Contact me ·  Browse archives ·  Search this site:  

Sunday · September 19 2004

The Saturday page at is worth reading if you're wondering how the presidential polls can report wildly different results and all claim to be within their margin of error.

It has been a long story today, but understanding the limitations of polling is very important. It is my own feeling that the mass media don't have a clue. They keep harping on the national polls, which are only peripheral, and rarely, if ever, discuss the normalization for political party issue discussed above. They give the (highly misleading) impression that a score of, say, 49% to 41% means that 49% of the people polled were for the former candidate and 41% were for the latter. In reality, that result is the consequence of a lot of statistical massaging of the data based on a model that is rarely openly discussed.

Archived: Political » September 2004
© 2004 Jason Keglovitz