here is a thing that happens in financial markets: someone buys a stock on Robinhood. they tap the button, the trade executes, and they feel like they participated in a market. and they did. but the thing they participated in was not exactly what they think it was. robinhood isn't sending their order to an exchange. it's sending it to citadel securities, which is a market maker, which pays robinhood for the order flow, which is why robinhood can offer "free" trading. citadel then decides when and how to execute the trade, subject to best execution requirements, which are somewhat loosely defined. the customer gets a price. citadel keeps the spread. this is fine! it mostly works! but the mechanism is nothing like "buyer meets seller on an exchange and they agree on a price."
the gap between the story people have about markets and how markets actually work is one of my favorite topics. the story is: supply meets demand, prices adjust, efficient outcomes result. the reality is: bid-ask spreads, order routing, settlement cycles, liquidity provision, and a lot of people extracting rents from the plumbing at each step. macro narratives — the fed raised rates, the market reacted — are mostly a useful fiction layered on top of a much messier reality. the mechanism determines outcomes more than the narrative does.
consider the difference between a centralized crypto exchange and a decentralized one. coinbase has a continuous order book. prices adjust efficiently because there are real buyers and sellers matching up. but coinbase also sees your order before it executes, knows your position, and runs a custodial business where they hold your assets. the microstructure enables things that aren't visible to you as a trader. uniswap does it differently: prices come from a formula (x * y = k, if you care), there's no order book, and every trade is visible on-chain. frontrunning still happens — bots watch the mempool and insert transactions ahead of yours — but it's at least legible. you can measure exactly how much value you're losing to it. on coinbase, you can't. one model embeds opacity; the other embeds transparency. neither is obviously better. they just have different failure modes and different distributions of who extracts value.
miner extractable value is the clearest example of microstructure as a policy choice. in ethereum's current design, whoever gets to order transactions in a block can extract profit by reordering them — putting certain trades before or after certain others to capture favorable prices. this isn't a bug someone forgot to fix. it's a consequence of the consensus mechanism. you can redesign the mechanism to make MEV harder to extract (some designs do this). the result is a different distribution of value. not because anything changed about ethereum's "fundamentals" — not because more applications launched or adoption increased — but because the plumbing changed. a rule changed. the plumbing is the policy.
settlement cycles are another underappreciated example. when you buy a stock, you don't actually own it the moment you click buy. the official settlement cycle in US equity markets is T+1 — trade date plus one day. it used to be T+2 until the SEC shortened it in 2024. before that it was T+3. before the 1990s it was T+5. this choice — how many days between trade and settlement — affects how much leverage is embedded in the system, what happens when a broker fails, what systemic risk looks like during a stress event. robinhood ran into exactly this problem during the GameStop short squeeze in 2021: its clearinghouse required it to post collateral for unsettled trades, the collateral requirement spiked, and robinhood had to restrict trading because it didn't have enough cash on hand. the microstructure created a constraint that looked, to users, like robinhood deciding to shut down their ability to buy stock. the mechanism bit them and everyone downstream of it.
market makers are the other thing nobody thinks about enough. a market maker is an entity that stands ready to buy what you want to sell and sell what you want to buy, earning the spread between bid and ask. this is genuinely useful — it means you can transact when you want to rather than waiting for a natural counterparty to show up. but the spread is real money leaving your pocket. a tight spread (say, a penny on a $50 stock) is fine. a wide spread is a significant tax on every transaction. the market maker's incentive is to set spreads wide enough to cover their risk and generate profit. the trader's interest is in tight spreads. the resulting spread is an outcome of negotiation between those interests, mediated by competition among market makers. you can affect this by changing the rules of the competition — by mandating certain disclosures, by permitting or restricting certain order types, by adjusting how exchanges pay for order flow. all policy choices. all plumbing.
the reason most financial commentary ignores this is that "interest rates up, stocks down" fits in a headline and "the structure of the settlement system and the incentive design of the market maker layer creates transaction costs that systematically disadvantage retail participants while concentrating returns in a small number of high-frequency trading firms" does not. but one of these things actually determines outcomes and the other is a vague handwave at a correlation that sometimes holds. i don't think most people who write about markets are lying. i think they've learned to talk about the thing that's legible and stopped noticing the thing that isn't.
i got interested in microstructure through prediction markets, which turn out to be a good illustration of how much mechanism matters. two prediction markets can be asking the same question but produce wildly different prices because one has better liquidity, a cleaner resolution mechanism, and fewer adverse selection problems than the other. the prices diverge not because people have different beliefs but because the structure creates different incentives for who shows up to trade. if you're building a prediction market and you understand the economics but not the plumbing, the plumbing will kill you. i've watched it happen.