I am guessing that in about half these cases, at the individual level, the contractors are former government employees who weren't getting paid their fair market value by the public sector. Given that a good IT worker is worth about 5 times a medioce one and 20 times a bad one, they're probably a much better value, on average, than those "left behind". Consulting budgets and the like also let huge bureaucracies get necessary work done that is internally impossible because it is "not in the budget".
The other half these cases, I am also guessing, will prove to be unnecessary wastes of money even worse than typical government IT initiatives.
To begin with, the financial sector has more or less always used scenario analysis which is of course distribution independent. The obvious flaw, of being only as good as the selection of chosen scenarios, is in evidence with the 2008 credit crisis. Nevertheless, the scenarios involved are based on human experience and intuition and as such are obviously independent of probabilistic models. In general, the financial sector does a better job of scenario testing than any other sector.
Even today you will of course find many risk computations done with multivariate gaussian models. The results are very useful for day-to-day management and even Mandelbrot would concede that on most days they are a darn good approximation. The exceptional days, the tail events, are where the gaussian models are nigh useless. Everyone knows that, and that's why the scenarios, position limits, and so on are put in place.
To make a car analogy, pretending the financial sector relies solely on multivariate gaussian models for controlling tail risk is like pretending that motorists control their vehicles with gas pedal and steering wheel only, eschewing the brakes. To expand the analogy to fat-tailed models mentioned elsewhere, I would say the fat-tailed models are like upgrading to a sport suspension. Nice, but you had still better keep the brakes installed.
You mention that you would
be very surprised if the quants these days work on fundamentally different assumptions from the academics
As it turns out, the quants tend to use the more simplistic stuff by far. A lot of the academic work ignores real-world problems, such as model calibration, parameter uncertainty, or computational efficiency.
Now, securities pricing models are a different kettle of fish from risk models. But since the topic was broached, mbkennel notes that
Neither academics or practicing quants actually use thin-tail Gaussian models for practical pricing.
This quite true, the models used are, in effect fat-tailed. You wonder
Who solved the problem that the parameters for fat-tailed models cannot be estimated from data?
For pricing, we have the luxury of inferring model parameters from market prices. The models used are typically not running with fat-tailed stochastic processes, but in practice the pricing and trading is fat-tailed because the parameters are allowed to vary in theoretically inaccessible ways that express fat-tailed outcomes. Implied volatility is the most famous example of this. If you work through the mathematics of what is really happening, you will find that the market outcomes are the same as if everyone was using a fat-tailed model to begin with.
Again, this latter bit is about pricing. The closest analogy for risk models would be using empirical return distributions, which is certainly done a lot but outside the scope of this discussion.
move their money into the stock market to avoid investing it into projects that create real value
The stock market is is exactly where that money can create real value. The stock market is not some wormhole shooting food and iPads into the sun, you know. Using $1 to buy stock in, say, Netflix puts another dollar toward video streaming technology, with all the rewards to society that come from that, including network admin jobs, computer equipment orders, lowered distribution costs, etc.
One has to consider the metric used to select the "best".
For any individual, the question is what health care system is best for someone like them. The U.S. system is expensive but very high quality for the mid-middle class and above. That is to say, it is excellent for those who have health insurance and little chance of losing it. Below that income level, the quality of health care ranges down to nearly non-existent, bringing down the aggregate statistics.
For society as a whole, the aggregate statistics and economic efficiency in achieving them is the most important. By the first measure, Cuba is roughly on par with the USA. By the second, it is far ahead.
it is not a mathematical modeling problem, but an epistemological one
Quite. I cringed when the grandparent brought heavy-tailed models into the discussion. Financial firms are well aware of this, and fight the epistemological problem as best they can using position limits and scenario analysis. Generally, they are more rigorous and systematic about that than businesses operating in other sectors.
The credit disaster of 2008 shows that, even using position limits and scenario analysis, absolutely idiotic things can happen. For me, though, the lesson was not that financial companies were doing too poor a job of perceiving, understanding and controlling risk. They do as good a job as anybody, i.e. not very good, and they always will as history has shown at least since the 1600s. Rather the lesson is firms should not be allowed to get so large, especially if they are to be associated with banks, and that banks themselves must be instantly and easily separable from riskier groups.
he has pointed out that the models cannot account for random political events, for psychological forcing, or natural disasters. The probability of these cannot be reliably assessed.
