International political events of the last fifteen years indicate deep popular skepticism about the egalitarianism of the welfare state: the latest dramatic example, for Americans, may be the Republican sweep in the recent congressional elections. The reasons for this skepticism are complex, but they are partly philosophical. Many people associate egalitarianism, and the policies of the welfare state in particular, with a rejection of individual responsibility. They accuse the modern welfare state of being a “Nanny State,” which seeks to take care of citizens — ministering to their needs, indemnifying them against all major harms, and relieving them of any personal responsibility to make their lives go well.

In this essay I aim to answer this charge. I will present a form of egalitarianism founded on the idea of equality of opportunity — the prevailing conception of social justice in western liberal democracies. According to this widely shared conception, society must provide a “level playing field,” and individuals should face the consequences of their own choices; those who do well are entitled to the fruits of their success, and those who fare poorly cannot ask for rectification. But exactly what is required to level the playing field?

According to one conventional answer, government should eliminate legal barriers to social mobility, require that employers use meritocratic hiring procedures to fill jobs, and that some other institutions, such as schools, use meritocratic procedures to fill slots.1 A second, more liberal answer, would require equal access to education for all citizens, and insist, more generally, that people receive equal amounts of various resources that society can provide. I shall argue that equality-of-opportunity often requires that people receive quite unequal amounts of the relevant resources — without denying the personal responsibility that is at the core of the “opportunity” idea. This proposal probably entails considerably more equality of income and wealth than currently exists in the United States. Before getting into the details, though, let’s consider why equal opportunity seems so attractive in the first place.

Freedom, Responsibility, Opportunity

Equality-of-opportunity is commonly contrasted with equality of outcome (or result). An advocate of equality of outcome must stipulate what that outcome should be: among the candidates are income, satisfaction, life expectancy, welfare, or the probability of leading a life deemed successful. Political philosophers have actively debated this over the past 30 years. Although this topic is extremely important, it is not my concern here, and so I shall simply assume that a particular result has been decided upon, let us say, the degree of success a person rates his life as having. Thus, if equality of outcome were our society’s theory of justice, then social policy would endeavor to distribute resources so as to equalize, as far as is possible, the degrees of success that people rate their lives as having; if equality of opportunity were our society’s theory of justice, then social policy would endeavor to distribute resources so as to equalize the opportunities people would have for making their lives successful, on their own accounts of success.

Suppose I have three children, and I wish to give them equal opportunities to achieve successful lives. I will allocate the resources at my command — my own parental time devoted to the children, the money I spend on their education — to achieve this goal. Perhaps one child is slow to learn: equality of opportunity among my children will require that I spend more money on her education than on the education of the others. Without such differential expenditure, her accessibility to success will probably be lower than for the other two. Suppose one of my children is not slow to learn but has an emotional cyclicity, clearly inborn, which makes it harder for him to carry out plans, or succeed in school. Then equality-of-opportunity may well require that I spend more time with this child than with the others.

Suppose, at any rate, that I successfully implement a strategy of equalizing the opportunities for my children until they each reach young adulthood: that is, I provide the appropriate compensatory education for the slow child, and the appropriate extra parenting for the emotional one. It turns out that the eldest, Alicia, develops a plan to be a music teacher; Bob, the middle child, although a poor athlete, develops a life ambition to reach the top of Mount Everest on foot; and Karl, the youngest, although no more musically talented than Alicia, develops an ambition to be a concert violinist. I, unfortunately, am at this time diagnosed with a terminal illness, and must make out my will. How should I divide my estate among the children, if I adhere to the equal opportunity principle?

