1. Fifteen Months

By the autumn of 1930, the American economy looked ready to turn a corner.

That was not a hindsight illusion, some rosy story later generations told themselves — it was a judgment the data of the moment actually supported. The contractions before this one had lasted fifteen months on average, and counting from the peak in August 1929, this one had, by that autumn, reached exactly fifteen months too. The curve was due to bend.

This detail is worth holding onto, because it shaped the form every argument that followed would take. Had that line actually turned upward in the autumn of 1930, today's economics textbooks would read very differently, and Keynes would likely be remembered as a British official who had written a few clever pamphlets. A contraction lasting fifteen months already counted as long by the standards of the day, and at the time nobody had any instrument that could tell them whether what they were living through was the same kind of thing as what had come before, or something else entirely.

In November, the bank panics began.

First a financial group in Nashville failed, taking its affiliated institutions down with it. On December 11, the fourth-largest bank in New York closed its doors. The runs spread from one town to the next. Within a few weeks, hundreds of banks had shut down.

These sentences read quickly. On the ground, every bank that closed meant a town's worth of passbooks turned to waste paper on the same afternoon, meant a round of already-negotiated loans that would never be disbursed, meant a payroll due the following month that would not be met. A bank is not a facility for storing money; it is the junction point for every payment relationship in a place. When the junction breaks, what breaks is not the savings — it is whether the people of that place can still get anything done with one another.

The line that should have turned upward instead went down for three and a half years.

By the business-cycle chronology drawn up later, the trough of this contraction fell in March 1933. From 1929 to 1933, real output in the United States fell by roughly twenty-nine percent, consumer prices fell by roughly a quarter, and unemployment reached roughly a quarter of the labor force in 1933.

These three figures need to be read together. Output fell by nearly thirty percent, prices fell by nearly a quarter, and joblessness reached a quarter. By the classical account, once prices come down, things should become affordable again, business should become viable again, people should be hired back. Prices did fall, and fell thoroughly. People did not go back.

And so-called recovery does not mean people get reabsorbed.

That sentence is the key to everything that follows. A curve can climb back to its old height while the people standing beneath it do not automatically climb back with it. Output is a quantity that can be restored. The eight years a person lost are not.

The 1933 unemployment rate stood at about 24.9 percent, with roughly 12.83 million people out of work. By 1935, about a fifth of the labor force still had no job. By 1937, industrial production had already surpassed its 1929 level, and 7.7 million people were still unemployed. That same year policy turned toward tightening, and 1937 to 1938 brought another recession, pushing unemployment back up toward twenty percent.

Among these figures, output, prices, and the money supply could all be restored. People could not.

This is not a lament — it carries a very technical meaning. Output falls and then climbs back, and on the books that is one entry out and one entry in, netting to zero; but a forty-year-old machinist who returns to the shop floor after four years of unemployment brings with him the skill he lost, the seniority that was broken, the savings that were spent, and everything that happened to his family across those four years — and none of that gets written off simply because output has returned to where it stood before. The restoration of the aggregate and the restoration of the individual are the same event on the ledger. In the world, they are not.

Across the fourteen essays before this one, the construct kept running into things it could not fit inside itself, and the remedies changed seven times over: loosened from outside by a larger construct, carried on under a new name, measured to the decimal without the question ever being asked, rewritten into a sum that could be amortized into cost, its discipline's boundary pulled in by one notch, given an orbit that circled the ledger, and required to produce proof.

This time was different.

This time what would not fit was people — several million of them, standing in the street, lined up outside relief stations, living in shantytowns on the edge of town, visible to anyone who looked. The construct had nowhere left to move them.

2. You Chose This

To see clearly what Keynes broke, you first have to see clearly how the thing he broke had described itself.

The classical tradition was never a single monolith, but it broadly shared one core conviction: a general deficiency of demand did not, in theory, amount to a fundamental problem. The very process of production generates income sufficient to buy back what has been produced; prices adjust, wages adjust, interest rates adjust, and resources eventually get reallocated to wherever they can be sold and wherever labor can be hired. The doctrine left behind one endlessly quoted slogan: supply creates its own demand.

