San Francisco Chapter Meeting: Speaker Steve Roemerman, Lone Start Analysis
December 15, 2020
2020 was a year of notable polling failures. Some professional pollsters say the year marks the end of their profession.
Why does polling fail? There are a number of reasons, rooted in the mathematics of sampling, in the psychology of biases, and the oddities of human behavior. Polling fails because it is much harder to do it right than incorrectly. In this talk we explore some challenges pollsters predictably face. An irony of the 21st century is that our connectedness has amplified some problems, and created new ones our 20th century predecessors didn’t face.
If Polling is so hard, why bother? Among political pollsters this is a serious conversation at the moment. But others need to do polling for many reasons. This talk will explore some examples of non-political polling and the methods used to address 21st century challenges. Among the responses to polling limitations is use of polling data with Decision Analysis methods, which better incorporates uncertainty.