Like bubbles, panics and crashes too have a long history. One of the earliest occurred in ancient Rome, which by AD 33 had become an extensively monetized economy.3 At the time, officials of Emperor Tiberius concluded that a boom in private lending had become excessive. It was then accordingly decided that mon- etary tightening, along the lines suggested by the edicts of Julius Caesar, was needed. Such tightening was enforced through regulations involving capital ade- quacy requirements for lenders. Lo and behold, it was soon discovered that most of the senators were in violation of the new regulations! Loans were called in. The property market collapsed. Bankruptcies spread far and wide. And the Emperor finally caved by implementing a massive bailout. This is by now a familiar pattern.
Still, there are nuances. A popular and common definition of a crash is a 20%
or so decline in a few days or at most over only a couple of months.4 The defini- tion used by Ursua and Barro (2009)—cumulative real returns of minus 25%
or worse—is similar.5
Such definitions, however, are not particularly useful unless they also specify the amount of time over which the decline occurs. Not only does a price drop of 50% (in an index, stock, or economy) arithmetically require a 100% gain to recoup the loss, it also requires an often lengthy expanse of time during which there is exposure to risks, new and unknown. This time aspect—in which the amount of time exposure to the market may be as critical to performance as is timing the market—is therefore of greater importance in describing crashes than bubbles. Inflation of bubbles can often extend over many months or even years. Crashes happen fast.6
In most other respects, though, the practical problems in defining and ana- lyzing crashes are in many ways similar to those encountered with bubbles.
Everyone, it seems, knows that a crash has occurred after they’ve experienced one but, as with bubbles, statistical features are often rather arbitrary or absent.7
Even so, Wilmott (2001, p. 394) provides an apt and precise description of what is special about a crash—which is when correlations (p. 340) go to one:
…Obviously a crash is a sudden fall in market prices, too rapid for the liquidation of a portfolio. But a crash isn’t just a rise in volatility. It is characterized by a special relationship between individual assets. During a crash, all assets fall together. There is no such thing as a crash where half the stocks fall and the rest stay put. Technically, this means that all assets become perfectly correlated…A high degree of correlation makes diversification impossible.8
As is illustrated in Fig. 3.1, higher historical volatility is associated with higher returns correlations for both bubbles (left panel) and crashes (right panel).
All of this is just another way of saying that once herding becomes prevalent, diversity of opinion—and also the theoretical notion of complete markets—
dissipates (i.e., diversity approaches a limit of zero). In crashes there is—much more so than in bubbles—also a contemporaneous collective reduction in time horizons for expected returns.
Diversity of horizons will normally tend to keep the markets relatively sta- ble. And trading experiences carried over from the preceding bubble may have deeply implanted “invest-for-the-longer-term” and “buy-on-dips” goals and strategies. But as people are forced to abandon these bull-market prescriptions, liquidity disappears and prices gap wide open to the downside.9 It becomes a time when you are often compelled to sell not what you want to sell but what you still can sell.10
Although these ideas reflect the essence of crashes (and also apply to panics), the relationship of bubbles to crashes (and vice versa) nonetheless remains rather nebulous. It is not at all clear, for instance, that bubbles must end in
crashes or that crashes require bubbles to precede them.11 Germany in 1927, for example, seems to have experienced a crash without a preceding bubble.12
The definitional line between crashes and panics is also a bit blurry as the two terms are often used interchangeably. Nevertheless there is a difference. A panic is a stampede out of a theater on fire. In the case of market crashes, the fire might be the collapse of a major bank, a fund, an insurance company, a brokerage firm, a foreign currency, or of any or all of these combined. In every such situation, mob psychology and flight-for-survival instincts and reflexes govern the action.
