Help! Still trying to get my head around it

nesf

JDLK wrote: »

I see, I need to provide academic qualification before I can enter into a discussion now?

Please feel free to lecture on the reliability of probability models (in the face of the economic disintegration of the worlds largest hedge fund and derivative advisors)

Of course we all know that these models only work when markets are "well behaved" and crumble when unpredictability enters the arena. Remember "Long-Term Capital Management". The fact is that quant theories-while mathmatically provable do not gaurantee success, the main problem is that over inflated share values who's only justifaction relies on a baynesian or Black Scholes probabaility calculation go to pot once a variable deemed too unlikely enters the equation

Eh, no mate academic qualifications in Economics normally wouldn't give you much background on the pricing of options/quant methods used in the financial markets. I only asked because I don't want to give you an answer that's full of jargon that means nothing to you. You most certainly don't need a qualification in Economics or Finance to grasp some of the problems that effected the Sub Prime crisis and the ensuing credit crunch.

For instance, quant approaches work quite well in some markets, there are specific reasons for why they didn't work with respect to the Sub Prime mess.

JDLK

nesf wrote: »

Eh, no mate academic qualifications in Economics normally wouldn't give you much background on the pricing of options/quant methods used in the financial markets. I only asked because I don't want to give you an answer that's full of jargon that means nothing to you. You most certainly don't need a qualification in Economics or Finance to grasp some of the problems that effected the Sub Prime crisis and the ensuing credit crunch.

For instance, quant approaches work quite well in some markets, there are specific reasons for why they didn't work with respect to the Sub Prime mess.

Go for it, assume I have no background in stats, quant or finance.

(though I believe it is moot- proving that something is mathmatically probable then staking massive amounts of finance on it only to be proved wrong (by any variable regardless of significance) pretty much conclusively proves the uselessness of the original estimation.)

My main point is that we must go beyond the mathematical- that dynamic global markets cannoy be accurately valued or predicted and todays finacial situation is case and point, and the only way to avoid a repition of reckless financial practices is not through increased reliance on probability models but through a more interventionist, here and now, common sence approach

The fact that nobody predicted the finacial collapse is a pretty big kick in the crotch of any probability proponent

silverharp

JDLK wrote: »

My main point is that we must go beyond the mathematical- that dynamic global markets cannoy be accurately valued or predicted and todays finacial situation is case and point, and the only way to avoid a repition of reckless financial practices is not through increased reliance on probability models but through a more interventionist, here and now, common sence approach

If you want common sense , how about breaking up the gov sponsored ratings agency. Would you trust a housing survey paid by the vendor?
This whole affair has a moral hazzard dimension to this , The LTCM bailout & Y2K & 9-11 Fed response all "interventionist" actions laid part of the foundations for this.

JDLK

silverharp wrote: »

If you want common sense , how about breaking up the gov sponsored ratings agency. Would you trust a housing survey paid by the vendor?
This whole affair has a moral hazzard dimension to this , The LTCM bailout & Y2K & 9-11 Fed response all "interventionist" actions laid part of the foundations for this.

I think its a pretty big reach to say that the reckless practices of the private sector finacial institutions are the fault of governement intervention.

Think about this- the nationalisation of banks (Anglo) is the closest thing the west has ever come to real socialism- and this occured in a period of little governement regulation- if anything common sence regulation is proably the only way to ultimately protect free trade

nesf

JDLK wrote: »

Go for it, assume I have no background in stats, quant or finance.

(though I believe it is moot- proving that something is mathmatically probable then staking massive amounts of finance on it only to be proved wrong (by any variable regardless of significance) pretty much conclusively proves the uselessness of the original estimation.)

Right, I've only as much time as until the toddler wakes up to write this so apologies for it not being a complete treatment.


Fundamentally the problem in options pricing is trying to calculate what the price of a particular financial instrument (i.e. product) should be given that it isn't directly visible. Futures are the simplest example to work with, they are just a contract to buy or sell something at a specified future date. We cannot know the actual price of this something at this date so we have to estimate it. If I'm selling a future to you that says that I will buy X bushels of wheat at the start of April I ideally want to agree a price that will be lower than the price that I expect wheat will be selling for in April on average. The core concept here is that I don't need to know exactly I just need to be able to put some rough probability on future price. Now for things like wheat it turns out to not be so bad, what is knowable is the price today of wheat and what is also knowable is the price everyone else is paying for wheat in April (because you're not the only one doing this). Quants know that the probability of prices isn't like what things like the Black-Scholes model tells us. They add in a lot of different items to account for the fact that prices aren't distributed "normally" (a particular type of probability distribution that is easy to work with mathematically and describes some kinds of physical facts like height in a population).

