- Email: [email protected]
Maximizing conservation and in-kind cost share: Applying Goal Programming to forest protection
Journal of Forest Economics 18 (2012) 207–217
Contents lists available at SciVerse ScienceDirect
Journal of Forest Economics journal homepage: www.elsevier.de/jfe
Maximizing conservation and in-kind cost share: Applying Goal Programming to forest protection Jacob R. Fooks ∗, Kent D. Messer University of Delaware, Department of Applied Economics and Statistics, United States
a r t i c l e
i n f o
Article history: Received 5 February 2011 Accepted 3 April 2012 JEL classification: C6 Q15 Q23 Keywords: Goal Programming Multi-objective programming Conservation optimization Forest conservation Environmental services In-kind cost sharing Matching grants
a b s t r a c t This research evaluates the potential gains in benefits from using Goal Programming to preserve forestland. Two- and threedimensional Goal Programming models are developed and applied to data from applicants to the U.S. Forest Service’s Forest Legacy Program, the largest forest protection program in the United States. Results suggest that not only do these model yield substantial increases in benefits, but by being able to account for both environmental benefits and in-kind partner cost share, Goal Programming may be flexible enough to facilitate adoption by program managers needing to account for both ecological and political factors. © 2012 Department of Forest Economics, Swedish University of Agricultural Sciences, Umeå. Published by Elsevier GmbH. All rights reserved.
Introduction Conservation continues to be a major policy objective in the United States and internationally. Examples of programs which have been implemented to achieve conservation objectives include the U.S. Department of Agriculture’s Conservation Reserve Program (CRP) which has yearly enrollment of over 30 million acres (USDA, 2009). U.S. states and county programs have spent over $2 billion protecting over 1 million acres with permanent agricultural conservation easements in a recent 25year period (American Farmland Trust, 2010), while the entire set of open-space referenda authorized
∗ Corresponding author. E-mail address: [email protected] (J.R. Fooks). 1104-6899/$ – see front matter © 2012 Department of Forest Economics, Swedish University of Agricultural Sciences, Umeå. Published by Elsevier GmbH. All rights reserved.
http://dx.doi.org/10.1016/j.jfe.2012.04.001
208
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
$31 billion between 1996 and 2009 alone (Trust for Public Lands, 2009). As of 2005, 37 million acres had been protected by the 1667 private land trusts in the United States (Aldrich and Wyerman, 2006). The most recent U.S. Farm Bill (2008–2012) allocates $11.7 billion for conservation1 including expenditures of up to $1.8 for the CRP in 2010 (USDA, 2009), while in the European Union an anticipated D35.4 billion will be spent on “agri-environmental” programs between 2007 and 2013.2 Perhaps the largest conservation program in the world is China’s Sloping Land Conversion Program with an estimated budget of $48 billion (Xu et al., 2010). Frequently, these conservation programs involve some type of in-kind cost share or matching funds component where one program seeks to leverage additional resources from partner agencies, organizations, or individuals to achieve their conservation objectives (Kotani et al., 2010). While researchers have long advocated for the inclusion of costs as part of the selection process through either benefitcost targeting (Babcock et al., 1997) or integer programming (Underhill, 1994), the literature has not addressed issues around how to best incorporate in-kind cost sharing from partners into the selection process. While in-kind cost share contributions are implicitly accounted for through the lower project costs facing the funding organization, this fails to take full advantage of the additional information potentially provided by the size of the in-kind cost share, such as the level of commitment of the partner organization and the political benefits of being able to use program funds to leverage resources from other organizations, agencies, and individuals. This research addresses this potential concern by developing a two- and three-dimensional Goal Programming (GP) model that seeks to optimize both conservation outcomes and partner in-kind cost share contributions. This model is applied to the 2008 applicants to the U.S. Forest Service’s Forest Legacy Program (FLP), the largest single program dedicated to preserving productive forest land in the United States. Results from the models show that GP offers results that are superior to the program’s current selection mechanism and can also yield cost-effective outcomes that may be more practical than traditional optimization approaches so that they could be more attractive for adoption for program managers. Literature review In spite of the substantial amount of money spent on land conservation programs, most conservation programs are typically faced with more potential projects than they are able to fund. A vast body of economic literature has grown up around conservation programs aiming to evaluate the efficiency achieved with the vast public and private resources invested into these programs. It has been widely acknowledged that the selection methods typically employed by land conservation programs are suboptimal. As part of the selection process, programs frequently assign parcels with scores based on a variety of scoring systems, such as ratings from expert panels, standardized scoring system such as the Natural Resource Conservation Service’s Land Evaluation and Site Assessment (LESA) score for agricultural land, or the Environmental Benefit Index (EBI) originally designed for CRP. These parcel-specific scores are then used by the selection mechanism, typically a Benefit Targeting (BT) algorithm, which is essentially a greedy algorithm where projects are funded from highest to lowest benefit score until funds are exhausted. This approach has been subject to substantial criticism as being very inefficient, and it has been repeatedly proposed that adopting alternative mathematical optimization methods would lead to superior outcomes (for example, Babcock et al., 1997; Polasky et al., 2001; Wu et al., 2000). One alternative to BT is Cost Effectiveness Analysis (CEA). With CEA projects are ranked in terms of environmental benefits achieved per dollars spent on the project. These cost effectiveness ratios are then ranked and projects selected by highest ratio until funds are exhausted. CEA performs quite well as compared to BT and are easy to implement, but can still achieve suboptimal results under certain
1 U.S. calculations based on data from Claassen R., Conservation Policy Briefing Room, Economic Research Service, U.S. Department of Agriculture. accessed February 2010. 2 E.U. calculations based on data for Axis 2, Section 214, of the second pillar of the Common Agricultural Policy from: European Union Directorate-General for Agriculture and Rural Development, “Rural Development in the European Union: Statistical and Economic Information Report 2009.” (EU, Brussels, 2009).
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
209
circumstances (Messer, 2006). Additionally, CEA does not allow for additional constraints to be easily placed on the problem. A second alternative is Binary Linear Programming (BLP), also referred to as Binary Integer Programming. BLP is a variation on linear programming problems where there is a binary (0/1) variable representing the decision to fund a project. BLP maximizes the benefit score of each project, which is multiplied by the binary variable representing the funding decision, subject to the project costs times the funding decision variables and any other constraints that are pertinent. Through the use of the branch-and-bound algorithm, BLP considers every possible combination of projects and selects the best combination. BLP typically performs somewhat better that CEA. In spite of the promise of substantial increase in environmental benefits offered by both CEA and BLP land conservation programs have been very hesitant to adopt either in their practices. To our knowledge, the only organization that has used either of these techniques is the Baltimore County’s Division of Environmental Protection and Resource Management (DEPRM), which has used CEA for its agricultural protection program since 2007. In the first three years of use, DEPRM protected an additional 680 acres worth $5.4 million relative to what they would have done with their previous BT given the same budget amounts (Kaiser and Messer, 2011). Researchers have speculated about the lack of adoption of cost-effective and optimization techniques (i.e. Prendergast et al., 1999; Pressey and Cowling, 2001). For example, Prendergast et al. (1999), as the result of an informal interviewing process with a wide sample of ecologists and land managers, identified three main barriers for adoption: lack of knowledge, lack of resources, and real or perceived shortcomings in the methods. These authors conclude that ultimately for conservation optimization to be widely adopted more work needs to be done to cater research to the needs of real life conservation practice. A recent survey of Maryland county level conservation program administrators suggested that managers not only lacked knowledge of BLP and CEA, but managers also reported not considering cost to be a major priority in the selection process and lacking incentives that would lead them to adopt it with their program (Chen, 2010). Another possible explanation for this lack of adoption is that managers need to be able to consider numerous political and strategic objectives in their decision making process. Their duties include not merely maximizing benefit scores, but also defending the “value” achieved from donor, funding agency, or tax payers’ money, ensuring that applicants get a fair deal from a transparent decision mechanism, and distributing funds in a manner perceived as equitable. BLP is unable to clearly address many of these duties. These can be thought of as secondary or operational objectives that do not immediately impact the primary goal of protecting high quality land but still may be important factors in the decision making process. One way to incorporate these into the problem is as constraints, but this offers little ability to consider sensitivity or alternatives to the single solution provided. This is the basis of the approach used in Önal et al. (1998) which considers both environmental and equity concerns for watershed management using a mathematical programming model. This model maximizes total profit across a watershed with a chance constraint on chemical runoff levels to take into account the stochastic nature of rainfall, and a constraint on the equity of program impact as measured by an index measurement of deviation from a uniform loss sharing level. These constraints are varied to examine the tradeoff between income, pollution, and equity losses. A second option is to format the problem as a GP problem with these secondary objectives included as weighted goals. There have been several applications of GP in this field, including balancing economic and biological objectives over short term and long term time frames in fishery management (Drynan and Sandiford, 1985; Mardle and Pascoe, 1999; Mardle et al., 2000) optimizing environmental, social, and economic goals in energy production (Silva and Nakata, 2009); and the management of public water resources (Neely et al., 1977; Ballestero et al., 2002). Önal (1997) considers an approach similar to GP in forest management. A model is employed which, instead of minimizing deviations from a goal as is done in GP, uses constrained deviations from a goal to maximize discounted future harvest value while maintaining a minimum value of a species diversity index. To our knowledge GP has not been proposed in the planning of land conservation decisions. In-kind cost sharing is an example of a secondary objective. According to Kathryn Conant, National Forest Legacy Program Manager, the US Forest Service is interested in approaches that could both maximize forest benefits and get the most from leveraged partner cost sharing funds (Conant, 2008).
