Published on Let's Talk Development

Inequality of opportunity: the new motherhood and apple pie?

This page in:

On the face of it, questioning the usefulness of “inequality of opportunity” seems about as wrongheaded as questioning the merits of family vacations, Thanksgiving or dessert trolleys. What’s not to like about it? Well, as we argue in a recent World Bank working paper, the idea is not quite as useful as it might at first glance appear, and is in fact rather dangerous. But turned upside down, it might yet be useful.

A simple idea – let’s see some numbers

The idea behind inequality of opportunity is simple yet powerful. Not all inequality is bad. The bad bit of inequality (‘inequality of opportunity’) is the part that emerges because of factors over which we have no control (our 'circumstances'). By contrast inequality that emerges because of our different choices and efforts (holding constant our circumstances) is fine, and to be encouraged.

For inequality of opportunity to be a useful idea for policy, we need an estimate of how much of observed inequality is due to inequality of opportunity – or at least an upper bound on this number. So, for example, if inequality in income is 50/100, we’d like to be able to say that at most 20 percentage points is due to inequality of opportunity. Policy should aim to get rid of at most these 20 percentage points, but shouldn’t try to eliminate the remaining 30 percentage points.

In the last few years an industry has emerged aimed at quantifying inequality of opportunity. Much of the work originates in the World Bank, with the 2006 World Development Report providing the impetus. A key piece of work in the genre includes a book on inequality of opportunity in Latin America. There’s also a nice two-minute video on the Bank’s YouTube channel.

Stripping away the technicalities, the exercise proceeds broadly like this. Take a household survey. Decide on an outcome – it could be immunization coverage, or educational attainment, or income. Calculate the amount of inequality in the outcome in the data — let's say it’s 50/100. Then identify in the dataset the variables that capture the influences on the outcome over which the person had no control – their ‘circumstances.’ Form groups (or ‘cells’) based on combinations of these circumstance variables. Calculate how much inequality there is across these groups.  Let’s suppose it’s 20. This is inequality of opportunity. The residual (30) is legitimate inequality due to differences in effort – differences within the cells of people in similar circumstances.

No man is an island

Our outcomes do, of course, reflect our own efforts and decisions. But they also reflect other people’s efforts and decisions, and we have little if any control over these. The outcome of an infant reflects entirely the efforts and decisions of her parents, but she has no choice in who her parents are. That means that all inequality in immunization coverage is unjust. For outcomes relating to small children then we don’t need to get into a complicated decomposition exercise to figure out the share of inequality of opportunity in total inequality – it’s 100%. So it’s odd the Bank’s empirical work in this area does just such a decomposition exercise.

As a child moves through infancy into childhood, parental influences diminish, but they don’t cease – our parents’ choices and efforts shape the way we think and behave, right through to adulthood in fact. And other people’s actions start to affect the choices and efforts that a child makes – their teachers and their classmates. Because children don’t select teachers or classmates, the choices and efforts of these people ought surely to be listed among the circumstances over which the child has no control. But in the empirical work on inequality of opportunity in educational attainment at age 15, they are not explicitly included in the set of circumstances – only implicitly insofar as they are captured by family income and a crude place-of-residence variable (urban vs. rural).  The estimates will thus miss out an important part of what inequality-of-opportunity advocates agree is part of inequality of opportunity.

Luck and risk

Most outcomes also reflect luck. The authors of the World Bank’s empirical work on inequality of opportunity say that luck belongs with effort: “In an ideal world, inequality in outcomes should reflect only differences in effort and choices individuals make, as well as luck.” (p15 of the Latin America book).

To be sure luck would belong alongside efforts if we could eliminate risk from our lives and choose not to do so. But we cannot. Much of the risk we are exposed to is linked to activities we have to engage in to get through the day, if not survive. Often the risks involved are not known with certainty by the scientific community; when they are known, they are not always disseminated in an accessible way, and there are commercial pressures to ignore them.

