Abstract
This paper considers what a tax-benefit system look like if it were explicitly based on a principle of partitioned responsibility, according to which: individuals should not be disadvantaged due to costly eventualities whose occurence is beyond their control (the brute luck principle), but should at the same time be required to pay for the costs of risks that they undertake, voluntarily, as a matter of life-style choice (the costs of choice principle).

Introduction

A multitude of pressures and demands exist on the tax-benefit system. The overwhelming purpose of the tax system is to generate revenue to pay for provisions and services judged advantageous to society. Though generally not framed in this fashion, most of these services are forms of insurance, or risk (and opportunity) spreading, such as in the form of health care (13.5 per cent of general government expenditure), education (12.4 per cent GGE), public order and security services (14.4 per cent GGE), and income provision in times of hardship or social security (34.2 per cent GGE). These four categories alone account for 75 per cent of general government expenditure.

Because of the complexity of the tax-benefit system and the multitude of demands and pressures upon it, we tend to loose sight of the overall rationale of the system. Indeed often we do not see it as a totality at all, but rather as a set of unrelated systems and practices without any overarching principles or objectives. This lack of coherence or agreement on principles is manifested in incoherent political debates and practices, and in inefficient, unjust and paradoxical tensions across the tax-benefit system. The purpose of this paper is to argue that the tax-benefit system, indeed the entire activity of the state, does have a meaningful and integrated role, and that it is important that we seek to make its guiding - but generally implicit - principles explicit. We explore one such principle, which we call the principle of partitioned responsibility.

I. Background pressures: three developments and attendant problems

The need to set out clearly the principles which ought to underlie the tax-benefit system is made especially urgent by three sets of developments.

(1) General confusion about the role of the state. Recent years have seen a general reconstruction and, to some extent, blurring of the division between the state and the private sector. Many areas of activity that had been assumed to have been the sole province of the state have moved either directly into the private sector or into the burgeoning grey area of the quango. This shift, which shows little sign of slowing down, is giving policy-makers on both left and right unsettling feelings of not quite knowing the ground that they are standing on. In particular, it raises the fundamental question, What is the state for?, or What is uniquely the role of the state?

(2) The crisis of the welfare state. The second increasingly familiar problem is that of the crisis of the welfare state. Governments, especially on the right, typically present this as a crisis of affordability, of overgenerous entitlements spiralling out of control as a result of labour market and demographic changes. However, this argument has been rather overstated and is not really the crux of the issue (Hills, 1996). Much more serious, we think, is the crisis of confidence in the effective functioning of the welfare state as manifested in concerns about the emergence of a dependency culture, of the welfare state as a poverty trap, and with the issue of moral hazard (the problem that insurance against a given costly contingency, like unemployment, may induce behavioural change in the insured agent which makes that contingency more likely to occur). Again, many of these concerns are shared across the political spectrum, and it is noteworthy that both left and right are to some extent coming to embrace similar rhetoric and even policies. Witness, for example, the move on both left and right towards the adoption of workfare policies. In essence, problems associated with the area of social security and income support are forcing politicians to ask the practical side of the What is the state for? question, in other words, What should, and should not, be the responsibility of the state?

(3) Developments in private insurance markets. The third set of developments are currently less salient in popular debate, but are become more pressing almost day by day. They relate to the private insurance market, a market which is becoming at once more complex, more sophisticated, and more ruthless. All three of these changes are potentially major sources of concern.

Complexity arises as an issue when individuals come to assess exactly which risks to insure, and exactly what insurance product to buy. It is becoming apparent that many individuals buy products that are unsuitable for them (see problems over pensions and mortgages) and, more generally, that people tend systematically to over-insure high to moderate risks while under-insuring low but potentially catastrophic risks. This in part a problem of time and information - it is extremely difficult and time-consuming for individuals to assess very low probability risks. In these circumstances, what is the responsibility of the state towards those who buy the wrong or no insurance? Would it in fact be more efficient for the state to handle much of this insurance rather than have every individual trying to replicate such activity with its high decision-costs?

The growing sophistication of the insurance market, secondly, raises another set of actual or potential problems. Perhaps the most serious of these concerns whether insurance activity could undermine basic market incentives, and with them, economic growth. This is now becoming a serious concern in the U.S.A. Consider, for example, the derivatives market. If products such as derivatives allow companies to insure against loss of profitability, then the incentives for that company to maximise efficiency and profit may be reduced. This is analogous to the problem posed to policy-makers by welfare dependency, but in this case applies to corporate managers. If the practice becomes widespread, then the incentives to efficiency and innovation will be reduced which will, in turn, impair economic growth.

The growing ruthlessness of the private insurance market is also a development that policy-makers will not be able to ignore in the coming years. Private insurance companies are under pressure to adjust premiums according to true risks, and in some cases these differentials will be very large, or high-risk groups may be excluded from insurance altogether. A large section of the general public already finds many of these differentials and/or exclusions extremely offensive on grounds of equity, e.g., higher differentials in health insurance for gay men (reflecting higher risk of HIV infection). As medical and other scientific knowledge advances, companies will become increasingly adept at assessing degrees of risk at the individual level, and we should expect this to reflect itself in more refined patterns of premium differential and exclusion in the supply of insurance products, patterns that offend even more acutely widespread notions of equity. As said, this problem is already to some extent with us, and though companies sometimes seek to reduce differentials through informal agreements among themselves (such as with insurance for mortgage payments), this practice will be hard to sustain under more intense market pressures. The problem once again raises the question of whether certain forms of insurance may perhaps be better handled outside of the market-place by the state.

II. Enter the state: the principle of partitioned responsibility stated Brute luck, the costs of choice, and the principle of partitioned responsibility

The developments described above all pose the question: what is the proper role of the state in risk protection?

