1. Introduction
Compulsory saving for retirement is a pillar of fiscal policy in developed economies. There is considerable variation as to how such policy is designed. Key differences include whether funds are obtained via taxation or employer contributions, whether birth cohorts fund themselves or the preceding birth cohort, and whether any redistribution within cohorts is pursued. Much current policy was designed in the mid 20th century. In the decades since, changes in demographic and economic conditions have introduced challenges for the continued funding of retirement. Demographic ageing has stretched the ratio between retired and working-age people. Growing inequalities in longevity have brought about increasingly unequal lengths of retirement. This has often meant discrepancies between what different people actually receive in retirement payments, even after paying in similar amounts over working lives roughly equal in length. Retirement funding is also threatened when prolonged inflation erodes the savings of retirees. At the same time, stagnating incomes and higher household debt of younger people limit their savings potential. Different models of funding retirement make different trade-offs in responding to this bundle of challenges.
Most recent philosophical discussion of retirement funding focuses on questions of justice. Such questions have been made more pressing precisely by the changes in demographic and economic conditions just mentioned.Footnote 1 Here there is disagreement as to what model of retirement funding is most compatible with justice. What’s less controversial, however, is that there exists a separate and (perhaps) more fundamental rationale for making retirement saving compulsory in the first place: By and large, there seems to be implicit agreement among philosophers that some sort of paternalistic rationale for compulsory retirement is cogent, even if considerations of justice might weigh against its implications. The idea here is that, left alone, people will not voluntarily save enough for their retirements. Consequently, many people will end up materially worse off than they could have been during their later years, with some severely impoverished. So, the state might have authority to coerce its citizens, during their years spent in the labour market, to contribute to some kind of savings scheme. This paper seeks to examine this paternalistic rationale, which has gone largely unexamined.
An important preliminary point is to establish what is meant by ‘paternalism’ for the purposes of this paper. Paradigmatically, paternalism involves some third party interfering with an individual with the aim of preventing, or at least discouraging, that individual from acting on some preference or disposition. Familiar examples of paternalism in law and policy include rules requiring cyclists to wear helmets, taxes on ingestible products known to increase bad health outcomes, ‘cooling off periods’ when consenting to major surgeries, and so on. Paternalism is often associated with measures that seek to offset individuals’ cognitive biases, or associated tendencies that lead them to act contrary to their interests. Among these is a tendency to discount future wellbeing. A failure to adequately save for retirement might be traceable to this. At the same time, failure to save optimally for retirement may happen because an individual’s preferences simply aren’t well defined or complete. In such circumstances, individuals may make choices that resemble the explicit discounting of future wellbeing even if this is not strictly what is happening. In some cases, a paternalistic intervention may be a response to an individual facing a choice that it is unreasonable to expect them to resolve on their own. As some recent work on paternalism has emphasised (Quong Reference Quong2025), paternalism of this sort need not involve any negative appraisal of the recipient’s beliefs or preferences, nor need it attribute cognitive shortcomings to the individual.Footnote 2 Rather it involves recognition that sufficiently strong preferences aren’t present and perhaps can’t readily be formed, and that individuals might therefore benefit from policy interventions that make up for this. The case of mandatory retirement might count as a case in point, since individuals might not – particularly when relatively young – have well-formed preferences about retirement.
So, this paper relies on a fairly capacious understanding of paternalism whereby an individual’s life is regulated in ways that aim at making this individual better off, over the course of their life, than if they were not so regulated. This is a fairly minimal definition that is neutral across a variety of more refined views of what paternalism is. But the intention is to capture conceptions of paternalism that emphasize some combination of cognitive bias, or difficulty in resolving a difficult problem of practical reasoning, in the event of an individual being left alone.
So understood, a paternalistic rationale for retirement funding would seem to have force. At least, it has features often thought to make paternalism easier to justify: The empirical evidence that people discount their future wellbeing, and are disposed against saving voluntarily, is well-established. Intuitively, there may be something especially concerning about poverty in old age: There might be distinctive burdens associated with being old, or perhaps distinctive goods that a funded retirement is necessary to secure (Overall Reference Overall and Wareham2022; Kazez Reference Kazez2023; Halliday Reference Halliday2024). Being old can mean being less able to escape poverty through the labour market. This is because finding paid work is often harder in old age than earlier in life. Compulsory retirement saving also seems to avoid some of the problems that make other paternalistic policy hard to justify: Retirement funding doesn’t appear to single out any individual group as especially in need of paternalism, and it doesn’t seem to discriminate against, or unduly promote, any particular conception of the good life. For these reasons, compulsory retirement saving has been contrasted with objectionable paternalism such as compulsory weight-loss programmes, and likened more to widely accepted paternalistic policy such as laws requiring the use of seat belts in vehicles (Hojlund Reference Hojlund2021).
An implicit assumption within political philosophy and relevant work in the social sciences is that some paternalistic rationale for compulsory retirement funding is forceful. Beyond this, there is little discussion that seeks to vindicate or even illuminate this rationale. In particular, little has been said about what policy should aim to do beyond making up for individuals’ tendency to not take care of their own retirement funding. Correcting for individuals’ bias against saving is a kind of negative paternalistic aim of compulsory retirement funding. Granting this aim on its own leaves doubt about what exactly the positive aim of retirement funding is, as in how far coerced saving should go and what sort of form it should take. With some exceptions (to be discussed below) I have found no really extended discussion of paternalism and retirement funding beyond the typical observation that people tend to be biased against saving and that the tendency for developed economies to run some kind of compulsory scheme offers a good means of overcoming this. But this is just to restate the negative aim. Typically, more attention is paid to how to make retirement funding compatible with requirements of distributive justice, as if the paternalistic rationale is a separate – and perhaps settled – issue. This is not a way of identifying the positive (paternalistic) aim, but instead an attempt to illuminate how retirement funding policy might be shaped by another sort of requirement.