The end result of all this mathematics and computer time is a market that tends to work relatively well. The prices that future contracts are originally sold for are rarely absolutely correct but they get it right enough of the time for it to be profitable and useful. Farmers get a guaranteed price for their harvest and can transfer the risk (or reward) of a worse (or better) harvest than expected onto a third party for a fee. These kinds of instruments work well in commodities, foreign exchange etc precisely because a lot of companies and people are happy to get a particular type of risk off their plate in exchange for a fee. The sellers of Futures can turn a profit once they are able to make a decent guess of future prices for the goods. People have been doing this successfully since the 18th century in Japan with rice futures iirc.



The problem with CDOs and the Sub Prime crisis (with respect to the above) was this:

Firstly: The financial instruments in question had a lifetime of up to 30 years! This longer the period of time the less and less accurate any estimation of value will be. 3 month or 1 year Futures markets work precisely because the periods of the estimation are relatively short.


Secondly: An individual CDO consisted of parts of hundreds of different mortgages in different regional markets. It was practically impossible for the end buyer in the market to assess risk correctly for them. This makes valuing them even more uncertain because you have to work out all the correlated and uncorrelated effects of different movements in underlying economic variables in the models.


Thirdly: CDOs with mortgages were based on historical data, not current market prices. The current Euro-Dollar exchange rate contains within it some view of what the market thinks will happen over the next 3-12 months. So there is some information in there that can be used to price a 3 month Future. There was no such equivalent with CDOs consisting of mortgages so analysis of the models were based solely on historical data of home prices and mortgage defaults which is a far more flawed system since it's extremely sensitive to what particular time periods you pick to model from, i.e. what do you consider to be likely to happen again or what do you consider to be a "once off"? Do you include data from the Great Depression or not etc?

Fourth: And possibly most importantly, rating agencies rated these products using overly simplistic models of risk (specifically models that ignored "tail risk"). The market is built on trust and market participants trusted the ratings these products were given but in reality the ratings agencies were not accurately rating the products which resulted in enormous market failure and the wiping out of billions and billions of dollars from bank and company balance sheets when the true risk of these products became evident.


CDOs were a completely different animal to the previous quantitatively heavy areas of finance like derivatives and futures and precisely because of the differences the methods didn't transfer and the market failed with the resulting credit crisis and other silliness. The underlying quantitative approach to finance isn't wrong necessarily it is just limited in application and here it was applied to something where it just doesn't work.

JDLK wrote: »

The fact that nobody predicted the finacial collapse is a pretty big kick in the crotch of any probability proponent

People have been warning that the quant models used in CDOs etc underestimated risk for years. No one listened because there was money to be made in selling the products.

JDLK

nesf wrote: »

Right, I've only as much time as until the toddler wakes up to write this so apologies for it not being a complete treatment.


Fundamentally the problem in options pricing is trying to calculate what the price of a particular financial instrument (i.e. product) should be given that it isn't directly visible. Futures are the simplest example to work with, they are just a contract to buy or sell something at a specified future date. We cannot know the actual price of this something at this date so we have to estimate it. If I'm selling a future to you that says that I will buy X bushels of wheat at the start of April I ideally want to agree a price that will be lower than the price that I expect wheat will be selling for in April on average. The core concept here is that I don't need to know exactly I just need to be able to put some rough probability on future price. Now for things like wheat it turns out to not be so bad, what is knowable is the price today of wheat and what is also knowable is the price everyone else is paying for wheat in April (because you're not the only one doing this). Quants know that the probability of prices isn't like what things like the Black-Scholes model tells us. They add in a lot of different items to account for the fact that prices aren't distributed "normally" (a particular type of probability distribution that is easy to work with mathematically and describes some kinds of physical facts like height in a population).

The end result of all this mathematics and computer time is a market that tends to work relatively well. The prices that future contracts are originally sold for are rarely absolutely correct but they get it right enough of the time for it to be profitable and useful. Farmers get a guaranteed price for their harvest and can transfer the risk (or reward) of a worse (or better) harvest than expected onto a third party for a fee. These kinds of instruments work well in commodities, foreign exchange etc precisely because a lot of companies and people are happy to get a particular type of risk off their plate in exchange for a fee. The sellers of Futures can turn a profit once they are able to make a decent guess of future prices for the goods. People have been doing this successfully since the 18th century in Japan with rice futures iirc.



The problem with CDOs and the Sub Prime crisis (with respect to the above) was this:

Firstly: The financial instruments in question had a lifetime of up to 30 years! This longer the period of time the less and less accurate any estimation of value will be. 3 month or 1 year Futures markets work precisely because the periods of the estimation are relatively short.