One answer, perhaps the one most people would agree to, is that I should divide my estate equally among the children. This, indeed, will have the effect of giving Alicia a much higher probability of achieving success in her life as she defines it than Bob and Karl will have — for they would require immensely more resources than Alicia to have decent probabilities of achieving success on their terms. (Bob will need to hire an army of sherpas and advisors to help him reach the top of Everest, perhaps many times; Karl will require an expensive 18th century violin, not to mention a conservatory education.) But equal division of my estate could be said to implement a program of equal opportunity for life success, if we believe that I have indeed equalized opportunities for my children in their formative years, and the life plans they have formed are the consequences of having had those equal opportunities. Under this view, my children are each personally responsible for their conceptions of success, and equalizing their likely success, under those conceptions, is unnecessary. In contrast, equality of outcome, or expected equality of outcome, would require me to give a large fraction of my estate to each of Bob and Karl, and a small fraction to Alicia.

As this example shows, equality-of-opportunity views are closely allied with a commitment to personal responsibility. Society, under such views, is not required to insure individuals against bad results, when they are the consequences of individual choices made after opportunities have been equalized. On the equality-of-outcome view, in contrast, society’s mandate is to render all lives equally successful, at least in so far as this is feasible. Thus, persons are effectively not held responsible for their choices.

I have here ignored certain important kinds of things that happen to people, namely, pure bad luck — bad luck that a person could have done nothing to avoid. The legal philosopher Ronald Dworkin calls this “brute luck.” Being hit by a truck which runs a red light while you are in the pedestrian crossing is brute bad luck. Being hit by a truck while you are jay walking is not: for in that case, you took a calculated gamble and lost, a gamble you need (and perhaps should) not have taken. Brute luck is to be contrasted with option luck, which is the luck of the voluntarily taken gamble. Even under an equal-opportunity view, we might well decide that society should insure its citizens against brute bad luck, but not against option bad luck. Under an equal outcome view, society must insure its citizens against bad luck of any kind, whether the consequence of voluntary gambles or not.

Citizens of western liberal democracies generally endorse equality of opportunity, I believe, because we think it is morally correct to hold persons responsible, at least to some degree, for their actions. This moral view about responsibility devolves in turn from the western liberal view of the value of individual freedom. If individuals are to be free to choose how to lead their lives, then they must be held accountable for those choices: otherwise such a freedom is vacuous. In economic phraseology, the cost of freedom is responsibility. If, in contrast, we thought that individuals were not free, that their actions were all part of God’s plan, for example, then it would not be so obvious that they should be held responsible for the consequences of those actions. In other words, a notion of individual freedom requires a concomitant view of personal responsibility, with two qualifications: that equality of opportunity has been implemented before responsible choices are made, and that society insure individuals against brute bad luck.

An equality-of-outcome view, on the other hand, can be justified if one believes that there is no such thing as real individual freedom, perhaps because of predestination, or because the actions of persons can always be reduced to causes over which they have no control. Suppose one believes that a person’s behavior is completely determined by a combination of her genetic make-up, and by influences upon her over which she has no control: the country and family into which she was born, the particular teachers and adults to whom she was exposed, etc. One could construct a tree of causes, so to speak, leading backward from any action the person takes, rooted finally in an initial set of genetic and circumstantial variables beyond the reach of her powers. Freedom requires that an alternative action be possible, which this tree of causes does not leave room for. On this view, an equality-of-outcome conception of justice would be morally appealing. How, indeed, could we justify society’s allowing persons to lead differentially successful lives if those lives are beyond the control of persons, and society could make lives more equally successful with a different social policy?

A Definition of Equality of Opportunity

Equality of opportunity, then, is closely connected to fundamental values of responsibility and freedom. But what exactly is it? My strategy in answering this question shall be first to sketch a rough account of equal opportunity, and then to proceed to a more concrete definition by reflecting upon some examples. Finally, I hope to provide a procedure by which a society can implement equality of opportunity as a social policy.

A person’s actions are determined by two kinds of cause: circumstances beyond her control, and autonomous choices within her control. Any particular action a person takes, and its associated consequences, are thus caused by a highly complex combination of circumstances and autonomous choices. I say that equality of opportunity has been achieved among a group of people if society indemnifies persons in the group against bad consequences due to circumstances and brute luck, but does not indemnify them against the consequences of their autonomous choices. Thus an equal-opportunity policy must equalize outcomes in so far as they are the consequences of causes beyond a person’s control, but allow differential outcomes in so far as they result from autonomous choice. When there is equality of opportunity, then, no one will be worse off than others as a result of factors beyond her control.

