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.
On 5: I see that you want to distance ‘mid-level’ principles from axioms. Let me say though that axioms may be highly conditional in their nature: e.g., ‘In circumstances X, such and such should hold’. So they can (and often do) express requirements that are only applicable in a narrow domain. But axioms do not tend to focus on mere proxies for what really matters morally, unlike the way you are describing mid-level principles.
It may be very fruitful to further pursue sustainability as a ‘mid-level’ moral principle, in your sense, where proxies are involved. (In our paper, by contrast, we stick to definitions of ‘sustained trajectory’ where the measure for each generation’s level of advantage is supposed to capture all that matters morally regarding that generation. We call it ‘generational well-being’ for short.) The appeal to proxies may capture more of the common usage of sustainability principles. Many people associate sustainability with maintaining healthy ecosystems, for instance, or maintaining very specific social institutions. Under the right conditions, adhering to such a principle is intuitively an effective way to promote the good in a more comprehensive sense. That would seem to fit your paradigm of a ‘mid-level’ principle.
Thanks Katie for the clarifications. Just two comments on 5:
I associate axioms with ground principles (but not necessarily universal ones) for which one does not have further justification. Mid-level principles can be axioms in this sense for some fundamental principles. For others, it can instead be something that is backed up by a distinct fundamental principle in combination with empirical assumptions. But this might just be a terminological issue.
Yes, mid-level principles with proxies are not captured in your model. But, as you say, they seem to be much closer to the sustainability principles that are are used in political debates. In fact, they might be the only politically feasible sustainability principles, the ones we can agree on, despite all the religious, political, moral, and cultural differences. Brundtland’s sustainability constraint is perhaps a case in point. Perhaps no other sustainability constraint would do better than Brundtland’s as a politically feasible proxy constraint. We know at least that it has gained and retained formal recognition. For example, it is still mentioned in Agenda 2030. But it is an open question whether it is the best proxy constraint.
Thank you again Krister and one last thing on 5 (perhaps!):
…Just about the Brundtland statement, which says: ‘Sustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs’. This statement has widespread popularity and it is much cited, to be sure. But the Brundtland statement itself doesn’t strike me as a paradigm ‘mid-level’ principle that appeals to proxies for moral value, or some limited range of goods. Indeed, we suggest in the opening of our paper that the Brundtland statement at least *appears* to be a general principle, albeit an ambiguous one, that refers directly to what has ultimate moral importance (depending on how one interprets the ‘needs of generations’). We treat the ‘hard sustainability constraint’ that we investigate in the paper as a potential interpretation of the Brundtland statement. (Above, I was thinking rather that spin-offs of the Brundtland statement – principles about maintaining waters supplies, or ecosystems or even particular industries – might be more like your idea of ‘mid-level’ sustainability principles.)
I think the needs referred to in the Brundtland principle can stand in for or be a proxy for a lot of thing, since what you need is always something you need *for something*, and that something can be different things in different contexts. I think you are right that there can be more levels than the fundamental and the mid-level. The water supply principles comes higher up than mid-level principles. Different mid-level principles can support the same water supply principle. You have a whole hierarchy of principles starting from the foundation. Lots of work in climate justice seem to assume this complex structure.
Hi Katie and Susumu!
Thanks for writing this excellent paper, and to Krister for the very helpful summary and comments.
I wonder how committed you are to Contraction Consistency and whether you think it might be worth relaxing in this context. After all, we can also get violations of Contraction Consistency from other moral principles that many find attractive. You mention egalitarian intuitions in the paper, but other violations arguably arise from the Mere Addition Paradox (it may be OK to choose Parfit’s outcome A from {A, A+, B} even though it wouldn’t be OK to choose A when the only alternative is A+), partial aggregation (it may be OK to save some number of people from moderate harm, when one could instead either save a much smaller number from a severe harm or a much larger number from a slight harm, even if we’d be required to save the larger number from the slight harm if that were the only alternative), and secondary permissibility in the ethics of killing (it may be OK to save five people by hurling a sixth in front of a lethal trolley if the alternatives are to either let the five die or to divert the trolley onto the one in a way that kills her in a much more painful way, even though it wouldn’t be OK to hurl the one if the only alternative were to let the five die)—perhaps there are other examples too.
Hard-Constraint Sustainability doesn’t have the same kind of intuitive pull (to me) as these other judgments. But I wonder, if one is already comfortable with violations of Contraction Consistency in those other contexts, how much of a problem for Hard-Constraint Sustainability one should take your result to be.