Secondly: An individual CDO consisted of parts of hundreds of different mortgages in different regional markets. It was practically impossible for the end buyer in the market to assess risk correctly for them. This makes valuing them even more uncertain because you have to work out all the correlated and uncorrelated effects of different movements in underlying economic variables in the models.


Thirdly: CDOs with mortgages were based on historical data, not current market prices. The current Euro-Dollar exchange rate contains within it some view of what the market thinks will happen over the next 3-12 months. So there is some information in there that can be used to price a 3 month Future. There was no such equivalent with CDOs consisting of mortgages so analysis of the models were based solely on historical data of home prices and mortgage defaults which is a far more flawed system since it's extremely sensitive to what particular time periods you pick to model from, i.e. what do you consider to be likely to happen again or what do you consider to be a "once off"? Do you include data from the Great Depression or not etc?

Fourth: And possibly most importantly, rating agencies rated these products using overly simplistic models of risk (specifically models that ignored "tail risk"). The market is built on trust and market participants trusted the ratings these products were given but in reality the ratings agencies were not accurately rating the products which resulted in enormous market failure and the wiping out of billions and billions of dollars from bank and company balance sheets when the true risk of these products became evident.


CDOs were a completely different animal to the previous quantitatively heavy areas of finance like derivatives and futures and precisely because of the differences the methods didn't transfer and the market failed with the resulting credit crisis and other silliness. The underlying quantitative approach to finance isn't wrong necessarily it is just limited in application and here it was applied to something where it just doesn't work.




People have been warning that the quant models used in CDOs etc underestimated risk for years. No one listened because there was money to be made in selling the products.

Surely the reckless behaviour outlined above just proves my point about the need for independant regulation

nesf

JDLK wrote: »

Surely the reckless behaviour outlined above just proves my point about the need for independant regulation

The regulators were already bringing in regulation to deal with the problems caused by CDOs being held off balance sheet* by banks in the proposed Basel II reforms in 2004.


The behaviour wasn't so much reckless as the left hand didn't know what the right hand was doing in the banks, which is far more worrying.


*This was an accounting trick that allowed banks to put aside very small amounts of capital to cushion against potential losses which aggravated the whole situation enormously.

JDLK

nesf wrote: »

The regulators were already bringing in regulation to deal with the problems caused by CDOs being held off balance sheet* by banks in the proposed Basel II reforms in 2004.


The behaviour wasn't so much reckless as the left hand didn't know what the right hand was doing in the banks, which is far more worrying.


*This was an accounting trick that allowed banks to put aside very small amounts of capital to cushion against potential losses which aggravated the whole situation enormously.

Here's the problem, the view that the finacial analysts can say "oh yeh we fu€ked up... but you know what???... the important things is; we know HOW we fu€ked up!!" This kind of justification through retrospection, being able to justify the validity of the past prediction of the current market is completely negated by the fact that the predicted market did not come to fruition- so what if the figures pointed to this or that, the reality is the predictions failed. This is a philisophical point of view that requires a leap for anyone within the "woods" of quantitative mathematics but the important and fundemental fact is that a market left to its own devices will not reach a natural equilibrium- maybe it would if you took the immeasurable human variable out of it but until we can measure the propensity to accumulate personal wealth we'll never accurately predict futures.

nesf

JDLK wrote: »

Here's the problem, the view that the finacial analysts can say "oh yeh we fu€ked up... but you know what???... the important things is; we know HOW we fu€ked up!!" This kind of justification through retrospection, being able to justify the validity of the past prediction of the current market is completely negated by the fact that the predicted market did not come to fruition- so what if the figures pointed to this or that, the reality is the predictions failed. This is a philisophical point of view that requires a leap for anyone within the "woods" of quantitative mathematics but the important and fundemental fact is that a market left to its own devices will not reach a natural equilibrium- maybe it would if you took the immeasurable human variable out of it but until we can measure the propensity to accumulate personal wealth we'll never accurately predict futures.

Ok, but you don't need to hold to neo-classical assumptions of markets reaching natural equilibrium on their own to believe that financial analysis can be useful. You don't need to hold to the Efficient Market Hypothesis either. Being a Keynsian or neo-Keynsian doesn't mean you reject all of financial analysis automatically.

JDLK

nesf wrote: »

Ok, but you don't need to hold to neo-classical assumptions of markets reaching natural equilibrium on their own to believe that financial analysis can be useful. You don't need to hold to the Efficient Market Hypothesis either. Being a Keynsian or neo-Keynsian doesn't mean you reject all of financial analysis automatically.

