Do you have an economic background?

nesf

Permabear wrote: »

This post had been deleted.

Krugman was talking about a very serious problem in neoclassical economic's assumptions about behaviour. I don't contest this*. But, considering there are many, many alternative branches of economics tackling different ways to approach modelling the problem I hardly think that it invalidates mathematical modelling as a whole within the discipline since it is not representative as a whole. It would be the same as me dismissing the entirety of work within Austrian economics because of the approach of Mises alone.


Edit:

*Mainly because from day one in my studies of economics these very problems with the neoclassical model were pointed out to me and there is a very big body of research within the discipline going back decades showing why the neoclassical assumptions don't hold. Beyond a few diehards it's extremely hard to find practicising economists who actually believe people behave rationally in the neoclassical sense.

nesf

Scofflaw wrote: »

That level in economic models seems to consist instead of a series of untested assumptions (slightly unkind, perhaps, but that how it seems to a field scientist) which economists in general are happy to accept because they're only interested in the big picture anyway.

Overgeneralise much? That's like me criticising all climate scientists because some of them have taken shortcuts in their modelling that are frankly insane from a mathematical point of view and yet ignore all the people working on correcting these shortcuts. Ditto weather modelling.

No one here is arguing that current modelling in economics is perfected or even very accurate. What's being argued is that giving up on modelling complex phenomena because they're complex is foolish.

Scofflaw

nesf wrote: »

Overgeneralise much? That's like me criticising all climate scientists because some of them have taken shortcuts in their modelling that are frankly insane from a mathematical point of view and yet ignore all the people working on correcting these shortcuts. Ditto weather modelling.

Fair enough - I have indeed overdone the criticism, but it is the impression one gets as a scientist. I'd have to point out that climate scientists don't make the kind of claims of accuracy made by economic modellers - climate scientists are currently very cautiously beginning to offer limited regional predictions, for example, hedged around with a lot of caveats, whereas the latest IMF report on Ireland is full of quite detailed discussions of improbably precise predicted outcomes for, for example, Ireland's debt trajectory.

nesf wrote: »

No one here is arguing that current modelling in economics is perfected or even very accurate. What's being argued is that giving up on modelling complex phenomena because they're complex is foolish.

I wouldn't disagree at all. After all, what constitutes a 'complex system' has moved nearly as much as what constitutes 'artificial intelligence', and for the same reasons.

cordially,
Scofflaw

nesf

Scofflaw wrote: »

Fair enough - I have indeed overdone the criticism, but it is the impression one gets as a scientist. I'd have to point out that climate scientists don't make the kind of claims of accuracy made by economic modellers - climate scientists are currently very cautiously beginning to offer limited regional predictions, for example, hedged around with a lot of caveats, whereas the latest IMF report on Ireland is full of quite detailed discussions of improbably precise predicted outcomes for, for example, Ireland's debt trajectory.

I'm not particularly interested in the IMF here tbh. I'm more interested in modelling as an end in itself and whether we can model human behaviour at all.

I'm not defending the end use of current models.

Scofflaw

nesf wrote: »

I'm not particularly interested in the IMF here tbh. I'm more interested in modelling as an end in itself and whether we can model human behaviour at all.

I'm not defending the end use of current models.

Ah - well, that was what I was opposing, not modelling as an end in itself.

cordially,
Scofflaw

nesf

Scofflaw wrote: »

Ah - well, that was what I was opposing, not modelling as an end in itself.

cordially,
Scofflaw

Well the argument is over the place of empiricism in economics.

Scofflaw

nesf wrote: »

Well the argument is over the place of empiricism in economics.

But the argument being put forward to support the claim that empiricism has no place in economics is the unreliability of current economic models...

cordially,
Scofflaw

nesf

Scofflaw wrote: »

But the argument being put forward to support the claim that empiricism has no place in economics is the unreliability of current economic models...

cordially,
Scofflaw

That's akin to arguing modelling has no place in weather because it can't tell you what the weather will be like 2 weeks from now with great accuracy.

Permabear

This post has been deleted.

nesf

Permabear wrote: »

This post had been deleted.

Well first off economics is a combination of qualitative and quantitative techniques. There are thing that are not tractable mathematically and different tools are used for this.

An example:

If we want to know how much a difference wealth makes to happiness we combine the two techniques. So we use psychological techniques to measure well-being and happiness and quantitative techniques to measure wealth and model wealth. What's found is that above a certain wealth level (around 30K a year) happiness is pretty static by these measures. Below that level happiness is negatively affected. This is modelling and investigating something intractable mathematically but economics can still investigate it.

Without empirical studies here this question cannot be answered satisfactorily. This is exactly what I'm talking about when I'm talking about the role of empiricism in economics, without going out there and measuring happiness at various wealth levels we cannot with any accuracy talk about how wealth and happiness are related.

