The power of mathematical models for better policy decisions

“Harnessing the power of mathematical models for better policy decisions” sets out four practical recommendations to help policymakers across a wide range of policy areas effectively capitalise on, and sidestep pitfalls of, using mathematical models for decision-making.

Decision-makers are often keen to “follow the science” in highly-charged contexts such as climate policy, pandemic response, economic policy and humanitarian crisis response. In situations like these, where decisions are informed by complex modelling or simulation and there is inevitable disagreement and uncertainty, it is unclear what “following the science” really means.
Some existing paradigms of model use and interpretation are beginning to reveal themselves as unfit for purpose in the present age. We can see this in the roots of the financial crisis in 2008 (over-confidence in models), public discourse on the developing climate crisis (no one quite knows how to interpret different model results), and the present pandemic crisis (where various models appear to be driving policy in conflicting directions). These are not coincidental: as computers become more powerful, and models become more complex, the role of the model within the scientific process has become larger and larger and so has the power of the model to influence decision-making.
Limitations and biases of the scientific approach in decision-making contexts have been studied extensively, but the specific role of mathematical models is becoming more and more important.

The results and outcomes of this project will be directly relevant to UK decision-makers in government, business and the third sector, with potential to be transformative for the use of models as decision support tools.

1. Communicate the purpose for seeking models
2. Promote organisational structures which facilitate dialogue
3. Understand key assumptions in the mode
4. Assess limitations of models for decision-making

Quality assurance guides outline the role of the ‘analytical assurer’ for translating between policy context and technical details.
Such translation is most effective and efficient when conducted through real-time dialogue among modellers, policy stakeholders, and knowledge integrators. Importantly, these are active deliberations and negotiations, for both the model users and producers to spotlight assumptions, align on trade-offs, and correct misinterpretations.
Such discussions are best done iteratively and are especially critical during initial technical scoping (to ensure the models are fit-for-purpose) and model output delivery (to ensure the relevant insights are distilled and recommendations for action are sound).

MODELS EMBODY EXPERT JUDGEMENTS
Mathematical models to forecast transport demand are one of many inputs to transport policymaking. As a simple example, one way of building the model may assume that people primarily respond to economic incentives and choose the lowest-cost method of transport. Another modelling approach may assume that access to certain services takes priority over cost, such as step-free access, internet connectivity or other safety and convenience considerations. Yet another assumption might be that the remote work trend will increase, reducing overall transport demand. Different mathematical models might embody one or several of these expert judgements, likely resulting in different pictures of future transport demand. They also carry different implications for policy decisions and metrics for policy success.