An opinion piece by David Menzies, Modelling Lead for the National Infrastructure Commission.

To a large extent, how you heat your home is your decision. For most people it’s also a decision taken when their existing heating system is at or near the end of its life.

Will you keep gas central heating? Will you decide to replace your oil-fired boiler with an air source heat pump? What do you know about alternative technologies, and what factors might you consider when deciding?

This uncertainty, driven by the diverse decision making of millions of individuals, means that modelling approaches which aim to identify optimal outcomes – the “best” route to deploy new heating technology, for example – don’t tell us what might happen, only what can happen.

After all, if your boiler breaks down, it’s unlikely you’ll consider what’s cost optimal for the energy grid when deciding on a replacement, or perhaps even what’s most cost effective for you over the long term.

You might take other factors into consideration – for example, the affordability of any upfront cost, how much you value environmental factors, how much you know about a technology, the availability of someone to install the technology, and how disruptive the installation will be.

Agent-based modelling – an alternative approach

At the Commission we believe in testing alternative approaches to exploring infrastructure challenges.

Agent-based modelling (ABM) is one such alternative approach, and in preparing for its second National Infrastructure Assessment, the Commission seized the opportunity to look into it.

The move to decarbonise the provision of heat to homes means that most of us will need to take a decision on how we will heat our home over the coming years. There are a lot of us too – there are around 27 million households in the UK.

ABM can be used to explore what technologies might be adopted by households as a result of individuals’ (agents’) decision making. It introduces interactions between agents – gives them the ability to influence each other and learn from experience – and lets us explore the outcome of these interactions.

Interactions can magnify the impact of decisions or lead to different outcomes to those expected. ABM can point us towards the areas of uncertainty which should be explored further, as well as help us understand whether there are a feasible set of circumstances under which an outcome might be reached or otherwise.

How influential on your decisions might the experience of other people be? Under what circumstances might there not be widespread adoption of electric heat pumps? Are there any scenarios where different technologies can co-exist? How large an impact might the availability and experience of installation engineers have?

These things are of course impossible to predict with any certainty. However, understanding how factors influencing decisions might affect outcomes, identifying feedback loops in the decision-making process and understanding the plausibility of desirable outcomes, can help us understand what we need to do, or shouldn’t do, now.

Three things we learnt from the ABM…

• Feedback effects: decisions taken by one group who have been targeted by an intervention (such as a marketing or price incentive campaign) can materially influence the decisions made by individuals beyond the group initially targeted. In one of our modelling scenarios an additional 50% of households were influenced beyond the initial group targeted. Exploring feedback effects can help to maximise the breadth of a policy’s impact and avoid designing a policy which may have unintended consequences.
• Distributional impacts: The modelling approach necessitates that we define how different households make decisions based on different factors. This allows us to explore how different groups might be affected by policy. In our modelling we separated householders that own their properties from those that rent. A key difference between the two is that the decision maker for the rented property is the landlord rather than the occupier, who will place a greater weighting on up-front costs than an owner-occupier.
• Barriers to change: the modelling shows that upfront costs and the “hassle factor” are likely to be a material barrier to changing heating technology. It might seem obvious but illustrating how significant affordability (or availability of credit) is compared to other decision-making factors can help guide policy.

…and a note of caution

I say “can help guide policy” quite deliberately. ABM should not be viewed as a predictive tool, but its outputs are better read as a menu of possibilities. Like all models, it relies heavily on assumptions inputted into the calculations. But ABM can help highlight areas that warrant further testing by other methods, and ‘sense check’ emerging policy ideas.

If what you’ve read here and the questions we have posed have piqued your interest, the Frontier Economics’ report contains much more information about ABM and the insights gained from the project.