| Saturday,
September 20, 2008 Wil McCarthy on the Black Swan (from
another conference):
At heart, THE BLACK SWAN is really a technical
treatise, not a social one. There's a lot of window dressing to make a full
book out of it, and Taleb clearly enjoys ripping on people in colorful ways,
which IMO makes the book a lot of fun to read. But in some ways this
detracts from the simplicity of his core message:
MISTAKING A POWER-LAW DISTRIBUTION FOR A GAUSSIAN ONE
GUARANTEES THAT COLOSSAL UNFORESEEN FAILURES WILL HAPPEN AT APPARENTLY
RANDOM INTERVALS.
It's one of those observations that, in hindsight,
seems blindingly obvious. And yet, for the financial community to get their
heads around it could take generations. Power laws are not taught in finance
class!
-- Wil McCarthy <http://www.wilmccarthy.com
>
Engineer, Columnist, Author, etc.
"Nature, to be commanded, must be obeyed" -- Francis Bacon
==
David Friedman on Taleb,the Black Swan, and current
economic conditions:
I asked my old friend David Friedman (an economist
professor of law at Santa Clara University) his views on the current
economic mess.
>Two queries: Have you any take on what's happening
in the >economics/finance world,
Not a lot. Part of it pretty clearly is what happens
when the government, at least implicitly, guarantees firms against the
downside. But my interests as an economist aren't mainly the economy, so I
don't know enough to offer predictions.
> >And > >What do you think of Taleb's works?
I haven't read them. Looking at the piece you link to,
I certainly agree that statistics is a tricky field and can be used to draw
mistaken conclusions or prove what the researcher wants to prove. I discuss
that a little in _Future Imperfect_, in the context of specification
searches--analyzing the data in lots of different ways, each reasonable, but
then only reporting the way that gives the result you want.
His turkey graph corresponds to a point I remember my
father [Milton Friedman, Nobel economist; JEP] making many years ago, I
think in the context of gambling on the future value of the Mexican
currency. Each year for many years, you could make money by betting on the
peso--buying one year futures, then selling them for a higher price when the
year was passed. The reason was not that the futures market was irrational
but that devaluation was a rare event. So you would make a little money each
year for fifteen years, then lose it all on the sixteenth when the Mexican
government devalued the peso. More generally, there are real problems in
using statistics to predict things if you don't know the form of the
underlying probability distribution, which is part of his point.
A different version of his point about binary vs
non-binary risk ... . A long time ago, a friend who ran an investment
newsletter discussed with me an idea for evaluating investment advisors. Let
each of them give the newsletter a series of buy and sell decisions, the
newsletter would keep track of the results and report which one would have
done best over (say) a six month period. I pointed out that the sensible
advisor wouldn't bet to maximize the expected return but to maximize the
probability of a high return--and the fact that one advisor won his bet
wouldn't tell you how good his advice was.
Taleb sounds bright and interesting, and his basic
point about what sorts of problems statistics is more or less useful for
seems right, but I would have to read more to say more. Since I have no
involvement in the sorts of applications of statistics he is criticizing, I
don't know how fair his criticism is. He never seems to say just what black
swan he believes showed up to cause the present problems.
At a slight tangent, involving a simpler error ... .
In my experience, most non-statisticians who talk about confidence intervals
and related ideas have them backwards. They think that a .05 result from an
experiment or statistical analysis means there is only a .05 chance that the
theory is false. In fact, what it means is that if the theory is false (in a
particular way, the null hypothesis), there is only a .05 chance that the
evidence it is true would be as good as it is.
My standard example is to pull a coin out of my pocket
and, without examining it, flip it once. The theory is that it's a double
headed coin, the null hypothesis that it's fair. It comes up heads. The
probability that it would come up heads, thus supporting the double headed
hypothesis, if it's a fair coin is only .5. It doesn't follow that the
probability that it is double headed is .5.
I've now read the essay you linked to, and see nothing
to disagree with, aside from the factual claims about how people have been
acting, which I don't know enough to judge. His point about moral hazard is
actually one I made in _Law's Order_, in explaining why a firm large enough
to self-insure might choose to buy insurance instead. The manager in charge
of a single factory has an incentive to take suboptimal precautions against
fire, since fire is a rare event. Ninety-nine times out of a hundred, the
$10,000 he saves makes him look good, gets him a bonus or promotion. One
time out of a hundred the factory burns down, the company loses $10,000,000,
and he gets fired and goes to do something else. By insuring the factory,
the company transfers at least some of the decisions to the insurance
company, which it can trust more than it can trust its own employee.
