WPS5484 Policy Research Working Paper 5484 Troubling Tradeoffs in the Human Development Index Martin Ravallion The World Bank Development Research Group Office of the Director November 2010 Policy Research Working Paper 5484 Abstract The 20th Human Development Report has introduced a a collapse in its health-care system could still see its HDI new version of its famous Human Development Index improve with even a small rate of economic growth. By (HDI). The HDI aggregates country-level attainments contrast, the new HDI's valuations of the gains from in life expectancy, schooling and income per capita. extra schooling seem unreasonably high--many times Each year's rankings by the HDI are keenly watched greater than the economic returns to schooling. These in both rich and poor countries. The main change in troubling tradeoffs could have been largely avoided using the 2010 HDI is that it relaxes its past assumption of a different aggregation function for the HDI, while still perfect substitutability between its three components. allowing imperfect substitution. While some difficult However, most users will probably not realize that the value judgments are faced in constructing and assessing new HDI has also greatly reduced its implicit weight on the HDI, making its assumed tradeoffs more explicit longevity in poor countries, relative to rich ones. A poor would be a welcome step. country experiencing falling life expectancy due to (say) This paper--a product of the office of the Director, Development Research Group--is part of a larger effort in the department to assess whether prevailing development indices are providing a reliable guide to assessing country performance and guiding policy making. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at mravallion@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Troubling Tradeoffs in the Human Development Index Martin Ravallion1 Development Research Group, World Bank 1818 H Street NW, Washington DC, 20433, USA 1 The author is grateful to Satya Chakravarty, Deon Filmer, Emanuela Galasso, John Quiggin, Suman Seth, Dominique van de Walle, Adam Wagstaff and staff of the Human Development Report Office, UNDP, for helpful comments and discussions. These are the views of the author, and need not reflect those of the World Bank or any affiliated organization. 1. Introduction The Human Development Index (HDI) aims to provide a broader characterization of "development" than is possible by focusing on national income alone. For this purpose, the HDI aggregates country-level attainments in life expectancy and education, as well as income. The index has been published since 1990 in the UNDP's Human Development Reports (HDRs). Each year's scores and rankings by the HDI are keenly watched in both rich and poor countries. The countries that do well on the index are congratulated by each new HDR.2 Politicians and the media often take note. The HDI aims not only to monitor human development, but to encourage countries to take actions that promote it. The latest (2010) HDR claims that the HDI and its various descendants "...yield many novel results--and insights--that can guide development policy debates and designs" (UNDP, 2010, p.8). UNDP (undated) documents numerous examples of the policy influence of the HDRs, including the HDI. As in any composite index, users should know what weights are attached to the HDI's dimensions, to properly judge if it has got the balance right.3 The weight in any given dimension can be defined as the index's first partial derivative ("slope") with respect to that dimension. Since the units of the index are arbitrary (the HDI is normalized to lie in the 0, 1 interval) what really matters is the relative weights of its component dimensions. In other words, we need to know the assumed tradeoffs, as given by the HDI's marginal rate of substitution (MRS), i.e., how much of one desired component of the HDI must be given up for an extra unit of another component, keeping the overall index constant. If a policy or economic change entails that one of the positively-valued dimensions increases at the expense of another dimension, then it is the MRS that tells us whether human development is deemed to have risen or fallen. 2 In answer to the question: "Which countries have been most successful in furthering the human development of their people?" (UNDP, 2010, p.41) the HDR looks at HDI indices over 1970-2010 for 135 countries and identifies the "top 10 movers," defined by the rate of increase in their HDI relative to its 1970 value; the countries are Oman, China, Nepal, Indonesia, Saudi Arabia, Lao PDR, Tunisia, South Korea, Algeria and Morocco. At the other extreme, the report identifies three countries for which the 2008 HDI is lower than its 1970 value: Democratic Republic of the Congo, Zambia and Zimbabwe. 3 On the importance of knowing the weights built into a composite index of development see Ravallion (2010a), which also discusses a number of other issues not touched on here, including the robustness of country rankings and whether aggregation of the core dimensions is useful for policy. 2 While the HDI has clearly aimed to influence policy makers, and appears to have had some success, the interest in identifying its tradeoffs does not rest on a view that the HDI is the maximand of some policy calculus. The interest stems instead from the need to understand the properties of the index. We can all agree that GDP is an incomplete metric of development.4 The real issue is how we form a better composite index, should we feel the need for one.5 What tradeoffs does the index attach to the various components? Only if we accept those tradeoffs can we be confident that the composite index is adequately measuring what it claims to measure. In common with other "mashup indices" (Ravallion, 2010a), the HDI's tradeoffs are not constrained by theory, though economic theory can offer some insights into how one might form a composite index of human development.6 The authors of the HDR set themselves free to pick the HDI's variables and weights. From 1990 to 2009 the HDI gave equal (linear) weights to three functions of its core dimensions for health, education and income.7 While the choice of variables and their weights can certainly be questioned, the HDI has at least appeared to be transparent and simple. That appearance is not quite so evident on closer inspection. Indeed, the HDI has never made explicit its tradeoffs across the core dimensions; users are only told the weights on its three derived functions of those core dimensions, even though the deeper tradeoffs between the core dimensions are clearly more salient. Since income is one of those core dimensions, the tradeoffs can also be monetized, which makes them easier to understand, and to assess whether they are appropriate by comparison with other research findings, including on the economic returns to better health and education.8 4 The claims regularly made by the HDR's that "development" is typically defined solely in terms of GDP have surely been exaggerated, as Srinivasan (1994) argued in an early critique of the HDI. 5 Saying that there is more than one relevant indicator of development does not in itself imply that we need to force them into one dimension; see Ravallion (2010a) on this point. 