Using Social Costs to Distinguish Among Alternative

Using Social Costs to Distinguish Among Alternative Land‐Use Futures Jae Hong Kim, University of Illinois at Urbana‐Champaign, ([email protected]) Varkki G. Pallathucheril, American University of Sharjah ([email protected]) Brian Deal, University of Illinois at Urbana‐Champaign ([email protected]) More than before, urban planners now have the ability to systematically consider and communicate the future consequences of alternative public policy and investment choices. Using a growing number tools—such as LEAM (Deal 2003), UrbanSim (Waddell 2001), and WhatIf (Klosterman 2001)—planners can simulate future land‐use change associated with policy and investment alternatives in relatively large regions. Deal and Pallathucheril (2007) argue that for these simulations to be useful in thinking about the future and planning, it must be possible to usefully distinguish among them. Distinctions can be relatively easily drawn using the amount of development in each simulation and how this is distributed across space. More meaningful comparisons, however, would consider impacts across a range of concerns: traffic, energy consumption, the regional economy, hydrology and water quality, habitat fragmentation, etc. See, for instance, Allen (2001). Deal and Schunk (2004) describe how monetary and non‐monetary costs can be assessed for simulations of future land‐use using a dynamic spatial model of land‐use change. They describe a sub‐model developed within the Land‐use Evolution and impact Assessment Model (LEAM) framework that is based on a cost‐accounting framework developed by the U.S. Department of Transportation’s Federal Highway Administration called the Social Cost of Alternative Land Development Scenarios or SCALDS (Conrad and Seskin, 1998). Deal and Schunk use this sub‐model to assess the development, individual, communal, and social costs of two different future growth scenarios (low‐density and high‐density growth) in Kane county, Illinois. They find significant cost differences between the two scenarios and argue that equitable development must find a balance among individual, communal, and development costs. While social costs represent an important means of distinguishing among scenarios, the SCALDS accounting framework as applied by Deal and Schunk cannot distinguish among future growth scenarios that are similar in development density but that locate growth differently across space. Since they use a single set of values—the average across the region—for computing the cost of development across a region, it is unlikely to distinguish between the costs of similar amounts of development but differently distributed among the parts of the region. In this paper, we look at the problem of distinguishing among the social costs of alternatives that have different spatial distribution of land‐use change but similar total regional population and employment. We describe a series of experiments conducted using three simulations of future land‐use change in the Columbus‐Fort Benning region, an eight‐county area that straddles Georgia and Alabama in the United States and includes at its center the Fort Benning military installation. We first present some background material on computing social costs, the Columbus‐Fort Benning region, and the three regional land‐use 1 futures used in our experiments. We then describe how we went about computing social costs in sub‐
regions (6 Public Use Microdata Areas from the 2005 American Community Survey) and then aggregating to the entire region. Based on these different applications, we arrive at a number of findings: for instance, that the largest differences are produced when the amount of land being consumed is different (about 13% decrease in costs for a 33% decrease in land consumption). We conclude with a discussion of the additional data that might help better distinguish among scenarios. We also consider the other cost parameters that would be of interest to planners in the region. Background In this section we provide some background material on computing social costs, the Columbus‐Fort Benning region, and the three regional land‐use futures used in our experiments. Computing Social Costs The term social cost is used here to describe the true or full cost of an action, which includes costs registered through market transactions (direct costs) as well as those that are not (externalities). Berger (2008) attributes the notion of social costs to K. William Kapp (1950) even though it is widely, and Berger claims incorrectly, attributed to Coase (1960). Berger also cites Kapp’s explanation of social costs, which suggests a narrower definition limited to externalities: “All direct and indirect losses sustained by third persons or the general public as a result of unrestrained economic activities. These social