Introduction
Traditionally, in economics at least, public policy is directed at the satisfaction of individual preferences. This is clear whenever the Pareto criterion is employed to judge policy interventions, and in this respect, the behaviourally inspired ‘nudging’ approach to public policy is no different. It assumes that people have some underlying, well-defined preferences that, as a result of well-known behavioural biases, they fail to satisfy as well as they might. A judicious public policy nudge can push them towards better preference satisfaction (see Sunstein & Thaler, Reference Sunstein and Thaler2008; Halpern, Reference Halpern2015). I do not wish to argue against the idea that people (often/sometimes) have preferences (that is, it makes sense to talk about preferences being more or less satisfied), and that nudges can be used to good effect. However, the starting point of this paper is that behavioural evidence also suggests that there are many occasions when we seem not to have well-defined preferences at all. In other words, it is not just that we have well-defined preferences but do not reveal them because our decision making is guided by heuristics and biases; we sometimes simply do not have the well-defined preferences in the first place.
This is a common interpretation of the behavioural evidence among psychologists (see, for example, Slovic, Reference Slovic1995). On this account, heuristics are as much involved in deciding what is valuable in the first place as in evaluating how an action contributes to the generation of such value (see, for instance, Simon, Reference Simon1978; Kahneman, Reference Kahneman2003).Footnote 1 Consider, as an illustration, the behavioural evidence on decoys: that is, the introduction, say, of a weekend in Paris where breakfasts are not included into a choice between a weekend with all-expenses paid in either Paris or Rome.Footnote 2 The introduction of the inferior Paris decoy typically shifts peoples choices towards the all-expenses paid Paris option and away from the all-expenses paid Rome alternative. This is extremely difficult to explain under any assumption that people have well-defined preferences over weekends in Paris and Rome. But the shift can be explained by the processes through which people come to hold a preference for one weekend location over the other [e.g. see Stewart's (Reference Stewart2009) account of decision by sampling, where the decoy increases the relative mental tally in favour of the all-expenses paid weekend in Paris]. The question that arises and that I address in this paper is: how should public policy be framed in these circumstances when, because we seem not to have underlying well-defined preferences, there is no possible standard of preference satisfaction?
My answer is constitutional or procedural. Public policy should be concerned with the character of the rules that constrain and enable behaviour (e.g. how procedurally fair they are and how much freedom they allow) and not simply the behavioural outcomes themselves (e.g. the extent of preference satisfaction).Footnote 3 This is not to say that outcomes do not matter – they do. Indeed, it is often the case that we judge different procedures by the outcomes they tend to produce. The point is that we have to think and make policy in relation to the procedures as distinct from the actual outcomes.
One way of understanding this argument is to position it in the debate over ‘libertarian paternalism’. That people might not have well-defined preferences is sometimes taken in this debate as a reason against policy intervention altogether, because ‘nudging’ can only be paternalistic in these circumstances (e.g. see Sugden, Reference Sugden2009). The nub of my argument is that, contrary to this, the lack of well-defined preferences does not count against intervention. Rather, it supplies reasons for policy intervention, but of a different kind: intervention at the constitutional level. Before I illustrate what this might mean in more detail, I sketch next how this approach has deep roots in the liberal political tradition. This is important because part of the enduring appeal of ‘individual preference satisfaction’ in economics (and more widely) is that it respects the individuality of the individual in liberal societies (‘de gustibus non est disputandum’). In particular, I present an account in the next section of how this alternative liberal approach points to the need for a constitution that encodes liberty and a redistributive bias. I illustrate how the constitution of liberty can be redistributive without compromising the constitutional approach to public policy in the sections entitled ‘An outline of a constitution for liberty and redistribution’ and ‘An illustration’. The illustration concerns the tax and benefit system. It has a tax system that is redistributive and constitutionally insulated from normal democratic tinkering and a welfare benefit system in the form of a basic income.