This definition suffices to convey the intuitive idea of equal opportunity. Now we need to make it more precise. That will require a way to decide which aspects of a person’s behavior are due to circumstance and which to autonomous choice. And, once we have such a way, we will need to see how to implement it — to separate consequences of behavior into those due to autonomous choice and those due to circumstance. The meat of my proposal, which follows, is a procedure for doing just this.

I shall proceed by example. Consider the problem of compensating persons for lung cancer acquired as a result of smoking. Although people in our society have all been intensively exposed to warnings about the dangers of smoking, many persist in smoking, and of those, a fair share develop lung cancer or other serious ailments that require costly medical care. Suppose we hold an equality-of-opportunity for health ethic. To what extent should that medical care be financed by society at large, and to what extent should the individual have to pay? If, indeed, we decided that an individual were entirely responsible for his choice to smoke two packs of cigarettes a day, then an equality-of-opportunity view would say that he should pay the costs of the consequent diseases.

The choice a person makes to smoke or not to smoke is in part determined by his circumstances — say, his economic class, his ethnicity, whether his parents smoked, and his level of education — and is in part a matter of autonomous choice. (One might question whether economic class and level of education are properly part of circumstances, since there is an aspect of autonomous choice in determining them. I shall assume that, for the purposes of analyzing the smoking problem, these are circumstances, in the sense that a person does not consider the effect of his “choice” of economic class and level of education on whether or not he will come to smoke.) I propose, first, that society make a list of the factors beyond a person’s control which it views as influencing the decision whether or not to smoke. Second, we divide society into groups or types according to individuals’ values of these factors (i.e., a group consists of all persons whose factors all have approximately equal value). Suppose, for example, that the list of circumstantial factors society decides upon is: age, ethnicity, gender, occupation. Then one type will consist of all sixty-year-old White, female college professors, and another of all sixty-year-old Black, male steelworkers.

Society wishes to decide the social compensation a person should receive, in the form of socially financed medical care, if he contracts lung cancer. Assume that the chances of contracting the disease increase with the number of years a person smokes. Within each group, there will be a distribution of years smoked: some people will smoke more, some will smoke less. This distribution is a characteristic of the type, not of any individual. The society that has chosen its list of circumstantial factors, determining type, should view the different locations of particular sixty year old Black steelworkers in their group as due to their responsible choice, for their circumstances have already been normalized by type, and similarly for sixty year old White female college professors. That a 60-year-old Black male steelworker is more likely to have smoked for thirty years than a 60-year-old White female college professor is a statistical fact not due to the autonomous choices of individuals, but to group: this is a characteristic of the smoking distributions of the different types, not of any individual. Thus, the distribution of years smoked within a group provides us with a way of calibrating the real opportunities of the members of a group. To take an extreme case, if all 60 year old steelworkers smoked for thirty years, I would say that the choice of “not smoking” was not accessible to 60 year old steelworkers: as a 60 year old steelworker, one would have had effectively no opportunity except to smoke for thirty years. Given one’s group, certain choices may be effectively, even if not physically, barred.

How, then, might one equalize the opportunity for a life free of lung cancer, or at least for a life free of shouldering the financial burden of contracting that disease? I propose that we seek a distribution of socially financed medical care which is equal, across groups, for all those who exercised a comparable degree of responsibility in regard to smoking. To be specific, consider the college professor and steelworker who each have smoked the median number of years for their types — let’s say 8 and 30 years, respectively. I view these two as having acted with comparable responsibility. Alternatively phrased, the act of smoking 8 years for a White, female college professor and the act of smoking 30 years for a Black, male steelworker are equally accessible acts, and this is because exactly one-half of White female college professors smoked less than 8 years, and exactly one-half of Black male steelworkers smoked less than 30 years (see figures 1 and 2 on which the median smokers of the two groups are depicted).