Mill left behind an even harder line, one Keynes would later take dead aim at: demand for commodities is not demand for labour.

In its own context, the sentence is not absurd. Mill's point was that when you buy a finished coat, the money you hand over pays for existing stock, and it is the capital committed to the process of production that actually hires the labor. The distinction has its logic, and it remains useful for certain questions even now. The trouble lies in how it got generalized: once read as buying things cannot create jobs, the entire line of inquiry into the relationship between demand and employment was shut off at the starting gate.

Within this framework, unemployment on a large and persistent scale is, in theory, impossible.

So what, then, were the more than ten million people standing in the streets in 1932?

This is not a rhetorical question. It is a classification problem, and classification problems have come up before in this series more than once. Essay 8 made the point that classification precedes reasoning: which column a thing gets placed in decides every inference that follows. Which column the people in the street got placed in would decide what government was supposed to do, what the discipline was supposed to study, and how those people were supposed to see themselves.

The classical side had a ready answer, and a fully worked-out one at that. Set aside frictional, temporary mismatches, set aside union restrictions, and a person is out of work, in the end, because he will not accept pay commensurate with his own marginal productivity. In other words: he is asking too much.

What appears here is the construct's eighth method for handling the remainder, and it is the least effortful of them all.

It does not deny that these people exist. It does not say they went unmeasured, and it does not sort them outside the gate. What it does is reclassify: it files the remainder under the remainder's own choice. You are not out of work because the books fail to balance. You are out of work because you chose not to work.

This move is harder to counter than any of the previous seven, for two reasons. First, it is logically airtight — there is no way to refute it from inside, because the possibility of some slightly lower wage always exists, and that possibility can always be invoked to explain why no deal has yet been struck. Second, it shifts the entire burden of proof onto the remainder itself: to prove that he did not choose this, the man without a job has to be the one to prove it. And a man without a job has no instrument in hand with which to do so.

Keynes's counterattack against this move was fierce, and it was methodological.

He began by conceding a fair point: classical theory is logically self-consistent within its own premises, and classical economists were not fools. The problem is that the theory applies only to one special case, and applying it directly to an age of mass unemployment is a category error, a mismatch of method.

Then he reached for a geometric analogy. When Euclidean geometry fails inside a non-Euclidean space, you cannot turn around and blame the straight lines for not being straight enough.

This analogy is worth remembering, because it names one of the construct's standing habits. When the model and the object stop matching, the model can be revised, or the object can be declared disqualified. Revising the model means rebuilding an entire chain of derivation. Disqualifying the object takes only one adjective.

And this particular adjective is a very respectable one. Voluntary sounds like it is honoring a person's choice, even carries a faint scent of freedom about it. Essay 13's produce your proof worked the same way — it too sounded like nothing more than procedural rigor. At this point the construct never needs to say anything harsh. It only needs one word that sounds reasonable enough to rewrite a matter from unresolved to resolved.

And the person rewritten this way ends up in a strange position. He goes out every day looking for work and comes home with nothing, and to himself this is a plain, unambiguous failure; but on the books, it gets entered as a choice he made. There is no channel of appeal between the two ways of recording it, because the ledger does not accept testimony from the person concerned — the ledger accepts only variables that fit into the model.

3. When Both Sides Want More

Keynes did not go and argue with the classical economists over whether those people actually wanted to work.

He did something else: he built a definition, one the classical model had no slot for.

The move itself is worth pausing on. The usual way to refute a theory is to find the point where its reasoning goes wrong. Keynes picked a different target: he went looking for the box missing from that theory's own table of classifications.