We all make predictions in our day to day lives- its simply a part of life to try to anticipate the future and educated guesses are often made, the problem in the financial markets is that predictions carry too much weight and effect values disproportinately to the vast variables which are external to mathematical quantification. This will always remain a fact but the main problem is that if, as you say, people had been predicting the finacial collapse that there was simply no mechanism to stem the spiral.

If indeed analysts knew this was coming (and I find that dubious considering the downfall of major financials)it is only less of a comfort that a system knows it is heading for disaster but cannot steer itself away. Almost like an overeater knowing they are going to die of obesity but simply cannot stop eating

Keyensian or not I believe mathematical probablities should be a part of future predictions not the whole- the shift toward over reliance on probabilities even when they are successful is the issue. When analysts are screaming to invest becuase they are 76% certain there will be a 6% return we need a mechanism to offset the mathematics with possible human risks and thus create a new probablity- that may be pie in the sky but there should at least be a mechanism to distinguish between inflated value and "real value".

But I suppose we have come full circle now from semantics to philosophy.

nesf

JDLK wrote: »

Keyensian or not I believe mathematical probablities should be a part of future predictions not the whole- the shift toward over reliance on probabilities even when they are successful is the issue. When analysts are screaming to invest becuase they are 76% certain there will be a 6% return we need a mechanism to offset the mathematics with possible human risks and thus create a new probablity- that may be pie in the sky but there should at least be a mechanism to distinguish between inflated value and "real value".

What's "real value" and do you accept my point on probability being useful in Futures markets et al without destabilising the rest of economic system?

JDLK wrote: »

But I suppose we have come full circle now from semantics to philosophy.

Yup.

JDLK

nesf wrote: »

What's "real value" and do you accept my point on probability being useful in Futures markets et al without destabilising the rest of economic system?


Yup.

My main point about "real value" is a philosophical view point (of which iIbelieve Keyens was a proponent)- that uncertainty can never be quantified and that all attempts to turn uncertainty into "risk" and therefore attach values to it -is moot in a situation of total system failure.

Saying the current finacial system is mathematically sound is moot as it lays in a smoldering mess. Its like communists saying that Marx is right but when you put it into practice it falls apart.

Like I said I think there is certainly a place for probabilities but it is the variables which are accounted for that are the issue- the current situation must prove that some variable was not included or was ignored, or undervalued. This requires you step outside the quantitative process and gain a more holistic view.

A view, in my opinion which must take into account the inequalities in gloabalisation. I beleive the quantification of probabilities at the national to international level (ie domestic product and export/import markets) are not the same when creating global stratgeies such as positioning manufacturing in offshore regions which are not providing fair/free trade. This then skews the redistribution of currency which is a future variable beyond the intital cost savings, yet is just as importnat because short term gains are not sustainable and the new conditions created therefore have a direct effect on future probablilites

nesf

JDLK wrote: »

My main point about "real value" is a philosophical view point (of which iIbelieve Keyens was a proponent)- that uncertainty can never be quantified and that all attempts to turn uncertainty into "risk" and therefore attach values to it -is moot in a situation of total system failure.

And your basis for uncertainty not being quantifiable is? (I'm not unsympathetic to this view, I just this this is a very broad and strong assertion)

JDLK wrote: »

Saying the current finacial system is mathematically sound is moot as it lays in a smoldering mess. Its like communists saying that Marx is right but when you put it into practice it falls apart.

I never said it was mathematically sound as a whole anywhere. I said that some markets seem to work fine with quantitative techniques, such as most Futures and Derivatives markets.

JDLK wrote: »

Like I said I think there is certainly a place for probabilities but it is the variables which are accounted for that are the issue- the current situation must prove that some variable was not included or was ignored, or undervalued. This requires you step outside the quantitative process and gain a more holistic view.

You're assuming that the quantitative process cannot do this. Why do you believe this?

JDLK

nesf wrote: »

And your basis for uncertainty not being quantifiable is? (I'm not unsympathetic to this view, I just this this is a very broad and strong assertion)



I never said it was mathematically sound as a whole anywhere. I said that some markets seem to work fine with quantitative techniques, such as most Futures and Derivatives markets.



You're assuming that the quantitative process cannot do this. Why do you believe this?