Yet each instance differs in its details: The mania and subsequent crash of Japan in the 1990s ensued without a panic because “depositors believed that government would socialize the loan losses.”13 However, the Panic of 1907, also known as the Bankers’ Panic, involved numerous runs on national and local banks and ruin for many companies.14 This panic was of particular impor- tance because of its intensity and speed—NYSE prices fell almost 50% from the peak over a three-week period—and because it eventually led to establishment of the Fed in 1913. But unlike panics of 1873, 1893, 1907, or 1931, the crash of 2008–2009 was for the most part centered not in retail but in wholesale banking that involved commercial paper, “repos”, and derivatives in the
“shadow banking” system.15
It is easy to understand why the term “panic” has been historically applied to sudden fear-driven runs (i.e., withdrawals) on bank deposits that have also often led to system-wide liquidity crises.16 Such crises can be caused by a wide variety of economic ailments and imbalances that might include hyperinfla- tions, currency debasements, external and domestic debt defaults, politically corrupted and/or otherwise compromised banking systems, and, as in 2008–2009, bank runs on other banks.17
0 10 20 30 40 50
0.0 0.2 0.4 0.6 0.8 1.0 UPCORR
HISTVOL3M
0 10 20 30 40 50
0.0 0.2 0.4 0.6 0.8 1.0 DOWNCORR
HISTVOL3M
Fig. 3.1 Historical volatility, based on three-month at-the-money call options for the S&P 500 related to positive and negative return correlations within the 500 stocks in the index, 2004:2017:09
But generally, as Redleaf and Vigilante (2010, pp. 4–5) write in reference to the financial stresses of 2008:
Like most grave financial crises, the mortgage crunch and the crash were crises of information. That’s what panics are…the real problem was not that some of the banks were broke but at the critical moment none of them could prove they weren’t.
It became impossible for either executives or regulators to fully understand the finan- cial condition of any great modern bank.18
Certainly, a sense of panic is elemental to a crash as it motivates hurried emotional rather than thought-based tactical selling. And through feedback effects (i.e., autoregressiveness), a crash can fuel further panic—which is symp- tomatic of a betrayal of trust.19
Yet history has shown that a crash may occur without a panic, even though it is conceivable that a crash might cause a panic or a panic might cause a crash.
One prominent example of a crash without a panic is the experience of Japan in 1990, a year when market prices halved (from around Nikkei 40,000 to 20,000).20
As Roehner (2002, pp. 143–4) explains:
A crash is a sudden price fall in a couple of days or weeks, while a collapse refers to a bear market which lasts at least several months and possibly a couple of years. A crash does not necessarily lead to a collapse nor is a collapse necessarily preceded by a spectacular crash…A crash is a microeconomic phenomenon, which results from a panic among investors; on the contrary a collapse is a macroeconomic phenomenon…
By all such accounts, it thus appears that crashes unfold or crystallize much faster (because of fear) than do bubbles and, as a result, their start and end points ought generally to be much more precisely defined (both visually and statisti- cally) than are those of bubbles.21 Analysis might thus intuitively begin by first comparing crash severities through measurements of peak-to- trough price per- centage declines and number of days expended in moving from high to low (sometimes converted to an average percentage loss per day).22 That there is substantial variation in peak-to-trough data appears in Table 3.1.
Some ambiguity also inevitably remains as to whether an episode can be classified as a crash—as in the brevity of the 1987 (38 days) episode—or as a Roehner-type collapse, as in the lengthy 1973 to 1974 (436 days) experience.
What, for example, was the NASDAQ decline from the 2000 high, which was greatest in terms of total (77.9%) percentage lost? The descent required 647 days for completion, and the average percentage loss per day was actually less than the average for the S&P episodes. On this basis it might well be argued that despite the large total percentage lost, the NASDAQ’s decline from the 2000 high was not a crash but a collapse.23
Yet there is also nothing that precludes a collapse from being punctuated fractal-like by a series of smaller-scale crashes (perhaps as from the alternative
lines of Table 3.1 for May 2008 to November 2008, in which the average losses per day were relatively large). Such definitional difficulties therefore prompt the search for a more objective, formal, and/or robust statistical approach to describing crash characteristics.24
