Page 2 of 3 Soaring, cryptography and nuclear weapons
By Martin Hellman
In spite of the similarity between the Cuban and Turkish missiles, Khrushchev
realized that America would find this deployment unacceptable and therefore did
so secretly, disguising the missiles and expecting to confront the US with a
fait accompli. Once the missiles were operational, America could not attack
them or Cuba without inviting a horrific nuclear retaliation. (The Turkish
missiles had a similar purpose from an American point of view.) However,
Khrushchev did not adequately envision what might happen if, as did occur, he
was caught in the act.
With respect to the Cuban missile crisis, the substates of Figure 3 which
brought us to the brink of nuclear war can now be
identified as:
Conflict between America and Castro's Cuba.
Russia demanding to be treated as a military equal and being denied this
status.
The Berlin crisis.
The Bay of Pigs invasion.
The American deployment of missiles in Turkey.
Khrushchev's deployment of missiles in Cuba.
The actors involved in each step did not perceive their behavior as overly
risky. But compounded and viewed from their opponent's perspective, those steps
brought the world to the brink of disaster. During the crisis, there were
additional, fortunately unvisited substates that would have made World War III
even more likely.
As just one example, the strong pressure noted by Burlatsky to correct the Bay
of Pigs fiasco and remove Castro with a powerful American invasion force
intensified after the Cuban missiles were discovered. But those arguing in
favor of invasion were ignorant of the fact - not learned in the West until
many years later - that the Russians had battlefield nuclear weapons on Cuba
and came close to authorizing their commander on the island to use them without
further approval from Moscow in the event of an American invasion.
Risk analysis
I have been concerned with averting nuclear war for over 25 years, but an
extraordinary new approach only occurred to me last year: using quantitative
risk analysis to estimate the probability of nuclear deterrence failing. This
approach is a bit like Superman disguised as mild-mannered Clark Kent but,
before I can explain why it is extraordinary, we need to explore what it is and
overcome a key mental block that helps explain why no one previously had
thought of applying this valuable technique.
To understand this mental block, imagine someone gives us a trick coin,
weighted so heads and tails are not equally likely, and we need to estimate the
chance of its showing heads when tossed. What do we learn if we toss the coin
50 times and it comes up tails every time? Statistical analysis says we can be
moderately confident (95% to be precise) that the chance of heads is somewhere
between zero and 6% per toss, but that leaves way too much uncertainty.
Thinking of the 50 years that deterrence has worked without a failure as the 50
tosses of the the coin, we are moderately confident that the chance of nuclear
war is somewhere between zero and 6% per year. But there is a big difference
between one chance in a billion per year and 6% per year, both of which are in
that range. At one chance in a billion per year, a few more years of business
as usual would be an acceptable risk. But 6% corresponds to roughly one in 16
odds, in which case our current nuclear strategy would be the equivalent of
playing nuclear roulette - a global version of Russian roulette - once each
year with a 16-chambered revolver.
Just as the overly simplified two-state model of Figure 2 hides the danger of a
nuclear war, the coin analogy hides the possibility of teasing much more
information from the historical record - the two-sided coin corresponding to
Figure 2's two states. Breaking down one large state of Figure 2 into Figure
3's smaller substates illuminated the danger hidden in the two-state model. In
the same way, risk analysis breaks down a catastrophic failure of nuclear
deterrence into a sequence of smaller failures, many of which have occurred and
whose probabilities can therefore be estimated.
Modern risk analysis techniques first came to prominence with concerns about
the safety of nuclear reactors, and in particular with the 1975 Rasmussen
Report produced for the Nuclear Regulatory Commission. In Risk-Benefit Analysis,
Richard Wilson and Edmund Crouch note "[The Rasmussen report] used event tree
analysis ... This new approach originally had detractors, and indeed the
failure ... to use it may have contributed to the occurrence of the Three Mile
Island Accident. If the event tree procedure ... had been applied to [the
reactor design used at Three Mile Island] ... probably the Three Mile Island
incident could have been averted." [Wilson and Crouch 2001, pages 172-173]
An event tree starts with an initiating event that stresses the system. For a
nuclear reactor, an initiating event could be the failure of a cooling pump.
Unlike the catastrophic failure which has never occurred (assuming we are
analyzing a design different from Chernobyl's), such initiating events occur
frequently enough that their rate of occurrence can be estimated directly. The
event tree then has several branches at which the initiating event can be
contained with less than catastrophic consequences, for example by activating a
backup cooling system. But if a failure occurs at every one of the branches
(eg, all backup cooling systems fail), then the reactor fails catastrophically.
Probabilities are estimated for each branch in the event tree and the
probability of a catastrophic failure is obtained as the product of the
individual failure probabilities.
Applying risk analysis to the catastrophic failure of nuclear deterrence, a
perceived threat by either side is an example of an initiating event. If either
side exercises adequate caution in its responses, such an initiating event can
be contained and the crisis dies out. But the event tree consisting of move and
counter-move can fail catastrophically and result in World War III if neither
side is willing to back down from the nuclear abyss, as almost happened with
the 1962 Cuban crisis. Each branch or partial failure corresponds to moving one
or more substates toward disaster in Figure 3.
Because nuclear deterrence has never completely failed, the probability
assigned to the last branch in the event tree (the final transition in Figure
3) will involve subjectivity and have more uncertainty. Confidence in the final
result can be increased by incorporating a number of expert opinions and using
a range instead of a single number for that probability, as well as providing
justifications for the different opinions.
