Data Science for Public Policy: How to Achieve Long-Lasting Impact

Executive Summary

Data science has thus far played only a minor role in public administration and governance, unlike in the private sector. However, data-driven approaches hold a lot of potential: data-driven tools can help ensure that administrative or political decisions are made on a better factual basis. Moreover, with the help of data tools, states can automate processes and utilise government resources more efficiently.

In Germany, the federal government has recognised the possibilities of data in governance. Its data strategy, published in 2021, sets a goal to significantly “increase innovative and responsible data provision and data use”. Among other provisions, the plan allocated around €240 million to establish data laboratories in all federal ministries and in the Chancellery itself by 2024.

However, for all of its potential, the establishment of data departments in administrative bodies also poses numerous challenges. Fundamentally, data scientists in this sector bear special responsibility: their work often affects large swaths of the population and can have an impact on critical areas of life over a long duration. In addition, there are hurdles to overcome in the practical implementation of data units. Data teams do not always succeed in finding suitable partners within their organization for the purpose of working together. Rigid administrative processes can also be burdensome when implementing data science techniques. And at the same time, high transparency standards accordingly impose constraints and limitations.

This impulse paper is meant as a guideline. In it, we address the question of how to successfully integrate data labs into public administration by compiling a series of concrete recommendations. Above all, we are guided by the maxim that data science in politics and administration should never be an end in itself. Rather, when developing data-driven products, data teams should focus on users and the concrete impact their products make in all phases of the process, from brainstorming to evaluation.

Data scientists in governance must fully understand their roles. They do not only have a technical task, as communication skills and interdisciplinary thinking are just as important for the success of their work. Data scientists in the public sector must be able to “translate” political and administrative problems into data science use cases that span a wide variety of users. They must be able to explain and evaluate their data products in an accessible manner. Furthermore, they need to have an awareness of the risks of their methods: undetected biases in datasets can have major consequences–to the detriment of entire segments of populations.

Our paper follows the development of data products along the four phases of Ideation, Prioritization, Implementation, and Evaluation. Each chapter includes a list of the key “Do’s and Don’ts” of building data labs. The final chapter focuses on recruiting and developing data teams. In the public sector, it can be difficult to find the right talent because data labs require a combination of skills and must compete with the private sector. We recommend making positions more attractive through cross-organizational dialogue and development. The search for talent should focus on candidates who have a strong orientation towards the public good and an interest in the greater social context of data work.