The persistent inefficiency of finance

To finance expansion plans in 1890, a US railroad company would typically hire a New York investment bank to underwrite the sale of stocks and bonds to investors. A prospectus would be printed with a mechanical press, probably using heavy paper stock and Victorian typography. There might be a fold-out map displaying the route of the tracks to be built. Many of these would be sent to London — by steamship — where partner banks would redistribute them to prospective investors, by hand or in the post. Interested investors would place their orders by handwritten note or in person; maybe over lunch in the City. The London bank would send its investors’ aggregated demand to New York by telegraph. The New York bank would have specialist engravers print ornate stock certificates to send back to investors. Ownership was recorded in a bound leather book and updated by hand whenever the securities traded. Final settlement occurred via bills of exchange — again by steamship, in the case of London investors. 

Early finance was slow and costly.

To finance a new datacenter in 2026, an AI company will typically hire a New York investment bank to underwrite the sale of stocks, bonds, and private credit to investors. The prospectus will be made in Microsoft Office and emailed to prospective investors around the world. Commitments to invest will be returned electronically. New shares and certificates are created digitally. Final settlement occurs by bank wire. Ownership is recorded in a database at the DTCC, updating automatically whenever the securities change hands.

Modern finance is fast… and surprisingly costly: despite the advent of information technology, intermediating between companies and investors is probably no more efficient today than it was in 1890.

To test this, finance professor Thomas Philippon undertook the herculean task of estimating how the “unit cost of financial intermediation” in the US has changed over time.

One might expect that computers, the internet, and digitized money would lower the cost of connecting companies with capital: surely sending a prospectus by email rather than steamship would make the finance industry more efficient?

One would be disappointed.

“The finance industry that sustained the expansion of railroads, steel and chemical industries,” Philippon wrote in 2012, “seems to have been more efficient than the current finance industry.” (Emphasis added.)

This is a strange result, because email is definitely cheaper than steamships. But Philippon finds that the finance industry still captures roughly 2% of the capital it intermediates.

In other words, for every $100 of financial assets raised or borrowed by non-financial entities, roughly $2 is transferred to the financial industry in fees, commissions, and trading spreads. 

Most surprisingly, unit costs have increased since the 1970s — the very time you’d have expected computers and then the internet to have lowered them.

“How is it possible,” Philippon asks, “for today’s finance industry not to be significantly more efficient than the finance industry of John Pierpont Morgan a century ago?”

The answer he speculatively proposes is intuitive: we trade too much.

“One possible explanation for this puzzle is that improvements in information technology have been cancelled out by zero-sum activities, perhaps related to the large increase in secondary market trading.”

It’s mathematically impossible for traders and investors (in aggregate) to beat the index. But the cheaper cost of trading makes us more inclined to try. Our efforts cancel each other out and the only thing accomplished is that we’ve transferred more of our wealth to the finance sector.

Philippon's metric is calculated simply as the finance industry's income divided by the amount of useful financial intermediation it performs. All our trading raises the numerator of that equation while leaving the denominator unchanged, causing the unit cost of financial intermediation to rise.

Philippon uses this methodology to further estimate that, as of 2015, investors' unproductive activity amounted to $280 billion a year of additional revenue for the finance industry — primarily in excess trading fees and management fees paid for active-management. 

Why, then, do we keep doing it? Philippon guesses it’s a mix of entertainment and overconfidence. We enjoy trying to beat the market, and we tend to overestimate our chances of success.

The unfortunate result is that the same technological advances that make finance more efficient also encourage us to pay more for it.

The yachts in New York Harbor keep getting bigger.

And none belong to the customers.

The next efficiency trap

The latest technology promising to make finance more efficient is tokenization: putting stocks, bonds, derivatives, and credit onchain.

SEC Chair Atkins has long said tokenization offers the “next step” in market efficiency because it “makes trading and holding securities easier.”

Similarly, BlackRock CEO Larry Fink says “tokenization can replace paper with code, reducing the frictions that make assets costly and slow to trade.”

No doubt true: tokenization would lower the cost of trading by cutting through the legacy tech-debt that makes settlement slow and costly.

But investors should not expect to see any of the savings. If securities get easier to trade, we will trade them more often — 24/7 even, because on-chain markets never close.

Some of us will win. An equal number will lose. 

The intermediaries will take their cut — and the unit cost of finance will rise.

— Byron Gilliam

_{Brought to you by:}

Robots and machines already outnumber humans. They will soon outnumber us onchain, too.

peaqOS turns machines into autonomous participants in the economy — a new, liquid asset class, with yield tied to real-world outcomes. peaq gives machines everything they need to do business on any chain and lets humans earn from automation.

Billions of machines. One layer.