However that is not exactly news, and financial firms were worrying about, and limiting, such risks long before the whole business became very mathematical. The true and good things in Taleb's books are not novel, and whatever is novel in them is not of high quality.
It is also untrue that
smart mathematicians and physicists in the world can only derive better models and better algorithms
Much as you would prefer to demonize them, for example with the psycopath label, they are humans and capable of quite wonderful things outside the domain of modeling. Being good at mathematics does not preclude being a good entrepreneur, businessman, or artist. Among physicists and mathematicians you will find many better at taking unquantifiable risks, such as political turmoil, into account than you are.
bugs are declared whenever any sequence of events is established that would, theoretically, lead to an unacceptable outcome. It's considered "too cowboy" to argue that a bug doesn't exist because the client hasn't get generated that sequence of events in the live system. Conversely, this is exactly how risk management works for financial risks.
That is a very interesting perspective. In this particular case, though, it is likely Adoboli bypassed risk systems, booked fake trades a la Nick Leeson, or set parameters to hide the risks. In this case the proper analogy is something closer to making systems robust to, say, SQL injection attacks, which of course we know not all programmers do even when they should.
Actually, you are incorrect. You are probably thinking of partners in private equity. Trader bonuses (and all others' cash bonuses) are taxed as ordinary income...I know this firsthand.
The tax dodge for private equity is that they don't get money, but rather stock whose appreciation is taxed at capital gains rates. Its typically stock they are not allowed to sell and turn into cash, and is therefore regarded as an "investment". They are taxed at ordinary (35%+$state) rates on the initial grant of, say, $1MM in stock but the subsequent 20% per year gains (if they happen) are taxed at the 15% capital gains rate.
Actually, in this case, Adoboli will go probably to jail and the compensation of bonus-eligible UBS employees will be reduced by a sizeable amount. That's basically like a fine, whether or not you think they make too much money in the first place.
People who think the financial sector is using only these multivariate normal risk models (if I correctly interpret what you mean by "sigma models") are incredibly naive. Many of them have read and believed the books by N. Taleb, which I find tragicomic.
You may have noticed that thousands of extremely smart mathematicians and physicists have been hired by the financial sector over the last decades. Do you really think all those clever people somehow missed something so blindingly obvious?
As I've said elsewhere, it is unlikely this whole situation arises from an unusual market event combined with normal operations at UBS. Far more likely, Adoboli was able to hack the risk system into assuming the EUR/CHF exchange rate was constant. This is known as operational risk and is not very amenable to quantification.
That proposal for a certain amount of debt issuance to be held by "an especially sophisticated tranche of institutional investors" was a smart one, and it angered me to no end that the banks weaseled out of it. It would have been a low-cost parallel form of regulation, and the prices of that debt would have been interesting indicators of institutional health completely independent from the ratings agencies.
There is a little bit of similar activity out there, such as the bonds issued (earlier this year) by Credit Suisse that turn into equity if the Tier 1 capital ratio falls below a certain level, but it isn't organized, widespread, or well-known.
To your point about regulations, I'll just note that we're kind of talking past each other here. I was mainly posting about Adoboli's presumed circumvention of internal risk controls, while you're taking a larger perspective on financial regulation. It's quite true that the finance industry fights regulations, in many cases to the detriment of the economy, because they want to make more money. The same stresses occur within a financial organization between its traders and risk control, but the difference to me is that a single company is a lot like a communist dictatorship (planned economy, central control) and imposes risk controls through a very different sociological process.
I would take care, incidentally, about lumping the finance industry all together in its attitude to regulation. Hedge funds and trading exchanges, for example, would love to see all CDS contracts traded on exchanges the way the SEC tried to mandate. The dealers in the investment banks fought that one off.
In contrast, equity short sale restrictions were really annoying to long-short asset managers (hedge funds, a few mutual funds, option market makers), but much less of a problem to equity market makers, brokers, and the big banks.
Finally, consider capital requirements. With the exception of the banks, everybody would like to see those go up because no one wants to have to worry to much about counterparty risk or their prime broker going bust. (I learned the word "rehypothecated" in 2008).