The answer we give to this question, and which we wish to explore in this paper, is as follows: it is the responsibility of the state to spread the costs of risks that may befall individuals but that are beyond their control, while at the same time requiring individuals to pay for the costs of risks that they undertake, voluntarily, as a matter of life-style choice.

We may refer to this as the principle of partitioned responsibility. It is really a combination of two subordinate principles: firstly, the brute luck principle, which says that people should not be disadvantaged as a result of factors beyond their control, matters of bad brute luck; and secondly, the costs of choice principle, which says that people should meet the specific costs associated with their life-style choices, and not displace those costs onto others. We propose that this principle be made more explicit, and that, in accordance with this principle, calculations of risks (population estimates of net costs/savings from given behaviours or activities) be used to calculate marginal differentials in tax-benefits.

We contend that the state ought to adopt this as a guiding principle because it is both just and (in general) efficient. As political philosophers such as Ronald Dworkin and G.A. Cohen have argued, it is unjust for individuals to suffer significant disadvantage relative to others through no fault of their own, a point of view which is shared amongst the general citizenry to the extent that they find some of the aforementioned patterns of premium differentiation and exclusion thrown up in private insurance markets inequitable (see especially Cohen, 1989). On the other hand, it is just that individuals meet any significant costs associated with their life-style choices, rather than being able to displace these costs onto others (see especially Dworkin, 1981). And, aside from this consideration of justice, it will also in general be more allocatively efficient if individuals make choices in full awareness of the respective costs of the options before them - if they make choices, that is, knowing the true underlying pattern of resource scarcities.

A third possible advantage of applying the principle is that, in some contexts, it may serve to enhance individual autonomy. There are some social costs which society currently tries to protect itself against through paternalistic legislation (e.g., we seek to protect ourselves from the health-care costs of road accidents by requiring people to wear seat-belts and crash helmets). A tax-benefit system organized in accordance with the principle of partitioned responsibility might be able, in some contexts, to protect us against meeting such costs by targeting the burden of meeting them more precisely on the shoulders of those who (choose to) incur them. This then removes one existing rationale for (some) paternalistic legislation.

We offer the principle of partitioned responsibility, therefore, as a principle that promotes justice, allocative efficiency, and, in some contexts, individual autonomy. It is important to emphasize at this early stage, however, that we do not offer it dogmatically, but in a spirit of suggestive inquiry. We believe that the basic moral intuitions underpinning the principle are strong and attractive ones. But, nevertheless, we may find on closer scrutiny that some of the policy implications of the principle are unattractive, or attractive only under very special, ideal circumstances that we have no chance of even approximating in the real world. We might then decide to limit the scope of application of the principle. It is only by exploring these policy implications in the first place, however, that we will come to discover to what extent this is the case, and so this is the task we set ourselves for this paper.

The state as insurance agent

Underlying this proposed principle, it can be argued, lies a particular conception of the state: the state as akin to a giant, comprehensive insurance company. Individuals - citizens - pay premiums in the form of taxes, and in return receive compensation or treatment in the event of misfortune. In principle, if people want better or fuller cover, then they can arrange this (through the political process) through an agreement to pay higher taxation. Like a private insurance company, moreover, the state as insurance company faces the problem of moral hazard: the problem that insurance against a given contingency induces behavioural changes in the insured party that increases the probability of the insured contingency occuring.

Private companies tend to handle moral hazard problems in two ways: (a) they seek co-insurance (the insuree has to pay some proportion of the costs if the contingency occurs, such as in the form of an excess); (b) they simply avoid insuring things that are heavily dependent on human agency (e.g., pregnancy-related costs). They also go to considerable trouble to ascertain if insurees are culpable for contingencies which eventuate, asking such questions as, Did the fire start by accident or was it arson? Insofar as the state handles moral hazard, it tends to use broadly similar strategies, for example, by setting benefits at low levels such that individuals share the costs of the contingency (co-insurance), by avoiding covering certain risks at all (e.g., loss of property), and by holding occasional enquiries into responsibility or cause.

We would argue, however, that the norms appropriate to the state as insurer differ in certain fundamental ways from those which govern the behaviour of private insurance companies. Consider, for example, the issue of insuring an individual against the possibility of developing an expensive medical condition. The commercial insurance company must simply calculate as accurately as possible the probability that an individual will develop the illness and the likely cost of the subsequent treatment. If it is known that the disease is mainly genetically determined and inherited, and it is known that both the parents had the disease (or even that the individual is known to carry the gene associated with the disease), then the company must reflect this risk in its premium. For if it does not, then its net premiums will have to be increased, and other companies may steal its custom by using the adjustment and offering lower premiums to those known to be at less risk. The private insurance company simply cannot afford to be equitable, and adjust its premium to reflect the fact that certain risk factors may be beyond the individuals control, e.g., that someone is just unfortunately born with a congenital illness that we can predict will require costly medical attention. In calculating the individuals risk and her appropriate premium, the question of whether a given risk factor is under the individuals control will thus be relevant only secondarily insofar as insurance of the contingency raises problems of moral hazard.

But now let us consider the state in the same situation. Acting on the principle of partitioned responsibility, which conjoins the brute luck and costs of choice principles, the states proper primary concern is not the predictability of a given risk, but the degree of personal responsibility for its eventuation. In our foregoing medical example, it is clear that the elevated risk is not the fault of the individual as there is nothing she can do to alter her own genetic make-up. Hence the state, unlike the private insurance company, and for the moment assuming away inequalities in earnings power between individuals, should apply flat premiums to this risk. Fortunately, the state is also in a unique position of being able to co-opt the entire population or community to share the risk of costs from genetic illness, and is therefore able to spread the cost without danger of an undercutting of premiums.