6 In the context of aggregating mean income and life expectancy, Dowrick et al. (2003) show how revealed preference theory can guide the methodological choices. Also see the more structural economic models in Becker et al. (2005) and Jones and Klenow (2010), and the latent-variable statistical model used by Høyland et al. (2010) to set the weights for a version of the HDI. 7 The income variable has been somewhat controversial, with some observers arguing against its inclusion in the HDI; on the case for including income, see Anand and Sen (2000). 8 Advocates of making human development the overarching development goal often reject monetary valuations. However the fact of using money per se as the metric of value cannot be objectionable; rather the issue is how we assess "value." For further discussion and references to the literature on money metrics of social welfare see Ravallion (2010a). 3 This paper examines the tradeoffs embodied in the latest version of the HDI, as presented in UNDP (2010).9 After summarizing how the index has changed, the paper turns to its valuations of longevity and schooling. The paper questions whether the HDI's implicit valuations are sending the right signals to governments trying to monitor and promote human development. Next the paper shows that the troubling tradeoffs found in the 2010 HDI could have been avoided to a large extent using an alternative aggregation function from the literature--indeed, a more general form of the old HDI, as proposed by Chakravarty (2003). A final section concludes. 2. The Human Development Index The three core dimensions of the HDI are life expectancy (LE), schooling (S) and income (Y). The changes introduced in the 20th Human Development Report (UNDP, 2010) concern the precise measures used for these core dimensions, and how they are aggregated to form the composite index. Life expectancy is the only core dimension that is unchanged in the 2010 HDI. Gross national income (GNI) has replaced GDP, both still at purchasing power parity (PPP) and logged. The two variables used to measure the third component, education, have changed. Literacy and the gross enrolment rate (as used in the old HDI) have been replaced by mean years of schooling (MS) and the expected years of schooling (ES), given by the years of schooling that a child can expect to receive given current enrolment rates. As in the past, the three core dimensions of the HDI are first put on a common (0, 1) scale. The rescaled indicators are: , 1.1 1.2 where the "max" and "min" denote the assumed bounds (in obvious notation). (Note also that S is itself a composite index of MS and ES, which I return to.) 9 In addition to its new HDI, UNDP (2010) introduced a new "multidimensional poverty measure," which raises a number of distinct issues, as discussed in Ravallion (2010a). 4 The bounds used in rescaling all three variables to common units have also been modified. It used to be assumed that life expectancy is bounded below by 25 years, and above by 85 years; in the 2010 HDI these bounds changed to 20 years and 83.2 years (Japan's life expectancy). In the 2010 HDI, GNI per capita is bounded below by $163 (the lowest value, for Zimbabwe in 2008) and above by $108,211 (for the United Arab Emirates in 1980). The new education variables are both taken to have lower bounds of zero with MS bounded above by 13.2 years (the US in 2000) and ES bounded above by 20.6 years (Australia, 2002). Figure 1 gives the density functions of the three scaled indicators across the 169 countries for which UNDP (2010) provides estimates of the HDI. The distributions are quite different, with notably higher values for life expectancy than schooling and (especially) income. There are also signs of bi-modality for schooling and life expectancy. There are 14 countries with life expectancy under 50 years and 13 with mean years of schooling less than three years. (Five countries are common to both categories.) An important change (in the present context) is in how the three scaled indicators are aggregated. The old HDI used their arithmetic mean: /3 2 The 2010 HDI uses instead their geometric mean: / / / 3 Similarly, the way the two education variables are aggregated has changed, so that the new HDI . . has / / . Using either (2) or (3) the HDI is automatically bounded below by zero and above by unity. Note that equation (3) embodies two distinct sources of nonlinearity in the income effect (unlike the one source in (2), namely through the log transformation). In there is both the log transformation of income built into IY and the power transformation in (3). On twice differentiating with respect to Y one finds that the 2010 HDI is still strictly concave in income. However, the combined effect of these two sources of nonlinearity is to impart a large positive income effect on the HDI's valuations of longevity and schooling, as we will see later. 5 Why did the 2010 HDR switch from equation (2) to (3)? The report offers the following explanation:10 "Poor performance in any dimension is now directly reflected in the HDI, and there is no longer perfect substitutability across dimensions. This method captures how well rounded a country's performance is across the three dimensions. As a basis for comparisons of achievement, this method is also more respectful of the intrinsic differences in the dimensions than a simple average is. It recognizes that health, education and income are all important, but also that it is hard to compare these different dimensions of well-being and that we should not let changes in any of them go unnoticed." (UNDP, 2010, p.15) These reasons are not as compelling as they may seem at first glance. It is true that the old HDI assumed that the scaled indices ( , and ) were perfect substitutes (constant MRS), but this was not true of the core dimensions. Since income enters on a log scale (and is only then rescaled to the 0, 1 interval), income and life expectancy (or income and schooling) were not in fact perfect substitutes even in the old HDI. And relaxing perfect substitutability between , and does not imply that one should switch to the form in (3); one can do so by using instead the generalized (old) HDI proposed by Chakravarty (2003). (I will return to Chakravarty's index in section V.) The other arguments made by the 2010 HDR for switching to the geometric mean are also less than fully compelling. It is not evident in what sense using the geometric mean makes poor performance more "directly" reflected in the HDI, or more "well rounded," or "more respectful of the intrinsic differences in the dimensions," or that using this aggregation formula means that we do "not let changes in any dimension) go unnoticed." Indeed, one can argue, to the contrary, that the HDI's new aggregation formula hides partial success amongst countries doing poorly in just one dimension. As dimension approaches we see that approaches zero no matter what value is taken by the other dimensions. Consider, for example, Zimbabwe, which has the lowest of 0.14 (UNDP, 2010)--and it is the lowest by far, at about 60% of the next lowest. Yet this is due to one component that currently scores very low, namely income; Zimbabwe's =0.01--the lowest of any country, and by a wide margin (60% of the next lowest value)--while =0.52 and =0.43, both well above the bottom. Indeed, there are 56 countries with a lower schooling index than Zimbabwe's, yet this relative success is hidden by the HDI's new aggregation formula, given its 10 A number of commentators in the literature have advocated a multiplicative form for the HDI, such as (3), including Desai (1991), soon after the HDI first appeared, and Sagar and Najam (1998) (although the 2010 HDR does not refer to these antecedents in the literature). 