Comments: Nicholas Barr

Ke-young Chu, Sanjeev Gupta, and Vito Tanzi
Published Date:
May 1999
  • ShareShare
Show Summary Details

It is a great pleasure both to be back in Washington and to share a platform with my former colleague, Amartya Sen. Sen’s key message is that equity cannot be captured as a single number, but is multidimensional and should be treated as such by policymakers—a point made also in Atkinson’s paper.

The point is perhaps best illustrated by analogy. If I go for a health check, the doctor has a long checklist of things to look at. At the end, she draws to my attention potential problem areas and ameliorative action. What she does not do is boil down 58 test results into a single number and tell me that, after adjusting for age differences, I am 5 percent less healthy than Sen.

Broader Dimensions of Equity

Sen reminds us that there are at least three strategic questions that policy analysts should ask about inequality. The first is, inequality of what? Inequality is conventionally measured in terms of income/consumption, or wealth.

As teachers, we always tell our students that, whatever else they do in the exam, they should answer the question. Like any good rule, there are occasions when it is right to break it, and Sen makes no bones about his “determined departure from the framework of questions that I have been asked to address.” Discussions of inequality, he argues, should include discussion not only of income or wealth, but also of:

  • health, including morbidity, mortality, nutritional state, and infant mortality;

  • human capital, for example, literacy rates;

  • freedom, democratic rights, and equality before the law, including—a point Sen rightly emphasizes—empowering citizens;

  • employment opportunities: “If unemployment batters lives, then that must somehow be taken into account in the analysis of economic inequality”; and

  • access to critical inputs, such as land and clean water.

The next question is, inequality between whom? Sen discusses inequality between men and women, between larger and smaller households, and between boys and girls. He cites figures showing that in both India and sub-Saharan Africa, male literacy, at 63–64 percent, is more than 50 percent higher than female literacy; in the worst Indian district, the literacy rate for men is nearly five times as high as for women (Table 4.2). He also discusses inequality between adults and children; for example, 63 percent of children under age 5 in India suffer from malnutrition.

The third question is, unequal for how long? The issue (raised also in Munnell’s comments) is that of mobility. Society A, where poor people remain poor to the nth generation, is in an important sense more unequal than society B, with the same number of poor people at any one time, but where individuals experience poverty mainly as a transient phenomenon.

Why a Multivariate Characterization of Equity?

Sen’s assertion that inequality is multidimensional is right, it can be argued, for two separate sets of reasons. Sen is correct if we agree with the value judgment that justice requires that we take account of outcomes, not inputs. Income and wealth are really inputs: outcomes relate to health, freedom to choose, security, and so forth. As Sen says, “It would be absurd to assume that the reach of economics does not go beyond income inequality,” and “there is no justification for ignoring the deprivation and inequality associated with illiteracy, morbidity, undernourishment, or infant mortality” merely because such phenomena may be hard to measure.

Once we accept this value judgment, we are bound to take account of multiple dimensions of equity. As a technical matter, however, it is not possible to boil them down into a unique number that enables us to say, for example, that Western Europe is more equal than the United States.

It is important to understand why the entire enterprise of boiling down different dimensions of equity into a unique, value-free scalar is misconceived. Suppose that we have a vector of measures of equity

where h1, h2,…, hn are measures of health outcomes, e1, e2,…, em measures of educational attainments, f1, are measures of freedom, and g1 and l1 measures of gender equality, the rule of law, and so on. The only way to convert this vector into a scalar is to multiply it by another vector

where w1 is the weight we attach to h1, etc. Saying the same thing in a less technical way, we could express each element in vector (1) as a score out of 100 and then calculate the average value. A simple average implicitly assumes that we weight all elements equally. Alternatively, we could take a weighted average.

