Every weekend we dig into the archives. This post by Stu was first published in November 2015.

In Matt’s recent post about MoT’s work on the future of transport, there was an interesting little side-discussion about transport models, and in particular the travel demand forecasts which emerge therefrom.

We’ve previously written about the accuracy of transport models when used for project evaluation purposes. Perhaps the most notable (notorious?) was this post on the Waitemata Harbour Crossing, where we discussed how NZTA’s business case used traffic volumes that were approximately 10% higher than actual volumes. Naturally this led to the benefits of the project being overstated.

More recently, this post on the SH20 Manukau Harbour Crossing found it was carrying approximately 45% less traffic than projected 10 years after it was complete.

Manukau-Harbour-Crossing-Traffic-Volume-actual-vs-prediction

Transport models are important not just for the purposes of project appraisal. The outputs of transport models are also used to forecast aggregate travel demands and determine policy at a much higher level. For example, the MoT and the NZTA use (different) transport models to forecast aggregate vehicle kilometres travelled, which in turn determine the funding that is available in the NLTF to fund the projects identified in the NLTP. Hence, our ability to plan ahead is influenced by transport models.

In this context, graphs such as the one below are something of a cause for concern.

But let’s not be to critical of the MoT and NZTA; they are not alone insofar as their transport models have consistently over-estimated demand. Graphs surprisingly similar to that above exist in the US and the UK; one such example is shown below.

In some earlier posts here and here, we’ve presented some possible reasons for what might be causing the transport models that are used for forecasting travel demands (both at the level of individual projects and in aggregate) to get it wrong. The systematic positive bias in forecasting errors has been the subject of formal academic research led by the Danish economist Bent Flyvberg. I presented a paper at last year’s IPENZ Transportation Conference in which I discussed some of these issues in more detail, while Peter wrote an excellent post on the topic here. His analysis of NZTA data suggests that New Zealand may not be immune to the same systematic biases found overseas.

This post, however, is not about the systematic positive bias into transport models. Instead, this post is about whether the mechanics of the models are capturing what they need to capture in order to formulate accurate predictions. I think this is a useful starting point for thinking about “transport futures”, as the MoT seem keen to do.

I also think it’s fair to say that the superficial explanation for the slowdown in the growth of aggregate vehicle travel is that per capita vehicle travel has been on the decline. That is, people (both in New Zealand and overseas) aren’t driving as they used to. But this doesn’t get us very insofar as predicting the future, i.e. observing that per capita demands are declining simply begs the question of why?

And this is exactly where things start to get interesting. In my time thinking about these issues the views that are expressed tend to be readily grouped into one of three broad categories, which I now present for you to consider.

First we have what I call the “establishment view“, which is led by the likes of the MoT, the Government more generally, and a number of consultants. This view argues that the decline in per capita vehicle travel primarily reflects higher fuel prices and reduced economic activity over the last 5-10 years. There’s obviously some logic to this view; it seems reasonable to suggest that both the cost of fuel and the state of the economy will impact on travel demands, at least in the short term (say 1-5 years). Where the establishment view struggles, however, is to explain why the slowdown in vehicle travel started so early (way back in 2005), and why volumes haven’t bounced back more strongly of late. The latter is particularly interesting given NZ’s robust levels of economic growth, strong population growth, and sustained low fuel prices.

While VKT is currently growing again, it doesn’t seem to be growing by as much as one might expect based on these factors.

Which leads me to the second view, which I call the “wider socio-economic view“. This is probably the position which best describes my own views, at least in terms of understanding travel demands in the medium term (say 5-10 years). This view looks beyond the hard economic factors considered by the establishment view and instead consider some wider factors that seem likely to impact on the demand for vehicle travel. People who subscribe to this view will often talk about the following issues:

  • Demographic factors, such as an ageing population and changes in the number of people with drivers licences;
  • Transport and land use factors, such as availability of public transport and the ongoing intensification of our cities; and
  • Vehicle substitutes, such as air travel and telecommunications.

The wider socio-economic view complements the “establishment view” in some senses, because it appeals to similar micro-economic mechanisms, but it does so in a way that allows for a wider range of factors to impact on the demand for vehicle travel. In doing so, however, the socio-economic view can lead to predictions that are quite different to those of the establishment view. Instead, the socio-economic view allows quite a lot of room for future growth in aggregate vehicle travel to differ from what we’ve seen in the past – which is something the establishment view struggles to incorporate.

Finally, we have what I call the “changing preferences” view. This perspective interprets the declining per capita demand for vehicle travel as a function of wider shifts in people’s underlying preferences. This view emphasises, for example, that young people are now placing a higher value on other forms of consumption, such as the connectivity offered by smart phones, than the personal mobility associated with owning and operating a vehicle – at least compared to their parents. The changing preferences view would seem to suggest the trends we’ve seen in the last 5-10 years are just the tip of the ice-berg, and that profound changes are just around the corner. Often people highlight that it’s not just technology which is driving these changes, but also awareness of the health and environmental effects of driving. Evidence for this view were recently summarised in this NZTA research report, which we previously discussed in this blog post.

The following figure is taken from this report.

Which view to believe? Well, personally I think all three have elements of truth to them. I think the establishment view is correct insofar as certain price and economic factors are likely to dominate changes in travel demands in the short term.

