Hosted on Acast. See acast.com/privacy for more information.
[00:00:21] [SPEAKER_02]: This is the Debunking Economics podcast with Steve Keen and Phil Dobbie.
[00:00:28] [SPEAKER_04]: Well that is a clip from a show called Breaking Vegas, retracing how two physics geeks managed
[00:00:34] [SPEAKER_04]: to beat the roulette tables of Las Vegas.
[00:00:37] [SPEAKER_04]: But that started a journey focusing on complex systems driven by data that can get to the
[00:00:42] [SPEAKER_04]: bottom of what's happening not just in Vegas, what could happen anywhere and what could
[00:00:47] [SPEAKER_04]: be more complex than understanding gambling.
[00:00:50] [SPEAKER_04]: Well what about the economy?
[00:00:51] [SPEAKER_04]: One of those men who beat the casinos is our guest this week on the Debunking
[00:00:56] [SPEAKER_04]: Economics podcast.
[00:00:57] [SPEAKER_04]: Me and Steve Keen as well of course.
[00:00:59] [SPEAKER_04]: Welcome along.
[00:01:08] [SPEAKER_04]: So, Making Sense of Chaos, that's the title of Doyle Farmer's new book, A Better Economics
[00:01:13] [SPEAKER_04]: for a Better World.
[00:01:15] [SPEAKER_04]: Doyne is Professor of Complex Systems Science at the Smith School for Enterprise and the
[00:01:20] [SPEAKER_04]: Environment at the University of Oxford.
[00:01:22] [SPEAKER_04]: And like Steve he believes a more complex approach to economics could make more accurate
[00:01:26] [SPEAKER_04]: predictions from climate change to financial crises.
[00:01:31] [SPEAKER_04]: So Steve here's another advocate for new economic thinking.
[00:01:34] [SPEAKER_04]: But for both of you, here's a question.
[00:01:37] [SPEAKER_04]: I mean it's a hard sell in this day and age isn't it?
[00:01:40] [SPEAKER_04]: Where people are oversimplifying things to actually say we need to make things more
[00:01:45] [SPEAKER_04]: complicated.
[00:01:46] [SPEAKER_04]: It's a bit of a tough sell isn't it Doyne?
[00:01:47] [SPEAKER_01]: Well let me distinguish two things.
[00:01:49] [SPEAKER_01]: Complex and complicated don't mean the same thing.
[00:01:52] [SPEAKER_01]: So you can have actually very simple models, very simple systems that can nonetheless be
[00:01:57] [SPEAKER_01]: complex because they display rich behavior, what we would call emergent behavior where
[00:02:03] [SPEAKER_01]: the qualitative properties of the system as a whole are different than those of the
[00:02:07] [SPEAKER_01]: components of the system.
[00:02:10] [SPEAKER_01]: So now it is a hard sell to some people, maybe to economists.
[00:02:15] [SPEAKER_01]: My experience is it's not a hard sell to a lot of central bankers and others
[00:02:20] [SPEAKER_01]: who are on the front line and have egg on their face when they get things wrong.
[00:02:24] [SPEAKER_04]: Right.
[00:02:24] [SPEAKER_04]: So really, so you think, because I mean a lot of people would say well central bankers
[00:02:28] [SPEAKER_04]: are quite shallow in their thinking but you think no they want to get the depth?
[00:02:32] [SPEAKER_01]: Well it depends on the central banker.
[00:02:35] [SPEAKER_01]: But I think one of the good signs is that a lot of our stuff is getting traction
[00:02:38] [SPEAKER_01]: with a variety of central banks.
[00:02:41] [SPEAKER_01]: You know the Bank of Italy and the Bank of Canada are now running an agent based
[00:02:45] [SPEAKER_01]: macro model that they run alongside their other models.
[00:02:48] [SPEAKER_01]: There's at least eight banks that are running versions of the housing model that
[00:02:52] [SPEAKER_01]: we created back in 2010, 2012.
[00:02:56] [SPEAKER_01]: So some of these tools are starting to take hold.
[00:03:00] [SPEAKER_04]: Steve you'd probably say yeah running alongside central bank models but those
[00:03:04] [SPEAKER_04]: models are wrong.
[00:03:05] [SPEAKER_03]: Oh yeah.
[00:03:05] [SPEAKER_03]: Doyne's the same.
[00:03:06] [SPEAKER_03]: I mean the dominant model in economics called dynamic stochastic general
[00:03:10] [SPEAKER_03]: equilibrium and I think Doyne would agree with me.
[00:03:13] [SPEAKER_03]: I argue they're neither dynamic nor general.
[00:03:15] [SPEAKER_03]: They're stochastic equilibrium models that are shocked so-called by
[00:03:20] [SPEAKER_03]: external events and then they have what they call frictions to explain why they
[00:03:26] [SPEAKER_03]: don't rapidly return to equilibrium because as we all know all of us live in
[00:03:30] [SPEAKER_03]: a permanent state of equilibrium except when disturbed by small external
[00:03:33] [SPEAKER_03]: things like meteor strikes or COVID.
[00:03:36] [SPEAKER_03]: It's total garbage but that's the way the models are built and they're
[00:03:39] [SPEAKER_03]: extremely complicated.
[00:03:40] [SPEAKER_03]: This is actually partly Doyne's point.
[00:03:43] [SPEAKER_03]: The models that the DSGE models I'll give an example of Michael Kumoff is
[00:03:49] [SPEAKER_03]: sort of a good mate of mine from the Bank of England who's very much a DSGE
[00:03:52] [SPEAKER_03]: thinker and we were talking about extending part of our models to include
[00:03:58] [SPEAKER_03]: I think bank lending to a bank selling bonds to non banks and he said it
[00:04:03] [SPEAKER_03]: would might take him a couple of months to do it in building his DSGE
[00:04:06] [SPEAKER_03]: model because they build the whole thing from the algebra right up until
[00:04:09] [SPEAKER_03]: they've got a statement of a partial and then general equilibrium model.
[00:04:12] [SPEAKER_03]: I said I think I can do that in 10 minutes.
[00:04:14] [SPEAKER_03]: So it's complicatedness versus complex and complex is the right way to go.
[00:04:20] [SPEAKER_04]: So those DSGE models Doyne that Steve thinks are too complex.
[00:04:27] [SPEAKER_04]: Is that the case?
[00:04:29] [SPEAKER_04]: That they are a case of...
[00:04:31] [SPEAKER_04]: Complicated.
[00:04:32] [SPEAKER_04]: Okay so they're too complicated so do they need simplifying?
[00:04:35] [SPEAKER_04]: Is that what we should be looking at?
[00:04:36] [SPEAKER_01]: Let's not get hung up on semantics.
[00:04:38] [SPEAKER_01]: Right I think the important thing is they have a they're based on a
[00:04:41] [SPEAKER_01]: completely different set of principles.
[00:04:43] [SPEAKER_01]: You know to build a DSGE model you assign a few representative agents
[00:04:48] [SPEAKER_01]: a utility function that is kind of scorecard that measures what they want.
[00:04:52] [SPEAKER_01]: Typically they want to consume as much as possible and you then calculate
[00:04:58] [SPEAKER_01]: the decisions those agents will make in order to maximize their utility.
[00:05:05] [SPEAKER_01]: As Steve mentioned you may throw in some frictions because few things they
[00:05:09] [SPEAKER_01]: can't do.
[00:05:10] [SPEAKER_01]: Firms can't change wages instantaneously so they throw in some assumptions like
[00:05:15] [SPEAKER_01]: that and then they calculate they assume supply equals demand and then
[00:05:20] [SPEAKER_01]: and then they look at the economic consequences of that in equations
[00:05:24] [SPEAKER_01]: and so those models are completely different than the models that we're
[00:05:29] [SPEAKER_01]: making where we have genuinely dynamic models.
[00:05:33] [SPEAKER_01]: We have agents who make who look at the information they have.
[00:05:37] [SPEAKER_01]: They make decisions based on rules of thumb or simple learning algorithms
[00:05:42] [SPEAKER_01]: that people really use.
[00:05:46] [SPEAKER_01]: That affects the economy.
[00:05:48] [SPEAKER_01]: That creates new information in addition some new information may flow
[00:05:53] [SPEAKER_01]: in from outside COVID or something.
[00:05:55] [SPEAKER_01]: The agents make another round of decisions and then we just go around
[00:05:58] [SPEAKER_01]: and around that loop so it's genuinely dynamic.
[00:06:02] [SPEAKER_01]: Sometimes those models settle into equilibrium sometimes they don't.
[00:06:06] [SPEAKER_04]: You're saying the models used by central banks and many economists
[00:06:10] [SPEAKER_04]: obviously are wrong in that they are they make too many assumptions which
[00:06:16] [SPEAKER_04]: are proven wrong and they assume those assumptions are just going to
[00:06:19] [SPEAKER_04]: keep on repeating themselves.
[00:06:20] [SPEAKER_01]: Is that their weakness?
[00:06:23] [SPEAKER_01]: I think actually more than that.
[00:06:24] [SPEAKER_01]: I mean you always have to make assumptions.
[00:06:26] [SPEAKER_01]: We make assumptions too.
[00:06:28] [SPEAKER_01]: I think the difference is that if you're trying to calculate the
[00:06:32] [SPEAKER_01]: optimum decisions of agents okay first of all I don't think that's
[00:06:35] [SPEAKER_01]: what we really do but even more important you can't make the models
[00:06:39] [SPEAKER_01]: very complicated.
[00:06:40] [SPEAKER_01]: You can't put in real world stuff because then you just can't solve
[00:06:44] [SPEAKER_01]: the models and so they're stuck.
[00:06:48] [SPEAKER_01]: And as Steve also emphasized those models have the property that
[00:06:53] [SPEAKER_01]: under certain conditions they always settle into what's called a
[00:06:57] [SPEAKER_01]: fixed point that is absent any disturbance from the outside world.
[00:07:00] [SPEAKER_01]: They just go to rest nothing changes anymore.
[00:07:03] [SPEAKER_01]: And so that means the only way to get dynamics is to be poking them all the
[00:07:07] [SPEAKER_01]: time right and some things happen that way.
[00:07:11] [SPEAKER_01]: COVID was a great big poke that knocked the economy out of equilibrium
[00:07:15] [SPEAKER_01]: but the 2008 financial crisis was a spontaneous event.
[00:07:19] [SPEAKER_01]: The economy generated it.
[00:07:22] [SPEAKER_01]: The market crash of a few days ago was a spontaneous event.
[00:07:26] [SPEAKER_01]: The market caused that crash due to causes that are internal to the
[00:07:33] [SPEAKER_01]: system.
[00:07:33] [SPEAKER_01]: And so it's what we and you know in dynamical systems theory would call
[00:07:37] [SPEAKER_01]: endogenous dynamics that is it's dynamics that comes from within
[00:07:43] [SPEAKER_01]: and it happens automatically from within the system.
[00:07:47] [SPEAKER_01]: So those are those are to me the key differences.
[00:07:49] [SPEAKER_04]: Well I mean these what we're seeing right now in the markets is just
[00:07:53] [SPEAKER_04]: insecurity about knowing the future isn't it?
[00:07:55] [SPEAKER_04]: That's part of the problem.
[00:07:56] [SPEAKER_03]: Well that's actually a very good point because they is these DSG
[00:07:59] [SPEAKER_03]: models assume we can predict the future accurately.
[00:08:02] [SPEAKER_03]: That's the sort of crazy stuff that people don't realize
[00:08:05] [SPEAKER_03]: are built into these models.
[00:08:06] [SPEAKER_03]: They think they're sophisticated and they include you know people's
[00:08:09] [SPEAKER_03]: decision making and what they call forward looking expectations.
[00:08:13] [SPEAKER_03]: When you unwrap that this is something a doing criticizes quite nicely
[00:08:16] [SPEAKER_03]: in the book as well.
[00:08:17] [SPEAKER_03]: They presume what they call rational expectations and we all want to be
[00:08:21] [SPEAKER_03]: rational don't we?
[00:08:22] [SPEAKER_03]: You look at how they define rational and if you told any ordinary person
[00:08:25] [SPEAKER_03]: on the street what the definition was and said what word would you use?
[00:08:29] [SPEAKER_03]: What word am I describing in the dictionary?
[00:08:31] [SPEAKER_03]: They wouldn't choose the word rational.
[00:08:33] [SPEAKER_03]: They'd choose the word prophetic.
[00:08:35] [SPEAKER_04]: And there's nothing rational about the way markets are behaving right now
[00:08:38] [SPEAKER_04]: because they are looking at one set of job numbers and seeing a weak
[00:08:42] [SPEAKER_04]: result so then they look at another set of job numbers which comes out
[00:08:46] [SPEAKER_04]: every week and get an oversized response to that.
[00:08:48] [SPEAKER_04]: I mean there's just a complete lack of logic and it's but it's just
[00:08:51] [SPEAKER_04]: punting isn't it?
[00:08:52] [SPEAKER_04]: It's just people punting on you know with the way they think the
[00:08:55] [SPEAKER_04]: future is going.
[00:08:56] [SPEAKER_04]: So no mathematical model can replicate that sort of behavior can it?
