By Steve MacFeely (Steve.MacFeely@oecd.org) and Ashley Ward (Ashley.Ward@oecd.org), OECD Statistics and Data Directorate

For many years Governments and international organisations have publicly espoused the advantages of evidence-informed decision making. But in recent years we have witnessed a subtle but noteworthy shift in rhetoric away from evidence-informed in favour of data-driven decision making. At first glance this evolution seems welcome – decisions are now driven by data, implying those decisions must be objective and scientific. It rings of progress and modernity.      

But is this the case? Evidence-Informed decision making acknowledged and made transparent the judgements and trade-offs involved in democratic decision making. Data-Driven decision making, however, can be interpreted in several ways. Do data drive policy debate, with the eventual decision being “data-triggered”? Or does data, in and of itself, take the decision, the eventual decision being “data-determined”? Recent rhetoric seems increasingly to indicate the latter, with decision-making resulting from a more technocratic, (datacratic if you will) approach, suggesting that data alone are sufficient to make decisions once guiding parameters have been set. It is in today’s parlance, an algorithmic approach.  

It is not always clear how those decisions are made, how they accommodate unusual circumstances, or what input data were used. In short, there are issues concerning transparency. As decision making is automated, the process may slip into the shadows, hidden from view, so much so that we may not even be aware of the decisions being made.  

This is concerning as algorithms may reinforce and amplify, often unconscious, biases and stereotypes. They may perpetuate prejudices and errors through the use of flawed, human-generated data. These biases, especially in a data-determined algorithmic world really matter, as they may hard code discriminatory decisions into public administrations and private sector organisations, without any transparency or recourse (for example, whether you are successful in securing a social protection benefit or affordable health insurance).  

Principle 1 of the United Nations Fundamental Principles of Official Statistics states that official statistics provide an indispensable element in the information system of a democratic society. The OECD Recommendation on Good Statistical Practice too considers that quality statistics are an indispensable tool for good analysis, transparency, accountability and for informed decision-making and the functioning of democracies. These statements rest on the Churchillian ideal that statistics should be the neutral arbiters in political debate and are an essential pillar in the deliberative public space.  

In a democracy, citizens elect leaders, entrusting them to make important decisions on their behalf. Official Statistics are one important tool in the armoury of those elected decision makers, providing an objective benchmark to ground political debate and assess alternatives. A data-driven world or a data-determined one, threatens to remove a fundamental step from that democratic process, by removing the decision maker from the decision. Algorithms obey programmatical codes, not codes of ethics.  

Official Statistics are produced to inform decisions, not to make them – they are produced to support democracy, not to replace it.