Mill on liberty
“The only freedom which deserves the name, is that of pursuing our own good in our own way, so long as we do not attempt to deprive others of theirs, or impede their efforts to obtain it” (Mill, Reference Mill and Collini1859, p. 16).
That it is our ‘own’ good’ and in our ‘own’ way proves to be crucial in this canonical statement at the beginning of On Liberty. That we ‘own’ what we do is important because Mill's defence of liberty turns later on the way that it contributes to the ‘free development of individuality’. A person becomes an individual (they obtain ‘individuality’ and they have their ‘own character’) because they have been free to think about who they want to be, to discuss freely with others what matters and to engage freely in ‘experiments of living’. This matters because such self-direction is what contributes to individual well-being.
“He is capable of rectifying his mistakes by discussion and experience. Not by experience alone. There must be discussion, to show how experience is to be interpreted” (p. 23).
“As it is useful that while mankind are imperfect there should be different opinions, so is it that there should be different experiments of living…” (p. 57).
Indeed, if only it “were felt that the free development of individuality is one of the leading essentials of well-being … there would be no danger that liberty should be undervalued” (p. 57).Footnote 4 The problem is the “disposition of mankind, whether as rulers or as fellow-citizens, to impose their own opinions and inclinations as a rule of conduct on others…” (p. 17). The result is clear: “One whose desires and impulses are not his own, has no character, no more than a steam-engine has a character” (p. 60).
Thus, Mill argues for liberty on grounds that are not dissimilar to another of the great defenders of liberalism, Hayek. In The Constitution of Liberty, Hayek (Reference Hayek1960) argues that freedom matters instrumentally because we have imperfect knowledge. It is the solution to the knowledge problem. For Hayek, the knowledge problem is ubiquitous. For Mill, it is narrower: the knowledge problem concerns our identity. To own our preferences is to have an identity and to own them requires that we know why we have them. Such knowing is a life project and liberty makes available the practices of thinking and acting through which we develop the ability to discern and desire what is best.
The contrast with the sense of respect for the individual in standard welfare economics could not be sharper. It comes in conventional welfare economics from satisfying an individual's preferences. This shows a liberal respect for the individual because individuals are identified with their preferences. An individual is an individual; they have an identity because they have reliably fixed preferences. As Stigler and Becker (Reference Stigler and Becker1977) famously put it:
“…one does not argue over tastes for the same reason that one does not argue over the Rocky Mountains – both are there, will be there next year, too, and are the same to all men” (p. 76).
Mill would have been shocked at the suggestion. For Mill, our preferences are in a state of flux, open to amendment in the light of experience, discussion and thought. This is how we come to have an individual identity, and it is little wonder that we often do not reveal a well-defined set of preferences in our behaviours.
Mill is useful not only for offering a different sense of individuality to value in liberal society; he directly connects the fluidity of individual preferences to a procedural or constitutional public policy. It is a constitution of liberty that allows individuality to develop. Public policy is not directed to achieve a specific outcome (although, of course, Mill is not immune to outcome orientations in his other writings, notably on utilitarianism); it is directed at getting the rules within which action takes place right. Getting the rules ‘right’, however, is not simply a matter of establishing a constitution that gives liberty to people so long as these liberties can both be shared by all and do not cause harm to some. Mill is clear that people need not only a constitution of liberty, but also resources for thinking and reflecting. One such set of resources are the institutions that promote reflection and public discussions: the education and media systems. Another is the mental capacities that come from age: “…this doctrine is meant to apply only to human beings in the maturity of their faculties. We are not speaking of children or young persons” (p. 13). Nor does it apply to societies that are so poor that they spend most of their time providing for subsistence: “…we may leave out of consideration those backward states of society…” (p. 13). This last point is important for the argument because although Mill makes it about whole societies, it would seem to apply as well to individuals within a society. Individuals need resources, so to speak, to engage in thought, discussion and ‘experiments in living’. Two questions arise: first, to what extent does this mean that a public policy of encouraging people's individuality must be concerned with inequality in these resources? At first glance, this would seem unavoidable, because if individuality is what counts, then all individuals would seem to count equally. Second, how can policy be directed at inequality of resources while remaining consistent with a constitutional approach? The apparent difficulty is that inequality refers to the pattern of outcomes, while the constitutional approach focuses on the rules under which the outcomes are generated, and not the outcomes themselves.