6 multiplicative form. Using the arithmetic mean instead (with other data unchanged), Zimbabwe still has a low HDI, but it ranks higher than six countries. (And Zimbabwe does even better using the alternative HDI discussed in Section V.) The rest of this paper examines the country-specific tradeoffs implied by the 2010 HDI, and how they have changed. On a priori grounds it is unclear what effect relaxing perfect substitutability between the scaled indicators ( , and ) would have on the tradeoffs in the core dimensions. Whether the MRS increases or decreases will depend on the data. This is illustrated in Figure 2, which plots the contours (holding the HDI constant) between log income per capita and life expectancy for both the arithmetic mean (the straight line contour) and the geometric mean (the convex one). As usual, the MRS is the absolute value of the slope of the contour. For convenience, countries A and B are taken to have the same HDI either way.11 For country A, the switch implies a higher MRS, while for B the MRS is lower. The fact that we are more interested in the MRS with the core dimension of income ("unlogged") adds further theoretical ambiguity to the effect of this change in the HDI. While the focus here is on the HDI's implicit tradeoffs, knowing those tradeoffs is clearly not sufficient for deciding whether policies that promote health care or education will promote human development. Even leaving aside the issue of whether the HDI is an adequate representation of that goal, we would also need to know the costs, assuming that it is national income net of those costs that is valued for human development.12 And those costs will vary across countries. The costs of lengthening life or raising school attainments are also likely to be higher in richer countries, given that health and education services are labor intensive, and (hence) will tend to be more expensive in rich countries where wages are higher. I will note some comparisons with the costs of increasing longevity, drawing on the literature. However, this paper's focus on the valuations built into the HDI is primarily intended 11 In other words, the scaled values of log income and life expectance are swapped between A and B. More generally, depending on the data and assumed bounds, the switch to the geometric mean may alter the HDI ranking of A and B. 12 For example, let VLE denote the monetary valuation (MRS) for longevity and let MCLE denote its marginal cost i.e., the income forgone for other purposes when life expectancy is increased by one year. Then higher life expectancy will increase the HDI if (and only if) VLE>MCLE. 7 to inform public understanding of the HDI, rather than to inform discussions of what policies might increase human development. In examining the implications of the changes to the HDI for its implicit valuations, I focus first on the HDI's valuation of longevity, after which I turn to its valuation of schooling. 3. The HDI's troubling valuations of longevity While the weights attached to the HDI's scaled indices ( , and ) are explicit, those on the core dimensions (LE, S and Y) are not, and arguably it is these weights that we care about in understanding the properties of the HDI.13 The HDRs have never discussed explicitly the valuations on its core dimensions, and they can be questioned. Ravallion (1997) pointed out the seemingly low monetary value implicitly attached to longevity in poor countries by past HDIs (using an earlier functional form). As we will see, it turns out that the changes introduced in the 2010 HDR have lowered the HDI's valuation of longevity in poor countries even further. The HDI's marginal weights can be readily derived by differentiating equation (2) or (3) with respect to each variable. The effect on these weights of switching to the new formula for the HDI is theoretically ambiguous, and will vary across countries according to: / , , 4 For longevity we find that for 164 of the 169 countries. So the new HDI has lowered the weight on longevity for all but five countries (using the new bounds). For the old HDI the marginal value on longevity was a constant, 3 0.0054. (Going from the lower bound of life expectancy of 20 years, as assumed by the 2010 HDI, to the upper bound of 83.2 years, adds 0.34 to the HDI.) This changed when the HDR switched to the geometric mean; the marginal weight on longevity then became: 5 3 13 The HDI is not alone in this respect. Ravallion (2010a) discusses a range of composite indices of development which tell their users little or nothing about the weights attached to their core dimensions. Their weights are made explicit, but not in what is (arguably) the most relevant space. 8 Figure 3 plots the new and the old weights on longevity against national income per capita (on a log scale to avoid bunching up at low incomes). It can be seen that a strong positive income gradient has been introduced, with markedly lower weights for poorer countries (in terms of GNI per capita). This pattern is not confined to income; the weight on longevity is also positively correlated with the (new) HDI (r=0.697; which is significant at 0.001level using a robust standard error) and life expectancy (r=0.347--also significant at 0.001 level).14 By contrast to longevity, the new formula for the HDI increased the weight on income for the bulk of the countries. In particular, one finds that for 148 countries. The HDI implicitly puts a monetary valuation on an extra year of life, where that valuation is defined by the tradeoff between longevity and income, i.e., the extra income needed to compensate for a year less of life expectancy, keeping the HDI constant. This is given by the ratio of the HDI's marginal weight on longevity to its weight on income. Denote this tradeoff by VLE. We have (in obvious notation): 6.1 6.2 It can be seen that is directly proportional to , given the bounds. The direction of the effect on VLE of switching from the old to the new formula for the HDI is theoretically ambiguous, and depends on both the data and the bounds used for rescaling the variables. Since the weight on longevity has fallen for the bulk of countries, while it has risen for income, we can also expect lower monetary valuations of longevity. More precisely, it is plain from equations (6.1) and (6.2) that if (and only if) . Out of the 169 countries, I find that the monetary valuations of longevity have been revised down for 158 countries (161 if one uses the new bounds). The Annex gives my calculations of the HDI's valuations of longevity for all 169 countries, as well as the 2010 HDI and GNI per capita in 14 Given the function form, the new HDI is strictly concave in its core dimensions, but this only tells us that the weight on x declines with x, holding the other two components constant. In these data-based comparisons, the other variables are not constant, and so their interaction effects come into play. 9 2008. Figure 4 plots the valuations against national income. (I return to explain the "marginal cost" series in Figure 4.) The HDI's