Now this principle - that the state should spread the cost of risks according to personal responsibility rather than predicability - can be applied quite widely to the state's activities. In the case of random or field-determined risks - of ill-health, unemployment, disability, and so on - the state simply spreads costs across the population. We can then all rest a lot easier at night in the knowledge that, having paid our taxes (premiums), we are all protected against the excesses of misfortune. This is a massive benefit that we receive from our fellow citizens - from each other - in return for our contributions.

In the case of these contingencies the state, following the brute luck principle, should apply flat premiums - at least as long as we continue to assume away inequalities in earnings potentials between individuals (we will relax this assumption later on and consider how the state ought ideally to respond to these inequalities).

But, leaving brute luck misfortune aside, how should the state treat intentional or chosen activities that involve higher risks of net costs - costs that the citizenry at large may be asked to cover? The answer comes in two stages. First, as a community we must decide whether we are prepared simply to leave individuals to cope with the eventuality by themselves. We can simply decide that we will not cover a risk at all, for example, the loss of personal effects (though an individual will remain free to purchase private co-insurance for this risk). Second, if we decide that the state should cover a risk, then we must decide at what level to set the premiums. Following the costs of choice principle, and acting in the interests of both efficiency and justice, we hold that the state should levy higher premiums or specific taxes in order to cover elevated costs and risks resulting from individuals life-style choices. The individual is then left free to decide whether to pay the higher premiums and continue with the activity or change his behaviour. If the premium has been correctly set, then the community should be essentially neutral towards the individual's subsequent choice.

In the next sections, we shall explore some of the apparent consequences of applying the brute luck and costs of choice principles to tax-benefit arrangements in the manner we have suggested: what, we shall ask, would a tax-benefit system based on the principle of partitioned responsibility look like?

III. Applying the principle

The example of green taxes

The case of environmental or green taxes provides a useful starting point for examining the practical implications of basing a tax-benefit system on the principle of partitioned responsibility. This is partly because they represent a relatively simple application of the principle - more especially of the costs of choice component of the principle - and also because the idea of green taxes is already familiar to most people. However, we will go on to show that green taxes are just one example of a much larger family of tax-benefits based on the partitioned responsibility principle.

The basic idea behind green taxes is simple: place higher taxes on environmental bads and lower taxes (or subsidies) on environmental goods. In this way, the prices of goods whose production and/or consumption involves environmental goods or bads can be adjusted so that they equate to the true costs and benefits caused by them (Taylor and Taylor, 1996). The price of a fuel, for example, should reflect not only the cost to the supplier of extraction, but also the cost to the community of the damage done by pollution from its burning to health, crops, buildings and damaged ecosystems. Through the tax, the cost of the likely externalities are charged to the fuel-user in the form of an additional tax in order to recompense the community for the damage being done. This provides the individual consumer with a meaningful price signal (information) and consequent incentive to reduce his harmful behaviour. The individual is then free to choose whether to buy the environmental good/bad or whether, given the additional cost, to change to some other activity or pattern of consumption (e.g., to cycle more and drive a car less). Presuming the individual does have a real choice, then this tax instrument is both more just and more efficient than a blanket regulatory solution to the problem (e.g., banning the use of cars). It is more just in that individuals pay in proportion as their behaviour damages the environment as opposed to some individuals disproportionately bearing the costs of the damage done by others. It is more efficient because it allows individuals in different situations to make optimum adjustments depending on the costs and benefits of consumption in their particular circumstances. Moreover, precisely because it allows people to decide just how much of the bad/good they wish to consume, given a more accurate reflection of its true price, the policy also respects individual autonomy in a way that blanket regulatory solutions do not.

Some may question whether green taxes are really just because they are potentially regressive in their impact, taking a larger proportion of the incomes of the poor (Turner et al, 1994). This is a serious and legitimate concern, and one which arises with many of the differential tax premiums we will discuss below, such as sales taxes on tobacco products. It would not arise as a concern if, as we have so far assumed, individuals had equal earnings potentials. But they dont, and, in the absence of an ideal system of compensation for these inequalities (which we will discuss below), policy-makers will have to find ways of compensating low-income households - or, more exactly, those who through no fault of their own have low earnings potential - for these kinds of taxes. The funds raised from green taxes could be used, for example, to finance cuts in taxes on other goods. The ideal policy, however, which would arguably apply even if we have otherwise controlled for unequal earnings potentials - is to use the revenues specifically to finance payment of a cash grant to each person: an ecogrant.

In his famous article, Equality of Resources, the philosopher Ronald Dworkin imagines a group of people ship-wrecked on a desert island (Dworkin, 1981). The group find the island cluttered with resources, and Dworkin poses the question: how can they organize the division of these resources amongst themselves so as to give equal concern and respect to each person on the island? His answer is that the islands resources should be sold off to the islanders in an auction in which each islander has an equal endowment of purchasing power (clamshells). This is effectively the same as giving each person a cash grant equal to a per capita share of the market value of these external, manna-like resources (Van Parijs, 1992).

The ecogrant idea can be seen as capturing a similar intuition about justice. It is a way of giving each person some equal initial endowment of environmental use-space. Those who then make relatively heavy use of such space, e.g., use relatively large amounts of fuel, will pay more in tax than they receive in grant, and this is perfectly just, for it means that they will be paying for the special environmental costs associated with their life-style choices, rather than displacing those costs onto others. To use more than their fair initial endowment of environmental use-space, they have to make a net payment to the rest of the community for it.