A traditional misgiving about paternalism concerns whether the intervening party really knows best when it comes to the interests of the affected individual. A narrower concern is whether it is clear precisely what the intervening party is trying to do. A version of this concern is what’s at issue in what follows. This paper is an attempt to show that the relation between compulsory retirement funding and paternalism is worthy of more attention than it usually gets. Rather than mount any sustained criticisms of specific proposals in the existing literature, it aims to show that more work is needed on paternalism and retirement, and to make a start on such work here. I shall argue that some ways of funding retirement, including some real retirement policy in some jurisdictions, constitute failures of paternalism rather than implementations of it. This becomes apparent when we distinguish more sharply between the idea of insurance against bad outcomes and the smoothing of wellbeing or resources over an extended timespan. Each of these concepts is routinely associated with retirement funding. But the differences between them have been overlooked. Moreover, distinguishing more sharply between these concepts reveals reasons to doubt the independence of paternalism-based and justice-based perspectives on retirement funding. More specifically, some of the ways in which paternalism can fail can also be viewed as introducing or compounding an injustice. In this way, thinking about paternalism both illuminates the paternalistic rationale itself, while also having something to offer more mainstream attempts to figure out a theory of retirement justice.
The key points in this paper, then, are as follows: First, I argue that we need to distinguish the idea of (counterfactual) insurance from (temporal) smoothing more sharply than has typically been done. Doing this shows that, accordingly, there exist somewhat competing conceptions of what mandatory retirement funding might be trying to do, even if there is some tendency for them to converge. Importantly, however, the idea of temporal smoothing is more coherent than insurance in the context of mandatory retirement funding. But once this is conceded, certain puzzles become more troubling than if it had been coherent to understand retirement funding as insurance. The upshot is that the paternalistic case for retirement funding has more limited force than it is often presumed to be.
As a guide, here is how the rest of this paper is organized. First, (section 2) I review an argument that is somewhat sceptical about the paternalistic rationale for compulsory retirement funding. This argument involves the presupposition that retirement funding should be understood as a certain kind of insurance. In response, I suggest that the sceptical rationale – though forceful – suggests that the concept of insurance is being conflated with that of smoothing. Section 3 then works out the distinction between counterfactual insurance and temporal smoothing more fully. Section 4 then concentrates on identifying the problems that emerge for a paternalistic rationale once mandatory retirement is understood as an attempt at smoothing wellbeing temporally. Because smoothing is more ambitious than insurance, in terms of the degree to which it seeks to move resources around a lifespan, it is more prone to involve a failure of paternalism. In particular, an attempt to smooth resources risks inverting the bias that paternalism is meant to mitigate. Individuals’ tendency to discount their future wellbeing is replaced by a different bias that imposes opportunity costs early in life in return for an excess of resources later on. This risk becomes larger the greater the quantity of resources, or average level of wellbeing, in an individual’s life. This makes smoothing more prone to inverted bias than mere insurance. A second problem focuses on the arbitrariness of targeting income from labour, while ignoring other sources of income or wealth, when it comes to mandating saving for retirement. Probably all retirement funding schemes exhibit this second problem. In section 5, I identify some reasons to see a stronger connection between paternalism and justice than has usually been assumed. In particular, paternalism that aims to smooth resources runs the risk of having anti-egalitarian expressive content, particularly about the importance of ensuring that upper middle-class status should be retained by those who have it until the end of their life. In addition, the presence of a savings bias compounds – albeit as a contingent matter – injustices relating to inequalities in inherited wealth. Section 6 concludes: I suggest that if retirement funding is to rely on a paternalistic rationale, then it needs to be more modest. Alternatively, the limits of the paternalistic case mean that ambitious retirement funding needs to be more responsive to justice considerations in order to correct for likely failures of paternalism.
2. Doubts About Paternalistic Rationales for Insurance
Mandatory saving schemes seem at least to conform to a paternalistic rationale. This is to say that an effect of such schemes is that individuals gain protection against being made worse off by their disposition to not save had they been given freedom to act on this disposition. The same is true of a host of other schemes that work similarly to protect individuals against bad consequences through some kind of mandatory contribution. Requirements to purchase health insurance, or at least to pay taxes to cover healthcare costs, are among the most familiar of these. These appear genuinely paternalistic, unlike (say) car insurance and professional liability insurance whose compulsory status draws on the need for regulation to protect third parties.
But conformity to a paternalistic rationale may conceal a justification that is not strictly paternalistic. Precisely this claim has been argued for with respect to at least some compulsory insurance requirements. According to Paul Bou-Habib, the justification for requiring people to take out insurance against the consequences of imprudent choices is to protect others from costs of rescue.Footnote 3 The key idea here is that to prevent an individual from undertaking risky activities, “we may force him to insure in order to prevent him from imposing on us the costs of performing our duty to him in the future” (Bou-Habib Reference Bou-Habib2006: 262). An obligation to rescue someone can persist, and be just as inescapable, when the individual requiring rescue is wholly to blame for their situation. Versions of this claim have been defended in a variety of contexts, sometimes under the description that justice prohibits “the abandonment of the imprudent” (Anderson Reference Anderson1999; Voigt Reference Voigt2007) when that would involve allowing anyone to languish below some basic social minimum.Footnote 4 Bou-Habib’s point is that although the rest of us have a duty to rescue the imprudent who find themselves in desperate need, imprudent people should not thereby gain an ability to “exploit” wider society by relying on the fact that others’ duties of rescue are morally inescapable. Mandatory insurance schemes solve this problem by ensuring that when duties of rescue need to be performed, the costs of undertaking them are absorbed, at least in large part, by the party being rescued. Since the justification for such schemes is ultimately about protecting innocent third parties from unduly having to absorb a cost, the justification does not rely on paternalism even if its effects are somewhat as if it did. And it brings the case for mandatory insurance against imprudent acts that harm oneself closer to types of insurance more obviously about protecting other people, such as driving and professional liability insurance.