value of longevity in the poorest country, Zimbabwe, is a remarkably low $0.51 per year, representing less than 0.3% of that country's (very low) mean income in 2008. Thus the 2010 HDI implies that if Zimbabwe takes a policy action that increases national income by a mere $0.52 or more per person per year at the cost of reducing average life expectancy by one year, then the country will have promoted its "human development." Granted Zimbabwe has an unusually low GNI. The next lowest valuation of longevity is for Liberia, for which the HDI attaches a value of $5.51 per year to an extra year of life expectancy; this is 10 times Zimbabwe's valuation, though it is still only 1.7% of Liberia's annual income. The value tends to rise with income and reaches about $9,000 per year in the richest countries (Figure 4). The highest valuation of longevity is 17,000 times higher than the lowest. Even dropping Zimbabwe's (exceptionally low) valuation, the differential is 1,600. The least-squares elasticity (the ordinary regression coefficient of on ) is 1.208 (with a robust standard error of 0.033; n=169). This is significantly greater than unity, implying that the HDI's valuation of longevity as a proportion of mean income tends to rise with mean income. The elasticity is also higher than most past estimates of the income elasticity of market-based estimates of the value of statistical life.15 The fact that the valuation of longevity as a proportion of mean income tends to rise with mean income is confirmed by Figure 5. (The highest value as a proportion of GNI turns out to be almost 16%, in Equatorial Guinea, though this is clearly an outlier.) By contrast, the old HDI had an income elasticity of unity, and (when evaluated with the HDI's new bounds) is almost exactly 10% of each country's annual income. The changes to the HDI have devalued longevity, especially in poor countries. Given the construction of the index, / is directly proportional to / (equations 6.1 and 6.2); the constant turns out to be 10.014. So by dividing the vertical axis of Figure 5 by 10 (noting that the axis is in percent), we can also read it as a graph for / . (Selected 15 A review of the evidence by Viscussi and Aldy (2003) concludes that the income elasticity is in the range 0.5-0.6. 10 points for / are indicated on the vertical axis in parentheses.) There was a roughly 25% downward revision on average (mean / 0.748). If one focuses on the poorest half of countries (GNI per capita below the median) then the average downward revision was close to 40% (mean / 0.620 ; n=84); for the poorest quarter, the valuation of longevity has been almost halved ((mean / 0.545; n=42)). Figure 6 provides a "blow up" of Figure 4 for the poorest half of countries (in terms of GNI per capita), as well as the values implied by the old HDI aggregation using the arithmetic mean. (I also give the old valuation of life using the arithmetic mean and old bounds.) It can be seen that changing the bounds alone in the old HDI would not have produced this large downward revision to the index's monetary valuation of longevity. Rather it was the combined effect of switching to the geometric mean, the form of the scales used and (of course) the data. Given the scales and aggregation formulae, the marked devaluation of longevity stems from the fact that for all except eight of the 169 countries, implying that . (This difference in the distributions was already evident in Figure 1.) Whether this holds depends on the assumed bounds built into the HDI. For example, a higher upper bound for LE would have lowered the value of life implicit in the old HDI; I find that would have been quite close to the level of at 100. The somewhat arbitrary and time varying choice of bounds has played an important role in the HDI's devaluations of longevity. The income gradient in the HDI's monetary valuations of longevity appears to be substantially greater than the gradient in the marginal costs of longevity. Dowrick, Dunlop and Quiggin (DDQ) (1998) estimate marginal costs of an extra year of life expectancy for 58 countries in 1980, which I have simply converted to 2008 prices using the US CPI. There are a number of comparability problems between the DDQ estimates and my calculations of VLE, so these calculations should only be considered as broadly indicative for the present purposes.16 The DDQ estimates are also given in Figure 4. Their estimated marginal cost of a one year increase in life expectancy is 400 times higher in the country with the highest cost (Denmark in their 16 Probably most importantly, updating solely for inflation in the US misses the structural changes in growing developing economies, which entail changes in their relative prices; in particular, we can expect that the cost of attaining higher longevity may have risen more in rapidly growing economies such as China than these estimates indicate. This is suggested by comparisons of PPP estimates across different rounds of the International Comparison Program; see Ravallion (2010b). 11 sample) than the lowest (Madagascar). This is far less than my calculations of the differential in the valuation of longevity implicit in the HDI. The DDQ estimates are only roughly similar to the HDI's valuations for the poorest countries, but the HDI's valuations greatly exceed marginal costs among most countries, and the gap is very large for the richest countries. Across individuals, one expects the value attached to extra longevity to rise with income. Even if (instantaneous) utility depends only on consumption, a high income allows more to be consumed in the extra years of life, giving higher expected utility.17 Similarly, one would expect people in rich countries to be willing to pay more for extra longevity, and they clearly do. However, such observations do not justify building an income gradient (let alone a steep gradient) into the valuation of longevity. The HDI is clearly intended to embody social values, which need not accord with private ones. With reference to the private valuations of "statistical life"--such as derived from contingent valuation questions in surveys or wage premia paid for risky jobs--Ackerman and Heinzerling (2001, p.18) note a similar concern: "Calculation of the link between average income and the value of a statistical life could, if applied indiscriminately, lead to the unacceptable implication that rich people, or residents of rich nations, are worth more than the poor." While the HDI is not deriving its valuations of longevity from such sources, the fact that it puts a higher value to an extra year of life for people in rich countries than poor ones is arguably no less of an example of the "unacceptable implication" that Ackerman and Heinzerling refer to. This troubling tradeoff in the 2010 HDI will clearly influence its rankings of performance in human development. However, a more worrying concern arises if the index influences (domestic and international) policy making. The HDI's embedded tradeoffs imply that, in the interests of promoting human development--or at least improving its HDI--the government of a poor country should not be willing to pay more than a very small sum (in $'s and as a percent of national income) for an extra year of expected lifespan for its citizens, while the government of a rich country would be encouraged to spend vastly more for the same gain in longevity--17,000 times more if one compares my calculation of for the richest country with the poorest. 