In the medium to long run, however, the other two views presented above seem to have more currency. I mean, if fewer people have drivers licenses then it seems plausible to suggest that there will be a reduction in per capital vehicle travel, ceteris paribus. Similarly, we can expect reductions in the cost of air travel to eat into the demand for long distance vehicle trips. In terms of preferences, these are critical to the accuracy of any model that seeks to forecast future demands based on past behaviours. If preferences changes, then our predictions are resting on a wobbly plank over shark-infested analytical waters.

So what’s the takeaway message? Well, I think it’s fair to say that we simply don’t know what to expect with regards to the future demand for vehicle travel.

In this context, I personally wouldn’t be investing in large transport projects, especially in rural areas, e.g. Puhoi-Wellsford, Waikato Expressway etc. As for the CRL, I think it’s likely to be a good project because of 1) patronage growth on the rail network, 2) Auckland’s growth as a whole, 3) the explosive growth in the city centre (both in terms of population and retail), and 4) wider trends in transport and land use policy, e.g. time-of-use road pricing and removing minimum parking requirements.

But I’m open to being convinced either way, and am interested to hear what others think on all this.

Share this

8 comments

  1. Three years later this post is still a relevant thread. I worked on such models earlier in my career, but no longer believe them either. I would add two other issues that I think hugely challenge the accuracy of the models – induced demand and travel time budgets. Both suggest that any bounce back in car usage after the completion of projects like Waterview is a circular logic at work. Stop building projects like Waterview and per capita car usage will continue declining. Once the road system reaches saturation (it has in Auckland) expanding it is futile.

    So what do we do? I say throw the models away and build the city we want and can afford, that maximises quality of life and housing affordability. Both of those objectives suggest building more PT and more walkable suburbs, linked to better housing options.

  2. It is a good article, but it is sightly wrong on the modelling of funding needs.

    The demand forecasts give context to project identification and appraisal, but are distinct from the revenue forecasts that are used to advise on tax rates.

    The revenue forecasts rely more strongly on assumptions about the relationship between transport growth and economic growth. The $500m ‘hole’ in transport revenue in 2013-15 was largely because the economic forecast data from Treasury thought the economic upswing would happen earlier than it actually did, with an unclear but apparently meaningful contribution from the lag in household income growth coinciding with a spike in the real price of fuel.

    Historic fuel and RUC purchases are the other important factors looked at in revenue forecasts, which are monitored (monitorable, anyway) on a monthly basis. The model itself tends to over-weigh historic purchasing patterns, which tends to pull the forecast ‘up’. This provides an opportunity to take advantage of the multiple models by creating a couple of moments where, first, MOT brings the NZTA, Treasury and occasional others into the room to agree downside scenario parameters and then, second, once the raw and downside forecasts are done, a choice is made around where in that range to pitch the ‘official’ result.

    That official result, on its best day, is only good out to a 3-5 year horizon. In fact it is redone every six months because the real world eats predictive models for lunch.

    In consequence, demand forecasts don’t really relate well to the revenue forecasts, and you can get quite different short-term demand stories from them: you cannot accurately recreate expected revenues by reverse engineering from the demand model estimates, for example.

    Of course, the purpose of the revenue forecast is to help people ask an answer the question “what do we need to do to get the revenue we thought we would need”? Quite different to asking “what will the transport task look like in 10-30 years?” In the latter case, I’d suggest monitoring and modelling is less relevant than vision and choice.

  3. A question for the GA authors, relating to the graphs and data presented here. Back in 2012 when the cover graph was put together, there seemed to be a clear link between traffic numbers being hopelessly different from the actual numbers being recorded (i.e. the black diamonds). I’d be interested to know if that is still the case now – given that we have been in an economic boom for the last several years, but in 2012 the economy was just coming out of the post GFC debacle.

    Were the vehicle figures (being so out of kilter with the predictions) like that just because less people had jobs i.e. less driving around, or has it spiked more lately with more people, more money etc, or is it an actual structural change in the way that people travel? There certainly does not seem to be any less importation of cars into the country…

    1. Average Human I think you are exactly right here – 2012 was the end of flat traffic numbers. If the diamonds for actual traffic were put on the first graph here I think they would follow the modelled pink lines pretty much spot on. So its really more that the traffic models got it wrong proceeding to the GFC and for the following 4 years. Perhaps the salient point is the proportion of influence economic factors has seem to be much stronger than the changes in people not driving for whatever reason. That said, its hard to see how the change to more inner city living, millennials not so interested in cruising the streets in cars etc can’t begin to have a significant effect.

    2. Could it be that vehicle numbers are increasing while VKT is not? This would suggest more stationary vehicles are taking up more room on streets, filling up more off-street parking facilities and occupying more space on private sections – all of which seems to be happening. Would be interested to see if stats support this.

      1. vkt numbers are rising for both local roads and state highways in Auckland, and we are at an all-time high for both. The figures are definitely independent of car import numbers. I don’t have car registration numbers at hand to compare. Some years the vkt has fallen slightly.

  4. What benefits did Peter use to create that Transport Benefit Overruns graph? Were they just the benefits included by NZTA in the PIR’s?

Leave a Reply