Bou-Habib’s position is correct as far as it goes.Footnote 5 That is to say, it vindicates some important uses of mandatory insurance in ways that avoid presupposing an endorsement of paternalism. Notably, however, Bou-Habib’s view has quite limited scope: It applies only to imprudent choices whose outcomes can be severe enough to push an individual’s wellbeing below a critical threshold. This is because it is only in these severe cases that a duty of rescue, as opposed to mere assistance, kicks in for other individuals.Footnote 6 Importantly, however, duties of rescue do not apply in all scenarios where a person has been left worse off. They do not obviously apply when an individual has simply been left only somewhat badly off in an absolute sense albeit perhaps considerably worse off relative to some counterfactual scenario in which that individual made better choices.
Importantly, mandatory retirement saving almost always aims to do more than keep individuals just above a threshold below which duties of rescue get generated. As it happens, the same can be said of other mandatory insurance schemes, such as compulsory health insurance. These often fund treatments for health outcomes that are not so severe as to trigger a duty of assistance. Arthritis treatment, cataract removals, and some prescribed medications are among cases where healthcare makes a significant difference to wellbeing but does not serve to rescue an individual from an especially perilous situation.Footnote 7 This reflects the fact that a threshold below which duties of rescue get generated is not very high in terms of absolute wellbeing (as opposed to danger of death). Typically, entitlements upon reaching retirement are considerably more generous than these healthcare entitlements, in terms of the level of absolute wellbeing that they aim to secure.
A rough way to gauge how generous retirement incomes are is to compare median retirement incomes with a measure of the local cost of living. Eighteen European countries reportedly have regular pension incomes that exceed the cost of living, some exceeding it by four or five times.Footnote 8 Admittedly these claims need to be qualified. The real value of a retirement income is offset partly by what other services retirees can access for free or at a discount, such as healthcare and public transport. Measures of the cost of living are in any case contentious. When used with reference to retirees, they often exclude housing costs: An assumption frequently made when measuring retirement incomes is that retirees have paid off a mortgage. This is of course inaccurate in lots of cases, though being retired might mean an opportunity to access cheaper housing when pressures to live near schools and jobs are relieved. Additionally, countries that have a self-funded retirement scheme, rather than a pension paying out a universal fixed amount, will be subject to large inequalities in retirement incomes, whatever the average.Footnote 9 I shall revisit some of these complexities later on. The point for now is just to stress that retirement funding often aims at, and often achieves, something beyond protection against poverty.
Real policies aside, it’s not obvious why any theoretical rationale for compulsory saving should focus narrowly on protection against outcomes severe enough to count as emergencies, or severe enough to place people below any particular absolute threshold. This is in fact evident in some formulations. Ronald Dworkin, for example, suggests that “mandatory minimum insurance … is openly paternalistic: a decent society strives to protect people against major mistakes they are very likely to regret” (Dworkin Reference Dworkin2002: 114, emphasis mine).Footnote 10 Dworkin then gives examples of seatbelts in cars and emergency medical care. Admittedly, these cases are associated with outcomes in which an individual may risk falling below a decent social minimum. But a “major mistake” can be measured counterfactually rather than in terms of where it leaves an individual absolutely: An individual whose wellbeing remains comfortably above an absolute threshold might still have made themselves worse off than otherwise, and that could have been due to a major bout of imprudence.
Consider one more remark from Bou-Habib’s discussion: “In asking how a just society must address the needs of imprudent people, then, I am asking for an account of how such a society addresses needs that arise … because of imprudent choices” (Bou-Habib Reference Bou-Habib2006: 246). The reference to “needs” suggests a connection between imprudence and being left badly off in an absolute sense, that is, where a person’s needs go unmet. It may be that the concept of imprudence is not quite the right one for thinking about how individuals might fail to save optimally if left alone.Footnote 11 But we also think of imprudence as involving risks where there’s a large relative difference between where one starts off and where one might end up as a result of the choice(s) in question. And an intervention to guard against imprudence measured in this way would, at least intuitively, still be paternalistic. This suggests that the default or intuitive understanding of a paternalistic rationale for mandatory saving may be not just about preventing against emergencies but about maintaining a certain level of evenness of wellbeing across a lifespan.
It could be that mandatory insurance is justified only when it protects third parties from the costs of rescue. But it could still be true that individuals are required to contribute to mandatory schemes that have the effect of protecting their future wellbeing, and that the reasons for this requirement include paternalistic considerations. It’s just that such schemes might not be properly described as insurance, but as something else. In short, arguments that attempt to establish something general about the relationship between compulsory retirement funding and paternalism appear to suppress important distinctions internal to the idea of paternalistic intervention itself. This calls for further attention.
3. Pooling, Insurance and Smoothing Across a Lifespan
To say that compulsory retirement funding seeks to protect or promote an individual’s wellbeing during later life leaves things rather open about precisely what the goals of any funding scheme should be. Any policy might have a clear and defensible negative aim – in this case that of counteracting a bias in individuals against protecting their future wellbeing. But as I have said, it doesn’t follow from this precisely what a policy’s positive aim can or should be. The task now is to get clearer on the range of positive aims that compulsory retirement funding might have.
It is common to see compulsory retirement funding described – and endorsed – as social insurance. What retirement insures against is the bad of poverty during old age, and it is social in the sense that everyone is required to make contributions earlier in life. It is also common to see retirement funding described as risk-pooling. These are distinct concepts. Distinct again is the concept of consumption smoothing over the life course. These three concepts are often not distinguished sharply enough.