210
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
The use of cost sharing (or matching funds) is very common in conservation and other publically funded programs. While the way in which such requirements implemented can vary, typical approaches include requirements that the participating organization cover some percentage of the total cost of the projects themselves (for instance, the USDA’s Conservation Reserve Program (USDA, 2010)), or a system where the amount of in-kind cost sharing offered weights the ranking for amount of funds awarded or priority to receive funds (such as is the case with some counties that apply for funds for water quality enhancements from the Pennsylvania’s Dirt and Gravel Road Program (Bradford County Conservation District, 2006)). Programs that require some sort of cost sharing include the USDA’s Conservation Reserve and Forest Legacy Programs, the U.S. Fish and Wildlife Service’s National Wetlands Conservation Grant Program, and numerous other federal, state, and private conservation funding programs. Several justifications are typically offered for requiring some sort of in-kind costs share in the selection mechanism for conservation programs. A primary justification is that by asking applicants to commit to pay for some of the costs themselves, this ensures that applicants are committed to the project and consider it worth the effort to conserve. In other words, the applicant is willing to “put their money where their mouth is” for the project. Another justification is that cost sharing helps extend the reach of the conservation program as it can be considered a type of discount on the total project price. From the point of view of the funding agency, this could mean that there would be more money available to fund additional projects and also help increase the number of projects funded with the program’s funds. In the literature, in-kind cost sharing can be considered a type of matching grant program where the granting agency agrees to pay some portion of costs up to a percentage of the total project costs. Theoretically, matching grants are seen as a mechanism to correct for externalities and spillovers in federalized agency structures. Oates (1999) points out that the local benefits from a project under consideration may not justify funding the project for the local agency. However, the spillover benefits to other areas make it attractive to society as a whole. The matching grant offered by an outside body would represent a sort of Pigouvian subsidy to pay the local agency for the external benefits obtained in other jurisdictions. Bucovetsky et al. (1998) offers an informational argument for matching grants similar to the “money where mouth is” justification. Ideally, a government should distribute funds to the regions which value public services the most. The matching grant severs as a mechanism to reveal the private value of funding to the applicants. Current evidence on the effect of matching grants and cost sharing has not offered strong results. Baker et al. (1999) found that instituting a matching grant system in the Canada Assistance Plan lowered expenditures growth by 8–9 percentage points as provinces became responsible for a portion of program costs. Using simulations, Borge and Rattsø (2008) found that compared to block grants matching grants decrease expense, but lead to unstable service provision over time. Chernick (1998) found that the conversion from matching grants to fixed block grants by U.S. federal welfare funding programs had substantial variation across states, but generally led to a reduction in benefits. For an in depth discussion of matching grants and their implementation see Boadway and Shah (2007). In the following section we develop a GP model which incorporates in-kind cost share as a secondary objective. Following that we apply this weighting scheme to the 2008 project set for the USDA’s Forest Legacy Program.
Model development The basic BLP program that seeks to maximize environmental benefits, B*: Max : B∗ = s.t. :
xi ∈ {0, 1}
bi xi
pi xi ≤ T
(1) (2) (3)
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
211
where xi is a binary variable indicating whether project i is chosen, bi represents the environmental benefit score as determined by the Forest Legacy expert panel, and pi the funding request for project i. T is the total budget available for the selection program for the year. Similarly, denoting ci as the dollar amount of cost share offered for project i, a BLP program can be set-up to maximize the amount of in-kind cost share, C*: Max : C ∗ = s.t. :
ci xi
(4)
pi xi ≤ T
(5)
xi ∈ {0, 1}
(6)
A two-objective (two-dimensional) GP model can be developed as an extension of these two problems as the set-up to minimize a weighted sum of percent deviations from the maximum feasible value of in-kind cost share, C*, and environmental benefits, B*. Using these two extremes as targets, the GP problem can be set up and then the weighting between the two can be varied parametrically to examine the tradeoffs. This problem is stated as follows:
Min : Z = s.t. :
dC−
C∗
+ (1 − )
dB−
(7)
B∗
bi xi − dB+ + dB− = B∗
(8)
ci xi − dC+ + dC− = C ∗
(9)
pi xi ≤ T
(10)
xi ∈ {0, 1}; di+ , di− , ≥ 0
(11) +/−
are chosen to measure the achieved level of objective N (in this case The deviation variables, dN ni xi , from N*. In the objective function, the negative deviations are either B* or C*), determined by divided by the target value so that each represents a percentage deviation, and thus are of comparable magnitudes. The objective is to minimize the total deviations from the joint optima based on the priority weighting. The weighting factor, , can be varied parametrically from 0 to 1. This represents the percentage of priority that is given to in-kind costs, such that when is 0, 100% of the priority will be given to maximizing environmental benefits and when is 1, 100% of the priority will be given to maximizing in-kind cost share. Thus when is 0.3, then 30% of the weight will be on in-kind cost share and 70% will be on environmental benefits. As will be discussed below in the result section, the model can be expanded to consider more than two objectives. Since managers of conservation programs typically are concerned about a range of objective which may have differing degrees of exclusivity, below we develop a three-dimensional model that incorporates total environmental benefits, in-kind cost share, and the total amount of acreage preserved.3
Min : Z = C s.t. :
dC− C∗
+ (B )
ci xi − dC+ + dC− = C ∗
bi xi − dB+ + dB− = B∗
dB− B∗
+ (A )
dA− A∗
(12) (13) (14)
3 Note that as the number of objectives increases constructing a systematic weighting scheme becomes rather complex and the solutions set is less amenable to visual representation. None the less, the technique can still be useful to explore alternate solutions and the effects of changing prioritization schemes.
212
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
ai xi − dA+ + dA− = A∗
(15)
pi xi ≤ T
(16)
xi ∈ {0, 1}; di+ , di− , ≥ 0
(17)
Case study: Forest Legacy Program The Forest Legacy Program (FLP) is a program administered by the US Forest Service to support the acquisition of conservation easements and other voluntary protection mechanisms on privately held forest land. It was established in 1990 and is the largest single protector of privately held forest land in the US. As of 2010, FLP has helped to protect nearly 2 million acres across 41 states and Puerto Rico. The FLP works primarily with state agencies to acquire easements on privately held land to protect timber supply, wildlife habitat, soil and watershed protection, aesthetics, and recreational opportunities. Participating landowners typically receive some percentage of the land value and a decrease in their tax burden in exchange for the right to develop the land. In order to participate in the program landowners must develop a comprehensive multiple resource management plan. The USDA requires states to provide at least 25% of the total project costs, however no cap is placed on the percentage of in-kind cost sharing that can be offered by the applicants and some applicants to the FLP in the past have submitted in-kind cost shares of nearly 90%. The average percentage of in-kind cost share offered in 2008 was 36.3%. Applicants receive no additional benefit from offering excess in-kind. FLP selects projects using an environmental benefit score determined by a panel of experts. Each panel member is asked to score each project from 0 to 30 based on importance, threat, and strategic value. Panel members are not allowed to evaluate projects from the states in which they work. The three scores are summed such that each project is receives a score between 0 and 90 from each reviewer. The FLP administrators discard the lowest and highest scores and then average the remaining scores to assign the project a final score. The projects are selected for funding using the BT algorithm. This research considers the 83 projects submitted to the FLP in 2008 from 44 states, Puerto Rico, and the US Virgin Islands. Table 1 presents summary statistics for the variables relevant to the model: the total project cost, the dollar amount of in-kind cost share offered, the percentage of total cost represented by the in-kind cost share offer, the average environmental benefit score, and the total amount of acres. The results of the BT and BLP selection mechanisms are displayed in Table 2. The Table 1 Descriptive statistics for the 83 Forest Legacy Program considered in 2008.
