Transcript
0:00 [Music] so what I want to cover now is an Introduction to System Dynamics using Minsky has anybody here installed Minsky okay a couple uh if you want to if you got a network connection you can install it now and perhaps try to do some of the things I do with it as I'm talking through I want to start from the um ontology side to some extent because there's a fairly you know we all don't need to talk about Tony Lawson and
0:25 critical realism we've got a fair few comment that so far and my I've written a critique uh Tony and our good friends so it's the sort thing you have a friendly disagreement but I think the critique that Tony's making about the nature of mathematical modeling and economics is specific linear models only uh and he has a quote from Hiser his contribution now what is neoclassical economics where he basically said anybody using mathematics is neoc classical which I took very well as you
0:54 can imagine um in a good sense uh is that when you say Adam's must be isolated from other countering factors to that could interfere with the outcomes and to guarantee truly predictable and or deductible and then deducible and then says that pretty much describes mathematical modeling in general well you can do a mathematical model of the solar system using Newtonian Concepts no nothing nothing from Einstein needed and that's a closed system model as you describe it in the
1:23 mathematical terms that that um that Tony thinks he critiquing but this is a modern astronomer talking about the Dynamics of the solar system now can somebody please move their pen move your pen pick up a pen move it move it okay you've just changed the location of of Jupiter in a billion years time perhaps in one side of the Sun to the other okay that is how sensitive complex nonlinear interacting systems are and in fact rather than
1:53 saying the university is predictable then I say well it's it's predictable over over a certain time Horizon but unpredictable in the genuine sense of the word and that's that the impact of interactive non-additive systems is something outside the critique that's been made by critical realism and I I just would like to see heterodox economists catch up with that because they're critiquing what NE classicals do when I Define neoc classicals anybody who works with equilibrium Concepts and leaves out
2:21 money um so I think that's the critique of critical realism fundamentally applies not to what's been done by heterodox economists uh but to What's Done by neoc classical where they try to make everything additive and they and they take system which are complex and say let's make them simple they actually make them complicated which is not by no means the same thing is complex so System Dynamics is fundamentally it's a system to visually represent differential equations where the
2:45 differential equations are describing processes of change in a complex system and the idea is if you describe the system structure you'll find the Dynamics arise from the structure it puts much less emphasis upon what individual behavior is than the neoc classical and then indeed the aine obsessions are uh compared to doing the same things with mathematical mathematics directly which you can do if you try to do symbolic Solutions everything you're forced to work in 2D systems okay once you get to 3D uh and
3:14 sudden you get to 5D there is no solution symbolically to any of these systems so what system ma mathematic mathematicians tend to work with what they regard as tractable and that's a tendency to work to reduce things to two dimensional systems or to take a 3D system and break it into into three 2D versions and try to combine it together System Dynamics began uh in actually solving a real world problem and tends to be what you can call structural
3:40 you're trying to describe the structure of the system you're looking at and many dimensional because there'll be many dimensions to virtually any system you try to describe in the real world and you use simulation and visualization to extend what you can do with mathematics alone and this has got much much more powerful over the time than it was when it was first invented now there's there's many programs available at least a dozen and there's many there basically
4:01 three dra broad groups what I call process control or engineering sociology and management um are dominated by two called then Sim Stella and I think and quite a few not a lot but a number of economists are using using ven or Stella to build economic models as well then there's engineering and the dominant program there is called simulink who who use mat lab here okay well looked at Sim linol side mat lab okay Vis which is my
4:30 favorite and that's that's a lot of I've borrowed a lot of ideas from Vis extended a few of ideas as well and then you have in mathematics and trying to model large scale systems there's modelica which is actually a language compilable computer programming language which actually has in its core core code the capability to simulate systems so you can actually call a routine like math mathematic as MD solved to numerically solve a system of differential equations which you state
4:57 in code and that is then compiled and run at machine code level so very high performance and system modeler and X cost are based upon that X cost being an open source program and msky is open source and it's if I wanted to put in any of those categories it's halfway between engineering and Mathematics it doesn't have a lot in common with the sociology and process control software packages and strictly speaking it's at A Primitive level of development because
5:21 the entire budget so far is under $300,000 us which is Trivial any any other program that list would be have 10 to 100 times as much development going into it so it's really the the infancy but it still has many advantages over those programs because I've I've stolen ideas uh from all sorts of computer software packages not just uh from The System Dynamics world so the idea of System Dynamics is to represent a dynamic set of equations as a system
5:48 of flowcharts where the flowcharts give you the causal Loops that are involved in the in the in the system itself and a guy called Jay Forester who's just died at the age of about 98 I I think invented this whole idea back in the 1950s and the stimulus for it was he was approached by a factory that couldn't understand why it had such huge Cycles in its manufacturing process and what he worked out was that there's a a series
6:13 of time lags and a series of amplifications between different stages in production what you call the bull whip effect was occurred so if you imagine a bull whip if you're wobbling if you ever anybody ever cracked a whip ever done that okay well you're moving a handle a small amount and the whip is going up down like crazy he said that's actually built into the time dynamics that exist in the manufacturing process so a blip in the orders by customers
6:38 causes a big increase in the warehouse purchases which bumps up what happen in the Distribution Center which meant the plant goes crazy okay so that was the original impetus for putting this down in in a flowchart format and ultimately who's said of the limits to growth who's read the limits to growth okay the rest of you search for limits to growth PDF you can download it take a look at it it's highly highly relevant for the modern day particularly since
7:03 we're here at the inauguration of a new president who doesn't seem to take climate change all that seriously okay so the intention of this stuff is to model all the things in their classicals AUM don't exist Orum away because they think they've got a AUM them the way to do modeling in the first place which was true in the 19th century and I really must compliment neoc classical Economist for keeping the 19th century alive for so long I'm trying to drag them into the
7:25 21st century you don't have to assume equilibrium to model a complex system in fact assuming it means you can't model it properly now one thing I've got this is actually feedback I got from u a good friend Tom Ferguson who's the research director for inet inet provided the first funding for Minsky and quote unquote why is Minsky so difficult to use well my answer is because economists aren't familiar with System Dynamics in the first place if they could see some
7:49 of the other programs they wouldn't be calling Minsky complicated it has a better interface I think than the industry leader venim I'm going to show you here uh a model of what's called this this is you've heard of Loren's Butterfly Effect this is lorenz's Butterfly Effect done in in um in venim and if I simulate it which I Do by clicking on this button then you see the result that was exciting wasn't it told you a lot and to see the
8:17 equations I've got to go inside the box here and first thing I get to do is is to say what shape it has well let's make it a hexagon why the hell am I changing formatting as my first step in defining an equ system well there's the equation Tab and the equation has DX or the initial value of initial X and somewhere inside that I've got to set up a set of equations okay all done in text so
8:40 that's that's venim and that's the industry leader now let's take a look at the same thing in Minsky okay and so what Minsky has and I can actually read the equations to you by working from system System Dynamics uses integration rather than differentiation as the first step because integration is a smoother operation than differentiation if you think about the shape of a hill it
9:11 changes quite radically the area beneath the hill if you you take like a if you take a cutway of the Hill the area doesn't change anywhere near as fast as the slope of the Hill does so integration is a smoother process so reading this backwards it says X is the differential of X is equal to a MTI y - x so the equations are obvious on the pallet as well and if I simulate you see it in real
9:36 time and you can see why when you hear about the butterfly effect the reason is when Loren simulated this in three dimensions or two two or three dimensions for the first time the pattern of the cycles of the unstable system looked like like the wings of a butterfly so and design divining a model in Minsky is quite straightforward if I wanted to make that time something else like I'll show you that a blank count rather than damaging this
10:02 model but it's very easy to define an equation on that screen so that's that's the the starting point of Minsky it's actually easier than most of the other System Dynamics programs to program in it's the Paradigm that's new to economists that they have to really learn because they've learned the wrong stuff equilibrium modeling difference equations and so on and the the Insight coming out of that little model there is is two is twofold that is an incredibly
10:28 simple system in fact I can show you the equations you won't necessarily see them from the back of the room this is saying DX DT a * y- d ydt = x B- z- Y and d z DT = x * y - c * z x y and z are the variables a a b and c are the parameters and their constants pretty hard to imagine a simpler model than that that's the behavior it gives you highly
10:56 complicated so complex Behavior out of simple systems is the complex system insight there and that's just again showing the equations in more detail so if you wanted to simulate a model again of course I'll give you copy of this presentation what you got to do is build that system and you'll have a demonstration of lorenz's model now there is plenty of advances in Minsky's user interface so I'll show you those just start a new model here okay and make it a bit smaller so I
11:27 can have that on screen while I while I talk so let's just shrink the window a bit okay you saw for a moment there what I've what I've shown you in that previous simulation is the graph is actually embedded in the canvas and it simulates in real time only one or two of the other programs does that this Sim does uh secondly you can directly type it variables on the on the canvas if I want to type the word Fred for
