I provided these remarks at the 54th Annual Conference of AARES (Australian Agricultural and Resource Economics Society) that I am now attending in Adelaide. It is in the main a simple argument for using adaptive management techniques for managing highly uncertain and complex environmental systems. Very provisional. Revised, comments welcome.
What economists learn from ecology is that living systems are complex. But that life is complicated isn’t often a particularly unexpected or fruitful insight. Even studying highly oversimplified predator-prey models in courses on dynamical systems establishes this proposition – predation, symbiosis, extinctions, limit cycles, deterministic chaos and so on are possible in a positive sense even without raising the more difficult normative questions about what should be done to manage co-existing species. I remember decades ago asking the famous animal population mathematician Jim Cushing whether he had any optimal control results on the management of interacting populations and he looked at me with incredulity – he saw enough complexity in analysing their positive dynamics!
What the ecologists who have spoken today have done today is flesh out the sources of complexity in various dimensions in the context of environmental management of the Murray Darling Basin.
I’d like to rephrase some of their conclusions using language more familiar to economists. They are for the most part comparing ecological reality with a driver/conceptual model/outcome theoretical schema that would result from use of a reductionist economic modelling philosophy. In fact I am a strong supporter of reductionism if, for no other reason, that I see no alternative – urging adoption of holistic approaches in itself is valueness jargon. We must simplify complexity to the point where we can understand things. Identifying issues as being complex is not enough. But that does not mean we need to rely exclusively on a tunnel-visioned view of the world that ignores ecological complexities.
My preferred language is that of control theory. This replaces the idea of drivers with that of ‘controls’, the conceptual model with the notion of ‘transitional dynamics’ or ‘laws of motion’ and the idea of outcomes with that of the value of ‘targets’ or ‘objectives’. This alternative language conveys the idea that controls are derived to meet objectives or targets accounting for a system’s laws of motion. Controllability is an issue that is not presumed – it is something that needs to be established. Natural systems have self-organising characteristics that are well-understood in dynamical modelling and which will play there part in determining the evolution of the environment.
As the ecologists have mentioned today, assigned objectives are often little more than an arbritrarily extrapolated trend. I would put the issue more forcefully and say that objectives are often highly uncertain. We often don’t know what we really seek in environmental management because of ethical and environmental valuation uncertainties. Moreover, our objectives change markedly over less than human lifespan length time horizons. Think, for example, about the attitudes of Australians to flora and fauna conservation and the eclipse, within a human lifetime, of moves to build up a more resilient and interesting biodiversity via the acclimatisation movement. Partly this has reflected developments in taxonomy but also in aesthetics.
Ecologists correctly claim that ecological ‘laws of motion’ are often only very partially understood because of their complexity. They are subject to high levels of uncertainty which makes for a difficult policy task when objectives are also uncertain. In my view the core issue that modellers need to avoid is exclusive reliance on tunnel vision models. This implies the need to manage adaptively or, again to use the language of control theory, to manage’ closed loop’ by focusing on new information as it becomes available. Information surprises and, indeed, serendipity become specifically important.
Analytically social investment tasks with uncertain dynamics, irreversibilities, non-linearities and catastrophic changes can be handled using investment theory adapted to deal with such features. The basic ideas are connected with work done 35 years ago by Kenneth Arrow, Anthony Fisher and Claude Henry and form the basis for an analytical theory of why, in ecological contexts, we generally should seek to be conservative in utilising potentially destructible environmental assets. In brief, we should be cautious about taking steps which reduce our future options both because we learn and because our objectives may change. In my view economics is the primary discipline in establishing a sound decision science of ecological conservation and in promoting, on positive grounds, the case for being cautious and conservative in managing such systems.
Unfortunately developing the AFH analysis from comprising a simple though useful parable to being an operationally useful model requires information on the probability distributions of outcomes which we almost never have. Indeed, most of the time, ecologists can’t provide us with such information needed to flesh out the complexity of the models we seek to use because they don’t have it. Standard procedures of guessing parameter values and doing simulations are a substitute for careful thinking not something that advances our knowledge markedly. Doing the job properly would call for an overly detailed refinement of our understanding of the biological dynamics we observe and this is typically impractical to obtain.
Using classical decision theory, which has the advantage of requiring less data than modern stochastic optimisation techniques, yields some insights. The sacrifice is much less specificity about system structure but the benefit is much more focus on the way decisions are undertaken. We know that ideas of ‘avoiding worst-possible outcomes’ (or minimax, or the precautionary principle that the Basin Plan apparently endorses) are unhelpful – they don’t handle well the situation where expensive policies fail. Minimax regret criteria are better – they suggest a type of generalised insurance principle that, even if you don’t have probability information, take insurance actions if the costs of doing so is low relative to the damages that might occur if you don’t. These insights help.
