Tuesday, 12 February 2013

Minsky, and an Appeal from Steve Keen.

Computer program for building & visually simulating dynamic, monetary economic models. A vital tool for a new approach to economics
  • Launched: Feb 9, 2013
  • Funding ends: Mar 17, 2013

Why Minsky?

In late 2007, the world walked blindfolded into the biggest economic crisis since the Great Depression. Economists, who should have warned about it, were in general useless. The vast majority expected economic tranquility ahead; only a handful warned of the crisis—and I was one of that handful.
The key reason why most economists were unable to see the crisis coming is that they have convinced themselves that banks are irrelevant to how the economy operates, and too hard to model anyway. I've argued for decades that banks are crucial to how the economy operates, and I've worked out an easy way to model them in my program Minsky.
Minsky is my attempt to reform economics, and I need your help to make it happen.

Why Minsky matters—for the public

In late 2007, financial markets and the economy crashed. Five years later, the global economy is still mired in the worst economic crisis since the Great Depression.
Conventional economists claim that this crisis was an unforeseeable event. But was it really impossible to foresee something so big?
Queen Elizabeth posed precisely that question in 2008, when she visited the London School of Economics. Though famously discreet, during a briefing on the financial crisis she put this challenge to the assembled economists: "If these things were so large, how come everyone missed them?
The answer she later received was that "no-one could have seen it coming".
That is manifestly false. With a realistic approach to economics, this crisis was entirely predictable. I am one of at least 12 economists who did see it coming, and who warned of it before the event. In 2010, I won the Revere Award from the Real World Economics Review for being the economist who most cogently warned of the crisis, and whose work will do most to prevent a future one (Nouriel Roubini came second, and Paul Krugman came seventh).
I was able to warn of the crisis because I have extended the realistic approach to economics developed by Hyman Minsky, which he called the "Financial Instability Hypothesis". This led me to focus on the role of private debt in economic activity, and I knew that a crisis would hit as soon as the rate of growth of private debt slowed down. It did in late 2007, and the crisis I had been expecting began.
Slowdown in growth of debt began the crisis
Slowdown in growth of debt began the crisis
Unfortunately the vast majority of economists don't even consider private debt, because they follow an unrealistic theory—known as "Neoclassical Economics"—which attempts to model the economy as if it is inherently stable, and as if banks, debt and money didn't exist.
Don't believe me? Here's what Paul Krugman said in a web debate with me in early 2012:
"Keen ... asserts that putting banks in the story is essential. Now, I’m all for including the banking sector in stories where it’s relevant; but why is it so crucial to a story about debt and leverage?"
The same inability to see that banks play a crucial role in the economy lay behind the Federal Reserve's failure to see this crisis coming. Here's Bill Poole, the Head of the St Louis Federal Reserve, speaking at the meeting where the Fed decides monetary policy, a mere two days before the crisis began:
"MR. POOLE: My own bet is that the financial market upset is not going to change fundamentally what’s going on in the real economy... this financial market upset is going to settle out and not have major repercussions on the real economy, putting the housing part aside." (FOMC Transcript, August 2007, p. 57)
Is it any wonder that conventional economists didn't see this crisis coming? Their models ignore banks, debt and money, but they are essential aspects of a market economy, and they have to be included in economic models if they're to make any sense at all. Modeling capitalism without banks, debt and money is like modeling birds without wings.
The only official economics body that gave any warning of the approaching crisis—the Bank of International Settlements—made precisely this point in a recent paper entitled "The financial cycle and macroeconomics: What have we learnt?":
"Think monetary! Modelling the financial cycle correctly requires recognizing fully the fundamental monetary nature of our economies: the financial system does not just allocate, but also generates, purchasing power, and has very much a life of its own."
I am one of a handful of unconventional economists who are trying to do just that, and I want to make it possible for many more people—and especially young economists—to join us. "Minsky" is my way of making that possible. It is a dynamic, visual, monetary alternative to the static, clumsy "intersecting lines" models of conventional economics. I hope that it will help reform economics by making it possible for young economists to do what old economists have for so long believed was impossible: to model the economy in time, out of equilibrium, and with money.
In a nutshell, Minsky is a visual way of modeling the economy as a dynamic system. It's a new addition to the family of system dynamics programs that enable complex systems to be designed using the metaphor of a flowchart—programs like Simulink and Vissim that engineers use, and Vensim, Stella and the like that are popular in management.
Current Minsky interface with a model being built and tested
Current Minsky interface with a model being built and tested
Before this metaphor was developed, simulation programs had to be written in a computer language like FORTRAN, and only another FORTRAN programmer could understand it. After the visual metaphor was developed (by Jay Forrester at MIT), a simulation model of a system could be designed by building a flowchart that described it. Then anyone who can read a flowchart could read and understand the model, and very complicated models could be broken down into sub-units that different teams could work on and then assemble later.
Minsky improves this approach by adding a way of accurately modeling financial flows. The next section (warning: it is geeky!) explains how Minsky does this, and demos here and in the "Economic Rebels" section show how to do it.