The Cuban missile crisis provides a good example of how to estimate that final
probability. Kennedy estimated the odds of the crisis going nuclear as
"somewhere between one-in-three and even". His secretary of defense Robert
McNamara wrote that he didn't expect to live out the week, supporting an
estimate similar to Kennedy's. At the other extreme, McGeorge Bundy, who was
one of Kennedy's advisors during the crisis, estimated those odds at 1%.
In a recently published preliminary risk analysis of nuclear deterrence, I used
a range of 10% to 50%. I discounted Bundy's 1% estimate because invading Cuba
was a frequently considered option, yet no Americans were aware of the Russian
battlefield nuclear weapons which would have been used with high probability in
that event. As an example of faulty reasoning due to this lack of information,
Douglas Dillon, another member of Kennedy's advisory group, wrote, "Military
operations looked like they were becoming increasingly necessary ... The
pressure was getting too great ... Personally, I disliked the idea of an
invasion [of Cuba] ... Nevertheless, the stakes were so high that we thought we
might just have to go ahead. Not all of us had detailed information about what
would have followed, but we didn't think there was any real risk of a nuclear
exchange." (Blight & Welch 1989, page 72.)
The sequence of steps previously listed as leading up to the Cuban crisis is an
example of an event tree that nearly led to a catastrophic failure, and
re-examining those steps in the light of similar current events will show that,
contrary to public opinion which sees the threat of nuclear war as a ghost of
the past, the danger is lurking in the shadows, waiting until once again it can
surprise us by suddenly leaping into clear view as it did in 1962:
Step 1: Conflict
between America and Castro's Cuba: The current
conflicts between Russia and a number of former
Soviet client states are similar. For example, as
noted earlier, Bush is pushing for Georgia to
become a NATO member even though Russia and
Georgia just fought a war over still unresolved
issues.
Step 2: Russia demanding to be treated as a military equal and
being denied this status: The same is true today. Even though Russia has 15,000
nuclear weapons, America sees itself as the sole remaining superpower, leading
even former president Mikhail Gorbachev to say recently, "There is just one
thing that Russia will not accept ... the position of a kid brother, the
position of a person who does what someone tells it to do." Repeated American
statements that we defeated Russia in the Cold War add fuel to that fire since
the Russians feel they were equal participants in ending that conflict.
Steps 3 and 4: The Berlin Crisis and the Bay of Pigs invasion:
Several potential crises are brewing (eg, Chechnya, Georgia, Estonia and
Venezuela) which have similar potential.
Step 5: The American deployment of IRBM's in Turkey: A missile
defense system we are planning for Eastern Europe bears an ominous similarity
to those Turkish missiles. While these new missiles are seen as defensive and a
non-issue in America, the Russians see them as offensive and part of an
American military encirclement. In October 2007, then Russian president and now
Prime Minister Valdimir Putin warned, "Similar actions by the Soviet Union,
when it put rockets in Cuba, precipitated the Cuban missile crisis." Two months
later, Gorbachev questioned America's stated goal of countering a possible
Iranian missile threat, "What kind of Iran threat do you see? This is a system
that is being created against Russia."
Step 6: Khrushchev's deployment of the Cuban missiles: While
there is not yet a modern day analog of this step, serious warning tremors have
occurred. In July 2008, Izvestia, a Russian newspaper often used for strategic
governmental leaks, reported that if we proceed with our Eastern European
missile defense deployment, then nuclear-armed Russian bombers could be based
on Cuba. During Senate confirmation hearings as air force chief of staff,
General Norton Schwartz countered, "We should stand strong and indicate that is
something that crosses a threshold, crosses a red line." While the Russian
Foreign Ministry later dismissed Izvestia's reports as false, there is a
dangerous resemblance to events which led to the Cuban missile crisis.
The fact that we are not yet staring at the nuclear abyss is little cause for
comfort. In terms of the sequence of events that turn a 99.9% safe maneuver
into a fatal accident, we are already at a dangerous point in the process and,
as in soaring, need to recognize complacency as our true enemy.
How risky are nuclear weapons?
Even minor changes in our nuclear weapons posture have been rejected as too
risky even though the baseline risk of our current strategy had never been
estimated. Soon after recognizing this gaping hole in our knowledge, I did a
preliminary risk analysis which indicates that relying on nuclear weapons for
our security is thousands of times more dangerous than having a nuclear power
plant built next to your home.
Equivalently, imagine two nuclear power plants being built on each side of your
home. That's all we can fit next to you, so now imagine a ring of four plants
built around the first two, then another larger ring around that, and another
and another until there are thousands of nuclear reactors surrounding you. That
is the level of risk that my preliminary analysis indicates each of us faces
from a failure of nuclear deterrence.
While the analysis that led to that conclusion involves more math than is
appropriate here, an intuitive approach conveys the main idea. In science and
engineering, when trying to estimate quantities which are not well known, we
often use "order of magnitude" estimates. We only estimate the quantity to the
nearest power of 10, for example 100 or 1,000, without worrying about more
precise values such as 200, which would be rounded to 100.
In this intuitive approach I first ask people whether they think the world
could survive 1,000 years that were similar to 20 repetitions of the last 50
years. Do they think we could survive 20 Cuban missile crises plus all the
other nuclear near misses we have experienced? When asked that question, most
people do not believe we could survive 1,000 such years.
I then ask if they think we can survive another 10 years of business as usual,
and most say we probably can. There's no
Head
Office: Unit B, 16/F, Li Dong Building, No. 9 Li Yuen Street East,
Central, Hong Kong Thailand Bureau:
11/13 Petchkasem Road, Hua Hin, Prachuab Kirikhan, Thailand 77110