Strictly speaking, pooling occurs when a group of individuals cooperate by contributing their own shares of some resource, where the contributions are put together and then reallocated to some subset of the contributors once some uncertainty has been resolved (Heath Reference Heath2006). This presupposes nothing about what kind of rule governs this reallocation, including whether priority is given to whoever is left worst off after the relevant uncertainty resolves. Pooling can be inefficient and even regressive. Commercial gambling, for example, when individuals bet differently on the outcome of (say) a sporting event, counts as pooling in the above sense, though it is not usually what people have in mind by pooling risks or indeed any idea of cooperative activity. Pooling can also be inefficient insofar it increases everyone’s expectation of wellbeing but with some risk that everyone will be worse off. This is true of some pooling in financial markets, for example mutual funds for investments. There is nonetheless a unity to these cases, in that an exposure to risk is altered by way of combining and then reallocating resources according to how that exposure imposes different results on different contributors. This is why they all count as pooling.
Insurance is strictly speaking a subset of pooling that aims to have a relatively progressive reallocation rule. More intuitively, insurance is pooling where the aim is to let individuals draw on each other’s resources according to who has ended up most badly off in some relevant sense and where a payout from the pooled resources can bring some relief. Importantly, the measure of who is badly off is on some pre-specified dimension that is made explicit, such as a car accident in the case of motor insurance. One can’t draw on an insurance pool just because one has ended up worse off with respect to some other (irrelevant) dimension. Insurance intuitively presupposes pooling: Bad outcomes are made less bad only because contributors can draw on each other’s resources in the relevant way. This explains why individuals typically don’t try to insure on their ownFootnote 12 or in very small groups. Insurance works best in larger groups because a greater number of contributors expand the resource pool, often enabling greater levels of assistance for those for whom the relevant adverse event obtains.
Insurance seeks to make sure that variations in wellbeing across a lifetime are not subject to very sudden drops, or at least that something is done to facilitate a quicker recovery of wellbeing after such a drop. In this way, insurance will tend to smooth wellbeing over time for those who participate in it. But it should not be confused with the stricter idea of smoothing wellbeing over the life course, or ‘temporal smoothing’. Insurance, as I have said, typically involves pooling: Individuals lack sufficient resources to absorb the impact of certain outcomes, and so they cooperate in order to call upon each other’s resources according to who falls foul of such outcomes and who does not. Temporal smoothing is less dependent on pooling, at least insofar as drawing on pooled resources is not necessary for what smoothing tries to achieve. This is because setting aside a person’s resources early in life can occur by way of simply dispossessing them of these resources for some period to prevent their being consumed or lost, though without doing anything else with them during this time. In practice, though, smoothing very often will involve pooling. The reason for pooling is that something ought to be done with these resources in the meantime, not least because their value might otherwise depreciate, and so it makes sense to invest them in some way. This of course increases the benefits to the individuals concerned, particularly given a long time horizon. But this is not constitutive of smoothing, rather a means of generating supplementary benefits as a result of setting aside resources early on.
The distinction here is between smoothing across times or life stages, and smoothing across possible worlds. The counterfactual dimension of insurance is intuitively obvious from how we often speak of insurance – as a device we use when we don’t know whether things will turn out this way or that way. Insurance is rational for individuals who believe that their actual situation falls within a range of possible worlds across which their fortunes vary. Paying insurance means incurring a cost if the actual world is one of the better worlds, and receiving a payout if the actual world is one of the worse worlds. This is the sense in which insurance involves smoothing over possible worlds. Strictly temporal smoothing, on the other hand, involves an arrangement whereby individuals do not waste, or prematurely consume, whatever amount of resources they already have, or are in the process of accumulating.Footnote 13 Although there will generally be uncertainty about which world is actual, the assumption behind smoothing is that an individual can be expected to live for a certain duration and that smoothing over time – in the sense of preventing waste of resources early on – makes rational sense because of what’s expected, not what’s uncertain.
Temporal and counterfactual smoothing can conflict in practice. Because length of lifespan is not known, temporal smoothing will tend to involve ‘transfer’ of resources from short-life possible worlds to long-life possible worlds. This may result in temporal smoothing in the actual world (if life is long) at the expense of smoothing across different possible worlds. It is also possible for an attempt at insurance to gradually develop into an attempt at smoothing over time. This is precisely what might happen as an insurance scheme gets altered so that contributions increase, to such an extent that the reallocation rules get modified so that payouts are triggered for less than catastrophic events. Indeed, the history of retirement policy suggests this sort of evolution from insurance to smoothing (Anton Reference Anton and Scarre2016). This is perhaps a further reason why insurance and temporal smoothing are only vaguely distinguished in existing discussions. But the distinction is nonetheless real: Temporal and counterfactual smoothing are simply different concepts, albeit ones with overlapping extension.
As I have said, insurance involves smoothing across a range of possible worlds whereas life-course smoothing applies to stages of a single life. There is a sense in which life-course smoothing can plausibly aim at being maximal whereas insurance cannot plausibly do so. This is the key difference. If the full set of life stages is known – that is, if the duration of life is known – then it makes sense to take all life stages into account when smoothing wellbeing.Footnote 14 This is just to say that temporal smoothing tends to apply to the whole life course. Importantly, no analogous point is plausible for insurance. This is because it is not plausible to include a maximal range of possible worlds in which a relevant individual exists (that is, worlds in which they have a counterpart person). To do this would be profoundly difficult, even as a conceptual matter. The number of counterfactual scenarios is infinite, and greatly varied. This explains why, as a general matter, deciding how much insurance to take out, and in particular what range of scenarios to seek protection against, is often especially difficult. Insurance, then, is subject to a threshold of what’s reasonable in terms of the range of possible scenarios that should be protected against.Footnote 15 Some scenarios will almost always be considered too remote even when within the range of possible worlds. Temporal smoothing is not limited in this way: No life stage is considered too remote once it is considered within the range of the same life. The difficulty instead is of predicting length of lifespan, which is a rather different difficulty.
