1. Nine Hundred and Eleven Contracts

In April of 1973, a new exchange opened in Chicago, built for one purpose: the trading of stock options.

On its first day, nine hundred and eleven contracts changed hands, on a roster of only sixteen listed stocks.

Before that exchange existed, American stock options were mostly struck over the counter. Buyer and seller found each other directly and negotiated the terms one contract at a time — the expiration, the strike price, the manner of settlement, no two deals quite alike. An option was a specific arrangement between two specific people, worked out between them, and what they agreed to was theirs to know; no one else needed to.

There was a logic to doing it that way. An option is a wager on what something will be worth on some future day, and whether anyone will take that wager with you depends on whether they trust you, on whether they are already holding the opposite position, on whether the two of you have done business before. Negotiating the terms contract by contract meant negotiating more than the terms.

What the new exchange did was standardize all of it. Strike prices were set out on a fixed grid; expiration dates fell on fixed months; the contract unit was made uniform; clearing was handled centrally; quotes were posted for anyone to see.

A transaction that had been intensely relational was remade into something comparable, calculable, and continuously priced — a market object.

Standardization came at a price: every contract now had to look like every other contract. If the expiration date you wanted fell between the lines of the grid, you took the nearest one; if the strike you wanted was half a dollar off, you took whatever was closest. The contract no longer bent to fit your needs — your needs had to bend to fit the contract. What you got in return was liquidity: a standard contract bought today could be sold tomorrow to a total stranger.

That same year, two scholars published a paper on the pricing of options. What it offered was not merely a formula but an idea: if a dynamic hedge could be built out of the underlying asset and a risk-free asset, then the option had a price that could be calculated — a single, theoretically consistent price.

The standardized contract and the calculable price were two halves of the same thing. One made the option comparable; the other made it computable.

Nine hundred and eleven contracts was the first sound the machine made.

The number is worth keeping in mind precisely because it is so small, so easy to set beside what came later. Within a few decades, the notional principal outstanding on derivatives traded over the counter worldwide would be counted in the hundreds of trillions of dollars. No single year saw a sudden leap. Each year simply did a little more than the year before.

This pattern will recur in every section that follows. The construct's expansion almost never proceeds by a single jump; it proceeds because each year does a bit more than the last, and each additional bit looks perfectly reasonable taken on its own. Look back afterward for the year someone should have hit the brakes, and there is none to find — because every year's increment sat comfortably inside what counted as reasonable that year.

The seventeen essays before this one have followed the construct doing one thing again and again: forcing what will not fit on a single scale into that scale anyway, so that the books can be made to balance. And what it has pressed into shape has always, until now, been something out in the world — a piece of land, a population, a stretch of time, a metal, a country's external accounts.

What happens across the decades that follow is different.

The ledger begins pointing outward less and less. It starts recording other ledgers.

That is not a figure of speech. Later in this essay there are exact numbers: in one market for repackaged products, close to eighty percent of a certain middle layer was bought up by the managers of comparable products, then folded back into the next round of packaging. By the time the construct reaches that point, what it is trying to bring under a common measure is no longer the outside world. It is its own output, one cycle back.

2. Almost Impossible to Value

On May 1, 1975, the American securities markets abolished fixed brokerage commissions. The date came to be called May Day.

Before that, commissions were fixed, and paid as a single lump sum. A regulator at the time put the reasoning plainly: under a system of fixed commissions, trade execution and research services were bundled together into one historically settled total fee, and the two were almost impossible to value separately.

Almost impossible to value — that phrase is worth writing down.

What it concedes is something quite ordinary. What a brokerage does for a client is not only route an order to the exchange for execution. It also shows you research, calls you with a read on the market, introduces you to people, keeps an eye out for opportunities on your behalf, does you a small favor when you need one. All of this mixed together makes up a relationship, and no one can really say what a relationship like that is worth.

When something cannot be said, the workaround is to give it one customary, agreed-upon price for the whole.

This workaround should not be mistaken for muddled thinking. It is a low-cost way of handling things: rather than itemizing and pricing every component, you quote one price for the whole bundle, and everyone saves the trouble. Essay 1 described a world before ledgers in which people's dealings with one another ran on exactly this kind of bundled, unitemized arrangement. It was imprecise, and it held up.

Abolishing the fixed commission did not simply lower fees. It pried the bundle apart and forced every component to be quoted on its own: so much for execution, so much for research, so much for advice, so much for the relationship itself.