My point is that there are players in the finance industry on either side of almost any regulation. The only counterexamples that come to mind right now are cross-border capital transfer controls.
I've got no answer for you there. All I can say for sure is that UBS runs VaR. How exactly the reports were ignored, deleted or altered might be broadcast someday, but there are lots of possibilities (e.g. the override I mentioned).
UBS runs VaR frequently. They have their own internal system, which isn't cutting edge but gets the job done.
The losses are $2B, but the positions that made those losses, especially when aggregated with other traders', may not have looked so large. But since this guy had also worked in the back office, another possibility is that he "hacked" the risk management system, for example by setting an override that would treat the EUR/CHF price as a constant during risk runs.
fraud... that doesn't mean we should structure the system to make it as easy for them as possible, which appears to be what we've done with trading.
Rogue traders and corruption have plagued large organizations basically forever. The structure of financial organizations actually makes it reasonably hard for them to operate, though obviously not impossible since one makes the papers every couple of years. Rogues have to "hack" the risk management system in one way or another; only a few succeed to the point of making spectacular losses. Nobody has an incentive to structure their risk management systems in such a way as to be easily hackable by rogues, I don't know where you get that idea.
A rogue almost always gets started with good (-ish) intentions. They fucked up, and want to fix the problem without anybody finding out. They fuck up more trying to to do so and the whole thing becomes an ugly spiral. We had the beginnings of such a situation at our large hedge fund. However, risk control detected the problem immediately and the fellow was fired before the day was over.
With purely electronic trading and algorithmic systems, by the way, it is (so far) unheard of for such monstrous losses to mount. This is not least because the algorithms lack necessary human qualities of a rogue, such as deliberate deception of coworkers, ability/motivation to forge false reports in an artful manner, and a profit motive.
I'm saying that Linux distros copied automount technology, mainly from Apple, and did it poorly. There's hardly a one-way pattern of Apple copying from Linux. In general, I think it goes the opposite direction.
And now what have you got? A career item literally millions of people have fantasized about, the ability to talk about it, interesting skills, and presumably some cool contacts.
I work in the finance industry, and I see plenty of people around here who have done better in dollars and worse in life.
How do you justify buying your kids replacements for stuff they don't take care of?
It is in the nature of children to be immature.
Adults tend to forget that fact, especially if they are (a) between 18 and 30 or (b) live in a relatively childless hipster haven such as the Bay area or Manhattan. Those demographics are relatively unexposed to children.
Immaturity manifests itself in many ways that, in adult behavior, is clearly irksome. These include loud speech, intransigence, rambunctiousness, lack of foresight, and reckless or dangerous behavior. If you hold children to adult behavior standards, you will be irked.
I try to see the good and ignore the faults in everyone. In the case of children, that means allowing for their immaturity (and even sometimes reveling in it). It's made easier because the good side of most kids includes such endearing traits as cuteness and enthusiasm.
Specific to something like a broken toy, I would tend to use its breakage as a teaching moment. But if the value proposition is still there, I would also replace it.
Re:Git could use revision numbers
on
The Rise of Git
·
· Score: 1
Three counterpoints:
(1) Facile typists can type "3.18" much faster than they can do two focus changes and two chorded keystrokes
(2) Humans are better at recognizing simple (and short) numeric patterns than alphanumeric strings.
(3) Some people still use their voices, e.g. in face to face conversations and on telephones.
Denigrating this as a "bullshit non-issue" is arrogant and ignorant of human psychology. Other posters suggesting the use of tags are on a more reasonable track.
I agree with everything you said. In particular, while I think you meant to write "risk-averse", your neologism of "risk-adverse" is far more accurate.
As pointed out in the Ars Technica review, the installer creates a small (1GB) new partition on your hard drive without destroying any existing data. It then uses this partition to bootstrap the remainder of the install process. (That's just the sort of approach I took with a Linux system years and years ago, though my reward was a whole weekend spent trying to fix a broken system and finally just erasing the HD).
Slashdot Mirror
I am guessing that in about half these cases, at the individual level, the contractors are former government employees who weren't getting paid their fair market value by the public sector. Given that a good IT worker is worth about 5 times a medioce one and 20 times a bad one, they're probably a much better value, on average, than those "left behind". Consulting budgets and the like also let huge bureaucracies get necessary work done that is internally impossible because it is "not in the budget".
