Combining scenarios with numerical models
We often find that we are asked to develop scenarios to get a longer term view of the future, and then use the scenarios to set values for parameters in an existing model to derive numerical values.
- Working with Helios for Eurocontrol, we were asked to derive scenarios for air traffic in 2050. The existing model for traffic to 2035 was to be used to provide the numerical estimates. Much of the model could be flexed in this way, but other areas needed a deep knowledge of the model to see how to incorporate (emulate) major changes in flying behaviour in relation to GDP, and increases in importance of “over-flights” for instance between Asia and Africa.
- Working with Experian for the Construction Skills Council, we were asked to derive quantitative scenarios for the construction work force in 12 years ahead. SAMI and Experian worked closely together and the scenarios were developed to provide the standard inputs to Experian’s existing short term model. A deep understanding of the model allowed data input changes to accommodate a few parameters that were beyond the model’s normal use. This was also aided by the relatively short timeframe of this project, and the reduced likelihood of extreme changes.
The practical problems that arise in such exercises are
- Access to a person with detailed knowledge of the way in which the model was created, and which parameters are important in determining the factor to be modelled – eg air traffic. It is very rare that a model is built so robustly that all of the underlying assumptions are clear to a reader who only has the spread sheet.
- Ability to change the model and check that it gives the same results as before for previous timeframes or domains. This is important for the credibility of the new results to stakeholders.
When building a model from scratch to explore a set of scenarios, the links between the scenarios and existing budgeting or forecasting systems present similar challenges.