The sleeping policeman: understanding issues of enforcement and compliance in conservation
Aidan Keane, Department of Life Sciences & Centre for Environmental Policy, Imperial College London, Silwood Park, Buckhurst Road, Ascot SL5 7PY, Berkshire, UK.
Rules governing human behaviour are at the heart of every system of natural resource management. Without compliance, however, rules are meaningless so effective enforcement is essential if conservation is to be successful. There is a large body of theory concerning enforcement and compliance with rules spread over several disciplines, including psychology, economics and sociology. However, there have been few attempts to extend this theory to conservation applications and there is little practical guidance for managers and conservation planners on the optimal design of enforcement programmes. We review approaches to understanding why individuals break rules and how optimal policy choices can reduce rule-breaking, highlighting research which has specifically dealt with natural resources. Because of the difficulty of studying rule-breaking behaviour directly, modelling approaches have been particularly important and have been used to explore behaviour at the individual, group and institutional levels. We illustrate the application of models of enforcement and compliance to conservation using the African elephant Loxodonta africana as a case study. Further work is needed to create practical tools which can be applied to the design of enforcement measures in conservation. Particular challenges include understanding the importance of violations of rationality assumptions and incorporating intertemporal choice in models of decision making. In conclusion, we argue that a new field of robust theory and practice is urgently needed to ensure that issues of enforcement and compliance do not undermine conservation initiatives.
Introduction: rules in conservation
Managing biological resources requires that rules of behaviour are followed. These rules might be agreed at any scale, from the international (e.g. EU fishing quotas) or national (e.g. National Parks) right down to the local (e.g. community reserves). They can involve a range of institutions, from governments to rural communities, and may be externally imposed or have evolved in situ. Whatever their provenance, rules and the management systems that depend on them, are worthless without compliance. However, compliance with the rules of resource management systems cannot be taken for granted. Resistance to conservation measures can arise because of differences in the spatial and temporal distributions of the resulting costs and benefits (Wells, 1992). For example, significant costs are often borne by local individuals who depend heavily on the resource, while the benefits arising from conservation may be less immediate and accrue to society as a whole (Balmford & Whitten, 2003; Chan et al., 2007). Successful management of natural resources therefore requires consideration of how rule-breaking behaviour can be discouraged in resource users. Despite its importance, this issue has not received sufficient attention in the conservation literature.
Enforcement – monitoring adherence to rules and agreements and punishing infractions when they are detected – is an essential part of successful conservation and natural resource management (NRM) (Ostrom, 1990; Gezelius, 2002; Walsh et al., 2003; Rowcliffe, de Merode & Cowlishaw, 2004; Gibson, Williams & Ostrom, 2005). Punishments may take various forms, from fines and prison terms to social sanctioning, depending on the enforcement system. Several studies of illegal hunting have shown that reducing the effort devoted to enforcement (e.g. lowering investment in equipment and training, or patrolling less frequently) increases the number of poaching incidents and can harm wildlife populations (Arcese, Hando & Campbell, 1995; Jachmann & Billiouw, 1997; de Merode et al., 2007). For example, investment in enforcement has been an important determinant of changes in the buffalo, elephant and black rhino populations in the Serengeti National Park, Tanzania (Hilborn et al., 2006). Similar effects have been seen in marine systems where effective enforcement of marine protected areas has been shown to reduce poaching-driven changes to reef fish communities (Walmsley & White, 2003; Floeter, Halpern & Ferreira, 2006; Samoilys et al., 2007).
Enforcement is costly, however, requiring investment in training, equipment and salaries. It may also have other costs. For example, enforcement activities can erode trust between local people and conservation authorities (Infield & Namara, 2001) and undermine traditional systems of resource management (Wilshusen et al., 2002; Horning, 2003; Gelcich et al., 2006). With limited resources available to conservation, particularly in the developing world (Balmford et al., 2002), enforcement at a level which produces no infractions can be prohibitively expensive. Techniques for optimizing enforcement strategies – maximizing benefit while minimizing cost – should therefore be of great interest to practical conservation.
