s t r a t e g y p r a c t i c e

JUNE 2009

‘Power curves’:
What natural and economic
disasters have in common
Parallels between the failures of man-made systems, such as
the economy, and of similarly complex natural ones offer fascinating
food for thought.

Michele Zanini
 1

Executives, strategists, and economic forecasters, somewhat sheepish after missing

the “big one”—last year’s global credit crisis—turned to the lexicon of natural disasters,
describing the shock as a tsunami hitting markets and as an earthquake shaking the world
economy’s foundations. Shopworn as these metaphors may be, they aptly capture the extreme
and unexpected nature of the circumstances. In fact, the parallels between the dynamics and
failures of man-made systems, such as the economy or the electricity grid, and similarly complex
natural ones are bringing new ideas to economic forecasting, strategic planning, and risk
management. This trend may have profound implications for policy makers, economists, and
corporate strategists alike.

Scientists, sometimes in cooperation with economists, are taking the lead in a young field that
applies complexity theory to economic research, rejecting the traditional view of the economy
as a fully transparent, rational system striving toward equilibrium. The geophysics professor
and earthquake authority Didier Sornette, for example, leads the Financial Crisis Observatory,
in Zurich, which uses concepts and mathematical models that draw on complexity theory and
statistical physics to understand financial bubbles and economic crises.

Sornette aims to predict extreme outcomes in complex systems. Many other scientists in the
field of complexity theory argue that earthquakes, forest fires, power blackouts, and the like
are extremely difficult or even impossible to foresee because they are the products of many
interdependent “agents” and cascades of events in inherently unstable systems that generate
large variations. One symptom of such a system’s behavior is that the frequency and magnitude
of outcomes can be described by a mathematical relationship called a “power law,” characterized
by a short “head” of frequently occurring small events, dropping off to a long “tail” of
increasingly rare but much larger ones.

The power law phenomenon, explored in recent bestselling books and observed by academics
for decades, seems to be applicable to a wide range of currently relevant economic outcomes,
including financial crises, industrial production, and corporate bankruptcies. It can even
describe how industry structures evolve.1

If, for instance, you plot the frequency of banking crises around the world from 1970 to 2007, as
well as their magnitude as measured by four-year losses of GDP for each affected country, you
get a typical power curve pattern, with a short head of almost 70 crises, each with accumulated
losses of less than 15 percent of GDP, quickly falling off to a long tail of very few—but massive—
crises (Exhibit 1). While the most extreme cases involve smaller, less developed countries, the
same distribution also applies to more developed ones—and with much larger absolute values
for GDP loss. Earthquakes, forest fires, and blackouts yield a similar power curve pattern—for
instance, from 1993 to 1995, Southern California registered 7,000 tremors at 2.0–2.5 on
the Richter scale, falling off to the 1994 Northridge earthquake, at the end of the tail, with
a magnitude of 6.7. The curve highlights a key property of the power law: extremely large
outcomes are more likely than they are in a normal, bell-shaped distribution, which implies a
relatively even spread of values around a mean (in other words, shorter and thinner tails).

1
See Michele Zanini, “Using ‘power curves’ to assess industry dynamics,” mckinseyquarterly.com, November 2008.
 2

Exhibit 1
The banking-crisis Power curves: banking crises compared with earthquakes

power curve Number of worldwide banking crises, 1970–2007

70
r2 = 0.97
Number of earthquakes in
Includes US Southern California, 1993–1995
savings-and-loan
60 crisis, 1980s 8,000
(4% GDP loss) r2 = 0.94

6,000
50

4,000

40
2,000

30 0
2–2.5 3–3.5 4–4.5 5–5.5 6–6.5 6.5+
Magnitude of earthquake

20
Includes
Sweden,
1991 (31%
Includes
GDP loss)
Includes South Korea,
10 Includes
Japan, 1997 (50%
Thailand,
1990s (18% GDP loss)
1997 (98% Cameroon, Niger,
GDP loss) GDP loss) 1987 1983

0
< 15 15–30 30–45 45–60 60–75 75–90 90–105 105–120 >120
Cumulative loss of GDP,1 %

r2 is the proportion of variance explained by a regression.

1IMF economists computed GDP loss by extrapolating real GDP from pre-crisis trend and adding the yearly percentage
difference between actual GDP and extrapolated GDP for first 4 years of crisis (for more details, see Luc Laeven and
Fabian Valencia, Systemic Banking Crises: A New Database, IMF working paper, November 2008).
Source: International Monetary Fund (IMF); US Geological Survey; McKinsey analysis

Similarly, the power law pattern can be seen in the frequency and magnitude of the monthly
swings, positive and negative, in US industrial production from 1919 to February 2009 (Exhibit
2). Negative swings of up to 4 percent in late 2008 and early 2009 were the largest since the
1940s, though much smaller than some of those from 1920 to 1946, at more than 7 percent (the
largest gyrations occurred at the end of World War II).