losses may take the form of damages to human health; they may find their expression in the destruction or deterioration of property values and the premature depletion of natural wealth; they may also be evidenced in an impairment of less tangible values.” (Kapp, [1963] 1977, p. 13) The notion of social costs is of great interest particularly in ecological economics. It has been used in a wide variety of situations: from computing damages to third parties from new electricity generation technologies (Burtraw and Krupnick 1996), to the value of forestry (Gale and Gale 2006), to the disruption of construction projects (Gilchrist and Allouche 2005), to siting of landfills (Sasao 2006). Gilchrist and Allouche (2005) posit that social costs can be valued using two types of methods: direct and indirect valuation techniques. Direct valuation techniques involve computing the cost of loss (productivity, human capital) or replacement (of a resource). Indirect valuation techniques include hedonic pricing (using market value as a proxy) and contingent valuation methods (using surveys of willingness‐to‐pay or willingness‐to‐accept). SCALDS (Social Cost of Alternative Land Development Scenarios) is an accounting framework for estimating monetary and non‐monetary costs associated with urban land development at the metropolitan scale (Conrad and Seskin, 1998). Sponsored by the U.S. Federal Highway Administration, SCALDS is a spreadsheet model that accounts for 1.) physical development, including land consumption, existing and projected housing mix, regional employment, and local infrastructure capital and operating costs, 2.) annual peak and nonpeak cost of travel on a passenger mile traveled (PMT) basis, and 3.) non‐
2 dollar denominated costs such as the air pollution and energy consumption. The primary impacts of urban land use transformation are captured through changes in the costs of providing infrastructure services to households, businesses, and governments. The inclusion of specific cost factors is based on their significance and on whether the factor can be attributed to and/or measure alternative development patterns. This is logic is represented in Figure 1 (Source: Deal and Schunk 2004). Deal and Schunk (2004) transform Conrad and Sesskin’s (1998) spreadsheet model into an integral part of LEAM. This enables social costs to be computed along with simulations of land‐use change rather than externally in the spreadsheet. They demonstrate that social costs computed using this approach for less and more dense development in a particular region are different and intuitively so. Since they used region‐wide values for many of the model parameters, however, this approach cannot distinguish between development patterns that involve similar amounts of land consumption but are very different 3 spatially. If social costs are computed at a finer spatial resolution, and if model parameters can reflect variations among spatial units at this resolution, then will this approach be able to make such distinctions? Being able to do so could provide valuable support for making public policy. The Columbus­Fort Benning Region To understand how a finer spatial resolution might affect computations of the social costs of different land‐use patterns, we conducted a series of experiments using three simulations of future land‐use change in the Columbus‐Fort Benning region. This is an eight‐county area that straddles Georgia and Alabama in the southeastern United States and includes at its center the Fort Benning military installation. (See Figure 1.) Muscogee and Lee counties are relatively urbanized, while the six other counties are relatively rural. In the year 2000, just over 408,000 people lived in this region. This population represents a 12% increase since 1990; the number of households increased at a quicker pace (22% over the same time period). Columbus, Georgia, is the largest city in the region, with a population over 185,000 in 2000. Some other regional characteristics in the year 2000 include: • Median household income in the region was $33,140 • Eighteen percent of regional household incomes were below the poverty level (44% higher than the national average). • Eighty percent of the population in the region had a high school diploma. • The average commute time to work was 25.5 minutes. 