Directly studying compliance and its determinants is problematic because rule breakers are usually unwilling to reveal themselves or to discuss their motivations freely for fear of punishment. Data on illegal activity are therefore potentially prone to unquantifiable biases. Consequently, much of our understanding of this topic stems from modelling studies which provide powerful methods for explicitly addressing these uncertainties and help managers and policy makers to predict the impacts of future changes to enforcement regimes. We review models of compliance with rules, focusing on those that have been applied to conservation and NRM. We structure our review according to the scale at which decisions are analysed, moving from the individual to the group and institutional levels. African elephants are used as a case study to illustrate how such approaches have been applied in practice. Finally, we highlight several areas where we feel modelling can contribute more to our understanding of rule-breaking behaviour by resource users.
Individual level models
Many theories attempt to explain why non-compliant behaviour occurs and how it can be discouraged. Psychological theories of compliance with rules and social norms often assume that the decision-making processes of rule breakers differ from those of other people. Cognitive theories of compliance argue that these behaviours stem from differences in personal moral development (e.g. Goslin, 1969; Tapp & Levine, 1977; Kohlberg, 1984). Another important group, the social learning theories (e.g. Sutherland, 1947; Burgess & Akers, 1966; Akers, 1985), suggest that individuals' decision-making processes are conditioned by interactions with their environment.
By contrast, sociological and economic theories of compliance assume that the decision-making process in rule breakers is not fundamentally different from that in other people. Normative theories argue that an individual's perceptions of the legitimacy and fairness of rules are crucial to decision making (Tyler, 1990). Instrumental theories, on the other hand, hold that acts of non-compliance occur because the benefits anticipated by the decision maker outweigh the costs (Becker, 1968).
The study of optimal enforcement has largely focussed on instrumental approaches, using economic models to answer the question of how best to modify individual incentives in favour of compliance. An individual's supply of offences may be modelled as a decreasing function of two factors of enforcement: the probability of an act of non-compliance being detected and punished and the severity of punishment that results (Becker, 1968). The process of detection is inherently costly, requiring law enforcers to be paid and equipped, whereas punishments may take the form of fines (assumed to be a costless transaction). This suggests that the optimal enforcement strategy is to reduce the amount of costly monitoring while increasing the size of penalty, thereby maintaining offences at an acceptable level with lower enforcement costs.
There are, however, several reasons why severe penalties may be undesirable. Extensions to Becker's model suggest that if sanctions are socially costly (Kaplow, 1990) or if corruption is present (Becker & Stigler, 1974), the optimal fine level may not be the highest possible. Similarly, if individuals are risk averse (Polinsky & Shavell, 1979), are imperfectly informed about their probability of being caught (Bebchuk & Kaplow, 1992), respond to penalties by trying to avoid detection (Malik, 1990), or vary in their wealth (Polinsky & Shavell, 1991) the optimal level of fines may be reduced. Severe penalties are also morally questionable and can lead to an increase in serious crimes relative to less damaging offences due to the loss of marginal deterrence (Stigler, 1970).
A large number of studies have attempted to empirically test the deterrence effect of enforcement measures upon crime rates in developed countries, with mixed results (see Cameron, 1988 for a review). Ehrlich (1996) argues that there is such an effect and that the probability of detection may be more influential than severity of punishment. However, issues such as the use of data at different levels of aggregation, uncertainty about the level of private protection and the difficulty in separating the influence of deterrence and incapacitation leave many studies open to criticism (Cameron, 1988; Ehrlich, 1996).
Early bioeconomic models of NRM assumed enforcement was costless and produced perfect compliance (e.g. Clark, 1990). The implications of imperfect enforcement in NRM were first explored in commercial fisheries. In quota-restricted single-species fisheries, for example, enforcement costs may be modelled as an increasing function of the stock size and the legal quota. Consequently, the larger the desired stock size (above the open-access equilibrium) the greater the necessary expenditure on enforcement (Sutinen & Andersen, 1985). More generally, the optimal level of enforcement is attained when the marginal cost of enforcement is equal to its marginal benefit (Becker, 1968; Sutinen & Andersen, 1985; Hallwood, 2005). Other models of fisheries enforcement have considered differences between input controls (e.g. time at sea, equipment) and output controls (e.g. landing quotas) (Mazany, Charles & Cross, 1989), and shown that avoidance behaviour affects the socially optimal level of enforcement (Anderson & Lee, 1986; Anderson, 1987).
