PEA Soup is pleased to host this Ethics discussion, on “What the Demands of Sustainability Can(not) Be,” by Susumu Cato and Katie Steele, with a précis by Krister Bykvist.
Over to Krister.
To say that it is important that we manage our resources in a sustainable manner does not raise many eyebrows. It is an ideal that we find in mission statements of universities, businesses, religious organisations, and national and regional governments. It is even written into the constitutions in some countries, such as Sweden. However, it is not easy to spell out what exactly it means to manage resources in a sustainable manner and why it matters morally.
In their excellent Ethics paper ‘What the demands of sustainability can(not) be’, Susumu Catu and Katie Steele zoom in on an influential take on sustainability: ‘(t)he present generation must choose in a way that leaves open the possibility that welfare (or whatever else matters morally) is maintained or improved from one generation to the next’ (808). They show that such accounts run into impossibility theorems, i.e., they fail to satisfy a set of very plausible requirements on a sustainability constraint. They then focus their discussion on what they think are the two best ways out of this impossibility.
To show the impossibility theorems, Cato and Steele introduce a simple but very useful model, in which here is a temporally ordered finite number (at least three) of generations. The generations are assigned comparable welfare levels, represented by real numbers. (A welfare level can be seen as any value of a life that matters for choice). An intergenerational trajectory is a sequence of real numbers, that represents the welfare levels of generations at successive times: (w0, w1, …, wn), where w0 is the wellbeing of the current generation, w1 the wellbeing of the next generation, and so on. A menu (set of options) is defined as a set of trajectories. Finally, there is a choice function that returns the permissible trajectories (or, more exactly, the trajectories the current generation’s choice has to be compatible with). All sustainability constraints are then formulated in terms of a choice function that privileges sustained trajectories, those trajectories whose welfare levels are non-decreasing between each pair of adjacent generations. This assumption about what it is to be sustained can be weakened without affecting the impossibility theorems. For example, one can instead say that a trajectory is sustained if each future level of wellbeing is at least as high as the wellbeing of the current generation.
Cato and Steele first show that it is a non-starter to say that only sustained trajectories are permissible, since there need not be any such trajectories. They also show even if we add the qualifier ‘if there is such a trajectory’, we are not out of the woods, since this qualified constraint will not privilege Pareto-efficient trajectories. For example, it seems evident that we should privilege (2, 3, 4) over (1, 2, 3). They argue that the best way to guarantee Pareto efficiency is to go for what they call Hard-Constraint Sustainability that says that if there is a sustained trajectory that is Pareto efficient, then all of the permissible trajectories are sustained and Pareto efficient. The first impossibility shows that this constraint cannot be combined with the following requirements: No Prohibition Dilemmas (there must be at least one permissible option); Contraction Consistency (if a trajectory is permissible in a menu, then it is permissible in any subset of this menu in which the trajectory is available); and Unrestricted Domain, (the set of menus contain all possible subsets of the set of all logically possible trajectories).
Is there a way out of this impossibility? Cato and Steele first consider the option of giving up No Prohibition Dilemmas, but they are not happy with this solution, since prohibition dilemmas paralyze choice, and, furthermore, giving up this condition does not help if we accept some other intuitive requirements, such as Path Independence. It is better to avoid the impossibility by rejecting Unrestricted Domain or revising the sustainability constraint so that it no longer gives all moral importance to sustainability. Regarding the first option, they show that one can prove Hard-Constraint Sustainability, if one makes some empirical assumptions about the economy, technology, or productivity. Now, some of these assumptions are not very realistic, which makes the proof limited in scope. The most realistic assumption does entail Hard-Constraint Sustainability, if we assume Anonymity (all permutations of a trajectory have the same value) and Paretian Ordering (a transitive and complete ordering that ranks a trajectory higher than another if it Pareto-dominates the other).
This last proof shows that sustainability has only derived significance, since it is derived from some general moral principles. Moreover, the sustainability constraint is not very discriminating, since it is compatible with a large set of different foundational principles, including utilitarianism and prioritarianism. This does not make sustainability useless, however. Cato and Steele argues that it can still be used as a heuristic: if your choice complies with the sustainability constraint, it ‘may conform with whatever is the privileged foundational moral theory’ (833).
