The State of Development Data

Imagine you've just been appointed Minister of Finance. Tomorrow morning, you must decide where to build schools and clinics, how much to borrow, which programs to prioritize. Every decision will affect millions of lives. To govern well, you need answers to basic questions: How fast is the economy growing? How many children are healthy? How many young people are unemployed?

The answers to each of these questions lie in national accounts, population censuses, health surveys, household surveys, labor force surveys, and other types of data. Now imagine opening your briefing folders and discovering that much of the available data are years old.

Every day, governments make decisions that affect millions of lives, such as setting budgets and regulating the scope and quality of services. Yet for many economies, the data guiding those decisions are years out of date, leaving millions of people invisible to policy makers.

Take data on unemployment, for instance. For the average low- and middle-income economy, the latest labor force survey is from 2019, the most recent poverty survey is from 2020, and the most recent health survey—measuring everything from child vaccinations to malnourishment and access to health care— is from 2015, more than 10 years ago.[footnote: The latest labor force survey is calculated as the median year for all low- and middle-income economies. An analogous approach is used for poverty, health, and other surveys.] So, most of the surveys informing our current understanding of labor markets and public health are from before the COVID-19 pandemic.

In more than 30 low- and middle-income economies, the latest labor force, health, business, agriculture, and[emphasis: ]poverty surveys [emphasis: are more than five years old]. In another 46 economies, four out five of these surveys are more than five years old. Together, this amounts to around 60 percent of all low- and middle-income economies. So, when these economies try to make data-informed decisions, they are charting their future using an outdated picture.

When survey data are missing, policy makers rely on models. But these, while useful, can miss the truth by a wide margin, as illustrated by [emphasis: Nigeria's] poverty data.[footnote: Poverty estimates are drawn from the World Bank's Poverty and Inequality Platform (PIP). The 2018 figure (34.2 percent) and the extrapolated 2022 figure (34.8 percent) are based on Nigeria's 2018 Living Standards Survey (LSS), with the extrapolation projecting poverty forward using national accounts (gross domestic product or GDP) data. The survey-based 2022 figure (41.8 percent) was incorporated in PIP's Spring 2025 update, following the release of Nigeria's 2022 LSS. The seven percentage point gap between the extrapolated and survey-based 2022 estimates illustrates the limitations of nowcasting as a substitute for timely household survey data.] In Nigeria’s case, any programs targeting households by poverty status would have missed millions of households.

The timeliness of statistics is not the only issue. In many cases, the methods and data used to calculate statistics are outdated, producing misleading estimates. An example from Ghana illustrates this starkly. When it updated its national accounts framework in 2010, incorporating new industrial census data, value added tax records, and household surveys and modernizing the way it classified the services sector, [emphasis: reported GDP rose by more than 60 percent].[reference: devarajan][reference: jerven_duncan] Three-quarters of that increase came from the services sector alone, which had been systematically undercounted under the old framework. But many economies continue to use old methodologies, with 52 still using a 1993 manual to report GDP statistics today.

The absence of data also severely limits an economy's ability to understand the needs of important populations, such as children, leaving policy makers unable to address the most pressing challenges they face.

Without these data, it is impossible to know how many children live in a country and whether they are healthy and learning. And without such basic information, how can a government plan where to build schools and clinics, how many teachers and health care staff to hire, or what kind of programs are needed to address nutrition or skill deficits before they become permanent?[footnote: Birth registration data measure the share of children under five whose births have been officially recorded with a civil authority. Stunting (low height-for-age) is the primary indicator of chronic undernutrition in children under five. Learning assessment figures refer to internationally comparable assessments conducted since 2019.]

The SPI measures the performance of statistical systems across five pillars using 51 indicators that cover everything from whether surveys are conducted to whether key data are produced and used, data are accessible online, and the legal foundations and financial resources for statistics are in place.[footnote: The five pillars are data use, services, products, sources, and infrastructure. For data to have an impact, they must first be used. The [emphasis: data use ]pillar also measures whether data meet the needs of key users of a statistical system, such as the Minister of Finance. The [emphasis: data services] pillar measures whether services, such as online accessibility, are available to connect users to data. The [emphasis: data products] pillar measures whether the system produces key statistics, including those concerning the well-being of children, to meet the needs of users. The data[emphasis: source]s pillar tracks the availability of foundational data sources, while the [emphasis: data infrastructure] pillar assesses the hard and soft infrastructure, such as statistical legislation, financing, and the methodologies that underpin statistical production.] It includes an overall score that measures statistical performance on a scale from 0–100, where 100 represents the best possible statistical system.[reference: dpss]

An economy’s income tends to be correlated with the performance and maturity of its statistical system. But this relationship is not deterministic. Many economies perform at the same level as far richer economies, providing valuable insights for how to improve statistical systems in the places where it is most needed.

The information available to guide policy is often incomplete, outdated, or uneven in quality. These constraints are not distributed randomly. They tend to be most binding in low-income and FCV settings where the demands on the state are greatest and the costs of policy error are highest. It is precisely in such settings that rapid shifts in economic conditions, prices, displacement, and service delivery can quickly render old statistics misleading.

But the evidence also cautions against a deterministic view of capacity. Several countries, such as [emphasis: Mexico], [emphasis: Burkina Faso], [emphasis: Senegal], [emphasis: Uzbekistan], and the [emphasis: Philippines], achieve levels of statistical performance comparable to peers with far higher GDP per capita.

The implication is straightforward but important: while resources matter, income alone does not determine statistical effectiveness. Institutions, incentives, and sustained investments in statistical systems are critical for good performance.

For policy makers and development partners, this finding points to a need to organize improvements in statistical performance around strengthening data sources. This includes traditional instruments, such as household surveys and administrative systems, and complementary sources, such as geospatial and other high-frequency data. At the same time, modernizing methods for integrating, validating, and using these inputs is vital.

Emerging technologies, including artificial intelligence, offer genuine opportunities to reduce the costs of data collection and processing and expand data use by improving discoverability, timeliness, and dissemination. But realizing these gains will require deliberate investments in governance, skills, and quality assurance, as well as technology.

National statistical offices can rarely meet these demands in isolation. Building durable foundational and frontier statistical capacity will increasingly depend on international partnerships that support training, peer learning, and the adoption of best-practice standards. For a finance minister tasked with making decisions under uncertainty, such investments are not ancillary. Rather, they are part of the core infrastructure of effective economic management.