Insurance, then, aims to protect a person against some possibilities, but typically not the most distant ones. Life-course smoothing aims to protect a person from being worse off later due to wastage or bad choices earlier on. But it does aim to protect a person across a full life, insofar as life’s duration can be predicted. In this way, smoothing is more ambitious than insurance in terms of what it seeks to achieve. From this, we can draw an important conclusion: If mandatory retirement funding really is a kind of insurance, then it will need to deal with a problem of where to locate the threshold for the range of scenarios insured against. Life-course smoothing does not face this problem. Even if a life’s duration is not certain, there is no threshold at which a life stage is too remote from the current life stage for it to be excluded, in the manner by which remote counterfactuals might be excluded from insurance.
Philosophers and economists sometimes describe retirement saving in ways that are apparently indecisive between whether they see it as involving smoothing or insurance. The political philosopher Michael Otsuka writes that “through the transfer of income from the middle to the later years of our lives, pensions provide a solution to the problem we would otherwise face of living so long that we find ourselves lacking sufficient resources to sustain ourselves and prosper through retirement” (Otsuka Reference Otsuka2023: 1). Similarly, economist Gregory Ponthiere writes that retirement funding can “constitute a good insurance against a long life (i.e. the risk of becoming poor in case of survival to old age)” (Ponthiere Reference Ponthiere, Adler and Fleurbaey2016: 881). Otsuka’s formulation is ambiguous – not being able to sustain oneself might be more severe than simply not prospering. Indeed, to prosper is intuitively to go beyond mere sustenance. Ponthiere’s formulation, though it embeds the term ‘insurance’ apparently conflates what it is to live a long life with what it is to live a long life and become impoverished as a result. Such indecision means that scholars writing about retirement policy don’t completely clarify what they are proposing, insofar as they see retirement funding as realizing some kind of insurance or (temporal) smoothing, but without really deciding which one.
There are two further ways in which treating compulsory retirement saving as insurance can be called into question. The first is that any retirement saving scheme does not typically pay out to the worst off among its contributors. Those who are most unlucky are, plausibly, those who die too early in life to reach the age at which they qualify for access to retirement funds. The greatest beneficiaries are those who live long lives and who would otherwise be badly off but for the pooling. This point is developed at some length in the literature on retirement justice. Here, much emphasis is put on its normative significance, though typically under the assumption that the schemes in question count as insurance (Ponthiere Reference Ponthiere, Adler and Fleurbaey2016; Valente Reference Valente2024). The second, and related, complication is that retirement funds pay out automatically once people reach the relevant age. This is unlike paradigmatic insurance schemes like those for healthcare, housing, unemployment and so on. For these schemes, one needs to demonstrate that one meets the criteria under which one qualifies for a payout. If retirement schemes really were insurance against poverty in old age, then strictly speaking they would not pay out – or at least would pay out less – to people who were not otherwise facing poverty once old.Footnote 16 This would be just like the way in which home insurance does not pay out to homeowners whose houses have not burned down, been flooded, or whatever. Pooling schemes whose payouts do not serve to offset some relevant misfortune are not easily intelligible as providing insurance, though they might still be perfectly intelligible as an attempt at temporal smoothing. The only way to remove this complication is to deny that retirement funding is insurance against the conjunction of poverty and old age, and to present it as insurance against a long life merely as such. But insofar as longevity as such is not really an adverse event, this is counter-intuitive, and departs from the common-sense understanding of retirement funding. Ultimately, the idea that retirement funding is insurance against the conjunction of old age and poverty might not be entirely coherent.
Neither of these observations presents any difficulty for a view that treats retirement funding as an instance of temporal, life-course smoothing. The fact that retirement schemes pay out automatically once an advanced age is reached is perfectly natural if the goal is to even out wellbeing across a life, particularly if the age at which one qualifies for access coincides with the age at which one ends, or greatly reduces, one’s labour market participation. The problem of inequalities in longevity is somewhat more persistent, given that attempting to smooth out resources across a lifespan requires some judgement about the duration of the life in question. But on balance, the idea that retirement funding serves to smooth wellbeing rather than provide insurance against sudden drops in wellbeing fits better with some features of retirement funding. Indeed, temporal smoothing is probably a more coherent way of understanding retirement funding than insurance.
To be clear: I am not insisting that prior discussion of retirement saving has suffered from any total failure to understand the concept of temporal smoothing and its applicability to compulsory retirement saving. My point is that typically life-course smoothing has not been distinguished sharply enough from insurance, or indeed from the concept of pooling resources. When we make this distinction more robustly, it becomes clearer that compulsory retirement saving is primarily about the pursuit of smoothing across the life-course. Life-course smoothing will tend to converge with insurance with respect to smoothing across sets of counterfactual scenarios within which lifespan is relatively long. Once the primacy of temporal smoothing is recognised, it becomes harder to downplay a paternalistic rationale than it might be if all retirement insurance were insurance against specific emergencies that might arise late in life. The question now is how much plausibility there is for a paternalistic rationale that is primarily a rationale for smoothing. (To save words, I will now use ‘smoothing’ as short for temporal or life-course smoothing.)
4. Some Problems for Paternalism-as-smoothing
The case for smoothing draws some independent force from influential views about the philosophy of wellbeing. Here it is commonly accepted that lifetime wellbeing is not simply the sum total of wellbeing at or during the various constituent parts of a life. It matters also how the pattern or distribution of wellbeing is realized across the lifespan as a whole (Velleman Reference Velleman1991; Fletcher Reference Fletcher2016: Ch. 7). Precise accounts of lifetime wellbeing differ. But there is general agreement that a life with a relatively smooth distribution of wellbeing is a better life overall than one in which level of wellbeing at a time is more volatile, as in it goes radically up and down over time, or one in which it gradually declines, holding fixed the sum total. Comprehensive accounts of justice across the lifespan seek to accommodate these intuitions (Bidadanure Reference Bidadanure2021: 56–62). What’s worth stressing is that even on the assumption that a person’s wellbeing will never drop below a basic minimum, it is intuitive to prefer a life where an otherwise uneven distribution of wellbeing is somewhat evened out: Even if we lived in a world in which really bad things never happened to people, the tendency for wellbeing levels to vary across the lifespan would make smoothing desirable, and thus could make some mandatory retirement saving justified on paternalistic grounds.