In The General Theory of Employment, Interest and Money, he defined involuntary unemployment this way. Take the current level of money wages. Men are involuntarily unemployed, he wrote, if, in the event of a small rise in the price of wage-goods relative to the money wage — that is, a small fall in the real wage — both the aggregate supply of labor willing to work at the current money wage and the aggregate demand for labor at that wage would be greater than the existing volume of employment. Those currently without work are, on this definition, involuntarily unemployed.

The definition is written in a roundabout way, and its weight is not obvious on a first pass. The whole of it turns on five words: both sides increase.

On the classical account, if what was truly jamming the gears was workers demanding too much, then a slight fall in the real wage should mean fewer people willing to work and more employers willing to hire — the two moving in opposite directions until they meet, and the market clears.

What Keynes wants is the scenario in which both sides move upward at the same time.

More people are willing to work, more employers are willing to hire, and these very people currently have no job. That tells you the jam is not located between the two of them at all. It is not that the price could not be agreed upon — it is that things never got as far as a negotiation.

Put it in more concrete terms. A factory in a town shuts down, and a thousand people go out looking for work. A new owner shows up and says he can pay twenty percent less. On the classical picture, some of the thousand should refuse the low offer while others accept it, and the factory should reopen with a smaller, cheaper workforce. What actually tends to happen is this: all thousand of them are willing to come, and the owner has no intention of opening the factory at all, because he does not know who he would sell the output to. The willing workers line up around the corner; the willing employer is nowhere in the room. Pressing wages down further only presses on the thousand people standing in line — it does nothing to summon the absent employer into existence.

This is a testable condition, not a complaint. It takes something observable and writes it into a definition, and once that condition holds, the classical explanation simply stops working.

This step deserves a note of its own.

In the essays before this one, the remainder was something pointed to, described, tallied up figure by figure by historians. It was real, but its reality depended on being narrated. Here, for the first time, someone wrote the remainder as a criterion: meet these conditions and it exists, fail to meet them and it does not. For the first time the remainder had a test of its own.

Giving the remainder a criterion is not the same thing as speaking on its behalf. Speak on its behalf, and someone can always say you have misapplied your sympathy. Once you have a criterion, anyone who disagrees has to show that the condition fails to hold empirically — and that is a matter to be settled with data. Keynes moved an argument about the human heart onto a table where things could be tested. That was not softheartedness. That was knowing which table you had to argue at for the argument to count.

What flipped next was the direction of causation.

On the classical side, the order runs: the real wage determines the volume of employment. Wherever the wage is set, that is where hiring stops.

Keynes's order reverses it: given the propensity to consume and the inducement to invest, the volume of employment is determined by effective demand, and the real wage is merely the outcome that corresponds to that volume of employment.

Translated into everyday terms: it is not that a wage capable of clearing the labor market comes first and hiring follows from it; it is that a firm's expectation of how much it can sell comes first, and only then does it decide how many people to hire.

The wage is not the entry point. Sales expectations are the entry point. And no matter how far down a man prices himself, there is nothing he can do, on his own, to conjure someone else's demand for what he makes.

There is another layer folded into this reversal. A firm's expectation of sales is a judgment about whether other people are going to spend money, and whether other people spend money depends on whether those other people have jobs. So every firm ends up waiting for every other firm to move first. Nobody has done anything wrong; everybody is making the most reasonable judgment available given his own position; and the sum of all those reasonable judgments is an outcome that nobody wanted. The place Essay 13 described — where all the reasonable dealings get added up and the party who should answer for the result simply disappears — turns up again here, in a different setting.

4. The Exception That Ate the Rule

The General Theory is not a pamphlet arguing that governments should spend more. It is closer to a machine that has had its causal order rebuilt, and the most important part of that machine sits in chapter three.

Keynes begins by separating two things: the aggregate supply price and the aggregate demand price. The first is the proceeds a firm must recover to make a given level of employment worth its while; the second is the proceeds a firm expects to receive at that level of employment. The firm's decision is simply a comparison of the two figures: if hiring one more batch of workers means expected revenue exceeds the required proceeds, the firm expands; otherwise it contracts.