This same move has appeared several times already in this series. Essay 5, on double-entry bookkeeping. Essay 10, on the marginal revolution. Essay 12, on housework. Each time it was a different version of the same question: should something that has always lived mixed into a relationship be pulled out and priced on its own.

This time, the answer was: pull it apart.

And this time the pulling-apart had a feature the earlier instances lacked. It was not started by someone who wanted more clarity. It was squeezed out by competition. Once the fixed commission was gone, whoever quoted the lowest price for execution got the order first. Which meant that once unbundling began, it could not stop, because any firm that refused to unbundle would lose business to the firm that did.

The unbundling had an immediate effect. Whatever could be priced on its own began to fall in price, because it now had to face competition. Whatever could not be priced on its own found itself in an awkward spot: it had no price tag, and it could not simply be given away free, so it was either folded back into the price of something else or it slowly vanished.

Once a thing is required to carry its own price, it must prove that it is worth that price. And there are things whose value lies precisely in never being priced on their own.

This is not nostalgia. A broker who is willing to call you the month you lost money, an analyst willing to share a judgment before it is even finished — these things happen partly because they were never priced, and so never had to answer for their return on investment. Once everything carries a price tag, everything has to survive the same question: on its own, is this one thing worth what it costs.

Moves in the same direction followed one after another in the years that came next. A law passed in 1980 loosened the constraints on depository institutions. In 1981, the World Bank and a computer manufacturer completed what would become the often-cited first formal currency swap: debts in different currencies, on different terms, carrying different funding costs, could now be recut, reassembled, and repriced through a single derivative contract.

From that transaction on, a debt was no longer only a debt. It became raw material that could be reassembled with other cash flows.

Reassembly is the key move of the decades that follow. A debt had originally been money one person owed another, and its term, its currency, its interest rate all grew out of that relationship. What a swap does is pull those elements apart: the maturity you want is yours to keep, the currency I want is mine, the interest recalculated on a different footing entirely. Once taken apart, the debt is still that same debt — but it no longer belongs to the relationship it was born in.

3. An Engine, Not a Camera

One researcher called the 1980s the decade of trading.

Working from American data, she found something specific: non-financial corporations' income from securities, measured against their cash flow, began climbing noticeably in the 1970s. The ratio of financial to non-financial profits, which had held fairly steady through the 1950s and 1960s, rose through the 1970s and then rose sharply through the 1980s, ending up, by the close of her sample, at roughly three to five times the typical levels of those earlier two decades.

What mattered more was the composition of that rise. The jump in securities income at non-financial firms was driven mainly not by capital gains from the stock market but by interest income.

In other words, manufacturers, industrial firms, retailers — companies that were, in name, part of the real economy — were becoming steadily more entangled in relationships of credit and debt. The factories kept running, the goods kept selling, and yet a growing share of the line at the bottom of the income statement was coming from money itself.

The effect inside a firm was concrete. Once a manufacturer discovers that putting idle cash out to earn interest is steadier, faster, and easier to show off in a quarterly report than opening another production line, the next time it weighs whether to open that line, it already has something to measure it against. Production does not have to be forbidden; it only has to sit on the same page as other uses of the same money and be found wanting.

Meanwhile, pricing models had begun doing something that went beyond description.

A sociologist later called models of this kind engines, not cameras. A model does not merely stand off to the side and photograph the market. It enters the market and changes it.

First the standardized option was created. Then came hedging. Then market-making. Then the curve of implied volatility. Then volatility itself became something that could be traded in its own right. Every new pricing formula gives birth to a new object of trade, and every new object of trade calls for a new formula.

The stock market crash of October 1987 showed the other face of this same fact.

A strategy called portfolio insurance was fashionable at the time. Its logic was clean: when the market falls, sell stock-index futures to hedge the loss away. One insurance policy, one model, one rule that executed itself automatically.

But when a great many people are acting on the same model at the same time, a falling market produces this sequence: the model calls for more selling; the selling pushes the price down further; the falling price triggers the model to call for still more selling.

A mechanism built to guard against decline became, in the middle of a decline, a mechanism that accelerated it.

The official reviews that followed, along with the research that came after them, pointed to exactly this dynamic hedging as a force that reinforced the downward, pro-cyclical selling pressure in a falling market. What shows up here is a pattern that would repeat again and again afterward: the model does not stand beside the market describing it. The model stands inside the market, driving the trading.

This tells us something important about the nature of the construct. In the essays before this one, the scale illuminates the remainder, the scale revalues an hour of someone's labor, the scale compresses a whole stretch of time into a single point — but what the scale does is always, in the end, to measure. Here, for the first time, the scale starts acting on its own. It reads out a number; that number sets off a trade; that trade changes the very thing that was being measured; so the next reading comes out different again.

