The Suprising Success of Railway Reopenings
and the crystal ball approach to infrastructure planning.
Over the past few months, a rail revival has been happening in a corner of North-East England. The Northumberland Line, closed in 1964 as part of the Beeching Cuts, has been partially reopened, connecting the coastal communities of northern Northumberland with central Newcastle.
After decades of campaigning by local authorities and communities, and after a few delayed opening dates, services finally recommenced along the line in December of last year. To describe it as a success would be an understatement: the project’s outline business case predicted 704,000 passengers in 2025, its first full year of operation1; in its first three months of operation, it has already served 250,000 passengers. Carry that forward over the year and it’s already ahead of forecast.
This success is even more astounding when you consider that the passenger forecast presumed that part of the line would have opened two years earlier in 2023 (and thus would already have built up a patronage), and second, that only half of the new stations on the line are currently open; final engineering works are still underway at Bedlington, Blyth Bebside, and Northumberland Park, which are due to open later in 2025.
On current growth rates, and with the eventual completion of the line, one councillor's wild estimate of passenger numbers five times greater than the forecast might actually be achieved in the next year 2, beating even the 2039 forecast of just under 1.5 million passengers a year.
Though potentially surprising to the project’s sponsors and economists, the dramatic under-forecasting of passengers on the railway isn’t unique to Northumberland Line.
A few dozen miles north is the Borders Railway. Also a victim of the Beeching Cuts, 57 km of railway between Edinburgh and Tweedbank was rebuilt between 2012 and 2015 by the Scottish Government. In its first year of operation, around 650,000 passenger journeys were forecast3. 125,000 people used the service in the first month, 20% of the annual forecast. Services were so overcrowded that passengers were unable to board trains at intermediate stations on the line, prompting Scotrail to lengthen services with additional carriages. The popularity of the service only continued through its first few years of operation, exceeding 2 million passengers a year in 2019 before the pandemic hit4.
In fact, almost every new railway opened in the UK over the last decade has beat its forecast. To give a few more examples:
Passenger numbers on the reopened branch railway to Okehampton were double the forecast, leading the government to announce a doubling of train services5.
London’s Northern Line Extension clocked up 5 million passengers in its first year6. In 2023, it had 12 million, exceeding projections of 8 million per year by 50%7.
The launch of new services between Oxford and Marylebone via a new line with stations at Oxford Parkway and Bicester Village performed ahead of expectations, and increased the size of the London-Oxford market by around 20% in just six months8.
The Elizabeth Line, built as part of the Crossrail project has already far exceeded revised forecasts produced during the pandemic, with passenger growth far in excess of the rest of London’s transport network9. Actual passenger numbers are likely to overtake the pre-pandemic forecast this year.
It’s not all rosy
Projects delivering far better outcomes than they were predicted to is obviously a good thing for those projects, but they also highlight systemic problems in how the UK plans and funds infrastructure.
Firstly, inaccurate forecasts mean that we may fund projects that we shouldn’t, and don’t fund projects that we should. Doubling the passenger forecasts within a business case doesn't just double the revenue received from those passengers. There are also secondary benefits from doubling the reduction in road congestion, more economic impacts from increased leisure travel and the take-up of better employment, environmental benefits from shifting to lower-carbon travel modes, and health benefits from active travel in the last-mile of a rail journey.
The UK has historically underinvested in urban public transport infrastructure compared to our peer countries, a problem that is particularly acute outside of London. Many transport projects that have been unable to justify Treasury investment might have gone ahead if the benefits of them had been substantially higher relative to the cost. When we ask why Leeds is the largest city in Europe without a tram or metro, perhaps the answer is simply that we fail to accurately model its potential benefits10. The Borders Railway, the Northumberland Line, and the Okehampton Line all faced decades-long campaigns to justify and acquire their funding; we might have delivered them and reaped their benefits much sooner if we knew just how good they would have been.
The second problem with relying so heavily on poor forecasts is that we misscope the infrastructure that we build. The Borders Railway was reopened as a single-track line with three passing loops to allow services to pass in opposite directions. As such, the frequency of services on the line is limited without additional rail infrastructure that could allow trains to pass each other along the entire route. That’s fine if you correctly planned for the number of people using the line, less so if you miscalculated by a factor of three. Whilst much of the line has passive provision for a second track, several key infrastructure elements such as bridges were built without the space for a second track and thus limit operationally flexibility. Any upgrades to new railways will also require line closures and cost more than if they had been implemented before the initial opening.