The second way out is to revise the sustainability constraint so that it allows for trade-offs between sustainability and wellbeing promotion. We can still assign fundamental importance to sustainability by stating that an unsustained trajectory is impermissible, if there is a sustained trajectory available that is a permutation of the unsustained trajectory. For example, (100, 100, 99) is impermissible, if (99, 100, 100) is available. But sustainability has only pro tanto significance, for we should add that a sustained trajectory, such as (99, 100, 100), is not permissible, if we also have the option of greatly increasing the wellbeing of any of the generations. For example, we may want to judge that (99, 100, 100) is impermissible, if (150, 100, 100) is available.
Cato and Steele have some doubts about this way out, for they wonder why we should care about the temporal positions of goods and say that e.g., (x++, x+, x) is always impermissible when (x, x+, x++) is available, when we do not care about the spatial position of goods.
The discipline should be grateful to Cato and Steele for introducing philosophers to models and results that rarely travel outside the economics departments. Unlike most economists, they also show much more clearly how these formal results help us decide on the moral significance of sustainability. Though I am in broad agreement with many of their points (something that is borne out by their references to my unpublished manuscript), I want to end with some comments that I hope can get the discussion started.
First, it is not clear from the model whether we should consider the past when assessing a trajectory. The starting point is time 0 and the choice function seems implicitly to be indexed to this time. Should the generation at t1 just consider their future trajectories and ignore the wellbeing level at t0? Or should they try to continue on a sustained trajectory that the generation at t0 started? For some accounts of sustainability this matters. If we need to make sure that all future welfare levels are at least as high as that of the privileged generation, we can get different verdicts depending on whether that generation is assumed to be in the past or in the present.
Second, the model assumes that all possible trajectories have the same length. But that does not seem to be a very realistic assumption. Some trajectories can be shorter than others due to future external events, such as meteorite collisions. Others can be shortened by the repeated mismanagement of resources by generations through unmitigated climate change. How should we extend the model to different-length cases? It seems intuitive to say that wellbeing is more sustained in (1, 2, 3) than in (1, 2), and more sustained in (2, 2, 2) than in (2, 2). After all, one way to sustain wellbeing is by making sure there will exist future people who have at least as much wellbeing as the current generation.
Third, I used the term ‘more sustained’ above, which seems perfectly fine. This use has intuitive support even for same-length cases. If what matters is that trajectories are continuously non-decreasing, it seems evident that (1, 2, 3) is more sustained than (1, 0, 3), which in turn is more sustained than (3, 1, 0). Perhaps being more sustained should be understood as coming closer to being sustained in an absolute sense. In any case, introducing a graded notion opens up the possibility of using graded sustainability principles instead of constraints.
Fourth, one can of course share Cato and Steele’s scepticism about giving any intrinsic significance to the temporal shape of welfare sequences, but that might be premature (as they themselves suggest in a footnote). After all, there seems to be something attractive about the view according to which an uphill life, which is constantly improving, is preferable to a downhill life, which is constantly getting worse, even if the latter is just a permutation of the momentary wellbeing in the former. Similarly, a relationship that starts well and ends poorly seems worse than one that starts poorly and ends well. The question is whether this idea can be applied to generations. It is perhaps not too farfetched to claim that humanity, seen as a species developing through time, can have an uphill or downhill trajectory and that the former is preferable.
Fifth, Cato and Steele suggest that, if a sustainability constraint has only derivative significance, one could still treat it as a heuristic. A better option is to see it as a mid-level principle, a principle that in certain contexts has the backing of a set of different fundamental principles. Mid-level principles are useful since they can be embraced by people who disagree about the fundamental principles. That also makes them more relevant for policymaking in societies with divergent moral views.
Many thanks to Krister Bykvist for his succinct summary of our paper and for raising excellent questions. We owe Krister extra thanks for early discussions and for work of his own that inspired our work on this project.
@Krister, let us respond to your questions in turn:
1. On ignoring the past:
This is a very good point about whether the past matters in assessing whether a choice satisfies the sustainability constraint. We had in mind that the past does not matter in our investigations of sustainability as a hard constraint. The present is always represented as the first generation in the trajectory, and the constraint concerns the pattern of generational welfare from the present onwards. But you are right that for some of the more general notions of a sustained trajectory that we later consider – in particular, the one where future generations need only have at least as high welfare as the present generation – ignoring the past is awkward as it can lead to dynamic inconsistency. What was once a sustained trajectory may be one no longer due to the reference level of welfare being raised over time. So even though our impossibility result holds for a range of notions of ‘sustained’ trajectory, some of these notions are problematic for other reasons – of dynamic consistency – when the past is ignored.