[00:09:00] [SPEAKER_01]: Well you could have different mathematical models so I think the
[00:09:04] [SPEAKER_01]: problem with mathematical models is it's hard to make hard to put in
[00:09:08] [SPEAKER_01]: all the real world stuff that needs to go in.
[00:09:12] [SPEAKER_01]: You know some of our models are mathematical too they're just based
[00:09:14] [SPEAKER_01]: on different kind of math.
[00:09:17] [SPEAKER_01]: I think it's important to emphasize that these DSG models do have
[00:09:21] [SPEAKER_01]: the future isn't completely certain in those models.
[00:09:25] [SPEAKER_01]: They do have probabilities of events but those probabilities they assume
[00:09:29] [SPEAKER_01]: events are known and they assume we know those probabilities and we
[00:09:33] [SPEAKER_01]: typically don't know those probabilities at all.
[00:09:37] [SPEAKER_01]: And you know one of the points I make in my book is that
[00:09:41] [SPEAKER_01]: behavioral there is a whole branch of economics now called behavioral
[00:09:44] [SPEAKER_01]: economics and the profession has not succeeded in melding behavioral
[00:09:50] [SPEAKER_01]: economics together with macro models that are used by you know
[00:09:55] [SPEAKER_01]: treasury departments and central banks at least the old fashion ones
[00:09:58] [SPEAKER_01]: and DSG models and so there's a mismatch there and so it's whereas
[00:10:07] [SPEAKER_01]: you know in our models we can because we're making more realistic
[00:10:12] [SPEAKER_01]: models of the way people behave we deal with uncertainty in a more
[00:10:17] [SPEAKER_01]: natural way.
[00:10:18] [SPEAKER_03]: There's actually an interesting comment from Keynes on the nature of
[00:10:22] [SPEAKER_03]: economic the time he wrote back in the 1930s of course well before the
[00:10:25] [SPEAKER_03]: computers and he said I accused the classical but which he meant would we
[00:10:30] [SPEAKER_03]: would now call neoclassical school of being one of those pretty polite
[00:10:35] [SPEAKER_03]: techniques designed for a well paneled boardroom which tries to deal with
[00:10:40] [SPEAKER_03]: the present by abstracting from the fact that we know nothing about
[00:10:43] [SPEAKER_03]: the future.
[00:10:44] [SPEAKER_03]: Now what Duane has done with the computer models he builds and what
[00:10:47] [SPEAKER_03]: I do with my non-linear complex system mathematical models is we
[00:10:51] [SPEAKER_03]: actually just have uncertainty generated out of the model itself
[00:10:55] [SPEAKER_03]: fundamentally was what the neoclassicals have done have deal with
[00:10:59] [SPEAKER_03]: it not by abstracting from the fact we don't know about the future
[00:11:02] [SPEAKER_03]: but pretending that we can predict the future accurately.
[00:11:05] [SPEAKER_03]: So what we have that people think are useful models that central
[00:11:09] [SPEAKER_03]: banks use and treasury dues and so on they're incredibly misleading
[00:11:12] [SPEAKER_03]: because they're based on the assumption we can predict the
[00:11:14] [SPEAKER_03]: future and that's madness and what you need to set as a sort of
[00:11:18] [SPEAKER_03]: multi-agent work that Duane does or the complex system stuff that I do top
[00:11:22] [SPEAKER_03]: down style which just acknowledges the future is uncertain.
[00:11:27] [SPEAKER_04]: Right but you do want to predict the future though don't you?
[00:11:29] [SPEAKER_04]: That's the reason why you're modeling to try and get some idea.
[00:11:32] [SPEAKER_04]: So Duane on this idea of multi-agents how different is it from
[00:11:35] [SPEAKER_04]: what I was doing 20 years ago when I was marketing for a large
[00:11:40] [SPEAKER_04]: internet company in Australia and we segmented the market into maybe
[00:11:45] [SPEAKER_04]: eight different categories and for each one of those groups we knew the
[00:11:49] [SPEAKER_04]: demographics that the makeup of them whether they were for example a
[00:11:52] [SPEAKER_04]: young family an old family retired couple a single person
[00:11:57] [SPEAKER_04]: particular income brackets and for each of those we knew their
[00:12:00] [SPEAKER_04]: propensity to buy our product and we knew what the buttons to press
[00:12:02] [SPEAKER_04]: to try and influence them and we knew the media that we could
[00:12:06] [SPEAKER_04]: use to reach them so we had a pretty good idea about what the cost
[00:12:09] [SPEAKER_04]: of acquisition was for them so we knew what our return was.
[00:12:12] [SPEAKER_04]: I mean that was you know reasonably complex modeling I guess
[00:12:15] [SPEAKER_04]: and we were trying to predict the future normally worked by the way
[00:12:19] [SPEAKER_04]: pretty well.
[00:12:20] [SPEAKER_04]: We had a pretty good idea about you know if we spent money on this
[00:12:23] [SPEAKER_04]: on this category this segment we're going to get a much better return
[00:12:26] [SPEAKER_04]: than spending the same amount of money on another segment.
[00:12:28] [SPEAKER_04]: So we talk about that sort of approach that you know marketing
[00:12:32] [SPEAKER_04]: smart marketing people have been taking and applying that to economics
[00:12:36] [SPEAKER_04]: is that sort of like the nub you know different agents behaving in
[00:12:38] [SPEAKER_04]: different ways?
[00:12:39] [SPEAKER_01]: What you're talking is part of it.
[00:12:42] [SPEAKER_01]: What you're talking about is close to what's called a micro simulation
[00:12:45] [SPEAKER_01]: model and those have been around for quite a while and a micro
[00:12:48] [SPEAKER_01]: simulation model those are used to understand the effect of say
[00:12:52] [SPEAKER_01]: a new tax regime and the key difference between those and what
[00:12:57] [SPEAKER_01]: we're doing is the dynamical part that is if we took your
[00:13:02] [SPEAKER_01]: model which you know we'd ultimately like to incorporate
[00:13:06] [SPEAKER_01]: things like that in.
[00:13:08] [SPEAKER_01]: We look at how people are consuming then we feed that back to what
[00:13:13] [SPEAKER_01]: people are going to produce and then go around and around the loop
[00:13:17] [SPEAKER_01]: so we get the feedback of the economy and we capture the kind
[00:13:20] [SPEAKER_01]: of Keynesian effects that Steve brought up like the fact that if
[00:13:24] [SPEAKER_01]: people become unemployed they don't spend which means people
[00:13:28] [SPEAKER_01]: don't produce which means more people become unemployed which
[00:13:32] [SPEAKER_01]: means less production happens so there are these feedback loops
[00:13:38] [SPEAKER_01]: in the economy that Keynes was the first to really home in on
[00:13:42] [SPEAKER_01]: that the kind of models that we're doing capture but a
[00:13:46] [SPEAKER_01]: component of our models is just what you say.
[00:13:50] [SPEAKER_01]: We have synthetic populations of agents.
[00:13:53] [SPEAKER_01]: The agents have different ages, different genders, different
[00:13:59] [SPEAKER_01]: incomes, different educations and so and we let all those agents
[00:14:06] [SPEAKER_01]: behave differently and we try and get their behavior as accurately
[00:14:10] [SPEAKER_01]: as possible.
[00:14:11] [SPEAKER_01]: They also by the way have jobs so they work and that's the
[00:14:13] [SPEAKER_01]: difference they have jobs they work for firms the firms
[00:14:16] [SPEAKER_01]: produce the stuff that they consume so if they're consuming
[00:14:20] [SPEAKER_01]: less the firms produce less which means they have to fire
[00:14:24] [SPEAKER_01]: people which means people put all of that happens inside our
[00:14:27] [SPEAKER_01]: models.
[00:14:27] [SPEAKER_04]: And is that largely just applying more detail to what have
[00:14:32] [SPEAKER_04]: been accepted sort of conventional economic wisdom or is
[00:14:37] [SPEAKER_04]: it taking you off in a whole new direction.
[00:14:39] [SPEAKER_04]: Have you discovered stuff which is which is so I know for
[00:14:41] [SPEAKER_04]: example you mentioned the word equilibrium before and
[00:14:43] [SPEAKER_04]: Steve is a big believer that there is no equilibrium that
[00:14:47] [SPEAKER_04]: you know that the system is forever moving.
[00:14:49] [SPEAKER_01]: Yeah so our models are agnostic about that question.
[00:14:52] [SPEAKER_01]: They they they run the dynamics and if in certain
[00:14:55] [SPEAKER_01]: situations we can set them up where they do go into an
[00:14:58] [SPEAKER_01]: equilibrium but but to be honest they well they often
[00:15:02] [SPEAKER_01]: don't.
[00:15:02] [SPEAKER_01]: In fact maybe most of the time they don't when we start
[00:15:05] [SPEAKER_01]: running more realistic situations and you know we
[00:15:09] [SPEAKER_01]: are building on a very different set of principles so
[00:15:11] [SPEAKER_01]: there is yes there is some accepted economic wisdom in
[00:15:15] [SPEAKER_01]: there.
[00:15:17] [SPEAKER_01]: We try and capture things that Keynes understand we're
[00:15:20] [SPEAKER_01]: trying to capture other effects that are well
[00:15:23] [SPEAKER_01]: documented in economics but we're doing it in a completely
[00:15:27] [SPEAKER_01]: different way.
[00:15:29] [SPEAKER_04]: So one of the three mentioned Covid and I know you did
[00:15:31] [SPEAKER_04]: quite a bit of work you write about it in the book
[00:15:34] [SPEAKER_04]: around Covid.
[00:15:35] [SPEAKER_04]: It seems curious doesn't it that during Covid governments
[00:15:39] [SPEAKER_04]: put a lot of money into the economy because they had to
[00:15:41] [SPEAKER_04]: because people needed to pay their mortgages they needed
[00:15:45] [SPEAKER_04]: to eat they didn't have work because we were all in
[00:15:48] [SPEAKER_04]: lockdown and then and then came inflation and it's
[00:15:52] [SPEAKER_04]: interesting if you look at you know central bank minutes
[00:15:56] [SPEAKER_04]: and discussions within governments and you know amongst
[00:16:00] [SPEAKER_04]: economists the idea that we'd have this sudden rampant
[00:16:04] [SPEAKER_04]: inflation wasn't even contemplated.
[00:16:06] [SPEAKER_04]: So did in your modeling did first of all why did that
[00:16:09] [SPEAKER_04]: happen and in your model do you see it differently to
[00:16:12] [SPEAKER_04]: you know to some of the wisdom that or some of the
[00:16:15] [SPEAKER_04]: discussion which is going now which is there's just
[00:16:17] [SPEAKER_04]: too much money in the economy and did you foresee
[00:16:20] [SPEAKER_04]: that in your in your modeling?
[00:16:21] [SPEAKER_04]: Tell you we'll talk us through that.
[00:16:22] [SPEAKER_01]: Yeah well let me first say there were a few people who
[00:16:25] [SPEAKER_01]: anticipated this Larry Summers for example yeah
[00:16:29] [SPEAKER_01]: kindly offered an endorsement of my book one of
[00:16:32] [SPEAKER_01]: the few mainstream economists who willing to be radical
[00:16:35] [SPEAKER_01]: enough to step up and do that and actually provided
[00:16:39] [SPEAKER_01]: useful input to my book.
[00:16:41] [SPEAKER_01]: He said inflation is a danger and we actually worried
[00:16:43] [SPEAKER_01]: about inflation too but our model we made it back at
[00:16:47] [SPEAKER_01]: the very beginning of the pandemic and we were quite
[00:16:50] [SPEAKER_01]: clear that we weren't tackling the question whether
[00:16:53] [SPEAKER_01]: there'd be inflation.
[00:16:54] [SPEAKER_01]: We said we don't think there's going to be inflation
[00:16:56] [SPEAKER_01]: for at least a year which was right.
[00:16:59] [SPEAKER_01]: That was a subjective question because our model
[00:17:02] [SPEAKER_01]: didn't have inflation in it.
[00:17:04] [SPEAKER_01]: It just looked at the way the shocks were propagating
[00:17:06] [SPEAKER_01]: around the economy and how they were going to damp
[00:17:09] [SPEAKER_01]: output in the economy and inflation is a trickier
[00:17:13] [SPEAKER_01]: problem.
[00:17:14] [SPEAKER_01]: It's one frankly I hope our macro model let me put in a
[00:17:20] [SPEAKER_01]: plug for my colleague Kars Hommas who's at the Bank of
[00:17:23] [SPEAKER_01]: Canada their agent based macro model has the ability to
[00:17:27] [SPEAKER_01]: have self-reinforcing expectations and when you do
[00:17:33] [SPEAKER_01]: that they actually get much more realistic predictions
[00:17:36] [SPEAKER_01]: about inflation than other models have been getting.
[00:17:40] [SPEAKER_01]: So there is some hint that we can do this better with
[00:17:43] [SPEAKER_01]: agent based models now.
[00:17:45] [SPEAKER_01]: It's still an unsolved problem.
[00:17:47] [SPEAKER_04]: And does it help central banks with the timing
[00:17:50] [SPEAKER_04]: and the lag effect because that's what everyone
[00:17:52] [SPEAKER_04]: is challenging challenged with right now.
[00:17:54] [SPEAKER_04]: How quickly do we move interest rates?