17 Suppose instead that (i) utility is strictly increasing in both life expectancy and income; (ii) the marginal utility of higher life expectancy does not fall with higher income, and that (iii) there is declining marginal utility of income. Then the MRS will be an increasing function of income. 12 Serious objections would naturally be raised to any proposal for public action within one country that rested on assigning a lower value to life to poor citizens than to rich ones, let alone a relative value that is such a tiny fraction. The same objections arise in a global context. One is led to question whether these valuations are consistent with promoting "human development." Yet, the 20 HDRs have largely avoided making explicit this potentially troubling tradeoff, although the basic problem was noted in early commentaries (Ravallion, 1997). 4. The HDI's valuations of schooling The fact that the HDI's education variables have changed is not of obvious concern in this context, so I will only use the new schooling variables in the 2010 HDI. Applying equation (4), I find that that the new HDI's aggregation method has put a higher weight on schooling for 119 of the 169 countries (i.e., all those with ). The ratio of the old and new weights in (4) does not depend on precisely how a gain in schooling is allocated between mean actual years of schooling and mean expected years (assuming, naturally, that it is allocated the same way for both calculations). However, in calculating the HDI's new valuation of schooling one does need to know that allocation. I shall assume that an extra year is added to both mean current schooling (MS) and the expected years of schooling (ES).18 While I will use the new education variables, I will keep their old aggregation function, including the use of the arithmetic mean of the (scaled) schooling variables. Then the HDI's implicit valuations of extra schooling are given by: /2 7.1 /2 7.2 Figure 7 plots and against (log) GNI. (Later I will explain the series in Figure 7 labeled "Chakravarty index.") Similarly to longevity, we see a marked income gradient, although flat at low incomes. The new HDI values an extra year of schooling at $1.68 per person per year in Zimbabwe, about 1% of mean income; the next lowest is for the Congo where 18 There is some support for this assumption in the data; the regression coefficient of expected schooling on mean current schooling is 0.88, which is close to unity, although it is still significantly less than unity (t=2.54, based on a robust standard error; n=169). 13 =$33 per year, or 11% of annual income. At the other extreme, rises to $53,000 per year in the country with the second highest GNI per person, representing 67% of that country's GNI. The valuation of schooling has increased in 94 countries, though the increase is more marked amongst high-income countries. Given the cross-country differences in schooling, the valuation of schooling as a proportion of GNI does not rise with GNI above some point; Figure 8 plots / against GNI. (The highest / is for Burkina Faso, but this is an outlier.) While the HDI's implicit valuations of longevity seem low, it's valuations on schooling seem high. In constructing a composite index such as the HDI, there is a (rather poorly- understood) issue about what dimensions are intrinsically, versus instrumentally, important. We can all agree that a longer life is valued intrinsically, independently of income. However, it is not quite so clear that education has such a large intrinsic value (as assumed by the HDI), rather than being (very) important instrumentally to income and (hence) welfare. In defense of the HDI, one might argue that the benefits of extra schooling are not fully reflected in current incomes; better educated parents pass advantages onto their children, leading to higher future incomes. (Possibly the new HDI's introduction of the variable for expected schooling is trying to capture this effect.) But it is a moot point just how much extra one would allow for such an effect, on top of the economic return to schooling. The HDI is presumably measuring a country's current human development not its future value. If we compare the HDI's valuations on schooling with the returns implied by earnings regressions, the HDI's valuations are clearly very much higher. The regression coefficient of log earnings on years of schooling is typically around 0.1; see Psacharopoulos and Patrinos (2002). So it seems that the HDI is putting a much larger value on the returns to schooling than is reflected in current earnings. Indeed, the HDI's valuation in developing countries appears to be roughly four times the labor market returns to schooling.19 Finally, Figure 9 compares the new HDI's valuations for longevity and schooling. What is most striking is how much higher the HDI's implicit valuation of schooling is than its 19 For high-income countries, the ratios of the valuation of extra schooling to mean income in Figure 6 are roughly seven times the coefficients on years of schooling reported by Psacharopoulos and Patrinos (2002, Table 4, p.14). However, Banerjee and Duflo (2005) question whether the evidence supports the claim that returns to education vary much across countries. 14 valuation of longevity. A shorter but better schooled life is preferred by the designers of the HDI. One is left wondering how many of the world's poor--many living short lives by rich- country standards--would agree. 5. Could the HDI's troubling tradeoffs have been avoided? Instead of using the geometric mean, suppose that the HDR's team had generalized its old additive HDI in the natural form proposed by Chakravarty (2003), giving the "generalized (old) HDI:" /3 8 where f is some smooth, twice-differentiable, concave function mapping from the [0,1] to [0,1] with f(0)=0 and f(1)=1. Chakravarty (2003) shows that the form in (8) satisfies three axioms: normalization (if all three components, , take the same value then that value is the HDI), consistency in aggregation (the HDI for a sum of component indices is equal to the corresponding sum of the HDIs across the components) and symmetry (the HDI is unaffected by permutations of its components).20 Consistency of aggregation forces the HDI to be linearly additive in the 's as in equation (8). Chakravarty proposed a parametric special case of (8) in which for 0 1 , giving an index that I will label . The old HDI is the limiting case when =1, and only then does the index impose perfect substitutability (constant MRS) between the s. I will present empirical results here for =0.5 and 0.25. With two further modifications, this special case of the Chakravarty index can take us a long way toward avoiding, or at least attenuating, the troubling tradeoffs in UNDP's (2010) new HDI. The first change is to replace lnY with Y in equation (1.2) so that / .21 This change is important, since it removes a source of the positive income effect on the weights implicit in the new HDI. The second change is to use the arithmetic mean of the two schooling variables, MS and ES (and their bounds), rather than their geometric mean. 20 Chakravarty (2003) actually proves a more powerful result: an even more general index will satisfy these three axioms if and only if it takes the form of equation (8). 