June 25, 2004 | Backgrounder on Smart Growth
Over the past several years there has emerged in the United States an influential political movement whose purpose is to severely limit, or even prohibit, further suburbanization. This "anti-sprawl" movement has received much attention and has been successful in implementing its restrictive land-use policies in some areas. Much of the justification for the current campaign against the low-density (sprawling) urban development that Americans and Western Europeans1 prefer is based upon assumptions that it is more costly than the more dense development of central cities.
Variously described as "smart growth," "growth management," or "New Urbanism," the movement would force people to live at higher densities, in multi-family units, townhouses, or clustered single-family developments--while placing significant restrictions on the expansion of suburban commercial development
The rationales offered for limiting suburban housing choices are many, various, and of questionable validity. At one point or another over the past half-decade, critics of suburban development have cited its adverse impact on "food security," wildlife, and air and water quality. Critics of suburban expansion even contend that suburbs contribute to serial killings, teenage angst, social alienation, low wages, obesity, asthma, and higher taxes. This last item, the belief that lower-density, "more sprawling" development fuels higher government expenditures, is the most common reason elected officials in many municipalities adopt measures to limit housing growth in their communities.
Typical of the concern that low-density development raises municipal costs--and therefore local taxes--is a contention in a recent, federally funded study of sprawl and costs that claims the United States "no longer can pay for the infrastructure necessary to develop farther and farther out in metropolitan areas."2
The U.S. urban planning community has adopted several assumptions about suburbanization and local government expenditures. These are outlined below and are referred to as the Current Urban Planning Assumptions in this paper.
Belief in these assumptions provides support to urban planners and others who are interested in limiting suburban development and, in extreme cases, outlawing development outside "urban growth boundaries" or designated "growth areas."
In fact, however, virtually all of the research on which the Current Urban Planning Assumptions are based is theoretical, projecting relative costs into the future without examining the actual expenditures that are being made today by municipalities of differing urban forms and ages. The analysis in this paper reviews actual municipal expenditure data in relation to the Current Urban Planning Assumptions. Among the findings:
By far the largest expenditure for municipal governments is employee compensation. There were no reliable data for including this variable in the econometric analysis. Yet our nominal analysis indicates that virtually all of the variation between municipal expenditures per capita can be accounted for by differences in employee compensation per capita.
Perhaps the most oft-quoted recent research attempting to estimate the relationship between sprawl and infrastructure costs was conducted by a team led by Professor Robert Burchell and funded under the auspices of the federal government's Transit Cooperative Research Program (TCRP). The project included two reports: Costs of Sprawl--2000 and The Costs of Sprawl--Revisited.3 The Costs of Sprawl--2000 projected that from 2000 to 2025, America would incur $227.4 billion in gross additional costs for what the study terms "uncontrolled growth" (less dense, more sprawling growth) versus "controlled growth" (more dense, less sprawling growth). This equates to approximately $9.1 billion in gross additional costs per year.
The figure of $227.4 billion may seem large.4 Yet in the context of 25 years and an average population of 115 million households, it is actually rather modest. The $227.4 billion would amount to only $80 per household annually, or $29 per capita. (See Table 1.) This includes:
Many growth critics have used these estimates as a measure of the cost that sprawl imposes on society. Advocates of smart growth policies have implied that these estimates represent the costs of continued low-density development on society. However, many of the purported costs are not imposed on society at all: They are private costs freely paid by the people who buy new houses.
The cost of sewer and water infrastructure in new developments is passed on to the buyer, and subsequent operation is typically funded with fees assessed on residents and businesses based upon use. Typically, new housing development infrastructure (local streets, curbs, sidewalks, storm and waste sewers, and water supply lines within the development) is paid for privately by the purchasers of new houses, having been built by developers or homebuilders. These are fully private costs that are paid for by persons who voluntarily move into new houses and apartments, having determined that they can afford such a move.
There are further indications that the projections from the Costs of Sprawl--2000 are not "unaffordable," and actually are modest in comparison to other costs in the economy. Specifically:
Because the authors of the Costs of Sprawl--2000 mix public and private expenditures that they claim are related to sprawl, the study's findings offer little guidance on one of the key public policy issues related to suburbanization: What are the actual, additional municipal costs that suburbanization imposes on the community at large, if any? This report will attempt to fill that void by conducting an econometric analysis (see box) of municipal spending patterns to determine what portion of municipal costs appear to be related to the impact of sprawl.