[00:17:57] [SPEAKER_04]: I mean Steve you might want to talk about the
[00:17:58] [SPEAKER_04]: impact of those interest rates are or not having
[00:18:01] [SPEAKER_04]: anyway but I mean how is that modeling helping
[00:18:05] [SPEAKER_04]: determine the timing of that because obviously if
[00:18:08] [SPEAKER_04]: you believe that they are bringing down interest
[00:18:10] [SPEAKER_04]: rates because they believe that that is going to
[00:18:12] [SPEAKER_04]: help boost the economy now inflation is contained
[00:18:15] [SPEAKER_04]: and there's the risk of jobs going if they don't
[00:18:18] [SPEAKER_04]: they've got to get the timing right from all of
[00:18:20] [SPEAKER_04]: that and the data is lagged.
[00:18:22] [SPEAKER_04]: So can the sort of modeling that you're talking
[00:18:23] [SPEAKER_04]: about does that help counteract the impact of
[00:18:26] [SPEAKER_04]: the lag.
[00:18:27] [SPEAKER_01]: Not yet.
[00:18:27] [SPEAKER_01]: I like to pick the problems I can solve.
[00:18:32] [SPEAKER_01]: So inflation is harder not to crack.
[00:18:34] [SPEAKER_01]: Steve you may have some things to say about
[00:18:36] [SPEAKER_01]: that so maybe I should pass it to you.
[00:18:38] [SPEAKER_03]: Well it is a time is actually the central part
[00:18:41] [SPEAKER_03]: of your work and my work.
[00:18:42] [SPEAKER_03]: So I work with systems of nonlinear differential
[00:18:44] [SPEAKER_03]: equations and of course when you're in
[00:18:46] [SPEAKER_03]: differential equations you're in time and the
[00:18:49] [SPEAKER_03]: dynamics are such that you know one event can
[00:18:52] [SPEAKER_03]: take time to pass through to another part of
[00:18:55] [SPEAKER_03]: the model even though you might not actually
[00:18:57] [SPEAKER_03]: have time lags built in.
[00:18:59] [SPEAKER_03]: This is one thing I find quite funny when I see
[00:19:01] [SPEAKER_03]: conventional economists trying to handle time
[00:19:04] [SPEAKER_03]: and they talk about things like being
[00:19:05] [SPEAKER_03]: co-integrated is one of the favorite terms
[00:19:07] [SPEAKER_03]: you'd know from econometricians and co-integrated
[00:19:10] [SPEAKER_03]: means they move at the same sort of cycle which
[00:19:13] [SPEAKER_03]: they say is not co-integrated well they
[00:19:15] [SPEAKER_03]: can't be causally related to each other.
[00:19:17] [SPEAKER_03]: But if you use sine and cosine which are both
[00:19:19] [SPEAKER_03]: functions of time they're always out of step
[00:19:22] [SPEAKER_03]: but they're absolutely integrated with each
[00:19:24] [SPEAKER_03]: other.
[00:19:25] [SPEAKER_03]: So when you work in with time as I do with
[00:19:28] [SPEAKER_03]: systems of differential equations and as
[00:19:31] [SPEAKER_03]: Doin does with multi-agents all of them
[00:19:33] [SPEAKER_03]: have to make decisions in a time framework
[00:19:36] [SPEAKER_03]: then you do get those time lags coming
[00:19:38] [SPEAKER_03]: naturally out of the internal dynamics.
[00:19:41] [SPEAKER_03]: You don't have to set them up they're created
[00:19:44] [SPEAKER_03]: by the way one factor influences another.
[00:19:47] [SPEAKER_03]: So in that sense we are based in time
[00:19:50] [SPEAKER_03]: whereas in a fundamental way even though
[00:19:52] [SPEAKER_03]: DSG models try to prediction predictions
[00:19:54] [SPEAKER_03]: through time they can basically get rid of
[00:19:57] [SPEAKER_03]: time by shaming equilibrium.
[00:19:59] [SPEAKER_03]: And the other big mistake they make which
[00:20:00] [SPEAKER_03]: is also very different to both my work and
[00:20:03] [SPEAKER_03]: Doin's is that they and you'll see Olivia
[00:20:05] [SPEAKER_03]: Lanshaw admitting this in a paper called
[00:20:09] [SPEAKER_03]: Do DSG models have a future and he said we
[00:20:13] [SPEAKER_03]: presumed everything was linear and that's
[00:20:15] [SPEAKER_03]: what's again Doin mentioned this earlier
[00:20:16] [SPEAKER_03]: that's actually forced on them by their
[00:20:18] [SPEAKER_03]: analytic approach.
[00:20:19] [SPEAKER_03]: It's almost intractable to do solutions
[00:20:22] [SPEAKER_03]: to systems of non-linear behavior so
[00:20:25] [SPEAKER_03]: they assume linear and that's a bit
[00:20:27] [SPEAKER_03]: like the old drunk saying I'm going to
[00:20:28] [SPEAKER_03]: look for my key under the light path
[00:20:30] [SPEAKER_03]: because that's where there's light.
[00:20:32] [SPEAKER_03]: I know I drop them back in the dark
[00:20:33] [SPEAKER_03]: there you know 30 or 60 meters back but
[00:20:36] [SPEAKER_03]: I can't see anything so I won't look
[00:20:37] [SPEAKER_03]: around there and that's really the
[00:20:39] [SPEAKER_03]: nature of their models whereas I'll
[00:20:41] [SPEAKER_03]: simply admit we live in time and time
[00:20:43] [SPEAKER_03]: gives you both dynamics and uncertainty.
[00:20:46] [SPEAKER_04]: Right and then those dynamics can change
[00:20:48] [SPEAKER_04]: of course over time against that
[00:20:50] [SPEAKER_04]: uncertainty and so is that part of
[00:20:52] [SPEAKER_04]: the problem with conventional models
[00:20:53] [SPEAKER_04]: that we tend to look at everything in
[00:20:55] [SPEAKER_04]: a linear way and there may be for
[00:20:58] [SPEAKER_04]: example tipping points within a model
[00:21:00] [SPEAKER_04]: where or you know or something changes
[00:21:03] [SPEAKER_04]: over time so the the multiply effect
[00:21:05] [SPEAKER_04]: of one particular variable might
[00:21:07] [SPEAKER_04]: increase over time or decrease or might
[00:21:09] [SPEAKER_04]: be how much it changes is dependent on
[00:21:13] [SPEAKER_04]: another variable that wasn't
[00:21:14] [SPEAKER_04]: associated with it before.
[00:21:17] [SPEAKER_04]: Is that the sort of thing we're
[00:21:18] [SPEAKER_04]: talking about and you know how to
[00:21:20] [SPEAKER_04]: arrive at the conclusions about what
[00:21:22] [SPEAKER_04]: influences what?
[00:21:23] [SPEAKER_01]: Well let's I think we should parse
[00:21:25] [SPEAKER_01]: out a couple of things so
[00:21:26] [SPEAKER_01]: so as a conventional DSG model is not
[00:21:29] [SPEAKER_01]: necessarily linear they do often
[00:21:32] [SPEAKER_01]: linearize them to make them simpler to
[00:21:34] [SPEAKER_01]: solve which is a bit like throwing out
[00:21:36] [SPEAKER_01]: the baby with a bath water if you want
[00:21:38] [SPEAKER_01]: to understand
[00:21:39] [SPEAKER_01]: a business cycle in my view but
[00:21:42] [SPEAKER_01]: but that's separate as I said before
[00:21:44] [SPEAKER_01]: the whole way the model is
[00:21:46] [SPEAKER_01]: constructed is different so first of
[00:21:48] [SPEAKER_01]: all our agent-based models part
[00:21:50] [SPEAKER_01]: them are written down in mathematics
[00:21:52] [SPEAKER_01]: part of them aren't they're computer
[00:21:55] [SPEAKER_01]: simulations
[00:21:56] [SPEAKER_01]: that's the key difference we actually
[00:21:58] [SPEAKER_01]: try and simulate the world
[00:21:59] [SPEAKER_01]: as it is getting at least the essential
[00:22:03] [SPEAKER_01]: elements
[00:22:04] [SPEAKER_01]: and and the essential interactions
[00:22:07] [SPEAKER_01]: and and correctly mimicking the
[00:22:11] [SPEAKER_01]: economic institutions
[00:22:12] [SPEAKER_01]: and the way those institutions actually
[00:22:15] [SPEAKER_01]: work
[00:22:16] [SPEAKER_01]: so just to give you example
[00:22:20] [SPEAKER_01]: in say our housing market model we
[00:22:23] [SPEAKER_01]: act you know
[00:22:24] [SPEAKER_01]: a conventional economist will say supply
[00:22:26] [SPEAKER_01]: equals demand
[00:22:27] [SPEAKER_01]: in a housing market number of buyers
[00:22:29] [SPEAKER_01]: equals the number of sellers
[00:22:31] [SPEAKER_01]: whereas in our model we mimic the way
[00:22:34] [SPEAKER_01]: houses actually get sold
[00:22:35] [SPEAKER_01]: by real estate agents what do they do
[00:22:37] [SPEAKER_01]: the real estate agent you go to the
[00:22:39] [SPEAKER_01]: real estate agent
[00:22:40] [SPEAKER_01]: you say I want to buy a house they look
[00:22:43] [SPEAKER_01]: at your house and they find some
[00:22:44] [SPEAKER_01]: comparables
[00:22:45] [SPEAKER_01]: they say well I think your house is
[00:22:46] [SPEAKER_01]: worth about this jack it up
[00:22:48] [SPEAKER_01]: maybe five percent put the house on
[00:22:51] [SPEAKER_01]: the market
[00:22:51] [SPEAKER_01]: the house doesn't sell put it down
[00:22:55] [SPEAKER_01]: market down by 10 percent
[00:22:57] [SPEAKER_01]: doesn't sell again market down again
[00:22:58] [SPEAKER_01]: by 10 percent you keep doing that
[00:23:00] [SPEAKER_01]: till the house either sells or the
[00:23:02] [SPEAKER_01]: seller says
[00:23:03] [SPEAKER_01]: sorry I'm not willing to sell it for
[00:23:05] [SPEAKER_01]: that low price
[00:23:06] [SPEAKER_01]: and in our agent-based simulation of
[00:23:09] [SPEAKER_01]: a housing market
[00:23:09] [SPEAKER_01]: we can actually mimic that process
[00:23:12] [SPEAKER_01]: which is called aspiration level
[00:23:14] [SPEAKER_01]: adaptation
[00:23:16] [SPEAKER_01]: we can mimic that process and much
[00:23:18] [SPEAKER_01]: more realistically capture the fact
[00:23:20] [SPEAKER_01]: that housing markets often supply is
[00:23:23] [SPEAKER_01]: way out of whack with demand
[00:23:24] [SPEAKER_01]: you can have 20 times as many people
[00:23:27] [SPEAKER_01]: trying to sell their house
[00:23:28] [SPEAKER_01]: as there are buyers or vice versa and
[00:23:31] [SPEAKER_01]: if you go through the
[00:23:33] [SPEAKER_01]: the housing market crash 2008
[00:23:36] [SPEAKER_01]: you see that flip happen leading up
[00:23:40] [SPEAKER_01]: to the crash
[00:23:41] [SPEAKER_01]: there's way more buyers and sellers
[00:23:43] [SPEAKER_01]: after the crash there's way more
[00:23:44] [SPEAKER_01]: sellers and buyers
[00:23:45] [SPEAKER_01]: so we can mimic that kind of thing
[00:23:47] [SPEAKER_01]: because we're just doing everything in
[00:23:49] [SPEAKER_01]: a completely different way
[00:23:51] [SPEAKER_01]: we're just simulating everything on a
[00:23:53] [SPEAKER_01]: computer
[00:23:54] [SPEAKER_01]: which means we can put in all the
[00:23:56] [SPEAKER_01]: realism we need
[00:23:59] [SPEAKER_01]: to mimic what the economy is
[00:24:01] [SPEAKER_01]: actually doing
[00:24:01] [SPEAKER_04]: right so long as you're drawing the
[00:24:03] [SPEAKER_04]: right conclusions obviously about what
[00:24:05] [SPEAKER_04]: is
[00:24:06] [SPEAKER_04]: driving that behavior for each of
[00:24:07] [SPEAKER_04]: those actors so maybe we'll
[00:24:09] [SPEAKER_04]: pick up on that when we come back
