Back to the Methodology Page
Questions; Infrequently and Frequently Asked
 
Data Sources
For those unfamiliar with the common sources of passenger statistics, there are five employed at different points in the Air4casts data bases:
◊ ACI
◊ ICAO
◊ CAA for UK data
◊ ATA for US airline data and
◊ airport's own web sites for the most up-to-the-minute data
GDP; Is It Included?
There is a view that all good air passenger forecasting systems must incorporate Gross Domestic Product as an input because there is an implicit assumption that the demand for air travel is a function of national income.
It is quite different to assert that there is a workable relationship between month to month GDP movements and month to month changes in passenger numbers.
While there is an association between economic well being, or at least the anticipation of economic well being, there is no measurable and useful correlation between period to period movements in GDP and changes in the demand for air travel.
The unit of measurement is the coefficient of correlation, expressed between -1 and +1, and the spotlight is on:
◊ USA and
◊ UK
◊ 1990 to 2000 in order to avoid the confusions of post 2001 passenger changes.
The co-efficients are:
Co-efficients of Correlation, 1990-2000
USA
0.56
UK
0.64
where 1.0 signifies a perfect positive relationship and (1.0) a perfect negative relationship.
Even if a workable relationship between the two variables had been identified there are practical considerations set to perplex the forecaster:
◊ economic forecasts are indicative, subject to frequent change and their inherent inaccuracy makes them less than useful as a forecasting input
◊ there are no consistently produced economic forecasts for over half the countries in this analysis
even if there were
◊ it is not clear how movements in GDP at the national level would be related to the demand for air transport at the regional level.
A better correlation is auto-correlation because it has practical applications; auto-correlation is the lagged relationship between past passenger throughput numbers.
Does it Take Account of the New Runway?
The only "knowledge" the system posesses is the passenger history and it bases all projections on that data.
To incorporate external variables on an airport by airport basis would demand an element of pre-judgement about capacity utilization and speed of project implementation.
What is clear is that as soon as the runway becomes operational and passenger numbers are affected, then the forecasting system reacts.
Does it Take Account of the World Cup?
Might appear as a parallel question while it is, in fact, very different.
Exceptional events are exactly that; there is no attempt at predetermination of air passenger demand levels and the one-off effects are soon absorbed into the trend.
One-off events are usually less significant than users imagine.
 
How Accurate are the Forecasts?
Forecast reliability is assessed through the standard deviation of forecast errors

The standard deviation of forecast errors which describes the spread of errors around the mean and can be readily used to assess error probability at different levels of confidence; traditionally 66% and 99%.
Subscribers who wish for further information in this are invited to contact us.
Generally, "accuracy" is in direct proportion to the month to month consistency of the data series; the more irregular the data, the greater will be the spread of forecast errors.
Why do the Forecasts Change from Month to Month?
Because the actuals change is too easy an answer.
The system reacts by giving greater weight to recent than to distant data with the result, on occasions, that it may under or over-shoot.
A useful tool is included on the Airports forecast pages which enables users to inspect the most recent actual data from the airport.
Can We Have Access to the Historical Data?
In principle yes, but not without cost. The costs would comprise royalty payments to Air4casts and third party data providers.
Users are better advised to work with the trend data which is not hugely different from the actuals and avoids the difficulty of omitted months which on average account for only 1% of observations over a fifteen year period but do occur more frequently in the case of individual airports.
 
The Forecast Rates Of Growth Are Lower Than The Historical; Does That Mean That The Industry Is Slowing Down?
A constant forecast rate of growth would mean an exponential series, the end point of which would be infinite increases. The important issue is the absolute size of the increases.
If the Forecasts Aren't Big Enough, Can I Send Them Back?
Is the sort of question which brings a little ray of light to the forecaster's day.