The problem is that banging on buttons has almost no cost. So an expert’s career is based on setting up a Twitter account and subtle marketing about what correct predictions one has. Even a few incorrect predictions are easy to talk down and thus people like Krugman still have a chance to make predictions. Paradoxically, then, it’s more interesting whether their prediction produces a clickbait article, and more valuable the amount of predictions that can be made into long articles for which the journalist is paid by character count, than their reliability.
This so much
I also sometimes see SN as some sort of prediction market: we're supposed to zap what we like, but the rewards are based on what most liked.
So if you want to maximize rewards, you need to predict what others like.