The thing being measured no longer holds still. It moves according to the reading the scale hands it.

This is the single most important change in the construct's shape across this whole trajectory. In the previous seventeen essays, there was always still some distance between the construct and the thing it measured: the ledger recorded what had already happened; the scale measured something that was already sitting there, and however badly it was measured, the thing being measured did not change simply because it had been measured. Here, that distance closes. The price enters the very event it describes and becomes part of that event.

By the 1990s, this whole way of speaking had spread everywhere. In 1999, the chairman of the American central bank said in a speech that the single most important financial development of the preceding decade had been the extraordinary growth of financial derivatives. The following year, testifying before Congress, he said that over-the-counter derivatives let people break risk apart and hand it to whichever investors were most willing, and most able, to bear it.

This is a very characteristic statement of what might be called efficiency theory, and it is not wrong. Risk does not disappear; it is reallocated to wherever it can be better carried. A farmer can pass off the price risk on his harvest to someone willing to take it on; a company can pass off its currency risk and put its full attention into the business it actually runs.

These benefits are real, and not minor. A farmer who can transfer away his price risk can put his mind fully into growing what he grows best, instead of spending half his attention guessing at prices; a company that does not have to watch the exchange rate every day can put its energy into making a better product. The division of labor can deepen only because someone, somewhere, is willing to pick up the piece of uncertainty that someone else does not want to carry.

This account frames financial innovation as a deepening of the division of labor: risk gets taken apart like a machine into its components, reassembled, and sold on to whoever is best suited to hold each piece.

It rests on one premise: that the components, added back together, still amount to the original thing.

For a long time, that premise went untested, for a simple reason — it held true on ordinary days. A risk split three ways and sold to three parties will, on an ordinary day, be handled by each of those three separately, with no bearing on one another. It is only on the day all three want out at once that anyone discovers the three pieces had been connected the whole time.

4. The First Duty

Financialization did not happen only on Wall Street. It moved inside the company as well.

In 1997, a statement on corporate governance from the Business Roundtable put it about as plainly as such a thing can be put: the paramount duty of the corporation is to generate economic returns for its owners.

The same document went further: the highest obligation of management and the board is owed to shareholders, and the interests of every other stakeholder are merely derivative of that obligation.

Translated into plainer language: a company's first duty is not employment, not the accumulation of technical capability, not the fortunes of the city it sits in. It is returns to shareholders.

What this way of talking was doing is easy enough to see. Inside any company there had always been a pile of goals with no common unit between them: jobs to protect, long-term research to fund, supplier relationships to maintain, a local reputation to look after, technical know-how to keep building. None of these share a common denominator; how much weight to give one over another can only be decided by a person's judgment.

Leaving it to a person's judgment has its own defects. It is slow, it is opaque, it depends entirely on who happens to occupy the seat, and it leaves management wide latitude to talk itself into whatever it likes. The language of shareholder value drew its force precisely from targeting these defects: with one shared number in hand, who was doing well and who was doing badly no longer required listening to anyone's explanation.

Shareholder value gave all those goals a single common unit.

In 2001, a scholar wrote the idea out more systematically. He argued that once every good has a price, social welfare is maximized when a firm maximizes the entire long-run market value of the firm — and that this total value is simply the sum of the market values of every financial claim on it: equity, debt, preferred shares, warrants, all of it.

The power of this formulation lies exactly here: it turns an enormously complicated organizational problem into something that looks like a single objective function, expressible in a market price.

And the overflow can be found inside that very same document.

The same governance statement that sang the praises of shareholder value, even as it pointed the corporation's whole purpose toward shareholder returns, also conceded that the truly decisive factors in governance — the caliber of the directors, the character of the chief executive and the board, whether anyone in the room is willing to stand up to the chief executive — are soft, subjective, and resistant to any tidy comparison.

A text more devoted than almost any other to reducing the corporation's purpose to a single number was, in its own words, admitting that the corporation cannot run without personal qualities and relational judgments that cannot be measured at all.

Law and financial practice pushed in the same direction. A rule adopted in 1982 gave issuers a safe harbor for repurchasing their own shares on the open market, and after that American companies' way of distributing cash shifted visibly toward buybacks. Among the companies in the S&P 500, the scale of share buybacks in 1980 came to roughly a tenth of what was paid out in dividends; by 1997 and 1998, buybacks had overtaken dividends.