Employment, then, is no longer determined by the labor market alone — it depends on whether firms, facing their sales prospects, judge it worthwhile to hire. The key term at this point is effective demand.

The word effective is doing specialized work here, not decoration. It refers to the kind of demand that actually turns into an order, not the kind of want that has no money behind it. Essay 5 carried a related line: a want with no purchasing power behind it will not throw the books out of balance. There, that line was the ledger-keeper's own excuse. Here, the same fact becomes the central working part of the whole machine. A crowd can be starving and the books can still balance — this is why.

Next comes the character of consumption.

Keynes does not mean that consumption is unimportant, only that it has one systematic feature: when income rises, consumption rises too, but typically by less than income does. So as employment expands, a gap opens up between total output and consumer spending. That gap has to be filled by investment. If investment falls short, firms will not go on offering higher levels of employment.

Pushed this far, the single most important sentence in the whole argument arrives.

Given the propensity to consume, the equilibrium level of employment depends on the current volume of investment; and the effective demand that corresponds to full employment turns out to be nothing more than a special case.

That sentence should be read slowly.

Full employment is not the natural resting state of a market economy. It is the accidental result that appears only when one particular set of parameters happens to line up.

At this point, for the first time, someone working from inside the construct, using the construct's own language, wrote non-closure as the default state. In the fourteen essays before this one, the construct always treated whatever would not fit as an exception, as error, as a piece not yet fully processed. Here the relationship is inverted entirely: fitting is the exception.

It is worth noting exactly where this inversion was carried out. It was not carried out by a critic standing outside, not by a victim, not by a revolution. It was carried out by a man who taught at Cambridge, who advised the Treasury, who defended the Bank of England and then opposed it, using this discipline's own methods of derivation. The hardest challenge a construct faces often comes from one of the very people who understand it best. Essay 9's Smith stood inside the construct; Essay 11's Marx stood outside it. Keynes is inside — and stays inside.

One thing has to be said at the same time, or this section will be read as a verdict.

This conclusion is not a philosophical proposition. It is a technical result derived from a set of assumptions about the propensity to consume and the inducement to invest, assumptions that have been disputed from the day they were written down right up to today. Keynes himself never said the market would never clear — what he said was that clearing is not automatic, is not something you get to assume by default.

Put plainly, his breakthrough compresses into one sentence: not every person willing to work can sell himself by cutting his price.

Because labor is not potatoes, and a wage is not an isolated price tag stuck onto goods. Cutting wages changes prices, changes the real weight of debt, changes expectations, changes consumption, changes firms' cash flow, changes banks' balance sheets — and loops back around to change whether a firm judges that hiring this particular person can even be sold at a profit.

So the labor market cannot be calculated in isolation. It has to be put back inside the whole monetary economy.

That sentence lands as a heavy blow against the construct, because the construct's most habitual technique is exactly partition: cut a complicated thing into pieces, price each piece on its own, settle each piece on its own, and add them up into a grand total. Partition is the precondition that makes commensuration possible at all. What Keynes is saying is that, in this particular case, the result you calculate for a piece cut off on its own and that same piece's actual behavior within the whole are not the same thing — sometimes they even point in opposite directions.

5. What the Banks Took Down With Them

The panic that began in November 1930 is worth examining for its mechanics, because it was not simply a matter of bad banks being weeded out.

At the time, the United States had more than eight thousand banks belonging to the Federal Reserve System, and close to sixteen thousand more that did not. That second, much larger group depended heavily on networks of correspondent banks to hold their reserves for them.

Money in transit through the check-clearing process would be entered on the books of two banks at once, counted as reserves by both. This was the so-called fictitious reserve. In ordinary times it worked, because not everyone demanded cash at the same moment; in a run, it fell short.

So country banks asked their correspondent banks for cash, and those correspondent banks were themselves being run on, and the whole chain — cash, notes, deposits, and the conversions between them — seized up all at once.