If the Current Urban Planning Assumptions are valid, the trends that Costs of Sprawl--2000 identifies--having been underway for at least five decades--should reveal clearly the differences in expenditures between less sprawling and more sprawling areas. This means that older, higher density municipalities should have lower costs per capita than newer, lower density, more sprawling areas. These differences should be evident in the present spending patterns of local governments.
For an issue that has galvanized public debate in many communities throughout the country, there is little comprehensive, academic research on the actual relationship between land-use patterns and local government costs. The most recent research was published some time ago and is based upon early 1980s data. It was conducted by Professor Helen Ladd at Duke University, who performed an econometric analysis of growth measures and the actual public expenditures of 247 counties. She found that per capita expenditures on public services tend to rise as density rises and that higher population growth is associated with lower per capita local government expenditures--precisely the opposite of Current Urban Planning Assumption #1, above.8
Because the data used in her study are now more than two decades old, there is a need for more contemporary research on the factors that drive local government expenditures, especially in view of the predominant influence of "smart growth" urban planning theories. The purpose of this paper is to fill that gap with municipal cost data drawn from the 2000 Census.
Although the term "sprawl" has no precise definition, its most fundamental characteristic is lower population density. Smart growth advocates presume that building less sprawling, higher density communities results in lower government expenditures. If this is indeed the case, an analysis of municipal spending patterns across the country should show that as population densities go up, costs go down--and vice versa.
In order to reliably capture the impact of density on local government spending, we analyzed data for municipalities (cities and towns) rather than counties9 because that is the level of government most affected by finance issues regarding utilities (wastewater and water) and general public services. The source of municipal financial data for the analysis in this paper is the U.S. Census Bureau government finance database for fiscal year 2000, which contains information for approximately 1,800 municipalities. Additional data for municipalities are available from other sources, such as the 2000 U.S. Census (demographic and density data) and the U.S. Department of Justice (crime rates). Another advantage of using municipal (rather than county) data is that municipal boundaries typically contain little rural space; therefore, the population density within those boundaries is generally similar to urban population density.
Because the current urban planning debate is principally focused on where development occurs within the nation's largest metropolitan areas, the municipalities analyzed in this study included only those within metropolitan areas of more than 1,000,000 residents in 2000.10 Consolidated city-counties were not included, because such municipalities provide both city and county services and would be expected to have inherently higher expenditures as a consequence.11
The analysis in this paper does not include primary and secondary education costs. Most of the nation's primary and secondary education is provided by independent school districts that seldom match municipal (or county) boundaries. As a result, there is little, if any, broad demographic data specific to the geographical areas served by such districts. Related research indicates that, contrary to Current Urban Planning Assumption #2, elementary and secondary education expenditures tend to be lower in school districts with the greatest enrollment growth, and highest where there is the least growth.12 Our research focuses on municipal costs in three categories:
For purposes of this study, three econometric models13 were developed to estimate the relationships between various factors and municipal expenditures:
The Wastewater Charges and Water Utility Charges models were developed to capture the impact of density, growth, and age of community on the cost of these functions. These are frequently cited in the urban planning literature as being upwardly affected by more sprawling development.
Another reason for analyzing utility charges separate from general government functions is that water and wastewater services are generally financed by user fees, rather than by the general tax revenues that finance most other municipal government operations. In fact, these utilities are not inherently government services: In many communities, regulated private companies provide such functions.14
The General Government Model estimates the impact of a number of factors on per capita municipal government expenditures, excluding utilities and education. The model uses 13 factors that would be expected to influence local government expenditures (current and capital expenditures)15 per capita. These include factors that test the Current Urban Planning Assumptions (population density, population growth, and community age as indicated by median house ages). There were sufficient data for 738 municipalities to be included in the General Government Model.16 Table 2 lists the variables included in the General Government Model.
