x1o'6 .4509x10’6 1.178 .2569 _4 _4 .291 Pov .3903x10 .4253x10 .8421 .2766 -11 -11 '370 Vden -.l445x10 .1397x10 .01069 -.0245 _3 -3 .919 (Constant) -.1189x10 .3403x10 .1221 .730 Summarnyable variable Multiple R R Square Simple R Overall F Significance Oldhous .5093 .2594 .5093 8.056 .009 Sa1ary .5641 .3182 .3106 5.135 .015 Pov .5878 .3455 .5075 3.696 .028 Vden .5881 .3459 .1611 2.644 .064 As in the full sample, the equation for ROADS falls far short of an adequate explanation of variation. None of the relationships is statistically significant, although 96 together they account for slightly over a third of the variation in ROADS. It may be logical to accept the position that ROADS costs are high in both the central city and the suburban sources of communters. If this is so, then apparently none of the variables used are sufficiently correlated with com- muters to produce a significant result. PROF to some degree had this quality in the full sample, but the same effect would not be expected to operate with a sample including only central and suburban cities. Here the cost for all should be fairly high, but PROF would vary. This would again produce a high standard error and non-significant result. Selected Expenditures By Category This section presents the breakdown by category of four selected costs: POLICE, FIRE, GENAD, and GOVT. The regression analysis is intended to provide trends and explanation of variation irrespective of classification. The group means and analysis of variance are used to deter- mine whether in fact the differences are really very impor- tant. If they are, than distinguishing between these groups may be important for both policy making and interest group formation. The four variables chosen are used because they are: (1) important parts of the local government's budget, and (2) widely reported (they have relatively few missing 97 cases). As previously noted, local governments may have difficulty in distinguishing GENAD expenditures and report these instead under GOVT. Therefore both categories are included. TOTEXP is not used because the smaller local governments face millage restrictions which effectively! limit the total budget. FIRE and POLICE have the largest number of reported expenditures for specific functions. Other categories such as sewers, parks and recreation and health expenditures are not consistently reported by smaller local governments. Highway expenditures are widely reported but the results are distorted by county and state highway operations. The four categories selected were felt most likely to yield the most accurate general picture of expenditures by category. The group means for GOVT indicate a fair degree of consistency between cities and townships.7 Rural municipal- ities have a much higher expenditure. This suggests that their size is below the point where economies of scale occur; Beaton (1974) and Netzer (1969) believe this point to be between 4,000 and 10,000 while the average pOpulation of rural municipalities in this study is only 1,855. How- ever, the sample size is too small (4) to make conclusions with any degree of confidence. Both central and suburban cities seem to spend about the same per capita, although the standard deviation is rather high. Thus for GOVT the urban-suburban dichotomy does not exist, but is rather a city-township split. 98 Table 22.--Breakdown for GOVT expenditures Category Mean Std. Dev. No. of cases Central city 24.31 11.19 13 Urban township 4.14 . 4.04 15 Suburban municipality 22.49 23.39 28 Suburban township 5.81 9.05 60 Rural municipality 36.70 27.60 4 Rural township 5.55 3.75 50 Tbtal 10.48 14.74 170 ANOVA Sum of Squares Degrees of Freedom Mean Square Between Groups 12402.24 ( 5) 2480.44 Within Groups 24320.50 (164) 148.29 TOtal 36722.74 (169) F 3 16.72 SIG. 8 .0000 GENAD presents the first case where central cities .clearly spend more than their suburban counterparts. As noted before, GENAD is a "paperwork” variable. It seems that central cities spend a good deal more than suburban cities or any township categories for this type of service. In common with GOVT, the cost of administration for very small cities is extremely high. GENAD presents the type of difference which provided the theory for the sample construction. That is, these cost items confirm that there is a real difference between the categories and that the difference is large enough to be con- sidered important. So far though, townships do not seem to differ much. Township administrative costs do not increase 99 Table 23.-~Breakdown for GENAD expenditures Category Mean Std. Dev. v No. of Cases Central city 25.74 6.87 13 Urban township 12.22 . 5.96 17 Suburban municipality 19.99 9.11 28 Suburban township 12.13 12.96 76 Rural municipality 52.20 51.10 4 Rural township 9.17 4.69 56 Tbtal 14.16 13.52 194 ANOVA Sum of Squares Degrees of Freedom Mean Square Between Groups 10253.3938 ( 5) 2050.6788 Within Groups 25042.5193 (188) 133.2049 Total 35295.9131 (193) F I 15.3949 SIG. - .000 very much as urbanization increases. This is probably due to the revenue limitations of township government. With the exception of rural municipalities, the standard deviation of GENAD is not unacceptably large. Thus the grouping (categories) seems to be a logical division. POLICE and FIRE are among the very few costs which can be described as "common functions" for all local govern- ments. Even in these categories quite a few localities did not report any expenditures. Surprisingly, many of those not reporting were in suburban rather than rural categories. These line functions of government provide increased differentiation between central cities and the suburbs. For POLICE the per capita expenditures for suburban cities fell 100 Table 24.--Breakdown of FIRE expenditures Category Mean Std. Dev. No. of Cases Central city 23.73 5.54 13 Urban township 5.21 4.27 17 Suburban municipality 9.69 6.54 23 Suburban township 3.79 3.05 - Rural municipality - - - Rural township 2.05 1.17 34 Total 6.38 7.27 139 ANOVA Sum of Squares Degrees of Freedom Mean Square Between Groups 5174.9964 ( 4) 1293.7491 Within Groups 2126.9972 (134) 15.8731 Total 7301.9936 (138) F = 81.5057 SIG. - .0000 Table 25.--Breakdown for POLICE expenditures Category Mean Std. Dev.' No. of Cases Central city 31.39 7.02 13 Urban township 6.72 6.13 17 Suburban municipality 18.87 10.46 28 Suburban township 2.30 2.80 49 Rural municipality 11.20 6.48 3 Rural township 1.35 1.31 28 Total 8.95 11.41 138 ANOVA Sum of Squares Degrees of Freedom Mean Square Between Groups 13186.3902 ( 5) 2637.2780 Within Groups 4659.6799 (132) 35.3006 Total 17846.0701 (137) F = 74.7091 SIG. = O 101 midway between central cities and townships, while for FIRE suburban cities are much closer to township expenditures. This coincides very well with the regression trends estab- lished for FIRE with BUSVAL and OLDHOUS. POLICE cost in townships rises in an almost linear fashion as urbanization increases. POLICE costs are again much higher for rural municipalities than townships. No rural municipalities reported FIRE expenditures, probably because these are either assumed by the township or provided by volunteer companies. On the whole, the cost breakdowns seem to provide some support for the sample categories. The statutory limits of townships place a low ceiling on township expendi— tures but there is still some urban-suburban-rural variation. The difference between central and suburban cities is much clearer. Groups Means and Analysis of Variance The purpose of presenting group means and analysis of variance is to test the sample selection and categori- zation and to determine what real differences exist between the categories. Variables presented fall into the general classes of structure, employment, income and wealth, and tax burden. These classes include some variables not used in the regression analysis. In testing the sample categories, the primary question is whether or not the categories are discrete. 