[00:24:11] [SPEAKER_04]: and you know how do you know what
[00:24:12] [SPEAKER_04]: to look for
[00:24:12] [SPEAKER_04]: how do you know which actors to
[00:24:14] [SPEAKER_04]: include how do you know which
[00:24:15] [SPEAKER_04]: variables to include
[00:24:16] [SPEAKER_04]: and can you make a model too
[00:24:18] [SPEAKER_04]: complicated so that
[00:24:20] [SPEAKER_04]: it becomes unwieldy or conversely
[00:24:23] [SPEAKER_04]: you know it's not complicated enough
[00:24:24] [SPEAKER_04]: in which case you might miss out some
[00:24:26] [SPEAKER_04]: fundamental factors and on those
[00:24:28] [SPEAKER_04]: you know how do you know what you
[00:24:29] [SPEAKER_04]: are missing i guess is always the big
[00:24:31] [SPEAKER_04]: question as well so we'll look at all
[00:24:32] [SPEAKER_04]: of that when we come back
[00:24:33] [SPEAKER_04]: join pharma is our guest on the
[00:24:35] [SPEAKER_04]: debunking economics podcast this
[00:24:36] [SPEAKER_04]: week he's got a new book out which
[00:24:38] [SPEAKER_04]: is called making sense of chaos
[00:24:40] [SPEAKER_04]: a better economics for a better
[00:24:41] [SPEAKER_04]: world join us after the break as we
[00:24:43] [SPEAKER_04]: explore this subject
[00:24:44] [SPEAKER_04]: that little bit more stay with us
[00:24:45] [SPEAKER_02]: this is the debunking economics
[00:24:48] [SPEAKER_02]: podcast with steve keen
[00:24:50] [SPEAKER_02]: and phil dobby
[00:24:53] [SPEAKER_04]: well we are looking at complex systems
[00:24:56] [SPEAKER_04]: on the podcast today much to steve
[00:24:58] [SPEAKER_04]: keen's delight because it's one of his
[00:25:00] [SPEAKER_04]: favorite subjects well it's what he
[00:25:01] [SPEAKER_04]: does basically so
[00:25:03] [SPEAKER_04]: it's it's good to have joined
[00:25:04] [SPEAKER_04]: pharma on the podcast as well
[00:25:07] [SPEAKER_04]: so and they both understand what
[00:25:09] [SPEAKER_04]: they're talking about i'm still
[00:25:11] [SPEAKER_04]: trying to really get to grips with
[00:25:13] [SPEAKER_04]: exactly what a complex system is
[00:25:15] [SPEAKER_04]: i sort of get it but not entirely
[00:25:17] [SPEAKER_04]: i like the example you gave of housing
[00:25:20] [SPEAKER_04]: but i still feel like that's supply
[00:25:22] [SPEAKER_04]: and demand all you've done is you've
[00:25:23] [SPEAKER_04]: said well okay there's a you know
[00:25:25] [SPEAKER_04]: there's a mismatch between supply and
[00:25:26] [SPEAKER_04]: demand in that you reach a point
[00:25:28] [SPEAKER_04]: where people can't afford
[00:25:30] [SPEAKER_04]: houses and so that all there's not
[00:25:32] [SPEAKER_04]: enough supply available it still
[00:25:34] [SPEAKER_04]: feels like a i mean maybe you're
[00:25:37] [SPEAKER_04]: you're testing the point at which
[00:25:38] [SPEAKER_04]: that mismatch starts to happen
[00:25:40] [SPEAKER_04]: and the housing market starts to
[00:25:42] [SPEAKER_04]: collapse what's what's the extra
[00:25:43] [SPEAKER_04]: dimension beyond supply and demand
[00:25:45] [SPEAKER_04]: that you've added in there what are
[00:25:46] [SPEAKER_04]: variables for example well so first
[00:25:48] [SPEAKER_01]: of all i'm not against supply and
[00:25:50] [SPEAKER_01]: demand i mean supply and demand is a
[00:25:51] [SPEAKER_01]: powerful force and if there's more
[00:25:55] [SPEAKER_01]: supply than demand prices go down
[00:25:57] [SPEAKER_01]: if there's more demand than supply
[00:25:59] [SPEAKER_01]: prices go up so there's a dynamics
[00:26:01] [SPEAKER_01]: that's driven by supply and demand
[00:26:03] [SPEAKER_01]: and one of the purposes of a good
[00:26:05] [SPEAKER_01]: economic model is to capture that
[00:26:07] [SPEAKER_01]: dynamics in a sensible way now i
[00:26:10] [SPEAKER_01]: was just describing our housing
[00:26:12] [SPEAKER_01]: market where what we do is we try
[00:26:15] [SPEAKER_01]: to actually mimic the way houses get
[00:26:17] [SPEAKER_01]: bought and sold so i i discussed how
[00:26:20] [SPEAKER_01]: we have buyers we have sellers the
[00:26:24] [SPEAKER_01]: the and we have households the
[00:26:27] [SPEAKER_01]: household might decide i'm just
[00:26:29] [SPEAKER_01]: going to rent or they might decide
[00:26:31] [SPEAKER_01]: i want to buy and so we have a
[00:26:33] [SPEAKER_01]: module that first of all tries to
[00:26:36] [SPEAKER_01]: capture that the the essence of
[00:26:39] [SPEAKER_01]: that decision what is it that make
[00:26:40] [SPEAKER_01]: people decide they want to rent
[00:26:42] [SPEAKER_01]: versus buy they have expectations
[00:26:44] [SPEAKER_01]: about the future those we would argue
[00:26:47] [SPEAKER_01]: are significantly driven by what the
[00:26:48] [SPEAKER_01]: housing market's been doing
[00:26:49] [SPEAKER_01]: housing market's been going up they
[00:26:52] [SPEAKER_01]: say hey my neighbor
[00:26:53] [SPEAKER_01]: bought their house their house is now
[00:26:55] [SPEAKER_01]: worth twice as much they just made
[00:26:57] [SPEAKER_01]: 100 000 bucks i'm a sucker i'm sitting
[00:26:59] [SPEAKER_01]: here renting
[00:27:00] [SPEAKER_01]: i need to go buy a house yeah so
[00:27:02] [SPEAKER_04]: you're missing out factor yeah right
[00:27:04] [SPEAKER_01]: so we try it we put in just that
[00:27:06] [SPEAKER_01]: kind of reasoning
[00:27:07] [SPEAKER_01]: into our model with all the
[00:27:09] [SPEAKER_01]: individual households
[00:27:10] [SPEAKER_01]: now then they you know they go to a
[00:27:13] [SPEAKER_01]: bank because they say they need a
[00:27:14] [SPEAKER_01]: loan the bank may or may not accept
[00:27:16] [SPEAKER_01]: them you know so we mimic
[00:27:18] [SPEAKER_01]: what the bank does what does the bank
[00:27:20] [SPEAKER_01]: do they look at their FICO score they
[00:27:21] [SPEAKER_01]: look at their
[00:27:23] [SPEAKER_01]: income they look at how big how
[00:27:25] [SPEAKER_01]: expensive a house they're trying to
[00:27:26] [SPEAKER_01]: buy they decide whether they're a good
[00:27:28] [SPEAKER_01]: risk
[00:27:28] [SPEAKER_01]: and they may or may not lend them
[00:27:30] [SPEAKER_01]: the money then if they do get the
[00:27:33] [SPEAKER_01]: money they do get the loan from the
[00:27:34] [SPEAKER_01]: bank
[00:27:34] [SPEAKER_01]: then they as i mentioned before they go
[00:27:36] [SPEAKER_01]: to the real estate agent who then
[00:27:38] [SPEAKER_01]: helps hook them up with the
[00:27:39] [SPEAKER_01]: sellers we try and match buyers and
[00:27:42] [SPEAKER_01]: sellers in the way it really
[00:27:43] [SPEAKER_01]: happens and so we run the whole thing
[00:27:45] [SPEAKER_01]: in a fairly realistic way mimicking what
[00:27:49] [SPEAKER_01]: actually happens
[00:27:50] [SPEAKER_04]: now for that to be for that to be
[00:27:51] [SPEAKER_04]: accurate you'd have to be looking at
[00:27:52] [SPEAKER_04]: that into in different income groups
[00:27:54] [SPEAKER_04]: and
[00:27:54] [SPEAKER_04]: different housing types and you know
[00:27:58] [SPEAKER_04]: different parts of the population that
[00:27:59] [SPEAKER_04]: segmentation i was talking about
[00:28:00] [SPEAKER_04]: earlier to be to be accurate
[00:28:01] [SPEAKER_01]: wouldn't we can do just that you
[00:28:03] [SPEAKER_01]: know we have
[00:28:04] [SPEAKER_01]: we there are various ways of doing it
[00:28:06] [SPEAKER_01]: from actually writing down all the
[00:28:07] [SPEAKER_01]: characteristics of the house
[00:28:09] [SPEAKER_01]: and having a model that you know
[00:28:11] [SPEAKER_01]: has some kind of valuation of the
[00:28:13] [SPEAKER_01]: house at what people want in our
[00:28:15] [SPEAKER_01]: earlier model we just
[00:28:16] [SPEAKER_01]: assigned houses qualities ranging from
[00:28:19] [SPEAKER_01]: high to low and and didn't even change
[00:28:22] [SPEAKER_01]: that and we still got
[00:28:23] [SPEAKER_01]: a lot of very interesting behavior so
[00:28:26] [SPEAKER_01]: one of the things we got
[00:28:27] [SPEAKER_01]: was we were able to simulate a bubble
[00:28:30] [SPEAKER_01]: we were able to actually
[00:28:32] [SPEAKER_01]: show depending on the way lending is
[00:28:34] [SPEAKER_01]: happening that
[00:28:36] [SPEAKER_01]: loosening lending policy was the
[00:28:38] [SPEAKER_01]: driver of that bubble it was that's a
[00:28:40] [SPEAKER_01]: much stronger driver
[00:28:41] [SPEAKER_01]: than changing interest rates which is
[00:28:44] [SPEAKER_01]: another suggested driver or
[00:28:46] [SPEAKER_01]: changing demography because we could
[00:28:48] [SPEAKER_01]: map that
[00:28:48] [SPEAKER_01]: so we put in all the realism we needed
[00:28:52] [SPEAKER_01]: to make the model match what was
[00:28:55] [SPEAKER_01]: actually
[00:28:56] [SPEAKER_01]: happening and make predictions about
[00:28:58] [SPEAKER_01]: what would happen
[00:28:59] [SPEAKER_01]: under different circumstances so for
[00:29:01] [SPEAKER_01]: the bank of england
[00:29:03] [SPEAKER_01]: they came we developed a model with
[00:29:05] [SPEAKER_01]: them
[00:29:05] [SPEAKER_01]: and and then they wanted to ask the
[00:29:08] [SPEAKER_01]: question
[00:29:09] [SPEAKER_01]: if we implement a restriction
[00:29:12] [SPEAKER_01]: because let me just say at the time
[00:29:15] [SPEAKER_01]: this was maybe six or eight years ago
[00:29:21] [SPEAKER_01]: england and london in particular was
[00:29:22] [SPEAKER_01]: in a major housing bubble
[00:29:24] [SPEAKER_01]: of course the bank of england didn't
[00:29:25] [SPEAKER_01]: want us to use that word bubble
[00:29:27] [SPEAKER_01]: anywhere
[00:29:27] [SPEAKER_01]: because they didn't want to scare
[00:29:29] [SPEAKER_01]: people but
[00:29:31] [SPEAKER_01]: but that was what's going on and
[00:29:32] [SPEAKER_01]: they wanted to damp the bubble
[00:29:35] [SPEAKER_01]: but they didn't want to pop the
[00:29:36] [SPEAKER_01]: bubble so much that there was a big
[00:29:38] [SPEAKER_01]: crash
[00:29:38] [SPEAKER_01]: and so they came up with the idea
[00:29:41] [SPEAKER_01]: let's impose a restriction on loan to
[00:29:44] [SPEAKER_01]: income ratio
[00:29:45] [SPEAKER_01]: so that banks can't loan to anybody
[00:29:48] [SPEAKER_01]: if the loan to income ratio
[00:29:51] [SPEAKER_01]: is more than
[00:29:52] [SPEAKER_01]: and this is annual income more than i
[00:29:54] [SPEAKER_01]: think it was three and a half
[00:29:55] [SPEAKER_01]: i might misremember the number
[00:29:58] [SPEAKER_01]: and so the question was what will it
[00:30:00] [SPEAKER_01]: what will happen if we do that so
[00:30:06] [SPEAKER_01]: we
[00:30:08] [SPEAKER_01]: reproducing a bubble and then we
