According to traditional ‘four step’ transport modelling (area-based modelling, which identifies ‘trip attractors’ such as workplaces, quantifies and distributes trips to those ‘attractors’, assigning them to modes, and working out routing), choice of mode (and route) is based around the ‘generalised cost of travel’. In other words, when deciding how to get to work, we trade off money and time. The ideal commute would not exist; travel time is seen as a pure loss, as equivalent to taking a certain amount of money away. (Of course, cycling again challenges this, with its physical activity benefits: but then, we also know many people value regularly travelling at least some distance, even just a short walk round the block ‘to get out of the house’.)
If the models just traded off time against money when predicting mode choice, cycling would likely come out quite popular, at least in dense and congested urban areas. In 1995 DfT tested out the Dutch modelling package QUOVADIS-BICYCLE, to model cycling in Ipswich; models used in the UK at that time would have been unlikely to include cycling at all. For Ipswich, “the model initially predicated that 30% of trips […] would be undertaken by bicycle.” This compared with an actual mode share for cycling of 6-7%. The model had to be adjusted to predict the actual levels of cycling in Ipswich, whereas in the Netherlands a mode share of 30% would be normal.
If saving time (and money) determines mode choices, the decision not to cycle appears irrational, as with the promotional material that tells us, for example, that the benefits of cycling are 20:1. However, it’s here that the ‘mode specific constant’ arrives and makes not cycling rational again – on the grounds that people just don’t want to cycle.
For time is not all the same. Different categories of time have different values; for example, waiting at a bus stop and walking to the bus stop are both ‘expensive’, sitting in the bus somewhat cheaper. (In a health conscious population, will the ‘value’ of walking time change?) This matters for cycling as a coefficient is given to different modes to reflect the actual mode choice within a population; ‘cycling time’ is thus categorised differently to ‘driving time’.
Wardman et al 2007’s study concluded: ‘Time spent cycling is valued almost three times more highly than travel time for the other modes’. In other words, twenty minutes cycling is as undesirable (travel time being a loss!) as 60 minutes in a car or bus. Hence while cycling may be substantially quicker in time terms, it can then still cost more in value-of-time terms. The undesirability of cycling (in current conditions) becomes built into the model, rather than something the model interrogates and challenges. Factors suppressing cycling then can’t be manipulated within the model and limiting the ability to predict and model for change. This affects other modes too; the growing use of smartphones and laptops on the move has transformed the ‘value’ of time spent on public transport.
As an aside, it’s odd that the high ‘value of time’ spent cycling rarely seems to translate into a policy interest in speeding up cycle journeys, although travel time savings are often where much of the benefits come from when schemes are appraised. But then traditionally, highway scheme appraisal has focused on benefits to drivers using a proposed road. Although now all modes are supposed to be considered, where cycling is concerned, data quality and modelling capacity may preclude its serious consideration. For example, if models don’t accurately capture actual and potential cycling trips, and there are no cycle counts taken on roads that cross a proposed highway site, impacts of a new road on cyclists (including the possibility that cyclists will switch to other modes as their journey becomes too unpleasant and/or too lengthy) will not be well understood. Yet potentially, modelling can help us understand why cycling has gone up (often very dramatically) in some contexts and among some people, while remaining flat on a national level – and how existing rises in cycling could be continued and generalised.
Wardman et al (2007) found evidence that cycle tracks and cycle lanes would make cycling more attractive, both through observing actual routes and by asking about preferences. Values derived from some of their work do appear in WEBTAG (DfT’s Transport Analysis Guidance), with caveats: ‘Note that the impact of a wide variety of different changes can be calculated but that these results should only be regarded as very approximate in general application. This is pending further research in the field and the potential derivation of coefficients for other purposes or the development of a more sophisticated model.’ DfT also stresses that the infrastructure factors are only suitable for looking at short trips and single journey purposes.
So, there is an understanding that route quality matters for cycling, but its inclusion in modelling and appraisal is still seen as experimental. The WEBTAG guidance quoted above, which says further research is needed, dates from 2010, and as I understand it, the infrastructure values it given there are currently little used. In terms of understanding mode choice and routes, I have been told that even relatively good models tend to lump cycling and walking together, and don’t construct a separate ‘cycle network’ layer for cycling route choice. Other models don’t specifically model cycling and walking, but allocate ‘left over trips’ to those modes. In other words, even when modellers are trying to include them, cycling and walking are usually seen as what doesn’t quite fit in and cannot be easily modelled in the same way that car use and, later, public transport have been modelled. Do we need new models, perhaps?
Forward to Part 3: Traffic Flow and Junctions