Government Model Results
The results indicate that the 13 factors in the General Government Model explain approximately 29 percent of the variation in municipal expenditures, as revealed in Table 3.17 This means that 71 percent of the variation in total expenditures is not explained by the factors included in the model, but rather by other influences which cannot be quantified or for which there are no available or accurate data. The conclusion is that, contrary to the theory, comparatively little of the variation in municipal costs is associated with the Current Urban Planning Assumptions. Other factors, which have not been identified, are more important.
As Table 3 reveals, 8 of the 13 factors were reliable predictors of either higher or lower municipal spending per capita (at a statistically significant 95 percent level of confidence). These factors are: poverty rate, local/state expenditure ratio, state and federal aid, density, persons per household, owner-occupied housing (percent of housing units occupied by owners rather than renters), median house value, and crime rate.
As the discussion in the box indicates, statistical significance does not necessarily denote practical significance. A factor may be a reliable predictor of an impact, but the impact itself may be small. Among the 13 growth-related factors analyzed in the General Government Model, practical significance varied widely. The local-to-state expenditure ratio18 had the highest practical significance (a 100 percent increase in the ratio of local government spending to total state and local government spending would be associated with a 55 percent increase in per capita expenditures--a practical significance of 55 percent). The other variables with comparatively high practical significance were persons per household (-42 percent), owner-occupied housing (-30 percent), crime rate (+25 percent) and median house value (+25 percent). The other reliably predictive (statistically significant) factors had practical significance less than 15 percent.19
The results derived from the General Government Model are consistent with the Current Urban Planning Assumption #1 that associates higher densities with lower municipal government expenditures--but only weakly so. The relationship was statistically significant (99 percent), but there was little practical significance, which would indicate that higher population density is associated with only a small downward variation in municipal costs per capita. The mathematical significance or elasticity was 0.146: Each 10 percent increase in density could be expected to produce a 1.46 percent decrease in municipal expenditures per capita. For the average municipality, each 1,000 increase in population per square mile20 is associated with a $43 per capita reduction in municipal expenditures. This is a minuscule expenditure decrease compared with the substantial increase in density required to trigger it. In other words, a virtually unprecedented increase in population density in an already urbanized area would trigger an decrease in expenditure equal to the price of dinner for two at a moderately priced restaurant.21
Population growth, the factor associated with Current Urban Planning Assumption #2 was not statistically significant and could not therefore be practically significant. Thus, the results from the model do not support Current Urban Planning Assumption #2, indicating no significant relationship between higher population growth and higher municipal expenditures per capita.
Median house age, the factor associated with Current Urban Planning Assumption #3 was not statistically significant and could not therefore be practically significant. Thus, the results from the model do not support Current Urban Planning Assumption #3, indicating no significant relationship between newer communities and higher municipal expenditures per capita.
Interestingly, the inclusion of the three factors that measure the impact of the Current Urban Planning Assumptions add little to the explanatory value of the General Government Model as here specified. Only one--population density--was found to be statistically significant (and of little practical significance). Excluding these three variables (population density, population growth, and median housing age) and re-running the model with the remaining ten factors yields an R-squared of 0.24, meaning that the model as so specified explains only 24 percent of the measured expenditure variability from one municipality to another. Adding the three growth-related variables to these ten factors brings the R-squared up to only 0.29, meaning that the inclusion of the growth variables improves the explanatory value of the model by only five percentage points. This is not much of an impact for issues that are alleged to be having important effects on government costs in growing communities.
Neither the Wastewater Charges Model nor the Water Utility Model indicated strong relationships between the identified factors and user charges, as Tables 8 and 9 demonstrate (see Appendix). The Wastewater Charges Model explained 12 percent of the variation in wastewater user charges per capita, while the Water Charges Model explained 8 percent of the variation in water charges per capita.22 Thus, the Wastewater Charges Model failed to explain 88 percent of the variation in wastewater charges, and the Water Utility Model failed to explain 92 percent of the variation in water charges. This suggests that influences other than those variables included in the model explain much of the differences in utility costs from one community to another.