Later still, researchers examined the four hundred and forty-nine companies that stayed continuously in that index from 2003 through 2012, and found they had put fifty-four percent of their profits into buying back their own stock, and another thirty-seven percent into dividends.

One researcher's account of this history is unsparing: before the 1970s, the strategic center of gravity at large American corporations leaned toward retaining earnings and reinvesting them; after the 1980s, it revolved increasingly around downsizing, distributing cash, and what came to be called shareholder value.

The decisive change was not that executives had simply grown greedier. It was that the visible scale inside the company had changed.

Long-term research and the slow accumulation of a workforce's skills are hard to discount into a present value on demand, while share price, earnings per share, buybacks, acquisition premiums, and the rate of return of every division can be ranked, held accountable, and rewarded every single day.

Production did not vanish; it simply came to need financial language to justify itself. The factories, the retail networks, the patents, the brands, the customer relationships were all still there — they just had to pass through discounted cash flow, cost of capital, shareholder return, and market multiples before they were granted legitimacy.

Whatever did not absorb easily into this common scale was treated, inside the company, as a cost center, and, in the capital markets, as a reason to discount the share price — and was then outsourced, trimmed, securitized, or sold off.

The sequence here is worth noticing. It was not that someone first decided to cut these things away. A scale appeared first; these things produced an ugly reading on it; and only then did cutting them become an obvious decision. The person making that decision did not feel he was sacrificing anything — he felt he was improving a number. Essay 13 made a related point: when every single part is handled on its own, each act of handling looks perfectly reasonable, and yet add up all those reasonable acts and the position itself has ceased to exist.

5. Seven Links in a Broken Chain

The same logic of taking things apart reached its furthest extreme along one particular line: the home mortgage.

In a traditional bank, a mortgage loan was one continuous affair. The bank examined your income and decided whether to lend. The bank lent you its own money. The bank kept watch on whether you were repaying it. And if you fell behind, the bank was the one that had to deal with the house. All of these things stayed tangled together, start to finish, on a single balance sheet.

That tangle served a purpose. Because the money was the bank's own, it had reason to scrutinize you closely; because it meant to hold the loan to maturity, it had reason to keep watching you; because the house would land back in its own hands, it had reason to work something out with you before simply taking it away. These things stayed bound together not because no one had thought to separate them, but because bound together, the people doing each of these jobs had a reason to do them well.

Securitization took that single affair and cut it into an assembly line.

Researchers who have taken the subprime securitization chain apart in detail count at least seven distinct categories of critical information friction, spread across eight roles strung along the chain: the mortgagor, the loan originator, the arranger, the warehouse lender, the rating agency, the servicer, the asset manager, and the investor. Theoretical remedies existed for these: due diligence, repurchase guarantees, haircuts on warehouse lending, credit ratings, credit enhancement, servicing agreements.

The whole apparatus of remedies can be summed up in a single sentence: replace a reason that never needed to be written down with something that can be written down instead. The bank used to watch you because the money was its own; now the servicer watches you because a contract says it must. The old reason grew directly out of a stake in the outcome. The new reason lives on paper. Paper works too — but paper needs a person to carry it out, and that person has calculations of his own.

Shadow banking as a whole took the same shape. Researchers have broken its credit intermediation down into seven steps: loan origination, loan warehousing, the issuance of asset-backed securities, the warehousing of those securities, the issuance of re-packaged products built out of them, securities intermediation, and wholesale funding.

A traditional bank kept deposits, lending, maturity transformation, liquidity support, and capital buffers all under one roof. Shadow banking sliced these functions apart vertically and handed them out to different nodes: mortgage companies, special-purpose entities, broker-dealers, money-market funds, the repurchase market.

The scale involved was not small. By June of 2007, on one flow-based estimate, the total liabilities of the shadow banking system stood at close to twenty-two trillion dollars, against roughly fourteen trillion dollars for the traditional banking system over the same period. In that same stretch, the American repurchase market ran to somewhere around twelve trillion dollars — on the same order as the ten trillion dollars in total assets held by the American banking system.

In other words, a very large share of the most critical credit activity in the economy was no longer taking place in the deposit-taking and lending that ordinary people recognize. It was taking place in wholesale markets, collateralized by highly rated securities, kept alive by funding rolled over every night or every few days.

Hot money flowed toward this same chain. In 2006, the origination of non-agency mortgages in the United States reached one point four eight trillion dollars, more than forty-five percent above the volume of agency mortgages; issuance of non-agency mortgage-backed securities came to one point zero three three trillion dollars, likewise ahead of the nine hundred and five billion dollars issued through agency channels.