11:55 example then I've typed a variable onto the campus on onto the the canvas which is your design Po and I can move it around which any of the other programs you can do the same sort of thing but you can't type directly onto the canvas what you do in other programs is click on something like a variable entry give it a name um and a value and then then you've got to place it on the canvas I
12:17 think it's just FAS to be able to do stuff by typing directly on the canvas if I type A divide by key I've got the divide by symbol and if I then say this is equal to another another entry uh then I wire that up and if I want to say for example there's a sign wave coming in behind Fred so Fred is actually the sign of something and you have t is as time uh then I do that and
12:43 then I can put the graph on there and I don't need I mean all everything actually even with the graph I've forgot direct the direct command is but you can type a graph directly onto the plet as well you don't need to go up here all the time I find it annoying going up and using having to click on a on a all the time to create stuff I'd rather work straight to the campus canvas rather
13:03 that's what's feasible with Minsky that can be done and then you've got your model defined in that sense and so I'll just stop that at that stage um and it supports formatting and another thing none of the other programs do this is quite unique in Minsky is they don't support anything other than raw text whereas the Minsky if I wanted for example uh let's get just delete Wilmer there and make this instead say Lambda and it's call it Lambda
13:31 um Z just for the heck of it then what I get is a Greek letter and I'll just make it larger so you can see that the Greek letter with zero subscripted it's we have we've got a little hassle in making superscripts because the superscript C is actually also used for a power but if I type that and put another title there then I can do a fair amount of formatting I'm Maring around there obviously but if you
14:02 want to use mathematical logic in your text of a set of equations and then simulate the same thing with the other programs you got to make a compromise and rather than typing the Greek letter Lambda you've got to Type L mbda and that's what you're stuck with with us you use latex commands and you can generate any formatting that latex supports I'll just uh do one more thing here I'm going to make that into a slider and uh let's see the value ranges
14:30 for the slider I'll make the start with a value of one but have a maxim between 10 and minus 10 with a step size of one and then while I'm running the simulation of course at the moment that's going to give me no out hang on dividing by zero that's not that shouldn't be working let's see what have I done wrong okay now I'm dividing it as I go so I I can change parameter values as
14:57 I'm running a simulation and actually I think I've caused a I've caused a crash there by going back to zero again so that's that version is going to crash pardon me while I should it down by the way we make next statement now Minsky is as I said has only had about $300,000 worth of development time so far every computer program has bugs you'll find Windows and Microsoft Word and Excel crashing on you occasionally didto for for Minsky it
15:24 will crash so my little advice is what I call the Tam Hall approach to computer programming save early and save often okay and save with different names all the time too in case anything actually does go wrong uh and we have a few other tricks that the other programs haven't done again this is easy to do it still amazes me the other programs haven't done it well let's just bring up another copy of Minsky here by the way if you
15:49 want to uh you can run multiple copies of Minsky you don't have to have just one so if you want two or three running at once you can do that um one thing that other programs if you want to add two numbers together so I have say Fred and Wilmer here for some reason I always revert to The Flintstones when I do this sort of stuff then if I wanted to add a third variable there with the other programs I
16:12 need to add another input to the plus block but what we thought well there's no point in that because um why don't we overload the operator so if you want to add multiple ones together you just keep on going and so it supports overloading in that sense where it makes sense okay equally for plus and minus divide by etc etc that's that's feasible there and the most important thing this is why I developed it in the first place you can
16:43 build stock flow models of financial system using what we call Godly tables and um that's just I think in overall the on the left hand side here I've got the interface for for um this xost which is a of simulink notice that thing there saying that that says plot that's an icon you don't actually see the plot in the in the model itself and it doesn't change on screen when you change the model you're paying if you
17:12 buy if you buy mat lab commercially you're paying at minimum $2,000 for that okay so for some reason they have never got around it's been around for 20 or 30 years they haven't got around to making graphs run dynamically on screen and that's the interface to Stella which I think ultimately comes a bit of a mess very very hard to read and actually talk about them being spaghetti diagram so Minsky is not hard to use uh it's the
17:35 it's the it's the Paradigm which is a new thing for economists to learn and lots more that I want to do if I can get fing that's why I W this little window up there that that links to my website and link to get funding from the public to keep Minsky development going because about half money I've got so far has come from the public uh rather than from I've got no money at all from Government
17:57 funding by the way for Minsky th far any government anywhere uh what Minsky can't do is wonderful stuff like this yet there's lots of fantastic software that exists out in the marketplace including things like this one uh which lets you design an environment and then see how people will move in that environment all this sort of stuff now is involved in designing offices designing hospitals and so on so Minsky doesn't support that it's by no means the ants pants and
18:24 computer simulation but it does things which are the the main advantages capacity to model Financial Dynamics I'm going to take you through uh that after I show the basic Paradigm as well like we call them Godly tables for obvious reasons I want to keep when godley's name alive so making it part of a software package is a pretty good way to do that and what you can do using double entry bookkeeping is maintain that very consistent view of financial flows and
18:50 you can Shuffle around the definition of a financial system in a way that's much easier than working with the flowchart Paradigm for the same purpose if I get the funding ultimately probably from the public rather than from government grants then I'm going to make it much easier to Define parameters and constants than currently occurs I'll show you one in a moment I want to have two-way design so you can go from mathematical equations to the flowchart
19:12 and back at the moment you only go from the flowchart to the mathematical equations support Vector operations so we can multiple model model input output Dynamics which currently can't be done I want to enforce dimensional analysis so we got flows of money and flows of goods and stocks which are dollars and flows which are dollars per year I want to enforce dimensional analysis in the program so makes it easy to follow the logic and I want to be able to clone the
19:38 model of a single economy to automatically create a world model with export and import Dynamics Financial flows and labelings all automatically provided by the program just as soon as you say want to have more than one economy in the model and also we are we do import data right now and there there's some I'll show you an example at the end of Incredible use of Minsky to uh use empirical data as part of generating a model of a of an actual
20:02 economy but I want to be able to do that on very Grand scale import large amounts of data and then use the data itself to generate the parameter values for simulation that's a long way in the future and then most importantly get to the stage where we can have this as a have multiple users operating at once on one Minsky model using it to try to control a virtual economy and see what happens when they do things like trying
20:25 to run a surplus I think we it's about time we got politicians to experiment on an artificial economy rather than the real one be much safer that's the intention overall now the flow chart side of Minsky I'll illustrate using a a mo reduced version of the of the Goodman model done by Michel and Englebert stockhammer from my own University and that has if you you know this expression dydt over y so the percentage
20:55 rate of change of Y is 1us W percentage rate of change W isus C+ R * y well to put that in sort of System Dynamics form what you first do was express it as a pure differential equation so I multiply across by Y and W that's all that's going on there and then you integrate the operation so what you do is say having defined d ydt y * 1us W you say Y is equal to the integral of Y multip 1us
21:23 W and then you put that in a flowchart form and then you simulate now when you do that you're going to model it looks like this so I'll just bring up that particular model this is a I call it they call it a Pudo good model and I agree it's pseudo but it's very a very nice simplified version of of good one so let's just make it a bit larger so you can see it drag it over here uh and
21:46 then do your simulation and you get the classic good one Cycles coming out of that okay so that's the basic idea with Minsky now it's Pudo because you're not actually modeling the structure of the economy you're modeling y as an index of GDP rather than actual GDP you're modeling W as a wages index rather than W as the actual wage and the key AR thing in System Dynamics is use it to show the underlying structure and most
22:17 of the time that will lead to cyclical behavior in fact as I said that's how the Paradigm came about why does our Factory have cycles and to explain it that's how the actual Paradigm is in Ed by Forester in the first place and you start with an incredibly simple model and you that add complexity over time so let's actually build that the goodwi model in System Dynamics form so I'm going to start off by having investment
22:40 and by the way I've seen a lot of people write that the the goodwi model is a is a supply driven model investment is part of demand okay all it's it's it's the Goodman model has a specification of investment it doesn't have a specification for consumption therefore consp is a residual but it's possible to extend it and include stock Dynamics Mets cycle stuff uh by specifying C and then stocks become your residual rather than consumption being the residual but
23:10 it is a demand driven model so investment determines Capital stock Capital stock determines output output determines employment the rate of employment determines the rate of change of wages profit is output minus the wage Bill and all profits are invested actually I'll build that I think I've got hope I've got time to do I'll start building the model here let's just um go with a new system again not save it so to Define that model I start saying
23:38 here's investment and I'm going to Define that is later the Amper sand looks like an integration block so I just type ersan to get the the first integration block there and I'll call that K and give it a value of 300 let's make it a bit easier to see that you guys are working and then say Okay investment determines Capital stock and then if I divide Capital stock by output by output accelerator divide Capital by output I get