With respect to the Current Urban Planning Assumptions, only density was found to be statistically significant, but of little practical significance. In the Wastewater Charges Model, density exhibited a practical significance of minus 18.0 percent, consistent with Current Urban Planning Assumption #1. Similarly, density's practical significance of minus 12.5 percent in the Water Utility Model was consistent with Current Urban Planning Assumption #2 (Tables 2, 8, and 9). However, this translates into only small impacts on consumer costs. A 1,000 person-per-square-mile difference in average population density is associated with a $6 difference in annual wastewater charges per capita, or fifty cents per month. In other words, a 1,000 person-per-square-mile difference is associated with an annual water charge difference of $4 per capita, or thirty-three cents per month--less than a penny per day. Obviously, such trivial savings in waste water and water-related costs cannot justify public policies that would force major changes in existing lifestyles or land-use patterns.
It is particularly significant that none of the Current Urban Planning Assumptions were associated with a statistically significant relationship with the variation in Wastewater Charges or Water Charges. These infrastructure functions are among those cited most often in claims that suburbanization imposes additional costs.
Another way to analyze the same data is to rank it by categories that reflect varying degrees of difference in some of the key independent variables (such as density) and relate these categories to the different cost measures that comprise the key dependent variables. The existence or absence of any obvious trends indicates how strong or weak the relationships are. Using the same Census data, a nominal (ranking) analysis by quintiles (20 percent rankings) was performed on the sample to determine whether the statistical relationships that the Current Urban Planning Assumptions would predict are actually evident in the data (Table 4).
As the nominal rankings reveal, none of the growth-related variables show the relationship with municipal expenditures that is predicted by the Current Urban Planning Assumptions. This confirms the findings of the econometric analysis, which was only able to explain a relatively small fraction of the cost differences among communities, and where only one of the growth-related variables (population density) was found to be statistically significant, but not practically significant, at conventional confidence intervals.
The most dense municipalities (quintile #1) also failed to have the expected lowest wastewater charges per capita or the lowest water charges per capita. Quintile #1 municipalities did, however, have lower than average wastewater charges, though only of $7 per capita per year--hardly rising to the level of "unaffordable." There was little difference between the quintiles in water charges per capita. (See Tables 10 and 11 in Appendix.) Thus, the pattern in the nominal data (actual ranked data) for utilities was different than predicted by the econometric analysis.
Thus, the actual expenditure data reveal that more dense, slower growing, and older municipalities do not have lower expenditures per capita--the opposite of what would be expected if the Current Urban Planning Assumptions were correct. 24
The fact that the econometric analysis explains so little of the variation in municipal costs per capita, combined with the fact that the highest density, slowest growing, and oldest communities do not have the lower expenditures per capita predicted by the Current Urban Planning Assumptions, would seem to indicate that other factors are more important drivers of variation in municipal costs between communities.
The most obvious place to look is local government employee compensation. Employee compensation is by far the largest expenditure function for most local governments, consuming, on average, 64 percent of total current expenditures.25 Employee compensation is approximately 3.5 times capital expenditures.26
Employee compensation varies significantly between jurisdictions. Census Bureau information indicates that local government average wages and salaries for similar positions and skills vary by as much as 93 percent between some states.27 These cost disparities are not necessarily explained by regional differences. For example, in the Denver metropolitan area the municipality with the highest wages and salaries per capita pays nearly 1.5 times the area average, and more than five times the municipality with the lowest wages and salaries per capita. Further, there are also significant differences (up to 123 percent) between the percentage add-on of employer-paid employee benefits costs among local governments by state.28
There are other factors that could be responsible for such large variations. There could be significant variations between the numbers of hours actually worked by government employees. This is evident at the state level, where differences of up to 38 days annually have been shown.29 Thus, it seems likely that differences in municipal government employee compensation per capita could be an important factor in explaining differences in municipal expenditures.30 Finally, there could be significant variations in the number of employees, or in employee productivity.
Although the available data cannot be used to econometrically test the impact of public employee compensation on municipal costs, the nominal ranking analysis used in the previous section can be extended to include a review of government employee compensation. 31 Table 5 provides an estimate of per-capita municipal employee wages and salaries for each set of quintile rankings for the three urban planning, growth-related variables.
As the data in Table 5 illustrate, virtually all of the difference between the highest municipal expenditure quintile and the lowest is accounted for (or more than accounted for) by the difference in municipal employee compensation per capita. This indicates that differences in employee compensation--not growth factors--may be the strongest driver of municipal expenditures.