The hottest money in housing finance was spilling out of the system built around more uniform underwriting standards and into the private machinery that depended instead on the securitization chain, on ratings, on underwriting, on warehouse funding and repo funding.

On its surface, this whole system was intensely anonymous.

Savers held shares in a money-market fund, not deposits in a bank. Money entered the system through commercial paper or repurchase agreements, never crossing a teller's counter. Collateral took the form of rated securities. Risk was managed through credit default swaps, insurance, and liquidity-support clauses. Counterparties spoke to one another through standardized master agreements and collateral schedules, not through personal acquaintance.

Essay 6, on the paper currency called jiaozi, observed that stripping the metal out of money is the same as pressing its entire weight onto personhood. What happens here is the opposite effort: stripping personhood out of every link in the chain, and replacing it with documents, ratings, haircuts, and standard terms.

And it cannot be stripped out.

The arranger still has to prove itself to the warehouse lender upstream, still has to stake its own reputation with the rating agencies and the investors, still has to sign representations and warranties. The servicer still has to be trusted to handle delinquent loans in a way that lines up with its own incentives.

Beneath the anonymous prices, the relationships had not disappeared. They had only been broken apart and tucked into every joint of the chain.

The difference between hidden and gone does not show up on an ordinary day. A chain with seven joints has, at every joint, a small amount that can only be carried across on trust, and on an ordinary day each of those small amounts is too small to notice. Only on the day all seven are questioned at once does their sum become visible.

6. The Books Begin Reading Each Other

This is the point at which the single most important phenomenon in this whole story can be seen clearly.

Once a mortgage loan has been securitized, what results is not one complete security but something sliced into tiers: the senior tier is paid first and absorbs losses last; the next tier down absorbs more; the tiers at the very bottom absorb losses first of all. Each tier carries its own rating, its own spread, its own price.

There is a logic to this slicing in itself. Different investors have different appetites for risk, and cutting a loan into tiers lets the cautious buy the top and the risk-tolerant buy the bottom — which is exactly the work of allocating risk to whoever is most willing to hold it.

And this slicing solved a genuine problem. A single mortgage loan, thirty years long, worth some hundreds of thousands of dollars, owed by one specific person, is almost impossible to sell on any market. Bundle several thousand such loans together, then slice the bundle into tiers ranked by who absorbs losses first, and what could not be sold before now finds buyers; capital can flow from those willing to hold it for the long run to those who need money to buy a house. This is what commensuration, at its best, actually achieves.

The trouble lay in where the middle tiers ended up.

The commission that investigated the crisis, looking into the market for re-packaged products in 2006 and 2007, turned up a detail that is hard to look away from: in some of the samples it examined, close to eighty percent — in some cases as much as eighty-eight percent — of the mezzanine tier was bought up by the managers of other, similar products, to be stuffed into yet another round of packaging. In the most extreme individual cases, every single buyer of a given product was itself a manager of some other, similar product.

That sentence deserves to be read slowly.

A machine that was supposed to disperse the risk of home mortgages out into the wider world was, increasingly, devouring and multiplying itself among its own products.

The middle tier from the first round of packaging could not be sold, so it was folded into a second round. In that second round it was sliced again into upper, middle, and lower segments, and the upper segment of that second slicing won back the highest possible rating. The same risk had gone in a circle, changed its name, and come out rated higher than before.

No one, in any of this, was committing fraud. Every step had a model behind it; every rating at every step was calculated by the methods accepted at the time; every step was set down in a document somewhere. The reason a higher rating could be reached at all rested on one specific assumption: that the things stuffed into the same package would not run into trouble at the same time. The assumption was written into the methodology, available for anyone to read. Not many people read it.

The ledger was no longer recording only the cash flows from home mortgages. It had begun recording other ledgers.

In the seventeen essays before this one, whatever the construct was doing always had an object outside itself: a piece of land, a population, a stretch of labor, a metal, a country's balance of payments. There was tension between the construct and that object, there was overflow, there was a portion that would not balance — and that object always stood apart from the construct, out in the world.

Here, for the first time, the construct begins to take itself as its own object. It prices its own product, then prices that price, then prices the contract written on top of that pricing. Every layer is calculated with perfect clarity, every layer carries a market quote — and at the very bottom of the whole chain sit hundreds of thousands of individual mortgage contracts, and hundreds of thousands of people who owe a payment every month.

How many layers separate those people from the top of that chain is something the people at the top can no longer work out.

