Data for sustainable development

Over 1,500 data experts from more than 100 countries will gather at the UN World Data Forum in Bern in October 2021. They will include representatives from national statistical offices (NSOs), the private sector, academia and civil society activists. Over discussions lasting four days, they will debate one key question: which data, statistics and methods can they employ to not just measure but also help achieve the 17 SDGs (see box)? The eight Millennium Development Goals (MDGs) from 2000 to 2015 contained 21 targets and 60 indicators.

The SDGs for 2015 to 2030 already have 169 targets and 231 indicators. These are intended to track global development more comprehensively and with greater granularity than ever before. At the same time, they take into account that sustainable development can only result from the confluence of many factors.

There is, however, one problem: most countries do not regularly collect data for even half of the indicators, as is stated in the introduction to the UN Sustainable Development Goals Report 2019. "In some countries, even the most basic indicators, such as total population and child mortality, are not measured regularly," says Francesca Perucci, assistant director at the UN Statistics Division (UNSD) in New York. There is a huge data vacuum: 18 low-income countries did not conduct a census or population survey from 2009 to 2018, making it impossible to assess progress on poverty.

Just 50% of countries calculate GDP growth on the basis of updated benchmarks, leading to economic stimulus programmes being based on false assumptions. Data gaps are even more glaring for sustainable development indicators relating to the environment, such as water quality or deforestation, despite the fact that the availability of high-quality data is itself part of the 2030 Agenda (in target 18 of SDG 17).

This has far-reaching implications. "Lack of data most affects the most vulnerable," says Perucci. A birth or marriage certificate is usually the basis for a legal identity, which opens doors to government services, such as healthcare and financial support. "Street children, persons with disabilitieis, small-scale farmers in remote areas, indigenous peoples – all these groups are not reflected in national statistics even today." They are buried within aggregated data and the averages derived from these datasets, explains Perucci.

The living conditions of such groups, e.g. difficulties in access to water, healthcare or education, remain invisible. This is in stark contrast to the 2030 Agenda's commitment to leave no one behind. Critical to this endeavour is the availability of reliable data that show where there is need for action and where interventions would be most effective. And yet, "Investments in statistical tools are often not immediately visible in the number of lives saved," says Perucci. "So for many decision-makers they are not a priority."

The magic bullet for better representing the most vulnerable groups in statistics is disaggregation. Data need to be broken down for sub-groups with different circumstances depending on gender, income, education and geographic location. However, this requires expertise, good infrastructure and financial resources. PARIS21, an international organisation that promotes statistical capacity building in developing countries and is co-funded by Switzerland, has calculated that USD 700 million will be needed annually for developing and emerging countries to develop their national statistical systems to a reliable level.

It is a paradox: there was never as much data available as there is today, and yet in many contexts there is a lack of essential information. Governments have long ceased to be the only producers of data about their populations and administrative territory. Equally, there are private actors such as mobile phone companies, search engine providers and social media platforms. The patterns of our shopping, travel, online searches, reading habits, film preferences, emails and social media posts all leave 'digital footprints' on servers around the world. Suitable algorithms can sift through this ocean of unstructured data for patterns and information. Although somewhat controversial, such big data analytics are already being used to strengthen healthcare systems, combat pandemics (see box on COVID-19), optimise public transport and detect financial crimes.

Big data analytics is also highly promising for the 2030 Agenda. It can help close some of the large data gaps in national statistics. Recognising this potential, the UN Global Pulse initiative has set up its own centres for developing new tools to harness big data and artificial intelligence (AI) for sustainable development. For instance, by analysing money transfers and purchases through mobile payment services. In Kenya, almost half the population uses M-Pesa, a private digital payment platform. If used properly, the data so collected could supplement statistical surveys on income and poverty (SDG 1/No poverty).

By collating satellite images with media and eyewitness reports, deforestation (SDG 13/Climate action) can be better tracked and expressed numerically. For such applications, the experts at Global Pulse work with data from social networks, mobile network providers, transport and postal companies as well as satellites. The latter are especially promising for environmental monitoring. In a detailed report in 2018, the European Space Agency (ESA) presented ideas and examples for using satellite data to measure the SDGs.

A UN global working group with Swiss participation has been working for some time on the potential uses of big data. Experts in the group discuss questions relating to methodology, quality, technology, access, legal aspects, privacy, management and funding. In its Big Data Project Inventory, an online catalogue, it has listed over 100 projects by national statistical offices, universities, UN agencies and other multilateral organisations, e.g. the World Bank, that are relevant for the 2030 Agenda. The potential advantages are obvious: lower costs compared with conventional statistical methods, real-time data collection, automation and more granular data.

Nevertheless, there are also reservations about big data analytics. The mere fact that 'only' 55% of the world's population has access to the internet is problematic. Women's use of mobile internet continues to be far lower than that of men. This digital divide, which is even more pronounced between genders, income groups and the global north and south, could result in important population groups being completely ignored. But this is precisely what the 2030 Agenda is trying to avoid at all costs.

In an article that appeared last year in Global Policy, Steve MacFeely, head of Statistics and Information at the United Nations Conference on Trade and Development (UNCTAD), drew attention to the fact that big data, including credit card, mobile phone and search engine data, is often proprietary and owned by private companies. If the UN or national statistical offices wish to use this data, it could turn out to be very expensive, or the organisations run the risk of violating property rights. Moreover, the algorithms that often underpin big data are extremely valuable commercially, and their owners are usually not interested in transparency.

There is another important factor: statisticians love continuity. When private companies adapt algorithms for their own interests, statisticians can lose the ability to measure indicators over long time periods. This kind of dependence is risky, especially because data production on the internet is heavily concentrated in a few platforms. In 2017, Google had an 88% market share in online searches, Amazon had a 70% share in e-book sales and Facebook had a 77% share in mobile social media. There is consequently a risk of manipulation and abuse. Yet for national statistical offices, independence and public trust are prized assets. In times of fake news and post truth, the integrity and credibility of such institutions is more important than ever.

Georges Simon Ulrich, Director-General of the Swiss Federal Statistical Office (FSO), is critical in his views as well. "99.8% of available data are not standardised. They do not provide the information that we as statisticians seek," he says. Data must be reliable and valid, and they must provide stable and comparable results over the long term – only then are they of statistical value. This is not a given for many big data applications. Ulrich is responsible for measuring Switzerland's progress on the SDGs and for coordinating internal data flows to the UN under the 2030 Agenda. His office is at the same time involved in a series of activities for international exchanges between statisticians.

Interoperability is an important quality criterion for data. However, as Ulrich notes, there is often no guarantee that data for different stakeholders can be used at different levels. "I would therefore want the UN to continue expanding its role as trusted global data broker and custodian." Ideally, he would want independent and competent national statistical offices, such as the FSO, to ensure the interoperability of public data, and for the UN to assume the same role at a global level, while also being responsible for setting meta-standards.

However, the UN needs to step it up in terms of speed, flexibility and communication, otherwise it risks that data sovereignty will be lost to private entities such as Facebook and Google. "The UN possesses the best data in the world today," Ulrich is certain. "Unfortunately, very few people are aware of this." At the UN World Data Forum in Bern in October 2021, he and hundreds of colleagues from all over the world will have an opportunity to change that.