Min Max Median Mean St dev
Total cost
In-kind
% in-kind cost share
Environmental benefit
Acres
$60,000 $8,100,000 $2,200,000 $2,311,389 $1,610,741
$20,000 $33,000,000 $1,000,000 $2,148,838 $4,298,501
13% 88% 28% 37% 18.05%
49 82 70 69 7.48
5 108,463 1060 4635 12,909.90
Table 2 Comparison of BLP and BT selection methods.
Funds spent Environmental benefit score In-kind cost share Acres protected Projects funded Avg. parcel environmental benefit score Avg. parcel in-kind cost share Avg. parcel acres protected
BT
BLP
$52,958,583 1396 $88,210,369 221,647 18 77.60 42.62% 12,313.72
$52,898,650 3024 $46,515,001 100,975 45 67.20 36.65% 2243.89
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
213
Table 3 Result set of the two-objective GP model.
Benefits
Total in-kind cost share
Total project cost
% cost in-kind
Number of projects
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
1534 1919 2169 2301 2604 2689 2753 2753 2937 2997 3024
$124,272,474 $123,103,470 $121,081,918 $119,440,251 $112,792,251 $109,685,751 $105,830,251 $105,830,251 $84,600,001 $70,600,001 $46,515,001
$177,268,827 $176,071,810 $173,985,818 $172,414,151 $165,751,151 $162,564,651 $158,829,151 $158,829,151 $137,473,651 $123,573,651 $99,413,651
70% 70% 70% 69% 68% 67% 67% 67% 62% 57% 47%
22 28 32 34 39 40 41 41 44 45 45
Total acres 196,554 197,455 193,528 192,933 195,287 194,405 193,120 193,120 87,628 88,743 100,975
increase offered by BLP represents a 120% increase in environmental benefits achieved over BT and 155% increase in the number of projects funded. This substantial improvement should be attractive to the program managers; however we also see a 53% decrease in in-kind cost share contributions and a 56% decrease in the number of acres protected. While these factors are external to the primary objective of the program this might still be a cause for concern, especially considering that there are no other options offered. In order to explore what other options are available consider the results of the GP method, varying from 0 to 1 in increments of 0.1. The results of this are displayed in Table 3 and Fig. 1. Starting from the BLP solution shown in the upper-left corner of Fig. 1 (or = 0 in the GP context), as the priority of in-kind cost sharing in the decision process is increased with increases in there are large improvements at relatively little cost in terms of environmental benefits. For example by moving
3,100 λ = 0.0
λ = 0.1
λ = 0.2
2,900
λ = 0.3, 0.4 BLP Result
Environmental Benefits
2,700
λ = 0.5 λ = 0.6
2,500 2,300
λ = 0.7 λ = 0.8
2,100
λ = 0.9
1,900 BT Result
1,700 λ = 1.0
1,500 1,300
$45
$65
$85
$105
$125
In-Kind Cost Share (in Millions) Fig. 1. Results from BT, BLP, and the result set from the two-objective Goal Programming problem.
214
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
Table 4 Result set of the triple objective Goal Programming model. Environmental benefits
% of max. env. benefits (B*)
In-kind cost share
% of max. in-kind cost share (C*)
Acres
% of max. acres (A*)
Number of duplicates
1436 1534 1565 1607 1664 1682 1697 1745 1805 1907 2030 2070 2091 2143 2144 2157 2205 2210 2226 2263 2266 2315 2325 2342 2367 2395 2395 2445 2455 2469 2506 2513 2561
47.5% 50.7% 51.8% 53.1% 55.0% 55.6% 56.1% 57.7% 59.7% 63.1% 67.1% 68.5% 69.1% 70.9% 70.9% 71.3% 72.9% 73.1% 73.6% 74.8% 74.9% 76.5% 76.9% 77.5% 78.3% 79.2% 79.2% 80.9% 81.2% 81.7% 82.9% 83.1% 84.7%
$80,565,891 $105,359,422 $97,901,426 $105,337,422 $97,906,922 $87,307,647 $85,232,647 $88,944,397 $90,971,922 $105,303,115 $97,952,115 $89,378,816 $87,777,066 $90,116,316 $89,268,264 $55,851,541 $98,526,563 $53,387,739 $106,255,588 $108,385,588 $89,904,063 $87,848,264 $88,423,264 $99,000,588 $90,319,063 $58,863,264 $110,390,588 $99,048,921 $110,458,418 $108,564,251 $99,116,751 $90,251,421 $110,486,751
64.8% 84.8% 78.8% 84.8% 78.8% 70.3% 68.6% 71.6% 73.2% 84.7% 78.8% 71.9% 70.6% 72.5% 71.8% 44.9% 79.3% 43.0% 85.5% 87.2% 72.3% 70.7% 71.2% 79.7% 72.7% 47.4% 88.8% 79.7% 88.9% 87.4% 79.8% 72.6% 88.9%
300,781 288,544 295,312 288,494 295,204 299,987 300,291 299,242 297,737 283,686 289,540 295,985 295,765 293,990 294,399 295,125 284,553 294,190 273,049 267,952 290,877 289,612 289,056 277,841 285,704 290,392 257,145 272,430 252,587 255,303 267,872 278,392 243,799
100.0% 95.9% 98.2% 95.9% 98.1% 99.7% 99.8% 99.5% 99.0% 94.3% 96.3% 98.4% 98.3% 97.7% 97.9% 98.1% 94.6% 97.8% 90.8% 89.1% 96.7% 96.3% 96.1% 92.4% 95.0% 96.5% 85.5% 90.6% 84.0% 84.9% 89.1% 92.6% 81.1%
2 2 4 4 5 5 3 6 6 4 2 12 1 2 4 1 5 1 2 3 11 1 4 1 1 9 2 4 1 1 2 8 2
Maximum 2561
Maximum $110,486,751
Maximum 300,781
Maximum 12
from = 0.0 to = 0.2, the program could increase its in-kind cost share by 82% (from $46.5 million to $84.6 million) while only decreasing environmental benefits by less than 3% (from 3024 to 2937).4 Interestingly, if the program manager also cared about the number of acres protected then a good result might be = 0.3. In this case, there would be a 127% increase in in-kind cost share (from $46.5 million to $105.8 million) and a 91% increase in total acres protected (from 100,975 acres to 193,120 acres), with only a 9% decrease in total environmental benefits (from 3024 to 2753). Table 4 and Fig. 2 show the results of the GP model which has been extended to include the total number of acres protected as a third objective. The weightings here were generated by allowing the weight for acreage to vary between 0 and 1, and applying the full set of weightings from the twoobjective problem to each level, to give a total of 121 weighting combinations.5 As might be expected the set shown in Fig. 2 appears to represent a convex hull. This is not precisely the case as the full set
4 From a static overall welfare perspective this outcome is likely not optimal – the program achieves a slightly lower level of benefits while imposing high costs on successful applicants. It does, however, put the program in a position where they can publically demonstrate the program’s effective leveraging of funds and a higher level of local private and public investment from the applicant, and hence higher likelihood of successful completion of the funded projects. 5 Given the specification used here, acreage will have at most a weight of 50%. This was done in the interest of accessing a high resolution of possible combinations near the set of solutions from the two-objective problem. A more complete alternative
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
215
Fig. 2. Result set from three objective Goal Programming problem.
