The United Nations Environment Program did a very unusual and far-sighted thing last year: it asked the CSIRO to produce a version of its biophysical model of the Australian economy for developing countries, and to pair that with an economic model which had to be non-equilibrium in nature.
The Stocks and Flows Resource Modelling team at CSIRO Sustainable Ecosystems, which maintains this biophysical model, approached me in July to see whether I would be willing to attempt the development of that nonequilibrium multisectoral economic model.
The CSIRO’s biophysical model uses a software package called WhatIf? that is designed for multi-dimensional aggregation of data and extrapolation of the trends that data displays. Resource data is stored at as disaggregated a level as possible (population, for example, is disaggregated by age, sex and location), and then aggregated to the national and global level. A similar exercise applies for consumer demand, which is driven by an age-based profile of the population, allied to the consumption patterns that apply for each age cohort across all industries in an economy, and the resource requirements of output in that industry.
As I document in this paper, the project to develop a monetary, multi-sectoral, nonequilibrium model of production was successful, while the CSIRO successfully adapted their Australian stocks and flows model to the available Asian data. The CSIRO and I presented our results to the UNEP group in November.
The video quality is pretty terrible–I’ve learnt a bit about filming since then–but nonetheless here they are. The videos “speak for themselves”, so I’ll let them do just that.
Franzi Poldi’s explanation of the CSIRO’s of the Asia-Pacific Stocks and Flows Model:
Steve Keen's Debtwatch Podcast with Stuart Cameron
My explanation of my multisectoral monetary model (and the PPT file for my talk):
Steve Keen's Debtwatch Podcast with Stuart Cameron
January 30th, 2010 at 1:19 am
I’m about halfway through the second of these videos. One of the most interesting and informative I’ve seen yet, so would be nice if you could improve the quality since you’ve “learnt a bit about filming since then”. I expect you’ve plenty of spare time … ?
January 30th, 2010 at 4:47 am
Mr Keen, thank you for your blog.
For a decade or more I have been thinking about and writing about many of the concepts you have developed. I am neither as educated nor as brilliant as you are. Logic, common sense and observation have led me to the conclusions that you so eloquently have developed through mathematics, history and research.
You do not mention Kondrateiff. It would seem to me that his wave theories along with the Elliot wave theories move in somewhat the same direction as your conclusions. According to my interpretation of Kondrateiff’s theories we are entering into what some refer to as the K wave or entering the Winter of his hypothesis. Just wondering how of if his theories fit in with your conclusions and work.
A clearer version of this latest presentation, where I could actually see the visuals would be nice. This is much better than no presentation though.
I want to thank you again for your blog and your educational service to the world. You are doing a great job of putting history and theory together to make sense of the world we live in and allowing the rest of us to gain knowledge and hopefully wisdom from all your hard work.
Best Regards and may you stay healthy
January 30th, 2010 at 10:31 am
Thanks economicminor.
One reason I don’t mention Kondrateiff is that my Minsky model with government (in the 1995 JPKE paper) generated “long waves” with absolutely no long-run dynamic in the model: instead the actual long run fluctuations in growth rates were the product of the complex dynamics of the system. So I have a similar feeling about K waves: while they might describe the actual shape of the trade cycle over the long term, they are due to the short run dynamics of the system and do not represent any separate long run causal mechanism.
I could be wrong of course, but with that perspective and my time constraints I haven’t paid great attention to Kondrateiff’s arguments.
January 30th, 2010 at 10:35 am
Hi doggett, if someone can invent a means of extracting 48 hours from a day, then I might just find that spare time somewhere–but I suspect it’s in a Black Hole!
I can’t improve the quality of that video, but I will probably get a chance to present that paper again and at that point the filming will improve.
I’ll also put up a separate post shortly with the PPT file; should have done this with the post but I forgot. Must be a time thing….
January 30th, 2010 at 10:50 am
Hi Steve,
First I want to thank you for you amazing efforts. As someone who is just about to finish a master’s degree in economics your work has been a major inspiration to move beyond the neoclassical models we are constantly fed in class.
One thing I find extremely interesting is the idea that equilibrium (if there is one at all) should be a result of the actual behavioural assumptions of the model, not something that is imposed on the model. I find it hard to understand how economists who always argue in favour of the self-stabilizing properties of capitalism find themselves in a position where they are not able to construct a model that exhibits these properties.
I do hope that I will get the opportunity to contribute to the exiting field of non-equilibrium modelling, but as someone who is just about to apply for a PhD program in economics I am not too hopeful.