The view that uncertainty is unquantifiable comes from the very basic viewpoint that the current situation was not predicted (though I know you say it was). This uncertainty (the shock of finacial collapse) is tackled by both classical and keyensian economists- classical view it as a managed risk an ability to attach a level or value to uncertainty so that we can be X% certain that this will happen- but it ignores the very real (as we can see) possibility of all consuming unpredictability- instead it rationalises it as the flip side to probablity ie if we are X% certian that this will happen then the remaining % is allocated to the possibility that it will not happen- but the possibilities of something not happening are not necessarily equal to the remainder in % of the proabailty that they will happen. Again this is reinforced by the lack of quantification in the proabability of the current situations previous finacial predictions. How do we value uncertainty- is it the probability that something will happen or is it the possible damage of something if it does/doesnt happen?

My point is that it is not being quantified (whether possible or not)- or if it is it is being recklessly ignored

Was the possibility of finacial collapse added as a variable when Anglo or AIg or Lehmann analysts were predicting futures and advising clients?

on the one hand we hear analysts saying- "nobody could have predicted this" but then on the other hand they are telling us they are so good at predicting futures that we should entrust our savings to them- there is a disparity hear which cannot be ignored nor can it be brushed off. Analysts cannot have ot both ways

nesf

JDLK wrote: »

The view that uncertainty is unquantifiable comes from the very basic viewpoint that the current situation was not predicted (though I know you say it was). This uncertainty (the shock of finacial collapse) tackled by both classical and keyensian economists- classical view it as a managed risk an ability to attach a level or value to uncertainty so that we can be X% certain that this will happen- but it ignores the very real (as we can see) possibility of all consuming unpredictability- instead it rationalises it as the flip side to probablity ie if we are X% certian that this will happen then the remaining % is allocated to the possibility that it will not happen- but the possibilities of something not happening are not necessarily equal to the remainder in % of the proabailty that they will happen. again this is reinforced by the lack of quantification in the proabability of the current situation predication previously.

Maybe it is quantifiable however my point is that it is not being quantified- or if it is it is being recklessly ignored

I'm asking a more fundamental question than that, why do you think that uncertainty in economic systems is unquantifiable and do you this it is totally unquantifiable or that it can only be partially quantified? These are very fundamental and broad ranging questions that underlay what you're asserting above.


As an important semantic point, I didn't say that this specific situation was predicted, I said that the underlying problems that turned this from a small mess into a big mess were known and people were trying to fix them prior to this blowing up in everyone's face.

JDLK

nesf wrote: »

I'm asking a more fundamental question than that, why do you think that uncertainty in economic systems is unquantifiable and do you this it is totally unquantifiable or that it can only be partially quantified? These are very fundamental and broad ranging questions that underlay what you're asserting above.


As an important semantic point, I didn't say that this specific situation was predicted, I said that the underlying problems that turned this from a small mess into a big mess were known and people were trying to fix them prior to this blowing up in everyone's face.

I added some more to that last post. Again Im not saying that uncertainty is not being quantified but that the value of mathematical probability is actually more than its real value, so when something is X% mathematically probable it is in actuality less than that.

If I could actually work out this disparity i would be the new Keyens, but it is defintely there because the futures mathematical proabilites of say 10 years ago for 2009 did not come to fruition- this proves that something was not quantified, proving that the probabilities were in fact false (or incomplete). I do believe it can be worked out though be revisiting the probabilities and then adding what we now know as the real values. This is retrospective but I believe it can give us at least a starting point to view the disparity.

this might lead to a "best guess" scenario but it is still shaky because the best guesses have proved to be wrong, but then this argument i suppose is easily negated by saying that "oh well we all knew this was going to happen", its just a pity nobody told the rest of us then (of course then it could be said they did but we ignored them)

nesf

JDLK wrote: »

I added some more to that last post. Again Im not saying that uncertainty is not being quantified but that the value of mathematical probability is actually less than its real value, so when something is X% mathematically probable it is in actuality less than that.

If I could actually work out this disparity i would be the new Keyens, but it is defintely there because the futures mathematical proabilites of say 10 years ago for 2009 did not come to fruition- this proves that something was not quantified, proving that the probabilities were in fact false (or incomplete).

this might lead to a "best guess" scenario but it is still shaky because the best guesses have proved to be wrong, but then this argument i suppose is easily negated by saying that "oh well we all knew this was going to happen", its just a pity nobody told the rest of us then (of course then it could be said they did but we ignored them)

That doesn't answer my question though. There are three basic options to choose to subscribe to (I'll use future prices as the example but you can use any economic quantity):

1) Future prices are knowable and we can and do predict them absolutely given all the correct data.

2) Future prices are knowable but at best we can only predict a range of possible values and the probability of the price being one of these values at a certain date, given all the correct data.

3) Future prices are unknowable and cannot be predicted even in the sense of predicting a range of possible values even if we are given all the correct data.