And not being able to work it out is no problem at all, in good times. Because there is a price.

This sentence is the spine of the whole essay. Not being able to calculate something and not knowing it are two different conditions. When you cannot calculate something, all you need is a price, and you can stop asking. Price stands in for the asking, and stands in for it thoroughly, because a price looks more objective than a question, more immediate, and much easier to justify to somebody else.

In January of 2006, an index tracking a basket of subprime securities began trading. Built on a set of credit-default-swap prices, it gave what had previously been an extremely opaque pocket of risk a thermometer that could be watched continuously.

From that point on, the future of tens of thousands of individual mortgage loans no longer showed up only as a delinquency rate or a foreclosure rate. It was folded, instantly, into the level of an index.

This was another victory for commensuration, and a genuine one. What that enormous pile of loans was actually worth had been anyone's guess; now there was a number every single day.

The more transparent a price becomes, the easier it is to mistake a visible price for a knowable value.

And outside this whole apparatus of unified measurement, certain things had never been counted at all.

Whether the borrower understood the contract he had signed. Whether the house would be kept up. Whether the servicer's incentives had been bent out of shape. Whether an entire region's home prices might fall together, all at once. Whether borrowing anew to pay off the old could go on indefinitely. Whether a market maker would still be willing to take on inventory on an ugly day.

None of this was imperceptible. It simply had no column of its own on the ledger.

Things with no column of their own share the same fate: they do not show up on the sheet as a zero, they do not show up at all. A person looking at a report does not feel that anything is missing, because there is no blank space on the page to remind him. Essay 12 observed that where the scale casts no light, people assume there is nothing there. Here the situation goes one step further: the scale shines so brightly that where it fails to reach, not even a shadow can be made out.

The single most important thing left in the dark had a name: correlation.

Looked at on its own, every loan's probability of default could be estimated. Looked at on its own, every tier's distribution of losses could be calculated. Looked at on its own, every price matched the market. But whether they would all default at once was not a property that belonged to any single asset.

It lived between the assets.

Researchers looking back summed it up bluntly: in the years before the crisis, rating agencies, risk managers, investors, and regulators alike badly underestimated how correlated the prices of structured securities would become under extreme conditions.

The remainder had changed its address. In earlier essays, the remainder sat on the excluded side of things — outside the register, inside a relationship, in a central bank's vault. This time, the remainder was not lodged in any single line item at all. It lived on the connections between line items, and a ledger, by its very nature, is kept item by item.

A book that records the world row by row has no way to open a column for the relationship between one row and another.

This is a structural limit left behind by double-entry bookkeeping, and it has not changed in over five hundred years. Essay 5 described that self-verifying ledger: every entry must have its matching entry, debit equal to credit, and the book closes. The power of that method lies in anchoring every single item in place. What it cannot do is open an account for the line connecting two items to each other. And in these particular decades, the real risk was growing precisely along those connecting lines.

7. Ninety-Seven Cents on the Dollar

Between 2007 and 2008, this machine ground to a halt, and the manner of its stopping tells you most of what you need to know.

Researchers have summarized the panic of those two years as a run on the repurchase market.

This kind of run looks nothing like a traditional bank run. Investors did not line up at a teller's window. What they did instead was raise the haircuts demanded on collateral, refuse to roll over financing, demand more high-quality collateral, shorten the terms on offer, and widen the spreads they charged.

On the surface, it looked like securities prices simply moving around. Underneath, it was a wholesale collapse of confidence — in counterparties, in market-makers' willingness to hold inventory, in the reliability of ratings, in the quality of collateral, in who, in the end, would be left holding the position.

Put another way, and it becomes clearer: on an ordinary day none of these questions need to be asked, because there is a price, a rating, a standard contract. A run simply means that people have suddenly started asking these questions again. And once they start, not one of the questions can be answered, because the entire system had been built for the specific purpose of letting people go without asking them.

There had already been a rehearsal for this, and a remarkably thorough one. In 1998, a hedge fund came close to collapse. The investigation that followed described it without any softening: on four point eight billion dollars of capital, the fund had built positions in over-the-counter derivatives with a notional principal exceeding one trillion dollars, plus a further hundred and twenty-five billion dollars in securities positions — and its principal counterparties and its regulators had not known, beforehand, the true scale of what it was carrying.

That episode was not the direct cause of the crisis that came a decade later. But it had already staged the whole problem once: high leverage, a web of over-the-counter contracts, opaque counterparties, and a faith that markets would discipline themselves.

