The big topic from this morning was demand forecasting. To my knowledge, the main principles for demand forecasting have been around for some time. However, there are different new possibilities around. I'm expecting them to have an impact on the process. What is your opinion? What is the current dynamic in the field and what will happen in near future? 

Graham: I think one of the great things new technologies are giving us is better data, more accurate data, particularly on travel patterns and travel flows in time and space. However, the principles of demand forecasting are the same. We just adopt that data because it's better than the travel survey data we've been collecting using expensive methods. But that's not to say that travel forecasting isn't changing. 

One of the biggest changes is activity travel modelling, which is a much more detailed and personalised approach to certain groups of people, making certain types of trips. The model links their activities together with the travel route that they take. So, it's a contrasting level of depth beyond the four-step transport models, which are the fundamental basis of forecasting in cities. Modern data can assist with this because it's more personalised and detailed. But, in addition, there is a picture which has not been developed yet, which is about individualised forecasting for individual people. One of the great things that personal information can do is it can be relevant to the person involved, and therefore travel planning and trips for that individual can be personalised. There are, of course, personal security issues about that and freedom of information issues about individual tracking and so forth.  

What about behavioural economics? There has been recent work in behavioural economics, awarded with Nobel Prizes, that has changed the world. Do you see any impact from there through how people choose things, how people behave, how people travel? 

Graham: The only Nobel Prize in transport was awarded to McFadden (Daniel McFadden) who is an economist who did work on logic curves and the discrete choice model. That's the fundamental basis of logic curves. It was some 20–30 years ago. His work is the basis of constraint modelling. Behavioural economics beyond then – all that it is really doing is adding more accurate parameters to understand behaviours. The field I talked about this morning, for example, with soft factors, is recognising we can value quite intricate items now.  

Also, the world of travel has changed. Real-time information is an example of that, making travellers make different choices to the ones they made before.  How we value real-time information is very important. Such information is not only relevant to travel at a particular time, but access to information removes a degree of uncertainty. There are wider benefits than just having the information – it affects our perceptions of unreliability. And yes, this information is valued. But you see, people’s decision-making data is being fed into a forecasting process, but the process itself has remained virtually the same over the years. We're just affecting the generalised cost parameters. Whereas mobile use data, for example, is really affecting the quality of the origin-destination matrices which are going into these models. So, I think that forecasting hasn't changed a lot other than employing new methods, like activity travel modelling and potentially individualised forecasting, which is still a long way off. 

Now that you mentioned mobile data, the question is where can the process mostly benefit from this type of data? What do you think? 

Graham: As I said, I think the biggest implication is better quality data, cheaper and at volume. Currently, the number one methodology of getting passenger information into models is household travel surveys. And typically, cities in the Netherlands have a whole process of sampling. Sampling about 1% of all households every year. And from that, the abstract travel behaviours for a weekday and a weekend day. Often, it's less than that – one weekday. And then they extrapolate that result to represent the other 99 percent of households. The quality of that data is terrible. It's acceptable for major transport planning decisions at an aggregate level, but it's quite weak at the individual travel level. Now why do we do only one percent? 

The reason we do that is because we can't afford more. It's tremendously expensive to get all that information from households. We are finding ways of reducing those costs, by the way, using sample data, using mobile phones to improve household travel surveys. But the idea of using mobile data, as long as we can get accurate mobile data and mobile data that's correct – it can be much cheaper. There are privacy concerns with mobile data, and there is the problem that mobile data can require an opt-in. There are apps that now allow you to collect mobile data, and privacy concerns can be depersonalised. So that's the hope, I think, that you can get better quality and better volume of data about the user's input going into travel planning processes. The planning processes are just going to have the same foibles that they have now, they just got better quality data in them.