Which out of 1, 2 or 3 best describes what you believe and why do you believe this over the other two options? Note: None of the above options says whether we can actually access or tell what all the correct data is, that is a separate philosophical question here.

JDLK

nesf wrote: »

That doesn't answer my question though. There are three basic options to choose to subscribe to (I'll use future prices as the example but you can use any economic quantity):

1) Future prices are knowable and we can and do predict them absolutely given all the correct data.

2) Future prices are knowable but at best we can only predict a range of possible values and the probability of the price being one of these values at a certain date, given all the correct data.

3) Future prices are unknowable and cannot be predicted even in the sense of predicting a range of possible values even if we are given all the correct data.


Which out of 1, 2 or 3 best describes what you believe and why do you believe this over the other two options? Note: None of the above options says whether we can actually access or tell what all the correct data is, that is a separate philosophical question here.

Im going to go with option 3 here as I cant really understand option 2, for me if something is unknowable then it is unknowable and best guesses are moot in a clinical environment- i believe this because when we attach a probability to something and it does not happen the actual probability of it not happening does not match the residual % in the probability of it happening (in my opinion), however option 3 poses a problem in that it says "even if we are given all correct data"- this I believe is the impossibility which is core to my argument - which pretty much negates all options

I think that it is too easy to get lost in the quantitative woods, the reality of the current situation vehemently rebels against any logical probability prediction of the past, I believe within the gap of the mathematical proability and the cold reality of todays situation is not just a variable which has not been quantified and undervalued but that unpredictability itself is a variable with its own value, because unpredictability effects human behaviour- most notably confidence- which has a very real economic value

As mentioned earlier the two thoughts on this is that we can manage long term risk (classical) or we cannot and at best we can react to dynamic markets(Keyenes)- I would be closer to Keyens and I think in the common sence view of the economic situation it becomes apparant that all predictions were undermined. though i think a value can be attached to uncertaintyl but only retrospectively

nesf

JDLK wrote: »

Im going to go with option 3 here as I cant really understand option 2, for me if something is unknowable then it is unknowable and best guesses are moot in a clinical environment- i believe this because when we attach a probability to something and it does not happen the actual probability of it not happening does not match the residual % in the probability of it happening (in my opinion), however option 3 poses a problem in that it says "even if we are given all correct data"- this I believe is the impossibility which is core to my argument - which pretty much negates all options

I think that it is too easy to get lost in the quantitative woods, the reality of the current situation vehemently rebels against any logical probability prediction of the past, I believe within the gap of the mathematical proability and the cold reality of todays situation is not just a variable which has not been quantified and undervalued but that unpredictability itself is a variable with its own value, because unpredictability effects human behaviour- most notably confidence- which has a very real economic value

I'll try and explain 2 better.

1) is a deterministic system, i.e. all physical processes in the world above the quantum level. If you have perfect knowledge you can predict any physical event. You can't always have perfect knowledge which is why it's very complicated to predict anything more than a very simple system exactly.

2) is a probabilistic system where there are random effects but these are systematic and predictable. The classic example is quantum mechanics, quantum effects are random but the probability functions describing this randomness are knowable. This means that we cannot ever know the exact price of a future good but we can know the probability distribution describing it's possible values if we have perfect knowledge.

3) is a completely indeterministic system where even if you know everything you can predict nothing.


3) is honestly nonsensical because it implies that economic movement is entirely random. The Foreign Exchange Futures market for instance is very predictable. Covered Interest Parity pretty much determines the price of a future contract.

JDLK

nesf wrote: »

I'll try and explain 2 better.

1) is a deterministic system, i.e. all physical processes in the world above the quantum level. If you have perfect knowledge you can predict any physical event. You can't always have perfect knowledge which is why it's very complicated to predict anything more than a very simple system exactly.

2) is a probabilistic system where there are random effects but these are systematic and predictable. The classic example is quantum mechanics, quantum effects are random but the probability functions describing this randomness are knowable. This means that we cannot ever know the exact price of a future good but we can know the probability distribution describing it's possible values if we have perfect knowledge.

3) is a completely indeterministic system where even if you know everything you can predict nothing.


3) is honestly nonsensical because it implies that economic movement is entirely random. The Foreign Exchange Futures market for instance is very predictable. Covered Interest Parity pretty much determines the price of a future contract.