This result is fairly robust across a range of different countries too.

andrew

I read an article a while ago, I wish I could find it. I think it by Von Neumann, or maybe it just referenced his work a lot. Anyway, it talked about economics as a science and, by extension, the power of modelling. The general point was that economics is still a young science (or 'science') if you prefer. It's only been around formally for around 200 odd years, and the idea of economic modelling has been around for only a fraction of that time. In contrast, the 'hard' sciences have been around for twice, or three times as long as that. And in the case of the hard sciences, it also used to be thought that some things were irreducibly complex, and impossible to model. But beginning with Newton,* and through centuries of observation and testing, the patterns and general rules which determine the physical world began to take shape and emerge, such that a mathematically sound picture of the universe has been largely created. Now the article didn't say that the same complete picture was open to economics, but the point is that we're only just beginning to conceive of human action as something which can perhaps be modelled and simplified. It's far too early to write off modelling as impossible or ultimately fruitless, and it'll be maybe centuries before we really get a grip on how it should be modeled.

What kind of models will obtain at the end? Well, even in the perfect case of Physics, there's a lot of uncertainty. The weather can be predicted, but not with a lot of certainty. And that's without the problem of models whose units are aware they're behaving like a model and can then cease to behave in the way the model thinks they will behave. But perhaps at some stage in the future economics will be able to simplify aggregate human actions in ways which can't even conceive yet, in the same way that the physical modelling done today by supercomputers would be beyond the comprehension of scientists from the 1600's.

*well, I think it was Newton. I'm open to correction.

Solair

Studied public finance and economics.

Permabear

This post has been deleted.

nesf

Permabear wrote: »

This post had been deleted.

It's similar problem really. What changes in economics over time is the speed of change not really the type of change itself. IT has resulted in information spreading faster than ever before but when you get down to it it's the same kind of reactions to information that we're getting that we got 50 or 100 years ago. Most of the basic elements of economic modelling were present centuries ago, we're just dealing with them on a much bigger and much faster moving scale now.

You can apply modern network theory to any point in human history for example, what's different is the speed of propagation within the network of information. We have new technologies sure but have they really added to the fundamentals of the market and behaviour or have they just made alterations like information propagation speed like I talked about above?

SupaNova

nesf wrote: »

More I'm arguing that you don't know what you're arguing against here.

I'm ignorant of modelling, but since you have had long conversations about it, can you give me some brief description of these models, if that's not possible, can you at least tell me some of the aggregate statistics used in these models?

Sure. And it came from a fundamental mistake in his assumptions. But empirical evidence proved him wrong and economics learned from the mistakes and didn't go on making those same assumptions. i.e. empiricism was useful here, actually necessary here, and you're arguing it doesn't have a place!

I haven't argued that empiricism is of no use at all, but defended that in economics you can theorize by making deductions from true statements. I am not really arguing whether the Austrian School is great at everything or whether all their statements and deductions are correct, only that their method has merit. If i can get you to agree that i at least gave one example of that, you can agree that the Austrian Method is valid and useful.

nesf

SupaNova wrote: »

I'm ignorant of modelling, but since you have had long conversations about it, can you give me some brief description of these models, if that's not possible, can you at least tell me some of the aggregate statistics used in these models?

Most of my interest is in the mathematics and I've a bad memory for details and these conversations were three years ago!

The approach of professionals in this area is one of "try anything and see if it gives reasonable results", guided by theory obviously. So we have them trying out mathematical techniques from applied maths, climate modelling etc for dealing with complex systems and applying them. If they don't give sensible results the models are thrown away. They don't restrict themselves to the kind of stuff you'll see in economics class and they definitely don't restrict themselves to the prevalent neoclassical approach found in mainstream economics. They also don't have one model of inflation, they have many and combine the results of these to get some kind of idea of the expected range of inflation going forward. Definitely far from perfect but it's a pretty tried and true method in modelling across all the sciences.

SupaNova wrote: »

I haven't argued that empiricism is of no use at all, but defended that in economics you can theorize by making deductions from true statements. I am not really arguing whether the Austrian School is great at everything or whether all their statements and deductions are correct, only that their method has merit. If i can get you to agree that i at least gave one example of that, you can agree that the Austrian Method is valid and useful.

My issue is with the idea of having true statements that aren't tautologies that you cannot test empirically. I mean, take the rationality set of assumptions from neoclassical economics, we can test all these using small scale experiments and show many flaws in said assumptions. I really doubt you can actually make such a statement that is useful without it being in some way at least indirectly testable.

ILikeBananas

Permabear wrote: »

This post had been deleted.

No, but we don't have to. The kinetic movement of individual gas molecules in a vessel are inherently random but when taken in bulk, order emerges which can be modelled effectively.

Likewise, we're never going to be able to create economic models that take into account individual free will. We can and will however see models based on the wisdom of crowds.

nesf

ILikeBananas wrote: »

The kinetic movement of individual gas molecules in a vessel are inherently random but when taken in bulk, order emerges which can be modelled effectively.

Indeed, if you try and model the behaviour of individual molecules within a gas for a volume of reasonable size you end up with a problem that not all the computers on earth could solve in our lifetime. Also relevant is the three body problem and other such intractabilities.