Suppose, then, that society decides it should pay all the medical-care costs of the median professor who contracts lung cancer. Then it should also pay all the medical care costs of the median steelworker. For the differences in the number of years smoked by these two smokers, each at the median of her type, are due entirely to circumstances that society has decided are beyond their control. Of course, society will end up paying much more in the treatment of steelworkers’ lung cancer than college professors’, because those who have smoked for thirty years will have a far higher incidence of disease than those who have smoked for eight.

Now I have taken the median smokers of their types as an example: but my equality-of-opportunity procedure requires that we treat persons from different types in the same way wherever they are located in their type distributions. Thus the White female college professor who is at the 80th centile of the smoking distribution of her type should also be provided with socially financed medical care to the same extent as the Black male steelworker who is at the 80th centile of the smoking distribution of his type.2 Though this uniform treatment across groups is attractive, it turns out to be impossible to achieve for all centiles simultaneously. So some kind of compromise proposal for the social allocation of that resource to lung cancer victims has to be used. I have such a compromise proposal, but I will not present the details here. The salient point is that, for every centile of the smoking distribution, the social transfer is approximately equalized across groups, but within each group, those who exercised more responsibility will be given more favorable treatment. People are indemnified against the consequences of being in a particular group, but not against the consequences of their autonomous actions within that group.

One aspect of the procedure I’ve described needs further justification. Suppose Fernando and Gabrielle belong to different groups, and that each has smoked the median number of years for their group. Why do I consider them comparably responsible for their actions, and hence, why do I say that they should receive equal social compensation for their bad health outcomes? The justification lies in the observation that the frequency distribution of years smoked for a group is a characteristic of the type, not of any person. People are not responsible, by hypothesis, for the group they are in; hence they cannot be held responsible for the frequency distribution of years smoked that is characteristic of their group. Exactly at what point in the distribution Fernando sits, however, is, by definition, a consequence of his autonomous choice: for society has already factored out everything that it considers to be beyond Fernando’s control, in so far as his smoking behavior is concerned, by assigning him to a group. Thus if exactly half the people of Fernando’s type have smoked less than he, and exactly half the people of Gabrielle’s type have smoked less than she, then it is reasonable to say they have exercised comparable degrees of will power, or have taken comparable degrees of responsibility, for their smoking behavior.

Before leaving the smoking example, I should perhaps say that, in the last few months, new information has emerged relevant to the assignment of personal responsibility in decisions to smoke. We have learned that cigarette manufacturers discovered, many years ago, that nicotine is addictive, and that many or all of them add nicotine to natural tobacco to enhance this addictive effect. Thus, another factor in a person’s circumstances, which I omitted in the example as described above, is the biological predilection of a person to nicotine addiction. Suppose our medical technology were sufficiently advanced to ascertain the degree of this predilection for all persons. This, indeed, should also be a component of type, as it is clearly beyond the person’s control. We might find that this new factor explains a good deal of the variance in years smoked within the types into which we initially partitioned society. As this example indicates, the development of medical technology, in many cases, will cause society to add new components to the list constituting the definition of type. As this happens, some actions that formerly appeared to be matters of personal responsibility come to be seen as due to circumstances beyond the person’s control.

Consider a second example, one which is perhaps more important than the smoking case, and more likely than health to be viewed as subject to an equal-opportunity ethic, rather than an “unqualified right” ethic. In our society, one’s income, and probably also one’s success in life, depends positively on the amount of education one receives. Income is strongly correlated with years of education, and, although I am not familiar with the survey data on this question, I would conjecture that the degree to which people rate their lives as successful is also strongly correlated with the number of years of education they receive. Let us, at any rate, assume that this is so. Now the years of education a person acquires depend, as always, on two kinds of factor: circumstances beyond the person’s control, and autonomous choices within her control. Suppose we, as a society, wish to implement a social policy of equality-of-opportunity for income; to do so, we shall concentrate on the relationship between years of education acquired in youth and income earned in later life. The general principle I have been describing says that we should design a social policy which indemnifies individuals against the low incomes which are the consequence of poor or insufficient education in so far as that insufficiency is caused by the individual’s circumstances, but not indemnify her against the income consequences of insufficient education to the extent that that insufficiency is a result of autonomous choice. For simplicity of exposition, I shall assume that income in society is exactly determined by the number of years of education one receives.

