Problems with transport modelling

The Department for Transport has issued a draft National Networks National Policy Statement (NNNPS) for consultation . It covers major investments on the road and rail networks in England. The draft is intended to replace the version issued in 2015, before the government’s commitment to Net Zero and publication of the Transport Decarbonisation Plan. The House of Commons Transport Committee has announced an inquiry into this draft.

The DfT states that the 2015 NNNPS shall apply to projects already selected for public examination, so the new NNNPS will apply only to applications accepted after it is implemented, following the consultation. It therefore looks as though the Lower Thames Crossing tunnel, which has been accepted by the Planning Inspectorate for consideration, will be subject to the old guidance, despite construction being deferred by two years as announced in the recent Budget, which seems odd.

The purpose of such National Policy Statements is to provide guidance for decision-makers on the application of government policy when determining development consent for major infrastructure. The intention is to remove the need for consideration of fundamental national policy questions at planning inquiries. Those subject to this guidance are the scheme promoters (National Highways for most road proposals), planning inspectors, and the Secretary of State when granting Development Consent Orders.

The important question is how investment in new road capacity could be reconciled with the government’s legal commitments to achieve Net Zero greenhouse gas emissions by 2050, meeting the requirements of both the Climate Change Committee’s Sixth Carbon Budget that has been agreed by the government and the intentions of the DfT’s Transport Decarbonisation Plan. (Rail, already substantially electrified, is less of a problem.)

The draft opens by rolling the pitch, stating that the government sees a compelling need for the development of national networks (para 3.22), such that there is a presumption in favour of granting Development Consent Orders (para 4.2), while at the same time recognising the need to move away from ‘predict and provide’ (para 3.44). This is very different from the new approach of the Welsh government, which does not see a compelling need to develop its national road network.

Scheme proposals are to be supported by assessments of whole life carbon emissions, to ensure minimisation as far as possible (para 5.29). The draft states that, in reaching a decision, the ‘Secretary of State should be content that the applicant has taken all reasonable steps to reduce the total greenhouse gas emissions from a whole life carbon perspective. However, given the important role national network infrastructure plays in supporting the process of economy wide decarbonisation, the Secretary of State accepts that there are likely to be some residual emissions from construction of national network infrastructure’ (para 5.36). Moreover, a net increase in operational greenhouse gas emissions [from more traffic] is not, of itself, reason to prohibit the consenting of national network projects or to impose more restrictions on them in the planning policy framework (para 5.37). So in policy terms, additional road capacity is more important than decarbonisation.

Importantly, the application for development consent orders applies to individual schemes. There appears to be no requirement to estimate the impact on carbon emissions from an investment programme, such as the planned five-year Road Investment Strategy 3 (RIS3) due to start in 2026. Accountability scheme by scheme is not so very different from the present practice whereby National Highways argues that each individual scheme makes only a de minimus contribution to national carbon emissions, which can therefore be disregarded.

The DfT’s Transport Decarbonisation Plan made broad-brush estimates of carbon reduction from policies and programmes, for instance 1-6 MtCO2e from increased active travel over the period 2020 to 2050, and 620-850 MtCO2e for electrification of cars and vans over the same period. It is inconsistent not to recognise offsetting carbon increases from investment in new road capacity, likely to fall somewhere between the above ranges, and certainly not de minimis for the programme as a whole.

There is also a problem of modelling future carbon emissions arising from road investment. Transport models are complex and opaque, with many parameters, the value of which requires expert judgement. In consequence, the are two types of protagonist: experts who have a good working understanding of transport models because they earn their living from building and running such models; and non-experts, who are interested in the output of models but are not able to understand the assumptions, simplifications and judgements that the experts must make. Non-experts include decision makers in national and local government who have prior expectations of the economic value of particular road schemes, and whose test of a good model is that it delivers outputs, comparing with- and without-investment cases, consistent with these expectations. Other non-experts are those opposing road schemes at public inquiries, who are faced with modelled outputs as part of the promoter’s proposal that are not open to detailed scrutiny. Inspectors at planning inquiries are also non-expert in this sense.

The NNNPS requires projects to be supported by a local transport model, but planning inspectors and the Secretary of State do not need to be concerned with the national methodology and national assumptions around the key drivers of transport demand (para 4.7). In practice, most schemes on the Strategic Road Network employ local versions of a set generic traffic and economic models, typically SATURN for network traffic modelling, the outputs of which are inputs to the TUBA economic model. So, as it appears, consideration of the predictive validity of these models for projecting carbon emissions need not be considered either at a public inquiry or by the Secretary of State. One can understand why a planning inspector should not be burdened with a task for which they are not professionally trained. Nevertheless, the question is where in the decision-making process the validity of the supporting modelling might be assessed.

The need to assess the predictive validity of transport models is pointed up by the failure of standard models to project fairly short-run traffic flows in two cases of motorway widening, on the M25 and the M1, as I have recorded previously. This does not increase confidence in the ability of such models to project economic benefits and carbon emissions out to sixty years.

One particular problem of transport models is that they are largely used to justify new investment, in which context the saving of travel time is supposed to be the main economic benefit. Yet average travel time, as estimated by the National Travel Survey, has changed very little over fifty years, excepting the period of the coronavirus pandemic. The implication is that people take the benefit of faster travel as enhanced access to desired destinations, people, places, activities and services, for the opportunities and choices on offer. Travelling further, rather than using travel time savings for more productive work or agreeable leisure, means more externalities related to vehicle-miles travelled, carbon emissions in particular.