SupaNova

Nesf how big a part does quantitative prediction play in adjusting these models? And is better quantitative prediction an explicit aim of economic modelling as is the case with modelling the weather or kinetic movement of gas molecules?

nesf

SupaNova wrote: »

Nesf how big a part does quantitative prediction play in adjusting these models? And is better quantitative prediction an explicit aim of economic modelling as is the case with modelling the weather or kinetic movement of gas molecules?

In adjusting two main things are looked at: Does the model accurately predict the future in the past (i.e. feed in data up to 1995 and see if it predicts 1996 well, or up to Q1 2011 and see if it predicts Q2 2012 well, and then repeat this for the entire dataset) and then if it passes this test start looking at known "break points" in the data where something changed in the economy that causes some models that worked before the break to no longer work after the break and so on. The second thing that's done is taking that model and applying it to different countries as above to see if it is a good general model or whether it favours certain kinds of countries over others.

If it passes all these then it would be added into the mix but its results wouldn't be taken that seriously for a good long while until it's been shown to give good predictions (or more accurately prediction ranges) for a certain amount of time.


Better prediction is always an aim in any kind of modelling in any area that does it. But there is a synergy here, usually models start hitting walls where previous theory no longer explains what's going on and at that point the ball is tossed back to the theorists to explain what's showing up in the data and they come back with something new and the cycle begins again.

Day-to-day science is very iterative. Small steps etc. Every so often some big discovery is made that dramatically furthers the science and its ability to predict and explain certain phenomena but this doesn't happen very often and usually takes a good while for this new method to be accepted. So while you have mainstream guys working away on current models you also have heterodox guys trying bizarre new approaches to try and find that elusive breakthrough.

I hope that answers your question. Bear in mind that I did not work in this job so my knowledge is second hand at best. Where I was doing research a few years ago was extremely heterodox, i.e. I could count the number of people doing it in the world with my fingers.

SupaNova

I think there is a key difference between predicting the weather and economic activity. In predicting future weather conditions it seems possible to give more and more accurate and exact predictions. Predicting ever more exact prices and dates of a crash seem completely illusory. For example if a model could predict stocks will crash in February by 30%. As soon as market participants had knowledge of this, they would look to get out of stocks in January, more cautious people would look to get out of stocks in December, and this would bring the onslaught of a crash much sooner, making the original prediction wrong. Lets say a higher power gave us a model that could predict everything exactly; as soon as people would see these predictions they would change their behavior. This problem doesn't exist when predicting the weather as we don't have the power to change weather conditions we don't like.

Scofflaw

SupaNova wrote: »

I think there is a key difference between predicting the weather and economic activity. In predicting future weather conditions it seems possible to give more and more accurate and exact predictions. Predicting ever more exact prices and dates of a crash seem completely illusory. For example if a model could predict stocks will crash in February by 30%. As soon as market participants had knowledge of this, they would look to get out of stocks in January, more cautious people would look to get out of stocks in December, and this would bring the onslaught of a crash much sooner, making the original prediction wrong. Lets say a higher power gave us a model that could predict everything exactly; as soon as people would see these predictions they would change their behavior. This problem doesn't exist when predicting the weather as we don't have the power to change weather conditions we don't like.

Something perhaps worth pointing out is that weather prediction cannot go past a certain point, because weather is a chaotic system - that is, it contains a lot of feedback loops, which means that initially tiny errors in your data blow up fairly rapidly into huge errors in your predictions.

Climate prediction, on the other hand, is a lot easier, because you're attempting to predict not the weather on any given day but the general pattern of weather over longer periods - if you like, the general framework within which each individual days weather will happen.

Economic modelling, it seems to me, suffers the same limitation in respect of the kind of prediction you've suggested - that of a market crash of size x on a specific day. Apart from the problem you've raised - that "markets" consist of self-aware agents capable of factoring predictions into their behaviour - markets are also chaotic.

So the proper use of economic models, however quantitative, is more similar to that of climate models rather than weather forecasting - and even that can suffer from the problem of self-aware agents. Interestingly, the problem of self-aware agents seeking to take advantage of the predictions of economic models actually grows as models improve, because the greater the confidence which can be placed in the predictions, the more agents will seek to extract advantage from the prediction, either reinforcing it or acting against it. There might be, therefore, a built-in limit to the reliability of economic modelling as you suggest - as long as the predictions of the modelling are public knowledge.

cordially,
Scofflaw

nesf

SupaNova wrote: »

I think there is a key difference between predicting the weather and economic activity. In predicting future weather conditions it seems possible to give more and more accurate and exact predictions. Predicting ever more exact prices and dates of a crash seem completely illusory. For example if a model could predict stocks will crash in February by 30%. As soon as market participants had knowledge of this, they would look to get out of stocks in January, more cautious people would look to get out of stocks in December, and this would bring the onslaught of a crash much sooner, making the original prediction wrong. Lets say a higher power gave us a model that could predict everything exactly; as soon as people would see these predictions they would change their behavior. This problem doesn't exist when predicting the weather as we don't have the power to change weather conditions we don't like.