Diet is a good example. We have to eat to survive. Yet the 2010 Global Burden of Disease (GBD) study finds that dietary risks account for more deaths worldwide than alcohol and tobacco combined. Commercial pressures encourage a default diet that poses risks to health, and we have to make a conscious and determined effort to eat in a way that lowers health risks. Doing so isn’t straightforward. Given the attention they receive in the media, one might imagine the big culprits in relation to diet are too little polyunsaturated fatty acid, and too much processed meat, trans-fatty acids, sugar-sweetened beverages and red meat. Yet these are not according to the 2010 GBD study actually the biggest causes of diet-related deaths worldwide: over six times as many deaths are attributable to people consuming too much sodium, and too little fruit, nuts, seeds, vegetables and whole grains.

True there are examples of where people unnecessarily and knowingly expose themselves to risk. Smoking is the classic example: information on the risks is widely disseminated and in a way that’s very intelligible; and people smoke only for pleasure, not because it is essential to their survival. Shouldn’t inequalities in health caused by smoking decisions be classified then as just? Not necessarily. There is a school of thought that says that people should not be held accountable for bad luck (‘brute luck’ as Dworkin calls it) but only for unnecessary fully-informed risky behaviors (‘option luck’ in Dworkin’s terminology). That might mean taxing tobacco, alcohol, sugar-laden beverages, red meat, etc. at a rate that generates enough revenue to cover the extra expected health care costs, but making sure that everyone – smokers and nonsmokers – receive whatever health care they need to prolong their life and increase its quality.


Like luck, talents play a big role in shaping outcomes. The World Bank inequality-of-opportunity team puts talents in with efforts as well: “Success in life should depend on people’s choices, effort, and talents, not on their circumstances at birth.” (p1 of the Latin America book)

This is also a controversial choice. We can think of ourselves as starting life endowed with innate talents, which we can cultivate during our life. The talents we start our life with affects where we end up. We’re unlikely to become an accomplished concert pianist if we start life with no musical talent. Since we have no control over our innate talents, by lumping talents with effort the Bank’s empirical work will, yet again, end up underestimating the true amount of inequality of opportunity.

Giving the emperor some clothes

In a survey, we will inevitably capture only a small subset of the influences on a person’s outcomes that were genuinely beyond their control. We have no hope of capturing innate talents, the influences and choices of parents, classmates and teachers, and the effects of brute luck (as distinct from option luck).

So although we started off with the observation that some inequality is good, and we wanted to be able to come up with a statement to the effect that the bad part of inequality (inequality of opportunity) is at most x% of total inequality, we end up being able to say only that inequality of opportunity is at least x% of total inequality.

If we’re honest, we actually have no idea how big the good bit of inequality is. Even if we present the number x to the policymaker with the caveat that it’s a lower bound estimate of inequality of opportunity, the policymaker will likely think of it as a point estimate. And given the whole thrust of the inequality-of-opportunity agenda is that not all inequality is bad, the message the policymaker will likely take home is that 100-x percent of inequality in her country is good.

This is a huge limitation of the inequality-of-opportunity work. But it does not mean the exercise was all for naught. What if we turned the thinking upside down? What if instead of pretending we can give policymakers a sense of how much inequality is justifiable, we used these exercises to give policymakers a sense of how much inequality is unjustifiable? True we’d have to be honest and say that at least x% is unjustifiable. But that’s good to be going on with. This is in fact just what researchers in education and health have been doing for years; the authors don’t claim to be measuring all inequality of opportunity but aim simply to parcel out economic differences in outcomes.  

By identifying the part of inequality linked to observable characteristics in household surveys, we also have a chance of developing practical policies to combat these inequalities. Yes, we may agree that inequalities reflecting differences in innate talents, brute luck, and so on are also unjust. But those differences are harder to neutralize through policy. If we can eliminate inequalities due to observed circumstances, we won’t have eliminated all inequality of opportunity, but we will have eliminated at least some of it. And what’s not to like about that? 


Adam Wagstaff

Research Manager, Development Research Group, World Bank

Ravi Kanbur

T. H. Lee Professor of World Affairs, Cornell University

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000