This assumes, of course, that individuals do indeed have room to make genuine choices in this area. To the extent that they do not, then, following our principle of partitioned responsibility, further refinements would have to be made to the tax compensation package to prevent some individuals being inequitably burdended by the system of green taxes. Individuals who are bound to have to consume more because of factors that are genuinely out of their control, such as for historical or geographical reasons, would be compensated at a higher rate. (They would, in effect, be given a higher initial endowment of environmental use-space.) It might be argued, for example, that individuals who live in colder parts of the country (such as Scotland) have to use more fuel simply because of the climate, a field variable over which they have no control. Of course, it could be argued that they do have some degree of control in that, in principle, they could move to a warmer area. But we might nonetheless judge that at least some proportion of the average differential in fuel use is not their responsibility and that a corresponding proportion of the average differential in fuel tax paid should therefore be refunded, e.g., through a higher ecogrant. A similar, perhaps more powerful argument might be made for age-related differentials in compensation.

In summary, the main point about green taxes is that, providing individuals have genuine choices, they are just, efficient for the community, and respectful of individual autonomy in a way that regulatory approaches are not. Yet even in this relatively simple example, it can be seen that their construction, and the details of the compensation packages required to make them work fairly, draw heavily on informed (and often controversial) judgements about the relative explanatory importance of personal responsibility and brute luck - about the extent to which an individual can be held responsible for their actions versus the extent to which their behaviour and prospects are constrained by the situation or the field. We will now examine some wider applications and implications of the principle of partitioned responsibility.

The extension of the concept to other risks and behaviours: a general formula?

Many types of behaviour that individuals engage in put them at risk of needing expensive treatment or intervention by the community. Consider, for example, smoking. It is estimated that around one-third of cancers are caused by smoking - including 85 per cent of lung cancer - and around one-fifth of heart disease. Indeed, smoking is often described as the number one preventable cause of death (Kaplan et al, 1992). Why should the community care about this? After all, isnt this an area where mature, adult individuals should be left free to choose whether or not they want to take the risk?

There are two issues that make smoking of wider concern. The first is the cost imposed on the wider community as a consequence of smoking. The costs of treating smoking-related diseases are extremely high (CHD and cancers are the leading causes of death in the industrialised West) and, given the limited resources available to the NHS, considerable controversy consequently surrounds the repeated treatment of smoking-related diseases. There are also various externalities associated with the loss of productive life as a result of the sickness and early death of smokers; with the direct harm done to others in the form of passive smoking; and even with the physical measures designed to separate smokers from others (e.g., the provision of separate smoking areas in workplaces). It should be recognised, however, that there may also be some savings to the community as the result of smoking, the most obvious and brutal of which are the savings in pensions from those who die young.

The second issue is the extent to which individuals are coopted or coerced into smoking - the extent to which tobacco companies draw young people into smoking, deliberately make products addictive, so undermining individual agency in the decision to smoke (and thus sharing some responsibility for the individuals behaviour). This is, of course, an issue that is increasingly finding its way into the courts as diseased smokers bring law-suits against tobacco companies. But it also has implications for public policy insofar as we want to protect people from significant disadvantages attributable to factors beyond their control.

In principle, it should be possible to calculate the net costs of smoking to the community quite accurately. This total net cost can then be divided by the number of cigarettes consumed over a given period to give the average net cost to the community of each cigarette smoked. Insofar as the decision to smoke is taken by the individual smoker (subject to the second issue described above), it is reasonable that the community should charge the smoker a premium for this risk equal to this average net cost. Once again, this is a just policy in that it makes individuals pay for the costs associated with their specific life-style choices, rather than displacing those costs onto others. And it is an efficiency-promoting policy, because it provides the individual with a clearer price signal concerning the costs of their behaviour, and an incentive to adjust this behaviour so as to achieve a fall in the total net costs to the communty from smoking.

This mechanism also has a positive political and administrative spin-off in that, provided the costs have been correctly estimated, it implies that no risk selection on the basis of smoking behaviour has to occur - or should occur - at the point of service provision: a population of smokers who need operations for smoking-related diseases have already paid for them through the years of premiums paid - the cost of each cigarette can be thought of as a down-payment for this eventual probable need. We can thus get away from the controversy involved in making difficult decisions about the relative deservingness of smokers, non-smokers, ex-smokers, etc., at the point of service provision.

Of course, we do already charge smokers an extra premium, though curiously, we do not appear to justify tobacco taxes in these terms. Furthermore, we have not actually gone through the exercise of calculating true net costs; instead premiums appear to be plucked out of the air according to no rationale other than what the Chancellor thinks that he can get away with. Consequently, we do not really know whether the community is currently under- or over-charging smokers: it could be that the community is continuing to subsidise smoking, or it could be the other way around.

One convenient thing about the example of smoking is that is quite easy to set the premium fairly precisely according to the risk that the individual chooses to expose himself to. The risk is thought to be proportional to the amount smoked, and as the levy is charged per pack of cigarettes the premium paid should vary proportionally with the risk. Of course, we can see that the match is probably rather imperfect - for example, the premium should also vary systematically according to the level of tar (if this is the true source of risk) rather than simply the number of cigarettes in the pack. Nonetheless, the premium does vary fairly systematically with the risk. However, it will not always be possible to proportion premiums to risk very precisely, and sometimes second (or third, or...) best solutions must be sought.

Consider, for example, the risks and costs associated with hill-walking and climbing. Every year there are news reports of hill-walkers and climbers going missing and of the extensive searches that follow. These reports are frequently followed by comments about the high cost of these searches both in terms of the time of volunteers and of the professional emergency services, and subsequently by complaints about the actions of careless or even reckless walkers and climbers. How should the community, and therefore state, handle this risk behaviour and the costs associated with it?