Each of these conclusions works strongly against what one might expect from the Current Urban Planning Assumptions. This is illustrated by reviewing the data for the quintiles under each Current Urban Planning Assumption that would be expected to have the lowest expenditures per capita. Table 6 indicates that differences in employee compensation alone are more than sufficient to account for the differences in municipal expenditures per capita--whether by density, population growth, or municipality age.
In fact, the impact of increases in local government employee compensation has been far greater than the sprawl-based costs projected in Costs of Sprawl--2000. From 1980 to 2000, the gross additional local government employee compensation alone in the United States was nearly $2.2 trillion (in 2000 dollars)--or more than $105 billion per year. This is approximately 12 times the $9.1 billion average annual additional cost projected in Costs of Sprawl--2000.32
The generally higher spending levels of the older municipalities may be due to a process of "political entrenchment" that occurs with the passage of time. The large impact of local government employee compensation indicates that internal employee interests may be a principal factor driving municipal expenditures per capita. According to the nominal ranking analysis presented in Table 7, there appears to be a strong relationship between higher employee wages and salaries per capita and higher density, lower population growth rates, community age, and higher population--all of which are in opposition to what would be expected if the Current Urban Planning Assumptions were correct.
Perhaps reflecting such entrenchment, older municipalities have often been notably resistant to cost-effective management innovations such as privatization, competitive contracting, more flexible labor arrangements, and innovative management techniques.33 For example, the oldest quintile of municipalities had a general government expenditure level 23 percent higher than the youngest (Table 5).
It must be pointed out, however, that employee compensation is not likely to be the only cost function that could be exercising undo special-interest influence on the costs of local governments. Other political interests not quantified (and perhaps not quantifiable) may also exercise an impact on municipal spending.
Larger governmental units--which also tend to be more dense and older34--may be inherently more susceptible to special-interest capture, whether employee, business, labor, or other. Generally, it can be expected that the influence of individual voters would be less in larger jurisdictions and that special interests would be more likely to exert control. Larger jurisdictions would seem to provide economies of scale for lobbying. It would seem reasonable that where there is greater opportunity for special-interest control, government costs are likely to be higher. The data in Table 7 indicate that the highest wages and salaries quintile (quintile 1) has an average population that is more than 50 percent larger than average, and that the average population of each succeeding quintile is lower. The lowest wages and salaries quintile (quintile 5) has the lowest population--approximately one-half the average. This finding is counter to another widely held urban planning assumption: that larger units of government are more cost effective due to economies of scale.
All of this seems to indicate that municipal costs are more susceptible to overwhelming influence by political interests than they are to economics. Theoretical studies, such as Costs of Sprawl--2000 may suffer from what might be called the "length of pipe fallacy"--the assumption that labor rates, cost of materials, and the costs associated with apparently similar projects is the same in every local government jurisdiction in a metropolitan area.35 In fact, older, inner-city government labor rates are often higher than suburban rates: Overheads may be higher and certainly the operating environment can be more challenging. For example, expansion of an inner-city sewer system is likely to be far more costly than laying a new one in a greenfield area.
"Entrenchment" may have first been noted by Adam Smith in The Wealth of Nations. He pointed out that historical control of guilds in the older cities had produced a situation in which prices were lower in the suburbs, which were beyond the reach of the guilds. This kept prices in the older cities above market levels.36 Economist Mancur Olson similarly postulated that, as time goes on, political and special interests become more entrenched in older national governments.37 Stronger bureaucracies, more powerful employee organizations, strong local business interests, political interests, and more rigid operating procedures may have developed over a longer time period. These may force costs in older municipalities to be higher than they would be in newer municipalities.
An "entrenchment" theory of municipal finance would be consistent with the findings of economist Charles Tiebout, who argued that people tend to "vote with their feet"--to move to newer communities that better meet their desires and needs. Relative tax levels were an important component of this thesis, which characterized the new suburban communities as competing with one another for new residents.38
Our analysis indicates that the Current Urban Planning Assumptions are of virtually no value in predicting local government expenditures per capita. The lowest local government expenditures per capita are not in the higher density, slower growing, and older municipalities.