These issues can be set out in a setting where ecological complexity provides the uncertainties. But the same analyses apply except that in such contexts you are more likely to be seeking specific policies about ways of thinking about policy action rather than grand insights. There is some provisional attempt to understand structure.
This can become waffly – the language of ‘systems science’ can become an empty set of platitudes – so let me consider a particular stylised environmental task which suggests what I think we should do in an ecological context where ecological complexity drives pure Knightian uncertainty but where the prospects of getting it wrong raises the possibility of severe long-term costs.
Consider a stylised two conservation zones problem where complexity stems from the migratory/invasive potential of a mix of species that are subject to climate change pressures and the core issue is whether or not connectivity should be enhanced to enhance conservation outcomes. The details of the specific prototype are less interesting than the way one comes to think about policy.
Suppose there are two conservation zones (‘sets of parks’) separated by developed landscape that provides a migratory barrier between the two zones. Suppose the ecological policy objectives are provisionally to seek to conserve the species that exist in these zones. Parks at B might be north of parks at A and, in response to an exogenous shock – a climatic change – species might (might!) be expected to seek to make a latitudinal relocation from B parks to A parks suggesting the case for a habitat linkage. Alternatively the species survival might be enhanced by providing them with greater resilience by increasing conserved land area sizes in parks at A and B. There are other options too such as using captive breeding programs and relocating targeted species from B to A.
At least three things might happen:
(i) Species might migrate from B too slowly to accommodate rapid climate change. If we knew that this was a likely outcome ’assisted relocation’ policies or, if these were infeasible because they were far too expensive to implement successfully, policies of doing nothing might make sense.
(ii) Alternatively species might successfully relocate from B to A without driving ecological in-balances at A so that comnnectivity would enhance sustainability. In this case excessive species predation at A does not occur and the relocations are not unbalanced in the sense that certain species do not develop to become local or exotic pest species. If this was known to be likely then the preferred policy would be to establish a habitat corridor between the zones.
(iii) Species might relocate but drive imbalances at A that wipe out species there or lead to the destruction of targeted species from B. Ecological imbalances then emerge creating pest species issues. Alternatively a corridor might enhance bush fire risks and damages. The preferred policy in this case is to not establish a habitat corridor but to increase the resilience of parks in the original zones by, for example, increasing their sizes.
What policy should be adopted here? One answer is to try all three and eventually select that subset of policies which works best. This, however, might be unfeasibly expensive because of indivisibilities. Another approach is to try to pick what are a priori seen as ‘best bet’ policies (to utilise the ‘best mathematical or conceptual model available’) but to be prepared to backtrack if emerging evidence (on invasiveness, capacity to relocate or cost) suggests that particular policy is proving less successful than anticipated. This is simply policy adaptation and implies not only a case for taking a broad view of the initial policy mix but also the need to carefully monitor emerging policy outcomes and the flexibility to revise and implement alternatives.
Of course one is monitoring not only to check if a particular policy is working well or not but more broadly to check for the unexpected. It is not presumed that the eventual preferred policy is known in advance. The monitoring needs to be part of a research program that sees potential value in serendipity. The optimality of doing nothing and of admitting policy failure are important emerging possible responses to unexpected events.
If there are no investment cost indivisibilities or time imperatives then the initial policy selections might take the form of experimental trials to reduce cost. But the core issue remains to be broad-based in initially selecting policy options, in retaining the need to be adaptive, to monitor with an open-minded research focus and to display policy flexibility based on realism not optimism.
In the Murray Darling Basin the environmental policy choices under the emerging Basin Plan are made more difficult than in this type of prototype because there is the possibility of catastrophic and irreversible environmental collapses contingent on both the level and time structure of agricultural water diversions. There remains however the possibility of learning and adaptation issues here – for example from the pre-existing Lower Lake experiences. I’d like to see how ecologists would flesh out such a scenario.
This general viewpoint has even greater validity if we admit uncertainty about objectives due to environmental valuation uncertainty or ethical uncertainties. Or if we admit that, in the future our policy predispositions are likely to change. Then we uncover an additional reason for taking a broad view and keeping options open.
With such philosophies using reductionist models is not necessarily being inconsistent with ecological realities. It is simply accounting for what we know and can understand but being broad minded enough to understand that we might initially get it at least partially wrong and thus might need to modify our management approaches.
In the Murray Darling Basin we do not need to know details of economic valuation – such as absurdly manufactured ‘environmental valuations’ to get started with worthwhile policy reforms since we know a priori that we are well-below expending enough money and effort in promoting sustainable use. In this case it might make sense to put ecological considerations (e.g. building resiliance) as the main starting points in establishing where we wish to proceed before we move towards evaluating alternative policies in conventional economic terms. Again adaptation and flexibility in assessing the quality of policy responses will be a core ingredient in managing system complexity.
Recent Comments