the efficient market hypothesis roughly holds in liquid markets and roughly fails everywhere else. that's the real version of the debate. academic markets with large numbers of traders, tight bid-ask spreads, and continuous price discovery tend toward efficiency. everything else drifts from it.
the grossman-stiglitz paradox captures the core tension. if markets were perfectly efficient, nobody would have an incentive to spend resources on research and trading. but if people don't do that, prices won't be efficient. the paradox suggests that some level of inefficiency is required to sustain the research activity that creates efficiency in the first place.
in practice, this creates a spectrum. equity markets in developed countries are relatively efficient because millions of traders watch them, information spreads quickly, and arbitrage is easy. you can't consistently beat the market through public information. that's roughly true.
but move to less liquid corners and the picture changes. venture capital markets are notoriously inefficient. real estate markets are inefficient. private company valuations vary wildly for similar firms. why? smaller number of traders, slower information flow, higher barriers to arbitrage. a startup raising Series A at 2x valuation from a different investor might be the same company as one raising at 0.5x elsewhere. the prices don't converge quickly.
crypto markets sit somewhere in between and it's revealing. liquidity on major exchanges is deep, so large-cap cryptocurrencies show some efficiency characteristics. but move to mid-cap or low-cap coins and you see sustained mispricings. pump-and-dump dynamics. manipulation. the market structure isn't dense enough to enforce efficiency.
what this suggests: efficiency comes from market density, not from the inherent rationality of participants. make it easy for many traders to participate and easy for them to arbitrage, and you'll get something close to efficient prices. restrict access or add friction, and inefficiency persists.
prediction markets are interesting here. they're a laboratory for testing emh in a bounded domain. when prediction markets have real money and good liquidity, they forecast events more accurately than expert surveys. this looks like efficiency. but even here, you see biases. overestimation of tail events. political bias. the market improves on single experts but doesn't achieve true accuracy.
maybe the right frame: emh is a statement about the speed at which prices incorporate information, not a claim about perfect accuracy. in dense, liquid markets, prices adjust quickly to new information. in sparse markets, they don't. the hypothesis survives if you're willing to be fuzzy about what counts as efficient.
the practical takeaway is simpler. if you're trading in highly liquid markets with many participants and tight spreads, assume prices are right. don't try to outsmart the crowd. but in illiquid, restricted markets, there's real room for edge. information, patience, and analytical skill can add value. the market isn't efficient there.