4 Future Development Scenarios Regional growth in population and jobs were forecast using a regional econometric input‐output model (Sarraf et. al, ). These forecasts are based on how the national and regional economies are expected to grow over the next thirty years, and include an expected increase in troops and civilian workers at Fort Benning as a result of proposed realignments and closures of military bases in the US. Employment is expected to increase by 125,000 and population is projected to increase by 113,000, for a total regional population of 522,000 by the year 2030. Using LEAM (see Appendix A for a brief description of LEAM), future land‐use patterns in the region were simulated for a number of public policy and investment choices. Of these, three were selected for this experiment and are described below. Baseline Development was selected to provide a frame of reference. ACUB Protection was selected because a similar amount of land is consumed as in Baseline Development but is distributed across the region in a different way. Compact development was selected to examine the effect of higher density development, and essentially replicates the comparison made by Deal and Schunk (2004). Scenario I: Baseline Development This scenario assumes that current development trends will continue to play out into the future. Based on population forecasts—along with forecasts of average household size, housing vacancy and abandonment, and change in average housing lot size—over 43,000 acres of new residential development are expected by the year 2030. At the same time, approximately 10,000 acres of new commercial development are expected. Figure 2 shows future land consumption (aggregated by square‐
mile section) forecasted to occur in the region by 2030 (the darker blue areas indicate where more growth is expected). Municipal boundaries are outlined in red. The map indicates that there are several significant growth areas in the region, including north of the installation (along US Route 80), north of Columbus (along US Route 27/State Route 85 and State Route 315), southwest of Columbus (along US Route 431 and State Route 165), and the Auburn/Opelika area. 5 Figure 3 shows how land consumption in each of the eight counties is projected to grow over this time period. A significant jump in urban growth is expected in the region in 2008‐2009 as more troops and civilian workers relocate to Fort Benning. Following this jump, Lee County continues to have tremendous growth. 6 Over one‐third (38%, over 17,000 acres) of the new development in the region from 2000‐2030 is projected to occur in Lee County, and another 20% in Muscogee County. Harris and Russell County are also expected to have a significant share of regional growth (19% and 15%, respectively). Scenario II: ACUB Protection This scenario assumes that Army Compatible Use Buffers (ACUBs) are designated to prevent development in some areas adjacent to the military base so that they can be used for military training and protecting wildlife habitat. The baseline scenario shows a significant amount of land consumed in these buffer areas. If, as is assumed in this scenario, the amount of land consumed remains the same, then this policy will move growth outside of these buffer areas to other attractive areas in the region. Figure 4 compares growth in this scenario with the baseline scenario. Blue indicates areas where more land is consumed in the baseline scenario, while brown indicates areas where more land is consumed in the ACUB protection scenario. (Again, darker areas indicate greater consumption.) As a result of ACUB protection, growth shifts from buffer areas, particularly the one along the northern border of the installation, to parts of the Columbus and Auburn/Opelika areas. (Interestingly, growth does not shift elsewhere around the base, to areas that are not protected by the buffer, except near the northwest boundary of the installation.) As a result of these shifts, Lee County sees a greater share of land consumption in this scenario than in the baseline scenario, as do Harris and Russell Counties; Muscogee County experiences a smaller share of land consumption. 7 Scenario III: Compact Development This scenario assumes that policies encouraging higher density development are established, leading to one‐third less development acres to meet the needs of the same amount of new households and jobs. Policies could include encouraging smaller lot sizes, clustered development, and more multi‐family housing. Figure 5 compares growth in this scenario with the baseline scenario. Blue areas have more land consumption in the baseline scenario. (Darker blue areas indicate greater consumption.) There are no areas in which more land is consumed in this scenario. As a result of compact growth policies, there is one‐third less growth in the region but this decrease is not uniform across the region. Muscogee and Russell counties have a greater share of growth, while Lee, Harris, Talbot, and Stewart Counties have smaller shares. Summing Up Table I compares each county’s share of land consumption in each of the three scenarios. The shares of regional land consumption vary in a very tight range in most counties: from about 4 percentage points in Muscogee county to about half a percentage point in many others. 8 Baseline
Development
ACUB
Protection
Compact
Development
