More thoughts on Prediction Markets

I have been carefully following the markets on Inrade.com and the Iowa Electronic Markets. I posited earlier that there was a "gambling" effect that skewed them.

I noticed that the McCain-Win-Contract peaked at 60 in the early voting hours of Jan 15 and then fell to 50 range before plunging as some numbers began coming in. (See 1st Chart)

At 60 the contract was overpriced from the polling expectations. The price should have been around 50 if one were to regard the final polls as accurate. The price only moved to the polling expected level a few hours before results could be reported or ascertained.
Romney-to-Win contract moved in essentially the opposite direction with a slight bias at 55 at the hours just before reports vs. McCain's 50. (See 2nd Chart)

What does this mean (as you think I am an idiot for stating the obvious)? I believe there is a "Gambling" effect that skews the polling favorite higher by at least 10 points on a close field (within margin of error). Also I believe the prediction markets return to polling results at the last moment. If this is the case, then the prediction markets have no more predictive effect than the polls and they are essentially an exaggerated reflection of the polls.

I will test this theory as we go forward.