I was reading George Soros' book, the Alchemy of Finance, and was impressed by the fact that it has a Foreward to the second edition, a foreward to the First Edition, a new Preface, and a new Introduction. That's a lot of throat clearing.
Getting to his big insight, he posits to key 2 assumptions to reflexivity:
1) markets are biased in one direction or the other
2) markets can influence what they predict
Now, reflexivity is Soros's Big Idea, but I find it rather unexciting. Point #1 basically says markets are unbiased. Unless one can identify the characteristics of an upward bias, as in contrast to a downward bias, that is just a description of asset price fluctuations. That an efficient market researcher would say the asset price has a return equal to the risk free rate plus white noise, is observationally equivalent.
Many critics of efficient markets point to price changes as evidence this theory does not work. Only if one is offering a theory on the particulars of a bubble ex ante, without hindsight. Otherwise, its just another lament on our inability to predict price declines.
Observation #2 is true enough, just think about asset prices falling, leaving to insolvent banks, which creates a multiplier effect. But many models have that mechanism, where a crisis in confidence create self-fulfilling implosions. I think Soros' own experience pushing down the pound during the 1992 European Union currency crisis makes him think this is a bit more radical than it sounds. There are 'bear raids', and these happen, but they usually happen only when there is fundamental weakness. That is, Soros selling the pound in 1992, or John Paulson selling mortgages in 2007, would not have worked if these assets were not also inherently overvalued at that time. I think he gives his correct calls too much credit in creating their own success, as opposed to 'merely' forecasting them.
So, the latter observation could be interesting, but his mechanism for how speculators can create what they predict needs more structure. It's plausible, and surely every price bubble has its share of enthusiastic bulls on the way up and bears on the way down. But it is not clear to what degree they are conspicuous statistical correlates irrelevant to the process.
Lots of really smart people have pet Big Ideas. These are usually pretty lame. Of course, it's a high standard, a Big Idea, as there aren't a lot of new, true and important ideas out there.