102 The analysis of variance indicates whether the variance between groups is greater than the variance within groups. This yields the chance, as given by significance, that the result could have been achieved by random distribution of variation. If categories are discrete for a number of varied elements, it may be concluded that using this technique for differentiating among local governments is a fruitful approach, and one that will provide an improved base for urban policy decisions. Table 26.--Category mean populations Category Population Central city 81,180 Urban township 18,851 Suburban municipality 18,878 Suburban township 7,240 Rural municipality 1,855 Rural township .3,010 Group mean populations are presented as background for the sample structure. They depict the relationship of community size to the classification and provide some per- spective for population density. Of note are the mean pOpulations of the last two groups (rural). These fall within the size range where economies of scale are still said to occur, meaning that average costs should fall as size increases in this group. All of these communities are within normal commuting distance of the nearest central 103 city, and may therefore be described as direct competitors for residential and at least some business location. The physical structure of communities is measured by population density, value density and percent of indus— trial property value of total property. Most studies in the past have used employment rather than value. These two alternative measures will be compared. Table 27.--Population density by category Category Mean Std. Dev. No. of Cases Central city 4989 1387 13 Urban township 795 617 18 Suburban municipality 3436 2292 28 Suburban township 417 1174 77 Rural municipality 2041 227 4 Rural township 99 164 62 Total 1098 1906 202 ANOVA F value = 63.78 significance = .0000 Table 28.--Residential value density by category Category Mean Std. Dev. No. of Cases Central city 10100000 3790000 13 Urban township 2660000 3110000 18 Suburban municipality 10300000 7660000 28 Suburban township 938000 1930000 77 Rural municipality 3900000 1050000 4 Rural township 231000 4990000 62 Total 2820000 5070000 202 ANOVA F value = 52.01 significance = .0000 104 Table 29.--Industrial value density by category Category Mean Std. Dev. No. of Cases Central city 8140000 4860000 13 Urban township 619000 1340000 18 Suburban municipality 2690000 3580000 28 Suburban township 313000 790000 77 Rural municipality 725000 1060000 4 Rural township 40900 217000 62 Total 109000 2780000 202 ANOVA F value = 45.96 significance = .0000 The data on density follow the anticipated rural to urban pattern. One exception to this in all density tables is the lower density of urban townships. Two alternatives may explain this. First, since the sample criteria were somewhat arbitrary, some suburban townships may have entered the urban township category. Secondly, the urban townships may not be as highly urbanized as was originally thought. Industrial density in particular is much closer to the sub- urban township level than to the central city. This fits fairly well with the theory of urban townships as "near suburbs" that compete for residential value, but not so well 'for those who hold that ”near suburbs" are attracting industry. Central and suburban cities appear to be qualita- tively much different. Population density is much higher for the central cities, yet residential value density is 105 much lower. Industrial density is again higher in central cities. Unfortunately, an oversight in the study omitted commercial value density. This would have provided another valuable indicator of qualitative differences. Density does contribute to the understanding of urban townships. Their density is lower than any category of city, yet they are much larger in population and denser than suburban townships. Although industrial density is lower, the study is a static analysis and does not indicate how value density is changing. Table 30.--Industria1 as a percent of total value Category Mean (%) Std. Dev. No. of Cases Central city 32.6 13.3 13 Urban township 10.3 14.0 18 Suburban municipality 17.5 18.3 28 Suburban township 9.7 14.6 77 Rural municipality 8.4 11.1 4 Rural township 3.3 6.4 62 Total 10.3 14.9 202 ANOVA F value = 12.70 significance = .0000 Industrial as a proportion of total property value again indicates that the urban townships are primarily residential and possibly commercial. Their percentage is much closer to suburban townships than to either central or suburban cities. 106 Suburban cities are still well below central cities in percentage industrial. This underscores the qualitative difference between the two. But this difference, combined with the residential value density may indicate the decline of housing in central cities relative to both suburban housing and industrial value in the cities. Table 31.--Percent housing built before 1940 by category Category Mean Std. Dev. No. of Cases Central city 59.58 14.46 13 Urban township 21.26 8.12 18 Suburban municipality 31.68 18.46 28 Suburban township 33.64 14.83 77 Rural municipality 48.89 12.00 4 Rural township 47.72 15.93 62 Total 38.56 18.04 202 ANOVA F value = 17.19 significance = .0000 Table 32.--Overcrowded housing by category Category Mean Std. Dev. No. of Cases Central city 10.46 11.41 13 Urban township 6.46 2.86 18 Suburban municipality 8.39 18.03 28 Suburban township 8.16 3.93 77 Rural municipality 7.35 3.14 4 Rural township 8.47 4.73 62 Total 8.27 8.09 202 ANOVA F value = .3855 significance = .8544 107 Aged housing and crowding are both structural vari- ables in which high levels of both are commonly ascribed to the central cities. Our sample indicates that for aged housing, this is indeed the case.’ Even the more stable rural areas do not approach the central cities in age of housing stock. The range between central and suburban cities explains the significance of this variable in the cities regression while the rural areas dilute the variation for the full sample. On the other hand, crowded housing reveals the cause of that variable's non-significance in the regressions. There is very little variation across the whole sample, although the group standard deviations seem to indicate that there are some extreme individual cases. One problem with CROWD may be that the Census definition is not extreme enough either to explain social problems or reveal differences between categories. Raising the measure to 1.5 persons per room might accomplish this. Types of employment provide a useful comparison with property value, especially since so many studies use employment to characterize cities. Given the realities of commuting, value is likely to give a much more accurate picture of the city's tax base and physical attributes. The "employment” variable states how the residents are employed, not where they work. With properly drawn occupa- tional categories this would be a useful social class variable. 108 Table 33.--Percent employed as professional, technical and managerial by category Category Mean Std. Dev. No. of Cases Central cities 19.70 , 9.98 13 Urban townships 26.33 11.99 18 Suburban municipalities 28.02 12.57 28 Suburban townships 19.17 8.95 77 Rural municipalities 19.30 4.32 4 Rural townships 13.63 4.65 62 Total 19.37 10.07 202 ANOVA F value = 12.79 significance = .0000 Table 34.--Percent employed in manufacturing by category Category Mean Std. Dev. No. of Cases Central city 34.74 10.98 13 Urban township 34.55 8.55 18 Suburban municipality 32.36 10.07 28 Suburban township 37.72 9.027 77 Rural municipality 34.37 7.46 4 Rural township 41.04 9.70 62 Total 37.46 9.81 202 ANOVA F value = 4.05 significance = .0016 The comparison between white- and blue-collar employment is quite interesting and does not parallel the findings of value density. There is a very slight differ- ence in percentage employed in manufacturing. This seems 109 to indicate the tendency of blue- as well as white-collar workers to commute, and the possibility of commuting in both directions. PROF clearly indicates where the upper class is going, or at least is now. Of particular interest is the role of urban townships in competing with cities. This trend is repeated in per capita income and value. The standard deviations present in employment are worth noting, especially in PROF. The deviations for both central cities and suburbs admit some rather extreme cases of concentration, both on the high and low sides. Disaggre- gating PROF into more specific occupational groupings might shed some light on the causes and effects of concentration, but that is beyond the scope of this study. A major question is the effect that the previous differences have made in income and wealth. These two fac- tors have somewhat different fiscal implications. Table 35.--Percent of families below the poverty level Category Mean Std. Dev. No. of Cases Central cities 10.31 4.11 13 Urban townships 4.83 3.05 18 Suburban municipalities 4.03 2.26 28 Suburban townships 5.71 3.43 77 Rural municipalities 6.00 3.57 4 Rural townships 8.23 4.26 62 Total 6.47 3.98 202 ANOVA F value = 9.93 significance = .0000. 110 Table 36.--Percent of workers with incomes below $5,000 per year Category Mean Std. Dev. No. of Cases Central cities 19.22 , 5.14 13 Urban townships 9.98 5.27 18 Suburban municipalities 9.73 4.56 28 Suburban townships 12.51 5.75 77 Rural municipalities 13.36 7.69 4 Rural townships 16.57 6.05 62 Total 13.59 6.30 202 ANOVA F value = 10.61 significance = .0000 Tables 35 and 36 are included to indicate the slight difference between the two, as well as the differences between the categories. The percent of families living in poverty is the slightly lower income level. It is rela- tively higher (percent) in central cities than the $5,000 level. This may be a partial explanation for its greater effect in the regression equations. The tables very clearly depict the difference between the central cities and suburbs in terms of poverty and maldistribution of income. This fits closely with results for larger cities, although here they are not so extreme. In terms of poverty and income, the urban townships very closely resemble the suburban categories. The break- downs seem to indicate a rural-suburban difference, although it is less pronounced. A factor in this is that the sample 111 rural areas are not typical or representative rural commun— ities, even for Michigan. All are within commuting distance of cities and are more prosperous than the norm. Table 37.--Per capita income by category Category Mean Std. Dev. No. of Cases Central city 34.98 5.25 13 Urban townships 50.86 20.13 18 Suburban municipality 48.24 16.27 28 Suburban townships 37.73 9.20 77 Rural municipality 34.41 3.90 4 Rural townships 33.49 5.56 62 Total 38.81 12.21 202 ANOVA F value = 12.43 significance = .000 Table 38.--Fu11 per capita residential value Category Mean Std. Dev. No. of Cases Central cities 1997.04 279.95 13 Urban townships 3216.55 1615.75 18 Suburban municipalities 2898.30 988.18 28 Suburban townships 2375.05 1005.04 77 Rural municipalities 1897.25 328.69 4 Rural townships 1925.09 1216.70 62 Total ' 2350.67 1169.04 202 ANOVA F value = 5.83 significance = .0000 Income and wealth are not the same, expecially for local government. Only four cities of the entire sample 112 Table 39.--Per capita assessed value by category Category Mean Std. Dev. No. of Cases Central cities 5127.53 1000.67 13 Urban townships 5323.11 2268.29 18 Suburban municipalities 5324.62 1769.68 28 Suburban townships 5759.25 2987.73 77 Rural municipalities 3580.33 978.66 4 Rural townships 4518.81 1460.29 62 Total 5195.61 2295.92 202 ANOVA F value = 2.52 significance = .0306 are permitted an income tax. The rest tax on the basis of wealth. But residents feel the burden of the tax on income. Per capita income and residential value reflect very closely the same pattern. As in PROF, income and housing value are now located in the suburbs. Of interest is that the highest values of both are in the urban town- ships, while the highest total value per capita is in the suburban townships. This indicates that business value may be highest in suburban cities, while the highest residential values are outside cities, both central and suburban. Were tax rates for low-income areas significantly lower than those for wealthier areas, cities would to some extent be compensated. But the reverse is actually the case; tax rates for cities are much higher. This is broken down by category in Table 40. 113 Table 40.--Local prOperty tax millage Category Mean Std. Dev. No. of Cases Central cities 57.13 6.55 13 Urban townships 45.94 . 4.15 18 Suburban municipalities 54.45 7.41 28 Suburban townships 41.81 4.44 77 Rural municipalities 49.98 5.07 4 Rural townships 37.39 4.88 62 Total 43.72 8.31 202 ANOVA F value = 63.02 significance = .0000 This table also includes the county property tax of the county within which each local government is located. If this is taken as the total tax for local governments, the rate for central cities will be somewhat understated since four of the 13 have an income tax superimposed on the prOperty tax. The effect of tax on such factors as industrial localtion cannot be dismissed. While the tax rate differ- ential between central and suburban cities is about five percent, the difference between central cities and urban townships is a rather large 24 percent. This rate differ- ence and the proximity of urban (and suburban) townships to central city facilities makes them attractive sites for future relocation or expansion. An oft-heard convention is that cities provide many services that must be provided privately in suburban and rural areas. While this is true, cities in Michigan also tend to charge for those services. 