Regards,
Linus
January 30th, 2010 at 11:46 am
Thanks Linus,
It may actually be a good time to be aiming to do precisely that. Where are you applying to do the PhD?
The neoclassical equilibrium fetish is bizarre too, isn’t it? You’d think the fact that they can’t get even their models to exhibit endogenous stability might make them wonder whether the real world behaves the same way; but instead they are so beholden to the need to achieve equilibrium that they impose it by assumption instead.
January 30th, 2010 at 11:59 am
I’m moving to Australia and Canberra shortly (if I get my de-facto visa that is), so I have started looking into ANU and the universities in Sydney.
At the moment I am finishing my last master’s course in international macroeconomics/finance and trying to spend as much as possible of my spare time reading non-orthodox literature that can give inspiration to a research proposal.
January 30th, 2010 at 12:19 pm
Hi Steve,
The RBA has recently published its monthly debt stats. I’m curious to see whether or not we’re at an all time high on the debt to GDP charts that you publish.
Have our economic managers managed to reignite and keep alive our debt fueled growth?
January 30th, 2010 at 9:39 pm
Hello Steve,
thanks for the blog. I have been reading it sometime now, and it is really good.
I have some questions regarding your modelling:
how far did you already extend your models? the next points, i don’t know if you have to implement it as different states in your model, or if they are already part of the current states (I looked at the model you showed in this presentation).
- Consumers: they can have loans as well, is this part of the deposit? In other presentations you showed private debt growth to spend ourself out of a recession.
- did you add Government already as a state in your models: (if I understand Minsky correct) the size of this ’stock’ should scale with the accumulated investment, and than could dampen out osccillations if consumers stop spending.
- the lender of last resort. So if central banks reduce their interest rates to zero, how can you see this in your model? is this ‘hidden’ in the interest rate from the bank to the firm? or is it part of a system state (bank reserve)
- did you generally check if the debt in the model is reduced to zero, every other stock is reducing as well?
- and do you see in the end that the banks will have the only remaining stock?
Well anyway, keep up the excellent work.
Best Regards
Ignace
January 31st, 2010 at 6:11 am
Hi Steve,
Here’s a story describing banks not lending, and
shows how team Obama and their focus on the finance industry
(like you describe in the video)
instead of the Main Street economy means it will likely not turn out like Team O expects.
Especially when viable businesses lack the finance to stay in business, and then end up with a negative economic feedback loop (business failure to increased unemployment to decreased tax receipts for government operations, and repeat a few cycles).
http://www.nytimes.com/2010/01/31/business/smallbusiness/31order.html?hpw
What is the best advice for Team Obama when they realize a different approach is needed?
January 31st, 2010 at 8:36 am
Hi Linus,
You mention in your post:
“One thing I find extremely interesting is the idea that equilibrium (if there is one at all) should be a result of the actual behavioural assumptions of the model, not something that is imposed on the model.”
I’m actually painstakingly memorising all the mathematical derivations which begin with behavioural assumptions and with further assumptions end up with general equilibrium. I am doing a mathematics course (the exam is tomorrow) and I would not imagine economics students having to go through the pain of learning all that maths.
I think the equilibrium is derived and is based on a series of assumptions which are clearly unrealistic. The idea of equilibrium is that the economy is in a state where there is no access supply or demand and everyone has maxed out their utility. Maybe economists, dazzled by the math take equilibrium as given, but the equilibrium itself is originally derived, not assumed as far as I can tell.
January 31st, 2010 at 8:58 am
Hi TruthIsThereIsNoTruth,
It is of course true that equilibriums can be derived from a set of assumptions, however a large number of macro models that I have encountered impose equilibrium conditions, even though they are sometimes presented as behavioural assumptions they are in fact equilibrium conditions. When solving models it is for example common practice to assume that some variable or market is at it’s equilibrium position, and then use that in the derivation of the solution.
The models are then used to analyse what happens when the economy is ’shocked’ away from equilibrium, even though the solution by assumption is only valid for equilibrium. This always puzzled me. I could never understand how this was an appropriate use of mathematical methods.
January 31st, 2010 at 9:40 am
Linus,
Sounds like it is as I suspected, the equilibrium is taken as given when used by economists. I’m only learning the maths up to equilibrium and finish by replicating the Black-Scholes result. This is not to say GE is a necessary condition of BS but you can get to BS via GE. After that I am done with it!
Having to learn this in a condensed summer semester has certainly thrown me out of equilibrium.