Thanks Jake. In the paper we do speak of Contraction Consistency as a rather weak consistency condition, even though it has been challenged in cases where options may have menu-dependent properties, in addition to intrinsic ones, that are relevant to their choice-worthiness. You mention a nice variety of potential cases of this sort. They all arguably involve a menu-dependent property that has to do with whether persons are treated disrespectfully.
For the Parfit case, one might say that the availability of B means that choosing A is no longer disrespectful to the added persons in A+, since A compares favourably with B which compares favourably with A+. In the partial aggregation case, one might say that the availability of the possibility of curing a small number with severe ailments means that choosing some greater number with moderate ailments is no longer disrespectful to the many with minor ailments, since the moderate cure compares favourably with the severe cure which compares favourably with the minor cure. And for the tragic rescue case, one might say that the presence of the side-track with severe pain means it is no longer disrespectful to the one to hurl them into the train to save five, since the hurling is better than the side track for the one which is in turn better or justified when compared to letting the five die.
Now these verdicts are all contested. But the question you ask is: if one took the view that violations of contraction consistency were ok in these cases for roughly the reasons given to do with menu-dependent properties, could not the same thing be said in the case of our hard sustainability constraint? In the paper we suggested that the menu-dependent property in the sustainability case would be rather unnatural. But let’s consider it further.
Can one appeal to a similar menu-dependent property of ‘undue treatment of generation(s)’? Now already that does not look so compelling as a property to do with ‘disrespectful treatment of persons’. The idea might be that when, for instance, the three trajectories (0,15,30), (5,25,25) and (30,15,30) are available, it would involve undue treatment of generations to choose (0, 15, 30) since (30, 15, 30) Pareto dominates that trajectory. But when the choice is just between the first two trajectories, one might well choose (0,15, 30) since that reason of undue treatment does not apply. This is what we had in mind when we were querying Contraction Consistency in the paper. We said that this was not very convincing, however, since the menu-dependent property being invoked is more or less ‘(not) Pareto dominated by another trajectory’. But can one appeal to a more subtle construal of the menu-dependent property – one that follows the pattern of the cases above? Perhaps one could say something like: when there are two sustained trajectories, and they are each Pareto dominated by separate trajectories, then it is not ‘undue’ to choose either of the sustained trajectories since the reasons for choosing them are balanced. But when the menu contracts such that only one of the sustained trajectories is Pareto dominated, it is ‘undue’ to choose that one. This again seems rather tenuous however so far as a menu-dependent property goes… So for the time being, we stand with Contraction Consistency in the sustainability context.
Hi Jake,
Thanks for raising this important point. Personally, I am fairly committed to Contraction Consistency, at least in the present context. While I cannot say that I accept it in every context, I believe that our results offer some reason to question Hard Constraint Sustainability even for those who accept violations of Contraction Consistency elsewhere.
I want to mention a formal observation. The main impossibility result no longer holds if Contraction Consistency is weakened to base acyclicity (acyclicity at the level of binary choice): base acyclicity and Hard Constraint Sustainability are compatible. So the result is tight in this respect. However, adding Pareto Efficiency restores the impossibility. Both results are established in the Appendix. If we use a further/different weakening of Contraction Consistency, we may get a possibility result. But my own inclination is that, in the present setting, Contraction Consistency itself is still worth retaining.
That said, I need to be cautious here. A more difficult case involves comparing trajectories of different lengths; this possibility was suggested by Krister. We might have to compare a longer but less sustained trajectory with a shorter but more sustained one – here I am following Krister’s suggestion of degrees of sustainability. Such comparisons bring in the intuitions behind the Mere Addition Paradox, and the appeal of Contraction Consistency becomes less clear. This is one reason we focused on trajectories of the same length.
As your three examples (and the different-length case) show, the issue is closely connected to menu dependence, as suggested by Katie. In particular, when an option is permissible only in the presence of another, removing the latter can lead to a violation of Contraction Consistency. I found your third example especially interesting and compelling. I think it would also be interesting to examine these cases in terms of base acyclicity. More generally, whether these cases should satisfy some weaker form of choice consistency strikes me as an important question in its own right. My reaction may also differ across the examples: in the partial aggregation case, for instance, I am more inclined to retain Contraction Consistency.