This is not the place to attempt a detailed answer to the first question. I want to claim for the purposes of moving on to the second question only that, in the current circumstances, a liberal individualist of the Millian kind would require some redistribution of resources in the typical Organisation for Economic Co-operation and Development (OECD) country (i.e. there is a public policy interest in redistribution on these grounds). This is because, in many OECD countries, the living standards of the poorest have declined in absolute terms over the last 10–20 years at the same time as those of the rich have increased, and this transgresses a version of the no harm principle (see OECD, 2015). The exercise of freedom by the rich has, as it were, harmed the poor. The rich have engaged in mutually beneficial trades of one kind or another that have, not by intention, but as a matter of fact, made the poor worse off. This counts against the liberal ‘no harm’ injunction for the exercise of freedom. This is one argument. Another is that any practical argument for a liberal order must not only provide the case for liberty, but it must also plausibly show why such an arrangement would be stable in the sense that those living under liberal order would not want to change it. This is an invitation to think about the liberal order in contractarian terms, and when the order actually makes some people worse off, it is unlikely to satisfy the contractarian test.
The second question concerns me in the rest of the paper: how are we to enact public policy towards redistribution while maintaining the constitutional approach? The outline answer is that the constitutional rules should have scope for a redistributive bias.Footnote 5 I offer an illustration of how to give the constitution such a redistributive bias in the next section.
An outline of a constitution for liberty and redistribution
There are two elements to this illustration of a redistributive constitution: the tax and the benefit systems. The benefit system is, perhaps, the simplest to illustrate because there is a ready-made constitutional proposal in the form of a basic income. The public education and health care systems in many countries are also available, like the basic income, to all citizens independent of income. These are constitutional because they are built into the rules of being in a society. They are the backdrop against which action takes place. I focus on the basic income element of this because (unlike education and health care in this respect) it would represent a public policy change from the largely means-tested welfare payments (i.e. outcome contingent) that characterise many current benefit systems.
The constitutional character of the basic income appealed to Hayek in The Constitution of Liberty, just as it should to Mill in On Liberty. The fact that it is received as a right and not contingently means tested has the well-known advantage that it does not distort incentives to action. Mill would, perhaps, see this insight from a different angle. For Mill, people have to feel responsible for their actions if those actions are to be a source of learning. They will not feel responsible either if they do not seem to have the resources to make many effective choices or if the consequences of an action become blurred, as they often do under a contingent-based benefits system when the effective marginal tax/benefit withdrawal rates can be very high. Or, to put the point in a more behavioural way, attention is a scarce resource and it is only engaged when it has to be; that is, in this instance, when choices have material consequences. Basic income, for this reason, is more likely to engage attention than the current means-tested system of welfare benefits.Footnote 6
Until relatively recently, the attractions of a basic income in these respects have largely been trumped by the financial impracticality of introducing a basic income of sufficient size to fill this role in any remotely plausible manner. However, two things have changed the financial calculation in the UK, for example. One is the fall in the share of wages in GDP. This means that the average wage is now roughly the same as GDP per capita (whereas in the past it was higher). As a result, the average rate of tax on GDP to achieve a basic income that is equal to ‘x%’ of the average wage is also ‘x%’ (whereas in the past the tax rate had to be higher). The other is the increase in the value of the personal tax-free allowance (which is, in effect, like a basic income for those in work). The cost of all the current social security programmes is approximately 14% of GDP, and the cost of the personal tax-free allowance is 4–5% of GDP. Thus, it is now possible without significantly raising taxes to give everyone the equivalent of 20% of the average wage (which is the value of the basic State pension) as a basic income instead of the current social security programmes and the tax-free allowance.