This structure is worth pausing over. Fictitious reserves were not the product of anyone's fraud; they were the natural consequence of the fact that clearing takes time. As long as the money is still in transit, it is not, strictly speaking, in either place at once, and bookkeeping requires that it be counted somewhere. The system's solution was to let it count in both places. In ordinary times that was efficiency. In a run, it was a hole. The construct allowed a small duplication so the books would balance every single day, and that small duplication turned out to be empty on precisely the day it most needed to be real.

Up to this point, the losses can still be measured: how many banks closed, how much the money stock fell, how much was lost in deposits.

What genuinely resists measurement is something else.

What bank failures erased was not only the figures on the books but an entire body of local credit knowledge — which merchants in this town could be trusted, which farmers would be able to repay this year, which warehouse receipts were good collateral, which businesses that looked prosperous on the surface were actually rolling old debt into new. None of these judgments was written on any price list; they were stored in relationships built up over years of dealing between people, and the vessel that carried those relationships was the bank.

When the bank closed, that body of knowledge disappeared along with it.

A researcher later put this into more technical language: bank failures reduced the efficiency of the entire process of credit allocation, raised the cost of credit, compressed the availability of credit, and in doing so did further damage to aggregate demand.

This is another shape the remainder takes, and it differs from every one in the essays before this.

That body of knowledge never sat on any balance sheet to begin with, so its disappearance shows up on no balance sheet either. A decline in the money supply can be tallied month by month; a town suddenly not knowing whom to lend to has no unit that can measure it. On the books it registers as a number. In reality it registers as an entire street of people who no longer dare speak frankly to one another.

Push the point one step further. That knowledge was never on the books not because nobody wanted to record it, but because it could not be recorded. Whether someone can be trusted is not an attribute you can fill into a box — it is a judgment built up from countless small events over many years, and it is always shifting. To write it down, you would first have to convert it into a handful of scoreable indicators, and the moment it becomes an indicator, it stops being the thing it was.

Essay 14 concluded that the more a system builds itself upon impersonal footing, the more it depends on trust. Here is the other half of the same fact: that trust is not merely an attitude. It is, at the same time, a body of specific knowledge, living inside specific people, and it can be physically destroyed.

On March 12, 1933, Roosevelt spoke to the country over the radio about the balance sheets of its banks.

He explained that a bank does not keep everyone's money sitting in a safe; it puts that money to work in bonds, commercial paper, and mortgage loans. When a panic arrives, even a bank whose assets are entirely sound has no way to turn those assets into cash overnight, short of dumping them at panic prices far below their real worth.

Then he said two lines that would be remembered for a long time. The first: it is safer to keep your money in a reopened bank than under the mattress. The second: there is an element in the readjustment of the financial system more important than currency, more important than gold, and that is the confidence of the people themselves.

The Emergency Banking Act was signed on March 9. Starting March 13, sound banks in Federal Reserve cities reopened first, with the reopening extended gradually to other cities and towns after that.

By the end of March, the public had already redeposited roughly two-thirds of the cash it had withdrawn in the four weeks before March 8.

A head of state explained the asset structure of banks to the nation over the radio, and then asked people to send their money back. And it worked.

6. From $20.67 to $35

For a long stretch, the United States maintained a statutory gold price: twenty dollars and sixty-seven cents per ounce.

On April 5, 1933, an executive order banned the private hoarding of gold coin, gold bullion, and gold certificates, and required individuals and institutions to hand them over.

In 1934, the price of gold was reset at thirty-five dollars an ounce. The gold content of the dollar was thereby compressed to about fifty-nine percent of what it had been.

That fifty-nine percent is simple arithmetic: an ounce of gold once worth $20.67 was now worth $35, which means the same dollar now bought more than forty percent less gold than before.

Lay the sequence out and it says enough on its own: first the government collected private gold at the old price, and only then did it change the price.

No judgment is being passed here. There is only one thing to establish: this entire sequence was fully legal. The executive order had statutory authority behind it, the surrender came with compensation, and the repricing was backed by legislation. Essay 13's line can be lifted directly into service here, and this is its fourth appearance: the whole process can be entirely legal, and the result can still be an extremely asymmetric transfer.