I want to suggest two specific problems with smoothing when understood as subject to a paternalistic rationale. The first I call the problem of inverted bias. Recall that the case for paternalism draws some force from the case for offsetting cognitive relevant biases. Left alone, people are disposed to under-save for retirement. Most people have a bias towards promoting their short-term wellbeing, and discount their later wellbeing.Footnote 17 Compulsory retirement saving is supposed to counter this. The trouble is that compulsory saving risks doing too much.Footnote 18 In seeking to smooth out wellbeing over the lifespan, mandatory retirement policies force individuals to contribute resources early in life and then access them later on. But an effect of this is that many people will end up worse off with respect to their overall lifetime wellbeing. This is most obvious when people simply die too soon for access to resources later in life to balance out the cost to wellbeing during earlier periods of life. But it may also happen simply because, even for people who live to an old age, resources have been unduly skewed towards the later part of life. This means that the opportunity costs attached to mandatory saving earlier in life are not outweighed by the benefits of accessing resources later on. The more resources people are required to contribute during their years in the labour market, the more pronounced the problem of inverted bias can become.
The problem of inverted bias is more severe for smoothing than for insurance. This is because, as a general matter, insurance is limited in the way earlier described. Insurance stops at seeking to offset only sudden drops in wellbeing later in life, and does so by selecting a limited range of possible worlds that the actual world is taken to sit within. Smoothing, more ambitiously, seeks to alter the distribution of wellbeing within a life even when enough has been done to offset sudden drops in wellbeing. A life containing a very high level of average wellbeing can still be evened out well after enough has been done to ensure that wellbeing never drops below a critical level. To put it another way, insurance has more of a satisficing character than smoothing, where this reflects a judgement about which possible worlds are too remote to be worth including. An effect of this is that insurance may not seek to completely smooth wellbeing across life stages. This is why the risk of inverted bias is less severe for insurance.
As an empirical matter, many of the world’s actual retirement savings schemes exhibit a degree of inverted bias. This is most pronounced, or at least most visible, for policies where there is a strong connection between what individuals contribute during their years of working and what they access after retirement. This includes the kinds of compulsory savings schemes that are not pensions because contributors simply keep the entirety of what they contribute, with no transfers across individuals. A standout example is the Australian policy of superannuation, which began to replace a pay-as-you-go pension scheme (where each working-age birth cohort funds the pension of the older retired cohort via taxes) in the 1990s.Footnote 19 Because compulsory contributions are a fixed percentage across all salary levels, there are now retired Australians with several millions of dollars in their superannuation accounts. Realistically, this is more than these Australians are able to consume in the years they have left. And even if they do consume it, the benefits are likely less than the opportunity costs that were absorbed when making the required contributions. The Australian approach to retirement funding is sometimes presented, at least politically, as some kind of standout success. This is true to some extent, at least in that demographic ageing does not undermine the feasibility of the funding model in the way it does for pay-as-you-go pensions. But such appraisal privileges visible contemporaneous benefits over less visible opportunity costs absorbed decades earlier. It is likely that wellbeing among this cohort of Australians would have been made smoother across lifespans if mandatory contributions had simply been lower, at least for those Australians who had higher incomes.
Inverted bias is a failure of paternalism: If a paternalistic policy is justified according to its capacity to mitigate a cognitive bias that will distort individuals’ free choices, then it cannot be successful if it simply replaces that bias with some other bias in the opposite direction. It may even be that an inverted bias is qualitatively worse than the kinds of cognitive bias in individuals that it is meant to correct: After all, an individual who is worse off in old age because of their earlier sub-optimal saving may have at least lived an authentic life. An individual who suffers opportunity costs earlier in life for the sake of having more resources than they need later on has, instead, been the victim of a kind of irrationality that isn’t even their own. Even setting aside tricky questions about how to understand authenticity, the problem with inverted bias is severe, because the opportunity costs are very large. This reflects a more general principle that it can be objectionable to ‘back-load’ resources when trying to distribute goods over the course of a life (Jauch Reference Jauch2023). Even if an individual does not really notice the foregone income that their employer pays directly into a retirement savings account, being able to access (more of) this income at the time can make a huge difference. This will impact on life-changing factors such as qualifying for a home loan, avoiding insolvency,Footnote 20 being able to change careers, feeling able to afford to have (additional) children, and many others. Having access to cash during retirement does not necessarily make up for this, or result in a more even distribution of wellbeing over one’s lifespan, even if the effect of a savings bias is the attractive vision of older people enjoying high levels of financial freedom and comfort during their final years.
Lifespan smoothing comes with an increased risk of inverted bias, the greater the amount of resources in a person’s life and the greater the degree of inequality in longevity within people’s lifespans. Inequalities in longevity mean that even when smoothing is optimized for the typical or expected life length, a larger range of actual life lengths will tend to mean more savings bias. Inequalities in longevity are such that the median life length is subject to more deviation by way of especially short lives than especially long lives. Put more intuitively, a small number of people die much earlier than the average whereas a large number of people die just a little older than the average. This means that the short-lived will be more deprived than the long-lived will be benefitted, when it comes to any policy oriented around the average lifespan. Second, smoothing seeks to move around a greater quantity of resources than mere insurance. This just increases the margin for error when it comes to avoiding a bias one way or the other. Insurance enjoys a smaller margin for error because it seeks to do less.
