The program schedule for the International Symposium on Forecasting, to be held in Hong Kong on 21-24 June, is now available from the conference web pages.

"Rethinking the Ways We Forecast" is the feature section in the upcoming Summer 2009 issue of FORESIGHT. David Orrell and Patrick McSharry explain why we’ve seen huge breakdowns in our models when we try to forecast complex systems, such as the economy, the weather, and our genetic network. Needed, they argue, are 1) new methodologies that explicitly address key features of complex systems, and 2) a shift of emphasis away from single-point forecasts and toward scenarios of possible futures.

Commentaries from Roy Batchelor and from Paul Goodwin and Robert Fildes punch back, saying, in so many words, “Not so fast -- the new tools haven’t proven themselves and may complexify without improving our forecasts.”

Also in this issue of FORESIGHT:

  • Roy Pearson shows where to find Free and Easy Access to Monthly Forecasts.
  • Jim Hoover shows How to Track Forecast Accuracy to Guide Forecast Process Improvement.
  • Analysts at Hewlett-Packard show how to combine time series, regression, and life-cycle models for Forecasting Demand for Spare Parts.
  • A new column on Sales and Operations Planning, which is perhaps the most significant innovation of this generation in promoting collaborative forecasting.
  • Peter Sephton’s book review of Leonard Mlodinow’s The Drunkard’s Walk: How Randomness Rules Our Lives. “If we can better understand the role of randomness in bringing us to where we are now, we can possibly achieve better mastery over our future.”

For more information about Foresight, go to

The top menu bar now includes a "Principles Checklist" tab. The checklist originates in the Principles of Forecasting handbook as the "Forecasting standards checklist". Clicking on the tab will take you to a new page that provides the option to download any of three versions of the checklist for printing out or copying and filling-in electronically.

Our idea in providing the checklist was that it would be less time-consuming and in some situations more convenient than using the Forecasting Audit Software. The checklist doesn't include the full text of the principles, so users will probably want to have a copy of the Principles of Forecasting handbook handy or a printout of The Principles (see the adjacent tab).

What do you think? Please let us know whether you find this feature useful and how we might improve it.

We have instituted a new "Principle of the Day" feature to highlight a different one of the 140 evidence-based forecasting principles on the site each day. The Principle-of-the-Day is at the top of the right menu bar, just under the page banner. Move your mouse over it and you will see the full text of the principle.

We hope that this feature will help to introduce new people to the principles approach to forecasting, keep the principles fresh in the minds of people who are already familar with them (140 is a lot to remember!), and encourage people to conduct research on principles for which there is only weak evidence. We plan to add relevant links to the Principle-of-the-Day window at a later date. Please send us your comments and suggestions on this new feature.

Is carbon dioxide a good causal variable for forecasting global temperature? Have there been alarms in the past similar to the current alarm over dangerous manmade global warming and, if so, what happened? Can rule-based forecasting help forecast global mean temperatures? What do prediction markets reveal? These questions and more will be answered at a climate forecasting session at the International Symposium on Forecasting presenting work by Green, Armstrong, and Graefe.

To be useful, forecasts should be substantially more accurate than those from a simple benchmark method, for example the no-change model. We suggest taking the following self-administered quiz Please write down your estimate and then follow the link to find the answer.

Q. Assume that at the end of 1850 you started making 50-year-ahead no-change forecasts such that your first forecast was that the global mean temperature in 1900 would be the same as 1850's. By 2008 you would have accumulated 108 forecasts for which you knew the global mean temperature (i.e to 2007). What would be the mean absolute error of your 50-year ahead no-change forecasts in degrees-Celsius?

A. See the abstract of the Green, Armstrong, and Soon paper "Validity of Climate Change Forecasting for Public Policy Decision Making".