Ideally, the community would charge hill-walkers and climbers a small premium for each time they spent time in the mountains, and these premiums would be varied according to the known risks being taken (e.g., higher premiums would be charged on poor visibility days and to inexperienced climbers, assuming experience to be associated with lower risk). However, this is clearly impossible in practical terms, and so a second best solution would have to be developed.

The possibility of mandatory insurance for walkers and climbers has occasionally been suggested, but this might prove nearly as difficult to enforce as direct charging. A next best solution in this case might be to charge a premium on equipment near-universally used by walkers and climbers. Hence walking boots, climbing boots and climbing equipment could be subject to a modest surcharge equated to the average net cost of rescue operations. This would also have the advantage of approximately charging higher use, and therefore risk, individuals higher premiums. Of course, it is not perfect as a scheme - for example, some high risk users might avoid buying suitable equipment - but it would certainly be a major advance on the current situation where rescue operations may be subsidised by the entire community.

We must acknowledge, however, that there may be some cases where the approach is impractible - where not even a second or third best policy solution is available. Consider the health risks associated with unprotected sex, for instance. Here, as in the case of smoking, there is a clear risk of the individual acquiring a need for expensive medical treatment as a result of life-style choices. But it is hard to see what commodity we could place a tax on to function as a premium matching the special costs potentially associated with this life-style choice. In this case, we do therefore seem back in the unwanted situation of trying to morally discriminate against people who acquire diseases through unprotected sex at the point of service delivery, or else simply letting them displace the costs of their life-style choices onto others.

Where the principle of partitioned responsibility can be applied to set tax-premiums, however, then it is properly applied in accordance with the following general formula:

Differential tax/benefit = r . p . C

where r = personal responsibility, the fractional causality or responsibility attributable to the individual (ranges from 0 to 1); p = probability of outcome associated with target action (or proxy; ranges from 0 to 1); and C = the proportion of the total net cost/saving to the community associated with the risk behaviour (£s).

Notwithstanding the foregoing precautionary point about cases where application of the principle may be impracticible, the principle can be applied very widely. Other relatively clear examples of premiums that can be charged on behaviours significantly under the control of individuals but with costs born by the community include consumption of alcohol (premium charged on product), reckless driving (premium as surcharge on insurance of high risk drivers) and on certain dangerous pursuits (premiums charged on, for example, motorbikes, rollerblades, hangliders etc.). In each case, the premiums would be calculated to equate to the probability of extra net cost to the community of the risk behaviour, multiplied by the fraction of the responsibility or hazard that is judged to be under the control of the individual (on average).

We should also recognise that there are certain classes of behaviour that are associated with lowered risks of costs to the community. For example, it is thought that diet may account for up to 40 per cent of the variance in the occurrence of cancers, and that more specifically, some food-types may be associated with lower risks (such as fresh fruit and vegetables). If the eating of such foods can be shown to be associated with net savings to the community, then the formula above would indicate that these foods should attract a marginal benefit or reduction in base tax rate rather than a surcharge or premium.

Finally, we should keep in mind how this approach can be used to handle types of risky behaviour that we currently attempt to force people to avoid through laws and regulation. Some drivers and passengers have a strong dislike of wearing seat-belts, and some bikers wish to ride their bikes without helmets. At present, in part because we are wary of the expensive medical care which will be needed if they get into accidents, we legally require individuals to wear seat-belts and motorcycle helmets. But the community might alternatively allow these high-risk behaviours subject to a marginal increase in tax-premiums or, equivalently, upon payment for specific exemption licenses. (One might, for example, have an option between buying two driving car-driving licenses, a standard one, and a more expensive one that entitles its holder to drive without wearing a seat-belt.) These tax-premiums would be set to equate to the marginal increase in risk of expenditure associated with the behaviour (i.e., the higher probability of expensive medical care and benefits for the injured individual and dependents). Of course, one might decide against this option on the basis of difficulty of enforcement, though this problem is one that already applies to many examples of common practice (e.g., normal car insurance).

Specifying the personal responsibility term

Once the framework proposed here is adopted, estimating the personal responsibility term becomes an important and explicit activity of the state. Of course, it is already important to the state and community now, but it is concealed in the implicit judgements that lie behind policies. How can we make these calculations more systematic? And, given the inevitably controversial nature of the judgements involved, how can we ensure that operational specifications of the personal responsibility term have legitimacy amongst the wider citizenry?

To begin with, we note that considerable work exists within psychology, and to some extent elsewhere, in the calculation of personal responsibility terms, though it is not normally expressed in this form. For example, there are many studies which attempt to partition variance in behaviour and outcomes into person, situation and person-situation interaction terms. Similarly, work in medicine that attempts to partition the aetiology of illnesses into genetic, social and lifestyle factors; work in sociology that attempts to partition social outcomes according to social background versus individual factors; and work in law that attempts to partition negligence. All are effectively trying to partition causes. This work illustrates that partitioning is possible, at least to some extent, but also shows that judgements about the appropriate partition are conditional on circumstances and may change over time. This is perfectly acceptable, but does imply that responsibility partitioning is something that must be regularly reviewed.

In practice, a degree of subjective judgement is bound to enter into the calculation of responsibility partitioning. The best that we would be able to do would be to empirically derive responsibility estimates in limited range of important examples, and then use comparisons with these to judge them in other cases. In so doing, we should also attend to the detailed literature on attribution theory in psychology, and remember to correct for the predictable and systematic errors that are known to occur. For example, people show a strong tendency to overestimate the importance of person factors in the occurrence of other people's behaviour (the fundamental attribution error), overestimate the importance of field factors in their own behaviour, especially when it is not seen as socially desirable (the actor-observer bias) and overrate the representativeness of their own knowledge (the false-consensus effect).

