How Do Rail Passengers Respond to Change?

How Do Rail Passengers Respond to Change?


A Meaney, OXERA, UK; J Dargay, J Preston, TSU, University of Oxford, UK; P Goodwin, University of the West of England, UK; M Wardman, ITS, University of Leeds, UK


This paper presents the results of a study looking into the
dynamic effects on British rail demand of changes in price, timetables and
delays. It also synthesises the evidence on ramp effects in relation to new
stations and routes.


Oxera has been commissioned by the Passenger Demand Forecasting Council to
provide advice on dynamic effects on the demand for passenger rail in the
British market. The work is being carried out with a number of academic
advisors, including Dr Joyce Dargay, Professor Phil Goodwin, Dr John
Preston and Dr Mark Wardman. The analysis has been completed, and the final
presentation to the client is on February 2nd. Thereafter, the results will
be incorporated by Dr Wardman into his update of the Passenger Demand
Forecasting Handbook (PDFH), which the Council is responsible for
producing, and which is used by practitioners in the British rail industry
to forecast the demand for rail services.

Currently, all elasticities of demand presented in the PDFH assume that the
effects of changes in fares, timetables (including station-to-station
journey time, interchange and headway) and delays are fully incorporated
into passenger behaviour within one year. However, there is a growing body
of evidence, both for rail and other modes of transport, suggesting that
dynamic effects are longer lasting and more complex than the PDFH currently
assumes. In addition, the Handbook currently contains rather limited
evidence on ?ramp effects? - the time it takes for demand for new stations
or routes to build up to the expected equilibrium level. We have undertaken
two pieces of primary research - the first being the estimation of long-
and short-run elasticities of demand to changes in price, timetables and
delays in five different market segments, and the second being the
estimation of revised ramp-up factors for new rail services.

We have obtained a number of important results. Previous rail industry
studies have provided conflicting evidence on the speed of adjustment to
the long run. A meta analysis conducted for this commission by Dr Wardman
demonstrated that studies using industry-standard four-weekly data provide
evidence of faster speeds of adjustment in response to changes in price
than studies using annual data. Our econometric analysis suggests that
previous studies using four-weekly data have not taken account of
significant higher-order lags, either due to the use of a partial
adjustment model, or a simple error correction model. Using four years of
four-weekly data on over 700 flows, we have opted for a fixed effects,
unconstrained error correction model, with changes in volume being
explained by changes in and levels of price, delay and timetables, and lags
of up to three years of volume and the three explanatory variables, plus
GDP. In contrast to the previous studies using four-weekly data, which have
shown speeds of adjustment of up to one year, we find that adjustment to
equilibrium takes between one and four years, depending on the variable
being changed, and the market segment.

We have also noticed that timetable change elasticities used by the
industry are actually short-run elasticities, and that long-run
elasticities are rather larger - meaning that the response of demand to,
say, increasing journey times is likely to be larger than previously

Our work has tested for the presence of asymmetric responses to service
improvements and deteriorations. In order to model potential asymmetric
responses, the variable of interest (such as price) was decomposed into two
monotonic variables: a cumulating series of real price rises, and a
cumulating series of real price falls. The two series then replaced the
variable of interest in the error correction model, specified as above, and
the estimation re-run. A significance test was then applied to determine
whether the rise variable coefficient was significantly different from the
fall variable. We have found evidence of asymmetric responses across all
three variables under consideration.

In relation to ramp effects, we have tentatively concluded using the sparse
data available that it can take three to four years for new services to
reach their equilibrium volume. To arrive at this result, we used a
five-stage process. First, we obtained data on actual patronage for the
first three years of a service funded by the Strategic Rail Authority?s
Rail Passenger Partnership. Next, we identified long-run steady state
patronage using forecasts and service reviews reported in ex-post
evaluations of individual schemes. Steady-state patronage (ie, patronage
without ramp-up) in the early years of the service was then estimated
backwards from the long-run steady state using actual system-wide growth
rates. Annual and (by interpolation) six-monthly ramp-up factors were then
calculated by comparing actual patronage with the estimated demand without


Association for European Transport