Policy collision: a framework to identify where polycentric, multi-objective sustainability solutions are needed

The exploitation of ecosystem services, through processes like agricultural production, is associated with myriad negative environmental impacts, which are felt by stakeholders on local, regional, and global scales. The varying type and scale of impacts leads naturally to fragmented and siloed approaches to mitigating externalities by diverse governmental and non-governmental institutions. However, policies designed to address a single impact may worsen other negative impacts. As a result, even when groups have the expertise to design policy solutions in one dimension, policies addressing single issues may conflict and result in less than ideal outcomes in combination. In this paper, we present a conceptual framework and examples of this kind of ‘policy collision,’ where policies produce mutual negative interference so that policies designed independently may fail to achieve their goals. We argue that an integrated systems perspective is needed to overcome this problem and present several positive examples where this has been put into practice. Policy collision provides a useful framework for determining how each colliding policy should be modified in improve outcomes.


Introduction
One of the major justifications for public policy is to manage negative externalities, where at least part of the cost of an economic activity is borne by others. Whether policy is done through command and control, Pigouvian taxation, tradable permits, etc, the goal is to correct market failure by imposing a price reflective of the externality's social cost. In theory, an omnipotent social planner could simultaneously manage multiple externalities across multiple spatial locations. In practice, however, policy actors implement policies that are more constrained in scope, scale, and complexity, based primarily on harms to their constituents associated with a limited number of externalities they are authorized to manage. In this paper, we identify a circumstance wherein the advantages of coordination outweigh these pragmatic limitations on typical policy design. This occurs when independently developed policies 'collide,' hampering the positive outcomes associated with the other policies.
Classical environmental economics casts many externalities as having a single, uniform effect on an inexhaustible public good (e.g. one person breathing polluted air does not reduce exposure of others to the same pollution) (Bator 1958, Baumol andOates 1988). This simple case generates simple policy prescriptions (e.g. a uniform tax on all air pollutant emissions, or treating all air pollutant emissions as equivalent in a cap-and-trade system). However, many activities causing externalities have multiple impacts (e.g. economic activity may cause both air and water pollution) that differ across locations and time. These more complex realities produce policy solutions that also often involve more nuance. Globalization and differing impacts by locale make many externalities effectively shiftable and depletable when applying a larger system boundary. Consequently, policies designed to manage negative externalities associated with the production of goods in one locale have the potential to induce increases in those externalities elsewhere (i.e. 'leakage' or spillover effects). When marginal costs of production increase under the management policy, local supply contracts and production shifts elsewhere to meet demand.
The various types of pollution embedded in goods give rise to concepts like carbon and water footprints that give a full accounting of production externalities. This has led naturally to studies of virtual pollution trading, wherein pollutants associated with production are relocated spatially by importing goods rather than making them locally. Much of this trade is voluntary and Pareto-improving-for example, allowing water-scarce nations to access water-intensive staples like wheat (Allan 1997)-but management policies that reduce local production efficiencies of priceinelastic goods induce unintended virtual trading of pollution (Dabrowski et al 2009, Teymouri andDehghanzadeh 2021). This exemplifies the complexity of scale issues in the sustainability and governance of socio-ecological systems that exhibit a mismatch between the scale of resource governance institutions and the scale of the resources themselves (Young and Gasser 2002, Cash et al 2006, Olsson et al 2007, Termeer and Kessener 2007. The usefulness of multiscale, multilevel governance to achieve collective goals in complex socioecological systems is well established (Kooiman 1993, Rhodes 1997, Hooghe and Marks 2003, Ostrom 2007. Approaches generally fall into two types. In the first, coordination costs are limited by involving the minimum number of jurisdictions required to solve a problem (Oates 1972); the second directly minimizes the negative externalities and spillover between jurisdictions (Termeer et al 2010). This literature focuses on the motivation and means for coordination between policy actors to effect a common goal.