The other half these cases, I am also guessing, will prove to be unnecessary wastes of money even worse than typical government IT initiatives.
YOU-fee?
YOU-fi?
you-EF-ee?
I like "weffy".
The stock issuance does not occur in a vacuum.
To begin with, the financial sector has more or less always used scenario analysis which is of course distribution independent. The obvious flaw, of being only as good as the selection of chosen scenarios, is in evidence with the 2008 credit crisis. Nevertheless, the scenarios involved are based on human experience and intuition and as such are obviously independent of probabilistic models. In general, the financial sector does a better job of scenario testing than any other sector.
Even today you will of course find many risk computations done with multivariate gaussian models. The results are very useful for day-to-day management and even Mandelbrot would concede that on most days they are a darn good approximation. The exceptional days, the tail events, are where the gaussian models are nigh useless. Everyone knows that, and that's why the scenarios, position limits, and so on are put in place.
To make a car analogy, pretending the financial sector relies solely on multivariate gaussian models for controlling tail risk is like pretending that motorists control their vehicles with gas pedal and steering wheel only, eschewing the brakes. To expand the analogy to fat-tailed models mentioned elsewhere, I would say the fat-tailed models are like upgrading to a sport suspension. Nice, but you had still better keep the brakes installed.
You mention that you would
be very surprised if the quants these days work on fundamentally different assumptions from the academics
As it turns out, the quants tend to use the more simplistic stuff by far. A lot of the academic work ignores real-world problems, such as model calibration, parameter uncertainty, or computational efficiency.
Now, securities pricing models are a different kettle of fish from risk models. But since the topic was broached, mbkennel notes that
Neither academics or practicing quants actually use thin-tail Gaussian models for practical pricing.
This quite true, the models used are, in effect fat-tailed. You wonder
Who solved the problem that the parameters for fat-tailed models cannot be estimated from data?
For pricing, we have the luxury of inferring model parameters from market prices. The models used are typically not running with fat-tailed stochastic processes, but in practice the pricing and trading is fat-tailed because the parameters are allowed to vary in theoretically inaccessible ways that express fat-tailed outcomes. Implied volatility is the most famous example of this. If you work through the mathematics of what is really happening, you will find that the market outcomes are the same as if everyone was using a fat-tailed model to begin with.
Again, this latter bit is about pricing. The closest analogy for risk models would be using empirical return distributions, which is certainly done a lot but outside the scope of this discussion.
move their money into the stock market to avoid investing it into projects that create real value
The stock market is is exactly where that money can create real value. The stock market is not some wormhole shooting food and iPads into the sun, you know. Using $1 to buy stock in, say, Netflix puts another dollar toward video streaming technology, with all the rewards to society that come from that, including network admin jobs, computer equipment orders, lowered distribution costs, etc.
One has to consider the metric used to select the "best".
For any individual, the question is what health care system is best for someone like them. The U.S. system is expensive but very high quality for the mid-middle class and above. That is to say, it is excellent for those who have health insurance and little chance of losing it. Below that income level, the quality of health care ranges down to nearly non-existent, bringing down the aggregate statistics.
For society as a whole, the aggregate statistics and economic efficiency in achieving them is the most important. By the first measure, Cuba is roughly on par with the USA. By the second, it is far ahead.
it is not a mathematical modeling problem, but an epistemological one
Quite. I cringed when the grandparent brought heavy-tailed models into the discussion. Financial firms are well aware of this, and fight the epistemological problem as best they can using position limits and scenario analysis. Generally, they are more rigorous and systematic about that than businesses operating in other sectors.
The credit disaster of 2008 shows that, even using position limits and scenario analysis, absolutely idiotic things can happen. For me, though, the lesson was not that financial companies were doing too poor a job of perceiving, understanding and controlling risk. They do as good a job as anybody, i.e. not very good, and they always will as history has shown at least since the 1600s. Rather the lesson is firms should not be allowed to get so large, especially if they are to be associated with banks, and that banks themselves must be instantly and easily separable from riskier groups.
