The wider applicability of Soros's idea
George Soros noted that market prices of tradeable securities often deviated wildly from what could conceivably be justified by fundamentals. He argued that a feedback loop develops in which the reason for buying becomes the expectation of higher prices in the future. To be fair, it’s a lot more complicated than that. One extra wrinkle is that once a stock is sufficiently highly price, the issuing company can raise finance more cheaply than its competitors. This may be what has happened to Tesla, but understanding what is going on with that stock would probably baffle Soros in his prime, so I’m sure I’m not capable of understanding it.
Florian Kronawitter talks about reflexivity a lot, including in his most recent post. I think he is a lot easier to understand than Soros, either because he has an incomplete understanding of the concept, or because he’s better at explaining concepts or that Soros suffers from the ‘curse of knowledge’ that results in him being not a good person to explain the concept to mere mortals.
Kronawitter makes the point that it’s reflexivity that means that most consensus views about where the economy is going are wrong. He gives a lot of real-world examples that seem convincing to me, e.g.:
Financial Markets - Markets incorporate probable future outcomes in their pricing. This changes the probability of exactly these outcomes. Examples:
1. Whenever growth expectations decline, long-term Treasury bonds are typically bid, bringing down their yield. A lower yield on 30-year treasuries cheapens mortgages, which stimulates economic activity
2. If the market expects a recession, it often heavily shorts oil. This drives down oil prices which then lead to lower gasoline prices, leaving more money in consumers’ wallets which helps the economy. This happened in H1 ‘23
I think the general point is that consumers’ utility surfaces are not static, which means that demand schedules shift, not only in response to externally imposed constraints (e.g. inflation reducing real incomes) but also in response to consumers’ expectations about the future.
To be fair to the macroeconomists, it’s almost impossible to model this behaviour shift in response to expectations. For a start, the expectations themselves may depend on the output of the model. As Kronawitter comments, this is akin to an economic analogue of Heisenberg’s Uncertainty Principle.
What does all this mean for investing? Well, according to his son, Soros made his trading decisions largely on ‘gut feel’ (or, rather, in his specific case, ‘back pain’ feel), so maybe it doesn’t mean very much. Stan Druckenmiller, his protegé, has been incredibly successful as an investor, but he has had some spectacular failures, not least, betting heavily on the NDX just before the “dot com” crash.
The standard way of avoiding the pitfall of trading only on fundamentals is to wait for a trend to establish itself, which may explain the dedication that so many serious traders have to studying charts, whether they are signed up adherents to the dark arts of technical analysis or not. So, the sure-fire way to make money is to identify an undervalued stock, wait for it to start going up, then buy it, right? Well, clearly not. If you doubt me, try it yourself, but don’t bet the farm.
I think that the sort of self-reinforcing feedback loops that explain reflexivity can explain the sort of exponential excursion that is sometimes found in economic data. Macroeconomic models tend to behave as though economic systems are heavily damped and that output measures such as interest rate show strong mean-reversion. Clearly, in some cases this doesn’t work. I am no expert, but I am not aware that any economic model would back-test very well over the 1930s Great Recession, or the 1980’s inflation surge, or the move in the iTraxx in 2007/2008.
Serious economists tend to dismiss Soros’s ideas, mostly because they are difficult to test empirically. Some economists feel that the subject should aspire to be more like history than physics, in other words, seeking to explain, qualitatively, what happened after the event, rather than trying, fruitlessly, to predict the future path of economic statistics that are subject to such a lot of noisy, chaotic feedback.