In 2008 and the years just after it, a great many people learned a sentence they had never expected to say, which was that the house was worth less than the loan against it. Others learned a different one, that eleven years at a firm can end in a meeting that takes four minutes. The time before had been calm. Borrowing was cheap, houses had gone up for as long as most people had been paying attention, the arrangement had the endorsement of banks and regulators and the people on television, and the prudent thing to do, the thing the neighbors were doing, was to borrow and buy. Then it stopped, and the same arrangement that had looked prudent looked overnight like something everyone should have seen coming. The question that follows is rarely asked as an accusation, though it sounds like one. An economy that has lived through a dozen of these keeps assembling the next one. Why?
The answer this chapter follows runs through credit, and it starts from an observation that is easier to state than to accept: the calm and the crash are the same event, seen at two points in its life. What happens in the good years is not the absence of the crisis. It is the construction of it, carried out by people who are each behaving reasonably, in an environment that rewards them for it, using a mechanism that is the ordinary business of a banking system rather than a fault in it. To see that, it helps to start with the one thing that has been counted for long enough to show the pattern.
What a century and a half of credit looks like
Economic historians have assembled, from national statistical records, a long-run database of the advanced economies: the Jordà–Schularick–Taylor Macrohistory Database, which carries annual banking and output figures for 18 countries and dates the systemic financial crises each one suffered. The United States is the country whose lending series runs longest without a gap, from 1880 to 2020, and Figure 5.1 draws it: total bank loans to the domestic private non-financial sector, households and firms together, measured against the size of the economy, with the crises the database records marked on the same axis.
Two things in that picture are worth separating, and the first needs stating carefully, because the version of this figure that gets waved around claims more than the line supports. The marked crises are not each perched on a peak. Two of them are: the crisis coded at 1930 comes at the end of a climb that ran for most of fifty years, and the one coded at 2007 at the end of a climb that ran for sixty. The other three sit part-way up. 1893 and 1907 fall inside the first long expansion, and 1984 falls inside the second, with American bank lending going on rising relative to output for another four years after it. What the line supports is narrower than “every crisis tops a boom.” It is that the crises cluster along the two great expansions and not on the flat and falling stretches between them. The second thing is that this is a picture of one country, and a picture is not a finding. The finding belongs to the economists who built the database. Testing credit growth against later crises across 14 countries from 1870 to 2008, Moritz Schularick and Alan Taylor concluded that “credit growth is a powerful predictor of financial crises, suggesting that policymakers ignore credit at their peril.” That is a statement about a broad statistical tendency across a panel of countries, and it is theirs rather than this chapter’s. It is also, on its own, only a correlation with a direction. What it does not say is why the borrowing keeps building in the first place.
Why the calm does the building
The account that has done most of the work here belongs to Hyman Minsky, an American economist who spent decades arguing that a capitalist financial system does not need a shock from outside to produce a crisis, because it manufactures the conditions for one internally, and does so fastest when things are going well. His name for it was the financial instability hypothesis. Its claim is that the safety of the good years is the raw material of the bad ones.
Minsky’s move was to sort borrowers by a single question: can the cash coming in cover what has been promised? He identified, in his words, “three distinct income-debt relations for economic units, which are labeled as hedge, speculative, and Ponzi finance.” A hedge unit is one that “can fulfill all of their contractual payment obligations by their cash flows”: the rent covers the mortgage, the earnings cover the interest and the principal, and nothing has to go right for the arrangement to hold. A speculative unit can pay its interest out of income but cannot retire the principal that way, so it must, in his phrase, “roll over” its liabilities, borrowing afresh to meet debts as they mature. A Ponzi unit’s income covers neither the principal nor the interest, so it survives only by selling assets or borrowing more, which works for exactly as long as the assets keep rising and the lending stays open. Figure 5.2 sets the three out with the movement between them.
The engine in Minsky’s account is not appetite. It is experience. In a decade when nothing has gone wrong, the borrower who kept a cushion and the lender who asked for a deposit both find that they earned less than the ones who did not, and neither is punished for noticing. Every year the calm lasts is another year of evidence that the cautious were being paid to be wrong. Lending standards loosen without anyone abandoning prudence, because a run of good outcomes has quietly redefined what prudence means, and the loan that would have looked reckless in 1997 looks ordinary in 2005 because the thing it was afraid of has not happened in eight years. Minsky put the conclusion as a theorem: over a protracted period of good times, capitalist economies tend to move from a financial structure dominated by hedge finance units to one carrying a large weight of speculative and Ponzi finance. The structure gets more fragile as a direct consequence of its own tranquillity, and it does so through decisions that are individually defensible at every step.
What makes the resulting structure dangerous is that it has no slack. A system of hedge borrowers can absorb a bad year, because nothing depends on refinancing. A system with a large weight of speculative and Ponzi borrowers depends on the lending window staying open and on asset prices continuing to rise, and both of those can stop at once. When they do, the units that need to roll over cannot, so they sell what they own to raise cash, and because a great many of them are selling the same assets in the same week, the sale itself destroys the value they were selling into. That is the crash, and the reason it feels like a sudden derangement of a working machine is that the machine has been quietly reconfigured, over years, into one that only works while the music plays.
They’ve figured out how to prevent crashes now.