The problem of subjectivity also suggests, however, that the process of specifying and reviewing personal responsibility terms should not be confined to panels of (supposedly) enlightened experts. The wider citizenry who will have to live with these decisions must also be brought into the process. To this end, institutions such as citizens juries, deliberative and consultative opinion-polls, might all be used to bring citizens and experts together, experts providing evidence and pointing to ranges of reasonable disagreement, and citizens deciding, or at least indicating, what should be done given the existing range of reasonable disagreement.

IV. Extending the principle: the cases of employment choice and natural ability

Most of the examples described so far relate quite closely to taxes and benefits that already exist or are at least sometimes discussed. We would like now to consider some less familiar applications of the principle of partitioned responsibility - though one of them, concerning inequalities in natural ability and earnings potentials, is arguably a precondition for the legitimacy of applying the principle on any large-scale at all.

We begin with the case of employment choices. To some extent, the employment outcomes of individuals are attributable to factors outside of their control. The opportunities presented by the young person's family and local context may constrain and distort many of the options available. Parents and communities, whether intentionally or otherwise, can have a strong influence on whether a person becomes a lawyer, say, or a farm worker. One recent study showed how law students were 10 to 20 times more likely to have a parent who was a lawyer than would be expected by chance alone (Halpern, 1994). Choices are also constrained by abilities, and to the extent that these are determined by genes and social advantage through birth, these are also significantly field factors.

Nonetheless, to some extent individuals also genuinely choose their occupations. The next question is, is there a differential risk of cost to the community associated with these choices? The answer must be affirmative. Moreover, these differences are to a large extent known. For example, some forms of employment are associated with very low rates of unemployment, while others are known to be associated with very much higher rates: actuaries are hardly ever unemployed, but actors may be very frequently so. And yet the premium we officially pay in return for public coverage of the contingency of being unemployed does not seem to reflect these differences. Actors are not charged more by the state for unemployment insurance than actuaries, and to this extent those who choose to become actors are displacing some of the costs of their employment choice onto those who choose to become actuaries. On the face of it, our approach suggests that, in order to avoid this unfair displacement of cost, forms of employment reliably associated with low risks of unemployment should pay a lower premium than those at high risks: actors (when employed) should pay higher premiums than actuaries.

Imagine, however, that we proceeded to institute a system of public unemployment insurance in which premiums were linked to the relative degrees of risk of unemployment across occupations. (Put aside the question of whether we could in fact ascribe a particular occupation to everyone.) What would be the result? The result would be that many unskilled workers would have to pay relatively high insurance premiums because the risk of unemployment in many low-skilled occupations is in fact relatively high. But these occupations are also relatively poorly paid. So, once again, in trying to apply our principle we would be taxing the poorest hardest. This would not matter if those in these occupations were in them by choice - if, say, they were all well-qualified poet bohemians just doing low-skilled work because the nature of the work gives them more time to focus on their art than being a merchant banker or a lawyer would. But while a few of those in low-skill occupations may fall into this category of the non-needy bohemian (Arneson, 1990), the majority probably wont. They will be there because they are unable to get other kinds of jobs. Charging them relatively high premiums for unemployment insurance would be to compound their bad brute luck in just the same way as a private insurance company does when it charges someone with a genetically-determined higher probability of contracting some illness a correspondingly higher premium to cover the risk of the costs of contracting this illness.

Thus, once we allow for inequalities in natural ability (and in other field factors that affect employment choice), this complicates the picture very considerably. Admittedly, some of the required adjustment can be effected in our proposed approach by focusing more closely on the r term in making the calculation of the appropriate tax-premium. If the raw, unadjusted tax-premium is determined simply by the probability of being unemployed, conditional on being in a given occupation, mutliplied by the cost of unemployment, then the adjusted tax-premium would have to be multiplied again by a personal responsibility term, reflecting the probability of an individual being in that occupation by choice, rather than through bad brute luck in the natural and/or social lotteries. If, as we have speculated, the vast majority of those working in low-skilled occupations do so because they lack the natural ability, or otherwise lack the opportunity, to move into better occupations, then this r term will be very low (closer to 0 than 1), and the tax-premium paid by unskilled workers for their unemployment insurance will not then necessarily be higher than that paid by skilled workers.

However, the usual process of adjusting the tax-premium through the r term will not entirely meet our anxieties concerning background inequalities in natural ability. Imagine that we have two groups of workers, skilled and unskilled. The probability of being unemployed as a skilled worker is, say, 0.1, and the same probability for an unskilled worker, 0.2. The respective r terms, however, are 0.4 and 0.2, and so, for a given level of unemployment benefit, members of each group pay the same premium. However, the skilled can also earn twice as much in a given working week than the unskilled. Thus, just as in the case of the green tax, a flat rate tax-premium will still be regressive in effect, swallowing a higher proportion of unskilled incomes than of the skilled. And, on the assumption that members of the unskilled are in this group primarily through limited natural ability, and/or other field factors over which they exert no control, we must conclude that this is unfair. In contravention of the brute luck principle set out in Part I, the unskilled still end up worse off than the skilled, on average, through no fault of their own.

At the end of the day, if we wish to address these inequities, then we simply have to deploy the approach we have been outlining directly to differences in natural ability - or, more exactly, to differences in earnings potentials as determined by field factors which centrally include natural ability. Many of the distributive anxieties about things like green taxes would be dissipated if we could tackle this one issue - the issue of how to eliminate the rent of ability (Shaw, 1937) - in a convincing way.

