Fourteen years ago, as preparation for the 2002 Monterrey Conference on Financing for Development, two colleagues and I wrote a paper estimating the costs of achieving the Millennium Development Goals. While the paper was widely cited and our estimate—an additional $50 billion—was the same as that adopted by the conference (see for example Kofi Annan’s speech), I have been feeling increasingly uncomfortable with some of the methods we used to arrive at an estimate. A number of thoughtful people have also criticized us (here and here). I now think the whole exercise was wrong on at least three counts.
- It was the wrong question. When you set a personal goal, such as losing a certain amount of weight in a year, you start with all the actions you will take—joining a fitness club to exercise regularly, eating certain foods, and so on. Of course, some of these activities will cost money, but you never start with the costs. By contrast, with the MDGs, once the goals were established, the focus very quickly shifted to the costs—how much additional aid would be needed to achieve the goals? Everyone recognized that it would take more than money, but by starting with the costs, the discussion of all the policy and institutional reforms needed to accelerate progress—not to mention make the additional resources productive—became an add-on rather than the centerpiece of the debate. Furthermore, the policy and institutional reforms were difficult to measure and monitor, whereas it was easier to see if aid had increased by $50 billion. This gave the misleading impression that, if the additional resources were not available, the MDGs would not be reached. In the event, we learned that the countries that made the most progress towards the goals were not necessarily the ones that received the most additional aid. Rather, strong economic growth and improved policies and institutions were the main factors. Instead of “how much would it cost to reach the MDGs,” the right question would be: “If you reached the MDGs, how much would it have cost?”
- It was probably the wrong answer. When we were (ahem) invited to undertake the costing exercise, we wanted to avoid the double-counting that plagued previous exercises that calculated the cost of achieving each individual goal—and then added them up. So we calculated the cost of reaching the first goal, a 50 percent reduction in the 1990 poverty rate by 2015, only. The estimate was around $50 billion. We recognized that, in reducing poverty to this level, progress towards the other goals—primary education, child survival, etc.—would also be made. Nevertheless, we separately calculated the costs of achieving the education and health goals from the bottom up, by estimating the unit costs of children in school, health care, and so on, and scaling up to the amount needed to reach the goals. This also came to about $50 billion. Again, we recognized that reaching the primary education and child survival goals would go a long way to reaching the poverty-reduction goal. That’s when we decided that, since the goals were complementary, we would completely avoid double-counting (and satisfy the various constituencies for each of the goals) by concluding that the cost of reaching the goals was $50 billion. This may have been an underestimate, but we had not one but two different ways of arriving at the same figure.
That is not to say that each of these two methods was foolproof. We estimated the costs of reaching the poverty-reduction goal by first calculating the growth needed to reduce poverty in each country (applying an estimated growth elasticity of poverty), and then calculating the investment needed to achieve the growth using a fixed incremental capital-output ratio. Neither poverty reduction nor growth follows these linear patterns, but it was the best we could do for all developing countries.
Similarly, calculations of the cost of the human-development goals involved, in the case of child mortality, looking at cross-country regressions of the determinants of child survival, taking the coefficient on health spending, and essentially inverting it (if $1 of health spending leads to x children surviving, then to reach the goal of, say, 100 children surviving would require $100/x). The only problem was that, in almost all of these cross-country regressions, the coefficient on health spending was not significantly different from zero (the reasons can be found in this paper). So the confidence interval of the cost estimate could be between zero and infinity. (It could also have been negative, especially if the policy and institutional reforms involved cutting wasteful subsidies to improve service delivery, but we chose not to mention that.
It is the wrong way to approach development. Poor people are poor because they are stuck in a low-level political equilibrium. When teachers are absent from the classroom 25 percent of the time, students don’t learn. But the same teachers run the electoral campaigns of local politicians who, in turn, give them a job to which they don’t need to show up. Doctors are absent from public health clinics at least as much as teachers—and when present, spend an average of 39 minutes a day seeing patients—because they prefer to work in the fee-paying private practice, and medical unions are politically powerful enough to resist reform. Poor people have to pay high prices for off-grid infrastructure (water tanker operators, candlepower) because politicians perpetuate subsidized water and electricity in order to control whom the utility services. And they lack jobs because of exorbitant transport prices (thanks to protected trucking monopolies) or other monopolies granted to politically connected firms that stand in the way of export competitiveness.
In this setting, approaching development as a problem of finance—the amount of money it will take to achieve the goals—can be counterproductive. From the donor’s side, a focus on raising the $50 billion in resources distracts from investing in the knowledge assistance needed to help unblock these political equilibria. And from the government’s perspective, many of the reforms that are needed to accelerate poverty reduction are politically difficult. Discussions of financing needs enables policymakers to avoid these difficult reforms, while giving them an excuse for missing the goals (“the money was not enough”).
In short, by costing the Millennium Development Goals, I may have helped shift attention away from what is needed to reach the goals, and hence contributed to the perpetuation of poverty. Shame on me.