The worst polls are one's that don't reveal their internals. Such polls are completely meaningless but sadly get press all the time. In other cases, polls are published with their internals. This at least lets the poll be evaluated and examined by people who care to do such things. Again sadly it doesn't stop the quick-glance headline from giving misleading information.
As an example, take this poll by SurveyUSA concerning the Missouri Senate race between Talent and McCaskill. The poll results show the race in a statistical tie and allow the article to begin:
It continues to be one of the hottest, if not the hottest, Senate races in the country.However, Alexander McClure of politics.wizbangblog.com takes a look at the internals of this poll and discovers that Democrats have been drastically overrepresented (at the expense of Independents for the most part). Adjust the results for percentages from 2004 exit polls and Talent is ahead 51% to 44%.
Now, the truth probably lies somewhere in between. This is just one more case where some information is worse than no information. The average person doesn't have the time nor the background in statistics to care about the internals of polls. They read "race tied" and move on. One can only imagine how that affects their behavior on election day.