of solutions is not continuous, but still represents a decreasing return relationship. It should also be noted that out of the 121 weighting schemes that were considered there were only 32 unique solutions. Solutions that appear multiple times in the set suggest robustness to preference weightings so might be attractive to program managers. The “Number of duplicates” column denotes the number of times a unique solution appears in the set of solutions. Depending upon the preferences of the program administrators, this analysis reveals a number of attractive solutions. For instance, the final solution shown in Table 4 yields a high aggregate environmental score of 2561 (84.7% of the maximum possible) and in-kind cost share of $110,486,751 (89.9% of the maximum possible), while still protecting 243,799 acres, which is 81.1% of the highest number of acres identified in any of these solutions. Conclusion and discussion Optimization techniques have been frequently proposed to improve the performance of land conservation programs. There has been resistance to these by the conservation profession. One of the reasons for this may be that standard optimization techniques are one-dimensional. This paper outlines two- and three-dimensional models that use Goal Programming (GP) to consider the tradeoffs between environmental benefits and in-kind cost sharing in conservation programs. By applying this approach to the Forest Legacy Program, this research shows with a two-dimensional model that program managers can achieve substantially better results by considering such tradeoffs. For instance, by moving from a Binary Linear Program to a more flexible GP format, the U.S. Forest Service’s Forest Legacy Program can achieve a 127% gain in in-kind cost share at only a 9% cost in benefits. Alternatively, by considering a three-dimensional model, a solution can be identified that achieves over 80% of the maximum levels of environmental benefits, in-kind cost share, and acres. Thus by using GP, program managers can have more flexibility in knowledgably selecting projects based on their priorities. A potential limitation of including in-kind cost sharing in the decision process is that partner cost sharing can be implicitly integrated into the decision process in so far as it lowers the total amount that the funding agency is asked to pay. Assuming that the funding agency incorporates a notion of value into the decision mechanism through, for instance, cost-benefit ranking or budget constrained
if one were interested in the full range of possible solutions might be to consider the set of all 55 combinations of weightings in tenths that add up to one. This will offer insight into the behavior near all three of the absolute maxima.
216
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
maximization, the use of in-kind cost sharing can be thought of as incorporating an element of a “discount auction” where applicants’ bids are ranked in terms of the largest discount offered. An example of this type of scheme used by the Delaware Agricultural Land Preservation Foundation, which select which parcels to protect based on the landowner’s discount offer, where the higherdiscounted offers are selected ahead of lower-discounted offers, without regard to the quality of the land parcel (Messer and Allen, 2010). Even though the inclusion of in-kind cost sharing to the objective function of a constrained optimization problem may appear redundant, program managers may find two attractive reasons for this approach. One is political; by achieving increased in-kind cost share contributions from local private and public sources the agency is able to demonstrate a higher degree of leverage of their program budget. A second reason is that by explicitly including in-kind cost-share in the decision mechanism applicants will have a clear incentive to develop project that have a larger in-kind cost-share. Whether applicants would actually exert effort to raise their cost-share in response to increasing the likelihood of being selected for funding by the larger federal program is an area of potential research and might be a good place to apply experimental economics techniques.
References Aldrich, R., Wyerman, J., 2006. 2005 National Land Trust Census Report. Land Trust Alliance, Washington, DC. American Farmland Trust, 2010. Status of State PACE Programs, at www.farmland.org (accessed 17.03.10). Babcock, B.A., Lakshminarayan, P.G., Wu, J., Zilberman, D., 1997. Targeting tools for the purchase of environmental amenities. Land Economics 73, 325–339. Baker, M., Payne, S., Smart, M., 1999. An empirical study of matching grants: the ‘cap on the CAP’. Journal of Public Economics 72, 269–288. Ballestero, E., Alarcon, S., Garcıa-Bernabeu, A., 2002. Establishing politically feasible water markets: a multi-criteria approach. Journal of Environmental Management 65, 411–429. Boadway, R., Shah, A., 2007. Intergovernmental Fiscal Transfers. The World Bank, Washington, DC. Borge, L.-E., Rattsø, J., 2008. Local Adjustment to Temporary Matching Grant Programs: A Dynamic Analysis. Mimeo. Department of Economics, Norwegian University of Science and Technology. Bradford County Conservation District, 2006. Environmentally Sensitive Maintenance of Dirt and Gravel Roads Program Policies. Bradford County, Towanda, PA. Bucovetsky, S., Marchand, M., Pestieau, P., 1998. Tax competition and revelation of preferences for public expenditure. Journal of Public Economics 44, 367–390. Chen, Y., 2010. Developing a best practice framework for the Maryland Agricultural Land Preservation Foundation: why don’t conservation professionals use optimization? Master’s Thesis. University of Delaware. Chernick, H., 1998. Fiscal effects of block grants for the needy: an interpretation of the evidence. International Tax and Public Finance 5 (2), 205–233. Conant, K.J., 2008. Personal correspondence on September 30, 2008. Drynan, R.G., Sandiford, F., 1985. Incorporating economic objectives in goal programs for fishery management. Marine Resource Economics 2 (2), 175–195. Kaiser, H.M., Messer, K.D., 2011. Mathematical Programming for Agricultural, Environmental and Resource Economics. Wiley & Sons, Hoboken, NJ. Kotani, K., Messer, K.D., Schulze, W.D., 2010. Matching grants and charitable giving: why the match level matters. Agricultural and Resource Economics Review 39 (2), 324–343. Mardle, S., Pascoe, S., 1999. A review of applications of multiple criteria decision-making techniques to fisheries. Marine Resource Economics 14, 41–63. Mardle, S., Pascoe, S., Tamiz, M., Jones, D., 2000. Resource allocation in the North Sea demersal fisheries: a goal programming approach. Annals of Operations Research 94, 321–342. Messer, K.D., 2006. The conservation benefits of cost-effective land acquisition: a case study in Maryland. Journal of Environmental Management 79, 305–315. Messer, K.D., Allen, W., 2010. Applying optimization and the Analytic Hierarchy Process to enhance agricultural preservation strategies in the State of Delaware. Agricultural and Resource Economics Review 39 (3), 442–456. Neely, W.P., North, R.M., Fortson, J.C., 1977. An operational approach to multiple objective decision making for public water resource projects using integer goal programming. American Journal of Agricultural Economics 59, 198–203. Oates, W.E., 1999. An essay on fiscal federalism. Journal of Economic Literature 37, 1120–1149. Önal, H., 1997. Trade-off between structural diversity and economic objectives in forest management with environmental impact and income distribution goals. American Journal of Agricultural Economics 79 (3), 1001–1012. Önal, H., Algozin, K.A., Is¸ik, K.A., Hornbaker, R.H., 1998. Economically efficient watershed management with environmental impact and income distribution goals. Journal of Environmental Management 53, 241–253. Polasky, S., Camm, J.D., Garber-Yonts, B., 2001. Selecting biological reserves cost-effectively: an application to terrestrial vertebrate conservation in Oregon. Land Economics 77, 68–78. Prendergast, J.R., Quinn, R.M., Lawton, J.H., 1999. The gaps between theory and practice in selecting nature reserves. Conservation Biology 13 (3), 484–492. Pressey, R.L., Cowling, R.M., 2001. Reserve selection algorithms and the real world. Conservation Biology 15 (1), 275–277.
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
217
Silva, D., Nakata, T., 2009. Multi-objective assessment of rural electrification in remote areas with poverty considerations. Energy Policy 37 (8), 3096–3108. Trust for Public Lands, 2009. Annual Report 2009, at www.tpl.org (accessed 10.08.10). U.S. Department of Agriculture, 2009. Agriculture Secretary Vilsack Announces $1.7 Billion in Conservation Reserve Program Rental Payments. USDA Newsroom, at www.fsa.usda.gov/FSA/newsReleases?area=newsroom &subject=landing&topic=ner&newstype=newsrel&type=detail&item=nr 20091007 rel 0497.html (accessed 20.07.10). U.S. Department of Agriculture, 2010. Conservation Reserve Program. USDA Farm Service Agency Programs at www.fsa.usda.gov/FSA/webapp?area=home&subject=copr&topic=crp (accessed 10.08.10). Underhill, L.G., 1994. Optimal and suboptimal reserve selection algorithms. Biological Conservation 70, 85–87. Wu, J., Adams, R.M., Boggess, G., 2000. Cumulative effects and optimal targeting of conservation efforts: steelhead trout habitat enhancement in Oregon. American Journal of Agricultural Economics 82, 400–413. Xu, J., Tao, R., Xu, Z., Bennett, M.T., 2010. China’s sloping land conversion program: does expansion equal success? Land Economics 89 (2), 219–224.
