02.03.2013 Views

Downloadable - About University

Downloadable - About University

Downloadable - About University

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Is human probability judgment really so poor? 267<br />

meterological conditions. Furthermore, they have considerable practice<br />

in quantifying their internal state of uncertainty. These circumstances<br />

may well be ideal for the relatively successful application of judgmental,<br />

as compared with purely statistical, forecasting.<br />

Additionally, accurate probability judgments have been demonstrated<br />

in several real-world situations apart from weather forecasting. These<br />

include horse racing, 26 prediction of future interest rates by bankers, 27<br />

and prediction of the success of R&D projects. 28<br />

Human judgment is particularly widely used in sales and economic<br />

forecasting 29,30 although forecasts tend to be single point estimates<br />

(e.g. ‘September’s sales will be 450 units’), rather than probabilities.<br />

Judgment is used here either on its own or in combination with statistical<br />

forecasts and there is much evidence to suggest that it can be useful<br />

and accurate. 31 In reviews of research in this area, Bunn and Wright 32<br />

and Webby and O’Connor 33 suggest that judgmental forecasting will<br />

generally outperform statistical methods when contextual information<br />

is available. Webby and O’Connor define contextual information as<br />

‘information other than the time series and general experience, which<br />

helps in the explanation, interpretation and anticipation of time series<br />

behaviour’. Contextual information therefore embraces unusual events<br />

that are too rare to include in a statistical model, but which have a<br />

profound effect on the variable to be forecast (e.g. a strike at a rival<br />

company which will lead to a massive, but temporary, increase in our<br />

sales) and soft information (e.g. strong rumors that a rival is about to<br />

launch an expensive advertising campaign).<br />

A number of real-world studies support the conclusion of these<br />

reviews. These include studies of forecasts made by managers of company<br />

earnings, 34 workloads at a public warehouse 35 and the sales of<br />

a leading manufacturer and marketer of consumer products. 36 All of<br />

this, of course, does not necessarily suggest that these judgments did<br />

not suffer from the use of heuristics, and the consequent biases, that<br />

we discussed earlier, nor does it say that the judgments could not be<br />

improved. 37,38,39 It does, however, show that human judgment can have<br />

a valuable role to play in many real decision-making contexts, and that<br />

its accuracy can outperform that of statistical methods.<br />

7. People think in terms of frequencies not probabilities<br />

Recently, Gerd Gigerenzer has argued that, from a strong frequency<br />

view of probability, observed bias in probabilistic reasoning is not an<br />

error since probability theory simply does not apply to single events. For

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!