Staged once — and over the following decade, the very same thing grew several dozen times larger.

This deserves to be recorded on its own, because it reveals another layer of what the construct is. The construct is not incapable of learning. There really were reviews after that episode, investigations, discussions about counterparty transparency. But the conclusions those reviews reached were, almost without exception, technical: report a few more data points, hold a little more margin. Rarely did a review conclude that the whole approach was itself amplifying the very same fragility — because once that conclusion is accepted, the people making money from the approach have to go find another line of work.

On September 16, 2008, a money-market fund announced that its net asset value had fallen to ninety-seven cents, breaking below par.

Shares in money-market funds had always, up to that point, been treated as something close to cash. People put money into them and treated the money as a deposit — available at any moment, never at risk of loss.

What made breaking the buck so alarming was not that it broke a price. It broke an assumption.

And that assumption had never been formally promised by anyone at all. The fund's prospectus stated that net asset value would fluctuate, stated that principal was not guaranteed, stated that investing carried risk. Everyone had read those words, and everyone had treated them as boilerplate. What actually persuaded people to put their money in was a simple fact: for more than twenty years, the fund had never once broken the buck. A promise that was never written down, once it breaks a single time, can never be made whole again.

Within days, the United States Treasury introduced a temporary guarantee program for money-market funds, covering the balances that qualifying funds held as of September 19.

The instrument that looked most like a substitute for cash, the share that had been stripped of personhood most thoroughly of all, still needed the state, in the end, to step in and stand behind it.

This is the most ironic thing about financialization.

At the outset it seemed to stand for a thorough depersonalization: no longer dependent on whether the branch manager knew your face, no longer resting on a long relationship, but built instead on uniform pricing, uniform ratings, uniform collateral, and a standard contract.

And yet, at the moment of crisis, the questions the market rushes to ask are exactly the personal ones: who is still willing to finance you. Who still dares to underwrite your securities. Who is willing to make a final quote. Who can prove that your top rating is not just a thin shell. Who, in the end, can step forward and stand behind the whole thing.

Anonymous instruments of exchange never actually killed off personal, relationship-based credit.

They only pushed that personal credit upstream, concentrated it into a handful of key institutions, and finally tucked it behind the state.

The end of Essay 17 observed that what was cut open in 1971 was a covering — that the forces that had been operating all along were, for the first time, brought out into full view. What happens here is the sequel to that same line: once the covering was cut open, the personal element did not return to everyone equally. It was concentrated into a smaller number of nodes.

The harder anyone tries to escape personhood, the more certain personal guarantees have to be concentrated into fewer hands, and more powerful ones.

8. Too Big, or Too Dear

About what actually happened across these decades, researchers disagree sharply, and the disagreement is not a matter of taste.

One camp might be called efficiency theory. It does not deny that the financial sector swelled enormously after the 1980s, but it stresses that a great deal of that growth answered real needs in the economy. Derivatives break risk apart and reallocate it to whoever is best suited to hold it; a more developed financial system helps generate information, share risk, and allocate capital; the literature on finance and growth generally finds that the development of stock markets and banking correlates positively with growth. One scholar, in 2013, simply rewrote the whole question in terms of function rather than size: whether finance is too big is not, in itself, the most useful question to ask; what matters is whether finance's function is being performed well.

A second camp might be called extraction theory. It fastens on the structure of profit: profit was increasingly captured through financial channels, and even non-financial firms grew steadily more dependent on interest and securities income. One scholar offered a broader definition, describing financialization as the pressure that financial motives, financial markets, financial institutions, and financial actors exert on how the whole economy is run. Other research, approaching from the angle of corporate governance, points out that shareholder value is nothing like an old, self-evident common sense — it only rose to the status of a governing principle after the 1980s. On this reading, the expansion of finance was not a neutral deepening of intermediation but a redistribution of the social surplus and of corporate control, tilting steadily toward financial creditors, institutional investors, executive incentive schemes, and asset managers.

Wedged between the two camps is one especially consequential empirical dispute: did the financial sector balloon because there was simply more that could be intermediated, so that intermediation naturally grew larger — or was roughly the same service, or even less, being delivered at a steadily higher cost.

One scholar's research is cited by both sides at once, and precisely because his answer is complicated. He acknowledges that the financial industry's share of output was high in the 1920s, low in the 1960s, and rose again after the 1980s, and he estimates that the long-run unit cost of financial intermediation in the United States has held roughly between one and a half and two percent of the assets being intermediated — and yet, by his own measure, he also finds that this unit cost has actually risen over the past thirty years.