Why Minsky matters—for Kickstarters & Geeks

We've modified the system dynamics flowchart paradigm in two ways:
  • We've added a simple tool that models financial flows in a way that is consistent with the tried and tested method that accountants have been using for centuries: double-entry bookkeeping; and
  • We've improved upon the visual interface of existing models to make Minsky really visual.
Dynamics using Double-entry bookkeeping
Standard system dynamics use integration blocks in flowcharts to create dynamic models, and they work brilliantly at modeling physical systems. But they have an inherent weakness when it comes to modeling financial flows, for one simple reason: financial flows require two balancing entries in at least two different system states, and the standard systems dynamics flowchart method doesn't directly support that.
So though the flowchart paradigm can be used to model financial flows, it provides no check that each entry occurs twice, or that these entries have the right sign.
Accountants and bookkeepers, of course, have to deal with this problem when recording financial transactions, and the solution they came up with centuries ago was double-entry bookkeeping. The inspiration that led to Minsky was the realization that a similar system to what works for accountants with individual transactions would also work to build properly balanced dynamic models of financial flows at the level of an entire economy.
Rather than entering equations for financial flows in a flowchart, the flows are entered into rows of a table, where each source is a positive entry and each sink is a negative one. Assets are also shown as positive sums and Liabilities (and Equity) as negative, and every row sums to zero. The symbolic sum of each column gives all the flows into and out of each financial account, so the integral equations that describe the financial system are derived automatically from the table, and consistency of one system state with another is assured by the rule that all rows must sum to zero.
The result is that financial flows which would be incredibly messy "spaghetti code" in a flowchart—entries appearing as often as six times in different system states, wires overlapping everywhere, no checking on correct signs, etc.—are extremely easy to both create and interpret in Minsky.
Here's a really simple arithmetic example. If you buy a Kindle Paperwhite from Amazon, ultimately, your debt to your bank rises by $120 & Amazon's bank account rises by the same amount. Looking at this from a bank's point of view, its asset (your credit card debt to it) has risen by $120, while its liabilities (Amazon's savings account) has also risen by $120. In a standard system dynamics program, it looks like this:
Simple financial flow in standard system dynamics program Vissim
Simple financial flow in standard system dynamics program Vissim
That's fine, but there's nothing in the program to warn you if you fail to wire up one of the flows, or if you give it the wrong sign. The program wouldn't object if you made the mistake of not showing the bank's liability as a negative amount:
Simple financial transaction with a wrong sign in Vissim
Simple financial transaction with a wrong sign in Vissim
Here's the same example in our "Godley Table" component of Minsky (yes the visuals need some polish, and that's one of the reasons we're asking for support from Kickstarters!):
Same simple transaction in Minsky
Same simple transaction in Minsky
If you make the same mistake as above, by not showing the increase in the bank's liability (Amazon's savings account) as negative, the program warns you because the row doesn't sum to zero:
Minsky spots the error using its row sum checker
Minsky spots the error using its row sum checker
This is just the start of the ways we intend pushing the system dynamics paradigm:
  • Properly modeling production involves acknowledging that there are multiple goods being produced using each other (and energy and labor) as inputs, and multiple banks as well, so both our financial and production modeling have to be multi-dimensional rather than single-dimensional (as the current version of Minsky is, and almost all existing conventional economic models are). I've already done this in Mathcad and Mathematica, so I know it can be coded in Minsky—and with an intuitive user interface.
Multisectoral monetary model developed in Mathcad & Mathematica
Multisectoral monetary model developed in Mathcad & Mathematica
  • Once you've modeled one economy, you've modeled them all, so to speak—in that a structure that fully describes a single country's financial and physical economic systems can be modified to describe another's. Link the two via exchange rates, exports & imports and financial flows, and you have a dynamic, complete model of the global economy.