Essay 14 observed that the construct is most successful exactly when it makes people forget it was ever made. Here is the reverse of that observation, and reversed with total cleanness.

A scale treated as natural, objective, above politics, was rewritten by a single decree. And who rewrote it? The very same people, in the very same administration, of the very same country. In the sixty years before this moment, the entire authority of that scale had rested precisely on the fact that it could not be casually altered by anyone at all.

This point is worth restating, because it is the source of everything this section carries. The gold standard did not persuade people because it was more convenient, or because it was more precise. It persuaded people because it said: nobody decided this, this is simply the weight of gold, and gold sitting in the ground answers to no government. Precisely because of that, it let a borrower believe that what he paid back later would be the same thing he had borrowed; it let two trading houses on opposite sides of an ocean strike a deal on a piece of paper. Its entire credibility was staked on five words: not alterable by anyone.

Once the fact that a scale can be rewritten has happened a single time, it can never be taken back. From here on, every claim about a common measure has to face a new question: who set this scale, and when will they change it.

And the questioning will not stop at this one sentence. Everything the essays after this one have to cover — the international arrangement rebuilt after the war, its dissolution decades later by a single announcement, and the attempts that refuse to recognize any official scale at all — is chasing the same question. A construct can build a scale, and a construct can also announce that the scale no longer counts; and once people have watched the second thing happen, the first thing carries a tail behind it forever.

A word, in passing, about where the gold went afterward. It did not disappear. It withdrew to a different position: the vaults of central banks, a store of value in private hands, a haven in times of crisis. It stopped being the scale and became one more thing that gets measured. Something moving from measuring other things to being measured by others has happened more than once already in this series, and each time it was not refuted — it was routed around.

As for the effects of leaving gold, the empirical evidence is fairly consistent, though the interpretation remains contested.

The consistent part: countries that left the gold standard earlier generally recovered faster. After Britain suspended gold convertibility in September 1931, the pound devalued and monetary policy gained room to move; later cross-country studies have also found that leaving gold could lift inflation expectations, lower real interest rates, and help bring on recovery.

The contested part: whether it was leaving gold itself doing the work, or the monetary expansion, the improvement in exports, the shift in real wages, and the change in policy expectations that followed from leaving gold. On these mechanisms scholarship has never settled on a single account.

One point still has to be made clear, or this section will be read as a simple moral. Leaving the gold standard helped not because there was anything wrong with gold itself. It helped because, in those years, the cost of defending parity had come to fall entirely on domestic employment. Giving up parity meant moving part of that cost off the backs of the unemployed. The scale was not at fault. What was at fault was that someone had to keep standing under it, forbidden to move.

7. Who Pushed It Deeper

Why the Depression went so deep and lasted so long is a debate that has never converged, even now. Laying the arguments out side by side is more useful than handing down a verdict.

One main line of argument centers on money. Its core claim is straightforward: the great contraction of 1929 to 1933 was as large as it was chiefly because the money stock fell by more than a third; a mass of commercial banks failed, the money multiplier collapsed, and the Federal Reserve did not perform its function as lender of last resort. Along this line, the decision in October 1931 — after Britain left gold — to raise the discount rate in order to defend America's gold reserves counts as an especially critical policy error: it protected gold abroad while further compressing money and credit at home.

This line of argument has its internal critics too. Some researchers point out that the monetary explanation has long had a gap in it: exactly what mechanism carried a monetary contraction into the real economy was never adequately spelled out for decades, since nominal interest rates in the early 1930s were not on any steady, sharp climb, and high interest rates alone cannot carry the weight of explaining the Depression. What later filled the gap was a chain running through deflationary expectations: monetary contraction produces expectations of deflation, expectations of deflation raise the real interest rate, and the real interest rate suppresses spending and employment.