Still—for whatever are their immediate causes—panics, crashes, and col- lapses are universally at their core crises of the dissolution of confidence and trust, in counterparty solvency, in creditors, in governments, and in clients, as much as of in capital.25 Trust and confidence are always what is then in shortest supply; it is a shortage seen time and again from the study of all such extreme events, no matter when they’ve occurred. “A company is only as solvent as the perception of its solvency.”26
Table 3.1 Crash or collapse? Important peak-to-trough moves (>10%), daily closing prices, S&P 500, 1962–2011, and NASDAQ, 1984–2011
S&P 500 Peak Trough Total # trading
days Avg loss per day
Points lost % lost Points %
9-Feb-66 94.10 29-Aug-66 74.53 19.57 20.8 139 0.14 0.15 14-May-69 106.16 26-May-70 69.29 36.87 34.7 260 0.14 0.13 11-Jan-73 120.24 3-Oct-74 62.28 57.96 48.2 436 0.13 0.11 31-Dec-76 107.46 6-Mar-78 86.90 20.56 19.1 296 0.07 0.06 28-Nov-80 140.52 12-Aug-82 102.42 38.10 27.1 430 0.09 0.06 10-Oct-83 172.65 24-Jul-84 147.82 24.83 14.4 199 0.12 0.07 25-Aug-87 336.77 19-Oct-87 224.80 111.97 33.2 38 2.95 0.87 16-Jul-90 368.95 11-Oct-90 295.50 73.45 19.9 62 1.18 0.32 24-Mar-00 1527.46 4-Apr-01a 1103.25 424.21 27.8 251 1.69 0.11 9-Oct-07 1565.15 10-Mar-08 1273.37 291.78 18.6 104 2.81 0.18 29-Apr-11 1363.61 3-Oct-11 1099.23 264.38 19.4 108 2.45 0.18 avgb 25.8 211 1.07 0.21 Alternatives:
24-Mar-00 1527.46 11-Mar-03 800.73 726.73 47.6 741 0.98 0.06 19-May-08 1426.63 20-Nov-08 752.44 674.19 47.3 130 5.19 0.36 19-May-08 1426.63 9-Mar-09 676.53 750.1 52.6 202 3.71 0.26 NASDAQ
27-Aug-87 455.8 7-Dec-87 293.70 162.10 35.6 70 2.32 0.51 17-Jul-90 469.5 17-Oct-90 325.10 144.40 30.8 65 2.22 0.47 27-Apr-94 800.39 27-Jun-94 694.16 106.23 13.3 69 1.54 0.19 21-Jul-98 2018.46 8-Oct-98 1420.94 597.52 29.6 56 10.67 0.53 10-Mar-00 5048.62 9-Oct-02 1114.11 3934.51 77.9 647 6.08 0.12 31-Oct-07 2859.12 17-Mar-08 2177.01 682.11 23.9 94 7.26 0.25 29-Apr-11 2873.54 3-Oct-11 2335.83 537.71 18.7 108 4.98 0.17 avgb 32.8 158 5.01 0.32 Alternatives:
5-Jun-08 2549.94 20-Nov-08 1316.12 1233.82 48.4 118 10.46 0.41 5-Jun-08 2549.94 9-Mar-09 1268.64 1281.3 50.2 190 6.74 0.26
aThe entire decline might also be measured to the post-terrorist attack low of 965.80 on September 21, 2001, which would be 36.8%. Thus the alternative row below
bThis average excludes the last row alternatives Source: Author’s calculations based on Yahoo.com data
Business Cycle Aspects
The downward sides of business cycles are often seen as causes for panics and crashes. But any such cycles are inevitable and unavoidable because that’s the nature of nature in life as well as in economies and markets.27
During the nineteenth and early twentieth centuries, Gorton (1988) found that banking panics were related to business cycles and were not random or
“sunspot” events.28 In the worst of such crises, it was the dissolution of trust that ultimately made financial institutions reluctant to lend “to each other or to anyone else at any price.”29
In reviewing eight centuries of financial crises initiated by a broad array of problems (high inflation rates, currency debasements, etc.), Reinhart and Rogoff (2008, 2009) also showed that serial defaults have been rather regularly experienced in many countries and at many times and for a variety of reasons, not all cycle-related.30 Such crises are seen as being “amplification mecha- nisms,” rather than triggers of recessions. A list of declines of 15% or more in real per capita GDP (Table 3.2) indicates that such occurrences were rather frequent and widespread prior and up to WWII.
As for the United States, the relationship between business cycle downturns and changes in stock prices is shown in Table 3.3. As Siegel (2008, p. 211), explains, “…out of the 46 recessions from 1802, 42 of them, or more than 9 out of 10, have been preceded (or accompanied by) declines of 8 percent or more in the total stock returns index.” And there have been instances (as in 1957, 1980, or 1990) in which the peak of the business cycle virtually coin- cided with that of the market.
A summary quoting the IMF (2003, p. 64) further suggests that “[F]our salient patterns emerge from the comparisons made over many crash episodes:
• Stock market crashes in both countries were frequent (10 in the United Kingdom, 13 in the United States).
• More than half of the crashes in each country were associated with recessions…
• Only about one-third of all crashes were associated with a preceding boom…
• Most of the crashes cum recessions were triggered by monetary policy tightening and also involved banking panics.”31