24:12 GDP this is a bug we haven't got over it yet but you notice that when I first write something on screen it's even though I've got it magnified it comes up in small size first of all divide output by population Sorry by by labor productivity ah I'm trying to move it okay there we go and divide the two and I'm going to have the employment rate which I I'm going to use Lambda for notice the circles there by the way
24:44 if you click in the circle you drag out an arrow to connect to somebody else if you drag click elsewhere you move the object around that's the basic design idea there and if I then subtract from I've done this so many times I think I can do it pretty rapidly uh just subtract a a rate at which workers accept no change in wage rises from the actual level of employment and multiply that by their reaction the slope of the reaction
25:10 function they have and then that's the modply block then that's my Phillips curve so I'll just call this P sorry pcore uh H for Phillips and then I need to that's going to give me the wage rate so I need to put another Block in there for wages let's say wages start at uh one unit of
25:43 wages um I now need to take a copy of that variable which is the wages so the rate of change of wages multiplied by the current value integrated is the wage rate that's the basic logic of a it's basically stating out a percentage rate of change function so I've now got the wage rate If I multiply the wage rate ah and I now say I've made a mistake here I've got labor productivity divided by um output divided by labor
26:11 productivity is employment rate no it's not it's the employment level so I'm just going to grab that lot group it move it sideways uh that's another bug ungroup it a got a bit of space there delete that wi oh damn I shot myself there didn't I was still affecting the group okay I'm going to stop at that point because it's I've now embarrassed myself on that one but what I was rapidly trying to design I think
26:38 you see how fast I was getting there is that model which is the Gooden model let's take a look at it make it a bit larger so that's your basic wooden model this time properly includ cluding labor the the employment rate dividing by population and hit run and you get your good one Cycles so without making a stuff up like that I can design that model in about
27:09 two minutes so it's fairly easy to work with much easier than people realize because they're just not used to using the Paradigm yeah you said something ACC not yet not yet it's coming up okay now so there's several tricks that make it much faster to design in Minsky than other software packages typing directly onto the canvas is much faster than having to go and click on a plette all the time um being able to write Greek characters which you
27:39 can't do at all in the other programs using underscore for subscript and the carrot for superscript which the moment only works in dialog boxes got to fix that using curly brackets to format text that's these are all latex commands and being able to curve the lines which is ridiculous ously simple um as you see that particular line as soon as the blue dot turns up you can curve the line if you if you need to for for appearances
28:07 sake and once you've done it once you can do it again and you can also it looks like a miss you can straighten it once more all simple things to do and if you want to get rid of a line just highlight it when you see the blue dots come up you can delete it so all these things are fairly straightforward in the program it's also designed to try to work out with the nearest pointer that's
28:27 not working all that well this time let's just try once more ah have I generated another bug ah pardon me there's a multiply window you can actually change the rotation all these things as well that's what was getting in the way there let's see bring it down release okay I've stuffed that one up as well this is getting oh that's why I'm trying to Def Define why it's actually W wouldn't let me because y was already
28:54 defined over here okay so now got it there now I can do curve it around and make it larger I just want to make a bit of space to uh make R to put the graphs in the middle as you see I tend to do there so and S I can delete something using delete key that's all incredibly simple stuff most the other programs don't make it that easy to make the changes so designing a model is pretty
29:22 easy but you've got to use a right click menu quite a bit by the way so get used to using your right mouse key now I'm going to add some realism to that I've got in that model at the moment that all profits are invested okay and I've got no depreciation so if I had depreciation and population growing over time and labor producty growing over time uh I can do that by simply saying well this variable here which is
29:47 currently a constant with a value of 120 if I add an integral ah how do I get W out of that Harding W let's just go back let's delete that one and add integral to this one okay I found another bug that's supposed to be finding it adding to n not to W I'll try this one
30:19 to a that worked so I'll just go with that so I've now made a which is labor productivity something which changes over time if I give a value to Alpha say 0.15 I'm saying the 0.015 I'm saying productivity labor productivity grows at 1.5% perom so you can start with a simple model where you have a constant
30:50 population and constant labor productivity and change it I'll just actually see how that works made a few oh there we go so I've now got the impact of rising labor turning up in that very simple model so that's the complete model if I've done it properly is here and now what I've got is depreciation over here so concentrated depreciation reducing the value the real
31:21 Capital stock and I've got labor product it growing over time which is down the bottom here and population growing over time very easy to add those elements to the model what I've done here is now because I've now got the wage rate Rising that's why these spirals are occurring here because GDP is rising and the wage is rising as well so you want to work in ratios I can just say well let's
31:51 actually Define a new variable uh which is going to be W / y so take a copy of that variable I'll just whack it up there for this for the heck of it a copy of Y divide wages by GDB P which now gives me the wages share of output call that Omega wire that up delete that and just
32:23 for the heck of it wire it to here and then having why there's a bit silly let's actually just go for for Lambda which I have defined up here copy of Lambda drag it down here by attaching I haven't actually explained this bit yet notice all the graphs have got four entries on the left hand side four on the right and eight down the bottom if you don't put anything down the bottom you get a simulation over time if you
32:52 put something on the the black on the vertical and the black on the horizontal you get an XY plot again trying to make it simple to design a complicated graph and you can have to have up to four XY plots on one plot and of course you've got two multiple axes X and Y axes there and I haven't got any labels there yet but I can easily by choosing options have uh put a label there and have uh what this is Wages
33:21 share and this is employment rate so you can then document fairly easily now what's Minsk is doing behind the scenes is building a system of equations for you and that's what it actually simulates as well the entire block of equations and from the fileman if you want to actually document what you're doing you can export that out in latex who uses latex here anybody good okay
33:51 supports latex and if you don't use latex if you use math math type in word you can cut and paste the equations directly into a word set of equations which is what I've done most of the time here you can also export the model to matlb so you can actually build a vis model visually in in spsky and then put it in matb for much higher level numerical analysis and you can export results of a simulation to a CSV file so
34:16 you can graph using other programs if you want to as well so it's got a lot of features for program that's only had about 3,000 hours of development time now one little extra piece of realism here Goodwin had capitalist investing all their profits now the way that I built my Minsky model in the first place was say well it's not true they do that they invest more during a burm and less during a slum and then so I I'm going to
34:40 add now an investment function just as a linear function of the rate of profit and I do that and nothing appears to change anybody tell me why nothing has changed so far apparently in terms of the qualitative dynamics of the model it still looks just like the Minsky the good one model beforehand why hasn't it changed it's still two dimensional I've added another cause there but I haven't add another differential equation to the system and I'll show you why that's
35:07 important in a moment now of course the insight to bring turn G to Minsky was to say Well when capitalists invest more than profits they borrow the money they borrow it from Banks and then of course uh they pay off debt when investment when design investment is less than profits and they've got to pay interest so you put that together and again that's a very simple extension of the model all I've done is take the same structure I've got here for the
35:32 Phillips curve and have it down here for an investment function get a rate of profit have a rate of zero rate of profit at which capitalists just invest what they earn uh and then a slope saying if they get you know profits 1% High how much more do they invest modify the investment function by output you get gross real investment if gross real investment exceeds uh actual profits there's a change in debt and you got to
36:00 pay interest on debt so you subtract interest on debt from output and there you've got the basics of a model involving debt inside then now you simulate this model and notice what's happening the Cycles in employment are getting smaller hey wonderful got a great moderation going on keep it going for a while and you have a breakdown
36:32 so a period of apparent moderation precedes a breakdown now that's what we're getting out of this model as simple as that and with no understanding necessarily of miny just saying let's add another element of realism to the model we're getting Minsky's Financial stability hypothesis yeah you turn it into a yeah of course of course it's just that's what I did when I when I first did my Minsky paper back
37:04 in '92 published 95 um but the the observation here is incredibly simple behaviors give you complex outcomes out of the structure of the model you don't need complex Behavior by your entities to get complex systems outcomes in fact it's pretty Kella who's probably the lead was the leading developer of nonlinear dynamical models for the last 20 or so years Carl used to insist using linear behavioral functions wherever
37:34 possible because said therefore you your your nonlinearities were intrinsic nonlinearities they come out of wages being multiplied by labor to work out the wage bill for example two variables multiplied together so my my preference is to start with the linear and then add linear after that where're again simple but realistic nonlinear functions so the basic shape is that Phillips argued for back in his papers before the the Philips curve empirical paper that there's a nonlinear relationship between
38:04 employment rate and wage demands and I put a nonlinear relationship between profit and investment but a linear one gives you the basic Cycles all the time yeah but what's actually it's actually explain it the interesting thing about doing a simple model like this is uh you have to it gives you a chance to say where is this complexity coming from because if you look at Minsky's explanation of why there'd be a