Another main line puts the emphasis on a collapse in spending. Researchers along this line argue that the onset of the Depression is hard to pin mainly on tight money: after the crash, the New York Federal Reserve Bank at one point bought large quantities of government securities, and both nominal and real interest rates fell noticeably from late 1929 into early 1930. The real problem, on this account, is that the crash and the extreme price volatility that followed it generated enormous uncertainty about future income, and spending on durable goods collapsed first. Along this line, monetary factors certainly mattered, but functioned more as an amplifier that later drove the downturn deeper, not as the initial blow.

The third main line is the gold standard, already covered in Essay 14. It situates the Depression within the interwar international monetary order: the contraction began in the United States and France, then spread worldwide along the channels of exchange-rate commitments, gold flows, and the obligation of central banks to defend their gold parities. This school, too, has disagreements within its own ranks.

A fourth main line is concerned with credit intermediation. It does not deny the importance of the monetary collapse; it only insists that bank failures did further damage on another level — degrading the efficiency of credit allocation — a level that changes in the quantity of money alone cannot capture.

Even the narrower question of why the banks failed has no consensus answer. One school emphasizes that most bank failures reflected deteriorating fundamentals and genuine insolvency rather than baseless panic; another school, after reorganizing the original data, holds that both liquidity crises and insolvency were real sources, and that in the earliest rounds of panic the contagion running through correspondent-bank networks and bank runs was very much in evidence; only once the Depression deepened further and asset prices worsened did insolvency increasingly become the dominant threat. In other words, panic and fundamentals are not an either-or matter — they operated together, interwoven across different periods, different regions, and different institutions.

Wages tell the same kind of story. The traditional textbook account often tells the Depression as a case of wages that would not fall, and so the labor market failed to clear. Later empirical research complicated that story: viewed across countries, nominal wages really did adjust quite slowly relative to prices, real wages rose, and this did depress output; yet fresh estimates for interwar Britain found that pay cuts were in fact not uncommon at the time — in 1931 alone, more than three million workers had their wages cut.

So the real point of contention is not whether wages could fall.

The point of contention is whether the interplay among wages, prices, expectations, debt, and demand would, at the aggregate level, automatically absorb unemployment on its own. The empirical answer the Depression gives is: it did not happen automatically.

Each of these explanations has its own evidence, and no verdict will be rendered here. But laid side by side, they share one thing in common: they are arguing over which chain pulled hardest, and not one of those chains was capable of putting those people back into the ledger.

This point matters more than settling on a winner. Suppose it were proven one day that the monetary chain pulled hardest — the thing to do would be monetary policy. Suppose it were proven that the gold-standard chain pulled hardest — the thing to do would be exchange-rate arrangements. And whichever chain is proven to have pulled hardest, none of it changes the fact of that decade: for ten straight years, a great many people were willing to work and had no place to stand, and every mechanism ever used to explain the Depression explains why they were left outside — not one of them explains how to bring them in.

One more thing has to be entered into the record here, because it is the strongest case against fatalism.

In this same crisis, the Federal Reserve Bank of Atlanta discounted fairly freely for its member banks, and even encouraged its member banks to help support non-member banks; the Federal Reserve Bank of St. Louis read its own mandate more narrowly, and refused to rediscount for the purpose of supporting non-member banks.

The aftermath diverged sharply. The Atlanta district's contraction slowed and began to recover. The St. Louis district saw hundreds of bank failures, shrinking loans, and rising unemployment.

Same country, same system, same year, two different styles of judgment, two different outcomes. This was not grand historical necessity. It was a fork produced by different decisions made under the very same set of rules.

8. The Side That Couldn't Find a Price

In the summer of 1932, a body of veterans of the First World War began marching on Washington.

They held a kind of bonus certificate that would not come due until 1945. And they were already too poor to wait thirteen years.

On June 1, about fifteen hundred of them reached the capital. Later arrivals brought the total to somewhere between ten and twenty thousand. They built a shantytown on the edge of the city and settled in to wait for an answer from Congress and the President.