Several studies have attempted to empirically test the predictions of fisheries enforcement models. Survey data from the Massachusetts lobster fishery show an increasing rate of compliance as the perceived probability of being caught increases (Sutinen & Gauvin, 1989). Similarly, data from Quebec fisheries show a greater influence of the probability of detection than severity of punishment on offences (Furlong, 1991). Data from federally managed US groundfish fisheries, on the other hand, suggest that a decline in compliance from 1982 to 1988 was best explained by poor stock conditions and high market prices, with enforcement having a negligible effect (Sutinen, Rieser & Gauvin, 1990).
The effects of the design of enforcement on poaching decisions have also been explored. Using a model of multi-species bushmeat hunting as a component of the household economy, measures targeting bushmeat sales were shown to be more effective than those targeting hunting directly (Damania, Milner-Gulland & Crookes, 2005). The benefits for different hunted species are complicated by technology switching (e.g. between snaring and gun hunting), however, and are therefore ambiguous. Clayton, Keeling & Milner-Gulland (1997) investigated economic deterrents to hunting two wild pig species in Indonesia, only one of which can be legally hunted. A fine on market dealers for selling the illegal species was shown to be most effective, and more equitable than other approaches considered because it does not affect the welfare of individuals who hunt the legal species.
Morality, equity and justice
Alternative models of compliance with regulations emphasize the role of normative factors, such as moral obligation and perceptions of fairness and justice. Normative factors have been incorporated into economic frameworks by assuming that an individual's utility is increased by performing actions that are socially acceptable or beneficial (Sutinen & Kuperan, 1999; Nielsen, 2003). The perceived legitimacy of rules, related to both the fairness and efficiency of the regulatory process and the justice and effectiveness of its outcomes, also affects their acceptance by resource users (Hønneland, 1999; Sutinen & Kuperan, 1999).
In Norway and Newfoundland, some small fisheries achieve high levels of compliance despite low levels of formal enforcement. Gezelius (2002, 2004) suggests that this results from informal sanctions based upon collective moral judgements. Non-compliant individuals are subjected to social opprobrium if their actions are perceived to confer unfair advantages or to be carried out for monetary gain, but not if they are necessary to secure an adequate basic income.
Quantitative empirical evidence on the influence of normative influences on compliance is, however, weak. Nielsen & Mathiesen (2003) identified factors which have a major influence on compliance in small Danish fisheries. The most important factors were instrumental: economic gains and deterrence measures, but normative considerations such as the fairness of rules were also represented. Hatcher et al. (2000) similarly reported a significant positive relationship between perceptions of fairness and participation and levels of compliance in the UK fishery, but Hatcher & Gordon (2005) failed to reproduce this result, finding instead that economic incentives dominate.
Group level models
In many situations an individual's costs and benefits are affected by the behaviour of others. Understanding decision making then requires a strategic perspective which has been modelled using game theory. For example, game theoretic approaches have been applied to study the interaction between an enforcement officer and a resource user. In ‘inspection games’ one player chooses whether or not to monitor the behaviour of another, who chooses whether or not to commit an offence (Tsebelis, 1989). Enforcers are treated as rational, utility-maximizing entities. If players interact only once, increasing the severity of penalty does not reduce the number of offences, but instead lowers the (costly) effort devoted to detection by enforcers (Tsebelis, 1989; Andreozzi, 2004). With repeated interactions, increasing the reward enforcers receive for catching criminals does not reduce the number of offences, and might increase it because enforcers can maximize their profit by monitoring less, reducing their costs and encouraging a greater number of offences and bonuses (Andreozzi, 2004). That these results are sensitive to the precise formulation of the game (see Weissing & Ostrom, 1991) highlights the complexity of modelling strategic behaviour in enforcement problems.