114 Table 41 presents a breakdown of miscellaneous charges by local governments. This includes license fees and service charges, such as refuse collection. It does not include charges by independent agencies, such as water or power authorities. Table 41.--Local government charges by category Category Mean Std. Dev. No. of Cases Central cities 119.24 91.05 13 Urban townships 14.68 14.14 15 Suburban municipalities 50.27 45.58 28 Suburban townships 6.54 13.16 69 Rural municipalities 77.94 61.28 4 Rural townships 3.33 4.41 60 Total 21.91 44.87 189 ANOVA F value = 32.29 significance = .0000 0n the whole, the group means present a very good picture of some qualitiative differences between community types. Only two of the variables, GROWD and manufacturing employment failed to show a significant difference. Employ- ment seems to be a variable suited perhaps to gross studies of larger cities, but not to comparisons between cities and their suburban neighbors. PROF in this context is as much a social class designation as an employment classification. Value by class of property seems to be a better measure of structure or ”city type," although it too has its drawbacks. 115 CROWD produces an interesting comparison with measures of wealth. Overcrowding in various forms has long been a concern of social reformers. The Census measure is perhaps the most common one used in studies of overcrowding. Yet the measure is almost completely unresponsive to the large income differences in the sample. This seems to point towards overcrowding being something of a false or past issue, with income and employment differences being the more important. V. CONCLUSIONS AND RECOMMENDATIONS The approach taken in this study has been one designed to interest planners in the implications of munici- pal expenditures for urban functions and to discover those variables exerting strong influences on the urban finance in southern Michigan. It examines some of the tools used in comparative expenditure analysis and combines these with a look at other basic characteristics of the sample commun- ities. The first section summarizes the study's findings and conclusions. The second section applies the findings and conclusions to public policy recommendations and the final section identifies pertinent areas for further research. The study began with three hypotheses describing the sample communities. The first was that Michigan's central cities are a distinct class, having more in common with each other than with other local governments. The second hypothesis was that characteristics which distinguish the central cities have a significant effect on the struc- ture of municipal costs. The third was that the differences 116 117 in municipal costs can be explained by a limited number of structural and socioeconomic factors which cannot be altered over the short run. The second and third hypotheses depended on the first being true and described the implications of these common characteristics on municipal costs. Findings The hypothesis that Michigan's central cities are a distinct category was substantiated for most variables investigated. Only in the cases of overcrowded housing and percent employed in manufacturing did the analysis fail to discover significant differences between categories. Elimd- nation of these variables still indicates that in structure, pOpulation, income and wealth, Michigan's central cities have much more in common between themselves than with their suburban neighbors. The classification system.went somewhat further and distinguished between types of suburban areas with a considerable degree of precision. The second hypothesis was that characteristics distinguishing the central cities have a significant effect on municipal costs. This was less definitively proven. Several variables had an important impact on variation in some costs. However, for several cost categories variation was substantially unexplained. The best evidence for sub- stantiating the hypothesis was that the majority of vari- ation in total expenditures was accounted for. Several 118 basic expenditure categories such as fire and police also had a substantial portion of variation accounted for. With respect to the third hypothesis, the degree to which expenditure variations can be explained by a limited number of structural variables is indicated directly by the equations. Most of the variation was accounted for by only two or at most three variables. All such variables were chosen because they are basic indicators of municipal struc- ture and cannot be altered over the short run. Therefore, to the extent that the regression equations are accepted, the third hypothesis is validated. When examining the results of the study, a note must be made of the perspective of the planner. Unlike the pure social scientist, the planner is not necessarily seek- ing to create a general model. This has important ramifica- tions concerning the statistical reliability of the study. In fact, the key to a "successful“ expenditure study probably lies in not creating a sample which represents anything other than the communities included. Drawing a representative sample gets the planner away from the speci- fic problemrsolving approach. From that perspective, the sample used is something of a compromise; central cities are completely represented, but the rest of the suburban- rural continuum has a representative sample. The classification system used seems to have performed fairly well. Most variables showed significant differences between groups. Refinements need to be made in 119 differentiating urban from suburban townships. There seems to be an important difference between the two, expecially in size and income. Yet the actual selection was somewhat arbitrary and probably included suburban as well as urban townships in the latter category. Ideally, performance criteria should be set down for any classification system, rather than relying on political divisions (school systems) as this study did. This should not detract from the major purpose, which was to indicate what differences do exist. Other studies have used popula- tion and density criteria (Williams, 1971) which are perhaps even more arbitrary. Such a classification system, especially one used for policy purposes, must take several factors into consideration. Size is one, but employment, income and wealth are equally valid. Currently, school aid and general revenue sharing legislation distinguish commun- ities by wealth, size, and tax rate. A primary purpose of the regression analysis was to point out factors that should be considered in classifi- cation. The relationship of value density to cost and the difference between residential and business values indicate that businesses do generate a large portion of costs. Age of housing is another significant factor in rising costs, yet there is very