January 31st, 2010 at 11:55 am
From Canada, and the rest of the world: Well done, Steve
Everybody who comes to your website could send an email to the newspapers, our leaders, TV, etc., pointing them here. How long would it be for us to get them to pay attention instead of just spending more of our money?
I believe Maple has an optimization routine. I assume Mathlab does also. If you put all your shaping constants in a table, and provide your model with a table of real-world historical data, and place your model’s output data in another table, then you can calculate the sum of squares difference between the two outputs, the optimizer can drive your model (via variations in the constants, using multiple search routines). Using this method on a reasonably fast computer and a weekend, can’t the model be automatically calibrated by itself?
January 31st, 2010 at 12:17 pm
Hi gcjblack,
That would work with a linear model; with a nonlinear one it would trap the model in the nearest local minimum to the initial conditions specified by your initial parameters. There’s an enormous literature on this from which a range of new techniques to avoid being trapped in local minima have developed–from genetic algorithms to simulated annealing. All of them still require enormous amounts of time and computing power. I marked a thesis recently where a ten parameter system still took a few weeks to solve on a standard PC.
January 31st, 2010 at 2:58 pm
Which version of MS Powerpoint do I need to open your ppt presentation?
My 2003 version can’t find a .ppt file after unzipping the download.
……../Chris
January 31st, 2010 at 3:02 pm
Hi Chris,
It’s using the 2007 format, but you can download a free converter/viewer from Microsoft.
January 31st, 2010 at 10:57 pm
[...] This post was mentioned on Twitter by greychampion, John Hacking. John Hacking said: CSIRO-UNEP Modelling: The United Nations Environment Program did a very unusual and far-sighted thing last year: i… http://bit.ly/dn2SGU [...]
February 1st, 2010 at 1:15 pm
Here is a few links supporting your work & conclusions. There is nothing like history to prove your case:
Inflation in France, a 77 page book, written in 1910’s in USA, discussing the 1750’s in France when they made 2 attempts at Fiat Currency, both of which were absolute disasters, in spite of trying everything to make it work.
http://mises.org/books/inflationinfrance.pdf
Link was originall discovered here: http://itulip.com/forums/showthread.php?p=146698#poststop
Another link:
http://www.theglobeandmail.com/report-on-business/commentary/brutal-economy-of-1700s-has-an-eerie-similarity/article1448623/
February 3rd, 2010 at 7:16 am
Steve,
I am extremely impressed by your work and truly hope your
approach gathers more interest from researchers inside and outside of the
economic community. As an engineer trained in dynamic systems and control theory,
but who has a personal interest in economics, your approach is very attractive.
In grad school I would always ask my econ friends about the lack of differential
equations in their field of study and your site has provided some answers.
Now on to some comments on your model:
I appears your multi-sector dynamic economic model can be
expressed as a nonlinear state space system
xdot=f(x,u)
where x is a nx1 vector of state variables, xdot is the time
derivative of the state vector x and u is a vector of inputs . Is it true that
you consider no exogenous inputs so that the right hand side of xdot is really
just f(x)?
If so it would be interesting to add an exogenous input of
money flow from the central bank into the Bank Reserve state variable (BR). I believe this would enter the equations in
the 1st column and last row of the flow matrix described in fig 20. Then one could start to analyze the dynamic
system with a central bank as a feedback control law i.e. u=fc(x) where this
law is limited in some way. This would be interesting because you could then
show the concept of “pushing on a string” dynamically. The CB could increase BR but only the algebraic
equations describing behavioral aspects of how much various sectors borrow
would determine if these dollars in BR actually get lent out.
Finally it seems to me that another way of describing the double
entry book keeping system is by introducing what we like to call an output equation
y=h(x)
where y is a mx1 vector of output variables and is an
algebraic function of the system state vector
x.
Now we can construct h(x) in order to describe as many things
about the system that we want. For instance we can define elements of the output
vector such that y1 = Firm Net Assets, y2=Firm Liabilities and y3=Firm Net
Worth. All of these output variables can be constructed as algebraic functions
of the system states without having to touch(augment ) the dynamic system
equations.
Finally when modeling physical dynamic system we like to use
the concept that the number of states in the system are related to the number of energy
storage elements. I completely agree with you that we should not blindly port
over concepts from physics like energy conservation. Money is not conserved. However
it is appealing to me that many of the state variables in your system can be
thought of as wealth storage elements. The states describing deposit accounts and
loans can all be considered assets of the entity that owns them and consequently
stores of wealth. However I am not quite
sure who really owns the BR storage element and how it is valued. I recall the bank can’t treat it as part of
their assets but it is of some value to them since they can derive future profits
from it if they can loan it out.