Measuring equity thus faces strategic problems. First, though in principle we can measure the different elements in vector (1), in practice measurement is difficult and far from definitive. Second, the elements of vector (2) (the weights) are explicitly a value judgment. Even in principle, therefore, there is no unique answer. Indeed, as Sen points out, there is a public choice aspect to the choice of weights. South Africa, for example, undoubtedly puts greater weight on improvements in racial equality than do some of the reforming postcommunist countries. Thus, there is no unambiguous answer; a complete ranking is not possible.

Where do we go from here? One solution is to give up, and stick to income (though, as we know, income inequality measures face their own major problems). As Sen points out, however, “The fact that there is no unique way of taking an important factor into consideration does not give us a licence to close our eyes to it.” The second solution, therefore, is to give up the Holy Grail of a single scalar measure. Instead, Sen argues, we should think of equity in terms of a checklist and use the results for “the identification of patent injustice.’ This is the analogy of a health check, which I used earlier. In sum, if we accept Sen’s value judgment, the logic leads inexorably to a multidimensional characterization of equity.

Implications for Policy

Equity issues are important for policy, first, as a simple value judgment that social justice is important. But equity is important for efficiency reasons also. This is the second core message in Sen’s paper. “Among the things we can learn from the history of economic development across the world… is to appreciate the tremendous power of human capital in fostering and sustaining economic growth…. [T]he enhancement of social opportunities for participation has played a major part in the successful economic development of Japan and East Asia.” Adjustment and growth that are equitable, in addition, are more likely to be politically (and hence economically) sustainable. Because I have been arguing for at least 15 years (see Barr, 1992 and 1998) that the welfare state is as much an efficiency device as an instrument of social justice, this argument—equity as an efficiency device—is one that is close to my heart.

Two comments in conclusion. First, another dimension of equity is security. Let σi2 = the variance of potential outcomes facing the ith individual. Some people face a higher variance than others, that is, there is inequality between people with larger σ2 and smaller σ2. This matters, because other things being equal a higher variance leads to a welfare loss (i.e., a less enjoyable life) for someone who is risk averse, and the effect is large (look at the amount people spend voluntarily on insurance to try to reduce the uncertainty they face). Where efficient private insurance is possible, the existence of risk does not necessitate state action. But for many economic risks—especially unemployment and the risk of postretirement inflation—private insurance is inefficient or nonexistent. Thus, public action may be needed not only for poverty relief, but also to assist insurance and consumption smoothing (Barr, 1992 and 1998).

As Sen notes, the different dimensions of equity in vector (1) can interact:

  • It is now generally accepted that higher unemployment is causally related to poorer health outcomes, including premature death.

  • Remarkable evidence by Richard Wilkinson (1996) shows a strong association between greater income inequality and poorer health outcomes. It is not just the level of income that determines health outcomes: for a given average level of income, a wider dispersion of that income is associated with poorer health outcomes.

  • Though the jury is still out, there is early evidence (Shapiro, 1995) that stress can lead to adverse health outcomes. Russia experienced a catastrophic decline in male life expectancy between 1990 and 1994—from 63 years to 57 years; female life expectancy fell by about half that amount (World Bank, 1996, p. 128).

The importance of the psychological dimension is reinforced by large-scale studies of British civil servants, which show that lack of control contributes to poor health (junior civil servants arguably face less stress than their seniors but have less control over their working lives; it is the junior civil servants who experienced poorer health outcomes) (Marmot and others, 1991, pp. 58–59; North and others, 1993, pp. 361–66).

What are the key messages for policy design?

  • Concern with equity should consider its multiple facets—physical (e.g., health), intellectual (literacy), and psychological. The Russian story illustrates the potential importance of uncertainty and stress. Policy should consider not only poverty, but also insecurity.

  • Equity is important not only because of the intrinsic importance of social justice but also as an important component of efficiency. Investing in social development, as Sen points out, is not just a matter of justice (important, though, that it is), but is a critical ingredient in economic growth.

  • And, perhaps, the most important interaction of all—that between the level of well-being in a country and the distribution of that well-being.


    Other Resources Citing This Publication