Two Percent Forever: The Middleman Does Not Die, He Rebrands

The unit cost of financial intermediation in America has sat near two percent for 130 years, through the telephone, the mainframe and the internet, which is why the right demand on any intermediary is not death but justification.

A toll booth is honest about what it is. It stands on a bridge, it takes a coin, and everyone crossing knows what they paid and roughly why. Most intermediaries are not built that way. They sit between two parties who would otherwise deal directly, they take a percentage rather than a coin, and the percentage is folded into a price neither side sees itemised. Every intermediary is a charge on human coordination. A charge is not a crime. It is a claim, and claims can be examined.

Here is the number that should govern the examination. The economist Thomas Philippon measured the unit cost of financial intermediation in the United States over roughly 130 years and reported, in the American Economic Review in 2015, an annual cost of 1.5 to 2 percent of intermediated assets. Revisiting the series after the financial crisis in his later work on fintech, he found it had declined only marginally: improvements in information technology, he wrote, "have not been passed through to the end users of financial services." Across that window communication costs fell by orders of magnitude. The toll held at two percent.

Take the strongest defence seriously, because it is older than the complaint. In 1937 Ronald Coase asked why firms exist at all if prices already coordinate production, and answered that "there is a cost of using the price mechanism": most obviously, he wrote, "that of discovering what the relevant prices are." In 1970 George Akerlof showed that a market in which buyers cannot tell good from bad can unravel to nothing, and observed that "numerous institutions arise to counteract the effects of quality uncertainty," listing guarantees, brand names, chains and licensing. The sharper line comes in his conclusion: "Informal unwritten guarantees are preconditions for trade and production." On this account the middleman is not a parasite on exchange. He is infrastructure for trust. Remove him carelessly and you do not get a cheaper market; you get a thinner one, or none.

Grant all of that, and the argument still does not end where the incumbent wants it. Coase and Akerlof explain why an intermediary should exist. Neither explains why its price should be fixed. If the fee buys verification, and verification gets cheaper, the fee should fall. Where it holds steady while the cost of verifying falls away, what is being sold has quietly stopped being verification and started being position. American merchants paid $187.2 billion in card processing fees in 2024, or $1.57 for every $100 they accepted, according to the Nilson Report, in a payments industry that has been fully electronic for a generation.

Migrant remittances make the comparison clean: the same money moves down the same corridors by different routes at wildly different prices. The United Nations reported that the global average cost of sending $200 was 6.7 percent in the second quarter of 2024, "more than double the SDG target of 3 per cent." Inside that average sits a spread no story about trust can absorb. Banks charged 13.4 percent in 2024. Mobile operators charged 3.9 percent and handled less than one percent of the volume. Nearly ten percentage points, for moving the same money. Part of that gap is the mix of corridors each channel serves. Not ten points of it. The rest is not the price of verifying a stranger. It is the price of the customer having nowhere else to stand.

American residential brokerage then ran the experiment for us. From 17 August 2024, under a settlement with home sellers, the National Association of Realtors' rules changed: offers of compensation to buyers' brokers could no longer be published on multiple listing services, and buyers had to agree their agent's fee in writing before touring homes. The mechanism said to hold commissions up was dismantled. Redfin's tracking of what followed is deflating. The average buyer's agent commission fell to 2.36 percent in the quarter the rules took effect, drifted back up, and stood at 2.42 percent in the third quarter of 2025. The rule changed. The price did not. Whatever pins that rate just above 2.4 percent was never the paragraph in the rulebook, and an argument that cannot tell a rule from a structure will keep winning cases and losing outcomes.

Which is where the slogan about killing the middleman collapses. Disintermediation almost never ends in nobody. It ends in somebody else, usually larger. The web dissolved the high street travel agent, and Booking Holdings reported $26.9 billion of revenue in 2025, most of it from arranging stays in rooms it does not own. The taxi dispatcher was replaced by an application, and in the fourth quarter of 2025 Uber reported Mobility revenue of $8.2 billion on Mobility gross bookings of $27.4 billion, revenue equal to thirty cents of every dollar booked. Part VI of these papers set out the general form of the failure: the internet decentralised publishing and never decentralised trust, so identity, discovery and payment were rebuilt at the application layer and rented back to us. Removing a middleman without answering the question he answered does not delete the booth. It moves it and repaints it.

The cases where the toll genuinely fell are instructive precisely because nobody was eliminated. In the same United Nations accounting, digital remittance channels averaged 5.3 percent against 7.2 percent for non-digital. Those channels are intermediaries too. What changed was not their existence but their contestability: a new one could be found, compared and walked away from. Part V compressed Elinor Ostrom's work on institutions that stay accountable to the people inside them into four conditions, and they translate here without adaptation. Behaviour you can see. Verification you can perform yourself. Rules you can help change. Exit that is real. An intermediary that satisfies those four is a service. One that fails them is Part II's granary keeper with a better interface, and his power never came from the grain. It came from holding the only key.

So the demand is not death. It is justification, repeated, against a baseline that keeps moving. Three questions do most of the work. What uncertainty are you removing? What would it cost me to remove it myself this year, rather than in the year your fee was set? And can I leave? Put them to a bank, a broker, a platform, a court. Some answer well, and there are more of those than the slogan ever admitted; the clearing house and the licensing board earn their keep daily and invisibly. The rest are collecting on a bridge that was rebuilt underneath them while they stayed in the booth. Naming which is which, case by case and with figures attached, is the work this publication has set itself.