Public programs are only as good as their data.

Data scientists will 2023 will be a good year as governments invest heavily in the application of AI and algorithms in public policy. The European Commission has allocated 1.3 billion euros ($1.38 billion) for research and innovation under the Digital Europe programme. The UK government is funding £117m ($143.6m) for doctoral students in artificial intelligence, and is now in its second year on its 10-year plan to “make the UK into a global AI superpower”. Examples of current initiatives include the use of AI by the National Health Service to detect abnormalities in CT scans and the Department of Labor and Pensions’ efforts to detect fraud in universal loan applications.

While the promise of these technologies is impressive, new tools will only be useful if the data they enter is accurate and complete. However, in 2023, most government data will still be inaccurate or full of holes.

For example, outside of the UK census year, there is no accurate data on population size, the extent of immigration or the nature of inequalities affecting groups such as ethnic minorities and the LGBTQ+ community. More than 15 percent of land owned in England and Wales remains unregistered, which means we still don’t know who owns large swaths of the country. The UK statistics agency removed the status of “national statistics” from reported crimes because the measures used to track them were so inaccurate. Similarly, there is still no consensus on how to measure poverty in the country, which makes it difficult to solve this problem.

Bad data has also caused many major political failures, waste of public funds and harm to people’s lives. Bad data is why people in the UK have been wrongly deported and accused of being illegal immigrants, as happened during the Windrush scandal. False data was behind the Dutch child benefit scandal, where benefit recipients were falsely accused of fraud because a government algorithm was programmed to identify dual nationals as the most likely to commit a crime.

The reality is that when it comes to collecting and analyzing national statistics, many governments around the world are severely under-resourced. Worldwide, one in four children “does not exist” – their births have never been registered. Only eight of Africa’s 54 countries have fully accurate mortality data. Large parts of the globe remain unmapped digitally; in India, only 21 percent of the road network is digital. More than half of the world’s countries still don’t have up-to-date data on eight of the 17 Sustainable Development Goals—targets to improve people’s lives that all UN countries have agreed to try to achieve by 2030. Without data, progress is impossible.

The prospects for AI and big data analytics in areas such as healthcare will be greatly diluted if existing government data is outdated and of poor quality. Personal intent data, such as data from mobile phones and internet traffic, can fill in some of the gaps, as happened with governments during the Covid-19 pandemic. But private company data is itself flawed and generated without the transparency and accountability the government promises. For example, when the Israeli government began using mobile phone records to track people’s movements in order to better understand the spread of Covid-19, its Supreme Court found the initiative a breach of privacy.

However, there will be gradual progress in 2023. For example, the UK National Health Service has announced a project to address data gaps in ethnicity. The Democratic Republic of the Congo will also conduct its first census since 1984, a challenging task that will provide valuable information about some of the world’s poorest people. These are steps in the right direction, but there is a long way to go.