Sure, sure, you can encode difficult computational problems into markets in convoluted ways, but this is uninteresting unless you're into that kind of thing.
When people usually say that markets aren't efficient, they don't mean that optimal resource allocation is a computationally intractable problem. They're saying that they see, clear as day, obvious inefficiencies that aren't being corrected. They're saying that there are easily noticeable inefficiencies that aren't being corrected. And finding easily noticeable inefficiencies isn't NP-hard, by definition of easily noticeable.
A better version of the efficient market hypothesis would be that markets are inexploitable. There is no easy action I can take that corrects a market inefficiency and makes me money. This is the version that comes up in cocktail party conversations and it's also the one that Dan Luu is talking about. "If XXX is systemically undervalued by the market, why can't I start a company that specialises in using XXX?" The discussion following this question is much more relevant and interesting than encoding 3SAT into economic models.
> A better version of the efficient market hypothesis would be that markets are inexploitable.
No, that's absolutely not true and relies on what I think is the most perniciously false assumption behind market economics: that all market participants only make moves within the market itself.
If you're trying to win at chess then learning to master the rules and strategy of chess is where you focus your attention. But if you're trying to defeat your opponent (or not get defeated by them), you'll probably do better to bring a gun and/or body armor. If you focus 100% of your attention on the chessboard, you are completely opening yourself up for exploitation by an opponent who is willing to play the metagame.
And, indeed, in market economics, actors invariably do play the metagame. This is why we get cabals, trusts, rent seeking, regulatory capture, monopolies, price fixing, price dumping, lobbying, etc.
The ultimate goal of market actors is not to be maximally efficient. It's not even to win the market game. It's to make the most money. And often the best way to make the most money is to rig the game, cheat, or get the rules changed in your favor.
Then couldn't you reframe things to include the entirety of human civilization as a kind of market, and still show that there are unexploitable inefficiencies?
Sure, but then your definition of "market" has no meaning.
A book on "chess strategy" would be very different if it's definition of "chess" included the whole of human conflict and warfare. It probably wouldn't spend much time talking about rooks and pawns at all.
That's not what I meant. I mean there's no rule that you have to include only what we call "the market". I mean the "real" global market includes politics, crime, etc. Basically anything humans do to get their hands on money.
It's only uninteresting because economics refuses to be based on anything but the empirical (leading us into the situation we are in now), and so considers things like "math" to be mostly uninteresting "unless you are into that kind of thing".
I looked at markets for a while, this is what i saw. Therefore, it's true.
As I said, there is no reason those efficiencies should be easily correctable - because markets aren't efficient! That's the whole point. It's only in an efficient market that they would be correctable. This is the same as lots of computational and other problems. Lots of instances are easy and heuristics can often work very well. But you will still come up with situations where obviously broken things happen and are hard to make algorithms work.
It's sort of like having a cocktail party conversation about why you can't exceed the speed of light. Except because it's economics, you get stuck because there often isn't any real rigor behind it that you can push on. Just the empirical. Which I get why it's fun to talk about (really!) - without any meaningful rigor, anyone can participate and have fun - got a crazy empirical story? Awesome, that's all you need to prove something in economics! It makes for fun discussions where most people can participate and feel like it's not too hard.
I have no issue with that - my issue is of course that the cocktail party is not distinguishable from the field ;)
Beyond that, arguing they are inexploitable/unpredictable seems equivalent to whether they are efficient (and in most papers is considered equivalent to the efficiency hypothesis). I'm really unsure how you are trying to distinguish it.
Maybe you could formulate it for real and show how it is not equivalent to the efficiency question?
Inexploitability is not efficiency. Back in the 2000s, plenty of financiers knew the subprime market was overpricing assets for years before anyone figured out how to profitably short the market.
Hard to deflate asset-price bubbles are a fact of finance.
When people usually say that markets aren't efficient, they don't mean that optimal resource allocation is a computationally intractable problem. They're saying that they see, clear as day, obvious inefficiencies that aren't being corrected. They're saying that there are easily noticeable inefficiencies that aren't being corrected. And finding easily noticeable inefficiencies isn't NP-hard, by definition of easily noticeable.
A better version of the efficient market hypothesis would be that markets are inexploitable. There is no easy action I can take that corrects a market inefficiency and makes me money. This is the version that comes up in cocktail party conversations and it's also the one that Dan Luu is talking about. "If XXX is systemically undervalued by the market, why can't I start a company that specialises in using XXX?" The discussion following this question is much more relevant and interesting than encoding 3SAT into economic models.