21 Note that this still allows diminishing marginal returns to income; the new HDI's functional form--in which income is logged within the scaled index, and then the index is raised to the power of 1/3--is arguably an "overkill" since one only needs one source of nonlinearity. 15 With these modifications, we can avoid the troubling property of the 2010 HDI in Figure 3, whereby the marginal effect on the index of an extra year of life rises with national income per capita (and the HDI itself). Indeed, we now have the reverse slope, with higher weight on longevity in poorer countries; Figure 10 gives the weights on longevity implied by , for r=0.5 and 0.25. Instead of a higher weight on longevity in richer countries, we now find that the weight rises from 0.0026 in one of the richest countries to 0.0042 in the poorest (r=0.5). The pattern is similar using r=0.25, though the negative gradient is less steep. The implied tradeoffs with income are given by: , 9 We still find higher monetary valuations on longevity and schooling in richer countries, but gives higher valuations for poor countries than and the troubling income gradient is much attenuated. Figure 11 compares the valuations on longevity in with those implied by for r=0.5 and 0.25; Figure 7 gives the corresponding valuations for schooling (only for r=0.5 to avoid cluttering up the graph; the series for r=0.25 is similar to the pattern in Figure 11). In both cases, the implied valuations rise with income per capita, but much less steeply than implied by the 2010 HDI. The lower value of reduces the income gradient. The Chakravarty index also puts higher valuations on schooling than longevity, similarly to the 2010 HDI. This property appears to be hard to avoid given the differences in distributions noted in Figure 1 and the assumed bounds. Of course, increasing the weight on relative to that on will narrow the gap in the valuations of schooling and longevity. However, I found that on even doubling the weight on the life expectancy component (equally weighting the other two components) the valuation on schooling still exceeded that on longevity. Figure 12 compares with . (The Annex also gives . by country.) The overall means are similar for r=0.5 (an un-weighted mean of 0.643 for . , versus 0.637 for ), but higher for r=0.25 (mean of 0.773). Switching to increases the index for low HDI countries, and decreases the upper values for r=0.5, and so gives lower overall inequality in the HDIs across countries; for example, the CV falls from 0.291 to 0.194 for r=0.5 and 0.121 for r=0.25. While it is clear that the two HDIs in Figure 12 are highly correlated (r=0.980 r=0.5 and 16 0.987 for r=0.25), there are some large changes. For r=0.5, Zimbabwe's index rises by over 300%, from the lowest value (by far) of 0.140 based on to 0.454; it also rises relatively, th to be the 12 lowest--reflecting the fact that the additivity property of the Chakravarty index puts a higher premium on Zimbabwe's schooling attainment. Using r=0.25, the upward revision to Zimbabwe's index is even more dramatic, with . =0.583. The largest decrease is that for New Zealand, for which the index falls by 0.094 in switching to . , and the ranking falls from third place to 18th. The differences are small at high HDIs using r=0.25 (Figure 12). 6. Conclusions The Human Development Index was introduced in 1990 as an alternative to using national income per capita as the metric of development success. Until 2010 the index was an equally-weighted mean of scaled attainments in three dimensions: life expectancy, education and income. The simplicity of the HDI gave it a transparency that was clearly appealing to many users, although the HDI was never quite as simple as one might think at first glance, given the transformations embedded in its components. Over 20 years, the Human Development Reports (and numerous offshoot reports at national level) have applauded those countries that do well in the HDI, and offered advice to others on how they might do better in the HDI stakes. A new version of the index was introduced in the 2010 edition of the HDR. The main change was to switch from the original additive aggregation function (the arithmetic mean of the three components) to a multiplicative function (their geometric mean). The main reason given for this change was to allow for imperfect substitutability between the HDI's three components. However, good intentions alone do not make for good measurement. The 2010 HDI is both more complicated and more problematic in its tradeoffs across core dimensions. Longevity in poor countries has been substantially devalued, though it seems unlikely that this was intended. The HDI's valuation of longevity in the poorest country is now a mere 0.006% of its value in the richest country--a far greater difference than in their average incomes (for which the poorest country has 0.2% of the national income per capita of the richest). A poor country experiencing falling life expectancy due to (say) a collapse in its already weak health-care system could still see its HDI improve with even a small rate of economic growth. By contrast, 17 the valuations of extra schooling have risen for most countries and they seem high--some four times higher than the valuations typically placed by the labor market on extra schooling. There are some contentious value judgments buried in the maths of the HDI. It can be granted that a rich person will be able to afford to spend more to live longer than a poor person, and will typically do so. But that does not justify building such inequalities into our assessment of progress in "human development." Given what we know about the marginal costs of extending life expectancy, if one accepted the tradeoffs embodied in the new HDI, one would be drawn to conclude that the most promising way to promote human development in the world would be by investing in higher life expectancy in rich countries--surely an unacceptable implication of the HDI's tradeoffs. And it is unclear why we would want to put so much higher a value on schooling than implied by its economic returns. Setting the tradeoffs in a composite index is never going to be easy and it is ultimately up to users to judge for themselves if they accept the HDI's valuations. However, the troubling tradeoffs in the new HDI are not in fact essential to relaxing the perfect substitutability property of the old HDI. The less appealing properties of the new index could have been avoided to a large extent, while allowing imperfect substitutability, by using an alternative aggregation function already found in the literature--in fact a straightforward generalization of the old HDI. An important lesson for future composite indices is the need for transparency about the implicit tradeoffs, especially in more complicated indices. Those tradeoffs are the key to understanding the properties and implications of the index. I would hazard to guess that if the authors of the 2010 Human Development Report had calculated the tradeoffs implicit in their index they would have had second thoughts about it, and looked for alternatives. 