Contents lists available at SciVerse ScienceDirect
Journal of Forest Economics journal homepage: www.elsevier.de/jfe
Maximizing conservation and in-kind cost share: Applying Goal Programming to forest protection Jacob R. Fooks ∗, Kent D. Messer University of Delaware, Department of Applied Economics and Statistics, United States
a r t i c l e
i n f o
Article history: Received 5 February 2011 Accepted 3 April 2012 JEL classification: C6 Q15 Q23 Keywords: Goal Programming Multi-objective programming Conservation optimization Forest conservation Environmental services In-kind cost sharing Matching grants
a b s t r a c t This research evaluates the potential gains in benefits from using Goal Programming to preserve forestland. Two- and threedimensional Goal Programming models are developed and applied to data from applicants to the U.S. Forest Service’s Forest Legacy Program, the largest forest protection program in the United States. Results suggest that not only do these model yield substantial increases in benefits, but by being able to account for both environmental benefits and in-kind partner cost share, Goal Programming may be flexible enough to facilitate adoption by program managers needing to account for both ecological and political factors. © 2012 Department of Forest Economics, Swedish University of Agricultural Sciences, Umeå. Published by Elsevier GmbH. All rights reserved.
Introduction Conservation continues to be a major policy objective in the United States and internationally. Examples of programs which have been implemented to achieve conservation objectives include the U.S. Department of Agriculture’s Conservation Reserve Program (CRP) which has yearly enrollment of over 30 million acres (USDA, 2009). U.S. states and county programs have spent over $2 billion protecting over 1 million acres with permanent agricultural conservation easements in a recent 25year period (American Farmland Trust, 2010), while the entire set of open-space referenda authorized
∗ Corresponding author. E-mail address: [email protected] (J.R. Fooks). 1104-6899/$ – see front matter © 2012 Department of Forest Economics, Swedish University of Agricultural Sciences, Umeå. Published by Elsevier GmbH. All rights reserved.
http://dx.doi.org/10.1016/j.jfe.2012.04.001
208
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
$31 billion between 1996 and 2009 alone (Trust for Public Lands, 2009). As of 2005, 37 million acres had been protected by the 1667 private land trusts in the United States (Aldrich and Wyerman, 2006). The most recent U.S. Farm Bill (2008–2012) allocates $11.7 billion for conservation1 including expenditures of up to $1.8 for the CRP in 2010 (USDA, 2009), while in the European Union an anticipated D35.4 billion will be spent on “agri-environmental” programs between 2007 and 2013.2 Perhaps the largest conservation program in the world is China’s Sloping Land Conversion Program with an estimated budget of $48 billion (Xu et al., 2010). Frequently, these conservation programs involve some type of in-kind cost share or matching funds component where one program seeks to leverage additional resources from partner agencies, organizations, or individuals to achieve their conservation objectives (Kotani et al., 2010). While researchers have long advocated for the inclusion of costs as part of the selection process through either benefitcost targeting (Babcock et al., 1997) or integer programming (Underhill, 1994), the literature has not addressed issues around how to best incorporate in-kind cost sharing from partners into the selection process. While in-kind cost share contributions are implicitly accounted for through the lower project costs facing the funding organization, this fails to take full advantage of the additional information potentially provided by the size of the in-kind cost share, such as the level of commitment of the partner organization and the political benefits of being able to use program funds to leverage resources from other organizations, agencies, and individuals. This research addresses this potential concern by developing a two- and three-dimensional Goal Programming (GP) model that seeks to optimize both conservation outcomes and partner in-kind cost share contributions. This model is applied to the 2008 applicants to the U.S. Forest Service’s Forest Legacy Program (FLP), the largest single program dedicated to preserving productive forest land in the United States. Results from the models show that GP offers results that are superior to the program’s current selection mechanism and can also yield cost-effective outcomes that may be more practical than traditional optimization approaches so that they could be more attractive for adoption for program managers. Literature review In spite of the substantial amount of money spent on land conservation programs, most conservation programs are typically faced with more potential projects than they are able to fund. A vast body of economic literature has grown up around conservation programs aiming to evaluate the efficiency achieved with the vast public and private resources invested into these programs. It has been widely acknowledged that the selection methods typically employed by land conservation programs are suboptimal. As part of the selection process, programs frequently assign parcels with scores based on a variety of scoring systems, such as ratings from expert panels, standardized scoring system such as the Natural Resource Conservation Service’s Land Evaluation and Site Assessment (LESA) score for agricultural land, or the Environmental Benefit Index (EBI) originally designed for CRP. These parcel-specific scores are then used by the selection mechanism, typically a Benefit Targeting (BT) algorithm, which is essentially a greedy algorithm where projects are funded from highest to lowest benefit score until funds are exhausted. This approach has been subject to substantial criticism as being very inefficient, and it has been repeatedly proposed that adopting alternative mathematical optimization methods would lead to superior outcomes (for example, Babcock et al., 1997; Polasky et al., 2001; Wu et al., 2000). One alternative to BT is Cost Effectiveness Analysis (CEA). With CEA projects are ranked in terms of environmental benefits achieved per dollars spent on the project. These cost effectiveness ratios are then ranked and projects selected by highest ratio until funds are exhausted. CEA performs quite well as compared to BT and are easy to implement, but can still achieve suboptimal results under certain
1 U.S. calculations based on data from Claassen R., Conservation Policy Briefing Room, Economic Research Service, U.S. Department of Agriculture.
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
209
circumstances (Messer, 2006). Additionally, CEA does not allow for additional constraints to be easily placed on the problem. A second alternative is Binary Linear Programming (BLP), also referred to as Binary Integer Programming. BLP is a variation on linear programming problems where there is a binary (0/1) variable representing the decision to fund a project. BLP maximizes the benefit score of each project, which is multiplied by the binary variable representing the funding decision, subject to the project costs times the funding decision variables and any other constraints that are pertinent. Through the use of the branch-and-bound algorithm, BLP considers every possible combination of projects and selects the best combination. BLP typically performs somewhat better that CEA. In spite of the promise of substantial increase in environmental benefits offered by both CEA and BLP land conservation programs have been very hesitant to adopt either in their practices. To our knowledge, the only organization that has used either of these techniques is the Baltimore County’s Division of Environmental Protection and Resource Management (DEPRM), which has used CEA for its agricultural protection program since 2007. In the first three years of use, DEPRM protected an additional 680 acres worth $5.4 million relative to what they would have done with their previous BT given the same budget amounts (Kaiser and Messer, 2011). Researchers have speculated about the lack of adoption of cost-effective and optimization techniques (i.e. Prendergast et al., 1999; Pressey and Cowling, 2001). For example, Prendergast et al. (1999), as the result of an informal interviewing process with a wide sample of ecologists and land managers, identified three main barriers for adoption: lack of knowledge, lack of resources, and real or perceived shortcomings in the methods. These authors conclude that ultimately for conservation optimization to be widely adopted more work needs to be done to cater research to the needs of real life conservation practice. A recent survey of Maryland county level conservation program administrators suggested that managers not only lacked knowledge of BLP and CEA, but managers also reported not considering cost to be a major priority in the selection process and lacking incentives that would lead them to adopt it with their program (Chen, 2010). Another possible explanation for this lack of adoption is that managers need to be able to consider numerous political and strategic objectives in their decision making process. Their duties include not merely maximizing benefit scores, but also defending the “value” achieved from donor, funding agency, or tax payers’ money, ensuring that applicants get a fair deal from a transparent decision mechanism, and distributing funds in a manner perceived as equitable. BLP is unable to clearly address many of these duties. These can be thought of as secondary or operational objectives that do not immediately impact the primary goal of protecting high quality land but still may be important factors in the decision making process. One way to incorporate these into the problem is as constraints, but this offers little ability to consider sensitivity or alternatives to the single solution provided. This is the basis of the approach used in Önal et al. (1998) which considers both environmental and equity concerns for watershed management using a mathematical programming model. This model maximizes total profit across a watershed with a chance constraint on chemical runoff levels to take into account the stochastic nature of rainfall, and a constraint on the equity of program impact as measured by an index measurement of deviation from a uniform loss sharing level. These constraints are varied to examine the tradeoff between income, pollution, and equity losses. A second option is to format the problem as a GP problem with these secondary objectives included as weighted goals. There have been several applications of GP in this field, including balancing economic and biological objectives over short term and long term time frames in fishery management (Drynan and Sandiford, 1985; Mardle and Pascoe, 1999; Mardle et al., 2000) optimizing environmental, social, and economic goals in energy production (Silva and Nakata, 2009); and the management of public water resources (Neely et al., 1977; Ballestero et al., 2002). Önal (1997) considers an approach similar to GP in forest management. A model is employed which, instead of minimizing deviations from a goal as is done in GP, uses constrained deviations from a goal to maximize discounted future harvest value while maintaining a minimum value of a species diversity index. To our knowledge GP has not been proposed in the planning of land conservation decisions. In-kind cost sharing is an example of a secondary objective. According to Kathryn Conant, National Forest Legacy Program Manager, the US Forest Service is interested in approaches that could both maximize forest benefits and get the most from leveraged partner cost sharing funds (Conant, 2008).
