Good predictions don't necessarily correlate with access to secret data, in other words. Indeed, said Philip Tetlock, a University of Pennsylvania professor who heads the project, too much information can sometimes overwhelm analysts and decrease their forecasting accuracy. This problem of separating the essential "signal" from the chatter of "noise" has been well known to scholars for decades. But it seems to have been forgotten by modern-day intelligence agencies in their push to collect.
The NSA obviously operates on the theory that more data are better.