Modellers who aimed to model such access benefits, and the resulting changes in land use and value, would not be appreciated by the economists who are wedded to travel time saving as the main economic benefit of investment, nor by decision-makers who are well used to conventional economic investment appraisal. So modellers must fix their assumptions, simplifications and parameters to get outcomes that satisfy a ‘realism test’ of prior expectations, subject to conformity with unspecified standards of professional respectability.

The upshot is that the modelling of the impact of new road investments will systematically underestimate carbon emissions from the additional (induced) traffic. This makes it easier to appear to comply with the pathway to Net Zero, but means that the outcome is likely to fall short of that pathway.

Some further light is shed on this matter by the cost-benefit analysis the DfT has published in support of options to implement the Zero Emissions Vehicle Mandate, the legislative framework to fulfil the government’s objective to phase out the sale of internal combustion engine cars and vans by 2030. The need for this cost-benefit analysis is not stated, since the timing of the phase out is largely for negotiation between the government and the motor manufacturers. Perhaps the Treasury wish to be assured that this route to decarbonisation represents good value compared with other possible decarbonisation measures. Or perhaps the DfT economists wish to parade their competences after cost-benefit analysis failed to be supportive of a number of major rail and road investments.

The modelling assumes that that the switch to ZEVs could result in increased mileage per ZEV driver because electricity as fuel is cheaper than petrol or diesel (which begs the question of whether some new charge for EVs might be introduced, as I have suggested). This extra driving, a ‘rebound effect’, is supposed to lead to more congestion delays, with a very substantial cost impact: for a central sensitivity case of the preferred policy option, the abatement cost of the ZEV Mandate for cars and vans estimated as £12/tCO2e excluding the rebound effect, and £100/tCO2e including it (Tables 61 and 62).

So, the DfT thinks it would be much more costly to reduce CO2 emissions by means of the Mandate if the lower operating cost of EVs led to greater distances travelled. However, in my view, rebound of the magnitude modelled is unlikely, quite apart from the possibility of a road user charge for EVs. The per capita distance travelled by car depends on three main factors: speed of travel, time available for travel, and household car ownership. None of these are affected by the switch to electric propulsion. Vehicle operating costs have a second order impact at best, witness the growth of SUV ownership despite higher fuel use.

Paradoxically, the DfT modellers postulate additional traffic from reduced vehicle operating costs arising from electrification (mistakenly, in my view), while being in denial about the additional traffic arising from road users taking the benefit of investment in increased capacity as enhanced access involving more travel (again mistakenly).

The ZEV Mandate cost-benefit analysis states that the preferred policy option is expected to achieve emission savings of 415 MtCO2e in the period 2020-2050 (Table 29). This is substantially less than the savings from switch to electric propulsion of car and vans of 620-850 MtCO2e projected in the Transport Decarbonisation Plan, mentioned above. No clear explanation for this discrepancy is given; it may be because the present Mandate is for the period to 2030, with a further Mandate promised for 2031-35; or it may reflect the sensitivity of model outputs to input assumptions.

More generally, modelling for the ZEV Mandate exemplifies how modelling outputs can be very sensitive to input assumptions that are made in the absence of firm evidence of future travel behaviour. This is a caution that applies to most transport modelling, not least to the projections of transport sector carbon emissions to support decisions necessary to achieve reductions required by the legislative framework to achieve Net Zero.

We have been before in a situation in which there have been doubts about approaches to transport economic analysis endorsed by the DfT. Good work was done by SACTRA – the Standing Advisory Committee on Trunk Road Assessment – an independent body created by the DfT, that issued two influential reports in the 1990s. One confirmed the importance of induced traffic arising from new road construction, a view that had been resisted by the DfT since such traffic added to congested and reduced travel time savings. The other report recognised the wider economic impacts of investment, beyond the conventional time saving, vehicles operating costs and those externalities to which monetary values could be attached; estimation of such wider impacts, such as agglomeration effects, now forms part of the standard approach to investment appraisal.

Although SACTRA, by its very name, was intended to remain in existence, at least until formally stood down, it seems to have fallen into that state by not receiving new commissions. There is a need, in my view, to reconstitute it, or some similar body of independent experts, to look at the suitability of the current body of official guidance on transport economic analysis and modelling in an era when decarbonisation is a national policy priority. As it is, however, the people in DfT and their consultants, who naturally wish to please their clients, are talking to each other in an echo chamber, from which interested outsiders are excluded.

Other departments do better. The Treasury’s model of the UK economy has long been available to independent forecasters. The Energy Department collaborates with academic energy modellers and makes available the online Mackay Carbon Calculator that allows users to explore the options for reducing carbon emissions. Modelling of the coronavirus pandemic was largely carried out collaboratively by academic groups whose models and outputs were public for all to debate. And the modelling of climate change is carried out openly, collaboratively and internationally as input to the reports of the Intergovernmental Panel on Climate Change.

Th DfT instigated a move to update the National Transport Model to generate a new version, NTMV5, intended to be open to other users, but this seems not to have worked out in that the National Road Traffic Projections 2022 employed the previous version (as I have noted). The DfT should explain what went wrong, and should engage openly with those beyond the Department and its immediate advisers on how best to model the decarbonisation of the transport sector.

This blog post is the basis for an article in Local Transport Today 23 May 2023.

My written evidence to the House of Commons Transport Committee inquiry into the National Networks National Policy Statement is based on this blog.