As a cross-check on how this might broadly affect those in receipt of benefits under the current UK regime, it is worth noting that a family of four in receipt of this basic income would receive close to what is the current benefit cap. The obvious losers from this arrangement would be those on disability benefits and those in receipt of housing benefits in households with fewer than four people. Since the basic income on this calculation is payable to everyone, it would be possible to address the greater needs of the disabled through increasing their basic income, covered by reducing the basic income of those under some age (e.g. those under 16). The deficit for those on housing benefit is best addressed by unwinding the favourable tax treatment of housing that has contributed to its high cost. I take this up as part of the tax system next.
The taxation dimension of the redistributive constitution is both simple and complex. It is simple in the sense that the rules of taxation should be redistributive and they should be changed in this respect only from time to time (i.e. in light of the experience of how they seem to have contributed to everyone being able to engage in the development of individuality). They should only change from time to time to preserve their constitutional status. They need to be taken as rules that constrain and enable action, rather than the objects of actions themselves (as when they could be changed continuously through an unfettered process of democratic politics). Of course, there are other reasons for having taxes, like the macroeconomic management of aggregate demand, and so some aspects of the tax system will need more frequent change. The point here concerns preserving the rule-like quality in relation to their redistributive stance.
The immediate practical difficulty of changing taxes periodically to achieve a desired redistributive bias is that the current tax system is so complex that there is neither a single tax that can be used for this purpose to track the redistributive stance, nor do we really know how redistributive the tax system is as a whole. Complexity has become the enemy of doing anything with the tax system, not just the creation of a redistributive constitution. Consider, for example, greenhouse gas emissions in the UK. There are many taxes that impinge on emissions and they often pull in different directions and interact with other taxes, with the result that the overall effect becomes difficult – if not impossible – to assess. How does the encouragement to consumption (because saving is double taxed) weigh against taxes on motoring, and what is the net effect of the tax breaks for renewable energy when set against the inducement to invest in agriculture that comes from its inheritance tax exemptions? In addition, this is just the beginning of a very long list of such questions one would have to address in checking all of the nooks and crannies of our tax system for an answer.Footnote 7 This is, in effect, the lesson of complexity theory and the analysis of complex systems: the effects of interventions in such systems are difficult to predict. About the only thing we can say with much confidence about our currently complex tax systems is that they spawn an industry of tax accountants, lawyers and lobbyists who game the system for the benefit of their clients, who are mainly rich. This, of course, is one reason why it is also difficult to know with confidence what would be the redistributive effects of the tax system. Other reasons relate to the complexity that comes from treating individual receipts from employment, capital gains and self-employment differently and from making the allowances also vary with how the receipts are used. The two are connected and explain why simply making income tax more progressive under the current system has uncertain effects: increased progressivity encourages the rich to employ accountants to exploit the loopholes in such an overall complex system.
This is a fundamental point. It means that something has to be done to tackle the complexity of the tax system if it is to be used to achieve any first-order political end, like controlling greenhouse gas emissions or redistribution. This is the same whether tax policy is directed at actual outcomes or the rules (as I am suggesting here). It is also a point, however, that adds weight to the argument for the constitutional or rules approach. This is because a plausible account of the growth of complexity is the one advanced by the Public Choice School, whereby the normal operation of democratic politics is for small groups to secure sectional benefits that are tangible for members of those groups and that come at the expense of most tax payers, who do not notice the individually much smaller costs of the intervention (see Butler, Reference Butler2012). Complexity in the tax system is exactly what you would expect from an unfettered democratic politics on this account, and the solution is constitutional: there have to be constraints on what governments can do with respect to taxes.
What follows is therefore an illustrative proposal for a constitutional tax system that tackles complexity and allows a democratic influence through the determination of the constitution and specific democratic discretion to change tax for macro-management purposes, and yet still preserves the rule-like quality of the system in relation to redistribution. This is made possible by having taxes that are each targeted at a specific objective and allowing the tax directed at macro-management to be adjusted more frequently than the tax targeted at redistribution.