We define 'policy collision' to be when policies enacted by two or more entities, each designed to manage a particular negative externality, fail to achieve their goals due to mutually destructive interference and a lack of coordination. Identifying policy collisions during the policy design process can enable egocentric, myopic institutions to better meet their own goals by indicating when coordination is necessary. By acknowledging the impact of other selfinterested policy actors on one's own domain, policy actors become more likely to succeed because externalities are accounted for and internalized as part of the policy design. Coordination between actors creates a larger policy space that enables policy solutions that are not possible when actors are siloed; by sacrificing some degree of de jure policy sovereignty, the cooperating actors improve their de facto control (i.e. their ability to effectively achieve targets through the use of policy instruments) (Cooper 1968, Chang 2006, Mayer 2009). Identification of policy collisions thus reveals an appropriate choice of scale and polycentric governance for the management of potentially multiple externalities (Ostrom 2009, Carlisle andGruby 2019). We provide three case studies illustrating different types of policy collisions, explaining how greater coordination would improve the achievement of each policy actor's goals, and showing success stories in overcoming collisions.

Conceptual underpinnings
Policy collisions can be found in a range of sectors and is characterized by a system with two or more policy instruments, each designed to influence distinct policy target variables, but which have a secondary impact on the other policy's target variables. In a simple two-policy example, assume that policy i targets policy outcome y i using instrument x i , where the relationship between them is governed by some function f i . The outcome y i is also influenced by another instrument x j (i ̸ = j), so y i = f i x i , x j . Without loss of generality, suppose the target variable is to be minimized; we thus assume that y i is a decreasing function of x i , while in the case of a policy collision, y i is an increasing function of x j in equilibrium. Generically, we can view x i as the degree of control exerted by actor i, and therefore assume x i ≥ 0 (where 0 represents a status quo or future without action).
In some cases, the policy design problem is to identify the least-cost means to achieve some outcome level Y i , i.e. solving f i x i , x j = Y i . An example would be regulated limits on chemical concentrations in water to protect health. Successful design of the policy hinges on knowledge of x j , but naïve actors may be unaware that another deliberative body is considering a policy which would impact their own system. Under this lack of awareness, each actor defaults to solving for Under coordination, the two actors instead simultaneously solve for x i = X * i and x j = X * j such that: Because the policies collide, X * i > X 0 i , meaning that additional control is necessary to achieve the policy target. In other words, when ∂f i ∂x j ̸ = 0 and ∂f j ∂x i ̸ = 0, the policies that achieve target outcomes assuming the siloed sectoral approach and the integrated approach will be different, with the integrated approach demanding greater costs but achieving the environmental targets.
Because it can be hard to coordinate policy across sectors, there may also be a cost of integration. Integration is preferred when the improvement in net benefits with integration are at least as large as the cost of integration, which will occur in situations with large interactions between sectors and where it is not too difficult (i.e. costly) to coordinate. Of course, institutional barriers and other political considerations may rule out the possibility of coordination, e.g. if legislators reject the importance of the colliding policy's objective(s) or are unwilling to take the time required to resolve the collision.
The expansive nature of this concept means that policy collisions can occur across a range of spatial and/or temporal scales. They can involve the same externality being managed in different jurisdictions, different externalities being managed in the same jurisdiction, or different externalities and different jurisdictions. They may also involve more than two policy actors. Another variation could see one policy impacting multiple outcomes targeted by other policies.
Finally, another situation would be one in which, rather than minimizing the control or cost needed to meet fixed environmental targets, the colliding policies' objectives are to maximize their net benefits. As an alternative to the construction above, imagine y i and y j represent monetized benefits for both policies, and that the policy instruments have implementation costs c i (x i ) and c j x j . If the policy actors are aware of each other but do not coordinate, siloed policy design would see each sector choose a policy to maximize its own net benefits without considering their impact on the other sector: This yields outcomes characterized by policy instruments that equate sectoral marginal benefit with marginal cost: An integrated approach would account for impacts across sectors: This yields outcomes that fully account for crosssectoral impacts: , the solutions to the siloed sectoral approach and integrated approach will be different, with the integrated approach delivering greater net benefits.
