Scott Armstrong and Kesten Green are seeking suggestions of relevant experimental evidence that they have overlooked in their new working paper, "Demand Forecasting II: Evidence-based methods and checklists". They describe the problem that the paper addresses as follows:
Decision makers in the public and private sectors would benefit from more accurate forecasts of demand for goods and services. Most forecasting practitioners are unaware of discoveries from experimental research over the past half-century that can be used to reduce errors dramatically, often by more than half. The objective of this paper is to improve demand forecasting practice by providing forecasting knowledge to forecasters and decision makers in a form that is easy for them to use.
The paper is available from ResearchGate, here.
Scott Armstrong presented a talk at Heartland's Twelfth International Conference on Climate Change (ICCC12) on March 23 in Chicago that summarised his research on forecasting climate and the effects of climate policies with Kesten Green.
The talk asked the question, "Are long-term forecasts of dangerous global warming scientific?", and concluded...
- the only 2 papers with scientific forecasts found no long-term trends
- IPCC methods violate 81% of the 89 relevant scientific principles
- IPCC long-term forecasts errors for 90-100 years ahead were 12 times larger than the no-trend forecasts
- tests on three other data sets, one going back to 112 AD, found similarly poor accuracy
- the "long-term global cooling" hypothesis was twice as accurate as the dangerous global warming hypothesis
Also "no" because the warming alarm...
- ignores all 20 of the relevant Golden Rule of Forecasting guidelines; the AGS scientific forecasts violated only one
- violates Occam's razor
- fails to comply with any of the 8 criteria for scientific research
- fails to provide scientific forecasts of harm to people
- fails to provide scientific forecasts that "solutions" will work
- fails to meet any of the 10 necessary conditions for successful regulation
- is similar to 23 earlier environmental alarms supported by the government: all lacked scientific forecasts and all were wrong."
A video of his presentation and a copy of a more complete set of slides with links to evidence, is available here.
Andreas Graefe presented a post-mortem of the forecasting of the 2016 U.S. Presidential Election at TEDxMünchen. Andreas discussed why most of the methods used to predict the Clinton-Trump election result were more wrong than usual, and introduced the PollyVote. As regular visitors to this site know, PollyVote combines many forecasts in order to incorporate more knowledge and information, but also to reduce the chances of being badly wrong. You can watch Andreas's talk, which was posted on 11 January 2017, here.
The ForPrin.com Delphi panel freeware had been causing users some problems recently, apparently due to the age of the software. We commissioned an update and repair of the software, and the repaired version is now online on the Software page of this site. Please let us know how the new version of the software is performing, so that we can continue to improve it.
There are at least seven ways to forecast political elections. The most popular among the public is polling of likely voters. As it happens, polls are actually the least accurate method to predict election outcomes. We could combine all of the many polls to improve accuracy, but that brings the polls only up to a tie with econometric models as the least accurate methods.
The solution is PollyVote.com. Launched in 2003, PollyVote provides real time forecasts to demonstrate simple well-tested methods designed to improve forecast accuracy in virtually any area...