One of my phrases for my kids is "redundancy is your
friend."
-- David Friedman
www.daviddfriedman.com
==========
September column part 1
I liked _Old Man's War_ too. He doesn't feel like an
imitation of anyone else--there are faint echoes of Heinlein, but Scalzi is
very much his own voice.
He did make one historical error, or at least one of
his characters did. Don't know if you noticed.
On the subject of tools for poor writers ... . The
least expensive version of the Acus Aspire One costs $330, with linux, word
processor, WiFi, a usable keyboard, and it weighs two pounds. At some point
I'm going to get one--probably the HD version--for traveling.
-- David Friedman
The Aspire does look like a good minimum system. I
continue to like: the HP TabletPC, and the Mac Book Air, with the Air the
best production portable writing machine I have ever used. Both are of
course high end tools for successful (or those who are sure they will be
successful) writers. They aren't cheap, but they're sure good. For research
you can't beat a good TabletPC with OneNote.
====================
Dear Dr. Pournelle,
In regard to the issue of automatic train control you
wrote “there must be 20 small game design companies that could build it
nearly in their sleep”
Railway safe working involves a whole slew of issues
far removed from game design. Operating a model railway is, from a safe
working point of view, very different from the real thing. A model railway
operator generally has a bird’s eye view of the track, communication with
other operators (if any) is not much of an issue, the trains stop instantly
when power is removed and the worst that can happen is a damaged model.
On the prototype the driver often has a very limited
view, communications can be a nightmare, the braking distance may be several
miles and worst case involves hundreds of deaths.
A train control system must offer an incredibly high
level of reliability in a physically harsh environment, deal with breakdowns
and vandalism (signals have shown false greens because shooters have put
bullets through the electronics), fail safe (i.e. stop trains if in doubt)
and yet still allow a high level of traffic. This translates into a lot of
money and writing the software is only a very small part.
Even the dead mans handle, supposedly a tried and true
technology, has major issues as illustrated by the accident at Waterfall
(NSW) in January 2003 (see http://www.cityrail.info/general/waterfall.pdf
for details). If you require the driver to continually press against a
spring you risk him getting RSI – not a good idea in today’s litigious
society. Make the spring weak enough to avoid RSI and the device won’t
trigger when a heavy driver collapses.
Locomotives in NSW are fitted with a vigilance control
which requires the driver to press a button once every 60 seconds. Some
ingenious souls got around this by building a mechanical device (powered off
the windscreen wipers) to press it for them.
Let’s look at stopping a train at a signal. You can
fit a device at the signal to apply the brakes if a train passes it at stop.
Brunel’s GWR was doing this back in 1902. This is a good added safety
feature but it isn’t perfect. You can still collide with an obstruction
which is within braking distance of the signal.
You say “Okay, I’ll put the sensor far enough in front
of the signal that the train will stop by the time it reaches the signal”.
This raises all sorts of issues if your railway carries trains with widely
varying braking characteristics but let’s put that aside and assume all
trains are similar. You have now effectively double blocked your line since
each block now has a trailing block which must be kept clear of traffic if
the first block is occupied. This means you’ve just halved your carrying
capacity – probably not a good idea if it’s a densely trafficked commuter
line struggling to cope with peak times.
How about we put a really smart computer on each
train, give it full knowledge of the track (grades, curves etc) and fix it
so it can stop the train at a signal? Such systems require input so they
know the status of signals ahead. This input can be either intermittent
(which leads to reliability problems – if I don’t hear anything do I assume
nothing’s changed) or continuous (very expensive). Sending signals through
the rails might reach the wrong train because the target train is on a rusty
section of track. Radio signals can be lost if the train is in a black spot.
By the way, you’ll need military grade encryption for
your communication system or some hacker will cause a train wreck just for
the fun of it.
Queensland Rail has been experimenting for quite a few
years now using a Swedish developed system called Ericab. Ericab’s problem
(common to all such systems) is that it must err on the side of caution when
braking a train and so tends to undershoot the stop. This is a major issue
when pulling into a passing loop as the rear of the train may foul the main.