Game theoretic approaches have also been used to study common-pool resources (CPR) where monitoring and enforcement are not the preserve of specific individuals or designated agencies but are instead carried out by the resource users themselves as part of a cooperative effort to manage a natural resource (see Heckathorn, 1996, for a review of games which display the properties of cooperative systems). Early paradigms in the analysis of CPR (e.g. Olsen, 1965; Hardin, 1968) were formalized as a prisoners' dilemma game (Dawes, 1973) and dealt with open access situations where there is little incentive for rational, self-interested individuals to moderate their exploitation in anticipation of future benefits since others may not follow suit. Cooperation can be achieved in the prisoners' dilemma under certain conditions (e.g. relatively small group sizes) by allowing repeated interactions (Axelrod & Hamilton, 1981). As described above in small fisheries (Gezelius, 2002, 2004), cooperation can also emerge and persist under less restrictive circumstances if, despite incurring a cost, individuals enforce rules by voluntarily punishing non-cooperators: the strategy of ‘altruistic punishment’ (Fehr & Gachter, 2002; Fowler, 2005).
Game theoretic models also allow the long-term stability of cooperative agreements to be assessed. Mesterton-Gibbons & Milner-Gulland (1998) model a cooperative NRM system to identify conditions under which a community who do not poach and monitor each others' compliance can be stable against invasion by individuals who poach and do not monitor. They find that people must be paid to monitor, even in the absence of poaching. Shared benefits are not sufficient to motivate protection of a communal resource without incentives for enforcement. Furthermore, cooperation breaks down at small community sizes because the likelihood of an infraction being detected becomes too low.
Institution level models
Some aspects of enforcement are better explored from the point of view of an institution, rather than individuals. For example, the ability of a private authority with legal harvesting rights to prevent poaching has been modelled under different property structures and economic environments (Skonhoft & Solstad, 1998). With well defined but imperfectly enforced rights, the effective property structure and long-term stock levels are affected by economic variables such as the profitability of alternatives, cost of enforcement, owner's discount rate and the resource's market price and non-consumptive value. Some effects are surprising. For example, Skonhoft & Solstad's model predicts that a government intervention to lower enforcement costs would not raise the optimal wildlife stock but allows the owner increase legal harvest because the illegal harvest can be further reduced for the same expenditure.
In many cases, management agencies may be required to cover a proportion of their operating costs. Where non-consumptive uses of wildlife such as tourism are not viable, revenue might be generated by selling permits to hunt common species and fining unlicensed exploitation. A model of a Western African wildlife department suggests that this approach could indeed benefit endangered species (Robinson, 2004). However, hunters using non-selective technologies (e.g. snares) may not be able to restrict their catch to legal species (Bowen-Jones, Brown & Robinson, 2003). Punishing the capture of rare species therefore risks causing significant waste by encouraging hunters to discard animals that were killed illegally rather than risk sanctions (cf. bycatch in quota-limited fisheries).
A case study: elephants
In the 1970s and 1980s high levels of poaching, stimulated by high ivory prices, threatened the survival of the African elephant Loxodonta africana and prompted much debate about how illegal hunting should be controlled given the resource constraints of governments in the elephants' range states. The species therefore provides an illustration of how models of enforcement and compliance have been used to inform conservation policy.
The elephant population of the Luangwa Valley, Zambia, has been particularly well studied. From 1972 to the mid-1980s the area lost c. 75% of its 100 000 strong population (Leader-Williams & Albon, 1988). Although anti-poaching patrols received significant investment from 1979 they largely failed to prevent further decline (Leader-Williams & Albon, 1988). Data from 1979 to 1985 suggest that although these patrols were well motivated and effective, they were not sufficiently numerous to control illegal hunting over the entire area (Leader-Williams, Albon & Berry, 1990).