[00:30:10] [SPEAKER_01]: implemented that policy we suddenly
[00:30:12] [SPEAKER_01]: switched it on
[00:30:13] [SPEAKER_01]: when the market was in that was
[00:30:16] [SPEAKER_01]: in that bubble and we showed that
[00:30:19] [SPEAKER_01]: that actually flattened the bubble out
[00:30:20] [SPEAKER_01]: that seemed to be just about right
[00:30:22] [SPEAKER_01]: not too hot not too cold
[00:30:24] [SPEAKER_01]: to
[00:30:25] [SPEAKER_01]: to not neither cause a crash nor
[00:30:28] [SPEAKER_01]: keep let the bubble keep rolling and
[00:30:30] [SPEAKER_01]: we presented that to them i don't
[00:30:32] [SPEAKER_01]: know how decisive our
[00:30:34] [SPEAKER_01]: analysis was but they implemented
[00:30:35] [SPEAKER_01]: that policy and
[00:30:37] [SPEAKER_01]: that was pretty much what happened
[00:30:39] [SPEAKER_04]: right and yet
[00:30:40] [SPEAKER_04]: it's way more than three and a half
[00:30:41] [SPEAKER_04]: times now
[00:30:43] [SPEAKER_04]: the ratio but yeah so steve this sort
[00:30:47] [SPEAKER_04]: of goes with
[00:30:47] [SPEAKER_04]: your thinking that you know we
[00:30:50] [SPEAKER_04]: talked about this many times that
[00:30:52] [SPEAKER_04]: house prices are largely determined by
[00:30:54] [SPEAKER_04]: what banks are prepared to lend
[00:30:56] [SPEAKER_03]: i'd just totally leverage driven and
[00:30:58] [SPEAKER_03]: the interesting thing is the joint
[00:30:59] [SPEAKER_03]: gets that result out of multi-agent
[00:31:01] [SPEAKER_03]: models
[00:31:01] [SPEAKER_03]: i get the same amount of aggregate
[00:31:03] [SPEAKER_03]: models and
[00:31:04] [SPEAKER_03]: the unifying thing first of all we both
[00:31:06] [SPEAKER_03]: abandon
[00:31:08] [SPEAKER_03]: any imposition of equilibrium on the
[00:31:10] [SPEAKER_03]: system whereas the mainstream
[00:31:11] [SPEAKER_03]: basically imposes equilibrium and then
[00:31:14] [SPEAKER_03]: the mathematics is how do we
[00:31:16] [SPEAKER_03]: make the make the system converge to
[00:31:18] [SPEAKER_03]: equilibrium despite shocks
[00:31:19] [SPEAKER_03]: we sort of say let's just put all the
[00:31:21] [SPEAKER_03]: maths together whether that's
[00:31:23] [SPEAKER_03]: top level math that i'm doing or
[00:31:25] [SPEAKER_03]: multi-agent behavior in a computer
[00:31:26] [SPEAKER_03]: system
[00:31:27] [SPEAKER_03]: with thousands of agents as doing
[00:31:29] [SPEAKER_03]: and trying to reproduce the actual
[00:31:31] [SPEAKER_03]: structure of markets
[00:31:32] [SPEAKER_03]: we watch and see what happens and i
[00:31:34] [SPEAKER_03]: think the thing i most
[00:31:35] [SPEAKER_03]: i like a lot about your book doing but
[00:31:37] [SPEAKER_03]: the i like the opening definition you
[00:31:39] [SPEAKER_03]: had of complexity
[00:31:40] [SPEAKER_03]: as being a complex system is one that
[00:31:42] [SPEAKER_03]: has emergent properties
[00:31:43] [SPEAKER_03]: in other words things happen in the
[00:31:45] [SPEAKER_03]: model that you don't program into it
[00:31:47] [SPEAKER_03]: or that i don't write in the
[00:31:48] [SPEAKER_03]: mathematics and yet they come out of
[00:31:50] [SPEAKER_03]: the model
[00:31:50] [SPEAKER_03]: and that's that that that
[00:31:53] [SPEAKER_03]: emergence of things which where the
[00:31:55] [SPEAKER_03]: whole is more than the sum of the
[00:31:56] [SPEAKER_03]: parts
[00:31:57] [SPEAKER_03]: that's the essential definition of
[00:31:59] [SPEAKER_03]: complexity but how do you get to a
[00:32:00] [SPEAKER_03]: point where
[00:32:01] [SPEAKER_04]: it still is manageable because if you
[00:32:05] [SPEAKER_04]: develop a model which starts to
[00:32:08] [SPEAKER_04]: develop its own complexity
[00:32:09] [SPEAKER_04]: in a way as it starts to realize what
[00:32:12] [SPEAKER_04]: factors are
[00:32:13] [SPEAKER_04]: influenced by what other factors how do
[00:32:15] [SPEAKER_04]: you keep control out of all of that
[00:32:16] [SPEAKER_04]: isn't there
[00:32:17] [SPEAKER_04]: the danger that you create a
[00:32:18] [SPEAKER_04]: monster that ultimately can't
[00:32:20] [SPEAKER_04]: understand
[00:32:21] [SPEAKER_01]: you know i'm living on a monster as
[00:32:23] [SPEAKER_01]: supposedly einstein said apparently
[00:32:25] [SPEAKER_01]: other people said this first
[00:32:27] [SPEAKER_01]: but a model should be as simple as
[00:32:30] [SPEAKER_01]: possible but no simpler and that's the
[00:32:33] [SPEAKER_01]: key to find the sweet spot
[00:32:35] [SPEAKER_01]: where you put in the things that
[00:32:36] [SPEAKER_01]: matter and you leave out the stuff
[00:32:38] [SPEAKER_01]: that doesn't matter
[00:32:39] [SPEAKER_01]: and will just confuse everybody and
[00:32:41] [SPEAKER_01]: secondly
[00:32:42] [SPEAKER_01]: in a agent-based model you know they
[00:32:45] [SPEAKER_01]: can be complicated
[00:32:46] [SPEAKER_01]: so you have to really build the
[00:32:48] [SPEAKER_01]: model with lots of diagnostic
[00:32:50] [SPEAKER_01]: indicators built in
[00:32:51] [SPEAKER_01]: so you can see what's happening and
[00:32:54] [SPEAKER_01]: you have to be able to strip the
[00:32:55] [SPEAKER_01]: model down
[00:32:56] [SPEAKER_01]: you have to be able to flexibly
[00:32:58] [SPEAKER_01]: change the model so that you can
[00:33:00] [SPEAKER_01]: pinpoint
[00:33:01] [SPEAKER_01]: what's causing what switch things on and
[00:33:04] [SPEAKER_01]: off
[00:33:05] [SPEAKER_01]: and and so there's a whole art to doing
[00:33:08] [SPEAKER_01]: that
[00:33:08] [SPEAKER_01]: well that is part of what we learn
[00:33:12] [SPEAKER_01]: as scientists
[00:33:13] [SPEAKER_03]: how do you actually don't if i can
[00:33:15] [SPEAKER_03]: sometimes in the met
[00:33:16] [SPEAKER_03]: my system's all your differential
[00:33:18] [SPEAKER_03]: equations and i've got direct
[00:33:19] [SPEAKER_03]: control over parameters so i can
[00:33:20] [SPEAKER_03]: change parameters of simulation see what
[00:33:22] [SPEAKER_03]: happens
[00:33:23] [SPEAKER_03]: what interface do you use your
[00:33:25] [SPEAKER_03]: modeling and which programming
[00:33:26] [SPEAKER_03]: languages are you using to build them
[00:33:28] [SPEAKER_01]: we're building most of the models in
[00:33:30] [SPEAKER_01]: python some of the parts get sped up
[00:33:33] [SPEAKER_01]: okay by using c plus plus and
[00:33:37] [SPEAKER_01]: you know we're using standard python
[00:33:39] [SPEAKER_01]: tools we're not using anything
[00:33:40] [SPEAKER_01]: very special to be honest so that is
[00:33:43] [SPEAKER_03]: the sort of thing you can show to a
[00:33:44] [SPEAKER_03]: politician or is something a
[00:33:45] [SPEAKER_03]: politician
[00:33:46] [SPEAKER_03]: uh would rather run a mile from so i
[00:33:49] [SPEAKER_01]: didn't quite understand the question
[00:33:50] [SPEAKER_03]: that's that's what the politician
[00:33:51] [SPEAKER_04]: would say
[00:33:51] [SPEAKER_04]: i mean is it all is is it all too
[00:33:54] [SPEAKER_04]: complicated for politicians i think is
[00:33:56] [SPEAKER_04]: what steve's saying
[00:33:56] [SPEAKER_01]: yeah ah ah well we're trying to make it
[00:34:01] [SPEAKER_01]: um uh through the use of visual tools
[00:34:04] [SPEAKER_01]: um so you can visualize what's
[00:34:06] [SPEAKER_01]: happening by
[00:34:08] [SPEAKER_01]: doing a good job of explaining things
[00:34:10] [SPEAKER_01]: making it so
[00:34:11] [SPEAKER_01]: anyone can play with it um and
[00:34:16] [SPEAKER_01]: uh we're trying to make it
[00:34:18] [SPEAKER_01]: accessible it's interesting when you
[00:34:19] [SPEAKER_01]: start talking let me let me say one
[00:34:21] [SPEAKER_01]: more thing
[00:34:21] [SPEAKER_01]: let me say one more thing a key is
[00:34:23] [SPEAKER_01]: that in our models
[00:34:24] [SPEAKER_01]: all the pieces are recognizable you know
[00:34:27] [SPEAKER_01]: here's a firm okay there's a firm what
[00:34:29] [SPEAKER_01]: is it doing it's producing stuff
[00:34:31] [SPEAKER_01]: and and here's another firm oh it's
[00:34:33] [SPEAKER_01]: producing different stuff
[00:34:34] [SPEAKER_01]: here's a household so you can burrow
[00:34:36] [SPEAKER_01]: in and everything looks like the
[00:34:38] [SPEAKER_01]: real world
[00:34:38] [SPEAKER_01]: it's a replica of what's happening in
[00:34:40] [SPEAKER_01]: the real world so from that point of
[00:34:42] [SPEAKER_01]: view
[00:34:42] [SPEAKER_01]: it's much easier to understand than
[00:34:45] [SPEAKER_01]: these agents that are all
[00:34:46] [SPEAKER_01]: taking each other into account and
[00:34:49] [SPEAKER_01]: maximizing their utility which you
[00:34:50] [SPEAKER_01]: can't directly measure
[00:34:52] [SPEAKER_01]: and um uh and and a lot of other stuff
[00:34:57] [SPEAKER_01]: isn't visible in the model but there
[00:34:58] [SPEAKER_01]: is politics if there's a firm there's
[00:35:00] [SPEAKER_01]: typically
[00:35:01] [SPEAKER_01]: a representative firm there's a single
[00:35:02] [SPEAKER_01]: firm that's producing
[00:35:04] [SPEAKER_01]: all goods and services i mean in ours
[00:35:07] [SPEAKER_01]: we've got lots of different kinds of
[00:35:09] [SPEAKER_01]: firms producing lots of different stuff
[00:35:11] [SPEAKER_04]: unfortunately we might end up with
[00:35:12] [SPEAKER_04]: one firm and that might just be
[00:35:13] [SPEAKER_04]: amazon but yeah
[00:35:14] [SPEAKER_04]: i take take your point but it's the
[00:35:17] [SPEAKER_04]: isn't the politics involved in here
[00:35:19] [SPEAKER_04]: it's interesting talking about talking
[00:35:20] [SPEAKER_04]: to politicians because
[00:35:21] [SPEAKER_04]: they could say well okay but i don't
[00:35:23] [SPEAKER_04]: agree with the assumptions you made
[00:35:24] [SPEAKER_04]: about the behaviors
[00:35:25] [SPEAKER_04]: i don't believe that agent will behave
[00:35:27] [SPEAKER_04]: in that particular way for example
[00:35:29] [SPEAKER_04]: that's where politics gets in
[00:35:30] [SPEAKER_04]: i mean that's what that that is why
[00:35:32] [SPEAKER_04]: politics is so entwined in economics
[00:35:34] [SPEAKER_04]: because there's disagreement about the
[00:35:35] [SPEAKER_04]: way
[00:35:35] [SPEAKER_04]: we behave or what people will do for
[00:35:38] [SPEAKER_04]: example will
[00:35:39] [SPEAKER_04]: you know you you're sending people
[00:35:41] [SPEAKER_04]: off to raranda it's going to stop
[00:35:42] [SPEAKER_04]: people migrating to australia or
[00:35:44] [SPEAKER_04]: trying to try to make their way to
[00:35:45] [SPEAKER_04]: australia everyone's got
[00:35:46] [SPEAKER_04]: you know you've you've got a
[00:35:47] [SPEAKER_04]: solution perhaps so everyone's good to