The driver, of course, can restart the train and creep up to the signal but
as it costs around $400 to stop and restart a typical freight train this is
to be avoided if possible.
QR’s solution is to add a switch so that the driver
can disable Ericab and avoid the undershoot. So having spent millions of
dollars on an automatic train control system we add a $2 switch to allow
people to turn it off. (Use of which led to a head on collision in 1994)
There’s also the question of what happens when foreign
trains want to use your tracks (never mind the enthusiast’s special hauled
by a vintage steam loco). They might get a touch upset at having to mount
expensive boxes on their locos and get the crews certified in their use –
especially if they already have boxes for their own system, one which is
incompatible with yours.
Oh yes, most of these systems are protected by patents
so be careful choosing a supplier. If you get locked into his system he’ll
be able to charge you whatever he likes. Unless you are going to roll your
own, in which case you’ll need a multi million dollar research budget and be
very careful you don’t infringe existing patents.
Finally, the key issue in most railway accidents is
not technology corporate culture. What is management’s attitude to the
accident that started this discussion? If it’s “People have died. How
dreadful! What can we do to ensure this doesn’t happen again” then upping
the technological safeguards might be part of a solution.
On the other hand, if it’s “People have died. Quick!
Ring the lawyers, we need to cover our butts!” then I submit no technology,
no matter how sophisticated, will be of much use.
Yours faithfully,
Graham Saint
Golly. You mean the map is not the territory?
At the moment they have a block signal system that
can tell when a block is occupied but cannot tell if there are two trains on
the block, and they say that GPS is not accurate enough, so they shouldn't
change what they have. Everything they might add to the system won't be
tested and proper and perfect.
And the result was the crash.
As I said, that accident could have been prevented
with a Mac Book and GPS in each train. That's what, $2,000 a train plus a
server that can run what amounts to World of Warcraft.
The perfect is always the enemy of the good.
As to the politics of the situation, I always
thought one should have a candidate to change to before rushing out to say
"I want change." Railway safety isn't really my specialty, but common
sense is; and if I can see solutions that are nearly good enough, and those
are being ignored, I do wonder if it's not time for citizens to enter the
discussions even if we're not experts and we don't see how to do it
perfectly.
The perfect is always the enemy of the better.
===========
A Bad Thing: First Chinese Carrier Aviators
http://www.strategypage.com/htmw/htnavai/articles/20080919.aspx
September 19, 2008: China announced that its first
class of carrier aviators had begun training at the Dalian Naval Academy.
The naval officers will undergo a four year course of instruction to turn
them into fighter pilots capable of operating off a carrier. China already
has an airfield, in the shape of a carrier deck, built at an inland
facility. The Russians have warned China that it may take them a decade or
more to develop the knowledge and skills needed to efficiently run an
aircraft carrier. The Chinese are game, and are slogging forward. Earlier
this year, the Russian aircraft carrier Varyag was renamed the Shi Lang
(after the Chinese general who took possession of Taiwan in 1681, the first
time China ever paid any attention to the island) and given the pennant
number 83. The Chinese have been refurbishing the Varyag, one of the
Kuznetsov class that Russia began building in the 1980s, for several years
now. It is expected to be ready for sea trials by the end of the year. The
Varyag has been tied up in a Chinese shipyard at Dailan since 2002. While
the ship is under guard, it can be seen from a nearby highway. From that
vantage point, local military and naval buffs have noted that some kind of
work is being done on the ship. The only visible signs of this work are a
new paint job <http://www.strategypage.com/htmw/htnavai/articles/20080919.aspx##>
(in the gray shade used by the Chinese navy) and ongoing work on the
superstructure (particularly the tall island on the flight deck.) Many
workers can be seen on the ship, and material is seen going into (new stuff)
and out of (old stuff) the ship. The new contracts are believed to be for
more equipment for the Varyag, in addition to the non-custom stuff already
going into the ship. Originally the Kuznetsovs were conceived of as 90,000
ton, nuclear powered ships, similar to American carriers (complete with
steam catapults). Instead, because of the cost, and the complexity of modern
(American style) carriers, the Russians were forced to scale back their
goals, and ended up with the 65,000 ton (full load ) ships that lacked steam
catapults, and used a ski jump type flight deck instead. Nuclear power was
dropped, but the Kuznetsov class was still a formidable design. The thousand
foot long carrier normally carries a dozen navalized Su-27s (called Su-33s),
14 Ka-27PL anti-submarine helicopters, two electronic warfare helicopters
and two search and rescue helicopters. But the ship can carry up to 36
Su-33s and sixteen helicopters. The ship carries 2,500 tons of aviation fuel
<http://www.strategypage.com/htmw/htnavai/articles/20080919.aspx##> ,
allowing it to generate 500-1,000 aircraft and helicopter sorties. Crew size
is 2,500 (or 3,000 with a full aircraft load.) Only two ships of this class
exist; the original Kuznetsov, which is in Russian service, and the Varyag.
