The Identikit fallacy ... or the problem with 'Phil and Jenny'

Pollsters seeking to advise political parties, survey analysts seeking to inform academic research, and journalists seeking to convey the gist of poll results and other survey findings to their audiences need to summarise their data. One method that has proved attractive, especially in Australia, involves an inspection of the marginal frequencies in relation to a particular dependent variable (say, "the swinging voter"'), the identification of the modal response for selected independent variables (demographic, psychographic, or any other), and the listing of these 10 create an identity for the "typical" case. This paper describes such moves, analyses the fallacy involved, and traces its real world consequences. For the media, consequences may include not only a massive distortion of the information being reported but, in applying the methods to their own industry, an overly narrow understanding of the backgrounds of the journalists who report the news, and the caricaturing by management of their own target audience.