You cannot predict crashes. You cannot predict booms. With accuracy. This is impossible due to their nature. What you can look at is more general things like how often major events occur and how to work out probability distributions to account for this in making predictions. At best we can say there is x% chance next month of a stock market movement beyond size y. No one with any sense claims any more than this. You can also study booms and busts to see what factors are common in them and by this get warning signs of booms and busts in the future, yet still not be able to accurately predict them.

This doesn't render modelling this kind of stuff useless however, you just need to account for the things you can't model and adjust expectations accordingly of what the models are capable of doing. The latter is a serious failing in politicians.

nesf

Scofflaw wrote: »

So the proper use of economic models, however quantitative, is more similar to that of climate models rather than weather forecasting

This is pretty much it for macro-economics. Work on a big enough scale and with large enough time segments that a lot of the day-to-day chaos is reduced to something more manageable. There's a hard limit on the resolution of these models (you can't drill down beyond a certain point, so for instance you could have a good model for say Dublin going forward to 2013 but it might not be able to say anything about how a part of Dublin might do during that period).


Even if we had a super accurate modeller of individual behaviour we wouldn't be able to model countries with it because of the level of computation that would be involved. There's a reason why, despite having an excellent understanding of small scale weather behaviour that weather forecasting is done on a big scale dividing that scale into very large areas (i.e. low resolution). Otherwise by the time you started doing your prediction the weather would have happened before you got a result out of your model.

This is really nuts and bolts stuff and isn't really discussed outside of the modelling community to be honest but computation power puts a hard limit on what's possible even with perfect models, never mind imperfect ones.


Edit: As an aside that I think ye might find interesting, there's this thing called the Three Body Problem in physics. Basically, despite gravitation being extremely well understood, you cannot calculate the exact position of three bodies that are moving and who gravitationally influence each other. You can only solve it if one of the bodies is not moving. And this is only with three objects interacting in one way! Imagine the problem of a million people all interacting with each other and you get an idea why a micro model doesn't scale up into something usable and why this modelling can be so complex.

Permabear

This post has been deleted.

nesf

Permabear wrote: »

This post had been deleted.

My point isn't that there hasn't been massive structural change in economies over the past 200 years, that would be a nonsense argument. What I'm arguing is basic human behaviour is still the same since we're genetically still the same creatures and still have minds evolved for life on the Savannah. We find the same behaviour in experiments in primitive societies and advanced modern democracies (see for instance the Ultimatum Game in game theory*). This indicates that some things don't change and these things we can work models off that will be robust going forward.

I'm also not arguing that economics is a hard science. I've done research in both Physics and Economics and they are very different animals.


Edit:

*This experiment as an aside is actually one of the better arguments against rationality. People don't behave "rationally" in it, they generally offer more than a rational person would and if a poor offer is made it is refused where a rational person wouldn't. Again, empiricism showing its place at testing really basic assumptions in Economics.

For reference: http://en.wikipedia.org/wiki/Ultimatum_game

Edit no.2: To address the latter part of your post. Don't get me started on microeconomics. There's a damn good reason that I went straight into macro and never looked back.

Eliot Rosewater

Mathematical chaos theory is well worth exploring as it has consequences in many areas such as modelling (as here) and philosophy (it lends a lot of credence to the determinism argument).

The prime mathematical example of it is the series of numbers [LATEX]x_n[/LATEX] described by the recurrence relation [LATEX]x_{n+1}=4 x_n (1-x_n)[/LATEX], with the series started with any number between 0 and 1. It's a very simple recurrence relation but the numbers it produces are chaotic and highly complex. If you start two series with two numbers as close you like (say 0.500000001 and 0.500000002) the two series you make will eventually diverge and be totally different. To relate it to weather systems: any small change in our observations of the weather today will eventually result in totally different weather in later days, hence why it's so difficult to predict the weather. (The equation also supports Scofflaw's argument that the climate over long time-periods is predictable: though the series above are unpredictable term-by-term their overall behaviour can be predicted nicely.)

But anyway, that such a simple equation could produce such unpredictable dynamics is a note of caution when modelling anything.

nesf wrote: »

My issue is with the idea of having true statements that aren't tautologies that you cannot test empirically. I mean, take the rationality set of assumptions from neoclassical economics, we can test all these using small scale experiments and show many flaws in said assumptions. I really doubt you can actually make such a statement that is useful without it being in some way at least indirectly testable.

That's basically Karl Popper's falsification criteria: a statement or claim is only scientifically legitimate if it is possible to be disproved, or falsified. You can contrast theoretical physical models, which can be falsified if the data doesn't fit, with a statement that God exists, which can never really be falsified, as examples of a scientifically legitimate and scientifically non-legitimate theories.

nesf

Eliot Rosewater wrote: »

That's basically Karl Popper's falsification criteria: a statement or claim is only scientifically legitimate if it is possible to be disproved, or falsified. You can contrast theoretical physical models, which can be falsified if the data doesn't fit, with a statement that God exists, which can never really be falsified, as examples of a scientifically legitimate and scientifically non-legitimate theories.

Basically, yes I'm a disciple of Popper in this, though there are problems with his theory and I recognise that.