Currently, the Kuznetsov is operating in the Mediterranean. The Chinese have
been in touch with Russian naval construction firms, and may have purchased
plans and technology <http://www.strategypage.com/htmw/htnavai/articles/20080919.aspx##>
for equipment installed in the Kuznetsov. Some Chinese leaders have quipped
about having a carrier by 2010 (this would have to be a refurbished Varyag).
Even that would be an ambitious schedule, and the Chinese have been burned
before when they tried to build new military technology in a hurry.
John
Sorry about the formatting, but I don't have time
to reformat today. I usually just ignore really badly formatted mail unless
it's important (and frankly I don't read a lot of it). The proper format is
given here many times: two carriage returns between paragraphs. I don't care
if you mark line ends, but it's better if you don't. But double space
between paragraphs.
==============
Fannie, Freddie, Bear Sterns, and AIG
Since the government has taken over these firms, their
executive team now works for me, the American taxpayer. Paulsen has already
said he intends to replace the management of these firms. I don't want these
greedy sob's to get one more cent of my money for years of mismanagement,
and for running their companies into the ground. How about if we promise
them immunity from prosecution in return for tearing up their severance
packages?
Rich Graff Troy, OH
============
Ships That Won't Sail.
<http://www.defensetech.org/archives/004431.html>
-- Roland Dobbins
We used to be able to build a ship a day. And they
worked, or we sent them out and the officers and crews made them work. But
that was in a less diverse country I guess.
==========
Rosenberg sons acknowledge
dad was spy.
<http://www.msnbc.msn.com/id/26761635/>
-- Roland Dobbins
It has long been the view of those best informed
that Julius -- who proudly called himself Stalin's Soldier -- was guilty of
high treason, but Ethyl was less so; and that the death sentence for Ethyl
was intended to force Julius to cooperate with the counter-intelligence
authorities in exchange for a pardon for her. That offer was given to Julius
right up to the time of the executions, but he was adamant.
I was concerned about the ethics of that situation
at the time and I signed a petition to commute their sentences to life;
which got me a question on my security clearance for Top Secret some years
later, but that's another story.
==
Case closed: The Rosenbergs were Soviet spies.
http://www.latimes.com/news/opinion/
la-oe-radosh17-2008sep17,0,490961.story
-- Roland Dobbins
==========
Viking Age triggered by shortage
of wives?
<http://www.msnbc.msn.com/id/26755692/>
- Roland Dobbins
Well -- yes. But there were unforeseen
consequences. Vikings raided Ireland for slave girls. Nest thing they knew
-- and they never quite knew how it happened -- the Irish girls were wives,
not slaves, and shortly after that a celibate priest was telling them when
they could sleep with their wives. It was called civilizing the Vikings, and
it worked so well that the Normans -- Frenchified Danes (from Skene which is
now part of Sweden) hired by the French to keep the other Vikings out of the
land that became known as Normandy -- became one of the most religious
people of the age, and endowed monasteries and other religious institutions.
The shortage of women was probably more related to
"childbed fever" and the general state of hygiene and health in Viking
villages than female infanticide; so far as I know, Vikings were overly
protective of women and there is no archeological evidence I know of for
female infanticide. There were Viking heiresses, and they where highly
sought after, but most men went a-viking in order to find wives because
there wasn't a lot of land. Or so I have been told about my ancestors.
============
We have seen this story before, but it does no harm
to repeat it:
Why the wall street problems scare me
Dr. Pournelle,
This old explanation of the U.S. tax structure and
liberal politics explains exactly why the Wall Street problems scare me,
even in what *should* be an “up” economy, absent the mortgage crisis.