[00:35:48] [SPEAKER_04]: different
[00:35:48] [SPEAKER_04]: idea about what the impact of that
[00:35:50] [SPEAKER_01]: is going to be so two things
[00:35:53] [SPEAKER_01]: if if the politician says i don't
[00:35:55] [SPEAKER_01]: believe that people behave this way
[00:35:56] [SPEAKER_01]: and they look at this part of the
[00:35:57] [SPEAKER_01]: model say
[00:35:58] [SPEAKER_01]: i don't i don't believe this part
[00:36:00] [SPEAKER_01]: well we go okay if we think it's a
[00:36:02] [SPEAKER_01]: plausible
[00:36:03] [SPEAKER_01]: criticism we'll go okay let's change
[00:36:05] [SPEAKER_01]: that part it's easy to change stuff
[00:36:07] [SPEAKER_01]: in agent-based models if we really
[00:36:10] [SPEAKER_01]: think they've got that wrong when we
[00:36:11] [SPEAKER_01]: pull out
[00:36:12] [SPEAKER_01]: you know some psychological studies or
[00:36:14] [SPEAKER_01]: some data to say actually
[00:36:16] [SPEAKER_01]: this isn't the way people the people
[00:36:18] [SPEAKER_01]: are behaving the way
[00:36:19] [SPEAKER_01]: we think they're behaving they're not
[00:36:20] [SPEAKER_01]: behaving the way you're behaving
[00:36:22] [SPEAKER_01]: and the other point maybe even a
[00:36:24] [SPEAKER_01]: bigger point is
[00:36:25] [SPEAKER_01]: you're when you talk about what'll
[00:36:27] [SPEAKER_01]: happen when
[00:36:28] [SPEAKER_01]: we make this political decision will
[00:36:29] [SPEAKER_01]: people migrate here or there
[00:36:31] [SPEAKER_01]: to me those are hard questions i
[00:36:33] [SPEAKER_01]: actually think economics a lot of
[00:36:35] [SPEAKER_01]: economics
[00:36:35] [SPEAKER_01]: is a lot easier because the kind of
[00:36:38] [SPEAKER_01]: behaviors we're putting into our
[00:36:40] [SPEAKER_01]: models
[00:36:40] [SPEAKER_01]: are things that are generally
[00:36:43] [SPEAKER_01]: generally hard to argue with like
[00:36:45] [SPEAKER_01]: do value investors
[00:36:47] [SPEAKER_01]: buy undervalued assets you know
[00:36:51] [SPEAKER_01]: do financial
[00:36:53] [SPEAKER_01]: when people are trying to estimate
[00:36:55] [SPEAKER_01]: their volatility in a financial market
[00:36:57] [SPEAKER_01]: do they i mean we assume they take a
[00:37:01] [SPEAKER_01]: moving average of the last couple of
[00:37:02] [SPEAKER_01]: years of volatility
[00:37:03] [SPEAKER_01]: and use that to estimate future
[00:37:05] [SPEAKER_01]: volatility well that's what people
[00:37:07] [SPEAKER_01]: do
[00:37:08] [SPEAKER_01]: do they use value at risk to decide
[00:37:11] [SPEAKER_01]: what leverage to use
[00:37:13] [SPEAKER_01]: well the the bank of international
[00:37:15] [SPEAKER_01]: settlements tells them they have to
[00:37:16] [SPEAKER_01]: do that
[00:37:17] [SPEAKER_01]: so those are the kind of behaviors that
[00:37:20] [SPEAKER_01]: we put in
[00:37:21] [SPEAKER_01]: you know do people consume more or less
[00:37:23] [SPEAKER_01]: under these circumstances
[00:37:25] [SPEAKER_01]: so those aren't actually political
[00:37:27] [SPEAKER_01]: decisions
[00:37:29] [SPEAKER_01]: so most of the behaviors in the
[00:37:31] [SPEAKER_01]: model are not things
[00:37:33] [SPEAKER_01]: that republicans and democrats have a
[00:37:35] [SPEAKER_01]: different view about
[00:37:38] [SPEAKER_01]: so now as i said
[00:37:40] [SPEAKER_01]: if there are controversial bits we
[00:37:42] [SPEAKER_01]: can pull those out and say well look
[00:37:44] [SPEAKER_01]: does it really depend on that but let
[00:37:46] [SPEAKER_01]: me say one more thing here
[00:37:48] [SPEAKER_01]: a lot of the behavior of this these
[00:37:49] [SPEAKER_01]: models is surprisingly independent
[00:37:52] [SPEAKER_01]: of the assumptions about behavior as
[00:37:54] [SPEAKER_01]: long as
[00:37:55] [SPEAKER_01]: you have more or less a certain kind
[00:37:58] [SPEAKER_01]: of assumption
[00:37:58] [SPEAKER_01]: then you're going to get behavior that
[00:38:02] [SPEAKER_01]: typically is not too sensitively
[00:38:04] [SPEAKER_01]: dependent and part of what we do
[00:38:06] [SPEAKER_01]: in when we build the models as we
[00:38:08] [SPEAKER_01]: try and home in
[00:38:09] [SPEAKER_01]: where are the sensitive bits that
[00:38:11] [SPEAKER_01]: really matter
[00:38:12] [SPEAKER_01]: where are the parts that don't
[00:38:13] [SPEAKER_01]: matter so much
[00:38:13] [SPEAKER_01]: and that's already very informative
[00:38:17] [SPEAKER_01]: because well that's interesting in and
[00:38:19] [SPEAKER_01]: of itself
[00:38:20] [SPEAKER_01]: and secondly then we at least know
[00:38:23] [SPEAKER_01]: what are the bits that we really have
[00:38:25] [SPEAKER_01]: to get right
[00:38:26] [SPEAKER_01]: what are the bits that actually don't
[00:38:27] [SPEAKER_01]: matter very much
[00:38:28] [SPEAKER_01]: what about the bits that you've
[00:38:29] [SPEAKER_04]: missed out that you don't know
[00:38:30] [SPEAKER_04]: about
[00:38:31] [SPEAKER_01]: ah well there's always that
[00:38:33] [SPEAKER_01]: possibility one of the big
[00:38:34] [SPEAKER_01]: advantages of agent-based models
[00:38:36] [SPEAKER_01]: is it's easy to add new things
[00:38:39] [SPEAKER_01]: if if i have my you know
[00:38:41] [SPEAKER_01]: complicated dsg model
[00:38:42] [SPEAKER_01]: and somebody says well banking is
[00:38:45] [SPEAKER_01]: actually important
[00:38:46] [SPEAKER_01]: well it takes somebody a few years of
[00:38:48] [SPEAKER_01]: work and they have to probably
[00:38:50] [SPEAKER_01]: simplify other parts to put that in
[00:38:52] [SPEAKER_01]: whereas
[00:38:54] [SPEAKER_01]: for us things are pretty easy to add
[00:38:58] [SPEAKER_01]: and
[00:38:59] [SPEAKER_01]: and i think actually that's the main
[00:39:01] [SPEAKER_01]: virtue of these kind of models
[00:39:03] [SPEAKER_01]: is that we can put
[00:39:05] [SPEAKER_01]: we can put all the bits that matter
[00:39:07] [SPEAKER_01]: in
[00:39:09] [SPEAKER_01]: and and without having problems
[00:39:12] [SPEAKER_01]: solving the model
[00:39:13] [SPEAKER_03]: and this is because yeah that's
[00:39:14] [SPEAKER_03]: different to the neoclassical
[00:39:15] [SPEAKER_03]: because they're trying to actually
[00:39:16] [SPEAKER_03]: reach a predetermined output
[00:39:17] [SPEAKER_03]: they're
[00:39:18] [SPEAKER_03]: directly trying to show it reaches
[00:39:19] [SPEAKER_03]: equilibrium
[00:39:20] [SPEAKER_03]: so a lot of their assumptions are to
[00:39:21] [SPEAKER_03]: generate that outcome
[00:39:22] [SPEAKER_03]: whereas what you're trying to do
[00:39:24] [SPEAKER_03]: and i'm doing the same bit again
[00:39:25] [SPEAKER_03]: from top down rather than
[00:39:27] [SPEAKER_03]: bottom up multi-ocean
[00:39:28] [SPEAKER_03]: just say let's just reproduce as
[00:39:30] [SPEAKER_03]: close as we can the actual system
[00:39:32] [SPEAKER_03]: and see what happens
[00:39:33] [SPEAKER_03]: is there a danger though that
[00:39:37] [SPEAKER_04]: so we see in in in the stock market
[00:39:39] [SPEAKER_04]: or in you know in all sorts of
[00:39:40] [SPEAKER_04]: different asset classes but
[00:39:41] [SPEAKER_04]: particularly in the share market
[00:39:43] [SPEAKER_04]: where you have algorithmic trading
[00:39:46] [SPEAKER_04]: where
[00:39:46] [SPEAKER_04]: at some point everyone's starting to
[00:39:48] [SPEAKER_04]: almost follow the same algorithm
[00:39:50] [SPEAKER_04]: and it all of a sudden
[00:39:51] [SPEAKER_04]: pinpoints a particular
[00:39:53] [SPEAKER_04]: asset or group of assets
[00:39:55] [SPEAKER_04]: and you see
[00:39:56] [SPEAKER_04]: you'll see them all of a sudden
[00:39:58] [SPEAKER_04]: running out of control
[00:39:59] [SPEAKER_04]: because everybody has used modeling
[00:40:00] [SPEAKER_04]: to draw the same conclusion
[00:40:02] [SPEAKER_04]: isn't there a danger that if you
[00:40:04] [SPEAKER_04]: use
[00:40:05] [SPEAKER_04]: modeling like this not just on the
[00:40:06] [SPEAKER_04]: share market but on the broader
[00:40:07] [SPEAKER_04]: economy
[00:40:08] [SPEAKER_04]: something could run out of control
[00:40:10] [SPEAKER_04]: because everybody's running
[00:40:11] [SPEAKER_04]: similar systems and drawing the same conclusions
[00:40:14] [SPEAKER_04]: and they create havoc
[00:40:15] [SPEAKER_01]: well let's let's make the difference
[00:40:17] [SPEAKER_01]: between the models
[00:40:18] [SPEAKER_01]: the traders are using
[00:40:19] [SPEAKER_01]: and the models
[00:40:20] [SPEAKER_01]: the regulators are using
[00:40:22] [SPEAKER_01]: right as a regulator
[00:40:23] [SPEAKER_01]: you'd like to understand
[00:40:25] [SPEAKER_01]: what is happening in the market
[00:40:27] [SPEAKER_01]: and when we're in a danger
[00:40:28] [SPEAKER_01]: when are we in a danger zone
[00:40:30] [SPEAKER_01]: and so
[00:40:32] [SPEAKER_01]: I talk
[00:40:33] [SPEAKER_01]: been part of the book about
[00:40:35] [SPEAKER_01]: the financial models we're building
[00:40:36] [SPEAKER_01]: and we for example built
[00:40:38] [SPEAKER_01]: simulations
[00:40:38] [SPEAKER_01]: of financial markets
[00:40:41] [SPEAKER_01]: where we
[00:40:43] [SPEAKER_01]: we try and mimic the sort of algorithms
[00:40:45] [SPEAKER_01]: that trading firms actually use
[00:40:49] [SPEAKER_01]: and we then simulate
[00:40:52] [SPEAKER_01]: the way prices are getting formed
[00:40:54] [SPEAKER_01]: and what we can show
[00:40:55] [SPEAKER_01]: is that when certain things happen
[00:40:58] [SPEAKER_01]: markets can become unstable
[00:41:00] [SPEAKER_01]: and behave badly
[00:41:02] [SPEAKER_01]: so if you have too many trend followers
[00:41:05] [SPEAKER_01]: as we did say in the tech bubble of 2000
[00:41:08] [SPEAKER_01]: the market you get a bubble
[00:41:11] [SPEAKER_01]: if you have too much leverage
[00:41:13] [SPEAKER_01]: then you know what once again
[00:41:15] [SPEAKER_01]: you get bubbles
[00:41:16] [SPEAKER_01]: you get you get crashes
[00:41:18] [SPEAKER_01]: you get instabilities in the market
[00:41:21] [SPEAKER_01]: why because in both of those cases
[00:41:23] [SPEAKER_01]: you end up with situations
[00:41:25] [SPEAKER_01]: where when the market falls
[00:41:27] [SPEAKER_01]: that causes people to sell
[00:41:29] [SPEAKER_01]: which then causes the market to fall even more
[00:41:32] [SPEAKER_01]: which causes people to sell
[00:41:33] [SPEAKER_01]: which causes the market fall etc etc etc
[00:41:36] [SPEAKER_01]: and so you get these unstable feedback loops
[00:41:39] [SPEAKER_01]: and so what I'm advocating is that regulators
[00:41:43] [SPEAKER_01]: who have access if they want to
[00:41:45] [SPEAKER_01]: the regulators can know
[00:41:47] [SPEAKER_01]: who's using what kind of strategy