Looking forward to hearing your thoughts.
February 3rd, 2010 at 9:36 am
Oh yeah,
It would interesting to formalize bank lending being capital vs reserve constrained. These would introduce saturation nonlinearities in the RHS of the state space system i.e. inside f(x).
February 3rd, 2010 at 10:30 am
Re #20 CSIRO-UNEP
Thanks mmb,
Yes that’s generally true (no exogenous inputs) but I have modelled that in two papers (One Two) to assess the impact of an “exogenous” input of size X and its impact on the economy depending on whether X is given to bankers (BR) or to debtors (FD). That shows that it’s much more effective to give bailout money to debtors.
There are algebraic relations you in my system as well, though not in the bank accounts section. I’m working with a European colleague to develop a new dynamic modelling tool, and algebraic relations aren’t yet possible there; but they will be.
The BR element is something that gets me in fights with Circuitists (and not just Chartalists!) because they have developed the convention that money is destroyed when debt is repaid. I instead argue that repayment of debt replaces one bank asset (debt) with another (which I label BR for Bank Reserves but which JKH might suggest would be better called Bank Equity). This is similar to what Keynes called “the revolving fund of finance” and I model it as such in this paper.
Sorry for a staccato reply but (as others may have noticed!) I’m busy as hell now and letting a lot of comments go because I don’t have spare time.
February 3rd, 2010 at 10:35 am
Re #21. Yes mmb, that is an interesting additional issue that my systems engineer colleague Trond Andresen has tackled in a very interesting set of papers.
February 4th, 2010 at 5:51 pm
Re #22. Thanks for the links I’m reading them now. The BR state variable is indeed a tricky one I am still trying to wrap my head around. I am having difficulty resolving BR as a bank asset or bank equity with the seigniorage issues you mentioned in previous papers/lectures.
With regards to my notion of an output equation and “algebraic relations”, I was just noticing how additional ways of looking at the dynamic system can be accomplished using this state space theory construct. For instance one could take various elements from your flow matrix and duplicate them in the output function y=h(x). The output function is just “accounting” to me or a particular way of looking at the underlying dynamic system.
February 4th, 2010 at 6:26 pm
Re #23. I have indeed found Trond’s Basel paper very interesting. His treatment of a central banks (CB) and capital ratios appears clearer to me after seeing it represented in equations!
In comparing your multi-sector model paper with his I have noticed that he creates an algebraic relationship between aggregate bank assets (A) and liabilities (L) by constraining lending such that a fixed capital ratio is maintained.
One of the rules of state space theory is that in order for a variable to qualify as a system state, it cannot be expressed as an algebraic function of one or more other system states.
So aggregate bank assets (A) becomes a state variable while bank liabilities (L) become an output variable i.e. L=h(A)
In his treatment of bank reserves held at the CB he creates an additional constraint when he forces a certain central bank interest rate policy and government deficit spending to force an algebraic relationship between reserves (R) and assests (A).
I’m wondering if this treatment can be modeled so that the constraints are expressed as inequalities so A and L are both independent state variables as long as the system is not against the constraint captial ratio constraint.
It is all extremely facinating!
February 5th, 2010 at 10:34 am
Re #25. Interesting thoughts mmb. Trond is very approachable–drop him an email about it: trond.andresen at itk.ntnu.no.
The BR account is something that JKH might argue would be better called BE for Bank Equity.
The point about seignorage is that a lender can’t directly spend money it creates on commodities–though it can earn an income from that money via interest payments and spend that. Hence the initial separation of BR from BD (or BI–I’m still working out how to label these various accounts in my deliberately skeletal model): BR is not a source of spending by the bank whereas BD is.
February 5th, 2010 at 11:55 am
I’ll drop Trond a note as you suggested. Now regarding the BR account. If the bank can’t spend the money in BR and can only make money by lending it out, how is it valued? Is it considered an asset on the banks books but not valued at face value? It would seem to be a very strange asset that must be valued using some estimate of how much interest could be generated from lending it all out. Maybe a NPV of future interest payments?
It doesn’t seem like BR should be bank equity either. Isn’t bank equity = assets – liabilities? Where am I going wrong here? Not to bother prof Keen with such a newbie question but can anyone else clarify?
Thanks!
February 5th, 2010 at 12:48 pm
mmb
The BR account is a pool of credit lines. When a firm repays a portion of a loan, it restores its credit line.