little local government can do to alter this in the short run. Other variables such as professional employment are indicative of broader changes in quality of life that 120 probably cannot be accounted for in any formula. Yet they are indispensable knowledge for anyone involved in planning for the future of these communities. As the real growth rate of society slows, these questions of distribution will become much more pressing than at present. Classification systems which define and delineate that distribution will be vital in finding viable solutions to problems resulting from maldistribution of society's resources. Such classifications also firm up issues such as those involving cities and the quality of life. This has figured prominently in discussions of urban growth policy and city size: Conceptually, the underlying issue seems to be that of "quality of life.” A city of any given size is not a goal unto itself; rather, city size seems important as it contributes to or detracts from conditions of life which are thought to be desirable." (Stanford Research Institute, 1974, p. 80.) - This study has used in its "central city" category those which fit within the commonly accepted "Optimum" range of 25,000-250,000. Yet the differences in quality between these and the suburbs indicate fairly extreme differences in variables related to the ”quality of life." Another aspect of the ”quality of life" issue is what the ACIR terms as the "decline of the balanced city” (ACIR, 1976, p. 33). Traditionally cities had poorer and richer sections. Although services on the ”other side of the tracks" left much to be desired, the balance enabled the cities to remain financially afloat and support the 121 necessary range of public activity. Now the richer sections are moving to separate jurisdictions, in effect leaving the poor to fend for themselves. If society wishes to remedy the inequities inherent in this transfer, then new ways will have to be found to break down the expenditure and rights transfer barriers between communities. Policy‘Igplications and Recommendations 0n the basis of the indicated trends and differences between urban and suburban cities of the sample, certain implications and recommendations can be drawn. Some relate to land and development policy, others to taxation which affects location and land policy. Most if not all apply to redistribution between urban and suburban communities. Among cost variables, two distinct types of varia- tion are delineated. The first is cost variation between cities. The increased costs in this category, as depicted by the second set of regression equations, are largely explained by the age of cities and their housing. This does not necessarily conflict with the emergence of property density as the major explanatory variable in the first regression set. It rather reflects the situation of similar densities among cities, while greatly differing densities are more a characteristic of the first set of equations. The true significance of BUSVAL is that all the assorted theories of poverty, class and income add much less to the explanation of relative expenditure levels than the 122 simple fact of the city itself. BUSVAL is no less than a measure of the urban center, a quantitative indicator of the city's economic function. This suggests that the very structure of the industrial city-is responsible for the high level of municipal costs. Certainly within the cate- gory of city and for specific municipal functions, other more specific causes may be found. But these will probably be secondary in nature to the basic structure of industrial urban life. This finding carries significant implications for the options available to the cities and state. Foremost must be the realization that the structure of increased costs is basic to the nature of the city; the same density that accounts for cost increases in most basic spending categories is also the very heart of the city, and in a sense the reason for its existence. To some extent the age and structure of the city may be altered, but this is not likely to be accomplished over the short run. This implies that the increased and increasing costs of the central cities are likely to remain with us, and the solution will not be found by exhorting city managers and city councils to spend less on essential services. Nor will important and sustainable economies be realized by simply cutting back on basic operating budgets. The reason for this is not a question of ethics involved with income redistribution or of increasing social problems if such services are curtailed. It is rather that costs 123 will increase in basic services, without which the city would not exist. Temporary economies achieved by cutbacks in operating budgets will eventually increase the costs in basic services areas as the cities decline further. One possible explanation for the changing cost pat- terns may be that as suburban and rural areas begin to add urban-type physical facilities, state and federal assistance is more likely to pick up at least part of the cost. Thus the outlying areas are faced with a relatively smaller financial burden than the cities, which originally paid the full cost of providing the infrastructure and are now having to replace them at much higher cost. In the regression, this shows up as the smaller increased cost for sewers explained by PROF or mostly suburban areas. The maintenance and operating costs show up as an increase because this is not compensated for as a specific cost by aid programs, whereas capital costs may be handled in a number of differ- ent ways. The central cities face the situation where, even when the state assumes the cost of many social programs, the central cities will still face a higher per capita operating cost that will be reflected in higher personal and property taxes. Some general legislative approaches recognize this cost factor and attempt to compensate for at least some of its effects. One type of program is state revenue sharing. This allows the state to assume a portion of general operating costs on a scale proportional to tax 124 burden. Other approaches involve various forms of tax reform to expand the tax base of the cities and reduce the fiscal incentives to leave the cities. One of the most striking findings of the study is the variation in tax effort by community type. Programs need to be designed which will take this kind of difference into account. The present state revenue sharing formula is calculated to return money to communities with high tax rates, but ignores the crucial difference in the ability of communities to finance those taxes. The formula Operates on the basis of tax burden with respect to property value. While current income is not the sole determinant of ability to pay, it needs to be heavily considered along with valuation. The present form- ula could be modified to include a measure of income. This would decrease returns to wealthy communities with high‘ tax rates and also benefit communities which rely more heavily on industrial rather than high residential valuation. Since the federal government now requires income data by