I shall proceed just as before. Society’s first step is to make a list of factors that it deems to be beyond a person’s control and that affect the years and quality of education that he receives. Perhaps this list will consist in the following: the years of education his parents had, his parents’ income, his ethnic group, his natural intelligence (assuming society can agree on how to measure this), the number of siblings he has, and whether he was raised by a single parent or by two parents. All of these factors are beyond a person’s control, and they all arguably affect the number of years and/or quality of a person’s education. The next step is to partition society into groups, where each consists of all persons who share the same values of these six factors. Now each type will include a large number of persons; there will be a frequency distribution of years of education for each type, and, of course, these frequency distributions will differ across types. The frequency distribution of years of education is a characteristic of the group, not of any single individual. Since persons are not responsible for their type, they cannot be responsible for this distribution. Where, however, a person sits in the frequency distribution of his type is viewed as a consequence of his autonomous choice, because, in listing the six factors of circumstance, we have, by social decision, exhausted the conditions we regard as beyond a person’s control. So the differences in educational level reached within a type are due, by definition, to differences in autonomous choice, and hence, matters of personal responsibility.

My equality-of-opportunity proposal in this example is a policy that equalizes, through the tax-transfer system, the income across types of all those at a given location in the group’s frequency distribution of education. Concretely, if Alice has achieved the median level of education for her type (i.e., exactly half of the persons of her type have gotten less education than Alice), and if Bernard has achieved the median level of education for his type, then our tax system should attempt to equalize the income of Alice and Bernard, even though they may have acquired quite different amounts of education. It is important here to recall that I am assuming that income is completely determined by years of education.

A comment about my equality-of-opportunity proposal is in order. I have not attempted to provide a theory of what aspects of a person’s behavior really are beyond his control, and what aspects are really within the realm of autonomous choice. Each society, according to my account, decides this question for itself. Thus, different societies will generally choose different lists of factors comprising a person’s circumstances. An individualistic society like the United States would probably include fewer factors in the list of a person’s circumstances than a social-democratic one like Sweden. Thus my proposal is not metaphysical, in the sense of trying to solve the deep problem of what actually is beyond a person’s control; it is political in the sense that it depends on the current views of the society in question. According to the proposal, each society can implement equality-of-opportunity according to its own conception of what features of a person’s social and biological environment constitute factors beyond her control.


The conception of equality-of-opportunity that I have described is not the conventional view. It appears to support a far more egalitarian society than we now have in the United States. No doubt it will generate considerable disagreement. To forestall some of that, I will respond here to a few likely objections.

One criticism — suggested in the work of Ronald Dworkin — may be that the distinction between what a person is and is not responsible for is not the same as the distinction between what she has and has no control over. I have in fact assumed that those two distinctions are, by definition, the same. Dworkin’s view may seem paradoxical; I shall try to make it less so by example. Suppose a child, who grows up in a poor family, whose parents lack education beyond primary school, who is exposed to no books in the home or any kind of high culture, develops preferences in which education has a low value. He does not care to become educated, and feels education will not make his life more successful. He identifies with these preferences, views them as intrinsic to who he is. Then Dworkin, I think, would have to say that such a child does not require any social compensation for the low level of education he acquires, and the consequent low income he earns.

Dworkin places tastes with which a person identifies, and the choices that follow from them, within the realm of personal responsibility, regardless of whether those tastes were formed or induced by factors over which the person had no control. I, on the other hand, do not make the distinction between autonomous and non-autonomous choice depend on what the person thinks, but rather on what society deems to be within or beyond a person’s control. Thus, the unfortunate child I have just described, or the adult that child becomes, would be due social compensation under my notion of equal opportunity for income, but not under Dworkin’s.

