What does the smart mobility transition mean for modelling?
The opening plenary session at Modelling World asks whether rapid change means new models? In advance of my talk there I reflect on this question drawing on the work of the Commission on Travel Demand and the DEMAND Centre.
The first part of the question implies that we are in a world of rapid change. The rise of Uber, increasing autonomy of vehicles, electrification, on-line shopping… the list goes on. Change is happening for sure – but how rapid it is and how much of it is defined by changes to transport technologies seems much more open to question. I argue here that there are a number of trends, many of which our traditional approaches to understanding demand did not anticipate, that have been on-going slowly but for a period of time. Even when we see significant technological change coming to transport this will take some time to happen and when it does it will do so in ways which we currently would not anticipate. There is time to reflect on what sort of changes these new systems might engender but there is a need to start now given that we can be trying to support decisions that meet the needs of users 20 and more years into the future.
The second part of the question asks whether we need new models. Here I will argue that is the range, scale and nature of the uncertainties that should pre-occupy the discussion about the need for new models rather than a concern with rapidity of change. I will suggest that decision-making processes needs to accept that there does indeed exist a range of potentially different, yet still plausible, futures for how we get around in order for the focus of model development to address these challenges in a meaningful way. Models are decision-support tools and if we consider the questions which they are supporting to be different to those we have been examining then adaptation and/or innovation will be necessary. I will discuss why that is currently a challenge.
So, how rapid might change be and change in what? Looking back over the past decade, at an aggregate level as captured by the NTS for England, we have seen a decline in trip rates, trip distance and, more recently, the amount of time spent travelling.
Figure 1: Declining per capita annual travel trends (Data from NTS0101 2016 release)
There have been various important studies looking at the reasons behind these changes (see RAC Foundation, Independent Transport Commission and Department for Transport for start points). Some of the change is attributed to later uptake of driving licenses amongst younger people, some to rising insurance costs, increasing urbanisation, the changes to company car tax rules and a shift to virtual activities. However, none of these provides a complete explanation (see the work of Noreen McDonald), nor should we expect to find all of the answers within the transport sector. For example life expectancy is increasing by around 3 months each year and the average age at which women have their first child has risen by around one year each decade since 1975. If we look at wider social change which will impact on traffic patterns in the long-run then decisions such as raising the pension age, reducing pension levels, tightening immigration, healthcare innovations and continued growth in the pace and quality of communications infrastructure are all going to be very significant for mobility, but we cannot anticipate all of these changes by calibrating on historic data.
One interesting pattern to be emerging more strongly appears to be the decline in traffic in major urban areas excluding motorways (see Figure 2). People here seem to be able to assemble their patterns of mobility to meet their needs in less car intensive ways. Some of this results from growth in housing and employment in city centres which are associated with lifestyles which are less car dependent. Will these trends continue? Will they apply largely to those households without children? Does it mean car later rather than car never? Will car later mean car less (less intensive use of cars when they are owned)? It is in answering these sorts of questions that it becomes relevant to understand how Uber or Car Clubs are being incorporated into the everyday mobility patterns of people in different places. Many transport systems work much better with higher densities of demand and we might expect greater divergence in the degrees of multi-modality of people in different areas.
Figure 2: Changes in motor vehicle kilometres by area of Greater Manchester (1996 = 100). Source: TfGM submission to the Commission on Travel Demand
Given some of the gaps in understanding of how travel patterns have been changing, it seems important to try and understand how any new mobility options might be being accommodated in how people conduct their everyday life. If there is already evidence of travel being put together in different ways to those expected then we should avoid simply bolting on new elasticities and modal preference constants as a means of representing these changes. Questions such as whether people will send children to school in autonomous pods seem at least as important as to whether they can surf the web whilst on the motorway and therefore have a different time penalty on travel. It is important to remember that early adoption is not the same as mass adoption and we should be investing more now in understanding how they might influence what people do.
