What good are polls?

Opinion

There have been a number of complaints on conservative websites about the polls.  A perusal of the polls leads me to the thought that things may be getting a bit wacky out there.  I wrote before that I thought people should stop watching the polls too much.  Here's a reason why--- they can't be very useful because too many factors can alter the results so as to render the outcome of the polls meaningless.

One example was in the presidential election  in 2000.  Bush led by a margin greater than the margin of error, but a last minute disclosure supposedly changed enough opinions so as to make the outcome very close.  The polls in that case were meaningless for predicting that outcome.

This leads me to the next point.  How accurate can they be?  Can they really predict within 5% of the actual results?  Even if they could, this would still be meaningless.  To a politician, 5% may be close enough to try an October Surprise, but as far as predicting a victory or defeat, it is meaningless.  For 5% in a presidential election means millions of votes.  How can you sway that many people that quickly?  It may be possible, but it isn't necessarily a given.  The last minute disclosure of a DUI against Bush may not have been a reason it is was so close.  There may have been other reasons completely unrelated to this.

The methodology of the polling differs as well.  You can have some polls that sample more of one group than another.  You can have polls of adults, registered voters, and likely voters.  Each of those groups can yield differing results.  The demographic makeup of a poll can vary from one poll to the next.  How can anyone be sure that a poll reasonably and fairly reflects all these factors?  It may not be possible to get all these correct, so that the statistical margin of error is only an estimate and it may not even be a good estimate.

My opinion is that polls are only useful for entertainment for those who watch them closely.  For others, such a politicians, they may have some limited value in determining how to run their campaigns.  Aside from that, they probably have little predictive value.