1 The Challenge of Degraded Environments: How Common Biases Impair Effective Policy Alan Berger, Case Brown, Carolyn Kousky, and Richard Zeckhauser ABSTRACT Economic activity can damage natural systems and reduce the flow of ecosystem services. The harms can be substantial, as our case studies vividly illustrate. Most degraded landscapes have at least some potential to be reclaimed. However, uncertainty plagues decision making regarding degradation and reclamation, in relation to the extent of the damage, the success of reclamation, and how exposure will change in the future. We examine how a range of observed decision biases can lead to far-from-optimal policies regarding how much degradation to allow and when, as well as how and how much to reclaim degraded sites. Despite our focus on degraded landscapes, we believe these are generic biases present in a wide range of risk situations. Our three case studies show these biases at work. The first two studies are of mining operations in the United States and Canada, and the third is of climate change. KEY WORDS: reclamation, decision making, biases, neglects, mining, climate change Forthcoming, Risk Analysis: An International Journal2 1. INTRODUCTION Natural systems produce ecosystem services—the benefits people obtain from ecosystems [1, 2]. Economic activity can sometimes impair natural systems or subject them to risks, reducing the expected amount and type of services flowing from them. In extreme cases, degradation can turn a valuable system into one that generates harms on net. After such degradation, it may be possible to reclaim the site, reestablishing or enhancing a flow of ecosystem services. Reclamation refers to any process that moves a site from degradation to an improved state. In practice, decision makers are often uncertain about how much reduction in value or how much damage an activity will cause, as well as the extent to which reclamation activities will be successful. This paper looks at the decision making surrounding degradation and reclamation of altered or degraded landscapes. When and how should we reclaim a degraded area? How much reclamation should be done? If it is known before the degradation occurs that reclamation is feasible, how does that alter optimal levels of alteration? In these situations, decisions should be made according to the rational-choice model, which involves using probabilities (usually subjective assessments), valuing and including all the benefits and costs flowing to others as well as to selves, accurately employing statistics to assess data and make predictions, and considering all possible alternatives, with no thumb on the scale for the original situation. This paper focuses on a set of seven common biases that impede our thinking about degradation and reclamation decisions, often leading to sub-optimal outcomes. We have referred to the first five previously as “neglects” [3]. The first bias, probability neglect, is the failure to consider appropriately the probabilities of consequences. Decision makers focus too heavily on outcomes, ignoring their likelihood of occurrence. The second bias, consequence neglect, is its3 complement: decision makers focus on the probability without considering the impact of the outcomes. The third bias, statistical neglect, occurs when decision makers fail to use data to predict the distribution of outcomes. The fourth bias, solution neglect, refers to the tendency to fail to consider the full range of possible solutions to a problem. The fifth bias, external risk neglect, applies when agents do not consider the benefits or costs their actions impose on others. Sixth, we consider status quo bias, a seemingly undue preference for the current situation, even when change may be preferable. A specific form of status quo bias relevant to reclamation decisions is authenticity bias. Authenticity bias reflects a preference for a former status quo, even when the original state has been lost. Loss aversion can contribute to status quo bias, as individuals calculate the loss of what they have as greater than the gain to be received, leading them to reject any changes from the status quo [4]. A particular form of loss aversion, multi-attribute loss aversion, arises when evaluating alternatives with multiple dimensions. In this case, individuals will not accept a loss from their current position on any one aspect, even if in doing so, they could move to a superior position overall by gaining more of another attribute that is of greater value. Multi-attribute loss aversion often prevents the adoption of socially superior solutions that impose a loss (even if compensated) on a subset of individuals. The seventh and last bias we discuss is faulty treatment of tail probabilities. Unlike a normal distribution, a fat-tailed distribution occurs when extreme outcomes—those that are more than three standard deviations from the mean—have a non-negligible probability. The faulty treatment of tail probabilities is a common failing that arises when people simply assume that the relevant risk is normal, when the distribution is, in fact, fat-tailed. Statistical assessments are then based on assumptions that are appropriate for normal distributions but not for fat-tailed distributions. With a normal distribution, an event three standard deviations or more from the4 mean occurs once every 370 trials; a four standard deviation event happens once every 15,787 trials. But events this extreme occur much more commonly in environmental contexts, say on the size of oil leaks, or lives lost to natural catastrophes. Fat-tailed distributions apply. Once fat tails are present, further difficulties arise. Analysts often focus on the most likely or high, medium, and low outcomes. However, with fat-tailed distributions, extreme outcomes account for most of the expected value. Two other heuristics push in the opposite direction, toward overly cautious thinking. Drawing on common economic models of utility, such as a logarithmic utility function, people can mistakenly assume that extreme negative outcomes cause infinite or near-infinite losses. (The logarithm of 0 is – ∞.) Like the heuristic of relying on the normal distribution, this represents the failure to validate a commonly used function with empirical experience. Simply put, we are not willing to in practice, and should not be in theory, to sacrifice nearly everything to reduce a small probability, but