I still dont agree with 2 (though im sure its the best we can do to date) I think this is moving into the area of chaos theory

and 3 is only nonesensical because of the added "entirely", I dont beleive they are entirely random however i do believe there is a certain randomness which is undervalued- it is this randomness which can be valued retrospectively to explain the difference between past quantitative probabilities and current realities

nesf

JDLK wrote: »

I still dont agree with 2 (though im sure its the best we can do to date) I think this is moving into the area of chaos theory

and 3 is only nonesensical because of the added "entirely", I dont beleive they are entirely random however i do believe there is a certain randomness which is undervalued- it is this randomness which can be valued retrospectively to explain the difference between quatitative probabilities and current realities

Remember, 2 involves perfect knowledge. The current risk could be undervalued in either 1 or 2 due to incomplete knowledge. By the way, chaos theory (i.e. non-linear mathematics) would be Option 1). There is no pure randomness in chaotic systems, they are actually fully deterministic. They're just sensitive to initial conditions (i.e. require you to have very very accurate knowledge in order to make accurate predictions).

JDLK

nesf wrote: »

Remember, 2 involves perfect knowledge. The current risk could be undervalued in either 1 or 2 due to incomplete knowledge. By the way, chaos theory (i.e. non-linear mathematics) would be Option 1). There is no pure randomness in chaotic systems, they are actually fully deterministic. They're just sensitive to initial conditions (i.e. require you to have very very accurate knowledge in order to make accurate predictions).

The argument here is that the current situation came from a mistake in the probabilites but that the probabilites are sound given all the available data

but I think that probability risk is always overvalued, even when all variables/data are present and that uncertainity remians a dynamic value made up of sociological factors which do not yield to quantification.

This may not sound nice to someone who wishes to use a logical, transparent framework to accurately predict futures but how else can one explain the disparity that exists in todays situation with past probabilities?

Even if we find a convenient specific explanation that something went wrong in the data collection process or human error, the possibility of this in itself must be a variable of uncertainity- the measure of "what if". Even if we reduce the possibility of such an action happening again we are still left with the gap of "what if" the possibility/proabability of that which is beyond our logical prediction- this has to be the explanation for the difference between the analysis of the past and the reality of today- and if probability theories remain sound even if their users remain flawed then the same holds true for the uncertainty of what is beyond our comprehension

Basically there has to be a gap between the futures predictions of the recent past and the current economic realities this sounds obvious BUT this is not simply the residual %'s because if it were then we would have been able to predict the current situation. Instead the actual or real proababilty contains a gap of uncertainty which must have been alot larger than the % residual- it absolutley has to be becuase if it didnt then we wouldnt be in this situation

I have to go to the blackboard and work this out- Im sure there is a gap here- if you detach yourself from the quantitative explantions it becomes obvious that a gap exists which isnt being quantified

nesf

JDLK wrote: »

The argument here is that the current situation came from a mistake in the probabilites but that the probabilites are sound given all the available data

This sentence actually makes no sense to me. I think you mean to say models instead of probabilities but I could be wrong.

JDLK

nesf wrote: »

This sentence actually makes no sense to me. I think you mean to say models instead of probabilities but I could be wrong.

Sorry yes, I mean the the probabilities were wrong but the models were sound.

The models ARE sound mathematically, but viewing them mathematically only serves to justify the logic of using them even in the face of the system collapse of the system which they failed- this means there has to be a "soft" factor at play.

The whole purpose of probablities is to predict the future so the "soft" factor missing is of no use to any of its proponents as it can only be valued retrospectively by which time it is uselss to those managing risk BUT I beleive if we can get a probablity from the past and then recreate it with the current variables the gap between the 2 must represent the "real" uncertainty value- which has to be much higher than the residual probabality and MUST be high enough to predict the financial collapse- this isnt even mathematics its just common sence and it actually fits into the theory of number 2 in that it provides a logical reason for why the system failed.

If we take the point of view that the current probability models are sound but couldnt predict the current situation then this view surely fits in with the idea of complete randomness number 3- the view that all predictions are useless- because all predictions were useless otherwise we would have avoided the current situation.

nesf

JDLK wrote: »

Sorry yes, I mean the the probabilities were wrong but the models were sound.

The models ARE sound mathematically, but viewing them mathematically only serves to justify the logic of using them even in the face of the system collapse of the system which they failed- this means there has to be a "soft" factor at play.

The whole purpose of probablities is to predict the future so the "soft" factor missing is of no use to any of its proponents as it can only be valued retrospectively by which time it is uselss to those managing risk BUT I beleive if we can get a probablity from the past and then recreate it with the current variables the gap between the 2 must represent the "real" uncertainty value- which has to be much higher than the residual probabality and MUST be high enough to predict the financial collapse- this isnt even mathematics its just common sence and it actually fits into the theory of number 2 in that it provides a logical reason for why the system failed.