[00:41:49] [SPEAKER_01]: they could have simulations
[00:41:51] [SPEAKER_01]: which would be like
[00:41:53] [SPEAKER_01]: the equivalent for financial markets
[00:41:55] [SPEAKER_01]: of environmental impact statements
[00:41:56] [SPEAKER_01]: you know they could say
[00:41:59] [SPEAKER_01]: well oh we have this new thing
[00:42:00] [SPEAKER_01]: called mortgage-backed securities
[00:42:01] [SPEAKER_01]: what happens when we put this in
[00:42:04] [SPEAKER_01]: to the market
[00:42:04] [SPEAKER_01]: and then you'd be able to see
[00:42:06] [SPEAKER_01]: that if you have too many people
[00:42:08] [SPEAKER_01]: using mortgage-backed securities
[00:42:09] [SPEAKER_01]: and you have a housing crash
[00:42:11] [SPEAKER_01]: you're going to bring the whole economy down
[00:42:13] [SPEAKER_01]: so it's not it's a bit different
[00:42:16] [SPEAKER_01]: than what you said now
[00:42:17] [SPEAKER_01]: I think what you were worrying about
[00:42:19] [SPEAKER_01]: that there might be lock-in
[00:42:20] [SPEAKER_01]: because everybody's using
[00:42:21] [SPEAKER_01]: the same algorithm
[00:42:22] [SPEAKER_01]: yeah that's something
[00:42:24] [SPEAKER_01]: we really should worry about
[00:42:25] [SPEAKER_01]: we actually wrote a paper
[00:42:27] [SPEAKER_01]: about that for insurance markets
[00:42:29] [SPEAKER_01]: because in catastrophe insurance markets
[00:42:32] [SPEAKER_01]: you've got under solvency 2
[00:42:35] [SPEAKER_01]: the new regulation
[00:42:38] [SPEAKER_01]: insurance companies are required
[00:42:40] [SPEAKER_01]: to use certified models
[00:42:42] [SPEAKER_01]: when their pricing risks
[00:42:46] [SPEAKER_01]: catastrophe risks
[00:42:47] [SPEAKER_01]: and so there's three companies
[00:42:50] [SPEAKER_01]: that produce those certified models
[00:42:51] [SPEAKER_01]: and something like 80 percent
[00:42:53] [SPEAKER_01]: of the insurers use
[00:42:56] [SPEAKER_01]: the model produced by one company
[00:42:59] [SPEAKER_01]: now there's a worry that
[00:43:01] [SPEAKER_01]: just because of the kind of thing
[00:43:02] [SPEAKER_01]: you're worrying about
[00:43:03] [SPEAKER_01]: well what if that model's wrong
[00:43:05] [SPEAKER_01]: we know the models aren't perfect
[00:43:07] [SPEAKER_01]: if that model's wrong
[00:43:08] [SPEAKER_01]: in a particular way
[00:43:09] [SPEAKER_01]: and 80 percent of the insurance companies
[00:43:11] [SPEAKER_01]: are using that model
[00:43:12] [SPEAKER_01]: you could bring down
[00:43:13] [SPEAKER_01]: the whole insurance industry
[00:43:14] [SPEAKER_01]: and we did some simulations to show
[00:43:17] [SPEAKER_01]: actually it's better to have diverse models
[00:43:20] [SPEAKER_01]: even if those models
[00:43:22] [SPEAKER_01]: aren't quite as good
[00:43:25] [SPEAKER_01]: so it's we're trying to
[00:43:27] [SPEAKER_01]: we're actually showing your point
[00:43:29] [SPEAKER_01]: and showing the regulators
[00:43:30] [SPEAKER_01]: what they need to do
[00:43:32] [SPEAKER_01]: to not get stuck in that trap
[00:43:33] [SPEAKER_04]: the upshot is we want the regulators
[00:43:35] [SPEAKER_04]: to have a better model than everyone else
[00:43:37] [SPEAKER_04]: which seems to be what we're saying
[00:43:39] [SPEAKER_04]: and maybe they have more time
[00:43:40] [SPEAKER_04]: and more access to do that
[00:43:42] [SPEAKER_04]: and obviously they're probably
[00:43:43] [SPEAKER_04]: she's saying they're approaching it
[00:43:43] [SPEAKER_04]: from a different angle
[00:43:44] [SPEAKER_04]: they're trying to stop
[00:43:45] [SPEAKER_04]: what could happen
[00:43:46] [SPEAKER_04]: but what about big data as well
[00:43:49] [SPEAKER_04]: so is this the fact that
[00:43:50] [SPEAKER_04]: I know during covid
[00:43:52] [SPEAKER_04]: so I sort of work in this space
[00:43:53] [SPEAKER_04]: a little bit as well
[00:43:54] [SPEAKER_04]: you know a lot of people turn to
[00:43:56] [SPEAKER_04]: more immediate data
[00:43:58] [SPEAKER_04]: to try and make sense
[00:43:59] [SPEAKER_04]: of what was going on
[00:44:00] [SPEAKER_04]: so you know looking at
[00:44:01] [SPEAKER_04]: how many planes were in the air
[00:44:03] [SPEAKER_04]: at any particular time
[00:44:04] [SPEAKER_04]: or looking at traffic patterns in cities
[00:44:06] [SPEAKER_04]: there's a whole load of
[00:44:07] [SPEAKER_04]: you know near real-time data
[00:44:09] [SPEAKER_04]: which is increasingly
[00:44:10] [SPEAKER_04]: becoming available
[00:44:11] [SPEAKER_04]: so I guess if you start to use
[00:44:12] [SPEAKER_04]: more and more of that
[00:44:13] [SPEAKER_04]: then you can make these models
[00:44:15] [SPEAKER_04]: more real-time
[00:44:16] [SPEAKER_04]: so you can predict
[00:44:17] [SPEAKER_04]: you know if you've got a theory
[00:44:18] [SPEAKER_04]: about what might happen
[00:44:20] [SPEAKER_04]: if you've got the variables
[00:44:21] [SPEAKER_04]: and you can see how they change
[00:44:22] [SPEAKER_04]: day to day
[00:44:24] [SPEAKER_04]: you get a clearer idea of
[00:44:26] [SPEAKER_04]: you know when things could really
[00:44:27] [SPEAKER_04]: seriously go off the rails
[00:44:29] [SPEAKER_01]: yeah no that's certainly
[00:44:30] [SPEAKER_01]: an important goal
[00:44:32] [SPEAKER_01]: more and more people
[00:44:33] [SPEAKER_01]: are thinking about that
[00:44:34] [SPEAKER_01]: they're very remarkable correlations
[00:44:37] [SPEAKER_01]: you know I
[00:44:38] [SPEAKER_01]: somebody the Dutch statistics office
[00:44:40] [SPEAKER_01]: showed me a plot
[00:44:41] [SPEAKER_01]: they had taken
[00:44:43] [SPEAKER_01]: you know they put cables
[00:44:44] [SPEAKER_01]: across the roads
[00:44:44] [SPEAKER_01]: to measure how many cars
[00:44:45] [SPEAKER_01]: were on the road
[00:44:47] [SPEAKER_01]: and because it varies
[00:44:48] [SPEAKER_01]: depending on things like oil prices
[00:44:51] [SPEAKER_01]: and it's remarkably correlated
[00:44:53] [SPEAKER_01]: with Dutch GDP
[00:44:56] [SPEAKER_01]: and so there are
[00:44:58] [SPEAKER_01]: more and more people
[00:44:59] [SPEAKER_01]: are trying to do now casting
[00:45:00] [SPEAKER_01]: so that we get real-time measures
[00:45:02] [SPEAKER_01]: of what's actually happening
[00:45:03] [SPEAKER_01]: in the economy
[00:45:04] [SPEAKER_01]: that's not a main thing
[00:45:06] [SPEAKER_01]: that we're doing
[00:45:06] [SPEAKER_01]: but it is something
[00:45:08] [SPEAKER_01]: that we're going to incorporate
[00:45:09] [SPEAKER_01]: into our models
[00:45:11] [SPEAKER_01]: as we go along
[00:45:12] [SPEAKER_01]: because you know
[00:45:13] [SPEAKER_01]: we don't want feedback
[00:45:14] [SPEAKER_01]: in real time
[00:45:15] [SPEAKER_04]: yeah look it's
[00:45:17] [SPEAKER_04]: I feel like we're going to
[00:45:18] [SPEAKER_04]: scrape the surface here
[00:45:19] [SPEAKER_04]: before we go though
[00:45:20] [SPEAKER_04]: just tell me about
[00:45:22] [SPEAKER_04]: you know your
[00:45:23] [SPEAKER_04]: you started down this road
[00:45:25] [SPEAKER_04]: trying to beat roulette tables
[00:45:28] [SPEAKER_04]: and tell me a bit
[00:45:29] [SPEAKER_04]: about that
[00:45:30] [SPEAKER_04]: and where your shoe came into it
[00:45:32] [SPEAKER_01]: all right so first of all
[00:45:33] [SPEAKER_01]: we didn't try
[00:45:34] [SPEAKER_01]: we did it
[00:45:35] [SPEAKER_01]: well okay sorry
[00:45:37] [SPEAKER_01]: but but
[00:45:38] [SPEAKER_01]: well it's a
[00:45:40] [SPEAKER_01]: in a sense
[00:45:40] [SPEAKER_01]: it's a very simple idea
[00:45:41] [SPEAKER_01]: we were I'm a physicist
[00:45:44] [SPEAKER_01]: my friend Norman Packard
[00:45:46] [SPEAKER_01]: came up with the idea
[00:45:46] [SPEAKER_01]: he said look roulette tables
[00:45:48] [SPEAKER_01]: they're just
[00:45:49] [SPEAKER_01]: it's just a physical system
[00:45:50] [SPEAKER_01]: it's a rolling ball
[00:45:51] [SPEAKER_01]: on a circular track
[00:45:52] [SPEAKER_01]: with friction
[00:45:54] [SPEAKER_01]: so we should be able
[00:45:55] [SPEAKER_01]: to write down the equation of motion
[00:45:57] [SPEAKER_01]: that the ball is following
[00:45:59] [SPEAKER_01]: and make measurements
[00:46:00] [SPEAKER_01]: of the position
[00:46:01] [SPEAKER_01]: and velocity of the ball
[00:46:02] [SPEAKER_01]: and then predict
[00:46:03] [SPEAKER_01]: when it's going to fall off the track
[00:46:05] [SPEAKER_01]: and what numbers
[00:46:06] [SPEAKER_01]: it will fall onto
[00:46:08] [SPEAKER_01]: so we bought a roulette wheel
[00:46:10] [SPEAKER_01]: we solved the equations of motion
[00:46:14] [SPEAKER_01]: we built
[00:46:15] [SPEAKER_01]: the first wearable computer
[00:46:17] [SPEAKER_01]: went under one armpit
[00:46:19] [SPEAKER_01]: the other
[00:46:20] [SPEAKER_01]: with a pack of 12 AA batteries
[00:46:22] [SPEAKER_01]: under the other armpit
[00:46:23] [SPEAKER_01]: this was 1977
[00:46:26] [SPEAKER_01]: so early days of computing
[00:46:30] [SPEAKER_01]: contemporaneous
[00:46:30] [SPEAKER_01]: with the very first
[00:46:31] [SPEAKER_01]: Apple computer
[00:46:33] [SPEAKER_01]: we had switches in our toes
[00:46:36] [SPEAKER_01]: in our shoes
[00:46:37] [SPEAKER_01]: that we operated with our toes
[00:46:39] [SPEAKER_01]: to measure the position
[00:46:42] [SPEAKER_01]: and velocity of the ball
[00:46:42] [SPEAKER_01]: the croupier releases the ball
[00:46:44] [SPEAKER_01]: there's typically about 15 seconds
[00:46:46] [SPEAKER_01]: before the ball hits the track
[00:46:48] [SPEAKER_01]: and they don't close the bets
[00:46:49] [SPEAKER_01]: until about one second
[00:46:51] [SPEAKER_01]: before the ball exits the track
[00:46:53] [SPEAKER_01]: onto the wheel
[00:46:54] [SPEAKER_01]: and so we would click
[00:46:57] [SPEAKER_01]: when the ball passed
[00:46:58] [SPEAKER_01]: a reference point
[00:46:59] [SPEAKER_01]: click again
[00:47:00] [SPEAKER_01]: when it passed
[00:47:01] [SPEAKER_01]: that same reference point again
[00:47:03] [SPEAKER_01]: and therefore
[00:47:04] [SPEAKER_01]: we could measure
[00:47:05] [SPEAKER_01]: the velocity of the ball
[00:47:06] [SPEAKER_01]: we knew the position of the ball
[00:47:08] [SPEAKER_01]: we plugged that into the equations
[00:47:09] [SPEAKER_01]: we would predict
[00:47:11] [SPEAKER_01]: where the ball was going to exit
[00:47:13] [SPEAKER_01]: and that gave us on balance
[00:47:15] [SPEAKER_01]: about a 20% edge over the house
[00:47:17] [SPEAKER_01]: so is there a photograph
[00:47:19] [SPEAKER_04]: with all the security guards
[00:47:20] [SPEAKER_04]: every casino on the planet
[00:47:22] [SPEAKER_04]: to not allow you in
[00:47:23] [SPEAKER_04]: as a result of all of that?