A second difficulty is the problem of the arbitrariness of income from labour. One often neglected feature of fiscal policy is its tendency to rely on the flow of income from paid labour as a source of funding. Largely this is a point about taxation – the income tax is the dominant tax base in most developed economies, regardless of what its rate structure might be. The funding of retirement follows this trend – as with the taxation of income, contributions to retirement funds are made out of individuals’ paid income. Sometimes this is just by way of income tax itself, though sometimes it is by way of separate contributions made by one’s employer, and perhaps voluntary contributions from one’s pre-tax income.
What’s curious about this is why states will target income from labour when seeking to smooth the consumption of individuals, given that states rarely seek to target alternative sources, such as individuals’ wealth or incomes from sources other than that individual’s own labour. Very often these alternative sources just don’t exist – many individuals only have paid income as a source of wealth. But there is a non-trivial segment of the population that receives a substantial amount by way of inherited wealth, and this segment is larger than it used to be. Though this often arouses concerns about justice (I will come back to these) it is notable that receipt of inheritance does not trigger any kind of paternalistic intervention. This is despite the fact that inheritance is more likely to be received as a single lump sum than as a prolonged series of smaller payments in the manner typical of income from labour. A life containing a large inheritance is therefore more likely to be a life with an uneven distribution of wellbeing, just because it is easier to squander a pile of wealth that one receives all at once. From a purely paternalistic point of view, it is strange to target income but not inheritance, and indeed one might think that if there is a discrepancy then it might be the other way round. If people who receive income from labour should be required to set aside some of what they receive for later, then why shouldn’t inheritors be required to do the same, especially if the risks of imprudence may even be higher?
There may, admittedly, be reasons of some kind for the privileging of income from labour. Inheritance is often received in illiquid form, by way of housing wealth or other ‘hard’ assets. Unlike income from labour, which is almost always a pure cash flow, it is hard to design policies that slice off part of the market value of illiquid assets. This is one reason for why income persists as a dominant tax base and wealth, when it does fall within the tax base, is taxed usually at low rates.Footnote 21 But this is not a satisfactory explanation for exempting inherited wealth from paternalistic interventions. In many cases of inheritance a pooling requirement could still be put in place without requiring liquidation of an inherited asset in order to pay tax. Indeed, anyone who has a large stock of wealth could be required to purchase insurance against whatever events might cause that asset to be destroyed or lose its value. This could even be done while relieving such individuals of a requirement to contribute any income from labour to their retirement. But such policies don’t really exist and never have. People who have wealth are generally left alone to use it prudently or imprudently, and yet people with income from labour are deemed more appropriate subjects of paternalism. Perhaps this is just part of the state’s long-standing historical deference towards wealthy individuals that persists alongside unfair suspicion towards those who need to work for a living (Anderson Reference Anderson2023).
As the above remarks might suggest, the problem of the arbitrariness of income from labour would persist if retirement funding were limited to insurance rather than to smoothing.Footnote 22 But the point is that it is more pronounced given the ambitions of smoothing. If someone inherits wealth, then it may be that their protection against sudden drops in wellbeing has already been secured (in the relevant satisficing manner that excludes remote possibilities) by mandatory income contributions. So there would be no need to add anything from the inheritance. But if the goal is to smooth out this individual’s lifetime wellbeing as much as possible, then the case for a mandatory surrender of some of the inheritance will persist.
What goes for inheritance is really a more general point about wealth whose value can be put at risk by the imprudence of its owner. Generally speaking, states do not impose requirements on home-owners to insure their residential property against the kinds of accidents that dramatically reduce its value, or render it uninhabitable without major repairs. While such protection is generally what private home insurance offers, the point is that its purchase is not compulsory.Footnote 23 Depending on its potential for appreciation, owning a home can be a means of funding a retirement, so long as the home does not lose its value through befalling some catastrophic event in the years prior to retirement. But consider an individual who inherits a valuable home, which they thereby own outright. Imagine also that this individual does not have a particularly well-paid job, and could not purchase or even rent a comparably valuable property based on their regular income alone. In a scenario like this, the home can actually serve as a de facto retirement policy, insofar as it could be sold after a couple of decades’ appreciation, with enough funds remaining to fund routine expenses after the purchase of a more modest residential property.
To conclude, the paternalistic case for temporal smoothing suffers from two problems so long as it used to justify policies of making individuals make mandatory contributions linked to their income. The first is the problem of inverted bias that becomes worse as retirement funding departs further from the limited goal of insurance and becomes a more ambitious exercise of evening out larger total quantities of resources. A second is that if states do have authority to smooth out wellbeing on paternalistic grounds, then this would seem to suggest that the mechanism should not be confined to contributions from income from labour, but include other things, like inherited or simply possessed wealth especially in forms that work as a store of value. The focus on funding retirement from labour market income is arbitrary. But if we think that the state lacks authority to make paternalistic interventions on the possession of wealth, then it remains puzzling why we should think it has authority to do it for individual’s labour market earnings.
5. Implications for Retirement Justice
I have argued that the paternalistic case for retirement funding, understood as a relatively ambitious attempt to smooth wellbeing across the lifespan, faces problems. So far these have been problems internal to the idea of paternalism itself. If one thinks that the case for mandatory retirement saving has nothing to do with whether the state has authority to pursue paternalism, then nothing much is learned from problems with paternalistic rationales. As I said at the outset, the implicit assumption seems to be that paternalistic considerations add some general force to the overall case for mandatory retirement saving. But I also said that the discussion of paternalism would end up revealing some implications about justice. This adds further interest. I’ll now demonstrate how this is the case, albeit somewhat briefly.