In the following sections, we present several examples of policy collisions related to sustainability in the agricultural sector. We explain how these examples relate to the conceptual model presented above and why they matter from an empirical standpoint. The first outlines a potential collision of two hypothetical policies designed to manage water quality and water scarcity, chosen as a didactic example to illustrate the conceptual framework in action. The second describes a real, ongoing collision between US development assistance and food aid programs, and the third presents two success stories, from Costa Rica and China, where governments realized the need for integrated land use policies to avoid policy collisions.

Case study: water scarcity and water quality management
This case study illustrates a collision between policies designed to manage nitrate leaching and groundwater depletion in major agricultural regions of the United States. The Northern Gulf of Mexico suffers from extensive summertime hypoxia, a 'dead zone' in its bottom waters that contributes to benthic mortality and an increased risk of fisheries collapse (Rabalais and Turner 2019, US Environmental Protection Agency 2022b). The major driver of this environmental hazard is nitrogen loading into the Mississippi River from nitrate leaching associated with agricultural production in the basin, where over 90% of US agricultural exports are produced. Donner and Kucharik (2008) concludes that production of biomass to support liquid renewable fuels mandated by US Renewable Fuel Standards (Renewable Fuel Standard (RFS) 2021) is incompatible with achieving the reduction in size of the hypoxic zone targeted by the Mississippi River/Gulf of Mexico Hypoxia Task Force (US Environmental Protection Agency 2022a).
Partially overlapping with the Mississippi River Basin, the Ogallala Aquifer provides groundwater for irrigation of crops in parts of the High Plains region; it services more than 30% of US crops and livestock production (Dieter et al 2018) and uses about 30% of all the irrigation water in the US (Gowda et al 2019). Rapid expansion of acreage under agricultural production has led to the Ogallala being one of the most rapidly depleted aquifers in the world (Famiglietti and Ferguson 2021). While the annual rate of depletion is estimated to have peaked in 2006, depletion continues to accelerate in some states in the region; for example, the peak rate of depletion in Nebraska is not projected to occur until after 2110 (Steward and Allen 2016). The Mississippi River Basin and Ogallala Aquifer are both large regions encompassing a multitude of stakeholders and government actors with many objectives. Unless designed in concert, a policy enacted to reduce nitrogen fluxes to the Gulf of Mexico would collide with a policy intended to preserve current levels of groundwater storage in the Ogallala. Organizations like the Hypoxia Task Force demonstrate efforts at polycentric governance, but no large-scale interventions target nutrient management across the entire region. The scientific consensus is that no single policy instrument could achieve the targeted reduction in nitrogen fluxes (Rabalais and Turner 2019), but Liu et al (2022) estimates that taxing nitrogen leaching in the region at a rate of one dollar per ton of leaching would reduce nitrogen fertilizer use by 5.7% and total leaching by 8.8%. Part of this decrease is attributable to a drop in crop output because of the increased implicit cost of nitrogen fertilizer as an input. In equilibrium, some corn and soybean production shifts to other productive regions of the US, including parts of Kansas and Nebraska.
The sustainable yield of aquifers for groundwater irrigation, where pumping is balanced by natural recharge and/or discharge, depends on the volume extracted, the timing of extraction, and the locations of wells, recharge and discharge areas (Bredehoeft 2002, Sophocleous 2005. Because current pumping vastly outstrips recharge in the Ogallala, a sustainable restriction on groundwater withdrawals is generally viewed as too economically harmful to be feasible (Sophocleous 2000). However, for the sake of illustration, imagine a simple policy that allocates a water budget prescribing how many mm yr −1 of groundwater irrigation can be applied on each acre under production. A reduction of w mm/yr-ac, relative to current average irrigation, is calibrated to produce a sustainable regional balance, even accounting for changes in cultivated acreage and other practices induced by the policy (e.g. efficiency improvements, changes in crop mix). It is assumed that this would lead to sharply curtailed production in the region, but demand for the crops produced in that region would not substantially decline, so the region's production would be partially diverted to other areas not reliant on the Ogallala aquifer.
