I like Seth's response to this survey on cookie deletion from Atlas OnePoint: "Vindicated!".
For whatever reason, Eric Peterson from Jupiter (who the study that claimed 39% of web users were deleting cookies monthly) is rather circumspect about this new study. He's casting aspersion on the motivation and ethodology of the new survey.
I've looked at a lot of logs, and the data I see is a lot more in line with Atlas OnePoint's than Jupiter's results from a questionnaire. I also trust the examination of log files more than I do the questionnaire approach of first explaining what a cookie is, and then asking how often people delete theirs.