It seems that almost all of the studied phenomena had outcomes determined by other people's emotional responses (presidential primary, idol competition, stock market performance, movie success). These would be expected to correlate with the subjects' emotional responses, as they are likely similar.

This was noted briefly in the paper, but seemed to be largely ignored in the conclusions.

Although the weather study does support the hypothesis, it is a somewhat unfair example, as there is little to go on other than feelings without access to complicated simulation software.

I do believe the fundamental point has validity, but the paper does not seem to support it to anywhere near the level that is implied.

A concrete example of caveats being ignored, from the conclusion:

"The fact that this phenomenon was observed in eight different studies and with a variety of prediction contexts suggests that this emotional oracle effect is a reliable and generalizable phenomenon."