38:35 break down a large part of it talked about increasing the interest rate now I've got a fixed interest rate in that model so I can't explain it that way you also talk about Ponzi investors all the investment in that creates factories there's no Ponzi investment going on so I'm still getting the basic idea of a breakdown and I'm getting this amazing phenomenon of a moderation before the crisis so stripping it down to that Bare Essentials the answer to why the cycles
39:03 of curame income distribution and that is that if you have as the boom goes on wages employment uh Rises employment rate Rises wage demands increase more of the output goes to workers than capitalists expected fundamentally uh more also goes to Bankers because they've got interest on debt they accumulated during the burm so the bo actually results in having less uh less profit than they expected
39:34 because of income distributional effects therefore they reduce investment and therefore the cycle turns around so it's actually your explanation for the goodwi cycle and for the Minsky breakdown is all about the boom itself changing the distribution of income so you don't need interest rate if you add the interest rate changes in there you're being more realistic but you don't need them to get the basic cycle that's why again it's worth starting at a simple level and adding
39:59 the complexity later so and this is Minsky's basic Insight now it's crazy in some ways that something is regarded as as deep as Minsky which clearly it was Minsky's work was Drops Out of so simply out of such a simple model and equally there's an Insight that Minsky didn't have and that is that a great moderation precedes the crisis which is what we empirically observed in the 2007 and if you look at the data
40:30 for the Great Depression even though it's much rougher data the same sort of phenomenon is going on there as well so rather than a period of of moderation meaning oh great we've solved the the Burman bus cycle if you're ignoring level of private debt you're getting the beginning of a breakdown so it's a simple model with complex behavior and that's really why it's such so important to work in these nonlinear non-equilibrium way of thinking and the
40:56 nonlinearity in that model are entirely uh inherent you can't get rid of them it's not because I've got nonlinear behavior that I get nonlinear results I've got linear behavior and nonlinear results okay so it's inherent nonlinearities wages times labor determining the wage Bill and complexity arises in models with three or more Dimensions uh and the three dimensions of that model the wages share of GDP the employment rate and the debt to GDP ratio so adding in the third dimension
41:24 is what makes the behavior so very different and there's a technical issue going on there uh which comes out of something you can learn if you learn the properties of ordinary differential equations because if you with system of differential equations you can work out the what equilibrium May apply all the rates of change is zero and the stability of properties of those equilibrium mapped to the properties of polom generated around the equilibrium and in reduced form the good model has
41:52 two system States therefore with those two system States you get a you get a quadratic my model adds in the debt ratio therefore you have a cubic equation third order polom now the properties of a third of polinomial are much more interesting in the dynamic sense than 2D because if you have a a 2d model then it's either got real it's either got two real or two complex AR values and the real Parts can be positive or negative maybe one's
42:22 positive the other's negative so I get a saddle that sort of thing that's the end of the complexity of 2D model the 3D model there must be one real Aron value for the simple reason that a cubic equation from between Infinity to minus to plus infinity must cross the xais so therefore you're going to get one real argon value which might if it's negative you talking about a stable node if it's positive you got an unstable
42:48 node now the other two can be real or complex if they're complex and the real part of the complex in ion value can be the dominant ion in the system you're all used to the idea of values and so on okay it's a number to represent whether the system is expanding or Contracting space okay if you have a a three-dimensional model it's like me this an object here am I throwing it away or bringing it closer to me if I'm
43:16 throwing it away the dominant ion value is positive if I'm pulling it towards me the dominant ion value is negative so what you get with the model like like the the good model is that this if you this this is this particular polinomial crosses the xaxis three times so it has three real igon values so all it's all its terms are either negative which are pulling towards or positive pushing away but no cyclical Behavior this one as I
43:44 said it necessarily has to cross the xaxis so there's one realon value in there which happens to be negative in this particular uh cubic but the two complex ion values can be can have a negative real part which means they attractors a zero real part which means it's purely cyclical system or positive real part meaning they're repellers but they can be bigger or smaller than the the the real ion value so you get quite a wide range of potential outcomes in
44:12 the m in my model you can Define two effect there are three equilibria but two of them are significant one is a good equilibrium where you have a positive wage a share of GDP a positive employment rate and a finite debt ratio and the bad one has zero wages zero employment and infinite debt ratio that's a black hole of debt effectively um and of course I don't have if you wanted to make it more realistic it add bankruptcy as a way of
44:39 reducing the scale of the back black hole I haven't done that yet but somebody could easily do that now when you analyze it you find the realon value of both equilibri negative which means that they're attractors along that particular axis um but the complex one of the the good Aron value has zero real part for low levels of the reactivity of capitalist to the rate of profit so if I come down here and I change the slope
45:08 value of the slope Factor I've got here and I make that five rather than 10 and I then res simulate this model I get stability you can see employment rate converging debt ratio becoming constant so on and so forth okay if I increase that value what I'll do this time is make I'll make a copy of that whack it up here make it into a slider I've got to stuff around a bit unfortunately because there are some
45:39 bugs in the program so as I've mentioned I'm trying to make that into a parameter I can't do that for some reason the maximum value of ah maximum value of 10 minimum value of say one step size of one okay so I can now show that model with a low level and then increase the
46:11 reactivity of the profit function we'll change it back again so you can test out a model in real time in that sense but you can see the change from the low reactivity of capitalist to the rate of Prof to the rate of profit in terms of how much they invest of high reactivity takes you from a stable to an unstable system so the model has Cycles to equilibrium for low values of the reactivity of capitalists approaches
46:42 equilibrium for high values but then diverges as it gets nearer which is strange Behavior you cannot get that out of a linear model it's called a strange repeller and the model may have what's called a a hidden attractor I've linked to a mathematical AC paper on the particular types of models to that part of the presentation so you can dive in further now again this is an incredibly simple model but it captures the essence of the financial crisis in 2008 uh there
47:09 was this period of apparent Cycles to equilibrium followed by a breakdown great moderation followed by a breakdown that's what we saw now in the model itself what I'm getting is the as you increase the reactivity of capitalist to the rate of profit the complexon value goes from having no real part to positive real part and that give you the overall breakdown again I can explain that in more detail if you uh take a look in one my next book and also
47:42 um uh some of lectures now that's all stuff that I've shown you so far you can do with any other program um so what I'm going to do now in a moment I'm actually do you feel like a break happy to keep on going okay okay um that's can be done in any of the other programs me gives no intrinsic Advance over the other programs in that sort of modeling what it does add is the capacity to model the financial system
48:11 using double entry bookkeeping and the difficulty of doing that with the flowchart Paradigm is varies in terms of How likely you to make this mistake depending on which software package you're looking at but you can give the wrong sign to one of the flows you can have something adding to one flow adding to another rather than adding to one and subtracting from another um and you also want to show the same entity as an asset for one agent and a liability for the
48:37 other so effectively to account for any transaction you should actually show it four times two for for from going from an asset to a liability and two going from liability to an asset now if you do that using flowchart diagrams you get a spaghetti code mess you're looking at a bowl of spaghetti and are all the wise and stuff like that it's very very hard to work out what's going on whereas there is there has been a graphical user
49:03 interface for showing Financial flows half half a millennia now it's called double entry bookkeeping anybody here learn accounting I didn't i' learn it the hard way by designing Minsky so Minsky implements the tabular approach as an alternative way of building those differential equations now if you look at the history I recommend this book by Jane gon white on how dou entry came about and what its impact was socially and economically but the essential concept is you record uh the same
49:33 transaction twice on each row in a ledger book to make sure that you sum each row to zero therefore you've properly recorded the transaction once is a positive adding to some particular account and once is a negative and therefore a properly accounted row sums to zero that's that's the check that accounting gives you they have three types of accounts assets which are a positive for your whoever you're looking at liabilities which are negative and finally equity which is the net worth of
50:03 whatever agent looking at by subtracting liabilities from assets and the fun what they call the fundamental law of accounting is that assets minus liabilities equals Equity sometimes you see describe as capital and a minus Lal Z it's your basic equation well that therefore means Aus lus E equals z so in that sense that's that's the fundamental role you you must do that accounting every time otherwise you'll stuff up as economists have with things like a money money multiply model
50:33 account accountants use a Convention of debits and credits uh and Minsky can do that as well but I find it quite confusing so the default convention we've developed Minsky is that assets are shown as a positive which makes sense liabilities are shown as a negative again that makes sense but Equity is also shown as a negative now that's counterintuitive but what it we do it that way is because adding up every row therefore should give you zero so if you
51:00 have a a minus lus equal to Z when you def find all three as positive uh then if assets are defined to be positive uh and both liabilities and equities are defined to be negative then if you sum up a properly recorded row it's going to give you zero okay A Plus L plus C will give you zero so that's the basic convention and it's checked with every entry by Minsky so let's actually I'll show illustrate this one