In particular, the illustration has the Government decide periodically on the objectives of the tax system, and then the implementation of those objectives via the tax system is handed to an independent tax authority (ITA). The ITA is, in turn, also constrained in the development of the actual tax system by principles of taxation, which have themselves been democratically established. The arrangement is, of course, modelled on the delegation of the conduct of monetary policy to an independent central bank and is designed to achieve the same result of insulating the conduct of policy from the sirens that can inhabit parliaments.
An illustration
For the purposes of illustration, I will distil the findings of the most recent review of the tax system commissioned by the UK Government (the Mirrlees Review, 2011) into two guiding principles for an ITA.
Principle 1: neutrality
The tax system should not distinguish between different activities unless there are externalities. This applies equally to taxes on how we dispose of income as to those on the sources of the income/resources in the first place. Thus, there is no obvious reason under the current system as to why savings (unless shielded, such as in the UK, by the Individual Savings Account) are taxed twice; nor why income from profits is taxed twice (once via corporation tax and again when received by individuals through dividend and capital gain); nor why income from labour is treated differently from that which accrues from non-labour wealth (e.g. in the form of a separate capital gains tax, with a separate tax-free allowance); nor why some types of non-labour wealth are treated differently from others for capital gains tax (e.g. housing); nor why we tax gifts in life differently from those after death; and so on. Thus, unless there is a good reason – like that of an externality – to want to discourage or encourage one activity rather than another, then the same tax should be applied to all items of expenditure. Likewise, the default, in the absence of some reason for encouraging people to spend more time acquiring resources in one way rather than another (or claim for tax purposes that they do), should be that all sources of income – broadly understood to include income from employment and non-labour wealth, including capital gains, and gifts – are summed and taxed using the same tax code.
Principle 2: appropriate simplicity and comparative advantage
Appropriate simplicity means that there needs to be as many taxes as there are objectives, and comparative advantage in this context means that each tax should be targeted at the objective on which it has the greatest relative effect. This is an old insight from Tinbergen (Reference Tinbergen1952) [see Mundell (Reference Mundell1968) for a classic exposition of the 2 × 2 case].
To see what the ITA might do with these principles, I shall assume that the democratic government cares about three objectives: (i) the redistributive stance of the tax system; (ii) the climate change that arises from the externality of depositing greenhouse gases in the atmosphere (this aspect would obviously generalise to other external effects); and (iii) the overall level of taxation (given the scale of public expenditure and the requirement of managing aggregate demand).
An illustrative ITA proposal in these circumstances would have three taxes to address the three objectives, each targeted at the objective they have most effect upon: (i) a progressive tax on all individual income and income-like receipts (i.e. other sources of changes in net wealth, like capital gains receipts and gifts); (ii) a carbon tax; and (iii) a uniform VAT. The progressivity of tax (i) would be adjusted once in the life of a parliament to preserve its rule-like status, tax (ii) would be adjusted whenever the progress in combatting climate change is deemed either too slow or too fast and tax (iii) would be adjusted whenever macroeconomic management makes such adjustment sensible.
There are two points that should, perhaps, be made about the focus on individual receipts in this proposal. The first concerns the ‘individual’ part. This is natural in liberal societies because they care about individuals. But more generally, it is difficult to understand why, if we are concerned with inequality, we should care about any aspect of this other than individual inequality or inequality between groups of individuals (e.g. ethnic groups). Why, for example, would we have a tax on corporate earnings in pursuit of an inequality objective? Corporate profits affect individual inequality through dividend payments and capital gains, but the natural thing to do if one is concerned with individual inequality is to focus taxes on individual receipts for this purpose (and not on some part of the process generating them). Of course, there might be some other reason for taxing corporate earnings. For instance, it may be a way of taxing some of the profits made in a country that, with foreign ownership of the company, would otherwise escape taxation. This may be warranted when the corporation relies on infrastructure and other public investments to secure these profits. But if this is the case, then it would be better to have a tax on foreign ownership that directly addresses this objective. For a similar reason, why would we have a tax on the estates of dead people? The inequality between dead and living individuals is not something that is obviously important. Of course, like corporations, the estates of dead individuals affect inequality among the living, but when the objective is individual inequality among the living, then the tax should focus on the income and income-like receipts of those individuals. This is why gifts are included in the definition of receipts for the purposes of the ‘income’ tax.