The efficiency camp seizes on the first half of that finding; the critics seize on the second half. But the real weight of the finding lies elsewhere: even without accepting any of the more radical claims at all, the simple, common-sense notion that information technology has made finance steadily cheaper does not hold up as easily as it sounds.

There is also a camp that might be called instability theory. In 1992, an economist divided the financing relationships found inside the financial system into three kinds: hedge finance, speculative finance, and Ponzi finance. He wrote, without hedging the point, that where hedge finance predominates, an economy may tend toward self-restraint, and that the greater the share taken up by speculative and Ponzi finance, the more likely the economy is to become a system whose deviations amplify themselves.

He put it even more bluntly elsewhere: bankers and financial intermediaries are merchants of debt, and innovation guarantees profit.

What separates this camp from efficiency theory is not whether finance can transfer risk. It is something else entirely: success during a boom is precisely what tempts the whole system into shaving its margin of safety thinner and thinner, until even the assets that look most liquid, safest, and most easily measured turn out to be the most fragile of all.

There is disagreement, too, about where all this came from. Some researchers have argued that American financialization was never a goal policymakers set out to pursue from the start. It looks more like a string of emergency responses — politically convenient at the time, economically consequential for decades afterward — that the state improvised once growth slowed in the late 1960s and the 1970s, in order to sidestep conflicts over distribution, fiscal pressure, and pressure on its own legitimacy.

On this account, financialization is neither a conspiracy nor the execution of some genius's design. It is a path that took shape amid argument, trial and error, delay, and the failure of one alternative after another.

Even the official commission that investigated the crisis could not settle on a single narrative. The majority report emphasized regulatory failure, leverage, shadow banking, securitization, and derivatives; the dissent placed the global credit bubble, capital surpluses flowing out of developing countries, the underpricing of risk, and narrowing spreads closer to the center of the story.

No verdict will be rendered here.

But put these accounts side by side, and what they are really arguing about turns out to be three quite specific questions: whether price really does move closer to value; whether financial claims really do stay close to the underlying assets beneath them; whether an anonymous market is really more reliable than credit based on relationships.

Three questions, three separate lines of evidence, and not one of them yields a tidy answer.

Withholding a verdict matters especially here, because the answers to these three questions may well not agree with one another. That financial claims drifted farther from the underlying assets is, on the evidence, reasonably clear. Whether price moved closer to value depends entirely on which period and which class of asset you mean. And as for which is more reliable, the anonymous market or relationship-based credit, the answer these decades give is that both fail — just not in the same way.

And underneath all three of these questions sits something even more elementary.

At the very bottom of a mortgage loan is a person who owes a payment every month. He has a job, or once had one. He signed a contract he may not fully understand. He lives in a house, and whether that house gets taken care of depends on how much of himself he has left to give it.

None of this can enter the machine. What the machine needs is his credit score, his loan-to-value ratio, the ratio of his monthly payment to his income. These numbers genuinely do relate to whether he will be able to keep paying — and they are not him. What the construct does was never conjuring something from nothing; it takes a handful of parts from a person that can be reported as numbers, and then lets those parts stand in for him. Ordinarily, that is enough.

Between that man and the top of the machine lie seven links, several layers of tiering, several rounds of repackaging, an index, a great pile of models, and a stack of master agreements.

Every layer prices the layer below it, and what every layer sees is only the price of the layer just beneath — never the layer beneath that one.

In good times this is no problem, because every layer's price lines up with the next. When the crisis comes, everyone remembers, all at once, to look downward — and the road downward has already been cut into seven segments, each one managed by a different party.

In the end, the machine was stopped only because the state stepped in. And all the state could actually do was announce one sentence: these shares — I will honor them.

To honor a claim is the oldest move on this entire road. In Essay 2, the debts written on clay tablets had to be honored through the hierarchy of the temple. In Essay 6, the paper currency called jiaozi had to be honored by the government. In Essay 17, the dollar depended on whether a group of people remained willing to go on honoring it. After all these layers, after all these links cut apart, after all these standard contracts written up, someone still, in the end, has to step forward and say: I honor this.

A ledger can cut the world into extremely fine pieces — fine enough that every sliver carries a price, every sliver's price can be calculated, every sliver can be bought and sold on its own.

What it cannot cut apart is whatever lies between the slivers. On calm days, those things look as though they do not exist at all, because they carry no price and occupy no column. Only when every sliver falls at once do people discover that they had been there the whole time.

The ledger has not yet balanced. It is still being kept.