Bioeconomic modelling of individual behaviour provides a means of predicting how effective different approaches to tackling poaching in the Luangwa Valley would have been. One such model shows that a fine which increases according to the number of trophies in a poacher's possession is a more effective deterrent than a fixed fine, but that increasing the severity of punishment is less effective than increasing the effort devoted to detecting and prosecuting poachers (Milner-Gulland & Leader-Williams, 1992; Leader-Williams & Milner-Gulland, 1993). However, sensitivity analyses suggest that the returns to hunting were so high during this period that unrealistic increases in enforcement effort would have been necessary to reduce poaching to an acceptable level (Milner-Gulland & Leader-Williams, 1992).
By 1989. continuing elephant declines across the continent (Stiles, 2004) led to the species being listed in Appendix 1 of the Convention on the International Trade in Endangered Species of Wild Flora and Fauna (CITES). Banning the international trade in ivory was intended to depress demand at a global scale, reducing the incentives to hunt illegally and thereby facilitating the enforcement of national anti-poaching laws. However, the success of the ban is unclear. A series of models intended to assess the ban's effects on incentives to poach have produced ambiguous results, varying according to their particular assumptions and parameterizations (Stiles, 2004).
For example, Jachmann & Billiouw (1997) compare a set of institution-level models of investment in enforcement, arguing that the variation in elephant mortality observed in the Luangwa Valley between 1988 and 1995 can be explained by changes in enforcement, without any need to invoke the effect of the ban. Bulte & Van Kooten (1999), on the other hand, argue that within the range of parameter values estimated for the period 1979–1985, and assuming a discount rate >5%, the ivory ban should have increased elephant numbers. Their analysis also suggests that the response of elephant populations to changing enforcement levels is greater if trade is allowed than if it is not.
Expectation of future management policies can affect current prices and therefore influence incentives to poach. Kremer & Morcom (2000) warned that anti-poaching policies that are expected to reduce the future supply of ivory could raise incentives to poach by creating the expectation of price rises. Using a dynamic institution-level model Kremer & Morcom (2000) argued that if managers can credibly commit to tough enforcement should elephant populations fall, the incentives to poach caused by anticipated higher ivory prices may be reduced. Where tough enforcement is not credible, creating stockpiles of ivory and threatening to sell this should populations fall, may also be effective at reducing poaching. Bulte, Horan & Shogren (2003) counter that the CITES ban might create incentives for governments to harvest their elephant populations to extinction if the prices for stored ivory are sufficiently high and if extinction is expected to precipitate the lifting of the trade ban. Although limited by the availability of suitable data, an attempt to assess the effects of a ‘one-off’ sale of stockpiled ivory in 1999 using mortality data from Kenya and Zimbabwe suggests that it had little effect on overall elephant poaching levels (Bulte, Damania & Van Kooten, 2007).
Challenges for future studies of enforcement and compliance
The preceding sections have highlighted the strengths of modelling approaches as tools to inform debate about the design and implementation of enforcement measures. However, we feel there are several ways in which models of enforcement and compliance can be further improved. Our review demonstrates that models of enforcement have tended to focus on economic factors influencing decision making, with less emphasis on research from the fields of psychology and sociology. Below, we discuss the scope for creating richer models of human behaviour, relaxing common assumptions about rationality, uncertainty and intertemporal trade-offs by rule breakers. Both our review and the case study of elephant conservation highlight the sensitivity of model outputs to their precise specification. Models must therefore be developed with a good understanding of the realities of the system being studied and parameterized with appropriate data. We briefly discuss the challenges of collecting such data below.
Rationality and uncertainty
Decisions under uncertainty, such as whether to break an imperfectly enforced rule, have traditionally been modelled using the expected utility framework. Utility is a measure of relative satisfaction and expected utility is defined as the mean utility received under risk. However, the use of the expected utility framework to explain decision making under risk is undermined by experimental evidence that its core assumptions are frequently violated in practice (Schoemaker, 1982). For example, people have been shown to evaluate loses and gains differently, to make decisions based on reference points rather than absolute values and to be influenced by the framing of choices as well as their anticipated values (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992).
Indeed, although economic models of human decision making generally assume that individuals are rational and act to maximize their utility, much of the psychological research into decision making suggests that these assumptions are not realistic (McFadden, 1999). Instead individuals may have bounded rationality, limited by cognitive resources, and employ a variety of heuristic procedures to achieve outcomes that are ‘good enough’ rather than truly optimal (Conlisk, 1996). Differences in the decision-making processes employed by different individuals might arise from their previous experiences, as suggested by psychological theories of compliance, and render some more likely to break rules than others.