SupaNova

nesf wrote: »

You cannot predict crashes. You cannot predict booms. With accuracy. This is impossible due to their nature. What you can look at is more general things like how often major events occur and how to work out probability distributions to account for this in making predictions. At best we can say there is x% chance next month of a stock market movement beyond size y. No one with any sense claims any more than this. You can also study booms and busts to see what factors are common in them and by this get warning signs of booms and busts in the future, yet still not be able to accurately predict them.

This doesn't render modelling this kind of stuff useless however, you just need to account for the things you can't model and adjust expectations accordingly of what the models are capable of doing. The latter is a serious failing in politicians.

I fail to see how these predictions can be of any more worth than qualitative predictions.

Take a general qualitative prediction made by Austrian's as far back as 1999:
This credit induced boom will end in a recession, the longer this large credit expansion goes on, and this malinvestment of resources continues the worse the recession will be. The length and severity is at the whims of central bankers, and government planners.

A quantitative equivalent:
Previous credit booms lasted between a and b years in length, this boom has being going on for c years, based on models there is an x% chance we will hit recession within the next year.

If that's all these models can produce, there is a lot of time being wasted.

SupaNova

Scofflaw wrote: »

Economic modelling, it seems to me, suffers the same limitation in respect of the kind of prediction you've suggested - that of a market crash of size x on a specific day. Apart from the problem you've raised - that "markets" consist of self-aware agents capable of factoring predictions into their behaviour - markets are also chaotic.

So the proper use of economic models, however quantitative, is more similar to that of climate models rather than weather forecasting - and even that can suffer from the problem of self-aware agents. Interestingly, the problem of self-aware agents seeking to take advantage of the predictions of economic models actually grows as models improve, because the greater the confidence which can be placed in the predictions, the more agents will seek to extract advantage from the prediction, either reinforcing it or acting against it. There might be, therefore, a built-in limit to the reliability of economic modelling as you suggest - as long as the predictions of the modelling are public knowledge.

cordially,
Scofflaw

For my final year of my software degree, modelling and predicting the outcomes of football games was one of the potential projects on my list. It is a good example of the limits of such models. We don't have to make specific predictions like what minute someone will score, even making general predictions like territory, possession, likely number of shots on goal, if they became very accurate would factor in managers tactical decisions and render initial predictions wrong.

nesf

SupaNova wrote: »

I fail to see how these predictions can be of any more worth than qualitative predictions.

Why?

SupaNova

nesf wrote: »

Why?

I just can't see how. From the examples i gave, what can we do with the second prediction that would make it of more worth than the first?

Edit: Also if you agree that the length of booms and recessions are dependent on the whims of central bankers and government planners, would you have to model those individuals with the power, spend years coming up with a model Greenspan that predicts his next move, only for him to have to be replaced with a model Bernanke. Do you go through their life's work, and weight their tendency to be loose or tight, the likelihood of them employing measures such as QE, operation twist, and what ever other tricks they may have up their sleeves?

Permabear

This post has been deleted.

andrew

nesf

Permabear wrote: »

This post had been deleted.

I don't disagree but we're talking about academic economics not predictions made by people who make their money from book sales.

SupaNova

Has there been any achievement or furthering of our understanding of market phenomena because of these models?

If all these models, that academics dedicate years of their lives to, do is make loose X% chance of Y happening predictions, that aren't any more helpful than the conclusions economists can come to without these models, it's a giant waste of time. Maybe you can give a situation or potential situation where these models can be of more help.

daithimacgroin

SupaNova wrote: »

Has there been any achievement or furthering of our understanding of market phenomena because of these models?

If all these models, that academics dedicate years of their lives to, do is make loose X% chance of Y happening predictions, that aren't any more helpful than the conclusions economists can come to without these models, it's a giant waste of time.

=the main reason I won't be doing postgraduate work in Economics

It's mostly cloud cuckoo land at the advanced level

andrew

SupaNova wrote: »

Has there been any achievement or furthering of our understanding of market phenomena because of these models?

If all these models, that academics dedicate years of their lives to, do is make loose X% chance of Y happening predictions, that aren't any more helpful than the conclusions economists can come to without these models, it's a giant waste of time. Maybe you can give a situation or potential situation where these models can be of more help.

Then you're sort of getting to the question of what is a model? If two economists reach the same conclusion, one based upon a formal model, one based upon whatever else, is the latter economist not also adopting a sort of internal sort of model anyway? The difference being that the former economist explicitly knows and can describe in detail every element of the model in a consistent, mathematical way, while the latter economist has much greater difficulty in doing so (and is more likely to make a logical mistake), because he has to use spoken language. Further, the latter economist has a harder time testing his model, given it won't translate as easily into empirical testing, and may be impossible to test unless it's formulated mathematically. Maths in model building brings rigour, without which it's too easy to use language either to obfuscate elements of the model, or to miss inconsistent parts of the model. Given Austrians emphasize logic so much, it's sort of surprising that they don't embrace maths, given it's based in logical reasoning.

nesf

SupaNova wrote: »

Has there been any achievement or furthering of our understanding of market phenomena because of these models?