Taleb is correct when
he has pointed out that the models cannot account for random political events, for psychological forcing, or natural disasters. The probability of these cannot be reliably assessed.
However that is not exactly news, and financial firms were worrying about, and limiting, such risks long before the whole business became very mathematical. The true and good things in Taleb's books are not novel, and whatever is novel in them is not of high quality.
It is also untrue that
smart mathematicians and physicists in the world can only derive better models and better algorithms
Much as you would prefer to demonize them, for example with the psycopath label, they are humans and capable of quite wonderful things outside the domain of modeling. Being good at mathematics does not preclude being a good entrepreneur, businessman, or artist. Among physicists and mathematicians you will find many better at taking unquantifiable risks, such as political turmoil, into account than you are.
bugs are declared whenever any sequence of events is established that would, theoretically, lead to an unacceptable outcome. It's considered "too cowboy" to argue that a bug doesn't exist because the client hasn't get generated that sequence of events in the live system. Conversely, this is exactly how risk management works for financial risks.
That is a very interesting perspective. In this particular case, though, it is likely Adoboli bypassed risk systems, booked fake trades a la Nick Leeson, or set parameters to hide the risks. In this case the proper analogy is something closer to making systems robust to, say, SQL injection attacks, which of course we know not all programmers do even when they should.
Actually, you are incorrect. You are probably thinking of partners in private equity. Trader bonuses (and all others' cash bonuses) are taxed as ordinary income...I know this firsthand.
The tax dodge for private equity is that they don't get money, but rather stock whose appreciation is taxed at capital gains rates. Its typically stock they are not allowed to sell and turn into cash, and is therefore regarded as an "investment". They are taxed at ordinary (35%+$state) rates on the initial grant of, say, $1MM in stock but the subsequent 20% per year gains (if they happen) are taxed at the 15% capital gains rate.
Actually, in this case, Adoboli will go probably to jail and the compensation of bonus-eligible UBS employees will be reduced by a sizeable amount. That's basically like a fine, whether or not you think they make too much money in the first place.
People who think the financial sector is using only these multivariate normal risk models (if I correctly interpret what you mean by "sigma models") are incredibly naive. Many of them have read and believed the books by N. Taleb, which I find tragicomic.
You may have noticed that thousands of extremely smart mathematicians and physicists have been hired by the financial sector over the last decades. Do you really think all those clever people somehow missed something so blindingly obvious?
As I've said elsewhere, it is unlikely this whole situation arises from an unusual market event combined with normal operations at UBS. Far more likely, Adoboli was able to hack the risk system into assuming the EUR/CHF exchange rate was constant. This is known as operational risk and is not very amenable to quantification.
That proposal for a certain amount of debt issuance to be held by "an especially sophisticated tranche of institutional investors" was a smart one, and it angered me to no end that the banks weaseled out of it. It would have been a low-cost parallel form of regulation, and the prices of that debt would have been interesting indicators of institutional health completely independent from the ratings agencies.
There is a little bit of similar activity out there, such as the bonds issued (earlier this year) by Credit Suisse that turn into equity if the Tier 1 capital ratio falls below a certain level, but it isn't organized, widespread, or well-known.
To your point about regulations, I'll just note that we're kind of talking past each other here. I was mainly posting about Adoboli's presumed circumvention of internal risk controls, while you're taking a larger perspective on financial regulation. It's quite true that the finance industry fights regulations, in many cases to the detriment of the economy, because they want to make more money. The same stresses occur within a financial organization between its traders and risk control, but the difference to me is that a single company is a lot like a communist dictatorship (planned economy, central control) and imposes risk controls through a very different sociological process.
I would take care, incidentally, about lumping the finance industry all together in its attitude to regulation. Hedge funds and trading exchanges, for example, would love to see all CDS contracts traded on exchanges the way the SEC tried to mandate. The dealers in the investment banks fought that one off.
In contrast, equity short sale restrictions were really annoying to long-short asset managers (hedge funds, a few mutual funds, option market makers), but much less of a problem to equity market makers, brokers, and the big banks.
Finally, consider capital requirements. With the exception of the banks, everybody would like to see those go up because no one wants to have to worry to much about counterparty risk or their prime broker going bust. (I learned the word "rehypothecated" in 2008).