On July 28, the government sent in troops and police to clear the camp. They used tear gas, bayonets, and torches. Five tanks mounted with machine guns appeared in the streets. One veteran died from police gunfire.

There is no elegant diagram of market clearing here. There is only a state physically removing a group of people who could not find a price.

A few things need to be said precisely. They were not a mob; they were men this country had sent off to fight a war and brought back not long before. What they wanted was not something new — it was the cashing of an IOU the government itself had already written. They attacked no one; they built shacks and settled in to wait for an answer. And when the answer turned out to be tear gas, that fact does not describe a matter of public order. It describes something else: when a group of people can find no place in the market and no place in the budget either, there is a third place available to put them, and that place requires no ledger at all.

In those same years there was more scattered movement, and more scattered falling. Many people went to California looking for work, and once they arrived found jobs far scarcer than the advertisements had promised, wages and living conditions far worse, and the big corporate farms nothing at all like the small-farm world they had pictured in their heads.

Labor does not simply flow, in the abstract, from a place where it is cheap to a place where it is dear and get saved by the move. What these people entered was very often just another equally oversupplied market, and they entered it with a weaker bargaining position than the one they had left behind.

Even that twenty-five percent figure has edges of its own. The agency that later explained the number specifically cautioned that it did not include people who would rather have worked full-time but could only find part-time work, nor people whose skills had been forced downward. A machinist reduced to working nights as a watchman counted, in the statistics, as employed.

Between those two positions lie more than a decade of craft, an entire body of judgment usable only on a shop floor, and a man's own sense of what trade he belongs to. None of these differences show up anywhere in the unemployment rate, because that indicator only asks whether, never what. Essay 12 made the point that what the construct measures was never the thing itself but the way the thing had been organized; here the point is even simpler — the construct measures whether a job exists, not where that job leaves the person holding it.

Which is to say: even the single most central indicator still had, ringed around its edges, a whole remainder that never got counted at all.

The road not taken deserves to be recorded too. The roughly one-billion-dollar round of open-market purchases in the spring and summer of 1932 halted the wave of bank failures for a while, but it ended quickly. That January, Congress had created a corporation to extend emergency financing to banks and railroads, but borrowing from it carried a stigma, because the list of borrowers was made public, and the mere act of borrowing was read by the market as a signal of weakness, so many banks preferred to do without. Britain had already left gold in 1931; the United States dragged its feet until 1933. And in 1937, with unemployment nowhere close to absorbed, policy tightened again, bringing on that second recession.

The road toward the later apparatus of macroeconomic management was not a straight line, and it was not a single decisive victory. It was forced into being through round after round of hesitation, conservative retreat, institutional fracture, and political combat.

What forced it into being was not that the theory had persuaded everyone. It was that too many people were already standing on the side of the market where no price could be found.

There is an uncomfortable implication here, and it has to be spoken plainly. What got this framework adopted was not that it was correct — it was that it was correct at a moment when enough people needed it to be. Consider the alternate turn: had that wave of bank panics never happened, had the contraction simply stopped at month fifteen, the same reasoning, set down in the same book, would probably not have found many readers. A construct gets rewritten, as a rule, not because someone has refuted it, but because what leaks outside it has piled up too high to keep calling it an exception.

The classical ledger's arithmetic was, in truth, a simple one: everything has a price, and if the price is right, it will be bought; whatever does not sell is asking too much. This arithmetic holds for potatoes.

Where it fails to hold for people is this: a person can lower his price to nothing and still fail to sell, and the very act of lowering his price can make the people who might buy him even less able to afford him.

Every single person living in that shantytown on the edge of Washington in the summer of 1932 was willing to sell himself at any price at all. There was no place for them on the ledger — and it was not because they had asked too much.

What Keynes did, in the end, comes down to one thing. He took a column that had been recorded as voluntary and re-recorded it as unsettled.

After the rewrite, that ledger looked uglier. It also came closer to being true.

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