Sean
Suppose that every night, ten men go to their
favorite bar for beer. The tab for all ten comes to $100 for ten pitchers.
If they paid their bill the way we pay our taxes, it would go something
like this:
* The first four men (the poorest) would pay
nothing. * The fifth would pay $1. * The sixth would pay $3. * The seventh
$7. * The eighth $12. * The ninth $18. * The tenth man (the richest) would
pay $59.
So, that’s what they decided to do. The ten men
drank in the bar every night and seemed quite happy with the arrangement,
until one day, the owner threw them a curve.
“Since you are all such good customers,” he said,
“I’m going to reduce the cost of your nightly tab by $20.”
So, now drinks for the ten only cost $80. The group
still wanted to pay their tab the way we pay our taxes. So, the first four
men were unaffected. They would still drink for free.
But what about the other six, the paying customers?
How could they divvy up the $20 windfall so that
everyone would get his ‘fair share’?
The six men realized that $20 divided by six is
$3.33. But if they subtracted that from everybody’s share, then the fifth
man and the sixth man would each end up being ‘PAID‘ to drink beer!
So, the bar owner suggested that it would be fair to
reduce each man’s bill by roughly the same amount, and he proceeded to
work out the amounts each should pay.
And so:
* The fifth man, like the first four, now paid
nothing (100% savings). * The sixth now paid $2 instead of $3 (33%
savings). * The seventh now paid $5 instead of $7 (28% savings). * The
eighth now paid $9 instead of $12 (25% savings). * The ninth now paid $14
instead of $18 (22% savings). * The tenth now paid $49 instead of $59 (16%
savings).
Each of the six was better off than before. And the
first four continued to drink for free. But once drunk and outside the
bar, the men began to compare their savings.
“I only got a dollar out of the $20,” declared the
sixth man. He pointed to the tenth man “but he got $10!”
“Yeah, that’s right,” exclaimed the fifth man. “I
only saved a dollar, too. It’s unfair that he got ten times more than me!”
“That’s true!!” shouted the seventh man. “Why should
he get $10 back when I got only $2? The wealthy get all the breaks!”
“Wait a minute,” yelled the first four men in
unison. “We didn’t get anything at all. The system exploits the poor!”
The nine men surrounded the tenth and beat him up,
took the $10 and gave each person a dollar.
The next night the tenth man didn’t show up at the
bar, so the nine sat down and drank without him. But when it came time to
pay the tab, they discovered something important. They didn’t have enough
money between all of them for even half of the tab!
And that, boys and girls, journalists and college
professors, is how our tax system works. The people who pay the highest
taxes get the most benefit from a tax reduction. Tax them too much, attack
them for being wealthy, and they just may not show up to pick up the tab
anymore.
On the other hand, we have a situation of much
greater gaps between average pay and the pay of the top executives, who are
often rewarded even as they are fired for ruining the company (and the
investors, and those who worked for the company for normal pay). It is one
thing to defend property; it is another to set up a system in which those
who do most of the work for the company are ruined by the decisions of those
making 100 times as much as they do.
Too much concentration of wealth produces a very
unstable situation: who will defend the wealth? In old Rome the Emperor had
to distribute bounties to the soldiers at frequent intervals....
The alternative to government by force is self
government, but self government is not likely to be enthusiastic about
enormous discrepancies in wealth and reward. Taxes, on the other hand,
enrich government while giving to the masses bread and circuses and whatever
the management thinks they ought to get...
================
Bailout Costs
The S&L crisis and the 1990s Resolution Trust
Corporation cleanup cost the gummint $200 billion in today's dollars. Since
foreclosure rates now are about 5 times higher than during the S&L crisis, a
minimum estimate of $1 trillion is very reasonable.
And that straightline comparison is an optimistic best
case. It depends on the assumption this new RTC II will be operating in
economic conditions very similar to those of the 1990s. Those conditions
were the end of the Cold War, oil falling to the equivalent of $25/ barrel,
the mother of all stock bull markets taking off and the Boomer Generation
reaching the peak of its earning power.
Since the macroeconomic environment is a little more
challenging now (tongue in cheek), there are reasons to believe RTC II won't
recover as much money for the gummint as RTC I did. Taxpayer costs can
therefore be expected to reach higher than $1 trillion.
Mark
====================
TOP
CURRENT
VIEW Saturday |