Lee
Muscogee
Harris
Russell
Talbot
Stewart
Marion
Chattahoochee
36.7%
20.4%
18.6%
14.5%
3.1%
2.5%
2.3%
2.0%
37.8%
18.8%
19.0%
14.9%
2.9%
2.5%
2.3%
1.8%
37.2%
22.9%
17.3%
15.0%
2.4%
1.9%
1.6%
1.6% Since ACUB Protection appears to move a small share of regional development from one urbanized area to another (Muscogee to Lee) and to a relatively rural area (Harris), we expect that this policy choice would slightly increase the social cost relative to Baseline Development. Based on Deal and Schunk’s (2004) findings, we expect that Compact Development would have lower social costs than Baseline Development. Method Baseline
Development
ACUB
Protection
Compact
Development
To test these hypotheses, we computed two sets of social costs (using regional average and using spatially variable parameter values) for the three scenarios. This set up is represented in Table II. If social costs computed in I, II, and III are all greater than or all less than social costs computed in IV, V, and VI, then we can conclude that using regional averages instead of spatially varying parameter values does have an impact. If social costs in I and IV are greater than in III and VI respectively, then we have replicated Deal and Schunk’s (2004) finding. If social costs in I and IV are different from II and V respectively, then we can conclude that this approach does distinguish among scenarios that have similar land consumption but different spatial patterns. Regional
averages
I
II
III
Spatially
varying
IV
V
VI
Parameter
values
We derived parameter values for computations IV, V, and VI from the 2005 American Community Survey (ACS). Six ACS‐PUMAs (Public Use Microdata Areas) that completely include the region are shown in Figure 6; two of them each contain two of the region’s counties while the others contain a single county. The 2000 Census would have provided much of the data at a finer spatial resolution, but we decided to 9 use the 2005 ACS because it is more current. Two parameter values were imputed from 2000 data. Population and job densities are computed using land cover data; since the latter are only available for 2000, population and jobs were drawn from the 2000 Census. For the region as a whole, we took a weighted average of these parameter values. Table III shows the six PUMAs and how key SCALDS parameters vary among them. The land‐use parameters display a higher degree of differentiation as compared to the transportation parameters. So, for instance, population densities range from 2.7 in Steward to 10.7 persons per acre in Muscogee county; job densities range from 1.9 to 10.4 jobs per acre in the same two counties. Fifty‐six percent of the housing in Harris and Talbot counties consists of single‐family houses on lots greater than 1 acre, compared to 8.7% in Muscogee. Transportation parameters vary more tightly across the PUMAs. For instance, mode share of single‐occupancy vehicles range only from 82.3% in Lee county and 84.4% in Russell county. Average commuting times do not conform to this pattern in transportation parameters and range from 17.3 minutes in Muscogee to 26 minutes in Harris and Talbot. 10 Key SCALDS Parameters by Sub-Region
Lee
Russell
Harris-Talbot
Muscogee
ChattahoocheeMarion
Stewart
AL01700
AL01800
GA02900
GA03000
GA03100
GA04100
2000 Population Density
(Person/Acre)
8.95
6.82
3.72
10.66
3.47
2.71
2000 Job Density (Job/Acre)
7.71
4.16
6.09
10.35
3.27
1.89
SF-detached; 1acre or more
20.4%
35.7%
55.7%
8.7%
37.6%
40.8%
SF-detached; Less than 1 acre
42.9%
43.9%
32.2%
58.0%
40.8%
47.3%
5.2%
1.1%
1.2%
1.5%
1.4%
0.9%
MF; 2~4 apartments
10.3%
9.2%
5.1%
10.3%
11.3%
6.9%
MF; 5~19 apratments
13.9%
9.8%
4.2%
14.8%
7.8%
3.5%
MF; 20 or more apartments
7.3%
0.4%
1.6%
6.6%
1.1%
0.7%
2005 Avg Commute Time; All
Modes (Minutes)
20.4
19.8
26.0
17.3
19.9
20.8
2005 Avg Commute Time; Singleoccupancy Vehicle (Minutes)
20.3
19.9
25.3
18.3
22.4
21.8
Single-occupancy Vehicle
82.3%
84.4%
83.6%
83.1%
83.3%
82.4%
Multi-occupancy Vehicle
12.3%
13.1%
13.1%
11.9%
10.7%
13.3%
Transit
1.4%
0.1%
0.1%
1.3%
0.1%
1.3%
Bike+Walk
2.8%
1.3%
1.6%
2.1%
5.1%
2.0%
Others
1.3%
1.0%
1.6%
1.5%
0.8%
1.0%
Sub-Region
PUMA Code
2005 Housing Mix
SF-attached
2005 Commute Mode
We made a key assumption in using different parameter values in different parts of the region: new development in any location will take on the characteristics of that location. We believe this assumption is reasonable in the short term because absent public policy constraints market forces generally tend to perpetuate existing conditions; the relative homogeneity of large parts of US regions is a testament to this. In the long term, this assumption may not hold, but there is not a basis for systematically forecasting changes in these characteristics. These parameter values might change in a number of ways that would have to be covered using a variety of scenarios. Results Table IV shows a comparison of different types of social costs when spatially varying parameter values are used. As expected, the differences between Baseline Development and ACUB Protection are very slight, if negligible, and appear to support our hypothesis in this regard. The differences are almost identical across all cost estimates. At the same time, there are significant reductions in Compact Development compared to Baseline Development. This also is expected and replicates Deal and Schunk’s (2004) findings. 11 ACUB Protection