If all these models, that academics dedicate years of their lives to, do is make loose X% chance of Y happening predictions, that aren't any more helpful than the conclusions economists can come to without these models, it's a giant waste of time. Maybe you can give a situation or potential situation where these models can be of more help.

Those predictions were for models of extremely complex entities like entire market indices or economies. We've a hard limit on possible accuracy with such things no matter what way you approach them due to their very nature.

That and prediction is only one thing, often with models we're trying to better explain what's going on, not trying to predict the future. Because so much public facing economics is prediction based people often don't realise the latter is probably the more common of the two.

E.g. we see a certain kind of behaviour in a relatively small market. We try different rules on behaviour of agents and try to recreate the result "in the lab." This kind of thing is extremely useful in trying to work out how to design markets to more efficiently achieve a certain goal (i.e. what kind of auction should you set up to get the maximum price for a once-off sale of something like a new radio station licence or whatever).

The reason to use maths rather than natural language logic is that it is usually more precise to use mathematics and less open to misinterpretation. Maths can only be interpreted in the way the author interpreted it, natural language doesn't have this restriction at all which can cause a lot of problems.

SupaNova

nesf wrote: »

That and prediction is only one thing, often with models we're trying to better explain what's going on, not trying to predict the future. Because so much public facing economics is prediction based people often don't realise the latter is probably the more common of the two.

E.g. we see a certain kind of behaviour in a relatively small market. We try different rules on behaviour of agents and try to recreate the result "in the lab." This kind of thing is extremely useful in trying to work out how to design markets to more efficiently achieve a certain goal (i.e. what kind of auction should you set up to get the maximum price for a once-off sale of something like a new radio station licence or whatever).

Do you really need to spend years making models to come up with this??? How superior would it be to consulting with people experienced in auctioneering. And in making such a model would you not be consulting with experienced auctioneers to get their input in the first place?

That and how much do radio licenses sell for, it would have to be a hell of a lot to justify years of research to produce such a model.

The reason to use maths rather than natural language logic is that it is usually more precise to use mathematics and less open to misinterpretation. Maths can only be interpreted in the way the author interpreted it, natural language doesn't have this restriction at all which can cause a lot of problems.

Does the maths not describe what the author of the equation thought in the first place, and thus can be expressed in language, and what if this is completely flawed view in the first place?

Furthermore, as political scientist Bruno Leoni and mathematician Eugenio Frola pointed out,

"It is often claimed that translation of such a concept as the maximum from ordinary into mathematical language, involves an improvement in the logical accuracy of the concept, as well as wider opportunities for its use. But the lack of mathematical precision in ordinary language reflects precisely the behavior of individual human beings in the real world…. We might suspect that translation into mathematical language by itself implies a suggested transformation of human economic operators into virtual robots."


[...]

Although himself a mathematical economist, the mathematician son of Carl Menger wrote a trenchant critique of the idea that mathematical presentation in economics is necessarily more precise than ordinary language:

"Consider, for example, the statements:

(2)To a higher price of a good, there corresponds a lower (or at any rate not a higher) demand.

(2')If p denotes the price of, and q the demand for, a good, then
q = f(p) and dq/dp = f' (p ) <= 0

Those who regard the formula (2') as more precise or "more mathematical' than the sentence (2) are under a complete misapprehension… the only difference between (2) and (2') is this: since (2') is limited to functions which are differentiable and whose graphs, therefore, have tangents (which from an economic point of view are not more plausible than curvature), the sentence (2) is more general, but it is by no means less precise: it is of the same mathematical precision as (2')."

Andrew can the fact that Inflation occurs at different points and times at different rates altering the structure of production be translated into a mathematical equation, is the mathematical equation going to be more precise than the language used? And if we can translate this fact into an equation that doesn't lose precision, what then can we use this equation to do?

SupaNova

I’ve been catching up on the apprentice lately, just imagine the following.

Academics: “Lord Sugar, we have spent 5 years and millions of pounds to produce a model of how to structure your meetings with clients that will maximize your chance of getting a good deal, are you interested?”

Lord Sugar: “You wha??? What kind of idiot gave you millions of pounds and 5 years to come up with this crap??? Lol, actually stop wasting my time and get the hell out.”

[Turns to Karen and Nick]

Lord Sugar: “Lolol Are these guys completely taking the piss or what?”

nesf

SupaNova wrote: »

Do you really need to spend years making models to come up with this??? How superior would it be to consulting with people experienced in auctioneering. And in making such a model would you not be consulting with experienced auctioneers to get their input in the first place?

Actually, no because of the biases involved in human judgement about their own area. There's a lot of psychology behind this but if you take aside an auctioneer and ask them the best way to sell something all you'll get is a gut answer based on a limited sample of their experience combined with natural biases towards supporting personal preferences. You can see an extremely good example of this by reading this thread in a neutral state of mind and see my support for mathematical models (of a certain kind) and your opposition to them as products of our inherent biases coming into this debate.

SupaNova wrote: »

That and how much do radio licenses sell for, it would have to be a hell of a lot to justify years of research to produce such a model.