On the contrary, the actual data indicate that the lowest expenditures per capita tend to be in medium- and lower-density municipalities (though not the lowest density); medium- and faster-growing municipalities; and newer municipalities. This is after 50 years of unprecedented urban decentralization, which seems to be more than enough time to have developed the purported urban sprawl-related higher local government expenditures. It seems unlikely that the higher expenditures that did not develop due to sprawl in the last 50 years will evolve in the next 20--despite predictions to the contrary in The Costs of Sprawl--2000 research.
Wendell Cox, Principal of the Wendell Cox Consultancy in metropolitan St. Louis, is a Visiting Fellow at The Heritage Foundation and a Visiting Professor at the Conservatoire National des Arts et Metiers in Paris. Joshua Utt is a Ph.D. candidate in Economics at Washington State University and an Adjunct Fellow at the Discovery Institute in Seattle, Washington.
1. Urban sprawl is often thought of as an American phenomenon. In fact, sprawl has been occurring throughout the high-income world and much of the low- and middle-income world. Virtually all population growth in major Western European urban areas has been outside the urban cores for at least three decades, occurring mostly in suburban style settings.
9. This is not the case for other local units of general government, such as counties and townships. These generally include much rural (non-urban) land. As a result, density data for other local government units is not reflective of urban densities.
10. Metropolitan areas of more than 1,000,000 residents comprised approximately 58 percent of the nation's population in 2000 (2000 metropolitan definitions). The 49 such areas had a combined population of 163 million, out of a national total of 281 million (Table H-10).
13. This research uses multi-linear regression analysis. Independent variables (such as population density) were chosen. These were theorized to have some impact on municipal expenditures per capita (the dependent variable).
15. Current expenditures are the day-to-day costs of operations, such as employee compensation, materials and supplies, and professional service contracts. Capital expenditures are for construction and acquisition of assets, such as vehicles, data processing equipment, furniture, etc.
18. This factor (local direct general expenditures as a percentage of state government plus local government direct general expenditures) was included to capture the differences (by state) in expenditure distribution between state and local governments.
24. It has been suggested by some that older, more densely populated municipalities subsidize newer, more suburban municipalities. In fact, however, the nominal analysis indicates the opposite. The quintile of municipalities with the highest state and federal aid per capita average 45 years old ($852 annually per capita). This is nearly four times that of the second quintile (36 years and $218). The three lowest state and federal aid quintiles have average ages of from 26 to 31 years.
29. Wendell Cox and Samuel A. Brunelli, America's Protected Class III (Washington, D.C.: American Legislative Exchange Council, 1994), p. 29, Table C-3. No similar data have been published for localities.
30. It is also likely that differences in hourly employee compensation per capita would be an important determinant of differences in other government total expenditures, such as at the county, school district, township (and comparable governments) and special district levels.
31. Employee compensation is estimated using the gross local government wages and salaries data from the Census Bureau database, scaled downward to exclude utilities and education and increased by the average 24.5 percent cost of employer paid employee benefits. Because wastewater and water expenditures are small compared to overall municipal expenditures, it was not considered reliable to estimate wages and salaries for these functions using the same formula.
33. This is illustrated by the case of public transit. In 2001, none of the approximately 100 older transit systems (established before 1980 or descended from pre-1980 systems) in major metropolitan areas competitively contracted their bus systems. By contrast, 56 percent of the newer, largely suburban systems competitively contract their bus systems. See Wendell Cox, Performance Measures in Urban Public Transport, paper presented to the 8th International Conference on Competitive and Ownership in Public Transport, Rio de Janeiro, 2003, at www.publicpurpose.com/t8-gbc.pdf (June 15, 2004).
35. For example, in Los
Angeles--where many transit services are sponsored by newer,
suburban agencies--costs per hour of service are 46 percent lower
where provided under contract by agencies other than the core
transit system Wendell Cox, Competitive Participation in U.S.
Public Transport: Special Interests Versus the Public Interest,
paper presented to the 8th International Conference on Competition
and Ownership in Land Passenger Transport, Rio de Janeiro, 2003, at
t8-cc.pdf (June 15, 2004).