Baseline
Value
Value
Compact Development
Ratio to Baseline
Value
Ratio to Baseline
1. Residence & Production Related Costs
Increase in Annual Private Cost for Water
(2005~2030)
Increase in Annual Private Cost for Sewer
(2005~2030)
Increase in Annual Private Cost for Storm
Water Systems (2005~2030)
Local Infrastructure Costs … Streets,
Utilities, and Schools (2005~2030)
Increase in Annual Non-Transportation
Energy Cost (2005~2030)
Increase in Land Consumption for Urban
Uses in Acres (2005~2030)
$40,638,399
$40,668,710
100.1%
$35,381,862
87.1%
$40,768,931
$40,794,258
100.1%
$35,615,976
87.4%
$4,076,893
$4,079,426
100.1%
$3,561,598
87.4%
$2,131,277,241
$2,140,286,156
100.4%
$1,636,658,184
76.8%
$278,676,411
$279,159,994
100.2%
$224,106,714
80.4%
46,325
46,647
100.7%
27,470
59.3%
$1,313,600,908
$1,319,177,927
100.4%
$1,305,573,756
99.4%
$825,930,788
$829,411,083
100.4%
$820,918,021
99.4%
HC Annual Emission in Tons (2035)
4,112
4,129
100.4%
4,087
99.4%
CO Annual Emission in Tons (2035)
30,782
30,913
100.4%
30,594
99.4%
NOx Annual Emission in Tons (2035)
2,505
2,516
100.4%
2,490
99.4%
SOx Annual Emission in Tons (2035)
93
94
100.4%
93
99.4%
PM Annual Emission in Tons (2035)
133
134
100.4%
132
99.4%
2. Transportation Related Costs
Increase in Annual Transportation Cost
with the Value of Time (2000~2035)
Increase in Annual Transportation Cost
without the Value of Time (2000~2035)
All monetary values: in 2000 Constant Dollars
Table V shows social costs in Baseline Development when regional averages are used. Comparison with social costs computed using spatially varying parameter values indicates that using regional averages produce slightly lower estimates. Residential and production related costs are slightly lower (less than a percentage point) with the exception of local infrastructure (which is about 5 percentage points lower). Estimates of transportation‐related costs are much lower (from 4 to 5 percentage points). 12 Conclusion
While our experiments in general support our initial hypotheses, the magnitude of differences between
Baseline Development and ACUB Protection were not as large as might be expected. ACUB Protection
was selected for this experiment because it had the greatest differences from Baseline Development of
scenarios that involved similar amounts of land consumption. We probably should have generated a
scenario that had even larger differences.
Some additional considerations come to mind. First, this outcome could also mean that the ColumbusFort Benning region does not vary too much in parameter values in places where much of the
development is occurring. Second, the internal logic of SCALDS can be examined to see if this is
preventing greater differences emerging.
The most valuable finding from our experiments is that spatially differentiated computation of social
costs produces higher estimates of social costs. These estimates probably better reflect the true costs
13
because regional averages assume that the existing ratio between urban and suburban development will continue into the future. This is not a reasonable assumption and will cause some new urban development to be assessed as suburban development and as a result underestimate the social cost. Of course, the lack of data will likely prevent differentiation beyond a certain spatial resolution. The knowledge produced by computations of social costs can be useful in public policymaking and planning. For instance, implementing ACUB Protection is often resisted by the public, especially land owners with the buffers. The finding that preventing future development within these buffer areas would allow planners to make the case that ACUB Protection is desirable: the region benefits from having the installation meet its mission without increasing social costs. References: Allen, E. 2001. Berger, S., K. 2008. William Kapp's theory of social costs and environmental policy: Towards political ecological economics, Ecological Economics doi:10.1016/j.ecolecon.2008.05.012 Burtraw, D. and A. J. Krupnick. 1996. The second‐best use of social cost estimates. Resource and Energy Economics 18: 467‐89. Camagni, R., M. C. Gibelli, and P. Rigamonti. 2002. Urban mobility and urban form: the social and environmental costs of different patterns of urban expansion. Ecological Economics 40: 199–216. Coase, R. H. 1960. The problem of social costs. Journal of Law and Economics 3, 1–44. Conrad, L. M. and S. N. Seskin. 