51:26 that's start with a a new file this will complain about me not shaving changes that's okay notice the bank AR on there so to access that click on that drag it somewhere click on put it on the campus we can make that one larger as well and you either right either double click to bring up this table or right click and choose open Godly table to bring up a table now at the moment I know you can't see the text on screen
51:55 we're using a third party spreadsheet right now we can't magnify or or Shrink its size it's just that one stock size so that's another thing I want to change with with more funding uh so there's the table and notice that when I click uh when you look at the table first of all um it says flock flow variables with a V flip pointing down to show they're shown this way stock variables with an arrow pointing to the right hand side to show
52:23 they're shown in the columns and then this thing currently says is no asset class now what that is there for is that actually reflects my learning process as I develop Minsky because when I first did this stuff I was using single entry tables and therefore there's a little checkbox on the left hand side says double entry now we only brought in Double Entry power the program later in its development because I hadn't realized how important it was at the time so I've learned it by
52:52 building Minsky and we used to also had arbitrary column types we did actually specify assets liabilities and Equity so that little thing saying no asset class when you click on it it gives a drop down saying make it to an asset a liability or Equity now what I want to do um in future versions is first of all make enforced Double Entry no choice okay make it necessarily double entry and then pre preallocate one column to
53:23 assets one to liabilities one to equity so you have all three turning up there and if you add an additional column in any of those boundaries you get an additional asset or liability or perhaps an equity um at the present however what you've got to do is choose which account you're talking about from a drop- down menu so I'm going to define the simplest possible banking sector with loans as an asset for the banking sector deposits as
53:47 a liability I'm going to I'm going to name the Equity Bank so I'll just do that quickly so if I again you can't see I'm sorry I can't show you the scale L Lar so call this loans doing stuff live you always make typing mistakes you're looking at the screen rather than the keyboard notice these plus and minus key here when I click on plus I create a new column it defaults to being an asset but I want it
54:11 to be liability so now say type deposits here and then click again defaults to being liability make that into equity and call this Bank having done that if I go back to Minsky you now notice there two columns there should be a third it will turn up at some stage I'll just shut that shut that down for a second ah why is Bank not turned up I I'll add some more data let's say there's loans of
54:39 100 and deposits have shown as minus 90 and Bank Equity of shown as minus 10 having done that now you can see loans deposits and Bank were all turning up there as down the bottom of the bank those are the those are the flows that exist the stocks rather that exist in the system I don't know why I've got an error message there I made some sort of mistake okay so you get that sort of set
55:09 up to begin with and the thing at the top there the green the black the blue bit I just bring it back up here that's just a label so we're not using it yet but the intention is to ultimately use that as a as a as a class statement effectively so you've made working class class capitalist class Bank class that would be defined by by the element you gave to the to the icon here we also make the
55:34 icons different so households and factories and things like that rather than just Bank icons as we have at the moment and notice the summing to zero of not just the each row in terms of the flow operations but also the initial conditions again that's something which is very hard to do with a set of the standard differential equations or the flow chart because you could you you don't necessarily make them sum to zero you can make that error Minsky forces
56:01 you to get it right so the fundamental operations are lending repaying uh and sorry I've got the wrong labels there pay interest and repay debt so I'm going to enter those so back up to the table notice there's a plus key here that gives you an additional Row for operation so I say lend money and that's just a label at the moment too and type the word lend here and minus lend in the deposit column you can see Minsky's
56:28 added that to the bank icon and then pay interest of course the interest increases the equity to the bank so I'm going to go say int here and minus int over here and then repay loan so I have repay as a positive entry here because the positive entry reduces the size of the bank's liability and it becomes a negative entry on the asset side because it reduces the size of the bank's asset so now you've got those pre
57:00 operation this is another bug I've got to show this to Russell they they should all be the same size as these ones down here and they should be align so this is coming out of a a modification of the program we've just done a major what's called refactoring so when you write computer software like for example me learning about Double Entry okay and I've got all this code to support single entry and then you cludge and add a
57:22 double entry component to it but we also made it possible to resign um we could resize other objects before we could resize the um the umly icon got another bug just popped up there so I'll have to shut that one down for some reason I'm now caught in a dialog box I can't see so um as I said save early and save often there are bugs on the software and if you save it if you don't save it and a bug turns up
57:48 you'll hate me okay so save it beforehand so some sort of bug has got there and I don't know what it is and I can't see the dialog box I'll just shut this one down maybe I'll can if I tap around I can't see a dialog box turning up so some of the bug has popped up I'll have to I've lost what I've done there so I've got to I've got to choose another so I choose file yeah it's it's frozen
58:13 so I don't know what's going on that that's a serious bug I'll just find another window in a moment but that's once I've got it done properly and I've got this in another uh file of course there I've got your basic operations and you can see again those conventions positive entries increase assets negative entries reduce them that makes sense and it's obvious negative entries increase liabilities and positive ones reduce them it makes sense that it's not
58:42 obvious but what you're doing is adding a negative to a negative so you've increased the the magnitude and then negative increase in increase equity which is what's going on over here positive ones reduce it that's counterintuitive but it forces that rule that all row sum Z that's why it's done yeah function you would need for for your personal research is not included yet could you like program it your own
59:12 by yeah if you're you can write in C++ the code's open source so you can get in there and add the features you want um there's a there's a GitHub resource with that's managed get in touch with Russell if you intend giving that a try of course you can do it to your own version you can download the source code and change it in your own version if you want to commit back to the version We
59:32 distribute to other people Russell has to make sure it fits in okay definitely that's there and we'd love to have more people helping us do that um one other thing about the flows positives the initiator it's that's the source a negative is the recipient okay in virtually every case so what's happening behind the scenes is that set of differential equations is being defined by Minsky okay and the diff you know you know that these things going to be stock flow consistent
60:01 anything that increments one will decrement another and vice versa and also notice all these stocks are now shown is positives here there are no negatives there so a flow that increases a stock is shown as a flow so loan lending increases loans repayment reduces it lending increases deposits repayments reduce it um and what that does is you've got a mapping between a accounting view of the World which makes sense as in the design phase and a
60:30 mathematics view which makes sense in analyzing it later and then what it also does is automatically generate the multiple assets and liabilities view of each of each uh flow because if you take a look at the um I'll bring up another I another one here if I click on here and I'll just quickly call this let's say let's make this asset and I'll have Reserves and then
61:01 loans and then I'm going to have liabilities of safe firms and then workers and then finally the equity of the bank having done all that I've now got those should be five accounts there again there's a bit of a delay by going I'll put the initial amounts let say there's reserves of
61:32 100 and Loans of initially loans of 10 and the firms have got uh minus 60 and the workers have got say -40 and the bankers have got minus1 in equity having done all that if I now say it's bring down another banking icon here and call this say the uh workers and so they give me some assets here it's looking for things which are
62:04 liabilities elsewhere there's workers any liabilities look for an asset somewhere else loans or reserves that not an asset for them and finally click here and say what the equity is I can get the gap between the two at the moment you can see if I call this workers worker Equity then you notice that Minsky's got 40
62:35 here and not not a minus 40 over here it's telling me I've got to effectively give work workers value an extra value of 40 so in this particular model here I've got um the non-banking sector I've got the banking sector over here have lended interest and repay and Loans deposit in bank and then when I create a non-bank banking sector uh the deposits there are an asset of the bank they
63:06 liability of the banking sector over here they're an asset of the non-banking sector over here and the loans which were an asset of the banking sector are a liability of the non-banking sector so it's doing all that double entry matching for you and if I add interest as a flow for the equity for the non-bank sector that paying interest reduces the equity of the non-bank sector and notice I've got notice up here if I from that initial model I
63:37 showed you I had loans of 100 and deposits of 90 and Bank Equity of minus 10 which therefore means in System Dynamics terms positive equity for the banking sector to make that balance for the non-banking sector I've got to show that it's Equity is actually a posit entry of 10 meaning in real terms negative equity so the equity the B banking sector is positive the equity of the non-banking sector is negative that's
64:08 again the whole idea about somebody ask somebody's ask as being somebody else's liabilities but the whole system is integrated so that R has to oppose now that's one little thing I've left out of that um actually I'll show you a full model I put together very rapidly with all those elements to it so I have a banking sector a firm sector and households uh the operations I have here are lending to firms interest payments by firms debt repayment by firms deposit
64:38 interest paid by the banking sector to firms deposit interest by the banking sector paid to the households wages paid by the firms to the households consumption by the households consumption by the Banks that's enough to illustrate the Paradox of thrift why I've got the model set up because I've made SL is here for various things like the rate of lending the rate of repayment the interest rate on loans interest rate on deposits how fast the firm sector turns over money in its