18 References Ackerman Frank and Lisa Heinzerling, 2001, "If It Exists, It's Getting Bigger: Revising the Value of a Statistical Life," Global Development and Environment Institute Working Paper No. 01-06, Tufts University. Anand, Sudhir and Amartya Sen, 2000, "The Income Component of the Human Development Index," Journal of Human Development 1(1): 83-106. Banerjee Abhijit V. and Esther Duflo, 2005, "Growth Theory through the Lens of Development Economics," In: Philippe Aghion and Steven N. Durlauf, Editor(s), Handbook of Economic Growth, Elsevier. Becker, Gary, Tomas Philipson and Rodrigo Soares, 2005, "The Quantity and Quality of Life and the Evolution of World Inequality," American Economic Review 95(1): 277-291. Chakravarty, Satya R., 2003, "A Generalized Human Development Index," Review of Development Economics 7(1): 99-114. Desai, Meghnad, 1991, "Human Development: Concepts and Measures," European Economic Review 35: 350-357. Dowrick, Steve, Yvonne Dunlop and John Quiggin, 1998, "The Cost of Life Expectancy and the Implicit Social Valuation of Life," Scandinavian Journal of Economics 100(4): 673-691. _____________, ____________ and ___________, 2003, "Social Indicators and Comparisons of Living Standards," Journal of Development Economics 70: 501-529. Høyland, Bjørn, Karl Ove Moene and Fredrik Willumsen, 2010, "The Tyranny of International Index Rankings," mimeo, Department of Economics, University of Oslo. Jones, Charles and Peter Klenow, 2010, "Beyond GDP? Welfare Across Countries and Time." Mimeo, Stanford University. Psacharopoulos, George and Harry Patrinos , 2002, "Returns to Investment in Education: A Further Update," Policy Research Working Paper 2881, World Bank, Washington DC. Ravallion, Martin, 1997, "Good and Bad Growth: The Human Development Reports," World Development, 25(5): 631-638. ______________, 2010a, "Mashup Indices of Development," Policy Research Working Paper 5432, World Bank, Washington DC. ______________, 2010b, "Price Levels and Economic Growth: Making Sense of the PPP Changes between ICP Rounds," Policy Research Working Paper 5229, World Bank. 19 Sagar, Ambuj D. and Adil Najam, 1998, "The Human Development Index: A Critical Review," Ecological Economics 25: 249-264. Srinivasan, T.N. 1994, "Human Development: A New Paradigm or Reinvention of the Wheel?" American Economic Review, Papers and Proceedings, 84(2): 238-249. United Nations Development Programme, undated, Ideas, Innovation and Impact: How Human Development Reports Influence Change. UNDP, New York. _________________________________, 2010, Human Development Report: The Real Wealth of Nations, New York: Palgrave Macmillan for the UNDP. Viscusi, Kip and Joseph Aldy, 2003, "The Value of a Statistical Life: A Critical Review of Market Estimates Throughout the World," Journal of Risk and Uncertainty 27(1): 5- 76. 20 Figure 1: Densities of the three scaled indicators used by the 2010 HDI 3.0 Log GNI per capita Life expectancy 2.5 Schooling 2.0 Density 1.5 1.0 0.5 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Note: Kernel density functions using Epanechnikov kernel (calculated using Eviews 7). Source: Author's calculations from the data for 2008 provided in UNDP (2010). Figure 2: The effect on the MRS of allowing imperfect substitution depends on the country Log income per capita Country A Country B Life expectancy or schooling 21 Figure 3: Weights on life expectancy in the old and new HDI Marginal effect on HDI of one year gain in life expectancy .007 .006 Old HDI .005 .004 New HDI .003 .002 .001 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) Source (this figure and all following ones): Author's calculation from data for 2008 provided in the 2010 HDR. The fitted line is a locally smoothed (nonparametric) regression. Figure 4: Implicit valuations in HDI and marginal costs of an extra year of life expectancy Implicit valuation or marginal cost of an extra year of life 9,000 Qatar 8,000 Liechtenstein 7,000 6,000 ($ per year) 5,000 Implicit valuation in HDI 4,000 3,000 2,000 Marginal cost (Dowrick et al.) 1,000 Zimbabwe 0 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) 22 Figure 5: Value of an extra year of life expectancy as percent of gross national income 16 Equatorial Guinea Valuation of an extra year of life as % of GNI (1.5) 14 (New valuation/old valuation) 12 10 (1.0) 8 6 (0.5) 4 2 0 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) Figure 6: HDI's revised valuations of life expectancy in the poorest half of countries 800 Implicit value of an extra year of life ($ per year) New HDI 700 Old HDI (new bounds) Old HDI (old bounds) 600 500 400 300 200 100 0 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 Gross national income ($ per person per year; log scale) 23 Figure 7: Implicit monetary values attached to an extra year of schooling by the 2010 HDI Valuation of an extra year of schooling ($ per year) 60,000 50,000 New HDI 40,000 30,000 Old HDI 20,000 Chakravarty 10,000 index (r=0.5) 0 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) Figure 8: HDI's valuation of an extra year of schooling as a percent of national income 100 Value of an extra year of schooling as % of GNI Burkina Faso 80 60 40 20 0 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) 24 Figure 9: HDI's valuations of schooling and longevity across countries 60,000 Value of an extra year of schooling ($ per year) 50,000 40,000 30,000 20,000 10,000 0 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 Value of an extra year of life expectancy ($ per year) Figure 10: Weights on life expectancy in an alternative HDI using the Chakravarty index Marginal effect on HDI of a one year gain in life expectancy .0044 r=0.5 .0040 .0036 r=0.25 .0032 .0028 .0024 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) 25 Figure 11: Valuations of longevity in the 2010 HDI vs. alternative HDI 9,000 Monetary valuation on an extra year of life 8,000 7,000 ($ per person per year) 6,000 5,000 New HDI 4,000 3,000 2,000 Chakravarty 1,000 index (r-0.25) Chakravarty index (r-0.5) 0 5 6 7 8 9 10 11 12 Gross national income ($ per person per year; log scale) Figure 12: Comparison of the 2010 HDI with alternative HDI using the Chakravarty index Alternative index using Chakravarty's generalized HDI 1.0 r=0.25 0.8 r=0.5 0.6 0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 New HDI in 2010 Human Development Report 26 Annex: Human Development Index by Country and its Implicit Tradeoffs Implicit value of Implicit value Alternative Gross national an extra year of of an extra year HDI using Human income per life expectancy of schooling Chakravarty's HDI Development capita (US$ (US$ PPP, (US$ PPP, generalized rank Country Index 2010 PPP 2008) 2008) 2008) HDI (r=0.5) 1 Norway 0.938 58809.53 5676.31 23703.14 0.887 2 Australia 0.937 38691.71 3421.38 13962.64 0.856 3 New Zealand 0.907 25437.50 2119.82 8388.97 0.813 4 United States 0.902 47093.85 4478.40 19194.40 0.848 5 Ireland 0.895 33077.57 2913.07 12483.41 0.821 6 Liechtenstein 0.891 81011.42 8439.05 41445.94 0.899 7 Netherlands 0.890 40657.78 3719.45 16763.15 0.832 8 Canada 0.888 38668.37 3464.53 15830.11 0.827 9 Sweden 0.885 36936.27 3269.29 15057.89 0.822 10 Germany 0.885 35308.04 3153.65 13867.94 0.818 11 Japan 0.884 34692.46 2944.15 14268.21 0.817 12 Korea (Republic of) 0.877 29517.62 2566.25 11151.29 0.804 13 Switzerland 0.874 39849.09 3522.89 17750.42 0.824 14 France 0.872 34340.71 2980.43 14496.23 0.812 15 Israel 0.872 27831.34 2339.34 10600.29 0.797 16 Finland 0.871 33871.73 3008.29 14054.14 0.811 17 Iceland 0.869 22917.03 1826.53 8555.26 0.790 18 Belgium 0.867 34872.70 3103.36 14723.96 