Of course, an economy can't be modeled in the same way that an engineering system can: you can't predict precisely what will happen in the future, because human agency will intervene, and unpredictable extra-economic effects do occur. But Minsky will enable the modeling of the financial imperatives, physical constraints and nonlinear feedbacks that exist in the real world, and which are ignored in existing economic models and economic policy. It will also allow issues that current models completely ignore—such as the level and rate of growth of private debt—to be properly considered. That has to lead to a better economics.
Lastly, if you haven't studied economics, you might think that economists already use programs like Simulink or IThink to build dynamic models. They don't. They instead use either specialized programs for what they call DSGE modeling, or they use a purpose-designed toolkit in Matlab or the like. These programs assume that the economy has a stable equilibrium, and that "dynamics" is simply the movement of the economy back to this stable equilibrium after some non-economic shock.
A real dynamic modeling system can't simply assume that the economy is stable, and it has to be able to model far from equilibrium dynamics as Minsky does.
Really visual visualization
All system dynamics models are visual when compared to coding a model in a programming language, but how visual they are varies considerably. My favorite among the existing programs is Vissim, because its equations are 100% visual, it uses variable names as well as wires to pass signals, it gives a structured outline view of its groups, and you can embed graphs in the diagram. Here's one of my cyclical models rendered in Vissim, shown at the top level of the model.
Vissim example: all equations are visual
Vissim example: all equations are visual
By way of comparison, here is a marketing model in Vensim:
Top level of a model in Vensim
Top level of a model in Vensim
That's rather less visibly oriented than Vissim. Equation entry in Vensim also uses a dialog box, rather than the visual metaphor:
Vensim equations are entered as strings of text rather than visually
Vensim equations are entered as strings of text rather than visually
Minsky is entirely visual in how equations are entered—with the double-entry bookkeeping table being our special way of visualizing financial flows.
But we're not leaving it there: we're also trying to do better than current system dynamics programs in making modeling visual. For example, here's how we have implemented grouping, compared to how it's done in the most visual of existing programs, Vissim. Our groups are transparent: check the video to see how this works.

Why Minsky matters—for economic rebels

Non-orthodox economists from all other schools of thought—Austrians, Post Keynesians, Evolutionary economists, you name it—have been critiquing the Neoclassical mainstream for decades. And the mainstream has ignored them all. Now a crisis that the mainstream believed was impossible has happened.
We're still critiquing Neoclassicals after the crisis, and they're still ignoring us (though less effectively than they used to do). They are also adapting their models to at least mimic the crisis—which is arguably what happened in the Great Depression when Hicks developed the IS-LM model. Hicks later admitted that his model was based not on Keynes at all, but on Walras:
"the idea of the IS-LM diagram came to me as a result of the work I had been doing on three-way exchange, conceived in a Walrasian manner" (Hicks 1981, pp. 141-42)
This model derailed Keynes's approach, and swept aside Hayek's challenge as well. We don't want that to happen again. But unless we develop a viable alternative paradigm, it will happen again.
The time has come to move beyond critique alone. Minsky is not intended to be all of that alternative paradigm, but it is intended to be part of it by:
  • Providing a visual, dynamic, monetary, non-equilibrium way of seeing the economy that will be far more attractive to new students than the static, equilibrium barter vision they get in introductory Neoclassical courses; and
  • Enabling researchers who wish to do so to build dynamic models of the economy using a tool that is specifically designed to model banking as well as the physical dynamics of a production economy.
Minsky is also a platform for modeling—not a model in itself. This means you can model any vision of the economy you have—whether that's Austrian, Institutional, Post Keynesian—so long as that vision is dynamic. So it can be used by any economist, and it doesn't lock you in to a particular view of how the economy operates.