The significance of deviations from rationality assumptions for models of enforcement and compliance is currently unknown. Future research in this area could focus on identifying and testing the ‘rules of thumb’ used for decision making in specific situations. Work is also needed to assess how adopting alternatives to expected utility, such as prospect theory (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992), could affect model predictions.
Many choices made by individuals depend on how they trade-off costs and benefits at different points in time (Frederick, Loewenstein & O'Donoghue, 2002). Models of NRM have generally dealt with intertemporal issues in a simplistic manner, with individuals having a single, fixed discount rate for all situations. Such models have suggested that slowly reproducing populations are more likely to be harvested to extinction if hunters have high discount rates (Clark, 1973). High discount rates may also affect the perceived severity of punishments, with future consequences (e.g. the later portions of long prison terms) devalued relative to more immediate sanctions (Leader-Williams & Milner-Gulland, 1993).
In reality discount rates, as well as other factors influencing an individual's decision making, may change through their life and with their circumstances (Edwards-Jones, 2006) and are likely to vary between individuals. Currently, however, factors affecting intertemporal choices are poorly understood. For example, it has been claimed that poverty forces individuals to make decisions on a short-term basis, neglecting resource conservation, but there is evidence of desperately poor people choosing long-term gains (or long-term stability) despite a short-term cost (Moseley, 2001). Further work is needed in this area to study the determinants of discount rates in order to better predict how intertemporal trade-offs will affect rule-breaking behaviour.
Ultimately, models can only take us so far. Our case study of elephant conservation illustrates that while models can be powerful aids to decision making, the details of their implementation and parameterization are crucial to their interpretation. The development of a theoretical framework for enforcement must therefore be underpinned by good data if it is to provide a solid basis for action. Many attempts to validate theories of enforcement with empirical evidence have been unconvincing, often because suitable data are simply not available (Cameron, 1988; Ehrlich, 1996). In conservation settings, data on non-compliance are frequently a by-product of attempts to deter rule breaking, limiting their quality. However collecting more detailed data, such as spatial patterns of non-compliance and enforcement effort, poses serious logistical challenges and may not be justified under local conservation budgets.
In order to ensure research into rule breaking can be used in practical conservation, a close reciprocal relationship between models and data is needed. Models can guide data collection and help to determine the minimal data requirements for robust decision making. Salafsky, Margoluis & Redford (2001) have promoted an adaptive management approach to ecological monitoring and project appraisal. Such an approach could be taken with enforcement to allow data to be collected in a more targeted and systematic manner. Future research should also explore other potential avenues for the collection of data about rule breaking including novel interview methods for the collection of sensitive information, such as the randomized response technique (Solomon et al., 2007).
Ultimately, as models become more complex, their data requirements might render them impractical as tools for management decision making. Although this trade-off between complexity and reality is common to all modelling approaches, the paucity of data for many exploited species amplifies the problem in NRM (e.g. Smith, 1993) and good data are rarely available for threatened species. In some cases, therefore, it may be desirable to identify situations where rules of thumb can adequately inform day-to-day decision making.
Rules, whether implicit or explicit, are at the heart of every conservation and NRM system but compliance cannot be taken for granted. Success depends on the ability of managers to influence the behaviour of resource users, and enforcement therefore has a vital role to play in the conservation of natural resources. To date the literature on this issue has been scattered among a number of disciplines, and theoretical insights from other fields have not been fully and consistently applied to NRM. We believe there is a need to develop a new field of robust theory and practice for enforcement and compliance in conservation, building on the experience of others. Models of enforcement have been important in predicting how individual incentives can be modified to improve compliance with rules but further work is urgently required to broaden our understanding, to validate models with empirical data and ultimately to produce practical guidelines for the optimal use of enforcement measures in conservation. If conservationists are caught napping on issues of enforcement, both the natural resources that we set out to manage and those who depend on them may suffer.