Regional Rail Demand in Northern England – Explaining Growth in a Downturn World

Regional Rail Demand in Northern England – Explaining Growth in a Downturn World


Jon Crockett, Systra, Pedro Abrantes, Passenger Transport Executive Group, Mark Wardman, ITS


This study commenced from a need to better understand the causes of the observed rail demand growth in northern England.


The rail network in Northern England has continued to see substantial passenger growth, with demand doubling in the last ten years. This pace of change is relatively consistent across all segments, from commuters and peak period business travellers through to young persons and senior citizens. However, the backdrop to this has been the prolonged economic downturn since 2007/8, meaning that many of the explanatory factors normally associated with rail demand growth have followed the opposite trajectory. Employment, whilst not contracting as much as some expected during the downturn, has also not risen appreciably, whilst declines in disposable household incomes predate the downturn. Whilst there are some favourable trends for rail demand from competing modes, the elasticities to explain growth in rail demand would have to be of a magnitude well beyond anything previously identified. It was clear that there have been additional and/or more complex relationships and interactions at play than the UK’s existing demand forecasting guidance, the Passenger Demand Forecasting Handbook (PDFH), currently recognises.
This study commenced from a need to better understand the causes of the observed demand growth. With substantial investment planned in the network, at least in part designed to help stimulate economic prosperity, and the potential for devolution of franchising powers to local authorities, there are significant risks from the existing uncertainties. Future investment in the network is dependent on understanding where and when any growth will occur, whilst also helping to unlock further growth itself. With the potential devolution of powers comes revenue risk for the local authorities, which is magnified if borrowing is undertaken against future revenues for additional investment in new or additional rolling stock, station upgrades etc which are themselves predicated on expected growth from external influences such as the economy, lifestyle preferences, levels of service from competing modes etc.
A large multi-source dataset was collated for the whole of the North of England, covering 540 individual stations (potentially ≈300,000 station-to-station pairs, although many of these have zero orvery little demand) and ten years. Critically this sought the most spatially detailed data available and to interpolate/extrapolate datasets where they were only available for selected years. This represents an important development on many of the studies which have underpinned the current PDFH guidance, as they have tended to rely on more spatially aggregate datasets available across all years in the time-series, ie sacrificing spatial detail to ensure that consistent and reliable temporal data was deployed in model estimation. This work took more of a hybrid approach, combining cross-sectional and time-series data, with small area datasets from the Census and other sources for station catchments providing somevery localised inputs, but still reliant on some spatially aggregate time-series data for selected variables such as income, bus levels of service etc.
The approach taken threw into doubt the need for station-to-station specific constants. These have typically been used in time-series econometric models in order to explain variations across individual data points in a model which are not captured in the explanatory variable data, either because the data used is too spatially coarse, contains too much ‘noise’, or because there are influences at play which have not been [or cannot be] captured in datasets. When such constants are included alongside spatially detailed data, then they tend to ‘swamp’ effects such as differences in populations, employment, car times and costs etc, at or between stations which may well be important missing links in explaining the observed rail demand growth. Past analyses have tended to fix effects across variables which were not of a direct interest to the study, primarily due to issues of correlation in spatially coarse data, relying on past evidence to do so. This study also questioned this approach, instead relying on freely estimating all parameters
Our econometric analysis revealed that new demand elasticities could be estimated from the spatially detailed dataset, without the need for station-to-station specific constants or fixing specific estimates to prior values. The emerging elasticities are of the correct sign and have an intuitive magnitude across all ten market segments investigated. However, they only represent a partial step in helping to better explain past growth in rail demand, highlighting the need to look further at bringing in additional variables, including additional detail on the types of population at production stations, the types of jobs at attraction stations, coupled with a further investigation of gravity model effects which are commonly employed in multi-modal modelling, but which are currently absent from the uni-modal forecasting approach widely employed in the rail industry.


Association for European Transport