In this example, x 1 represents the tax rate of US dollars per ton of nitrate leaching, x 2 the prescribed reduction in irrigation water (mm per year per acre), y 1 the proportional reduction in nitrate leaching, and y 2 the change in average water level below land surface. Both outcomes are measured relative to a fixed baseline and formulated such that maximizing them is desirable. Then Liu et al (2022) estimates that f 1 (x 1 = 1, x 2 = 0) = 0.088, and w is chosen so that f 2 (x 1 = 0, x 2 = w) = 0.
In the Mississippi River Basin and Ogallala Aquifer, policies designed to manage each region's major environmental challenge would reduce regional crop output and give the other region the opportunity to capture the lost output. Because of the relatively inelastic demand for agricultural products, some production would likely shift from one region to the other when either policy is implemented unilaterally. This would result in greater-than-anticipated nitrogen leaching in the Mississippi basin, as well as greater-than-anticipated groundwater withdrawals in the Ogallala. Consequently, if the policies x 1 = 1 and x 2 = w are implemented, the outcomes would be worse than anticipated: f 1 (x 1 = 1, x 2 = w) < 0.088 and f 2 (x 1 = 1, x 2 = w) < 0. While each externality is currently being tackled by multi-stakeholder, polycentric governance groups (e.g. coalitions of individual irrigation districts in the Ogallala), each coalition's focus on a single objective leaves them blind to policy collisions that could be avoided by coordination. Resolving the collision would resemble considering both policies' simultaneous implementation to identify some x 1 > 1 and x 2 > w that produces the original targets f 1 ( x 1 , x 2 ) = 0.088 and f 2 ( x 1 , x 2 ) = 0. Conceptually, this demonstrates the value of integrated systems modeling of agricultural best management practices in a multi-objective framework.

Case study: policy collision in United States food aid and agricultural development programs
The United States government potentially enacts policy collision at the intersection of food aid and agricultural development programs. Specifically, the US Department of Agriculture and the United States Agency for International Development both distribute food aid abroad, and both implement the Feed the Future (FtF) Initiative. US food aid is either distributed to recipient countries in kind, mainly as wheat through the Food for Peace program, or sold locally through the Food for Progress or Food for Nutrition programs, where the money from monetizing commodities goes to support development projects. FtF is the main US government funding initiative with the objective to increase smallholder agricultural productivity in the developing world. FtF began in 2010, in response to the global food price spikes of 2007 and 2008. It is funded by the Global Food Security Act, which was most recently renewed in 2017 (US Agency for International Development 2022). The FtF program works with people in developing countries around the world to increase income, food security, and nutrition through agricultural development. One of the main objectives is to promote agriculturalled growth in developing countries by increasing the productivity and incomes of small-scale farmers.
In theory, the objectives of distributing food aid could have minimal collision with the FtF objectives of promoting agricultural-led growth, and the initiatives could even complement each other. For example, if food aid is targeted to disaster-affected areas in developing countries where people lack income to purchase food commercially and, at the same time, if agricultural assistance is targeted towards farmers in more productive areas of those countries, then the two policies may be compatible. However, in practice these policies often collide with one another. For example, there is evidence that the amount of food aid distributed around the world is more closely associated with surplus wheat production and low prices in the United States than with the food security needs in recipient countries (Kirwan and McMillan 2007). As such, at least some percentage of the influx of imported food aid ends up being sold on local markets in recipient countries, pushing down food prices there (Levinsohn and McMillan 2007). Lower prices benefit local consumers but they have a negative impact on the incomes of many small-scale producers. These people often struggle to farm profitably and already deal with a multitude of challenges including lack of access to markets, climate volatility, and declining soil fertility (Pingali and Feder 2017). Lower commodity prices reduce developing-country farmers' incentives to purchase productivity-enhancing inputs like inorganic fertilizer, improved seed and agro-chemicals. This contributes to locking many of them in to a lowinput use, low-productivity cycle year after year.