When you're talking about say a 3G mobile licence you're talking about many millions, so yeah it's worth hiring someone to model it for you. And the academic research in this area will happen anyway because of that thirst for knowledge and understanding of the world around us that drives some people. Personally I find market design fascinating just for its own sake and I'll most likely never be in a position to be asked how to design a market! "Reinventing the Bazaar" has a good chapter on this actually if you're interested.

In the end it comes down to: why not model it? Why not try to get a better understanding of it? Etc.

SupaNova wrote: »

Does the maths not describe what the author of the equation thought in the first place, and thus can be expressed in language, and what if this is completely flawed view in the first place?

Um, of course. Putting a bad idea in mathematical form will not change the fact that it is a bad idea, I thought this would be obvious? The benefit of maths here is that it's often a lot easier to show it's a bad idea if it's in mathematical form rather than natural language. With natural language you get argument and counter-argument spirals that can last decades, with maths a single devastating point can be made to show the absurdity of the model/idea.



Anyway, as an aside, I'm not really into mathematical models devoid of empirical support*, so this is getting a bit outside my areas of interest here.


*Which means I dislike vast swathes of microeconomics and some mainstream macroeconomics! A lot of which is what Krugman is complaining about in that article Permabear linked to. Maths for the sake of maths rather than maths for the sake of explanation. I've a serious problem with this.

SupaNova

The book sounds interesting, i will add it to my list.

nesf wrote: »

In the end it comes down to: why not model it? Why not try to get a better understanding of it? Etc.

I have no problem with trying to get a better understanding of markets. My question a couple of posts back was "Has there been any achievement or furthering of our understanding of market phenomena because of these models?".

EDIT: If the goal is to find ways of making markets more efficient, energy is completely in the wrong place if people are looking at the auction of radio licences rather than causes of economy wide distortions.

Um, of course. Putting a bad idea in mathematical form will not change the fact that it is a bad idea, I thought this would be obvious? The benefit of maths here is that it's often a lot easier to show it's a bad idea if it's in mathematical form rather than natural language. With natural language you get argument and counter-argument spirals that can last decades, with maths a single devastating point can be made to show the absurdity of the model/idea.

If the math and models are based on bad theory, you still have the same problem. People can still debate for decades whether the theory is actually bad or not. I'm pretty sure Roger Garrison has some good stuff on Keynesian equations that would satisfy us both.

nesf

SupaNova wrote: »

I have no problem with trying to get a better understanding of markets. My question a couple of posts back was "Has there been any achievement or furthering of our understanding of market phenomena because of these models?".

It's a complicated question that you're asking and I can only respond from my own poor memory here and this wasn't an area I was hugely interested in to begin with! Have a read of that book, it's an interesting economic history of markets.

SupaNova wrote: »

EDIT: If the goal is to find ways of making markets more efficient, energy is completely in the wrong place if people are looking at the auction of radio licences rather than causes of economy wide distortions.

Well, studying one thing doesn't exclude studying something else. There are a lot of economists so a lot of ground will be covered from big market stuff down to studying micro-markets in developing countries and everything in-between.

SupaNova wrote: »

If the math and models are based on bad theory, you still have the same problem. People can still debate for decades whether the theory is actually bad or not. I'm pretty sure Roger Garrison has some good stuff on Keynesian equations that would satisfy us both.

An example: there was a Keynesian model that implied that inflation and unemployment couldn't increase at the same time. Then the 70s happened and both increased at the same time. What was wrong with the model? Well, it worked well enough so long as nothing was imported into the economy. As soon as you brought in imports the model no longer worked and the empirical fact of stagflation proved this beyond doubt really and forced Keynesians to go back to the drawing board (which in some ways was a pity because instead of rejecting what didn't work, most of Keynesianism was rejected and we lost things like sticky prices for a very long time from the mainstream models of economies which was a remarkably dumb thing to happen).

Yes the theory will be debated endlessly but no one uses that particular Keynesian model anymore for serious prediction.

SupaNova

nesf wrote: »

Well, studying one thing doesn't exclude studying something else. There are a lot of economists so a lot of ground will be covered from big market stuff down to studying micro-markets in developing countries and everything in-between.

Yes a wide range of ground can be covered, but from some of the quotes Permabear posted there is a the belief that there is too much attention in the wrong place.

An example: there was a Keynesian model that implied that inflation and unemployment couldn't increase at the same time. Then the 70s happened and both increased at the same time.

Garrison mentions this in one of his lectures. He also has an interesting lecture comparing Keynes and Hayek. If Garrison's interpretation of Keynes theory is correct, the theory is completely flawed. If it is a correct interpretation, it would be interesting to know if the flaws are recognized by modern Keynesians and how much it has been modified:

http://www.youtube.com/watch?v=pNX1rMiCUO0&list=FLTA_SD9OCs8dqvLilYcwbIw&index=1&feature=plpp_video

nesf

SupaNova wrote: »

Yes a wide range of ground can be covered, but from some of the quotes Permabear posted there is a the belief that there is too much attention in the wrong place.