My point is that there are players in the finance industry on either side of almost any regulation. The only counterexamples that come to mind right now are cross-border capital transfer controls.
I've got no answer for you there. All I can say for sure is that UBS runs VaR. How exactly the reports were ignored, deleted or altered might be broadcast someday, but there are lots of possibilities (e.g. the override I mentioned).
UBS runs VaR frequently. They have their own internal system, which isn't cutting edge but gets the job done.
The losses are $2B, but the positions that made those losses, especially when aggregated with other traders', may not have looked so large. But since this guy had also worked in the back office, another possibility is that he "hacked" the risk management system, for example by setting an override that would treat the EUR/CHF price as a constant during risk runs.
fraud ... that doesn't mean we should structure the system to make it as easy for them as possible, which appears to be what we've done with trading.
Rogue traders and corruption have plagued large organizations basically forever. The structure of financial organizations actually makes it reasonably hard for them to operate, though obviously not impossible since one makes the papers every couple of years. Rogues have to "hack" the risk management system in one way or another; only a few succeed to the point of making spectacular losses. Nobody has an incentive to structure their risk management systems in such a way as to be easily hackable by rogues, I don't know where you get that idea.
A rogue almost always gets started with good (-ish) intentions. They fucked up, and want to fix the problem without anybody finding out. They fuck up more trying to to do so and the whole thing becomes an ugly spiral. We had the beginnings of such a situation at our large hedge fund. However, risk control detected the problem immediately and the fellow was fired before the day was over.
With purely electronic trading and algorithmic systems, by the way, it is (so far) unheard of for such monstrous losses to mount. This is not least because the algorithms lack necessary human qualities of a rogue, such as deliberate deception of coworkers, ability/motivation to forge false reports in an artful manner, and a profit motive.
I'm saying that Linux distros copied automount technology, mainly from Apple, and did it poorly. There's hardly a one-way pattern of Apple copying from Linux. In general, I think it goes the opposite direction.
And now what have you got? A career item literally millions of people have fantasized about, the ability to talk about it, interesting skills, and presumably some cool contacts.
I work in the finance industry, and I see plenty of people around here who have done better in dollars and worse in life.
On the other hand, it was a crapshoot until well past 2005 whether any given Linux distro would properly automount a FAT32-formatted USB stick.
Do the name-brand drives do this as well? I have suspicions about my HP v165w "16GB"!
I am kicking myself for posting this but....right now one of the tags on this story is the work "cromulant". That's spelled "cromulent", people!
How do you justify buying your kids replacements for stuff they don't take care of?
It is in the nature of children to be immature.
Adults tend to forget that fact, especially if they are (a) between 18 and 30 or (b) live in a relatively childless hipster haven such as the Bay area or Manhattan. Those demographics are relatively unexposed to children.
Immaturity manifests itself in many ways that, in adult behavior, is clearly irksome. These include loud speech, intransigence, rambunctiousness, lack of foresight, and reckless or dangerous behavior. If you hold children to adult behavior standards, you will be irked.
I try to see the good and ignore the faults in everyone. In the case of children, that means allowing for their immaturity (and even sometimes reveling in it). It's made easier because the good side of most kids includes such endearing traits as cuteness and enthusiasm.
Specific to something like a broken toy, I would tend to use its breakage as a teaching moment. But if the value proposition is still there, I would also replace it.
Three counterpoints:
(1) Facile typists can type "3.18" much faster than they can do two focus changes and two chorded keystrokes
(2) Humans are better at recognizing simple (and short) numeric patterns than alphanumeric strings.
(3) Some people still use their voices, e.g. in face to face conversations and on telephones.
Denigrating this as a "bullshit non-issue" is arrogant and ignorant of human psychology. Other posters suggesting the use of tags are on a more reasonable track.
I agree with everything you said. In particular, while I think you meant to write "risk-averse", your neologism of "risk-adverse" is far more accurate.
As pointed out in the Ars Technica review, the installer creates a small (1GB) new partition on your hard drive without destroying any existing data. It then uses this partition to bootstrap the remainder of the install process.
(That's just the sort of approach I took with a Linux system years and years ago, though my reward was a whole weekend spent trying to fix a broken system and finally just erasing the HD).