If we take the point of view that the current probability models are sound but couldnt predict the current situation then this view surely fits in with the idea of complete randomness number 3- the view that all predictions are useless- because all predictions were useless otherwise we would have avoided the current situation.

But my post about CDOs was about how they used the wrong models and the wrong assumptions in their models. So even if we lived in a world of 1) or 2) they would still have been wrong.

We aren't forced to accept 3) because they got it wrong which is what I'm pointing out. It's a bit like throwing the baby out with the bathwater so to speak.

JDLK

nesf wrote: »

But my post about CDOs was about how they used the wrong models and the wrong assumptions in their models. So even if we lived in a world of 1) or 2) they would still have been wrong.

We aren't forced to accept 3) because they got it wrong which is what I'm pointing out. It's a bit like throwing the baby out with the bathwater so to speak.

Regardless of the specific reasons in this specific scenario for failure, the fact that we have the ability to get things wrong is in itself a variable which isnt being measured properly- the CDO case proves it becuase the probability that CDO's would use the wrong models was not valued high enough, in fact it may not have even been considered at all originally, indeed it was assumed they would use the right values so the value attached to CDO process failure was low.

To me this fits in exactly with the core of what I mean- mathematics was use dto prove what we wanted to see, but we didnt rise above it and question which mathematics we were using and what we were judging- we made assumptions that ended with a CDO giving high probabilities but because of the lack of one (significant) variable the proababilities were rendered useless. Imagine if a CDO is predicting 3% GDP growth with a 75% certainity and 25% uncertainty but because they are using the wrong models the actual or "real" uncertainty is almost 100% which we can logically say at the very least the GDP growth probability is severley flawed

To say that we got things right but it was jusyt so and so's fault that it all fell apart I think is too convenient, especially when dealing with the gloabl economy. A football manager could blame an injury to his star striker for not winning the cup final but it doesnt really matter because he has still lost the cup- we cant adopt the same "well it would have worked if XYZ didnt happen" approach, in a way Im sticking up for predictions because Im saying that we need to be more accurate, that we need to look beyond the models that failed us, to see if we failed the models and ourselves. We dont have to get rid of any models but we do need to analyse how we are using them and the value we are attaching to them

nesf

I'm curious, how much mathematical modelling have you been exposed to?

JDLK

nesf wrote: »

I'm curious, how much mathematical modelling have you been exposed to?

See, I think that questions of credibility do not serve a discussion. It usually only serves to arbitrarily dismiss other people views.

When I did my masters degree we learned about people who broke through academic ceilings by challenging the mainstream beliefs- but the big irony was that once their views became mainstream students refused to challenge them. Academic models should be constantly challenged as new environments come about and nothing should ever be accepted as gospel- that attitude leads to academic stagnation and Im sure was at the forefront of the minds of CDO's appliying bad practices

nesf

JDLK wrote: »

See, I think that questions of credibility do not serve a discussion. It usually only serves to arbitrarily dismiss other people views.

When I did my masters degree we learned about people who broke through academic ceilings by challenging the mainstream beliefs- but the big irony was that once their views became mainstream students refused to challenge them. Academic models should be constantly challenged as new environments come about and nothing should ever be accepted as gospel- that attitude leads to academic stagnation and Im sure was at the forefront of the minds of CDO's appliying bad practices

It's not a credibility thing, I'm only asking because the attitude you describe doesn't fit with the kind of people I've met in the empirical side of academic mathematical modelling in economics and finance who routinely reject mainstream beliefs and try other ideas when they don't fit with the data. The people at the coalface of social science disciplines (i.e. testing the theories against the data) tend to be the least likely to consider the theory as gospel because every day they see how it doesn't square up exactly or even close sometimes.


Edit: I can see how someone might think that of quants after reading the headlines of their misbehaviour in the markets over the past few months but I can assure you that among academics interested in modelling reality you'll tend to find little dogmatism about theory because, as they say, the data is king.

JDLK

nesf wrote: »

It's not a credibility thing, I'm only asking because the attitude you describe doesn't fit with the kind of people I've met in the empirical side of academic mathematical modelling in economics and finance who routinely reject mainstream beliefs and try other ideas when they don't fit with the data. The people at the coalface of social science disciplines (i.e. testing the theories against the data) tend to be the least likely to consider the theory as gospel because every day they see how it doesn't square up exactly or even close sometimes.


Edit: I can see how someone might think that of quants after reading the headlines of their misbehaviour in the markets over the past few months but I can assure you that among academics interested in modelling reality you'll tend to find little dogmatism about theory because, as they say, the data is king.

Business Studies is my major background (masters level)- which I suppose tends to attempt to strike a balance between theory and application.