The second point is with respect to the ‘receipts’ part of this proposal. There are two ways of measuring the differences in the potential command over resources that individuals have: there is a stock measure at a point in time (wealth) and a flow measure over an interval of time (income). The two are related because wealth properly measured gives rise to income flows. Thus, to avoid double taxation of some kinds of wealth/income creeping into the tax system, it makes sense to tax either individual wealth or income, but not both. The proposal here selects income, again suitably defined to capture all income and income-like receipts that are generated by all kinds of wealth. I favour income because the component of wealth that generates labour incomes – human capital – is not typically measured (and nor would it be easy to do so). In contrast, measures of labour income are readily available. There may be other reasons for preferring a tax on individual wealth rather than income receipts. For example, the temporal dimension of inequality may make wealth more appropriate because it is a better measure of an individual's lifetime possibilities at a moment in time, whereas income focuses only on the possibilities, without depleting wealth, open to an individual over the interval for which income is measured. In the absence of a compelling reason of this kind, the practical one of data availability prevails.
Conclusion
I want to claim that behavioural economics has profound implications for public policy, but these implications are not simply those that surface around nudging. Rather, I argue that behavioural public policy should also be concerned with the character of the constitutional rules that constrain and enable action. The argument is for the constitutional approach to public policy. This is not a new suggestion (see Brennan & Buchanan, Reference Brennan and Buchanan1985). Nor are the specific illustrative constitutional recommendations that I develop for a basic income and a simplified, constitutionally constrained tax system especially new (although some details are). The argument for them is, however.
This argument turns on three points. First, the evidence from behavioural economics that we do not always have underlying well-defined preferences makes the identification of a person's individuality with their preferences problematic. Second, liberal political theory is not tied to such an identification (and the associated policy concern with preference satisfaction). Mill supplies a different reason for valuing the individual. People have individuality, which they acquire through the exercise of thought, discussion and experiments in living, and this is made possible by a constitution of liberty. Finally, the character of the constitution that matters for public policy purposes is not just the liberty it gives, but also its redistributive bias. This is because an individual's acquisition of individuality depends on both their freedom and their resources.
In other words, this is an argument that starts from behavioural economics, but that does not take public policy into some new arena, like nudging. Instead, it generates distinctive, constitutional public policy proposals for what are the traditional instruments of public policy – the tax and the benefit systems. Although the starting point of this argument is a particular view of the evidence from behavioural economics – that we appear, on occasion, not to have well-defined preferences at all – its final, concrete conclusion regarding the simplification of the tax and the benefit systems ought to have a wide appeal. This is because these traditional instruments of public policy intervention have become so refined and targeted as to create a system of such complexity that we are in serious danger of not being able reliably to manipulate them to achieve any first-order public policy end.
This should worry everyone (and not just those who are persuaded by the constitutional direction of the argument here). This is because the worry over the tax and the benefit systems arises from an older and foundational behavioural point, which was one of Simon's early insights that people respond to a complex environment by relying on heuristics for decision making (see Simon, Reference Simon1978). We know that people can be led astray by such heuristics and that, although the use of our capacity for attention can correct mistakes, people will not allocate the scarce resource of attention for this purpose unless the decision environment is sufficiently simple as to suggest that this would be a productive use of the scarce resource. On both counts, if we want to use the tax and the benefit systems for public policy, they need serious simplification.