community for federal revenue sharing purposes, adjustment for income level should be available to the state on a consistent basis and at reasonable cost. Another problem with the present formula is that it Operates from a fixed per capita floor with increments added as state funds permit. The progressive increment is allocated by tax burden. Thus, when the fund Operates near the floor, as it does at present, richer communities receive 125 almost the same per capita amount as the poorest. If this continues the relative tax disparity between central cities and suburbs will remain, as will fiscal incentives to move industry to the suburbs. This cOuld be remedied either by lowering the floor or increasing funds allocated to the revenue sharing program. Various tax proposals have also been made to redis- tribute income geographically. One form of this is tax-base sharing. A variant of this is being tried in Minneapolis and one has been proposed for the areas surrounding Balti- more (Lyall, 1975). A similar proposal for the Detroit metropolitan area was made in 1976 (Senate Bill 6010). This measure is expected to receive serious consideration as one long-term.solution to Detroit's fiscal problems. The statewide industrial tax is more a more radical change from the present tax structure. This prOposal would simply remove business prOperty from the local tax base and tax it at a single statewide rate. This is now done for some utilities' property and property such as rail lines. The statewide industrial property tax has several advantages. It would be both fairer and easier to admin- ister than the local property tax, and would remove from local consideration the question of valuation of large and unique industrial properties. Under the tax-base sharing proposals, differential assessment would still be a temptation. By removing the tax advantage of migration, 126 the state would foster a more efficient location of indus- tries with respect to transportation, employment and other imputs. This would allow the central cities to bring underutilized resources back into production. Another point related to the industrial tax question is that of the property tax credit currently granted on the state income tax. While this provision may grant some relief to urban residential property owners and renters, it does not affect income-producing property. The state, by using the tax credit to encourage localities to raise millage rates is in effect hastening the departure of industries from the cities to outlying tax havens. If the tax credit is to be used for its original purpose, then either some relief must be granted to industries or industry must be removed from the local tax base. The easiest and most logical means of accomplishing that would be through a statewide uniform tax. This would still retain the indus- trial tax base while reducing the unintended but damaging anti-urban bias of the tax credit. . In contrast to tax-base sharing a statewide indus- trial property tax would not require the delineation of ”urban-suburban” tax base areas. Were such areas to be established in Michigan, southern Michigan would require a minimum of 11 areas. The possibility of inequity and political conflict is much greater there than in the crea- tion of a single uniform tax base and rate. 127 The tax-base sharing concept also differs in that it is an incremental approach to changing the tax base. Under 8.8. 6010, only new property would be added to the ”tax pool” shared by communities. As time goes by, a progressively greater proportion of the industrial tax base would be included. But it would probably take in excess of 50 years to include the total industrial and commercial tax base. Once included in the pool, property is taxed at the average tax rate and the money is shared between communities in much the same manner as funds under revenue sharing. A concept related to tax-base sharing is that of life cycle financing. Part of the purpose of expenditures analysis is to reveal the fiscal implications of different stages of growth and decline. When this is delineated, financing and the tax base can be shifted with stages of the cycle to share tax burdens as each part of the urban structure moves through the cycle. As yet, however, not enough is known about the cycle to develop practical guide- lines for this policy. The analysis of expenditures by class has Opened the way for a new structure of advocacy planning. Central cities are themselves a class of clients. This study has indicated that central cities have a great deal in common and much to gain by banding together in support of common interests. The classification system is necessary to reveal this sort of relationship, and is quite possibly its most 128 important use. The original contractor of the study, Middle Cities Education Association, directly parallels this philosophy in the public education structure. The role of the planner as advocate for the central cities involves: (1) seeking to clarify the relationship between central cities and their surroundings, (2) emphasiz- ing those aspects which are perceived to be inequities; and (3) pointing out alternatives which may rectify the perceived inequities. In this view planning is nOt a science in which only Pareto-Optimal solutions are acceptable, but a tool which is part of a larger struggle for social justice. Recommendations For Further Research Although useful, the regression approach is obviously an imperfect tool for expenditure analysis, and probably for urban research in general. aIts wide use is probably explained by its relative simplicity compared to other more s0phisticated approaches such as path analysis. These more sophisticated methods, which are more capable of distinguishing between direct and indirect effects, need to be made more usable and more understandable to social scientists and planners. Working extensively with financial and urban statis- tical data brings with it the realization that most of our measures are fairly crude. In Michigan, for example, con- sistent data on property value by class was not available before 1972. In most states it is not yet available. As a 129 result, comparative data on the physical structure of cities is virtually non-existent and researchers have fallen back on less apprOpriate measures. One of the important sectors not included in the analysis is county government expenditures. Although some large cities comprise whole counties or collections of counties (such as New York) the more common situation in Michigan is one where the central city is contained in the same county as its inner suburbs. While ample data are available listing sources of county revenues, there is very little information on geographical expenditure patterns within county borders. A major unexplored issue is whether counties are using urban-derived revenues to finance subur- ban servcies. Some evidence suggests that for such services as police and highways, county governments are furnishing some services to urban and suburban townships that cities must provide for themselves. Additional research is needed to discover the real pattern of county