When choosing among intergenerational trajectories (of the same length), I am inclined to retain Contraction Consistency. Even those who reject it in other contexts might accept it in this setting, in which case our result gives us reason to question Hard Constraint Sustainability.
Thank you, Krister, for your great points. I found it encouraging that you’re concerned about dynamic inconsistency regarding the “ignoring the past” problem. When dealing with problems involving time, I feel that dynamic inconsistency is very important in addition to choice consistency such as Contraction Consistency. I haven’t had enough time to do so yet, but I’d like to think more about the implications of dynamic (in)consistency.
The discussions related to mid-level principles are inspiring for me. Related to this issue, I want to mention two problems with regard to (non-declining) sustainability as a mid-level principle.
The first problem is about (technological) sensitivity and uncertainty. For example, if there are exhaustible resources, the utilitarian principle tends to accept (or may accept) trajectories in which well-being gradually declines across generations. Under other economic conditions/constraints, this is not true. It may suggest substantial sacrifices by the present generation if there are sufficiently large gains for future generations. (Other welfare criteria would have similar problems.) Whether welfare theories permit non-declining trajectories seems very sensitive to the details of the economic constraints.
One important thing regarding intergenerational problems is that it is not easy to know the practical/economic constraints. In the case of budgetary constraints for current policymaking, the decision-maker understands their situation, at least roughly. But when it comes to intergenerational problems, it would be difficult to know the economic constraints (in the far future). So, it is not easy to judge when sustainability should be accepted as a mid-level principle.
The second is related to the proxy issue. I want to mention what is called the Hartwick rule. Roughly speaking, this rule says that all resource rents must be invested in reproducible capital in a dynamic economy with exhaustible resources. If this rule holds, we have constant consumption across generations under certain plausible assumptions, or perhaps we can say that well-being levels are constant across generations. This constant trajectory partly captures the idea of sustainability (it requires complete equality in well-being across generations). The point is that the constant-well-being requirement restricts trajectories directly, while the Hartwick rule provides a practical savings and investment rule, given some economic constraints. It seems that the Hartwick rule is expressed in terms of proxies or indicators. In this respect, the Hartwick rule captures the role of a mid-level principle. So my claim is that (i) the constant-well-being requirement could be regarded as a more general moral principle and (ii) the Hartwick rule is a mid-level principle derived from the constant-well-being requirement. So the non-declining requirement would be regarded as a fundamental/general moral principle, and there could be a mid-level rule derived from it that is somewhat similar to the Hartwick rule.
I agree that it may not be easy to know the economic constraints and thus not easy to know when sustainability should be accepted as a mid-level principle. But what that shows, I think, is that not all mid-level principles are policy-relevant. To be policy-relevant they need to be easily applicable by the relevant agents. Mid-level principlies do live a life outside policy-making, however, as one can see from the example of the climate justice debate where some of the discussed justice principles rely on uncertain empirical evidence. Mid-level principles can still be useful since they can provide evidence of convergence between different fundamental principles. This convergence can then have some epistemic relevance. Other things being equal, we have stronger reason to believe a verdict when it is backed up by a diverse set of fundamental principles than when it is only backed up by a few, very similar ones, holding the empirical assumptions fixed.
Hi Katie and Susumu,
Thank you for this interesting paper, and, Krister and Jake, thank you for this excellent summary and discussion.
I was hoping to get some clarity about the structure of your argument. Do you reject the desirability of sustainability because it is inconsistent with other desirable properties, or do you reject it on its own grounds? In other words, should I feel sad that I cannot accept the desirability of sustainability without rejecting these other principles? Should I feel conflicted when I am in a situation in which I am forced to choose between them? Or, do I still have a reason to respect sustainability principles when I am in a situation in which the domain of decision problems is sufficiently restricted so that there are no conflicts with other properties that I consider desirable?
To me, the more interesting argument is the stronger one: that sustainability is not a desirable property to begin with, independently from how nicely it plays with other moral principles. This cannot be established by pointing out conflicts, because by now we know that conflicts frequently arise between different desirable properties.
Thank you, Krister, for your response! I think I am beginning to understand your idea of mid-level principles.
Let’s say we have a betterness relation that is transitive (and maybe complete). This is a general moral principle. What we normally call axioms are properties imposed on this relation. Given a menu, we can think about choices — optimal options regarding the relation. With transitivity and some standard regularity assumptions, choices from menus would be well-defined. Each menu reflects empirical assumptions about what is actually available. (The relation is usually defined over all possible options. ) Mid-level principles sound like properties of choices from a given menu, which are shared by the choice functions induced by different relations (or general moral principles).