The reality is that food aid and agricultural development policies have multiple objectives. For example, US farmers and shipping companies are major beneficiary groups of food assistance programs. Their lobbying effort led to the continued policy of moving food aid from U.S. farms on U.S. ships to recipient countries. Qian (2015) reviewed the role of foreign aid in development and showed that the existing literature is inconclusive on its benefits to recipient countries. Specifically, there is evidence that increased food aid exacerbates the duration of civil conflict in recipient countries, particularly those with a recent history of civilian conflict (Nunn and Qian 2014). According to Nun and Qian, this occurs through food aid being confiscated by local militias in recipient countries who use the food aid to prolong their armed conflict. Conversely, the literature has consistently found positive returns to investing in agricultural research and development. For example, Fan et al (2008) found that in India, 1 rupee spent on agricultural research and development in the 1980s and 1990s generated close to 7 rupees in benefits. Fuglie and Rada (2013) found that international agricultural research and development across Africa has yielded a benefit cost ratio of 6.2:1. These findings suggest that investing in the development and dissemination of agricultural technologies such as new seed varieties can have substantial benefits for small-scale farmers in the developing world. Yet if we acknowledge that US food aid programs are in part intended to benefit domestic producers by boosting demand for their product, this implies that US programs advancing agricultural productivity abroad in turn have a negative impact on domestic producers by dampening the demand for aid.
Ethiopia is an interesting case study for this type of policy collision. The country has a large population of nearly 120 million, while its climate and agro-ecology makes it susceptible to both droughts and floods. Furthermore, its historical formation and ethnic tensions have led to frequent civil conflict. At the same time, Ethiopia receives both a significant amount of food aid and agricultural development assistance from the United States. Wheat is the main food provided to Ethiopia through food aid, but wheat is also grown by many small-scale farmers, so the incoming food aid has the potential to depress wheat prices and crowd out domestic production. Levinsohn and McMillan (2007) found that wheat imported into Ethiopia as food aid does in fact lower food prices in the country. They found lower prices benefited the urban poor and the rural poor who farmed but did not produce enough food to meet their own consumption needs (and thus also purchased wheat at market). Lower wheat prices hurt the smaller share of the rural population who were commercialized enough to sell wheat. Kirwan and McMillan (2007) investigated longer-term impacts of food aid on agricultural and economic development in Ethiopia. They found that food aid imports have been more closely tied to US wheat prices than to food supply in Ethiopia, and that over the longrun food aid has not delivered clear developmental benefits in Ethiopia. Conversely, there is evidence that agricultural development programs in Ethiopia have generated some positive returns. For example, the country's comprehensive safety-net program, the Productive Safety Net Program has been found to produce positive and significant longer-run impacts (Berhane et al 2014, Debela et al 2015. This program offers food and cash transfers for the very poor; food and cash for work for the able-bodied but landless households; and extension, credit, and irrigation services for small-scale producers with land and labor access. These findings suggest that appropriately targeted agricultural development assistance such as those funded by the US government under the FtF program can produce benefits for recipient countries that outweigh those of food aid in the longer run. This is not to say that there is no role for US food aid. As the climate worsens, there will likely be an increased need to distribute food aid for humanitarian reasons. However, food aid is not a substitute for agricultural development in the long-run, and the policies should be designed to avoid collision with one another. There are several potential solutions to avoid the collision between food aid and agricultural development that would make the system more efficient and effective for small-scale farmers and consumers in recipient developing countries. The first is to reform of the US food aid system. Experts have shown that procuring food aid locally or in the region where it is needed is more efficient than shipping it in from the United States, with a time savings of 14.2 weeks on average, a 62% improvement (Lentz et al 2013). The same authors found that procuring grains locally or regionally was 50% cheaper than procuring from the US on average. This time and cost savings translates into more mouths fed in a shorter time per dollar spent. The second resolution would be to focus on targeting food aid to places where it has minimal impact on depressing food prices. This means that food aid should not increase when commodity prices are low in the US, which likely means they are low in other parts of the world as well. Food aid is more of a necessity when global commodity prices are high, such as in the wake of Russia's invasion of Ukraine in 2022. In such situations, targeted food aid can help relieve the burden of people suffering the double burden from high food prices and climate change. The third solution is to recognize that the long run goal of foreign assistance should be to develop sustainable, resilient food systems in developing countries. This means that the US government should enhance investments in education, research, and extension to develop and disseminate new agricultural technologies, market and physical infrastructure development, access to financial services, and strategies to build resiliency and adapt to climate change. Ultimately, better coordination of US food aid and agricultural development assistance will make both programs more effective, thus using US tax-payer dollars more efficiently and increasing impacts for beneficiaries in recipient countries.