The contingency against which we are here trying to arrange insurance protection is the contingency of having, through no fault of ones own, an earnings potential (when employed) that falls beneath some morally pertinent threshold level. (Note: earnings potential here refers to the maximum a person can earn over some specified period while making reasonable use of her exogenously-given endowment of ability.) If our concern is to correct for brute luck inequality in its entirety, then average earnings potential looks like a natural choice for this threshold level. We would then transfer resources from those with above average earnings potential to those with below average earnings potential so as to give everybody equal earnings potential (Akerlof, 1978, Barry, 1995). Thus, if average earnings potential in the community is, say, $40,000 for a standard working year, and As earnings potential is $25,000, while Bs is $80,000, then we need to redistribute between the As and the Bs so as to ensure that both end up with $40,000 if they work a standard working year.

One possibility, often floated by economists, is a straightforward lump-sum tax or subsidy equal to the difference between actual earnings potential and average earnings potential. Thus, B would be subject to a lump-sum tax of $40,000 at the start of the working year, while A would get a lump-sum subsidy of $15,000. Such an approach may have its merits, but it would arguably lead to some inequities too. For instance, in our example, B would have to work well over half (11/16 of) a standard working year to achieve a level of after-tax income, $15,000, that A can achieve by doing no work at all. This does not look like a case of A being compensated for a brute luck disadvantage relative to B; rather it looks like a case of A living a leisurely life-style at the expense of poor B (violating our costs of choice principle). To overcome this objection, we might pay out the subsidy to A not as an unconditional windfall, but rather, as an earnings subsidy (akin in some respects to the USs Earned Income Tax Credit system). Thus, A would ideally receive a subsidy from the state of 60c for every dollar he earns, while B pays a special tax of 50c on every dollar he earns. Then, given the maximum respective wage rates they can each reasonably be expected to command, they will each have to work the same number of hours to achieve any level of after-tax income (for a related idea, see Chapman, 1996). What we are doing, in effect, is calculating the societally average maximum wage rate, reflecting the societally average ability endowment, and then converting each persons maximum market wage rate into an after-tax/subsidy maximum wage rate that equals this societally average maximum wage rate. The tax rate applicable to each person depends on her estimated earnings potential - the maximum wage she can reasonably be expected to command given her exogenously-given endowments of ability.

Talk of such a radically redistributive scheme will inevitably raise doubts in some minds about efficiency. Would such a system detrimentally affect incentives to work and/or to invest in human capital? As far as work incentives are concerned, the effect of the earnings subsidy scheme is to give everybody the same income-leisure budget constraint. Faced with this constraint, some people will work more than they would in the absence of taxation, others will work more. The aggregate effect on labour supply is theoretically indeterminate.

Nor should the scheme destroy incentives to invest in human capital, i.e., for people to cultivate their productive abilities. The subsidy/tax rate ascribed to the individual depends on an estimate of the maximum wage she can be expected to earn given a reasonable deployment of her natural abilities, and, in making this estimate, we are to take into account what she could potentially earn after training - that is, after undertaking the level and kind of training that are reasonable given her natural abilities. Concretely: if we think that someone has the natural ability to get a university education, then we estimate her earnings potential by reference to what university educated people can earn. She can, of course, choose to forego that education, but this will not affect her (relatively high) estimated earnings potential or the (relatively high) earnings tax she will have to pay. Thus, if she consequently earns well below her potential, she will simply end up with a relatively low after-tax income. If she wishes to avoid that outcome, she will have to undertake the educational investment required to bring her earnings performance in line with her earnings potential. While the earnings subsidy scheme suppresses differences in final incomes across the ability range, we thus see that it retains differences in final income between differentially educated people within a given ability range, and so does not deprive people within any ability range of a material incentive to get the most education and training they can. Indeed, quite to the contrary - the scheme actually condemns people to having quite low final incomes unless they undertake as much education and training as they can make use of (as much as raises their earning power to its reasonable maximum). The approach clearly does presuppose, however, that individuals do have access to the education and training needed to realize their full earnings potential, indeed, that the costs of this education and training are spread across the society as a whole - clearly a tall-order in policy terms.

At the lower end of the ability scale, we have become more than ready in recent years to insist that individuals who are capable of working, but who make no effort to work, should be subject to a lower benefit (higher charge) from the community. This is the main impetus between many workfare type policies. The system of earnings subsidies and taxes proposed here simply extends this logic to those higher up the ability scale. It presents those higher up this scale with the same just and efficiency-promoting choice that workfare policies typically present to those in the bottom end of the ability range: make reasonable productive use of your natural endowments or suffer a lower (lower then average) income if you do not. If we really are really determined to apply the principle to those at the lower end of the ability scale, elementary notions of fairness suggest that we try to extend it to all.

Another obvious objection to this approach is the difficulty of obtaining unbiased estimates of individuals respective natural ability-based earnings potentials. A first approximation might be sought in the use of IQ tests. However, apart from the statistical limitations of the reliability and validity of these tests, it might be thought that there would be obvious incentives for people to fake low scores so as to give low estimates of their ability and reduce their future tax liabilities.

We should be wary, however, of exaggerating these difficulties. If an individual fakes a low score, and his pre-tax earnings performance then systematically and significantly exceeds the earnings potential ascribed to him on the basis of the test (a fact which it should not be difficult for tax authorities to acsertain), this would surely be a reason to reevaluate the earnings potential and earnings subsidy/tax rate originally ascribed to him. Admittedly, under the lump-sum transfer arrangement, an individual could fake a low score, and then enjoy a life of greater leisure at lower levels of income by partly living off the lump-sum subsidy which he would be given as someone with an apparently low earnings potential. So long as he stayed at low levels of earnings and income, there would be no observable discrepancy between his actual and estimated earnings potential, and so he could continue with the subterfuge indefinitely. But under the alternative earnings subsidy arrangement we have outlined this possibility of living a life of subsidized, relatively low-income idleness is not available - the subsidy is conditional on work - and so the incentive to fake a low score in the ability test is reduced.