Indeed, utilitarianism, prioritarianism, leximin, maximin, and others (understood as betterness relations) yield non-declining trajectories when a menu satisfies a certain empirical/economic condition (time productivity). This is our Theorem 8. This non-declining property can be regarded as a mid-level principle.
Here is a potential reason why I thought of your mid-level principles as axioms. In our paper, a choice function (how we choose) is taken as a primitive. Our “axioms” are imposed directly on choices. Our “axioms” are not imposed on relations.
Your explanation of mid-level principles works nicely when we take a betterness relation as a primitive. In a relation framework, axioms are imposed on the relation, not on choices. Properties shared by the choices induced by different general moral principles could then be extensionally equivalent to what we call “axioms” on choices. Anyway, we may be starting from different primitives, which could explain why we use some terms (including axioms) differently.
Hi Maya,
Thank you for your important question.
The main message I take from our impossibility result is that there is something wrong with the non-declining view of sustainability that is commonly used.
I see your point. There are many impossibility results. For example, in the case of the Mere Addition Paradox, we use transitivity, completeness, the mere addition property, avoidance of the repugnant conclusion, and some other properties (e.g., non-anti-egalitarianism). Some argue that we should drop transitivity; others claim the mere addition property is wrong. There are scholars who say the repugnant conclusion is not repugnant. Which one should be rejected?
As you suggested, rejecting a concept on its own grounds is very powerful. The statement of our main theorem mentions four conditions but it is mainly about a conflict between Hard-Constraint Sustainability and Contraction Consistency. I think that Contraction Consistency is plausible in this context. I regard Hard-Constraint Sustainability as the substantive moral principle at issue here (and Contraction Consistency as a basic requirement of coherency). So I think it comes fairly close to the kind of internal criticism you have in mind.
I don’t think that the result alone demonstrates that sustainability has no value in itself, but it still shows that the non-declining view of sustainability has a serious problem.
Thank you Susumu for the clarifications! I am not sure I see relevance of whether the choice function is primitive or not. Suppose we start with a primitive choice function and define utilitarianism, prioritarianism, leximin, and maximin in terms of this function. Then your Theorem 8 shows that they all select the non-declining trajectories, given certain assumptions about productivity. Would you then not agree that the sustainability principle they converge on is a mid-level principle? Not one that is defined in terms of proxies, and thus not one that is easily used in policy-making. I realize now that throwing in the talk about proxies and policy-relevance in the discussion muddied the waters and made it difficult to see that mid-level principles need not be defined in terms of proxies or be policy-relevant. They can still be used for epistemic purposes, as I suggested earlier.
Thank you, Krister! Yes, I agree that the sustainability principle in Theorem 8 can be understood as a mid-level principle in your sense. (I emphasized the difference between primitives to explain our use of the term “axiom.”) Your clarification about proxies and policy relevance makes the idea of mid-level principles much clearer to me.
I guess many economists are, in this sense, working on deriving mid-level principles. For example, progressive taxation can be supported by utilitarianism, prioritarianism, maximin, and other theories under suitable economic assumptions. This could also be useful for epistemic purposes.
Thanks, Susumu! Sorry I was a bit unclear when I introduced mid-level principles.
I think this discussion of mid-level principles will help me think about a range of problems. Thank you, Krister!
@Maya, let me also thank you for your comment/query. Perhaps not surprisingly, I view the impossibility result in our paper in much the same way as Susumu. As you say, when one substantive moral principle conflicts with another substantive moral principle, it may not be clear what lesson to draw. But in our paper, we show that the hard sustainability constraint based on non-declining well-being conflicts with what we take to be a weak consistency constraint. (You may have a different interpretation of Contraction Consistency). That conflict reflects badly on the hard sustainability constraint itself, we say. It shows that it is not tenable as a fully general (and hence fundamental?) moral principle. The hard sustainability constraint may be reasonable under various restricted domains, but we suggest later in the paper that any restriction in scope in a sense downgrades the status of the constraint—it is better seen as a ‘mid-level’ moral principle (to use Krister’s term) than a fully general (and hence fundamental?) moral principle. (We also explore the prospects of a soft rather than a hard sustainability constraint based on non-declining well-being. We show that can define a soft constraint of this sort that is generally consistent.)