65:04 account which generates GDP how much of GDP goes to workers versus firms or capitalists how much fast workers consume and how fast the banking sector consumes so I'll simulate now that at the moment with lending exactly equal to repayment so no change in the amount of money let's imagine workers decide they want to save more money so rather than having workers consuming on a on a a weekly basis they go into to consuming on a fortnightly basis and
65:33 what happens yes they get more money in their account but their incomes fall okay you got a slow rate of turnover of money enabling them to accumulate more of the money that exists so the amount of money in the household account as you can see here has risen but it's risen at the expense of incomes falling including workers incomes workers wages have gone down now that's a really hard concept to illustrate with standard um economic thinking it's a breeze to illustrate
66:02 that with msky and I can also illustrate set this out of the way lending creating money because of the rate of lending increases and the rate of repayment slows down then you get an increased economic activity and a growth in money in the system again very easy to illustrate with
66:34 Minsky and that is the set of equations defined by that model so again msky is self-documenting and guarantees stock flow consistency yeah so do you get put yourself yeah where you put the parameter this what I've what I've defined here like all these things are those are flows these are stocks but I want to modply them by various parameters so I simply Define them on
67:05 the on the plette and put make them slider so I can change them as the model simulated we have a chance to to link this link M to real data yep not well done but this little icon up here is the data icon so if you click on that whack it somewhere on the screen and then uh choose choose import data you can specify the CSV file you're importing the data from and then have and you can
67:32 actually derive your parameters from the data if you want to I'll show you an example of somebody having done that at the end of the talk so you can actually do mixed um Godly table and flowchart modeling as well but at the moment it's not enforcing stock flow consistency between those two paradigms so you've got to work a bit harder to make sure you get that right again at the later stage I want to make sure that we en Force any
67:56 stock flow interactions between a Godly table system and the flow chart is correct but I'll quickly show you how I can use a model like this you all know Krugman and his loanable funds fixation okay well he and egson published a paper in the quarterly Journal of Economics where they had in in the appendix a really little model of a bank where the bank was a one sector Loan in in the consumption but of a dreadful bloody piece of crap but
68:24 anyway I thought what I'd do is take that model and say well strictly speaking if you actually that was actually the structure of capitalist economy if banks were just in majories you'd be right lending wouldn't have much effect on the macroeconomy so what I did was take that silly structure they had there and put it in Minsky and what I have here is much what I've shown you for the model we just saw a moment ago
68:45 where I have lending repayment interest payments but because lending is going between the consumer sector and the investment sector in krugman's model the bank is simply charging a fee the bank's income comes out of charging a fee to the consumer sector for being the introduction agency between the borrower the saor which is the consumer sector and the borrower which is the investment sector I call this the Ashley Madison vision of banking now they don't actually screw you they find somebody
69:11 else is willing to screw you pardon me I'm getting so sick of people Krugman pushing that model as though it's a description of reality anyway having done that I can now run a model where I'll just take take you through the VAR stages of the model there's lending debt repayment interest payments paying a bank fee H consumption sector hiring workers investment sector hiring workers the investment sector purchasing consumer goods the consumer sector purchasing investment Goods the workers
69:38 consuming the bank is consuming the bank is investing okay all basic simple operations all defined down here using the flowcharts to make various definitions uh now if I simulate this model notice the growth rate of the economy is zero GDP is lining at 200 and the debt ratio is climbing there's no impact upon the level of money in the economy and if I have a faster level of lending the rate of the growth of the economy actually dips a
70:07 bit and if I have very slow repayment it dips occasionally but always turns back to zero and you got a rising level of debt in the economy but no impact upon GDP now if you then go to have a a Plunge in the level of Landing much faster repayment pardon me the growth rate actually accelerates and much slower lending growth rate Rises once more you have a Plunge in the debt to GDP
70:39 ratio no particular impact on GDP it's G up a bit but it'll come back down to 200 ultimately so in that sense massive changes to finance buger roll changes to macroeconomics and that's the vision that neoc classicals have now one of the beauties of Minsky is I can say well that's interesting model unfortunately it's not the real world it's not true that debt is an asset the financial sector so I'm going to not not true the
71:05 debt is an asset of the consumer sector Banks lend money so let's just delete this fiction that the lending is from One non-bank agent to another so delete that column and then delete all the so-called lending and repayment operations between the consumer sector and the investment sector and come across to the banking sector and say well let's not treat reality debt is an asset of the banking sector and notice this drop down here because I've deleted debt as an asset of
71:36 the consumer sector it's still sitting in the system as a liability the investment sector so it's currently an unallocated liability so if I click on this little down arrow here and I'm in the asset column Minsky goes looking for any unallocated liability click on D and it brings across the operations that currently exist and I've deleted the um the interest and the fee operations are still there but I've deleted one of the two entries so the interest payments are
72:02 made to the bank and the idea of a bank fee itself is a fiction let's delete that there's more changes I need to make to the system to make it completely consistent let's go back to the initial setting where I had loans doubling every seven years and repayment Haring debt levels over 9 years and simulated it and I've got a positive rate of growth a growing GDP increasing debt causes increasing money supply if lending happens more
72:29 rapidly the economy grows more rapidly if repayment slower the economy grows faster still and if you have a desire to repay loans and a slow rate of lending you get a slump and try making that comparison that rapidly with any other software package okay it just shows get your structural reality right and Money Matters and you can't ignore it the way the neoc classicals try to ignore it so that's the sort of thing I wrote Minsky
72:56 to make feasible to model very very rapidly and literally I can make a change from the neoc classical learnable Visions version of the world to the indogen money real world in about 30 seconds now I've shown you so far as toy models and that's all I really have developed Minsky so far but in the last year I've been blown out of the water by two masters one Masters and one PhD student uh the PHD student was a PO and
73:23 came across to a uh Kingston for a year to sit in our courses as he developed his model he's done a complete model of the portra of the Polish economy using Minsky I I won't show you his I'll show you the the Portuguese one because he's now doing a PhD with me so this model is a model done in Minsky including importing data to fit to various data data series uh of the Portuguese economy it's a
73:50 mess okay no argument it's not a mess the reason all this crap down here is because of definitions so all these things coming together here I'll bring up the model and show it to you uh hang on got
74:21 to so that's the model and this one here for example is defining the government deficit and the deficit of the financial sector um including debt from the rest of the world interest rate from the government sector public dead interest rate all that's being defined together you can actually Define it using a set of equations like over here so I intend tidying up the definition process so you don't have to have all these messy definitions here in a later stage but
74:47 just to show you the scale of this model he has a government sector households productive sector financial sector and rest of the world and all the uh double entry bookkeeping is being done by Minsky in the background to make sure that stock flow consistent so it looks like a mess at that scale I want to drastically improve what msky can do at a later stage but this particular model is a more more successful estimate of the Portuguese economy than the model of
75:18 the Portuguese Central Bank okay so it can go a hell of a long way that's the system of differential equations and not just the differential equations not the um not the um parameters and and and variables coming out of it defined in that model by by Pedro using minskin I simply don't believe anybody could do that without a t like Minsky just too complicated maybe using the The Godly L approach UNS maybe you could do it but I
75:50 I don't think it' get there so that's that's what Minsky can do it's going to do a lot more get more development funding so if you want to donate to that website of mine please do it'll go to Russell uh but I want to do a few few more issues now and that's I'll finish up at about half an hour you need to break it or keep on going okay change the topic how do you do
76:13 macro now the neoc classicals think you've got to work from micro foundations by the way liia Blanchard and I've been in touch with each other recently getting a pleasant conversation I'm pleased to say if You' seen my sacal piece in the review of Kes economics you haven't okay I'll bring it up later um but I sazed Olivia a bit and he actually wrote back saying he thought it was funny so I'm pleased about that it's a good reaction there's there's a chance
76:36 for a dialogue happening now at NE classical which I think is a positive thing after all these Decades of the two areas ignor what the neoc classicals ignoring us and in talking about why DSG were so unsuccessful he still thinks we can use them and he thought well you have to start from micro foundations where else to start from and this is a what uh a physicist called uh Philip Anderson called a constructivist fallacy and this guy this
77:03 is a real this is a real pH this is a real Nobel Prize that gets in physics he wrote a wonderful paper called more is different which of course I've linked here and I recommend you taking a look at and he said reductionism has worked brilliantly for science over decade over centuries but getting to where Sciences now they they have to go beyond reductionism he said one thing reductionism doesn't mean is you can con have a constructionist approach to
77:26 science he said the fact that you can reduce everything to simple fundamental laws doesn't mean you can take those laws and build the universe okay if it did an obvious question a biology exam would be please take these chemicals and create life okay imagine that' be a great question wouldn't it but that's what economists are trying to do they think they should start from the micro and be able to build the macro from it that's like saying a biologist is only going to