Another problem with the scheme, however, is that, in its simplest form, it presumes that people can get jobs that fully reflect their earnings potential (Barry, 1995). It presumes full employment of a very demanding kind, under which not only can everyone get a job, but also get a job that makes the most of their marketable abilities. Clearly, this is a tough condition to satisfy. One response to the problem would be to continually adjust individuals respective earnings tax/subsidy rates to reflect job scarcities. How, more exactly, might this be done? The ideal policy would be to supplement unemployment insurance with a system of underemployment insurance. Those, who through no fault of their own, leave unemployment for a job that pays below their estimated earnings potential would then receive a benefit, again in the form of an earnings subsidy, equal to the difference between their current wage-rate and the maximum wage-rate they could reasonably be expected to command (on which the estimate of their earnings potential is based). Simply deduct this subsidy rate from the standard earnings tax rate which applies when they are able to get a job that reflects their full earnings potential, and you have the adjusted earnings tax rate applicable to them in their present situation of involuntary underemployment. Of course, those who are voluntarily underemployed would not be eligible for the subsidy, and would still be taxed at the standard rate.

Finally, we turn to a problem with the proposed earnings subsisy/tax system that perhaps most readers will have spotted from the start, a problem which applies even at the ideal level, before we even get to the point of considering whether a state could ever have the kind and level of information about true ability endowments and job scarcities which the approach presupposes. This problem is that the proposed earnings subsidy/tax system, in its pure incarnation, is not necessarily self-financing in the way that the alternative lump-sum transfer system (which we rejected as inequitable) is. In the lump-sum system, we impose lump-sum taxes on one group, and, given that they are calculated by reference to average earnings potential, these will necessarily sum to the total of lump-sum subsidies the state pays to the second group. Even if we assume that there is no underemployment, under the earnings subsidy/tax scheme, the sum of total tax liabilities and subsidy entitlements will only equate if the two groups, optimizing against the income-leisure budget constraint that is now common to them, choose to work a number of hours that cause these two sums fortuitously to balance out against each other. If individuals income-leisure preferences are identical, then this outcome is assured (for everyone will work the same number of hours). But if their preferences differ, this outcome cant be guaranteed. In this case, the community would either have to insist that people worked the same number of hours, a policy that would obviously violate individual autonomy; or, it would have to adjust the earnings subsidy and tax rates away from their ethically optimal levels until the scheme was brought into a stable self-financing equilibrium. Further adjustments would have to be made to allow for the impact on tax yields and subsidy entitlements of underemployment. Just how far away from the ethically optimal subsidy/tax rates we have to move to achieve a self-financing equilibrium is an open, highly contextual matter.

Doubtless there are many other objections that could be raised against this ideal policy response to the problem of natural ability-based inequality in earnings potentials. The informational demands it makes of the tax-benefit authorities are clearly huge, and perhaps insurmountable. We declare ourselves currently agnostic on the question of whether the proposed scheme is just too utopian for it even to be worth thinking about how we might approximate it in practice. What we must recognize, however, is that to the extent that we do fail to address this fundamental background source of inequality, then application of the principle of partitioned responsibility in other areas of the tax system, e.g., in the form of green taxes, or tax-premiums for unemployment insurance differentiated according to occupationally-specific risks of unemployment, must be moderated - or else, contrary to the very demands of this principle, we will end up unintentionally making some people worse off than others through no fault of their own. Moderation here would entail either not applying the principle at all in some contexts, e.g., not imposing this or that green tax because of its potentially regressive effects, or else finding case-by-case ways to compensate those on lower incomes for the potentially regressive effects of such taxes.

V. Conclusion

The principle of partitioned responsibility can be widely applied and has wide-ranging implications in tax-benefit policy. It provides a detailed rationale and method of calculation for the multitude of tax-benefits that tend currently to be seen as isolated and unrelated issues. However, in issues concerned with achievement, its applicability is limited by the difficulty of controlling for natural ability.

Perhaps most importantly, however, the application of the principle makes explicit the key responsibilities and role of the state as a discriminating spreader of risks, and it focuses attention on the systematic analysis and influence of acts of individual agency that affect the whole community. In so doing, it also exposes the marked inconsistencies and imperfections of much of the current tax-benefit system and, indeed, of the policies proposed by the major political parties.

References

Akerlof, G., The Economics of Tagging as Applied to the Optimal Income Tax, Welfare Programs, and Manpower Planning, reprinted in An Economic Theorists Book of Tales, 1984.

Arneson, R., Is Work Special? Justice and the Distribution of Employment, American Political Science Review 84, 1990.

Barry, B., Arguments for Basic Income, paper presented to the L.S.E. Rational Choice Group, November 1995.

Chapman, D., Reform of the Tax and Benefit System to Reduce Unemployment, paper presented at the 6th Congress of the Basic Income European Network (BIEN), September 1996.

Cohen, G.A., On the Currency of Egalitarian Justice, Ethics 99, 1989.

Dworkin, R., What is Equality? Part 2: Equality of Resources, Philosophy and Public Affairs 10, 1981.

Hills, J., Tax policy: are there still choices?, in D. Halpern et al, eds., Options for Britain, 1996.

Shaw, G.B., The Intelligent Womans Guide to Socialism, Capitalism, Sovietism and Fascism, 1937 (2nd edition).

Taylor, S. and Taylor, B., Transport and the environment, in D. Halpern et al, eds., Options for Britain, 1996.

Turner, R., Pearce, D. and Bateman, I., Environmental Economics: An Elementary Introduction, 1994.

Van Parijs, P. Arguments for Basic Income, Introduction to P. Van Parijs, ed., Arguing for Basic Income, 1992.