expenditures, and whether changing that pattern might be an important factor in equalizing tax effort and service structure. Research is also needed in additional areas. This study has dealt with the effects of current operating costs. While this is a major immediate concern, capital costs have a significant impact on municipal budgets. But capital expenditures by local government are also much more diffi- cult to measure consistently and accurately. Research on capital expenditures has lagged far behind that of current 130 expenditures. In particular, more information is needed on the useful life and replacement cost of major components such as public buildings and utilities. While the older central cities are the focus of this impact now, it may be expected that the more established suburbs will be facing similar problems in the near future. A final problem.which is almost unique to the cen- tral cities is the cost of population and industrial decline. In Michigan this problem is neither small nor isolated; between 1960 and 1970, 58 Michigan cities declined in pop- ulation. This is 26 percent of all Michigan cities over 2,500 population. Within the sample, only one central city, Ann Arbor, increased in population between 1970 and 1973. Yet the expenditures for these declining cities are appar- ently still increasing. Research needs to be done on ”recycling” abandoned urban areas and on ways to both mini- mize and equitably distribute the costs of redevelopment. While more equitable aid formulas and tax structures may help equalize the burdens in the urban areas, a long- term solution must be found in helping cities to deve10p a base of economic activity consonant with its cost and structure. To do so will require aid not only in improving the cities' immediate financial situation but in assisting them in developing a new, more functional economic base. Expenditures analysis has produced the base information and techniques of analysis that make possible new opportunities for more rational use of planning policies 131 and power. Among these is the greater integration of tax and fiscal structure into the planning process, and the more accurate anticipation of the effects of fiscal policies on land use. It is only when the true costs and benefits of major decisions are made known that rational goals and policies may be set. APPENDICES APPENDIX A APPENDIX A MEANS, STANDARD DEVIATIONS AND NUMBER OF VALID CASES FOR REGRESSION VARIABLES Variables for first set of equations: Variable Mean Standard Dev Cases PROF 19.37 10.07 202 LOWINC 13.59 6.30 202 OLDHOUS 38.56 18.04 202 CROWD 8.27 8.09 202 BUSVAL 2283505. 4890554. 202 FIRE 6.38 7.27 139 TOTEXP 71.36 97.01 202 GENAD 14.16 13.523 194 BLDGS 1.329 1.567 127 HEALTH .668 1.309 27 HIWAY 7.643 8.452 180 HUD 10.402 10.385 15 COP 8.952 11.413 138 SEWER 32.890 49.560 118 SAN 3.674 5.403 6 GOVT 10.481 14.740 170 DEBT 8.400 10.128 108 Variables for second set of equations: Variable Mean Standard Dev Cases POV 7.05 3.66 25 OLDHOUS 42.16 23.59 25 POPDEN 4190. 1905. 25 SALARY 838. 271. 25 ROADS .0008 .0005 25 VDEN 20244213. 8764191. 25 SEW .0004 .0003 25 GEN .0011 .0005 25 FIRE .0013 .0008 25 POLICE .0018 .0009 25 132 APPENDIX B B>00 mmovm. z¢m mommm.l Hmmwm.l mxmBOB momma. mmmom. hommH.I oommm. mmmmm. hmoam. mommo. Hoomm. omomv. onmao. MQSMm mambo. bommm. ommoo.l mMHh hmmmd. woman. wmmam. mmoao. momma. momma. mommm. mmvom. mahmv. mhmmh. whovm. cOAmmoummm madamm flash now udeaUamwOOU cowumamuuoo MUHAOm moomo. «homo. voamo. mmmmw. A¢>mbm ommbo. moron. mmovv.l hamom. mmomm. oommm. moomn. Haovm. mohmm. mmoam. hommh. mhvhm. ODE MHmHm. moomo.l o mwmoH.I odmom. ZMD> momeo. movvn. oomdh.l moods. Hmdoo. ommmm. mommm. vmmhm. mmmmn. hommm. hhHHm. Hmvhm. comma. ”dBHm homoo. mmomm. mammo.l hmmmm. mmmam. Haomo.l DSOMU mmamo.l vomdo. ommom. mmmHo. commo. momma. otho.l mmdmo. oodmo.l hmmoo. hmmmo. momao. mmomo. Hmoma. m xHDzmmm< maddflm momma. momma. ooooH.I woman. Homoe. mommm. mmmov. mDOmQAO moood.l momma. omnuo. MMHvH. Hmmha. vomnm. Hmomo. mommm. mmoho. mmaoo. chemo. mmoam. ombmo. mmomo.l «mono. mUQAm mmHoo. mmomo. mmmoH.I mmomm. mammm. omHhH. Hmmdm. oommm. UZHSOA momma.l ohomo. mhhmo. Hmoao.l vmvHo. Hombm. mmmno.l oovmm. vhmvo.l ovaH. mhhdo.l moon. Hommo. mmhmo. mmooa. mommm. Gdzmw mmomm. mommv. ammoo. oaoam. Hmoom. mmvmv. Ahmmm. Hhfimm. mmvmo. momm ohmvo. hmmHo. omnmo.| mmmhm. mmvom. ommmm.l mmooH. momHH.I mvaH. oomoo. mlom. thmH. onom. mmomo. mmmem.t mammo.l mommo.l Emma B>OU z¢m mMSHm WUHQOm com mndom mendflm mwoqm Rama B>OU zmam zma> Q3Qm0 moomnqo UZHSOQ 133 APPENDIX C POP OLDHOUS SALARY INCOME POPDEN VDEN PCSEV APPENDIX c SIMPLE CORRELATIONS FOR.CITIES ONLY VARIABLES .0980 .7926 .1224 -.6072 .1252 .0910 -.2351 POV .1237 .2577 .1553 .4352 .2471 POP .1381 .0568 -00009 OLDHOUS .1948 .3673 .6256 .1030 SALAR! 134 .0871 .1793 .7974 INCOME POPDEN VDEN APPENDIX D 1. 10. 11. APPENDIX D List of Data Sources Variable Full sample expenditure variables NO. employees for municipal functions Salaries for common municipal functions Area Population Per capita income Low income and poverty levels Housing Built before 1940 Percent employment by occupations Tax millage Assessed valuation by category 135 Source 1972 Census of Governments, Data tape, file A 1972 Census of Governments 1972 Census of Governments 1967 Area Measurement Reports, Bureau of Census, Ge-20 no.4 Bureau of Census Current Population Series P-25, 1973. Bureau of Census Current Population Series P-25, 1973 1970 Census, fourth count 1970 Census, fourth count 1973 Tax tables, State Tax Cbmmdssion 1973 Tax Comission working tables, State Tax Commission APPENDIX E APPENDIX E Sample of Municipalities Category One - Central City Bay City Benton Harbor Battle Creek Flint Lansing Jackson Kalamazoo Grand Rapids Muskegon Muskegon Hts. Pontiac Saginaw Ann Arbor Category Two - Urban Township Benton Hagar Battle Creek Emmett Pennfield Dewitt Flint Delhi Lansing Kalamazoo Grand Rapids Laketon Muskegon Bloomfield Pontiac Waterford Saginaw Ann Arbor 136 Category Three - Suburban Municipality Auburn Essexville Buchanan St. Joseph Marshall Springfield Grand Ledge Burton East Lansing Galesburg East Grand Rapids Kentwood Walker Wyoming Milan N. Muskegon Norton Shores Roosevelt Park Berkley Huntington WOods Lathrup Village Oak Park Royal Oak Southfield Sylvan Lake Frankenmuth Zilwaukee Ypsilanti Category Four - Suburban Township Beaver Frankenlust Hampton Kawkawlin Merritt Monitor Portsmouth Williams Bainbridge Baroda Bertrand Buchanan Lincoln Niles Royalton St. Joseph Sodus Bedford Watertown Delta Oneida Roxand Genessee Mt. Morris Mundy Meridian Blackman Henrietta Leoni Summit Waterloo Charleston Oshtemo Pavilion Ross Texas Convis Eckford Fredonia Lee Marengo Marshall Eagle Egelston Fruitland White Hall Commerce Highland Milford Royal Oak Southfield White Lake Blumfield Buena Vista Carrollton Frankenmuth Kochville Zilwaukee Tuscola Lodi Northfield Category Five - Rural Municipality Pinconning Potterville Clio Cedar Springs Category Six - Rural Township Bangor Fraser Garfield Gibson Mt. Forest Pinconning Berrien Pipestone Batavia Matteson Sherwood Union Athens Burlington Clarence Alpine Plainfield London Milan Blue Lake Cedar Creek Dalton Pittsfield Salem Saline Scio Superior Webster York Ypsilanti Comstock Flushing Montrose Thetford Vienna Bentley Bunkerhill Ingham Stockbridge Wheatfield White Oak Parma Springport Tompkins Brady Praire Ronde Leroy Tekonsha Benton Brookfield Windsor Oakfield Solon Spencer Iosco Unadilla Exeter Holton Independence Springfield Greenwood Allendale Tallmadge Birch Run Taymouth Mendon Park Augusta Bridgewater Freedom Lyndon Manchester Sharon 139 Schoolcraft Wakeshma Algoma Courtland Nelson REFERENCES 10. 