If the future growth of the line had been anticipated, then future improvements to service frequency could have been delivered at a lower cost. History may be repeating itself on the new East-West Mainline being built between Oxford and Cambridge, including the potentially expensive mistake of not building it with electrification or 8-carriage platforms from the start. In the last year alone, fundamental assumptions of the scheme have changed, with the future largest leisure attraction in the UK now planned to be served by a station in the middle of the route11, and the ambitions for the future scale of both Cambridge and Oxford have dramatically grown12.
Another way forward
Overcoming the problem of poor predictions is essential if we are to change our approach to infrastructure delivery, but forecasting is a notoriously hard practice. When it comes to passenger forecasts, commuters often make up a plurality of future ridership, and commutes are relatively easy to model. Journey times are predictable, and economists have data readily available to assess where homes and jobs are with ease. But even commutes can be somewhat unpredictable. The Borders Railway hasn’t just had patronage from passengers on the line taking up jobs in Edinburgh or deciding to switch to the car, it’s also benefited from people actively relocating to stations on the line to benefit from the new connections. These second order moves are harder to predict.
Leisure journeys are even harder. Part of the early success of the Northumberland Line is thanks to matchday crowds travelling into the city to watch Newcastle play, but how do you accurately predict passenger numbers when the model needs to be detailed enough to factor how likely a passenger is to be too inebriated to drive. Similarly, passenger numbers to Bicester Village station, connected to London Marylebone in 2015 and home to one of the busiest brand outlets in the UK, are affected by the strength of the Pound and UK tax policy for tourists. In the face of so many variables and assumptions, and resultant room for error, one might have the same amount of luck in making passenger predictions through the medium of witchcraft.
Any public finance hawks reading this will also be thinking that given the margin of error on forecasts, it’s preferable to err on the side of caution. Schemes that do proceed to construction are often finely balanced between costs and benefits, and whereas an under-forecast of passengers is an overwhelming success, an over-forecast can easily turn an investment into a white elephant. This is even more pertinent considering cost overruns only tip the business case one way.
Given the difficulties in improving the models then, we shouldn’t just update how we appraise projects, we should also reconsider how we finance and decide on them.
Wider adoption of value capture and local financing initiatives would create additional financial justification for transport projects by reducing the level of central government funding (or local government borrowing) required for them to go ahead, thus reducing the need for forecasts to be so accurate. Additionally, councils and combined authorities should be given greater discretion in what projects should go ahead, allowing them to fund and proceed with schemes that don’t look as beneficial on paper, if they wish. The one size fits all model led by the Treasury and the Department for Transport clearly doesn’t work, and projects in London will always outcompete regional projects for funding when comparisons are made at the national level against a single model. Devolution will give regional bodies the freedom to invest in projects that don’t quite hit national thresholds for investment.
As the UK Government talks about the importance of growth whilst overseeing a reduction in the transport investment pipeline, it should learn the lessons of transport investments past and present, and put in place policies that allows the country to deliver some more.
Though I note that the centralisation of transport investment decision making in Westminster is also a problem; low benefit schemes in the south have proceeded where high benefit schemes in the north have not
*** health benefits from active travel in the last-mile of a rail journey....how do you accurately predict passenger numbers when the model needs to be detailed enough to factor how likely a passenger is to be too inebriated to drive***
I think a very good reason to get better at basic macro-level estimates here would be discouraging the habit of everything-bagelling any conceivable effect in to help the numbers and the related super shonky nth-order modelling. We don't really know how much Schools And Hospitals will benefit from people actively travelling in the last mile cough having to walk the rest of the way, we shouldn't pretend we do.
There's something deeply ironic in the way all these approval processes were invented because Do You Want A Planned Economy? Eh? Eh? Are You Some Kind Of Commie Let Me Tell You About Hayek And Friedman and they've evolved into efforts to do socialist calculation way beyond anything Oskar Lange or Wassily Leontief dreamed of.
Thanks for writing this Harry! I shall try to circulate appropriately