Interestingly, ignoring the past might raise another difficulty to do with incentives, if one is concerned about the choices of the present as well as the choices of future generations, and the extent to which these choices conform to sustainability. Consider our initial notion of a sustained trajectory: non-declining well-being over time. On the one hand, it seems better to set aside the past. Suppose that there has been a decline in the past. If we include that decline when assessing whether trajectories are non-declining, every option in the menu is unsustained. So the sustainability constraint provides no additional guidance for that menu. But we still want to say something about the future. It therefore seems plausible to ignore the past decline and apply a sustainability constraint to the menu.
However, ignoring the past might create incentives for manipulation. A policymaker knows that a present decline will be ignored in the next period. Suppose that a country can no longer maintain its current welfare policies because of its debt burden. What should it do in this situation? A sharp decline in the present period, potentially sacrificing the present generation, followed by stable or improving welfare levels, could allow the trajectory from the next period onward to count as sustained. The policymaker might choose this over a trajectory where well-being levels gradually decline, without any sharp decline. This seems like a dilemma inherent in the notion of a hard sustainability constraint.
2. On whether it would be more realistic to allow ‘intergenerational trajectories’ to have differing length:
Yes, this is a very good question. We restricted our attention to the fixed-length case. Indeed, this is a very restrictive assumption. Cases involving different lengths are very relevant to policymakers. We can consider the implications of extending our framework to include different-length cases by distinguishing between impossibility results and possibility results. For our impossibility results, we can say that our results are robust. Even if we include trajectories of different lengths, trajectories of the same length are still available in the domain. So we can reproduce the same result. For possibility results, differences in length can be a new source of impossibility. So even if we have possibility results in the fixed-length case, impossibility results may hold if we impose some plausible axioms that connect trajectories of different lengths.
3. On whether the property of ‘being sustained’ should come in degrees:
This is an important point about the potential appeal of a graded notion of sustainability. For the most part in our paper, we treat ‘being sustained’ as a binary property of an intergenerational trajectory, and moreover one that is not dependent on the menu of alternative trajectories. That’s because we focus mainly on sustainability as a ‘hard constraint’ on permissible choice and this seems suitable. But later in the paper we consider that sustainability may be rather a ‘soft constraint’ or a pro tanto consideration for choice, to be weighed against other considerations. Then it is more natural to appeal to a graded notion of ‘being sustained’.
The graded measure that we introduce for our soft sustainability constraint is one where a trajectory is (un)sustained to the extent that there are drops in generational well-being between one generation and the next. More generally, a fundamental question one needs to ask is whether the ‘more sustained’ relation would be transitive and complete. For instance, in your example – where (1, 2, 3) is more sustained than (1, 0, 3), which is in turn more sustained than (3, 1, 0) – transitivity seems intuitive, but completeness is not obvious.
One could additionally investigate a hard sustainability constraint that involves a graded ‘more sustained’ relation. This would be an interesting line of inquiry that we don’t think has been undertaken. For instance, we could replace our hard sustainability constraint with a requirement that, whenever there are trajectories that are both efficient and most sustained, only such trajectories should be chosen. Presumably a version of our impossibility result would hold also for this kind of hard sustainability constraint, although that deserves further investigation…
4. On the prospects for the shape of an intergenerational trajectory having intrinsic moral significance:
It is true that we are skeptical about it being intrinsically better that humanity’s trajectory over time takes some specific shape (e.g., unbroken progress) rather than some other shape. The analogy with an individual person and their well-being might be compelling if one saw humanity as a collective person (a group agent). But if one thinks of a series of generations as just that – generations ordered in time – then the analogy with individual persons seems tenuous. (Our thinking is along the latter lines.)
That said, we did indeed remark in a footnote that there may be reasons why the shape of an intergenerational trajectory may indicate something of intrinsic moral importance, even if the shape itself does not have intrinsic moral importance. One idea is that it indicates a subtle aspect of well-being that is not already included in (what may be a narrow) measure of a generation’s well-being. This might be individuals’ sympathy for the plight of their descendants, or for future generations more generally. The idea is that each generation’s well-being depends on both the (narrow) measure of their wellbeing, and also on the extent to which future generations do at least as well as them. Another idea (which we find in a paper by Michal Masny) is that the shape of the intergenerational trajectory indicates that there are impersonal non-welfarist goods that endure over time, like beautiful artworks.