Minsky—The Rewards

The main reward for those who make a pledge to our project is the project itself. Minsky will help the world by allowing economists to build models take debt, banks, and money seriously.
The other rewards include
We hope that you will enjoy reading the Debunking, and wearing our T-shirts that advertise Minsky's quirky and insightful wisdom.
The first Minsky "Stability is ... Destabilizing!" T-Shirt design
The first Minsky "Stability is ... Destabilizing!" T-Shirt design
The second "Stability is Destabilizing" T-shirt design
The second "Stability is Destabilizing" T-shirt design

Minsky—how to support it

If you're new to Kickstarter, there are some things you need to know if you want to help Minsky develop:
  • You must make a pledge while the Kickstarter campaign is running; So if you want to help us, do it now: don't wait;
  • Your pledge isn't accessed unless we do reach our target, and until the Kickstarter campaign period ends;
  • Pledging is easy. If you click on one of the pledges, you will be taken to Amazon. If you've ever purchased anything from them before, your pledge will be made using your account. If you haven't you can set up an account there and then;
  • You don't need a PayPal account, or anything else complicated like that. If you have a credit card or bank account, you can set up an Amazon account; and
  • If you want to help, do it now. Not only does that get us closer to our target straight away, Kickstarter campaigns tend to be more successful the faster a project reaches its target. In effect, you can help start a "bandwaggon effect" if you pledge early rather than late. So please, do it now.

Show me the Minsky

Minsky is already up and running—though it's very basic in its visuals and capabilities compared to our ultimate ambitions, because only 1,000 programming hours have gone into its development so far. Here are the latest versions (and here is the Help File):
Though this is still a very early "beta" version of Minsky, it can do a lot already—as the next two videos demonstrate.
Help us take Minsky all the way. We've shown the concept works (click here to see 10 more simple examples); now we want to enable it to model the global economy in all its complexity—doing what meteorologists have done in the 50 years since they discovered complexity. This cartoon gives an overview of our development plans:
We need to put in at least 10,000 hours of programming to get it to that stage, and I have a dedicated team ready to roll to do this.

Minsky—The Team

Professor Steve Keen. I came up with the tabular concept on which Minsky is based, and I'm an expert on the "Financial Instability Hypothesis" developed by the late, great iconoclastic American economist Hyman Minsky. I developed the first complex systems model of Minsky's hypothesis back in 1992 (i.e., one with more than 2 nonlinear differential equations in it), and its striking behavior led me to expect that a severe economic crisis would be presaged by a period of apparent tranquility--which is what happened with what Neoclassical economists dubbed "the Great Moderation" preceding the crisis.
Stability leading to Instability in my 1995 Minsky model
Stability leading to Instability in my 1995 Minsky model
I started publicly warning that a crisis was coming in 2005, and established the Debtwatch blog in early 2007. In 2011, I published the second edition of Debunking Economics, which discusses the flaws of conventional "Neoclassical" economic theory in great detail.
I was also the Software Editor for two major Australian computing magazines prior to becoming an academic. That gave me a breadth of knowledge of computer software, which I'm now putting to use in designing Minsky.
Professor Russell Standish is an outstanding computational scientist who has developed Minsky to its current level. Russell has a PhD in Physics from the University of New South Wales (UNSW), and he was the founding Director of the High Performance Computing Unit at UNSW. His main interest has always been the analysis of complex systems, and he has built a multi-agent modeling environment called EcoLab. Russell and I have collaborated on research papers in economics over the years, including a critique of the Neoclassical theory of the firm. When I received funding to develop Minsky from INET (The Institute for New Economic Thinking), Russell was my first choice as its developer, and I was delighted that he was available.
Russell & one of his robots
Russell & one of his robots
Nathan Moses is a graduate computer programmer who did my subject Behavioral Finance in 2011, and from that moment was "hooked" on applying dynamic modeling to economics. In his final year, Nathan pulled together a team of students to develop a browser-enabled version of Minsky as a practical project, and at least he and one of this team (Kevin Pereira) will join us if we get sufficient funding. Here's Nathan explaining the grouping mechanism he's implementing for the Web and Tablet versions of Minsky:

Minsky—The Budget

Whatever funding we raise from Kickstarter will almost entirely go on salaries to keep these "gun" programmers working on Minsky. The amount of time they can devote depends on how much we can raise.

If we raise $50,000

$50,000 is 40% of the grant I received from INET that enabled Minsky to be developed to its current state, and it would purchase about 700 hours of total programming time by Russell, Nathan and Kevin. That's not a lot, but it will enable Russell to complete the "Petty" release of Minsky (for geeks: you can see the existing tickets for the Petty release on our Sourceforge page), and Nathan and Kevin to finish the basic port of Minsky to a web application.
Minsky will then be a very effective teaching tool for monetary macroeconomics, which could also be used to build reasonable scalar models of an economy—similar in their level of detail that conventional economic models have today, though far more realistic since they would be inherently dynamic, monetary and non-equilibrium.


About 1400 hours of total programming time will enable Russell to complete the "Munn" release, which will focus on improving the graphics and presentation aspects of the program.
Nathan and Kevin will also be able to develop a version of Minsky for iPad and Android Tablets.


With twice as much as the original INET Grant, we should be able to complete stage 2 of Minsky—the "Quesnay" release named in honor of the person I regard as the world's first dynamic economist, Francois Quesnay—in which the platform could support the construction of multi-bank model of the financial sector, and multi-commodity model of production.


We will add an OLAP back-end for storage & analysis of economic data and model calibration. As with Minsky itself, this data engine will use a novel visual metaphor to display and analyze multi-dimensional data.

With $1,000,000 or more

This is my dream goal. With $1 million, I can hire Russell, Nathan and Kevin for 3 years full-time. That will enable us to take Minsky all the way to stage 3—where it can model multiple countries, multiple banks, and financial and physical flows for multiple commodities, and where it can perform nonlinear parameter optimization to fit models to data.

Minsky in the News

The Economist recently had a feature on "Economics after the Crisis" called "New Model Army" which featured Minsky as an example of what the future of economics could be:
In Australia Steve Keen, an economist, and Russell Standish, a computational scientist, are developing a software package that would allow anyone to create and play with models of the economy that incorporate some of these new ideas. Called “Minsky”—after Hyman Minsky, an American economist celebrated for his work on boom-and-bust financial cycles—it places the banking system at the centre of the economy. (The Economist, January 19th 2013, p. 68)

Help Kickstart Minsky. Pledge today!

Risks and challenges Learn about accountability on Kickstarter

The project has already commenced, and a working if very early version of Minsky is now available for download:
There is also a SourceForge site for Open Source developers:
We have also developed the next stage of the project—multi-commodity modeling—on the mathematical programming platforms Mathcad and Mathematica. So we know that the next stage of the project can work.
The only real risk are:
(1) That we won't raise enough money to develop the program to its full potential, which will require 10-20 times as much programming time as we've already put in.
(2) That the program won't be adopted by economists, and they will continue to use their unrealistic equilibrium approach to modeling. The more we raise, the more attractive we can make Minsky, the more we can promote it, and the less likely that unsuccessful outcome will be.

No comments:

Post a Comment