210
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
The use of cost sharing (or matching funds) is very common in conservation and other publically funded programs. While the way in which such requirements implemented can vary, typical approaches include requirements that the participating organization cover some percentage of the total cost of the projects themselves (for instance, the USDA’s Conservation Reserve Program (USDA, 2010)), or a system where the amount of in-kind cost sharing offered weights the ranking for amount of funds awarded or priority to receive funds (such as is the case with some counties that apply for funds for water quality enhancements from the Pennsylvania’s Dirt and Gravel Road Program (Bradford County Conservation District, 2006)). Programs that require some sort of cost sharing include the USDA’s Conservation Reserve and Forest Legacy Programs, the U.S. Fish and Wildlife Service’s National Wetlands Conservation Grant Program, and numerous other federal, state, and private conservation funding programs. Several justifications are typically offered for requiring some sort of in-kind costs share in the selection mechanism for conservation programs. A primary justification is that by asking applicants to commit to pay for some of the costs themselves, this ensures that applicants are committed to the project and consider it worth the effort to conserve. In other words, the applicant is willing to “put their money where their mouth is” for the project. Another justification is that cost sharing helps extend the reach of the conservation program as it can be considered a type of discount on the total project price. From the point of view of the funding agency, this could mean that there would be more money available to fund additional projects and also help increase the number of projects funded with the program’s funds. In the literature, in-kind cost sharing can be considered a type of matching grant program where the granting agency agrees to pay some portion of costs up to a percentage of the total project costs. Theoretically, matching grants are seen as a mechanism to correct for externalities and spillovers in federalized agency structures. Oates (1999) points out that the local benefits from a project under consideration may not justify funding the project for the local agency. However, the spillover benefits to other areas make it attractive to society as a whole. The matching grant offered by an outside body would represent a sort of Pigouvian subsidy to pay the local agency for the external benefits obtained in other jurisdictions. Bucovetsky et al. (1998) offers an informational argument for matching grants similar to the “money where mouth is” justification. Ideally, a government should distribute funds to the regions which value public services the most. The matching grant severs as a mechanism to reveal the private value of funding to the applicants. Current evidence on the effect of matching grants and cost sharing has not offered strong results. Baker et al. (1999) found that instituting a matching grant system in the Canada Assistance Plan lowered expenditures growth by 8–9 percentage points as provinces became responsible for a portion of program costs. Using simulations, Borge and Rattsø (2008) found that compared to block grants matching grants decrease expense, but lead to unstable service provision over time. Chernick (1998) found that the conversion from matching grants to fixed block grants by U.S. federal welfare funding programs had substantial variation across states, but generally led to a reduction in benefits. For an in depth discussion of matching grants and their implementation see Boadway and Shah (2007). In the following section we develop a GP model which incorporates in-kind cost share as a secondary objective. Following that we apply this weighting scheme to the 2008 project set for the USDA’s Forest Legacy Program.
Model development The basic BLP program that seeks to maximize environmental benefits, B*: Max : B∗ = s.t. :
xi ∈ {0, 1}
bi xi
pi xi ≤ T
(1) (2) (3)
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
211
where xi is a binary variable indicating whether project i is chosen, bi represents the environmental benefit score as determined by the Forest Legacy expert panel, and pi the funding request for project i. T is the total budget available for the selection program for the year. Similarly, denoting ci as the dollar amount of cost share offered for project i, a BLP program can be set-up to maximize the amount of in-kind cost share, C*: Max : C ∗ = s.t. :
ci xi
(4)
pi xi ≤ T
(5)
xi ∈ {0, 1}
(6)
A two-objective (two-dimensional) GP model can be developed as an extension of these two problems as the set-up to minimize a weighted sum of percent deviations from the maximum feasible value of in-kind cost share, C*, and environmental benefits, B*. Using these two extremes as targets, the GP problem can be set up and then the weighting between the two can be varied parametrically to examine the tradeoffs. This problem is stated as follows:
Min : Z = s.t. :
dC−
C∗
+ (1 − )
dB−
(7)
B∗
bi xi − dB+ + dB− = B∗
(8)
ci xi − dC+ + dC− = C ∗
(9)
pi xi ≤ T
(10)
xi ∈ {0, 1}; di+ , di− , ≥ 0
(11) +/−
are chosen to measure the achieved level of objective N (in this case The deviation variables, dN ni xi , from N*. In the objective function, the negative deviations are either B* or C*), determined by divided by the target value so that each represents a percentage deviation, and thus are of comparable magnitudes. The objective is to minimize the total deviations from the joint optima based on the priority weighting. The weighting factor, , can be varied parametrically from 0 to 1. This represents the percentage of priority that is given to in-kind costs, such that when is 0, 100% of the priority will be given to maximizing environmental benefits and when is 1, 100% of the priority will be given to maximizing in-kind cost share. Thus when is 0.3, then 30% of the weight will be on in-kind cost share and 70% will be on environmental benefits. As will be discussed below in the result section, the model can be expanded to consider more than two objectives. Since managers of conservation programs typically are concerned about a range of objective which may have differing degrees of exclusivity, below we develop a three-dimensional model that incorporates total environmental benefits, in-kind cost share, and the total amount of acreage preserved.3
Min : Z = C s.t. :
dC− C∗
+ (B )
ci xi − dC+ + dC− = C ∗
bi xi − dB+ + dB− = B∗
dB− B∗
+ (A )
dA− A∗
(12) (13) (14)
3 Note that as the number of objectives increases constructing a systematic weighting scheme becomes rather complex and the solutions set is less amenable to visual representation. None the less, the technique can still be useful to explore alternate solutions and the effects of changing prioritization schemes.
212
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
ai xi − dA+ + dA− = A∗
(15)
pi xi ≤ T
(16)
xi ∈ {0, 1}; di+ , di− , ≥ 0
(17)
Case study: Forest Legacy Program The Forest Legacy Program (FLP) is a program administered by the US Forest Service to support the acquisition of conservation easements and other voluntary protection mechanisms on privately held forest land. It was established in 1990 and is the largest single protector of privately held forest land in the US. As of 2010, FLP has helped to protect nearly 2 million acres across 41 states and Puerto Rico. The FLP works primarily with state agencies to acquire easements on privately held land to protect timber supply, wildlife habitat, soil and watershed protection, aesthetics, and recreational opportunities. Participating landowners typically receive some percentage of the land value and a decrease in their tax burden in exchange for the right to develop the land. In order to participate in the program landowners must develop a comprehensive multiple resource management plan. The USDA requires states to provide at least 25% of the total project costs, however no cap is placed on the percentage of in-kind cost sharing that can be offered by the applicants and some applicants to the FLP in the past have submitted in-kind cost shares of nearly 90%. The average percentage of in-kind cost share offered in 2008 was 36.3%. Applicants receive no additional benefit from offering excess in-kind. FLP selects projects using an environmental benefit score determined by a panel of experts. Each panel member is asked to score each project from 0 to 30 based on importance, threat, and strategic value. Panel members are not allowed to evaluate projects from the states in which they work. The three scores are summed such that each project is receives a score between 0 and 90 from each reviewer. The FLP administrators discard the lowest and highest scores and then average the remaining scores to assign the project a final score. The projects are selected for funding using the BT algorithm. This research considers the 83 projects submitted to the FLP in 2008 from 44 states, Puerto Rico, and the US Virgin Islands. Table 1 presents summary statistics for the variables relevant to the model: the total project cost, the dollar amount of in-kind cost share offered, the percentage of total cost represented by the in-kind cost share offer, the average environmental benefit score, and the total amount of acres. The results of the BT and BLP selection mechanisms are displayed in Table 2. The Table 1 Descriptive statistics for the 83 Forest Legacy Program considered in 2008.
