Professional Documents
Culture Documents
Ma Traffic Forecasts Feb 02
Ma Traffic Forecasts Feb 02
ANNEX
Review of Manchesters traffic forecasts
Paragraph
1.1
1.3
1.9
1.13
1.19
1.21
Introduction
1.1
By their nature forecasts will evolve over time as new information becomes available
and this is especially true for air traffic forecasts at the present time. Manchester
Airport has not yet provided CAA with its definitive traffic forecasts, and this review
focuses on the forecasts underpinning its November 2001 projections. The CAA
expects that, as the impact of the September 11th attacks become clearer over the
next few months, revised forecasts from Manchester, the Competition Commissions
analysis and further CAA analysis will result in changes to the forecasts finally
adopted in setting the price caps.
1.2
The CAA has undertaken an initial review of the likely traffic growth at the
Manchester airport based on various sources, including Manchesters pre- and postSeptember 11th forecasts, forecasts by DTLR (previously the DETR), various
aviation bodies, and aircraft and engine manufacturers. While taking into
consideration these forecasts, the CAAs approach to forecasting traffic at UK
airports involves a combination of SPAM1 modelling, time series analysis, market
intelligence and judgement.
Two of the key inputs in the CAA forecasting process are the estimates of the
growth in the (unconstrained) demand for air travel to and from the UK, and the
assessment of future capacities of UK airports. The latter entails considerations of
runway capacity, terminal capacity and other infrastructure constraints.
1.4
The passenger demand growth estimates are based on the (unconstrained) passenger
forecasts that were published in 2000 by the DTLR2 and the growth rates implied by
SPAM is the Second Passenger Access Model that was developed by the CAA/NATS. NATS holds the
intellectual property rights to the model, but the CAA is permitted on a memorandum of understanding to use
and adapt the model for non-commercial purposes.
DTLRs forecasts and the description of its forecasting methodology are available from DTLRs publication
Air Traffic Forecasts for the United Kingdom 2000.
DTLRs central scenario were adopted by the CAA as part of the inputs in its
forecasting process before September 11th.
1.5
Apart from BAAs London airports which CAA has for the most part relied on
BAAs capacity assessment, the assumptions about runway and terminal capacities of
all other airports largely correspond to the assumptions used in the base case of the
Regional Air Services Co-ordination (RASCO) exercise. These form part of the
inputs for CAAs pre-September SPAM modelling.
1.6
The events of September 11th have had a significant impact on the air transport
industry and have injected additional uncertainty into the short and medium term
outlook for the global economy. Apart from exacerbating the already weakening
world economy prior to the attacks, there are also substantial negative effects on
consumers confidence to fly. The CAAs approach to the forecasting of postSeptember traffic at Manchester airport is based on an assessment of the combined
effect of the economic and confidence to travel factors and on airlines response
strategies on travel demand.
1.7
1.8
The negative impact on travel confidence is assumed by CAA to last for at least 16
months before demand begins to return to its previous trends in each of the market
segments (i.e. from the first half of 2003). Furthermore, the initial impact is assumed
to be the same as the percentage fall in traffic in November 2001 as compared with
the average growth rate between April and August 2001 in each traffic segment.
These assumptions on the duration and initial level of impact on confidence to fly
could be considered as conservative in comparison with the Gulf War experience
where demand was restored to its previous trends within 9-12 months.
The forecasting approach used by Manchester differs according to the length of the
forecast period. There are three different time periods, namely, the short term, the
intermediate term and the long term, and the overall planning process consists of
three stages:
Deriving short-term forecasts for the next three years using a bottom-up
analysis of airline plans;
3 The difference in the forecast UK GDP growth rates as published in the September 10th and November 12th
2001 issues of Consensus Forecasts.
Estimating the throughput for the intermediate term using interpolation from the
end of the short-term forecast period (currently 2004/05) to the 2015 target.
1.10
1.11
1.12
The basic framework for Manchesters development is laid down in its Long Term
Business Strategy (LTBS). Manchesters long-term forecasts have been developed
under various assumptions about underlying market conditions and strategic choices
made by the airlines, which could make a significant difference to traffic growth at
individual airports. In developing its pre-September 11th long-term forecasts,
Manchester has tested a number of scenarios with various assumptions regarding
airport capacity and surface access constraints, low frills operators stimulation effects
at some regional airports, modal shift to rail, economic growth and diversion of
transfer passengers from Heathrow and Gatwick to other European hubs etc.
1.14
Low-frills carriers stimulating the market by 4% above the 1997 DETR midpoint forecast;
1.15
The passenger throughput forecasts using the above scenario are produced using
three different forecasting procedures depending on the passenger type. According
to Manchester, international terminal and international interlining passengers are
forecast using a multinomial logit model called ARTFUL. However, domestic and
international-international interlining passengers and ATMs are forecast separately.
1.16
The process by which ARTFUL generates forecasts is summarised in the flow chart
at the end of this Annex. The model allocates aggregate forecasts by DTLR of UK
scheduled air traffic between the competing airports5 on the basis of surface access
time, service frequencies and fares; the charter algorithm uses access time only.
1.17
The frequency used in the model is related to but not the same as the physical
frequency of airline operations. Rather it is considered as a choice variable capturing
passenger perceptions of relative frequency levels at different airports.
1.18
Twelve UK airports are modelled in the present version of ARTFUL: Luton, Birmingham, Edinburgh, East
Midlands, Glasgow, Leeds/Bradford, Gatwick, Heathrow, Liverpool, Manchester, Newcastle, Stansted.
1.20
Throughput for the intermediate term is obtained by interpolating from the end of
the short-term forecast period to the long-term target in 2015. However, it is unclear
what method of interpolation Manchester has used in the various passenger
categories. Clearly the reliability of this method will depend not only on the forecast
levels of the end points of the interpolation period but also on whether any sudden
changes may occur which would disturb the continuity of traffic growth.
The main differences between the forecasts by CAA and Manchester are laid out in
the main paper under paragraphs 13.16 to 13.24 and are illustrated in Figure 13.1.
Although the divergence in forecasts between CAA and Manchester narrows over
the forecasting period, in the short term, CAA forecasts suggest that Manchesters
post-September base case forecast could be understated by up to 13% by 2004/05.
1.22
In CAAs view, this understatement is mainly due to the overly pessimistic outlook
by Manchester in its European scheduled market and in the scope for recovery in the
North America and Rest of the World markets in 2003/04 and thereafter.
Manchesters forecast of only half of its 8% lost traffic in 2002/03 in its European
charter market would be recovered in 2003/04 also seems to be unduly conservative.
Disaggregate by County
Apply county splits to aggregate forecasts
Disaggregate by Route
Apply route splits to county volumes
a) Scheduled Traffic
b) Charter Traffic
Calculate Demand
Dpd = f(K, T, P)
Is Dpd Threshold
Calculate Demand
Dpd = f(K, T)
No
Close route
Yes
Calculate Frequency:
F = pd+ pd Dpd
Recalculate Demand
Dpd = f(K, T, P, F)
Stop iteration ?
No
Yes
Model Output: Dpd, Fpd