Same can be said for every academic field really. One man's useless waste of time researching can be another's fascinating goldmine of information. Don't fall into the trap of thinking some research is pointless just because you disagree with it. Read up some papers on how you can't plan or manage research like a command economy if you want to dip your toe in this area.

SupaNova wrote: »

Garrison mentions this in one of his lectures. He also has an interesting lecture comparing Keynes and Hayek. If Garrison's interpretation of Keynes theory is correct, the theory is completely flawed. If it is a correct interpretation, it would be interesting to know if the flaws are recognized by modern Keynesians and how much it has been modified:

http://www.youtube.com/watch?v=pNX1rMiCUO0&list=FLTA_SD9OCs8dqvLilYcwbIw&index=1&feature=plpp_video

In the mainstream anyway the idea was abandoned (and replaced with a different model with it's own flaws, the neo-classical one, which was then replaced by a mixture of the best of both called the neo-classical Keynesian synthesis, or neoclassical synthesis for short). You'll almost certainly find some fringe economist who follows the old Keynesian doctrine in full but most Keynesian commentators that i've read don't seem opposed to bringing in good ideas from elsewhere to patch the holes in Keynesianism.

Some good things came from Keynes though. Sticky prices etc was very much pushed by Keynesians (not sure if you know about what I'm talking about, basically in many markets prices don't change very often because it usually costs money to change prices or advertise the change etc, so instead of a free floating price dictated by supply and demand as per the neo-classical model you get a price that can cause over or under supply or demand in the market because of this "stickiness"). Makes absolute perfect sense when you consider it for the likes of restaurants and other businesses that need to spend money on menus and the like. There are other reasons for stickiness too but you can look them up if interested, I'm far too tired to go digging around about it.

SupaNova

nesf wrote: »

Same can be said for every academic field really. One man's useless waste of time researching can be another's fascinating goldmine of information. Don't fall into the trap of thinking some research is pointless just because you disagree with it. Read up some papers on how you can't plan or manage research like a command economy if you want to dip your toe in this area.

Fair point.

In the mainstream anyway the idea was abandoned (and replaced with a different model with it's own flaws, the neo-classical one, which was then replaced by a mixture of the best of both called the neo-classical Keynesian synthesis, or neoclassical synthesis for short). You'll almost certainly find some fringe economist who follows the old Keynesian doctrine in full but most Keynesian commentators that i've read don't seem opposed to bringing in good ideas from elsewhere to patch the holes in Keynesianism.

Some good things came from Keynes though. Sticky prices etc was very much pushed by Keynesians (not sure if you know about what I'm talking about, basically in many markets prices don't change very often because it usually costs money to change prices or advertise the change etc, so instead of a free floating price dictated by supply and demand as per the neo-classical model you get a price that can cause over or under supply or demand in the market because of this "stickiness"). Makes absolute perfect sense when you consider it for the likes of restaurants and other businesses that need to spend money on menus and the like. There are other reasons for stickiness too but you can look them up if interested, I'm far too tired to go digging around about it.

Yeah i'm aware of sticky prices, i don't think Austrian's have much of a problem with it. More so with the idea that just because prices are not perfect or don't adjust immediately to changes in conditions, automatically justifies intervention.

nesf

SupaNova wrote: »

Yeah i'm aware of sticky prices, i don't think Austrian's have much of a problem with it.

They really wouldn't want to have a problem with it, it's kind of all around us and we can all see it in front of us. Would you believe that some people continue to model markets as if sticky prices don't exist? Believe me, when I defend modelling I'm most certainly not defending all of it.

andrew

SupaNova wrote: »

That and how much do radio licenses sell for, it would have to be a hell of a lot to justify years of research to produce such a model.

Aside from the general desire for knowledge, as Nesf said, radio liscenses attract a lot of money:

"From 2000-07-31 to 2000-08-18, the German government conducted an auction for 12 frequency blocks for the new UMTS mobile telephony standard. The total of the bids exceeded expectations by reaching the staggering amount of DEM 98.8 billions (EUR 50.8 billions). (See de:Versteigerung der UMTS-Lizenzen in Deutschland)"

Andrew can the fact that Inflation occurs at different points and times at different rates altering the structure of production be translated into a mathematical equation, is the mathematical equation going to be more precise than the language used? And if we can translate this fact into an equation that doesn't lose precision, what then can we use this equation to do?

To be honest, I don't have enough maths to be able to answer that question, so I'll go with maybe. But I don't see the point in asking such specific questions; within that question is the assumption that inflation does cause structural distortions which are significant enough to need modelling in the first place, and the assumption that if such distortions can't be modeled, then modelling is useless. Finally, you seem to be taking a very utilitarian approach to all of this; not every economic insight need have, or does have, a policy aspect. I've already explained above what models are useful for doing; examining an argument with more rigour, and removing the huge ambiguity language can bring when dealing with complex issues.

Scofflaw

SupaNova wrote: »

"Reinventing the Bazaar" has a good chapter on this actually if you're interested.

The book sounds interesting, i will add it to my list.

Do! It's an excellent book, and doesn't mistake mathematics for sense.

cordially,
Scofflaw