Min Max Median Mean St dev
Total cost
In-kind
% in-kind cost share
Environmental benefit
Acres
$60,000 $8,100,000 $2,200,000 $2,311,389 $1,610,741
$20,000 $33,000,000 $1,000,000 $2,148,838 $4,298,501
13% 88% 28% 37% 18.05%
49 82 70 69 7.48
5 108,463 1060 4635 12,909.90
Table 2 Comparison of BLP and BT selection methods.
Funds spent Environmental benefit score In-kind cost share Acres protected Projects funded Avg. parcel environmental benefit score Avg. parcel in-kind cost share Avg. parcel acres protected
BT
BLP
$52,958,583 1396 $88,210,369 221,647 18 77.60 42.62% 12,313.72
$52,898,650 3024 $46,515,001 100,975 45 67.20 36.65% 2243.89
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
213
Table 3 Result set of the two-objective GP model.
Benefits
Total in-kind cost share
Total project cost
% cost in-kind
Number of projects
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
1534 1919 2169 2301 2604 2689 2753 2753 2937 2997 3024
$124,272,474 $123,103,470 $121,081,918 $119,440,251 $112,792,251 $109,685,751 $105,830,251 $105,830,251 $84,600,001 $70,600,001 $46,515,001
$177,268,827 $176,071,810 $173,985,818 $172,414,151 $165,751,151 $162,564,651 $158,829,151 $158,829,151 $137,473,651 $123,573,651 $99,413,651
70% 70% 70% 69% 68% 67% 67% 67% 62% 57% 47%
22 28 32 34 39 40 41 41 44 45 45
Total acres 196,554 197,455 193,528 192,933 195,287 194,405 193,120 193,120 87,628 88,743 100,975
increase offered by BLP represents a 120% increase in environmental benefits achieved over BT and 155% increase in the number of projects funded. This substantial improvement should be attractive to the program managers; however we also see a 53% decrease in in-kind cost share contributions and a 56% decrease in the number of acres protected. While these factors are external to the primary objective of the program this might still be a cause for concern, especially considering that there are no other options offered. In order to explore what other options are available consider the results of the GP method, varying from 0 to 1 in increments of 0.1. The results of this are displayed in Table 3 and Fig. 1. Starting from the BLP solution shown in the upper-left corner of Fig. 1 (or = 0 in the GP context), as the priority of in-kind cost sharing in the decision process is increased with increases in there are large improvements at relatively little cost in terms of environmental benefits. For example by moving
3,100 λ = 0.0
λ = 0.1
λ = 0.2
2,900
λ = 0.3, 0.4 BLP Result
Environmental Benefits
2,700
λ = 0.5 λ = 0.6
2,500 2,300
λ = 0.7 λ = 0.8
2,100
λ = 0.9
1,900 BT Result
1,700 λ = 1.0
1,500 1,300
$45
$65
$85
$105
$125
In-Kind Cost Share (in Millions) Fig. 1. Results from BT, BLP, and the result set from the two-objective Goal Programming problem.
214
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
Table 4 Result set of the triple objective Goal Programming model. Environmental benefits
% of max. env. benefits (B*)
In-kind cost share
% of max. in-kind cost share (C*)
Acres
% of max. acres (A*)
Number of duplicates
1436 1534 1565 1607 1664 1682 1697 1745 1805 1907 2030 2070 2091 2143 2144 2157 2205 2210 2226 2263 2266 2315 2325 2342 2367 2395 2395 2445 2455 2469 2506 2513 2561
47.5% 50.7% 51.8% 53.1% 55.0% 55.6% 56.1% 57.7% 59.7% 63.1% 67.1% 68.5% 69.1% 70.9% 70.9% 71.3% 72.9% 73.1% 73.6% 74.8% 74.9% 76.5% 76.9% 77.5% 78.3% 79.2% 79.2% 80.9% 81.2% 81.7% 82.9% 83.1% 84.7%
$80,565,891 $105,359,422 $97,901,426 $105,337,422 $97,906,922 $87,307,647 $85,232,647 $88,944,397 $90,971,922 $105,303,115 $97,952,115 $89,378,816 $87,777,066 $90,116,316 $89,268,264 $55,851,541 $98,526,563 $53,387,739 $106,255,588 $108,385,588 $89,904,063 $87,848,264 $88,423,264 $99,000,588 $90,319,063 $58,863,264 $110,390,588 $99,048,921 $110,458,418 $108,564,251 $99,116,751 $90,251,421 $110,486,751
64.8% 84.8% 78.8% 84.8% 78.8% 70.3% 68.6% 71.6% 73.2% 84.7% 78.8% 71.9% 70.6% 72.5% 71.8% 44.9% 79.3% 43.0% 85.5% 87.2% 72.3% 70.7% 71.2% 79.7% 72.7% 47.4% 88.8% 79.7% 88.9% 87.4% 79.8% 72.6% 88.9%
300,781 288,544 295,312 288,494 295,204 299,987 300,291 299,242 297,737 283,686 289,540 295,985 295,765 293,990 294,399 295,125 284,553 294,190 273,049 267,952 290,877 289,612 289,056 277,841 285,704 290,392 257,145 272,430 252,587 255,303 267,872 278,392 243,799
100.0% 95.9% 98.2% 95.9% 98.1% 99.7% 99.8% 99.5% 99.0% 94.3% 96.3% 98.4% 98.3% 97.7% 97.9% 98.1% 94.6% 97.8% 90.8% 89.1% 96.7% 96.3% 96.1% 92.4% 95.0% 96.5% 85.5% 90.6% 84.0% 84.9% 89.1% 92.6% 81.1%
2 2 4 4 5 5 3 6 6 4 2 12 1 2 4 1 5 1 2 3 11 1 4 1 1 9 2 4 1 1 2 8 2
Maximum 2561
Maximum $110,486,751
Maximum 300,781
Maximum 12
from = 0.0 to = 0.2, the program could increase its in-kind cost share by 82% (from $46.5 million to $84.6 million) while only decreasing environmental benefits by less than 3% (from 3024 to 2937).4 Interestingly, if the program manager also cared about the number of acres protected then a good result might be = 0.3. In this case, there would be a 127% increase in in-kind cost share (from $46.5 million to $105.8 million) and a 91% increase in total acres protected (from 100,975 acres to 193,120 acres), with only a 9% decrease in total environmental benefits (from 3024 to 2753). Table 4 and Fig. 2 show the results of the GP model which has been extended to include the total number of acres protected as a third objective. The weightings here were generated by allowing the weight for acreage to vary between 0 and 1, and applying the full set of weightings from the twoobjective problem to each level, to give a total of 121 weighting combinations.5 As might be expected the set shown in Fig. 2 appears to represent a convex hull. This is not precisely the case as the full set
4 From a static overall welfare perspective this outcome is likely not optimal – the program achieves a slightly lower level of benefits while imposing high costs on successful applicants. It does, however, put the program in a position where they can publically demonstrate the program’s effective leveraging of funds and a higher level of local private and public investment from the applicant, and hence higher likelihood of successful completion of the funded projects. 5 Given the specification used here, acreage will have at most a weight of 50%. This was done in the interest of accessing a high resolution of possible combinations near the set of solutions from the two-objective problem. A more complete alternative
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
215
Fig. 2. Result set from three objective Goal Programming problem.