The power law dynamics that affect the overall economy and industries can generate staggering
outcomes for individual companies. Plotting the size of the biggest US bankruptcies from 2001 to
2008 by assets shows that the largest, Lehman Brothers, was twice the size of the second largest,
Washington Mutual, which in turn was three times the size of the third largest, WorldCom
(Exhibit 3).
 3

Exhibit 2
The industrial- Number of swings in US industrial production, 1919–Feb 2009
(monthly figures)
production power
800
curve r2 = 0.95

700

600

500

400

300

200
Includes All swings above
Dec 2008– 7% date back to
Jan 2009 1920–46
100 Includes
Sept 2008

0
<1 1–2 2–3 3–4 4–5 5–6 6–7 7–8 8–9 9–10 10–11 11+
Magnitude of month-to-month swing,1 %

r2 is the proportion of variance explained by a regression.

1Absolute value of change (ie, both positive and negative).
Source: US Federal Reserve; McKinsey analysis

These examples indicate that power law patterns, with their small, frequent outcomes mixed
with rare, hard-to-predict extreme ones, exist in many aspects of the economy. This suggests
that the economy, like other complex systems characterized by power law behavior, is inherently
unstable and prone to occasional huge failures. Intriguing stuff, but how can corporate
strategists, economists, and policy makers use it? This is still a young field of research, and
the study of power law patterns may be part of the answer, but it isn’t too early to consider and
discuss potential implications.

Make the system the unit of analysis. You can’t assess the behavior and performance of a
specific agent—for example, a financial-services company—without gauging the behavior and
performance of the system in which it is embedded. Proponents of a systemic financial regulator
that would span multiple subsectors and geographies are making a similar argument.

Don’t assume stability and do take a long look back. Major systemic imbalances and corrections
are highly likely, and everyone should be wary of new economic paradigms to the contrary. It’s
equally important to take a truly historical perspective and consider a system’s underlying
patterns. If you look at the sharp rise in US corporate profits from 1997 to 2007 in isolation, it
might seem like steady, sustainable development that can be justified by pointing to near-term
 4

Exhibit 3
The corporate- Size of 15 largest US public-company bankruptcy filings
by assets, 2001–08, $ billion (in real 2000 dollars)
bankruptcy power
600
curve Lehman Brothers r2 = 0.96
(2008)

500

400

300
Washington Mutual
(2008)

200

WorldCom
(2002)
100
IndyMac New Century
(2008) Financial (2007)

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Rank by size of assets

r2 is the proportion of variance explained by a regression.

Source: Bankruptcydata.com; McKinsey analysis

trends, such as globalization and productivity growth. Yet it becomes a striking departure from
the historical norm when you look back and find that profits last hit such a lofty percentage of
GDP more than 50 years ago and dropped shortly thereafter. Outliers such as these should not
be ignored but rather studied closely for clues that might help us understand current and future
events.2

Focus on early warning. The inherent uncertainty of complex systems makes point predictions
unreliable. Much as earthquake scientists are developing tsunami early-warning systems,
corporate strategists should monitor potential indications that economic stress might be
building in their industries. One indicator could be changes in the exit and entry rates in
a particular industry. It’s notable, for instance, that specialized US mortgage companies
experienced difficulties in late 2006 and that several went bankrupt long before the problems
spread to financial institutions with a strong mortgage exposure and then to broader financial
institutions and other major companies.

2
 CLA management professor Bill McKelvey has advocated making the study of extreme events in business and economics a core focus
U
of inquiry, much as seismologists study large earthquakes to understand geological systems.
 5

Build flexible business models. Corporate leaders might consider robust business models
Related articles
incorporating some slack and flexibility instead of the models most common today, which
“Using ‘power curves’ aim to optimize value in the most likely future scenario and thus leave companies exposed
to assess industry when conditions change dramatically. The offshoring of production or services to a number
dynamics” of continents and countries, for example, is surely more costly, under stable conditions, than
maintaining a single center but would protect a company in the event of an unexpected regional
“Taking improbable or national economic crisis.
events seriously: An
interview with the
Learn from scientists studying other complex systems. Strategists, economists, and others
author of The Black
Swan” should consider several other potential parallels. To take one example, what economic-policy
lessons could be drawn from the observation that efforts to put out small forest fires quickly
“A fresh look at strategy may in time lead to large-scale fires, because the rapid mitigation of small ones allows burnable
under uncertainty: An
interview”
undergrowth to accumulate? Q
Michele Zanini is an associate principal in McKinsey’s Boston office. Copyright © 2009 McKinsey & Company.
All rights reserved.