Oversimplified Moderate confidence
There is something in this claim, which is why the ruling is not that its reverse holds. A century of institutional building was followed by a changed record. Deposit insurance ended the sight of depositors queuing outside a failing bank as a routine event; a central bank willing to lend against good collateral in a panic removes one of the mechanisms by which a solvent bank used to die; and on the long-run record the United States goes from three systemic crises coded between 1873 and 1907 to one, coded at 1984, in the 50 years after 1945. Someone who answers that nothing was learned and nothing changed is contradicted by that record, and that is what keeps this ruling off backwards. What the claim gets wrong is the word “prevent.” The same series that shows the long quiet also shows what was accumulating during it: bank lending to households and firms climbed from 12 percent of American output in 1945 to 62 percent by 2007, a rise of about 50 percentage points of output over 62 years, against about 29 points over the 49 years into 1929. It ended in the crisis the database marks at 2007. Whether the quiet itself helped build that is the argument this chapter reports rather than settles: it is Minsky’s hypothesis, and it was pressed against this particular calm in 2005, before the crisis, by Raghuram Rajan, who granted that volatility had fallen and warned that “we should not be lulled into complacency by a long period of calm.” Stabilization is real and it is documented. A cycle that has been made less frequent has not been retired, and the record the claim rests on is also the record that refutes it.
Sources
- The crises kept arriving, and the build-up before the last one was the largest before any crisis in the series: Òscar Jordà, Moritz Schularick and Alan M. Taylor, Macrohistory Database, Release 6 — United States bank loans to the domestic private non-financial sector as a share of GDP, 1880–2020 (0.189 in 1880, 0.476 in 1929, 0.124 in 1945, 0.620 in 2007); systemic crises coded for the United States at 1873, 1893, 1907, 1930, 1984 and 2007. The construct is bank lending, not total debt; the series ends 2020, so it says nothing about credit since.
- Credit growth as a predictor, stated as its authors state it: Moritz Schularick and Alan M. Taylor, “Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870–2008,” American Economic Review 102(2), 2012 — “Credit growth is a powerful predictor of financial crises, suggesting that policymakers ignore credit at their peril.” That paper’s panel is 14 countries to 2008, a predecessor of the 18-country database above rather than the same file; neither one’s figures are attributed to the other.
- The mechanism, attributed to the people who argue it: Hyman P. Minsky, “The Financial Instability Hypothesis,” Levy Economics Institute Working Paper No. 74, 1992 — over a protracted period of good times an economy tends to move from a structure dominated by hedge finance toward one weighted to speculative and Ponzi finance. Raghuram G. Rajan, “Has Financial Development Made the World Riskier?”, NBER Working Paper No. 11728, 2005 — written while the calm was still running, and delivered at the Federal Reserve Bank of Kansas City’s 2005 Jackson Hole symposium, “The Greenspan Era: Lessons for the Future”: “we should not be lulled into complacency by a long period of calm. The absence of volatility does not imply the absence of risk, especially when the risk is tail risk, which may take a long time to show up.” The housing side of 2008 is treated at Countries, Chapter 12.
- Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Evidence directness is strong: the lending and output figures are counted from national records and the crisis years are the database’s own coding, so the claim that crises kept arriving is read straight off the series. The binding, weaker leg is construct match. “Crashes,” as the claim uses the word, means the thing that took the job or the house, while the series codes systemic banking crises in advanced economies, which is neither a stock-market crash nor a recession; the two overlap in 2007 and do not always. The claim is also about a capability, what policymakers can now prevent, while the evidence is an outcome record, which can show that crises happened without establishing what would have happened under other policy. The direction is not in doubt, since a marked crisis in 2007 is incompatible with prevention, and that keeps the ruling firm and off low; the distance between the claim’s subject and the series’ construct is what keeps it off high.
What the century of stabilization did, and did not, do
The picture that comes out of the two figures together is a machine that was genuinely repaired and genuinely not solved. The repairs are visible in the long gap in the middle of Figure 5.1, between the crisis marked at 1930 and the one marked at 1984, half a century in which the United States records no systemic banking crisis on this coding, and they were real repairs: insured deposits, a lender of last resort, supervision of the institutions that had failed most spectacularly. What the same figure shows underneath the calm is the second climb getting under way, slowly at first and then steeply, until by 2007 the American private sector owed its banks a larger share of national output than at any point since the series began.
Whether the calm caused the climb is the live question, and it is not this chapter’s to settle. The argument that it did is Minsky’s, made decades before the event, and it was applied to this specific stretch of quiet while the quiet was still running. In 2005, at the Federal Reserve Bank of Kansas City’s Jackson Hole symposium, convened that year under the title “The Greenspan Era: Lessons for the Future,” Raghuram Rajan granted the achievement in as many words, that the volatility of growth in the industrial countries had been falling, partly through better policies, and then said what he thought followed from it: that the nature of tail risks, especially those related to credit, is such that we should not be lulled into complacency by a long period of calm, and that the absence of volatility does not imply the absence of risk. Three years later the crisis arrived. Economists have argued about that sequence ever since, and a reader who wants the site’s own answer to whether it settles anything will not get one here, because the case for the connection is a case about a counterfactual and there is only one run of the experiment.
What the record does support is narrower and still worth having. Credit expansions of the kind that preceded 1929 and 2007 are visible in advance, in a series that any central bank can read; Schularick and Taylor’s panel finds credit growth to be, in their words, a powerful predictor of financial crises; and they build during the years when nothing appears to be wrong, which is the complacency Rajan was warning against. The boom is not the reward that precedes the punishment. It is the same process at an earlier stage, and the reason it keeps happening is that at no point in it does anyone have to behave badly.
The cycle in this chapter is a cycle in the price of credit and the value of assets, and it ends, however painfully, with the currency in your pocket still meaning what it meant. That is not the only way a financial order can fail. When the state itself runs out of ways to pay and starts settling its bills with newly printed money, the failure moves from the credit system into the unit of account, and the losses land in a different place and in a known order. That is the subject of the next chapter, which turns from the money that gets lent to the money itself.