77:52 be a biologist if they can tell you how to how to make out of out of in chemicals which ISS and he finished up by saying that in each level of analysis whether you're looking at biology or psychology or chemistry or physics is is at least as complex ways of thinking at each of those levels which you can't reduce to the level below but that's what economists have been trying to do for one and a half centuries been trying
78:19 to turn macroeconomics into applied micro so I want to show a different approach let say let's derive macro from macro so what happens if you take a set of definitions that simply are true by definition the employment rate is the number of workers with a job divided by population the wages share of GDP is Wages divided by GDP and the debt ratio is debt divided by GDP take those and differentiate them with respect to time you get a system of
78:48 differential equations put in the simplest possible linear definitions you can so constant Capital output ratio labor productivity a linear Phillips curve a linear investment function what do you get you get the model I've just shown you and that simple model working from macro definitions find something neoc classicals couldn't work out so it's just wrong to believe we have to start from the micro level and build up that's why again the reason I'm in favor of the
79:20 system approach build the structure of the model the Dynamics will emerge from it and this is again the argument that Anderson finished with because you can have a hierarchy of Sciences they can rank quantum mechanics as the fundamental level and then many body uh physics coming out of that and then chemistry coming out of many body physics and biology coming out of chemistry then psychology coming out of biology and and so on you said it doesn't mean science X is just applied y
79:49 um psychology is not Applied Biology nor is biology applied chemistry and macro is not applied micro so that's my overall philosophical approach let's build macro to macro you don't have to go to the micro level the question or okay okay another issue and this is one of you now I've been working on this for a long time the role of credit and macro economics and that's really the topic of the next book I've written which is coming out in April
80:15 it's not a one-word book but it could be and what I do is look at the overall credit Dynamics for major economies in the world and say you can actually predict the financial crisis by collapsing credit which is what I'm showing there for America but credit is is I've always had the gut intuition that it's part of aggregate demand it's also part of aggregate income now this is ignored because neoc classicals like Krugman obviously so you
80:44 can forget about it because one person's asset another person's liability who cares that's about as deep as Krugman gets uh but post cians haven't Incorporated properly either because they say that uh rather than looking at the asset as liability liability truism they say expenditure is income as a truism and therefore where's the role for credit I've had fairly intense conversations on this front I think you're aware uh which end up seeing there's no role for credit
81:12 in demand or income but there is and it's simple to show it and what I'm showing here is a imagine you divide the economy into three sectors S1 S2 S3 and a banking sector with both loans so the B B assets liabilities and Equity view this is looking now at as a system of an economy with three sectors uh where sector one spends on sector 2 and sector 3 so expenditure by sector one and sector 2 becomes income
81:37 for sector 2 and sector 3 ditto for sector 2 and dto for sector 3 the bank having no role at the moment I haven't got anything coming out of the bank the negative sum of the diagonal is aggregate expenditure the sum of the off diagonal elements is aggregate income so there's aggregate expenditure what is equal to there's aggregate income what's it equal to that's where the idea of the identity the two comes from so that's a
82:03 world where there's neither borrowing nor lending what about a loanable funds World which is the world Krugman inhabits well you have if sector one is borrowing L dollars per year from sector 2 then there has to be a stock of outstanding debt capital L and the reason sector one sector 2 is lending to sector one is because of the interest payments that requires sector one to make the sector 2 so I can bring that onto the model and say that now sector 2
82:32 is lending a flow to sector one so that's the little L here minus L here plus L there and then sector one has to pay interest to sector 2 that's row time capital L here is a minus and row time capital L here is a plus I'm not using M accounting conventions anymore I'm just looking at positive and negative flows decrementing the amount of money in each those accounts if you sum up aggregate expenditure there's now a role for the
83:01 financial flows involved the financial flows are part of aggregate expenditure and that's gross by the way if I had Banks paying interest I'd also have deposit interest times deposits as well as loan interest times loans as part of aggregate uh expenditure there's aggregate income and of course it's the same it includes income of the financial sector so again the identity is still maintained now the real world indigenous money so whether you don't borrow from SE sector one
83:32 doesn't borrow from sector 2 or sector 3 it borrows from the bank and sector one pays interest to the banking sector so I now have a borrowing of Elder little L per year from the banking sector and paying interest of row times L capital L to the banking sector so now I've got the banking sector involved there if I add up aggregate expenditure I get the flow of credit as part of aggregate expenditure and add up aggregate
84:03 income I get the flow of credit as part of aggregate income as well so it's completely valid to include credit as part of aggregate demand and aggregate income and of course aggregate capital gains as well when you include that so this is the factor we've left out even postan thinking except implicitly in Minsky's own logic read msky carefully you'll see he's got a clear role for credit but he couldn't quite articulate it to the stage where
84:30 Rec convinced his his post and followers so what you've got expenditure is fundamentally monetary and there are two sources of expenditure turnover of existing money which is what what all the a A's B's and C's are um and new expenditure Finance one for one by Deb because when you borrow money you're borrowing to spend it and using that you can that's what I've done in the book to identify the countries that I expect to have a
84:54 financial crisis because they both have high levels of private debt so this is a level of private debt excluding 1.5 times JDP here and then credit being responsible for substantial proportion of demand over 10% of GDP equivalent for the amount of money coming from credit and therefore very susceptible to a slow down in the rate of growth of credit so that's another issue I want I would like you guys to build on because you are the
85:21 success system of people like myself and Mark Lavoy now okay okay Mark and I used to sit around tables like you're sitting around now never thinking anything about the role we' have later in life now we both found ourselves Elders of the postan community it's weird okay but we're delighted to have you coming along to S like this because you're going to replace us and really there wasn't necessarily a feeling that was going to happen even 10 years ago but now there's
85:49 such a community of you people out there you're the ones who are to carry this work forward so that's one thing left half finished I'd like to see some of you work further on it another one I've just done and I want this also I'd like to see this taken further is that we've had no theory of no no school of economics has properly included the role of energy in capitalism which is ridiculous because we try to model production as if
86:15 you can produce output using labor and capital alone and that applies to everybody neoc classicals poian safian marxians the whole lot um postans have outut okay that's that's that's animated in the wrong Point that's neoc classical obviously neoc classicals have got the cob Douglas production function where they have output you know smoothly combining labor and capital postans have output using fixed proportions and I've had neoc classicals argue to me that we can just treat energy as a form of capital no you can't
86:47 Capital means produced means of production yeah am effectively yeah but I mean everything from the lunch you've just had meaning you haven't fallen asleep yet during my talk uh through to the through the oil you might pour into your car to head drive off from here back home so every form of energy with no intrinsic role for energy in any economic model because energy is something you can't produce okay okay if you believe you can
87:20 produce energy You Believe In Perpetual Motion machines and I've had neoclassical economists come out with comments like that without realizing what's it mean energy can neither be created nor destroyed you just find it and that's the first law of thermodynamics uh but you can't produce output without energy so I was a guy called Bob s has built a like a generalized version of cob Douglas where energy is another input like capital and labor and Bob's a
87:47 good friend but I was never happy with that rendition yeah so um so if you to if you to put energy into looking ATS of enery yeah you look what what I'd like to see done obviously have renewable energy the stuff that's actually coming into this room cesy of the sun despite the fact that we're in England okay renewable energy like that stored energy
88:20 both solar both um you know fossil fuels and nuclear and include and of course you're going to run down the stored energy the fossil fuels and the nuclear you're not going to run down the solar so you'd have and of course using the solar has got different applications to using fossil fuel there's a huge amount of detail we need to do to incorporate ecology properly into economics and vice versa but we hadn't until last year come
88:46 up with an equation where energy played an essential role and the little insight that made me work out what uh you need to do do this is that the whole idea of Labor without energy is is a fast if you can find labor without energy you found a corpse and if you have a machine without energy it's not a machine it's a sculpture you won't move okay so the Insight was let's define GDP as useful work and frankly that's in
89:15 some ways that's the most sensible definition I've ever found and I just made it up last year GDP fundamentally means useful work so you can actually put in that sense idea of GDP ultimately is a manifestation of useful useful energy useful application of energy and Define labor and capital as ways of harnessing energy so your basic equation becomes GDP which is a function of energy is itself produced by using Capital as an energy extraction mechanism and labor as an energy
89:47 extraction mechanism you go back far enough it's slaves okay you come up to the modern days it's the Falcon 9 rocket but fundamentally in both cases we're using those means of exploiting energy we find in the environment so actual work by done by capital and labor depends upon several things how many units you've got now it's easy to find units of unskilled labor it's problematic as we know in this room to Define units of capital but
90:15 I'm going to stick with that to begin with you multiply that by the flow of energy harnessed by each of them so how many calories can you guys eat a day four or 5,000 calories per day I reckon how many calories can a machine consume if you go back to the days of of of James Watt not very many you go forward to the days of er musk a hell of a lot multiplied by the ratio of Avail