0.810 19 Denmark 0.866 36404.41 3353.66 15402.79 0.813 20 Spain 0.863 29661.16 2518.96 12172.00 0.799 21 Hong Kong, China 0.862 45090.48 4055.56 21856.34 0.826 22 Greece 0.855 27580.38 2370.80 11034.24 0.790 23 Italy 0.854 29619.21 2507.96 12664.21 0.795 24 Luxembourg 0.852 51109.19 4902.85 25604.28 0.831 25 Austria 0.851 37055.90 3329.38 16972.02 0.806 26 United Kingdom 0.849 35087.16 3153.20 15864.71 0.803 27 Singapore 0.846 48893.19 4592.41 25475.05 0.827 28 Czech Republic 0.841 22678.39 1965.98 8231.51 0.769 29 Slovenia 0.828 25857.03 2227.93 11182.50 0.775 30 Andorra 0.824 38056.49 3413.24 19086.43 0.792 31 Slovakia 0.818 21657.78 1921.00 8138.31 0.755 32 United Arab Emirates 0.815 58005.80 5903.94 33267.11 0.823 33 Malta 0.815 21004.33 1699.77 8690.60 0.754 34 Estonia 0.812 17167.68 1487.76 5855.15 0.742 35 Cyprus 0.810 21962.46 1796.11 9358.27 0.753 36 Hungary 0.805 17472.12 1516.02 6172.92 0.739 37 Brunei Darussalam 0.805 49914.55 4973.92 29356.66 0.809 38 Qatar 0.803 79426.35 8783.13 53049.90 0.856 39 Bahrain 0.801 26663.87 2425.78 11974.10 0.758 40 Portugal 0.795 22105.19 1836.83 10251.37 0.751 41 Poland 0.795 17803.06 1492.03 6939.45 0.736 42 Barbados 0.788 21672.62 1836.05 9633.26 0.741 43 Bahamas 0.784 25200.64 2337.02 11208.73 0.743 44 Lithuania 0.783 14823.72 1283.76 5161.69 0.723 45 Chile 0.783 13561.02 1019.85 5140.75 0.721 46 Argentina 0.775 14603.33 1178.66 5649.88 0.720 47 Kuwait 0.771 55718.61 5613.54 39630.67 0.805 48 Latvia 0.769 12944.18 1068.14 4565.53 0.711 49 Montenegro 0.769 12490.82 993.23 4435.56 0.709 27 50 Romania 0.767 12843.70 1054.06 4539.88 0.709 51 Croatia 0.767 16388.59 1332.95 6944.98 0.719 52 Uruguay 0.765 13808.44 1081.52 5596.90 0.716 53 Libyan Arab Jamahiriya 0.755 17067.63 1456.92 7872.34 0.721 54 Panama 0.755 13346.85 1050.64 5316.92 0.704 55 Saudi Arabia 0.752 24726.01 2329.44 12592.96 0.729 56 Mexico 0.750 13971.41 1096.99 5897.98 0.704 57 Malaysia 0.744 13926.86 1131.77 5733.77 0.698 58 Bulgaria 0.743 11139.16 875.50 4088.97 0.692 59 Trinidad and Tobago 0.736 24233.27 2427.76 11862.34 0.714 60 Serbia 0.735 10449.37 799.50 3896.45 0.687 61 Belarus 0.732 12925.70 1139.08 4976.54 0.690 62 Costa Rica 0.725 10869.63 772.51 4680.14 0.684 63 Peru 0.723 8424.21 619.21 2941.36 0.677 64 Albania 0.719 7976.33 545.63 2871.56 0.673 65 Russian Federation 0.719 15258.16 1467.10 6371.23 0.687 66 Kazakhstan 0.714 10234.32 933.95 3452.71 0.673 67 Azerbaijan 0.713 8746.57 685.12 3042.79 0.669 68 Bosnia and Herzegovina 0.710 8221.59 580.86 3097.88 0.670 69 Ukraine 0.710 6535.14 496.17 1893.01 0.665 70 Iran 0.702 11764.21 969.75 5290.62 0.675 71 Macedonia 0.701 9486.86 706.88 3925.63 0.667 72 Mauritius 0.701 13343.58 1128.13 6353.35 0.677 73 Brazil 0.699 10606.97 836.84 4692.70 0.671 74 Georgia 0.698 4901.91 320.60 1350.18 0.657 75 Venezuela 0.696 11846.23 936.48 5891.66 0.677 76 Armenia 0.695 5494.61 356.40 1706.02 0.656 77 Ecuador 0.695 7931.24 556.11 3186.96 0.664 78 Belize 0.694 5693.06 355.77 1916.05 0.658 79 Colombia 0.689 8588.94 637.55 3567.00 0.661 80 Jamaica 0.688 7206.85 521.82 2587.66 0.653 81 Tunisia 0.683 7979.31 571.47 3465.21 0.661 82 Jordan 0.681 5955.98 403.56 2056.91 0.650 83 Turkey 0.679 13359.24 1127.01 7031.20 0.665 84 Algeria 0.677 8320.16 618.51 3540.00 0.653 85 Tonga 0.677 4038.39 248.82 1095.64 0.647 86 Fiji 0.669 4315.42 287.22 1183.58 0.641 87 Turkmenistan 0.669 7052.09 586.07 2368.95 0.641 88 Dominican Republic 0.663 8272.56 615.31 3713.07 0.645 89 China 0.663 7258.47 515.29 3035.41 0.642 90 El Salvador 0.659 6498.11 460.96 2545.60 0.638 91 Sri Lanka 0.658 4886.32 305.40 1703.59 0.637 92 Thailand 0.654 8000.62 631.83 3531.18 0.641 93 Gabon 0.648 12746.55 1344.54 5904.01 0.641 94 Suriname 0.646 7092.90 542.06 2963.10 0.631 95 Bolivia 0.643 4357.24 308.97 1299.76 0.626 96 Paraguay 0.640 4585.32 292.73 1618.00 0.626 97 Philippines 0.638 4002.08 244.75 1294.56 0.624 98 Botswana 0.633 13204.19 1633.21 5605.66 0.630 99 Moldova 0.623 3149.33 190.72 870.18 0.616 100 Mongolia 0.622 3619.27 237.33 1091.62 0.616 101 Egypt 0.620 5889.20 417.97 2584.28 0.615 102 Uzbekistan 0.617 3084.89 188.27 849.43 0.612 103 Micronesia 0.614 3265.55 199.79 971.52 0.610 104 Guyana 0.611 3302.06 207.51 988.85 0.608 105 Namibia 0.606 6323.11 549.78 2550.19 0.603 28 106 Honduras 0.604 3750.11 223.51 1419.53 0.608 107 Maldives 0.602 5408.10 361.96 2760.11 0.614 108 Indonesia 0.600 3956.84 245.08 1612.44 0.609 109 Kyrgyzstan 0.598 2291.23 125.07 567.96 0.606 110 South Africa 0.597 9812.13 1257.63 3947.21 0.603 111 Syrian Arab Republic 0.589 4759.93 293.98 2413.75 0.603 112 Tajikistan 0.580 2019.88 107.47 483.51 0.596 113 Viet Nam 0.572 2994.76 158.76 1214.34 0.593 114 Morocco 0.567 4627.57 298.70 2514.32 0.591 115 Nicaragua 0.565 2567.40 131.63 947.65 0.590 116 Guatemala 0.560 4693.74 310.64 2645.98 0.587 117 Equatorial Guinea 0.538 22217.60 3517.38 16938.12 0.594 118 Cape Verde 0.534 3305.62 191.53 1862.77 0.579 119 India 0.519 3337.37 227.19 1633.98 0.556 120 Timor-Leste 0.502 5303.20 439.10 4169.13 0.559 121 Swaziland 0.498 5132.03 656.42 2104.38 0.528 122 Lao PDR 0.497 2321.00 134.37 1008.58 0.544 123 Solomon Islands 0.494 2171.56 119.52 935.34 0.544 124 Cambodia 0.494 1867.66 108.01 625.40 0.541 125 Pakistan 0.490 2678.26 158.95 1321.20 0.535 126 Congo 0.489 3257.64 287.54 1351.59 0.524 127 Sao Tome and Principe 0.488 1917.63 102.58 788.84 0.545 128 Kenya 0.470 1627.74 105.34 464.00 0.522 129 Bangladesh 0.469 1587.24 76.96 600.51 0.532 130 Ghana 0.467 1385.47 79.92 362.56 0.525 131 Cameroon 0.460 2196.89 180.09 776.82 0.509 132 Myanmar 0.451 1595.53 85.15 655.08 0.521 133 Yemen 0.439 2386.63 146.08 1652.56 0.517 134 Benin 0.435 1499.11 78.61 659.13 0.514 135 Madagascar 0.435 953.06 40.80 245.20 0.523 136 Mauritania 0.433 2118.32 145.71 1063.74 0.498 137 Papua New Guinea 0.431 2227.10 140.01 1229.15 0.494 138 Nepal 0.428 1200.79 50.53 506.32 0.521 139 Togo 0.428 843.78 32.06 204.09 0.523 140 Comoros 0.428 1176.07 50.33 516.88 0.528 141 Lesotho 0.427 2021.15 196.36 688.58 0.487 142 Nigeria 0.423 2156.50 195.96 875.52 0.482 143 Uganda 0.422 1224.06 72.29 379.96 0.501 144 Senegal 0.411 1815.78 120.80 917.62 0.484 145 Haiti 0.404 949.00 40.11 292.97 0.496 146 Angola 0.403 4941.20 600.42 3825.19 0.462 147 Djibouti 0.402 2471.38 185.87 1585.79 0.469 148 Tanzania 0.398 1344.29 76.77 543.43 0.475 149 Côte d'Ivoire 0.397 1624.86 97.39 861.41 0.476 150 Zambia 0.395 1358.52 105.48 420.94 0.467 151 Gambia 0.390 1357.68 78.66 675.75 0.482 152 Rwanda 0.385 1190.34 76.06 465.36 0.481 153 Malawi 0.385 910.97 45.29 272.01 0.482 154 Sudan 0.379 2051.14 133.69 1491.15 0.460 155 Afghanistan 0.349 1419.08 124.65 654.38 0.437 156 Guinea 0.340 953.46 43.34 629.57 0.473 157 Ethiopia 0.328 992.03 49.59 715.36 0.460 158 Sierra Leone 0.317 808.72 45.86 314.82 0.431 159 Central African Republic 0.315 757.85 42.03 256.68 0.426 160 Mali 0.309 1171.31 79.04 983.35 0.434 161 Burkina Faso 0.305 1214.83 72.41 1184.36 0.429 29 162 Liberia 0.300 319.81 5.51 37.20 0.496 163 Chad 0.295 1066.75 68.59 831.27 0.414 164 Guinea-Bissau 0.289 538.09 22.48 177.42 0.437 165 Mozambique 0.284 854.09 49.86 672.17 0.426 166 Burundi 0.282 401.57 11.54 86.22 0.451 167 Niger 0.261 675.38 29.55 445.12 0.399 168 Congo 0.239 291.23 6.04 33.38 0.428 169 Zimbabwe 0.140 176.17 0.51 1.68 0.454 Sources: HDI and GNI from HDR web site; valuations are the author's calculations from the data on the same site. 30