Eight goals with deadlines

A numbered list that changed the architecture

The Millennium Development Goals were adopted by heads of state at the United Nations General Assembly in September 2000, in a resolution that drew on a sequence of earlier commitments — the 1996 OECD document Shaping the 21st Century among them — and on targets that had emerged from a run of UN conferences through the 1990s on health, population, food, and women. The consolidation into eight goals, eighteen targets, and forty-eight indicators was not a single act of design; it was a negotiated compression, finalized through inter-agency technical work in 2001 and 2003 before the framework stabilized.

The eight goals were: eradicate extreme poverty and hunger; achieve universal primary education; promote gender equality and empower women; reduce child mortality; improve maternal health; combat HIV/AIDS, malaria, and other diseases; ensure environmental sustainability; and develop a global partnership for development. Each had a deadline, explicitly or implicitly pegged to 2015, with 1990 as the baseline year — a choice that would shape how progress was measured and, in some cases, whether it could be measured at all.

Goal 1 set the most precise numeric target: halve the proportion of people living on less than a dollar a day. The threshold has been revised several times since, and the World Bank's published poverty counts use purchasing-power-parity adjustments that themselves changed substantially between 1993, 2005, and 2011 vintages; the $1.25 line used through most of the MDG period was not the line against which 1990 was originally benchmarked. Progress claims for Goal 1 depend on which line and which deflator the reader is applying, a fact often submerged in headline announcements.

What the clock measured — and what it did not

The 2015 deadline imposed a discipline on the field that earlier declarations had not. Agencies, donor governments, and recipient country statistical offices all had an incentive to produce numbers that tracked against the eight goals. UNICEF reorganized much of its monitoring around child mortality; the IMF and the World Bank tied the HIPC debt-relief process to poverty-reduction strategy papers that were expected to align with MDG targets; and the OECD tracked official development assistance volumes partly as a proxy for whether rich countries were meeting Goal 8's commitments.

The deadline also created a specific statistical problem: the baseline year was 1990, but the goals were agreed in 2000. Surveys conducted in the intervening decade were repurposed, modelled, and sometimes extrapolated backward to produce 1990 estimates. Where no survey existed for 1990 or nearby years, the baseline itself was a model output rather than a measured value. The baseline problem was structural, not incidental, and it meant that what looked like a measurement of change was sometimes a comparison between two estimates of different quality and provenance.

Household surveys are the primary instrument for tracking income poverty, child nutrition, and several other indicators. The reference period matters: a survey conducted in 2012 cannot be subtracted from a 1990 estimate as though both were readings on a stable instrument. Population growth, geographic redistribution, and changes in survey methodology all affect the comparison. National statistical offices, which bear primary responsibility for these surveys, operate under constraints — financial, political, and technical — that vary enormously across countries. The OECD's Development Assistance Committee data and the World Bank's PovcalNet (now the Poverty and Inequality Platform) aggregate figures that arrive from these offices at different intervals and with different vintage structures, and published global totals rest on imputed values for countries with no recent survey.

Goal 4, reducing child mortality by two-thirds between 1990 and 2015, illustrates the measurement chain more sharply than most. The UN Inter-agency Group for Child Mortality Estimation — drawing on UNICEF, the World Bank, and others — produces estimates that blend civil registration data, where it exists, with survey-based estimates using statistical models. Many of the countries with the highest child mortality rates have the weakest civil registration systems, so the model contributes more where the underlying data is thinner. The inter-agency methodology is documented and versioned, but the confidence intervals around country-level estimates are wide enough that a claimed reduction of forty percent and one of fifty-five percent may both lie within the plausible range for the same country.

The architecture the goals created

The MDG framework did something beyond setting targets. It built a monitoring architecture — a demand for comparable cross-country data on a fixed schedule — that had consequences for how statistical offices were funded, what surveys were commissioned, and which indicators attracted methodological investment. The UN Statistics Division curated the official indicator list and periodically revised it. The World Bank's World Development Indicators database expanded to carry the series. PARIS21, the statistical-capacity partnership established in 1999, became the vehicle through which donor funding flowed to national offices trying to meet the data demands the goals had created.

The architecture was not neutral in its effects. Indicators that were measurable by household survey attracted more attention and more investment than those that were harder to quantify. Environmental sustainability — Goal 7 — contained targets on biodiversity, sanitation access, and safe drinking water that relied on survey self-report for some elements and on satellite observation for others; the governance and policy dimensions of sustainability were largely absent from the measured framework. Inequality within countries was not a headline MDG target, though the disaggregation of indicators by income quintile, sex, and urban-rural location was technically possible and sometimes reported.

The selection of what to count was also a selection of what would get managed. Goal 5 — maternal health — was one of the weakest performers against its 2015 target in many regions, and it was also one of the goals for which measurement was hardest: maternal mortality ratios depend on vital registration systems that many low-income countries did not possess in 1990 or 2015. The confidence intervals on the UN estimates of maternal mortality published for 2015 were, in several country cases, wider than the change the goal required.

Eight numbered goals with a fifteen-year clock created accountability of a kind the field had not previously institutionalized. They also created a particular kind of data production — demand-driven, deadline-constrained, and global in scope — whose quality, uniformity, and reliability the scoreboard format could not itself guarantee.