5. On whether we should consider sustainability a ‘mid-level’ moral principle as opposed to a ‘heuristic’:
Thank you for this suggestion. We can see that ‘mid-level principle’ may be a better term than ‘heuristic’ for a sustainability principle that is a (robust) consequence of (various candidate) moral assumptions coupled with other empirical or scope assumptions. Their names suggest that a ‘mid-level principle’ is better defined and justified than a mere ‘heuristic’, and so perhaps the former is better terminology. Moreover, mid-level moral principles are discussed in the literature (not least by Bykvist and Roussos in a recent paper!)
We have some questions of our own about mid-level principles, however, by way of understanding them better. For example, consider a distribution problem involving two people at a given time. Suppose that a fixed total amount of well-being can be freely distributed between them. Under leximin, the optimal outcome is equality in well-being. Under prioritarianism, equality in well-being is also optimal. Would the principle of achieving equality in this case count as a mid-level principle? Do we call a principle a mid-level principle if it is satisfied by all the plausible theories under consideration? If so, a mid-level principle seems to be something like an axiom. Here it is an axiom that concerns how a fixed amount of well-being should be distributed. The Pigou–Dalton principle is a more familiar axiom that is satisfied by both leximin and prioritarianism. Would this count as a mid-level principle?
Thanks for your excellent replies, Katie and Susumu. Here are some quick replies to your answers.
1. I did not have the space to explain the possible problem with dynamic inconsistency, but that is exactly what I had in mind. Thanks for doing that. You are also right that it would be odd to just give up on choosing a path that will be sustained in the future just because there was a decline in the past. But it is also odd to say that sustainability considerations are not applicable, if every feasible *future* paths contains a slight decline between two successive generations. A graded notion of being sustained can deal with both problems, since a decline only affects the degree of being sustained (unsustained).
2. Yes, my comment on different-length cases does not affect your impossibility result. It just shows that there are many more problems to consider, if the model is made more realistic.
3. I agree that ‘more sustained’ may not be a complete notion (even though it seems to be transitive). That calls for some other principle to fill the gaps in a way that avoids intransitivity (not an easy task).
4. My idea was not that humanity must be seen as a collective agent in any demanding sense (e.g., have intentions and the ability to deliberate). You must also think the humanity *at a time* is an entity that can act and whose act can be assessed as right or wrong. That was the notion I had in mind, and I think that it is sufficient to make the analogy with individual lives prima facie plausible. More generally, we seem to prefer being part of a development that is improving rather than one that is deteriorating, other things being equal. But a development need not be the outcome of a collective agent in any demanding sense.
5. It is true that the notion of a mid-level principle is hard to pin down. I think the notion is not used for just one particular abstract case, however. It is rather that it is a principle that is backed up by different foundational principles for a large set of relevant cases. That is vague, of course. But if we think about policy-relevant mid-level principles, they are meant for contexts that are severely constrained. First of all, there are practical constraints; many decision-makers work in institutions that are quite restricted in what they can or cannot do given their resources, which to a large extent is set by budgetary constraints. Second, the decision-makers do not assess the consequences of their actions directly in terms of what matters fundamentally. Instead, they use indicators or proxies for what matters fundamentally. Third, decision-makers are bound by laws which constrain their set of permissible actions. These constraints create the kind of restricted domain in which mid-level principles can be fruitfully applied.
Thanks again for giving me the opportunity to comment on your great paper!
Hello again Krister. Some brief further replies:
On 1: I see your point better now that a graded notion of sustained trajectories may be important even in the context of a hard sustainability constraint. It would mean that the constraint is applicable to and hence can provide advice on more decision problems.
On 4: Fair enough, you were not suggesting that humanity as a whole might be a collective agent. Just that progress over time might make a social outcome better, if indeed progress over time makes an individual life better. I suppose the idea is that, in the case of an individual life, progress over time may be good for that temporally-extended person, perhaps in part because progress is preferred by the person at various different times in their life. One might be able to say something similar for an intergenerational trajectory: that progress across generations may be good, perhaps in part because it is preferred by various generations. More generally, perhaps the line we were trying to draw in our response above between the shape of an intergenerational trajectory having intrinsic moral significance and the shape merely being an indicator of something that has moral significance is not so sharp.