1998. The Costs of Alternative Land Use Patterns (SCALDS). Department of Transportation Federal Highway Administration, Washington DC. Deal, B. 2003. Sustainable Land‐Use Planning: The Integration of Process and Technology. Doctoral dissertation. University of Illinios at Urbana‐Champaign. Deal, B. and D. Schunk. 2004. "Spatial Dynamic Modeling and Urban Land Use Transformation: A Simulation Approach to Assessing the Costs of Urban Sprawl", Ecological Economics 51, 79‐ 95. Deal, B. and V. Pallathucheril. 2007. Developing and using scenarios. In Modeling to Link Forecasts, Scenarios, and Plans, Lew Hopkins, and Marisa Zapata (eds). Cambridge, MA: Lincoln Institute of Land Policy. Gale, R. and F. Gale. 2006. Accounting for social impacts and costs in the forest industry, British Columbia. Environmental Impact Assessment Review 26, 139–55. Gilchrist, A. and E. N. Allouche. 2005. Quantification of social costs associated with construction projects: State‐of‐the‐art review. Tunnelling and Underground Space Technology 20: 89–104 Kapp, K. W. 1950. The Social Costs of Private Enterprise. , Cambridge, MA: Harvard University Press. 14 Kapp, K. W. [1963] 1977. The Social Costs of Business Enterprise (second enlarged edition of The Social Costs of Private Enterprise (1950)). Nottingham, UK: Spokesman. Klosterman, D. 2001. Sasao, T. 2005. An estimation of the social costs of landfill siting using a choice experiment. Waste Management 24: 753–62. Waddell, P. 2001. Appendix A: LEAM LEAM has been described elsewhere (Deal, 2003; Deal and Schunk, 2004; Deal and Pallathucheril, 2007) and so we will simply quote from one of these sources: “In LEAM, a region is represented as a grid in which each cell is 30m x 30m, a convenient resolution for simulating land‐use change because each grid cell is the equivalent of a quarter‐acre lot. Simulations of change across this landscape are run in one‐year time steps. A discrete‐choice model controls whether land use in each grid cell is transformed from its present state to a new state (residential or industrial and commercial use) in a given time step. This transformation is based on a development score computed in each time step for each cell based on a number of factors called drivers because they affect whether or not land‐use change takes place. Some drivers enhance a cell’s development score (e.g., proximity to cities, employment centers, roads, highways), others might diminish a cell’s score (e.g., location within wetlands, floodplains), while yet others might have variable impacts (e.g., no slope or very high slope might diminish a score, while in‐between values might enhance a score). Characteristics of surrounding cells (e.g., extent of surrounding development, type of development) also alter a cell’s score. The impact of each driver on a cell’s development score is calibrated based on current land‐use patterns in the region. In a given time step, the regional demand for new development and the development score associated with the cell determine whether or not the land use of that cell is transformed. A cell is likely to be transformed if it is available for development and has a high enough development score to compete successfully to satisfy the regional demand for new development. Some cells with low scores might also be transformed because all transformations occur if the score is greater than a random number generated for each cell. Many factors interacting in complex ways produce outcomes that are plausible.” (Deal and Pallathucheril 2007, pp. XX‐XX) LEAM is used to simulate different land‐use futures associated with different public policy and investment choices in the following manner: “Since the factors that contribute to a cell’s development score represent many causes of land‐use change, LEAM can play out the consequences of changes in these factors. 15 For instance, changing the road infrastructure in a region alters a cell’s proximity to various attractors. This, in turn, alters the development score and thus the relative competitiveness of individual cells; the resulting future land‐use pattern is often very different from what results when the infrastructure remains unchanged. Consider another example: if maximum residential densities are enforced in parts of a region, then the development score of a cell in these parts is reduced if a certain number of cells in its vicinity are developed. This diminished development score again results in a different future regional land‐use pattern than would be the case otherwise. As different combinations of public investment and policy choices alter development scores in different ways, LEAM works out the land‐use consequences of these choices.” (Deal and Pallathucheril, 2007, pp. XX‐XX) 16