11. REFERENCES Adams, Robert F. "On the Variations in the Consumption of the Public Services," in Essays in State and Local Finance, ed. Harvey Brazer, University of Michigan, Ann Arbor, 1967. Advisory Committee on Intergovernmental Relations, City Financial Emgrgencies: The Intergovernmental Dimension. Report A-42, Washington, D.C. 1973. -—- Involving Urban America: A Challenge to Federalism. Report M-107, Washington, D.C. 1976. --- Measuring the Fiscal Capacity and Effort of State afid:Local Areas. Report M-58] Washington, D.C. I971. --- Metropolitan Social and Economic Digparities: Central Cities and Suburbs. Washington, D.C. 1965. Applied Decision Systems, Inc. The MSHDA Community Impact Model, Michigan State Housing and Development Authority, Lansing, 1972. Bab, Herbert J. "Gut Issues of Urban Financing," American City, July, 1969. Bahl, Roy. Metropolitan City Expenditures. Univ. of Kentucky Press, Lexington, 1968. Bahl, Roy. "The Determinants of State and Local Govern- ment Expenditures,” in Functional Federalism: Grants-in-aid and the PPB System, ed: S. Mushkin and J. Cotton, George Washington University, Washington, 1968. Beaton, W. Patrick. "Environmental Structure and Municipal Costs," in Municipal Needs, Seryiges and Financing, ed. W. Patrick Beaton, Center for Urban Policy Research, New Brunswick, 1974. Birch, David. Economic Future of Citypand Suburbs. Committee on EconomiE DevelOpment, New York, 1970. 140 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 141 Blalock, Hubert M. Social Statistics. second ed. McGraw—Hill, Inc. New Yoik, I972. Bradford, D. F. et. al. ”The Rising Cost of Local Public Services: Some Evidence and Reflections," in Municipal Needs, Services and Financing, ed. W. Patrick Beaton, CenteriiOr Urban Policy Research, New Brunswick, 1974. Brazer, Harvey. City Expenditures in the United States. Occasional paper no. 66, NationaiiBureau of EconomiE Research, New York, 1959. Campbell, Alan, ed. The States and the Urban Crisis. American Assembly, Prentice-Hal , Englewood Cliffs, 1970. Campbell, Alan, and Seymour Sacks. Metropolitan America: Fiscal Patterns and Governmental Systems, The Free Press, Glencoe, 1967. Cothan, James C., and Kenneth E. Quindry, eds. Needs Vs. Revenue: The Dilemma of Urban America. University of Tennessee, KnoxVille, I97 . Davies, David. ”Financing Urban Functions and Services," Law and Contemporarprroblems, Winter, 1965. Davis, Otto A., and George H. Haine Jr. ”A Political Approach to Public Expenditure: The Case of Munici- palities," National Tax Journal, Sept., 1966. Dusansky, Richard, and Lawrence Nordell, ”City and Suburb: The Anatomy of Fiscal Dilemma,” Land Econo- mics, May, 1975. ”Financing Our Urban Needs,” Nation's Cities, March, 1969. Gabler, L. R. "Population Size As a Determinant of City Expenditures and Employment--Some Further Evidence,” Land Economics, May, 1971. Gale, Dennis E. The Municipal Impact Evaluation System. ASPO special report no. 294, Sept., 1973. Ginzberg, Eli. The Futurelof the Metropolis: People, Jobs, Income. OIympus Publishers, New York, 1974. Governor's Advisory Task Force On Property Tax Revision, ProaertyTax and Municipal Finance, Lansing, Michigan, 1976. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 142 Greene, Kenneth. Fiscal Interactions in a Metropolitan Area. Urban Institute, Lexington Books, Lexington, 1973. Grieson, Ronald E. "On the Optimal Level of Taxation, Borrowing and Government Expenditures,” National Tax Journal, June, 1975. Hawley, Amos H. ”MetrOpolitan Population and Municipal Government Expenditures in Central Cities," in Cities and Society, ed. Paul K. Hatt and Albert Reiss, Free Press, New York, 1957. Hirsch, Werner, "Expenditure Implications of Metropoli- tan Growth and Consolidation," The Review of Economics and Statistics, August, . Isard, Walter, and Robert Coughlin. Munici a1 Costs and Revenues Resulting Frompgommunity Gigwt . Chandler-Davis, New York, 1957: Kasarda, John D. "The impact of Suburban POpulation Growth on Central City Service Functions,” in Municipal Needs, Services and Financin , ed. W. Patrick Beaton, Center ofiUrban PoIicy Research, New Brunswick, 1974. Liebert, Roland J. “Municipal Functions, Structure and Expenditures: A Reanalysis of Recent Research,” Social Science Quarterly, March, 1974. Lyall, Katherine C. ”Tax-base Sharing, A Fiscal Aid Towards More Rational Land Use Planning," American Institute of Planners Journal, March, 1975. Masten, John T. and Kenneth E. Quindry, “A Note On City Expenditure Determinants,” Land Economics, Feb., 1970. Muller, Thomas. Growing and Declining Urban Areas: A Co arison. Urban Institute, Washington, D.C., 1975. Muller, Thomas. Fiscal I act of Land Development, A critique of Methods an? Review of Issues. Urban Institute, Washington, D.C., 1 . Neenan, William. "The Suburban-Central City Exploita- tion Thesis: One City's Tale." National Tax Journal, June, 1970. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 143 Netzer, Dick. "Financing Suburban Development," Econo- mic Aspectgpof Suburban Growth: Studies of the Nassau-Spgfolk PlanningTRegio . ed. D. D. Zschock, Economic ResearcEIBureau, Stony Brook, N.Y., 1969. New Jersey Department of Community Affairs, Municipal Information Service. Respondin to Difficult Financial Conditions. Tenton,*1 75. Nie, Norman et. a1. SPSS: Statistical Package for the Social Sciences. second edition, McGraw-HilI} New Yor , l . Polley, John W. "Variations in Impact of Municipal Government on Ability to Support Schools," in New York State Educational Conference Board, A New Approach to_School Finance: 1961 Reviewof Fiscal Po 1oy, or Public Education in New York—State, staff studies, Albany, 196i. Scott, Claudia D. Forecasting Local Government S end- i g. Urban Institute, Washington, D.C., I972. Sharkansky, Ira. ”Some More Thoughts about the Determinants of Government Expenditure," National Tax Journal, June, 1967. Stanford Research Institute, City Size and the Quality of Life. Menlo Park, 1975. Stinson, Thomas F. ”Population Changes and Shifts in Local Government Finance,“ Municipal Finance. February, 1970. Vernon, Raymond. Metropolis 1985. Harvard Univ. Press, Cambridge, 1960. Walzer, Norman. ”Measuring Inflation and Municipal Expenditures,” Government Finance. August, 1975. Weicher, John C. ”Determinants of Central City Expen- ditures; Some Overlooked Factors and Problems," National Tax Journal. December, 1970. Williams, Oliver et. a1. Suburban Differences and Metro litan Policies. Univ. of Pennsylvania Press, PhiladeIphia, I97I. Wise, Harold. ”More Than Planning," National Civic Review. May, 1966. ”@111 411111611 11147 ’75