Case study: landscape planning to provide multiple benefits in Costa Rica and China
Several countries have adopted approaches that can at least in part overcome policy collisions in landscape planning. We briefly describe two such cases: Costa Rica and China. In both countries, there was a recognition that existing approaches were leading to rapid declines in natural capital and unsustainable outcomes. In each country there was a deliberate redesign of the policy framework in order to improve the conservation of natural capital and provide for a wider array of benefits.
Landscapes jointly provide multiple benefits including habitat for species, filtering nutrients for clean water, storing carbon, as well as producing agricultural crops, livestock and timber, among other goods and services (Nelson et al 2009, Watson et al 2019, Brauman et al 2020. However, benefits cannot all be simultaneously maximized. There are tradeoffs in land use, particularly between intensive human use to produce economic commodities (e.g. crops, livestock, timber) and environmental benefits that require more natural habitat (e.g. carbon sequestration, water filtration, biodiversity conservation). As demonstrated previously in the conceptual framework, having some decision makers focused on one set of policy variable targets while ignoring impacts on others, and having other decision makers doing the same focused on a different set of policy variable targets, results in suboptimal outcomes. If instead, decision makers adopt a more integrated policy approach to manage landscapes with multiple policy variables in mind, improved outcomes for all policy variables relative to the status quo can be achieved (Polasky et al 2008, Pennington et al 2017. Integrated decision-making that jointly considers multiple policy variables from landscapes can minimize policy conflicts, maintain vital natural capital, and improve outcomes for sustainable development (Díaz et al 2019, Watson et al 2019. Costa Rica had one of the highest deforestation rates in the world from the 1950s to around 1990 (FAO 1990, Sánchez-Azofeifa et al 2001. Forest cover dropped from 95% prior to European colonization to around 30% in 1990 (Umaña Quesada 2019). Costa Rica enacted various policies to slow deforestation and promote reforestation, including establishing national parks and providing incentives for reforestation. Despite these policies, however, high deforestation rates continued. In 1996, Costa Rica enacted Forest Law No. 7575 that set up a payment for ecosystem services (PES) program for the provision of four forest ecosystem services: carbon sequestration, provision of clean water, biodiversity conservation and scenic amenities. In principle, PES programs can align the incentives of private landowners with public incentives by paying landowners for the contribution of their land to the value of public goods. When this is done, private landowners face the correct signals of tradeoffs between conserving forests for ecosystem service provision and deforesting to put land into crops or pastureland.
Costa Rica has greatly reduced deforestation rates since the PES program was initiated. However, it is difficult to disentangle how much credit to give to the PES program versus other factors (Robalino and Pfaff 2013). Both macroeconomic changes as well as the institution of other policy changes (e.g. making deforestation illegal) are also likely to have contributed to the decline in deforestation (Pagiola 2008). The combination of policies that Costa Rica put in place, including government reorganization so that forestry, conservation, and energy development were under the purview of a single ministry (now the Ministry of Environment and Energy), along with financial incentives including PES, is likely to have led to increased conservation of forests and the set of ecosystem services they provide (Pagiola 2008, Daniels et al 2010, Umaña Quesada 2019. The country has greatly reduced its deforestation rate and protected much of its remaining high-value conservation land, while allowing development in other areas. The coordination of policy to balance public and private benefits to landscape planning have helped to resolve policy collisions associated with conflicts between agricultural production and regulating and supporting the provision of ecosystem services.
