That reminds me of a joke - what's the probability of meeting of a dinosaur in the street? 50% - either I meet him, or I don't!

Not every event with two outcomes is 50/50, you know :) And the article does point to both the subject of the bet - unemployment indexes, which are known to all participants and it's kinda the job of the economist to know what they mean - and why Caplan thought he would win. It may be, of course, that he won randomly and his theory is still wrong - but in general people that successfully predict real world outcomes are regarded better in most scientific endeavors than people that predict wrongly.

> It was 50/50.

1) The fact that there are two possible outcomes does not make the odds 50/50.

2) As others have pointed out, he has a history of winning the bets he makes, well above 50/50.

Considering that he's won 19 of these bets, he'd have to be a very, very lucky person if it was just a 50-50 bet.

They aren't all independent 50/50 bets. There are a few clear trends in the wagers (bullish on the US economy, Eurozone/EU will not completely go to shit), so predicting just a few trends correctly wins most of the bets (though being wrong on those same things would lose most bets) and there are some bets at very long odds (e.g. 200:1 odds on Ron Paul becoming president) so even the person on the other side of the bet doesn't think it's a likely outcome.

50-50 that he'd win the bet by almost 3 percent per year?

That's more like 1000-1 against.