Has the future arrived? Data science and choices in pension for the individual

The rise of data science has important implications for choices in pension for the individual. In this paper, we reflect on the possible benefits and pitfalls of using data science in the context of pension advice. In particular, we distinguish between different types of data, namely administrative indicators (capturing objectively identifiable data) and subjective indicators (capturing data referring to more abstract personal characteristics). We also discuss the privacy and data protection side of data science, as well as the assessment of different interests that seem particularly relevant when applying data science techniques. Four conclusions can be drawn from this research. First, there is much to be gained if pension advice or guidance for individuals when making choices for their pension makes use of linking different datasets. Whether or not such an approach always falls under the heading of data science seems less relevant. Second, caution is required when correlations are being used to inform an advice for pensions; the finding that two or more variables are correlated does not mean that they are also causally linked. Third, the decision making process for implementing data science requires good governance. It is particularly important to define clear criteria for the application of data science techniques, since often it will be very difficult to determine the desirability of applying such techniques beforehand. Finally, we argue that data science should not be treated as a black box. Indeed, a black box approach entails the risk of using variables in analyses that, in fact, serve as a proxy for other variables that would not be acceptable for use in data science analytics on ethical or legal grounds.