China has also adopted policies to achieve multiple benefits from land management. The government of China embarked on a policy of ecosystem restoration and reforestation following disastrous floods and landslides in the Yangtze River Basin in 1998, which were exacerbated by deforestation of steeply sloped lands. China adopted a wide range of policies to address ecosystem degradation and deforestation that include both centralized planning and government control through zoning of land use, and incentives to local people to adopt conservation practices in areas with large returns to conservation. China mapped areas in the country important for provision of ecosystem services including water retention, flood mitigation, soil retention, sandstorm prevention, and biodiversity protection, using data from the China Ecosystem Assessment, which was begun in 2000 and done every 5 years following (Ouyang et al 2019). On the basis of this assessment, the government designated key ecological function zones that are protected from development to maintain natural capital, while other lands better suited to other uses are designated for agriculture, forestry, and urban development. The use of data and scientific analysis to determine the best areas for conservation, agriculture, forestry, and development, along with government control of land allows for a degree of rational planning to achieve multiple benefits. In addition to zoning, the government of China also initiated the world's largest PES program, the Grain to Green Program (also known as the Sloping Land Conversion Program) to provide incentives to farmers and herders to restore perennial vegetation on steeply sloping lands. This combination of centralized zoning and decentralized incentive programs has made a difference. The provision of ecosystem services and the sustainability of rural landscapes has improved since 2000 (Ouyang et al 2016, Bryan et al 2018, while China still allocates its most fertile land for agriculture and land needed for development to urban expansion. The examples of Costa Rica and China show that it is possible to design policies to improve integrated land management for multiple benefits to overcome policy collisions from siloed management. In each country, recognition that current approaches were failing in important dimensions (e.g. rapid deforestation in Costa Rica and disastrous flooding in China) led the governments to try new approaches that recognized the need to conserve nature and the provision of ecosystem services while also promoting agriculture and development.

Conclusions
In the preceding sections, we have illustrated multiple examples where coordination could be, or has been, used to avoid policy collisions where independently designed policies would produce mutually destructive interference. Provided that the costs of coordination for all policy actors are outweighed by the potential disutility of failing to meet a policy target, all parties have incentive to coordinate. As a result, we recommend that any policy process consider, at the very least, any first-order spillover effects and to investigate whether other ongoing policy deliberations would impact their own system and objectives.
Identifying potential collisions, however, can be a challenge. While the case studies provided here primarily demonstrate tensions between two policies with two primary objectives, many other examples of collisions involve larger numbers of policies, policy actors, or objectives. The transaction costs associated with resolving collisions may be higher than is obvious, given the complexity of integrated, multiobjective modeling of enlarged system boundaries. In the realm of environmental sustainability, this typically would require a combination of ecological and economic modeling to overcome the many challenges described in other papers from this special issue. At the very least, decision makers should strive to understand any potential major negative externalities associated with their own actions, applying the growing literature on telecoupling (Eakin et al 2014, Hull andLiu 2018), global-local-global analysis (Baldos et al 2020, Hertel et al 2021, or other similar frameworks; the policy actors impacted by those spillovers are candidates for producing collisions. Whether a policy is designed to manage a single target variable or multiple, the benefits associated with avoiding potential collisions justifies exploring a fuller accounting of policies' multiple outcomes, objectives, stakeholders, and spillovers. Future research developing this framework could explore constructive synergies between policies, i.e. instances where exogenous policy actions instead enhance the effectiveness of contemplated policies. Additionally, we envision instances where policy makers may be certain that another policy would impact their own system, but the directionality and magnitude of that impact is uncertain. In these cases, the urgency or pressure to act may conflict with a desire to improve understanding about the system dynamics. Our exposition here has assumed a deterministic projection of policy impacts (or at least certainty about the expected impacts), but resolution of policy collisions may require other considerations when impacts are less confidently known.