A second objection comes from the vantage point of efficiency. It is all well and good, you might say, to attempt to equalize opportunity, and, you might agree that my proposal is the way to accomplish that goal. But you might further say that the cost of equality of opportunity may well be a substantial decrease in national income: for the incentive effects of the redistributive taxation that would be necessary to implement my program — say, in the second example of education and income — would be such as to decrease the labor supply of highly skilled, and hence high-income, individuals. If those who are in ‘fortunate’ types, and who earn large incomes, are taxed to increase the incomes due those in unfortunate types under my proposal, they might work less, and in that case national income per capita would fall, perhaps disastrously.

There are two responses to this objection. The first relates to the incomplete definition of the equality-of-opportunity proposal that I have here made. In fact, the social policy I advocate is the one that equalizes opportunities (for income, say) at the highest possible levels. It is not possible to make this precise without going into some mathematical detail, but the idea is, roughly, that if a taxation policy results in opportunities for income being equalized at a very low level, then it is not the optimal policy. It is true that, under my proposal, mean income, that is, national income per capita, may well be below what it would be without any redistributive taxation: the benefit associated with that reduction in mean income would be increased equality of opportunity.

The second response to the objection is that, if you think social policy should attempt to maximize national income per capita, then you simply cannot advocate equality of opportunity. These two goals are just not simultaneously achievable. If, indeed, the highly skilled would to some degree withdraw their talents from productive use if their incomes were highly taxed, then maximizing mean income in a society could only be accomplished at the expense of equalizing opportunities. Thus, to the extent that our society measures its economic success by the rate of growth of mean income (i.e., GNP per capita), it is not measuring success by the extent to which society achieves equal-opportunity. If we rigorously adopt an equality-of-opportunity ethic, then we must redesign our statistical measures of what constitutes economically successful social policy.

A third objection is of a deeper philosophical nature. It is a conservative, or more properly, a libertarian objection, and runs as follows. It may be that children are not responsible for their genetic and social/familial environments, but it does not follow, according to this objection, that the just social policy entails compensating persons for the consequences of the differences in their genetic and social circumstances. Persons legitimately deserve to benefit from their natural genetic endowments; and parents, as autonomous adults, are responsible for providing their children with opportunities. Parents, furthermore, can legitimately bestow on and bequeath to their children the wealth they have legitimately earned. Society’s legitimate intervention is restricted to providing, let us say, free public schools and enforcing anti-discrimination laws.

This position may be ethically coherent; perhaps it can be given a sound logical foundation. It is, indeed, the task of libertarian political philosophy to do so. But this position is not consistent with equality of opportunity. We just cannot say that Fernando and Gabrielle, or Alicia and Bernard, face equal opportunities when the success of their lives will be vastly different, and quite predictably so, on account of features of their environments over which they have no control.

I have sketched some of the reasons that lead western, liberal democrats to advocate equality of opportunity, and spelled out what I think equality of opportunity entails. Although I did not argue for equality of opportunity as an ethic, I do believe that equality-of-opportunity is a sound ethical position. Perhaps the procedure I have outlined for implementing it will convince some readers of that soundness. Others, perhaps, will be convinced that equality-of-opportunity entails what I say it does, and therefore will reconsider their advocacy of an equal-opportunity doctrine, because they believe the consequences of an equality-of-opportunity view are unfairly egalitarian. Finally, there will be those who believe that the equal-opportunity view I have outlined is unfairly inegalitarian, that people who suffer large harms should be indemnified by society, even when they are responsible. I do not claim to have resolved these disagreements, but to have clarified their terms and to have shown that we can be egalitarians without rejecting the fundamental ideal of personal responsibility.

1Affirmative action rules are exceptions to this policy, but I will not discuss them here.

2I do not necessarily mean that provision of social financing to the same extent means equal monetary provision, for we may also have to take account of the income differences between the steelworker and the college professor that are not due to autonomous choice.