February 5th, 2010 at 1:41 pm
Steve,
Back to the usual subject of credit money destruction:
Opening a line of credit is not the same as creating credit money. It is “a commitment from a lender to make available to a borrower a certain amount of credit”. Credit money is created when the actual loan is extended and it is extinguished when the loan is repaid.
http://www.creditorweb.com/definition/revolving-line-of-credit.html
The best example of such a line of credit is my credit card account. When I pay using the card I spend newly created credit money. When I repay the debt credit money is destroyed and my credit card account reverts back to the same state it had had before spending money. The initial state of the credit line is restored and I am again able to draw (borrow) the same amount of money – created on demand by the bank.
But I am not charged interests just for having a line of credit – this is another argument that this is not money as it cannot be spend, doesn’t create a bank asset and doesn’t generate income. Paying a fixed fee to the bank for maintaining the credit line doesn’t change anything.
In my opinion the only reason BR account may be included in the general model is to limit the maximum amount of credit money created – if the model is not driven by the bank pushing money. If we want to simulate the behaviour of the system during the deflationary phase we must assume that debt is being repaid – this is exactly what slowes down the economy by reducing the demand. Only during the expansionary phase the amount of credit money in the system may be close to the limits.
February 5th, 2010 at 5:16 pm
Steve, Warren and AK,
AK makes a good point but hints that this has been discussed before. Not wanting to duplicate discussion points, can someone list articles and post #’s where this was covered so I can review. I looked for a way to search the comments but could find none.
AK, I am intrigued by your comment regarding BR being used as a way to limit money creation. This seems to me like the right way to introduce this variable into the ODE’s and is likely related to the concept of bank lending being reserve constrained. The paper by Trond Anderson, that Steve pointed me to,
http://www.itk.ntnu.no/ansatte/Andresen_Trond/econ/basel-working8.pdf
describes such a constraint in terms of bank lending being capital/asset ratio constrained.
Trond uses an equality contraint in the form of A-L/A=k where A=bank assets, L=bank liabilities and k=the minimum capitol to asset ratio. He also suggests the equality corresponds to the case where the economy is not pessimistic (i.e. banks want to lend and borrowers want to borrow) In my opinion it is important to have the model describe the case where bank credit money creation (flow) is dynamically determined based on banker and borrower sentiment but subject to inequality capitol and reserve constraints. In the example of Tronds paper where he considered capitol/asset ratio constraints without a central bank, this would suggest state variables A and L should be free to wander around in the state space until they bump up into the constraint curve A-L/A=k. As the system tries to move beyond this constraint boundary (i.e. A-L/A<k) the constraint becomes "active" and overpowers the new credit money creation flows preferred by bankers and borrowers and drives the state (A,L) back inside the constraint. Once it is back inside, the constraint turns off and the state variables (A,L) are again free to roam around based upon the banker/borrower preferences. If for some reason banker/borrower preferences are always trying to push past the constraint boundary, the state will move parallel to the constraint much like an ant that wanders around your floor until he reaches a wall.
It would seem to me that bank reserve constraints could be handled the same way.
Can someone describe the acounting mechanism in which a bank can add or subtract money from its reserve account (not Steves BR but Tronds R) at the central bank (CB)? It seems to me it can only transfer money into its CB Reserve account from its own account (i.e. Steve's BI account) This is the account where the bank stores all of its interest income and money from the sale of assets it owns? Is my understanding correct?
Thanks everybody!
February 5th, 2010 at 6:29 pm
mmb,
I can recommend reading Steve’s papers and searching this forum for posts written by JKH. This will lead you to the relevant threads.
Do a following Google search:
“JKH site:http://www.debtdeflation.com/blogs”
(I’m not posting a full URL of the search because it may not survive).
Bank lending may not be reserve constrained:
http://www.debtdeflation.com/blogs/2009/12/23/mish-on-the-fictional-reserve-system/#comment-19129
February 6th, 2010 at 8:13 am
My latest 1 sector realization of Steve’s model.
If you compare with US Unemployment 1890-2009, it getting pretty clear that Steve is within an exogenous shock or two of having them completely surrounded.
February 6th, 2010 at 8:16 am
Notice the oscillations about Unemp=5% and compare with previous simulation/
February 6th, 2010 at 8:37 am
Re #32-33 CSIRO/UNEP. You’re making me jealous Warren–I wish I had time to work on my own models but right now I don’t, too busy with the blog I have to put together on the walk I’m doing to Mt Kosciousko. So many thanks to you for getting well and truly into the modelling.
February 6th, 2010 at 9:28 am
OK, so here’s a credit crunch 100<t<120
I'd like to join you on the walk. It's just the 18 hour
airplane flight that's putting me off.