We should also be mindful of the longer term trends which were shown in Figure 1. Why will new technologies encourage people to start making more trips or to travel for longer? What will they stop doing if they are spending longer in vehicles? Whilst the substitution narrative is appealing it is unlikely to represent the full range of adaptations that occur (Did people use mobile phones just to make calls they used to do on their landlines?). There is also a tendency and a temptation to focus on the importance of the changes to transport modes and to forget that the other influences which have seemed so important to recent trends may continue or change in equally important ways. Will these changes accelerate or slow the adoption of some innovations as their mobility needs they might fulfil change? All of this poses some really interesting challenges to one of the key tenets of our approach to modelling and evaluation which is a reference case or business as usual. This would seem to be becoming more unusual.
So, to the second question, do we need new models? The answer to this, in the light of the potential for a quite different type of transportation system and ways of living, would seem to be a self-evident yes. How could modellers fail to try and incorporate and represent some of these changes? However, I would argue that we need to ask what we are using our models for before we jump in to improving them. All models are attempts to represent a complex world in as simple a way as is effective. Whereas the dominant influences of the decades up to 2000 could be seen to be those surrounding the growth of automobile use this is no longer the case. Given the range of uncertainties there seems to exist a range of future pathways which could diverge (and have already done so) from previously anticipated travel demand growth pathways. What does better modelling mean?
I am currently working with Glenn Lyons to draw the work of the CIHT Futures project together with that of the Commission on Travel Demand. One of the key areas we are grappling with is what models are being used to inform. As in Figure 3, we would suggest that what we need to aim for is a set of tools which provides insights about a range of plausible futures (left most image) – i.e. the dots are spread quite wide but around the centre of the target. This would imply that we take decisions that are robust in a range of futures (or that decision-makers understand that there would be circumstances where a particular course of action was not sound). If it is accepted that there is a lot of uncertainty then to continue to focus on one future and to develop our tools solely with the aim of testing lots off sensitivities around this creates an impression of accuracy which could be misleading (the centre image). The image to the far right would be some kind of Nirvana of great foresight. Outturn results of existing exercises suggest that we have never really lived there.
Figure 3: What are we aiming for?
So, I would suggest that there are parallel (but not unconnected) streams to be pursued in the future development of modelling. The first is a strand which develops a range of plausible demand futures. These will each develop different notions as to how society will change and how new technologies might become embedded within that. Where there is a need for more in-depth modelling then this will involve further development of modelling tools. This may be relatively minor such as understanding how smart technology impacts on interchange or more fundamental such as developing models around on-demand mobility access rather than car ownership. However, there will be a need to take an honest look at the requirements which the different futures suggest in order to determine whether current tools will continue to remain fit for purpose in representing these. Again here, there will be a growing need for transparency in terms of areas of limitation and uncertainty in how these different socio-technical futures are able to be represented by the tools at hand.
However, not all of the innovation or case for change can rest with the modelling community. Models are developed to find answers to policy questions. The current emphasis on being able to justify the number of jobs which a scheme will create or the degree to which agglomeration benefits or land value uplift will accrue itself implies a clear handle on these relationships and how they will evolve in the future. I think even here policy is running ahead of where experts in those fields believe the evidence is today. If the policy-making environment is not able to relax its obsession with a single BCR approach (down to the nearest decimal point) and consideration that the main way of supporting growth is delivered by the type of investments which supported the current travel paradigm then it is difficult to see why the mainstream modelling community would shift its developmental focus. It is necessary for there to be sufficient professional consensus, and then policy acceptance, that some kind of scenario based approach to policy making is a necessary part of the decision-making process before the door to more significant change can be opened.
So, to come back to the question posed by the plenary, does rapid change mean new models? It is not the rapidity but the breadth and scale of the changes to how society operates and the kind of mobility systems that might exist to support this that matters. Acceptance of a wider palette of futures and a need for a different approach to decision-making seems critical to the extent to which the demands on the modelling community are incremental or more fundamental. I am delighted to see this debate alive and kicking as I believe that if we do not get to grips with the real changes that are happening and what they mean for transparent decision-support then, at some point, our professional credibility will be challenged and potentially undermined.
This article is reproduced here with the permission of Landor as part of Modelling World 2017