of solutions is not continuous, but still represents a decreasing return relationship. It should also be noted that out of the 121 weighting schemes that were considered there were only 32 unique solutions. Solutions that appear multiple times in the set suggest robustness to preference weightings so might be attractive to program managers. The “Number of duplicates” column denotes the number of times a unique solution appears in the set of solutions. Depending upon the preferences of the program administrators, this analysis reveals a number of attractive solutions. For instance, the final solution shown in Table 4 yields a high aggregate environmental score of 2561 (84.7% of the maximum possible) and in-kind cost share of $110,486,751 (89.9% of the maximum possible), while still protecting 243,799 acres, which is 81.1% of the highest number of acres identified in any of these solutions. Conclusion and discussion Optimization techniques have been frequently proposed to improve the performance of land conservation programs. There has been resistance to these by the conservation profession. One of the reasons for this may be that standard optimization techniques are one-dimensional. This paper outlines two- and three-dimensional models that use Goal Programming (GP) to consider the tradeoffs between environmental benefits and in-kind cost sharing in conservation programs. By applying this approach to the Forest Legacy Program, this research shows with a two-dimensional model that program managers can achieve substantially better results by considering such tradeoffs. For instance, by moving from a Binary Linear Program to a more flexible GP format, the U.S. Forest Service’s Forest Legacy Program can achieve a 127% gain in in-kind cost share at only a 9% cost in benefits. Alternatively, by considering a three-dimensional model, a solution can be identified that achieves over 80% of the maximum levels of environmental benefits, in-kind cost share, and acres. Thus by using GP, program managers can have more flexibility in knowledgably selecting projects based on their priorities. A potential limitation of including in-kind cost sharing in the decision process is that partner cost sharing can be implicitly integrated into the decision process in so far as it lowers the total amount that the funding agency is asked to pay. Assuming that the funding agency incorporates a notion of value into the decision mechanism through, for instance, cost-benefit ranking or budget constrained
if one were interested in the full range of possible solutions might be to consider the set of all 55 combinations of weightings in tenths that add up to one. This will offer insight into the behavior near all three of the absolute maxima.
216
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
maximization, the use of in-kind cost sharing can be thought of as incorporating an element of a “discount auction” where applicants’ bids are ranked in terms of the largest discount offered. An example of this type of scheme used by the Delaware Agricultural Land Preservation Foundation, which select which parcels to protect based on the landowner’s discount offer, where the higherdiscounted offers are selected ahead of lower-discounted offers, without regard to the quality of the land parcel (Messer and Allen, 2010). Even though the inclusion of in-kind cost sharing to the objective function of a constrained optimization problem may appear redundant, program managers may find two attractive reasons for this approach. One is political; by achieving increased in-kind cost share contributions from local private and public sources the agency is able to demonstrate a higher degree of leverage of their program budget. A second reason is that by explicitly including in-kind cost-share in the decision mechanism applicants will have a clear incentive to develop project that have a larger in-kind cost-share. Whether applicants would actually exert effort to raise their cost-share in response to increasing the likelihood of being selected for funding by the larger federal program is an area of potential research and might be a good place to apply experimental economics techniques.
References Aldrich, R., Wyerman, J., 2006. 2005 National Land Trust Census Report. Land Trust Alliance, Washington, DC. American Farmland Trust, 2010. Status of State PACE Programs, at www.farmland.org (accessed 17.03.10). Babcock, B.A., Lakshminarayan, P.G., Wu, J., Zilberman, D., 1997. Targeting tools for the purchase of environmental amenities. Land Economics 73, 325–339. Baker, M., Payne, S., Smart, M., 1999. An empirical study of matching grants: the ‘cap on the CAP’. Journal of Public Economics 72, 269–288. Ballestero, E., Alarcon, S., Garcıa-Bernabeu, A., 2002. Establishing politically feasible water markets: a multi-criteria approach. Journal of Environmental Management 65, 411–429. Boadway, R., Shah, A., 2007. Intergovernmental Fiscal Transfers. The World Bank, Washington, DC. Borge, L.-E., Rattsø, J., 2008. Local Adjustment to Temporary Matching Grant Programs: A Dynamic Analysis. Mimeo. Department of Economics, Norwegian University of Science and Technology. Bradford County Conservation District, 2006. Environmentally Sensitive Maintenance of Dirt and Gravel Roads Program Policies. Bradford County, Towanda, PA. Bucovetsky, S., Marchand, M., Pestieau, P., 1998. Tax competition and revelation of preferences for public expenditure. Journal of Public Economics 44, 367–390. Chen, Y., 2010. Developing a best practice framework for the Maryland Agricultural Land Preservation Foundation: why don’t conservation professionals use optimization? Master’s Thesis. University of Delaware. Chernick, H., 1998. Fiscal effects of block grants for the needy: an interpretation of the evidence. International Tax and Public Finance 5 (2), 205–233. Conant, K.J., 2008. Personal correspondence on September 30, 2008. Drynan, R.G., Sandiford, F., 1985. Incorporating economic objectives in goal programs for fishery management. Marine Resource Economics 2 (2), 175–195. Kaiser, H.M., Messer, K.D., 2011. Mathematical Programming for Agricultural, Environmental and Resource Economics. Wiley & Sons, Hoboken, NJ. Kotani, K., Messer, K.D., Schulze, W.D., 2010. Matching grants and charitable giving: why the match level matters. Agricultural and Resource Economics Review 39 (2), 324–343. Mardle, S., Pascoe, S., 1999. A review of applications of multiple criteria decision-making techniques to fisheries. Marine Resource Economics 14, 41–63. Mardle, S., Pascoe, S., Tamiz, M., Jones, D., 2000. Resource allocation in the North Sea demersal fisheries: a goal programming approach. Annals of Operations Research 94, 321–342. Messer, K.D., 2006. The conservation benefits of cost-effective land acquisition: a case study in Maryland. Journal of Environmental Management 79, 305–315. Messer, K.D., Allen, W., 2010. Applying optimization and the Analytic Hierarchy Process to enhance agricultural preservation strategies in the State of Delaware. Agricultural and Resource Economics Review 39 (3), 442–456. Neely, W.P., North, R.M., Fortson, J.C., 1977. An operational approach to multiple objective decision making for public water resource projects using integer goal programming. American Journal of Agricultural Economics 59, 198–203. Oates, W.E., 1999. An essay on fiscal federalism. Journal of Economic Literature 37, 1120–1149. Önal, H., 1997. Trade-off between structural diversity and economic objectives in forest management with environmental impact and income distribution goals. American Journal of Agricultural Economics 79 (3), 1001–1012. Önal, H., Algozin, K.A., Is¸ik, K.A., Hornbaker, R.H., 1998. Economically efficient watershed management with environmental impact and income distribution goals. Journal of Environmental Management 53, 241–253. Polasky, S., Camm, J.D., Garber-Yonts, B., 2001. Selecting biological reserves cost-effectively: an application to terrestrial vertebrate conservation in Oregon. Land Economics 77, 68–78. Prendergast, J.R., Quinn, R.M., Lawton, J.H., 1999. The gaps between theory and practice in selecting nature reserves. Conservation Biology 13 (3), 484–492. Pressey, R.L., Cowling, R.M., 2001. Reserve selection algorithms and the real world. Conservation Biology 15 (1), 275–277.
J.R. Fooks, K.D. Messer / Journal of Forest Economics 18 (2012) 207–217
217
Silva, D., Nakata, T., 2009. Multi-objective assessment of rural electrification in remote areas with poverty considerations. Energy Policy 37 (8), 3096–3108. Trust for Public Lands, 2009. Annual Report 2009, at www.tpl.org (accessed 10.08.10). U.S. Department of Agriculture, 2009. Agriculture Secretary Vilsack Announces $1.7 Billion in Conservation Reserve Program Rental Payments. USDA Newsroom, at www.fsa.usda.gov/FSA/newsReleases?area=newsroom &subject=landing&topic=ner&newstype=newsrel&type=detail&item=nr 20091007 rel 0497.html (accessed 20.07.10). U.S. Department of Agriculture, 2010. Conservation Reserve Program. USDA Farm Service Agency Programs at www.fsa.usda.gov/FSA/webapp?area=home&subject=copr&topic=crp (accessed 10.08.10). Underhill, L.G., 1994. Optimal and suboptimal reserve selection algorithms. Biological Conservation 70, 85–87. Wu, J., Adams, R.M., Boggess, G., 2000. Cumulative effects and optimal targeting of conservation efforts: steelhead trout habitat enhancement in Oregon. American Journal of Agricultural Economics 82, 400–413. Xu, J., Tao, R., Xu, Z., Bennett, M.T., 2010. China’s sloping land conversion program: does expansion equal success? Land Economics 89 (2), 219–224.