90:43 useful energy the energy you can actually exploit the energy you can't which is the basic idea that some energy has to be wasted because you're dumping into an environment with a background radiation level above absolute zero multiplied by how efficiently you're using that energy and that's your basic expression so I'm now saying GDP is a function of number of machines times energy uh per machine times the available energy to the total energy times the efficiency
91:15 and didto for labor as well and I can call this x l for X ratio for labor that's supp got the wrong labels that should be K there and should be L there so now I've got this expression and if I put that in the start with a neoc classical form which is the idea of constant returns to scale and multiplicative form I rework that what I get is the cobd ductus production function times the energy input for
91:44 labor times the energy input for Machinery yeah that standard me of energy in this yeah you'd have to use standard me measure so like I've talk about calories for work for for people and you know BTUs for for machines but you need to convert all to BTUs to do it sensibly but of course when you think about what's involved in labor um there there's your cap C Douglas function there's the energy for labor
92:15 now energy for labor your maximum amount of energy that labor consuming maybe maybe on average 4,000 calories okay that's a constant uh the efficiency the ex the available energy to what you can actually use as work may be of 0.5 ratio if you eat 4,000 calories you might to do 2,000 calories of useful work per day and your efficiency may be 0.5 so you can pretty much treat that as a constant okay and you raise it to a power well you know
92:44 it's going to be some number between 0 and 500 that sort of thing whereas for the machine the energy parameters for Capital are the the energy consumption per machine per day that's risen dramatically over time from trivial amounts back in the days of of what even more trivial amounts if you take it back to the days of um you know fantasies like the Game of Thrones um the and then the XK which is the available energy out of the energy put
93:13 into a machine and the efficiency their time varying they would have gone up and down at various times over history hopefully going up all the time but certainly you have to have a maximum below one and you're raising that to a power so what you get is this expression where I've just treated the labor components as a constant multiplied by labor raised to 1us Alpha time K to Alp time the energy exploited by a machine also
93:39 raised to the alpha compar C Douglas production function it implies that this which is What's called the solo residual when you actually empirically estimated is actually the energy contribution of machinery and looking at what a and cl put together they had this format where they used Alpha K to the alpha L the beta and e to the 1us alpha minus beta um for a start this form is more realistic it has one less degree of Freedom so it's actually harder to
94:11 estimate but you can actually put it in output per capita form and capital labor ratio form which you can't do the formulation you can also relate it to useful output per machine because if I divide all this by L uh I've now got this as a scaling Factor rather than a statement of marginal productivity and I can actually we don't know what K is okay that's one of the problems we don't know what energy per machine is but we do know energy
94:42 consumption by industry there are good stats on that so um so so your problem you're having right now is um in a sense you cannot um probably identified the measurement for energy in relation to Capital no not per machine but you can Define it in the app because there's empirical data on this if I say the sum of energy used by the industry sector it can be substituted by the number of two I've got number of
95:11 machines which we don't know that's a hypothetical thing because there's no such thing as a aggregate number of machines you know adding computers to to lasers uh but we do know the aggre amount of energy used by industry and in fact we know by each individual industry to some detail with the Department of energy statistics in America so I can substitute that in and what I get is this expression total output is some constant representing the energy input
95:38 for unskilled labor times employment to the 1 minus Alpha times the energy used in industry to Alpha times these two unknowns which are the ex relationship for energy and the efficiency with which energy is used so if I substitute the whole thing together I can also put it in per capita form divide through by so I've got GDP per head and I can now rearrange this and get this in terms of the employment rate so GDP per capita energy
96:10 contribution for labor times the employment rate times energy per capita again we have data for that times these two unknown factors over here so I've got data for this data for that data for this don't know these two put it together and here's what I want to fit there's GDP per capita in America since 1965 that's the employment rate since
96:40 1965 by the way this is why Donald Trump got elected okay that's how many less workers have got a job now and had it back in 2000 and 2007 okay yeah you speak up my if my hearing is bad yeah claiming
97:22 that I feel like kind of complex relationship among those I'm walking up my hearing is so bad at the distance your voice is hard for me to understand what I'm to say you considering en as a distinct can be a little I feel like because I see a lot of complex relation
97:52 among themselves labor inut labor capital and enery because although we cannot create energy we need labor and cap oh yeah to find it Rel resour resour have hierarchies of resources some resour some of them act on other ones to get other so there's that kind of complex inter
98:24 well that's what I want to do at some stage as well because if you think about shord critique of idea of marginal productivity of capital his critique involved production using labor and capital only solid critique because it shows the whole cambage controversy problems of the near classicals ignore but what I'm saying is if we we have actually to be genuine you have to have Labor capital and energy being used to produce next year's round both GDP and
98:51 capital so the redu uction s is done as incomplete we can do a reduction to energy as well and we can have the more specific the more we drill down the more specific that energy becomes you can actually have energy which is just mechanical energy or energy used in making Iron and Steel which is a different manifestation to energy used in making um tables okay so you can do all that sort of breakdown I think we
99:18 can actually that's that's that's a research agenda for the future which is why I'm showing the stuff right now take your point but it's to me it's a reason to go deeper into this but I want to show you what happens even at a very superficial level there's the imployment rate and now let's take a look at energy consumption per head which I was amazed to find peaked actually back in 1979 this is the act this is data from
99:41 the um Europa project but it's derived from the American Department of energy so there's been quite dramatic fall energy consumption per head which of course would actually be turning up largely in the efficiency stats that I don't know okay the efficiency those two efficiency components the ex relate ratio and the efficiency of energy I don't know those two unknowable but when you put those two together using that equation I go from a correlation I think of 0 59 for looking just at the
100:10 relationship between GDP per head and the employment rate and 62 at looking at GDP per head and energy alone to 78 in that combined equation so just with um um employment only at a correlation coefficient of 62 just with energy .59 with the 2.79 so it's very powerful even at that superficial aggregate level um but what I think you get out of
100:41 this is something that lets you look at the Dynamics of energy Innovation and finance and class struggle all in one because you've got the class struggle stuff turning up on the employment rate investment and Innovation turning up in capital and the amount of energy that those machines can consume finance and class struggle turning up in the whole thing so I think we can actually also link the ecology this way and that's again something which my generation of
101:08 of postans heterodox Economist haven't done it's for your generation to get it done and I think it can give us something which is incontrovertible the near classical can't argue against this without arguing against the laws of thermodynamics yeah can do me the same favor of my hearing sorry I strle how does it enter because that's in terms of the employment rate and if you think about what's going on technologically all technological change fundamentally has been cap labor
101:38 replacing now one reason behind that is because Capital if you can work make a machine that does the same job as a human the amount of energy that machine can harness to produce the output is far greater than the labor can can do so you think about the original spinning jenny that replaced a person operating with a spinning wheel one person one thread first spinning jenny I think was 16 threads together ultimate one's couple of hundred so the amount of
102:07 energy being used by the producer was enormous which made the in that class struggle sense kick the worker out get a machine that does it the individual capitalist benefits from a huge increase in their productivity the worker loses out but then ultimately the worker gets back in there again because you need work to operate the machines and income distribution Dynamics become important but the overall trend we're certainly at the very beginning of it now with robotics and 3D printing and so on it's
102:33 quite feasible to imine a world in which there's almost no unskilled labor okay there'll be unskilled labor not there'll be skilled labor in services but not in manufacturing which is huge change coming away in your lifetimes and we still haven't got this this I hope is the beginning of this away being able to analyze that so that's uh pretty much my presentation there I've whacked a few little exercises any wants to give it a
103:05 try those are some classic equations if you can't work them out I've cheated and whack the Solutions in there for you as well and if you want to have a real cracket using Minsky see if you can use it to show by running a government surface it's a really stupid idea which of the Minsky model is easy to illustrate that's it may [Music]
Summary
- Minsky is an open-source software for System Dynamics, enabling the modeling of complex economic systems through flowcharts and differential equations.
- The speaker critiques neoclassical economics for its reliance on equilibrium concepts and linear models, advocating for a more dynamic approach that incorporates money and nonlinear interactions.
- System Dynamics emphasizes the structure of a system over individual behavior, allowing for the emergence of complex dynamics from simple rules.
- Minsky supports double-entry bookkeeping, making it easier to model financial systems and ensure stock-flow consistency.
- The software's user interface allows for direct interaction with variables and equations, enhancing usability compared to other System Dynamics programs.
- The speaker highlights the importance of energy in economic models, proposing that GDP can be defined as useful work, integrating labor, capital, and energy into economic analysis.
- Minsky has been used to model real economies, demonstrating its capability to handle complex interactions and provide insights into financial crises.
- The presentation concludes with a call for future research to further integrate ecological considerations and the role of energy in economic modeling.