The first point extends the earlier discussion of inverted bias. When some people end up compelled to save an excessive amount of resources for their retirement, this tends to have unjust consequences. Depending on the details, a savings bias might compound prior distributive injustices or (depending on how things are described) have unjust knock-on effects. The most obvious of these is that when people die with unspent savings and they own their retirement funds, these will very often get passed on as inheritance. Empirically, most inheritance goes to a small number of offspring, with most jurisdictions nowadays taxing inheritance either not at all or very little. As a general matter, inherited wealth raises a range of questions about justice (Halliday Reference Halliday2018). Specific to retirement funding, the real ‘success’ of Australia’s superannuation policies has been in achieving a tax-free inheritance scheme for the children of high earners. A different injustice might exist in PAYG pension systems (recall, these involve working-age cohorts being taxed to fund the retirements of older, retired cohorts). In such cases, it is possible for a younger birth cohort to absorb an especially high tax burden to pay for pensions received by an older birth cohort. This might amount to a kind of intergenerational unfairness, either because of opportunity costs absorbed by the young (taxes could instead be used to fund their education, etc.) or the ability of some of the old to absorb a reduction in their pension, for example by drawing on housing wealth. I lack space to go through the various possibilities, but it is plausible that a savings bias in retirement funding may present distributive injustices when the providers of the funds are a different group from the recipients.Footnote 24
A second point relates to what exactly it is that retirement savings are meant to protect people from. So long as retirement policy remains limited to compulsory insurance, it saves people from especially bad outcomes such as poverty, starvation, or enduring inadequate shelter, health, or access to other basic goods. There is nothing especially problematic about placing importance on these goods, as they tend to have high value across a range of conceptions of the good life. But, as I have been at pains to argue, temporal smoothing does rather more than this. For individuals who have high incomes, and hence a lot of resources within their life as a whole, smoothing wellbeing is essentially a way of ensuring that a high average level gets maintained throughout a life. This is akin to suggesting that it would be especially bad for an individual to move from a higher to a lower socio-economic status later in life. This might be problematic. Some have suggested that the very idea that justice is compatible with preventing downward mobility (of the modest sort that doesn’t see people sink to a level of poverty) may be little more than an arbitrary protection for the middle and upper classes (Persad Reference Persad2018). For this reason, paternalism aimed at consumption smoothing runs the risk of expressing something anti-egalitarian, namely that there is something especially bad about dropping out of the middle class.
I do not mean to suggest that the expressive aspect of paternalistic smoothing policy is close to the most objectionable paternalistic policies. It may be that the anti-egalitarian content of suggesting that downward mobility (short of reaching poverty) is a personal disaster is minor compared with other policies. I expect the stigmatizing effect of weight-loss programmes, and possibly anti-tobacco policy, may be worse. A plausible theory of wellbeing might indeed suggest that downward mobility is generally a bad thing even holding fixed the sum of wellbeing during the various life stages. But it may simply be more respectful for a state to leave it to individuals whether they are to take the kinds of risks that promote youth or middle age over old age.
6. Conclusions: More Work for a Theory of Justice?
I have tried to shed light on a paternalistic rationale for compulsory retirement funding that has typically been assumed but not examined. I’ve argued that the more coherent paternalistic rationale for retirement policy, as well as the one that might fit best with the bulk of actual policy, is one that fits the model of consumption smoothing rather than insurance, strictly speaking. But despite being more coherent, there are problems with a paternalistic case for retirement funding once mandatory schemes are understood as aiming at consumption smoothing. Some of these problems stem from the broader idea of paternalism itself, particularly the idea that paternalism should not over-correct for whatever biases it seeks to offset. Other objections stem from considerations of justice, particularly the idea that retirement saving should not compound existing material inequalities, or become vulnerable to the kinds of expressive objections that afflict paternalism in contexts otherwise unrelated to retirement funding.
Of course designing a system of retirement funding that contained no over-correction or inverted bias is going to remain remarkably difficult particularly so long as people’s lifespans are of unknown length taken invidivually but also are bound to be of unequal length taken collectively. I don’t mean to downplay this. But if compulsory retirement funding were to draw support from paternalism, then most existing schemes would have to be reformed either by way of making them less ambitious (so as to be closer to insurance than to smoothing), and/or less tied to contributions from paid income alone (as opposed to contributions from, say, inherited wealth). It remains possible that a retirement funding model with poor paternalistic credentials could still be justified, all things considered, if it could promote justice. I suspect that to do this, the funding of retirement may need to become more redistributive within birth cohorts. If the compulsory funding of retirement can be shown to promote distributive justice, then it may not matter as much if any paternalistic rationale is – as I have argued – shakier than typically assumed. It may be harder to put together a robust paternalistic rationale for compulsory retirement funding than has typically (though often implicitly) been supposed. If so, then it may be that resolving questions of how to design retirement funding should be treated more as a question of justice, with less reliance on the supposition that a strong paternalistic rationale exists.
But these are all preliminary points. The more basic aim of this paper has been to suggest that in taking it for granted that compulsory retirement funding enjoys a more solid paternalistic rationale than some of the more dubious cases of state-imposed paternalism, we risk taking more for granted than we should. The paternalistic case for compulsory retirement funding is rather more fraught than it looks.
Acknowledgements
I would first like to thank David Birks for stimulating my interest in paternalism and helping me work out how to put the project of this paper together. Earlier versions of this paper were presented to audiences at Frei Universitaet Berlin and at the Hoover Chair in Social and Economic Ethics at Université catholique de Louvain. Much of the writing on this paper was carried out while occupying a residential fellowship at The New Institute, Hamburg, during the academic year 2024–25. I thank the fellows and the team at TNI for a supportive environment and many enriching interactions. I also benefitted from a series of conversations on this topic with Colin von Negenborn while in Hamburg. Thanks are also due to Marco Meyer and to Ben Sachs-Cobbe for separate written comments. Finally I thank two anonymous referees for Economics and Philosophy.
Financial support
Work on this paper was supported by Australian Research Council Discovery Award # DP210100924 Sharing the Wealth: Tax and Justice in the Slow Growth Era.
Daniel Halliday is Professor of Political Philosophy at the University of Melbourne. He works in political philosophy, particularly where it intersects with economics and legal theory. He is particularly interested in issues of justice regarding labour markets, taxation, credit and lending, housing, and old age.
Open access