[00:47:26] [SPEAKER_01]: Well we did have to make
[00:47:28] [SPEAKER_01]: fast exits from a few casinos
[00:47:31] [SPEAKER_01]: you know we never
[00:47:33] [SPEAKER_01]: we never really
[00:47:35] [SPEAKER_01]: we were a bit chicken to be honest
[00:47:37] [SPEAKER_01]: we didn't want our kneecaps
[00:47:38] [SPEAKER_01]: getting broken
[00:47:39] [SPEAKER_01]: and there were well documented
[00:47:41] [SPEAKER_01]: examples of people getting beaten up
[00:47:43] [SPEAKER_01]: in the back room
[00:47:46] [SPEAKER_01]: so we never really went
[00:47:48] [SPEAKER_01]: for the jugular
[00:47:48] [SPEAKER_01]: and made the big bucks
[00:47:49] [SPEAKER_01]: we'd hoped for
[00:47:50] [SPEAKER_01]: but we did have a lot of fun
[00:47:52] [SPEAKER_01]: and we did
[00:47:52] [SPEAKER_01]: and let me say
[00:47:54] [SPEAKER_01]: we also had a lot of hardware problems
[00:47:55] [SPEAKER_01]: we were really pushing the boundary
[00:47:57] [SPEAKER_01]: of what could be done in that era
[00:47:59] [SPEAKER_01]: but it was an amazing experience
[00:48:02] [SPEAKER_01]: that taught me a lot
[00:48:03] [SPEAKER_01]: and we had a lot of fun
[00:48:04] [SPEAKER_01]: Seems like one final question for you then
[00:48:07] [SPEAKER_04]: unless Steve's got one to add in
[00:48:09] [SPEAKER_04]: I'm pleased to Steve
[00:48:10] [SPEAKER_04]: because I feel like
[00:48:11] [SPEAKER_04]: you know you're not
[00:48:11] [SPEAKER_04]: had a chance to get a word in edgeways
[00:48:13] [SPEAKER_04]: not entirely my fault
[00:48:14] [SPEAKER_04]: I have to add
[00:48:16] [SPEAKER_04]: so it seems to me
[00:48:17] [SPEAKER_04]: that economics suffers from the problem
[00:48:19] [SPEAKER_04]: it doesn't have a fundamental law
[00:48:21] [SPEAKER_04]: like there's no gravity
[00:48:22] [SPEAKER_04]: there's nothing there
[00:48:23] [SPEAKER_04]: which is indisputable
[00:48:26] [SPEAKER_04]: and so it's open slather in a way
[00:48:29] [SPEAKER_04]: that there's no
[00:48:30] [SPEAKER_03]: In fact there's one thing
[00:48:31] [SPEAKER_03]: Doyne said in his
[00:48:32] [SPEAKER_03]: would actually totally
[00:48:34] [SPEAKER_03]: 100% consider with me
[00:48:35] [SPEAKER_03]: and he said double entry bookkeeping
[00:48:37] [SPEAKER_03]: is the conservation law
[00:48:39] [SPEAKER_03]: that applies
[00:48:40] [SPEAKER_03]: every transfer of money
[00:48:41] [SPEAKER_03]: from one point goes to another
[00:48:42] [SPEAKER_03]: you have conservation
[00:48:43] [SPEAKER_03]: not of money
[00:48:44] [SPEAKER_03]: because money can be expanded
[00:48:46] [SPEAKER_03]: or contracted
[00:48:46] [SPEAKER_03]: expand the assets
[00:48:47] [SPEAKER_03]: and liabilities side
[00:48:48] [SPEAKER_03]: but that's something
[00:48:49] [SPEAKER_03]: that I noticed in your book
[00:48:50] [SPEAKER_03]: which is 100% compatible
[00:48:52] [SPEAKER_03]: with what I've done
[00:48:52] [SPEAKER_03]: with my Ravel software
[00:48:54] [SPEAKER_03]: which includes double entry bookkeeping
[00:48:55] [SPEAKER_03]: at the integrated level
[00:48:56] [SPEAKER_03]: the aggregate system
[00:48:58] [SPEAKER_03]: so there are some of those
[00:48:59] [SPEAKER_03]: principles like that
[00:49:00] [SPEAKER_03]: I think there's a lack
[00:49:02] [SPEAKER_03]: of a sort of a grand narrative
[00:49:04] [SPEAKER_03]: in the multi-agent approach
[00:49:06] [SPEAKER_03]: that we use
[00:49:07] [SPEAKER_03]: but what you're doing
[00:49:08] [SPEAKER_03]: fundamentally is trying to actually
[00:49:10] [SPEAKER_03]: build a simulacrum
[00:49:12] [SPEAKER_03]: of the real world
[00:49:13] [SPEAKER_03]: and if you pull that off
[00:49:14] [SPEAKER_03]: then what happens in your model
[00:49:16] [SPEAKER_03]: pretty much captures
[00:49:17] [SPEAKER_03]: what happens in the real world
[00:49:18] [SPEAKER_04]: Yeah and that was going to be
[00:49:19] [SPEAKER_04]: the question is leading to
[00:49:21] [SPEAKER_04]: does that then mean
[00:49:22] [SPEAKER_04]: that you are in a position then
[00:49:24] [SPEAKER_04]: to challenge some of the falsehoods
[00:49:26] [SPEAKER_04]: that exist amongst economists
[00:49:28] [SPEAKER_04]: are we going to have an
[00:49:29] [SPEAKER_04]: you know is the ultimate conclusion
[00:49:31] [SPEAKER_04]: of what you're doing Joy
[00:49:32] [SPEAKER_04]: and that we have a
[00:49:33] [SPEAKER_04]: a whole new approach to economics
[00:49:35] [SPEAKER_04]: and we do start to get some
[00:49:36] [SPEAKER_04]: fundamental agreement
[00:49:37] [SPEAKER_04]: on what are the basics
[00:49:39] [SPEAKER_01]: Yeah that's the goal
[00:49:42] [SPEAKER_01]: you know we're creating
[00:49:44] [SPEAKER_01]: an alternative way to do things
[00:49:45] [SPEAKER_01]: that I think has the potential
[00:49:46] [SPEAKER_01]: to produce more reliable answers
[00:49:49] [SPEAKER_01]: so that we reduce
[00:49:51] [SPEAKER_01]: the amount of disagreement
[00:49:53] [SPEAKER_01]: and maybe then
[00:49:56] [SPEAKER_01]: can make better decisions
[00:49:57] [SPEAKER_01]: about how we should
[00:49:59] [SPEAKER_01]: go into the future
[00:50:00] [SPEAKER_04]: I'm not quite sure
[00:50:01] [SPEAKER_04]: what politicians do then
[00:50:02] [SPEAKER_04]: you're still taking their job away
[00:50:04] [SPEAKER_01]: because both sides
[00:50:05] [SPEAKER_01]: no no
[00:50:06] [SPEAKER_01]: both sides of the house
[00:50:06] [SPEAKER_04]: would have to agree with each other
[00:50:07] [SPEAKER_01]: there's a lot more
[00:50:08] [SPEAKER_01]: to a lot more
[00:50:09] [SPEAKER_01]: to being a politician
[00:50:10] [SPEAKER_01]: than economics
[00:50:10] [SPEAKER_01]: you know we still have to understand
[00:50:13] [SPEAKER_01]: social science
[00:50:14] [SPEAKER_01]: more broadly sociology
[00:50:16] [SPEAKER_01]: anthropology psychology
[00:50:18] [SPEAKER_01]: politicians are never going to
[00:50:20] [SPEAKER_01]: cover all of that either
[00:50:22] [SPEAKER_01]: but but it's really
[00:50:24] [SPEAKER_01]: about collective behavior
[00:50:26] [SPEAKER_01]: of human beings
[00:50:27] [SPEAKER_01]: economics is an important part
[00:50:29] [SPEAKER_01]: of guiding the collective behavior
[00:50:31] [SPEAKER_01]: of human beings
[00:50:32] [SPEAKER_01]: but so far I have to say
[00:50:33] [SPEAKER_01]: I think we're pretty bad
[00:50:35] [SPEAKER_01]: at guiding our collective behavior
[00:50:38] [SPEAKER_01]: something we really need
[00:50:39] [SPEAKER_01]: to get better at
[00:50:40] [SPEAKER_01]: but that's that's a long-range project
[00:50:42] [SPEAKER_01]: probably over my pay grade
[00:50:43] [SPEAKER_04]: What's so what is the
[00:50:45] [SPEAKER_04]: so when does the big change happen
[00:50:47] [SPEAKER_04]: when does all your work
[00:50:48] [SPEAKER_04]: accumulate to a point
[00:50:50] [SPEAKER_04]: where people are going
[00:50:50] [SPEAKER_04]: oh well you know economics is
[00:50:52] [SPEAKER_04]: there is some agreement now
[00:50:53] [SPEAKER_04]: on economics
[00:50:54] [SPEAKER_04]: we can agree on a few principles
[00:50:56] [SPEAKER_04]: beyond you know
[00:50:58] [SPEAKER_04]: and actually understanding
[00:50:58] [SPEAKER_04]: as you take double entry bookkeeping
[00:51:00] [SPEAKER_04]: you can understand why
[00:51:00] [SPEAKER_04]: that's important for example
[00:51:01] [SPEAKER_01]: yeah I don't think
[00:51:02] [SPEAKER_01]: it's a matter of principles
[00:51:04] [SPEAKER_01]: because I don't think
[00:51:05] [SPEAKER_01]: I mean I agree with Steve
[00:51:06] [SPEAKER_01]: that double entry bookkeeping
[00:51:07] [SPEAKER_01]: is a very important principle
[00:51:08] [SPEAKER_01]: that underlies everything
[00:51:10] [SPEAKER_01]: but but I think it's more
[00:51:12] [SPEAKER_01]: a matter of methods
[00:51:14] [SPEAKER_01]: that actually are effective
[00:51:16] [SPEAKER_01]: at answering cause-effect relationships
[00:51:18] [SPEAKER_01]: if we enact this policy
[00:51:20] [SPEAKER_01]: where will that take us
[00:51:21] [SPEAKER_01]: if we enact another policy
[00:51:24] [SPEAKER_01]: where will that take us
[00:51:25] [SPEAKER_01]: so that's the kind of goal I have
[00:51:28] [SPEAKER_01]: just better guidance
[00:51:30] [SPEAKER_01]: even though a lot of this may be
[00:51:33] [SPEAKER_01]: from tools that don't operate
[00:51:35] [SPEAKER_01]: under some grand principles
[00:51:38] [SPEAKER_04]: it's just better models
[00:51:40] [SPEAKER_04]: and do you think these models
[00:51:41] [SPEAKER_04]: are helping central banks now
[00:51:43] [SPEAKER_04]: as they come out of the crisis
[00:51:45] [SPEAKER_04]: we've seen are they
[00:51:46] [SPEAKER_04]: taking a more measured approach
[00:51:47] [SPEAKER_04]: as a result of this sort of modeling
[00:51:49] [SPEAKER_01]: well I haven't checked in
[00:51:50] [SPEAKER_01]: with my colleagues
[00:51:51] [SPEAKER_01]: at the Bank of Canada lately
[00:51:52] [SPEAKER_01]: but they were quite happy
[00:51:54] [SPEAKER_01]: that their agent-based model
[00:51:56] [SPEAKER_01]: was giving better inflation predictions
[00:51:58] [SPEAKER_01]: than the other models
[00:51:59] [SPEAKER_01]: the Bank of Canada is using
[00:52:00] [SPEAKER_01]: which means you know
[00:52:03] [SPEAKER_01]: the people who are running
[00:52:04] [SPEAKER_01]: the Bank of Canada
[00:52:05] [SPEAKER_01]: start to pay more attention
[00:52:07] [SPEAKER_01]: to their model
[00:52:07] [SPEAKER_01]: and so my prediction
[00:52:09] [SPEAKER_01]: is over the next 10 years
[00:52:11] [SPEAKER_01]: we're going to see a flowering
[00:52:12] [SPEAKER_01]: of these kind of models
[00:52:13] [SPEAKER_01]: and we're going to see
[00:52:15] [SPEAKER_01]: more and more attention
[00:52:16] [SPEAKER_01]: getting paid to them
[00:52:18] [SPEAKER_01]: by decision makers
[00:52:19] [SPEAKER_01]: who make the decisions that count
[00:52:22] [SPEAKER_04]: so I'm going to leave
[00:52:22] [SPEAKER_04]: the final word with Steve actually
[00:52:24] [SPEAKER_04]: if you don't mind doing
[00:52:25] [SPEAKER_04]: because I know you hate central banks
[00:52:26] [SPEAKER_04]: that's quite right
[00:52:27] [SPEAKER_04]: might you like them a bit more
[00:52:29] [SPEAKER_04]: if they use these models
[00:52:29] [SPEAKER_04]: a bit more
[00:52:30] [SPEAKER_03]: well see this is the trouble
[00:52:31] [SPEAKER_03]: if they're like for example
[00:52:32] [SPEAKER_03]: of this Bank of England
[00:52:35] [SPEAKER_03]: before the financial crisis came out
[00:52:37] [SPEAKER_03]: I was visiting the bank
[00:52:38] [SPEAKER_03]: and shortly afterwards as well
[00:52:40] [SPEAKER_03]: and they'd appointed some post-Keynesians
[00:52:42] [SPEAKER_03]: this is actually under
[00:52:43] [SPEAKER_03]: I've forgotten his name right now
[00:52:45] [SPEAKER_03]: but the guy now on the RSA
[00:52:47] [SPEAKER_03]: these days
[00:52:48] [SPEAKER_03]: Andy Haldane
[00:52:51] [SPEAKER_03]: yeah very very funny guys
[00:52:52] [SPEAKER_03]: you're probably aware
[00:52:54] [SPEAKER_03]: so Andy brought in
[00:52:55] [SPEAKER_03]: a lot of physicists
[00:52:56] [SPEAKER_03]: he brought in people
[00:52:56] [SPEAKER_03]: doing non-linear dynamics
[00:52:57] [SPEAKER_03]: and so on
[00:52:58] [SPEAKER_03]: and that actually gave us
[00:52:59] [SPEAKER_03]: an alternative picture
[00:53:01] [SPEAKER_03]: but there's a huge amount
[00:53:02] [SPEAKER_03]: of pushback from the neoclassicals
[00:53:04] [SPEAKER_03]: and the thing is
[00:53:04] [SPEAKER_03]: they really don't want other people
[00:53:06] [SPEAKER_03]: like you and me on their turf
[00:53:07] [SPEAKER_03]: and they don't want to continue
[00:53:08] [SPEAKER_03]: reaching conclusions
[00:53:09] [SPEAKER_03]: about capitalism
[00:53:11] [SPEAKER_03]: that they think aren't ideological
[00:53:12] [SPEAKER_03]: which fundamentally aren't ideological
[00:53:14] [SPEAKER_03]: about it being a way of utility
[00:53:15] [SPEAKER_03]: maximizing and reaching equilibrium
[00:53:17] [SPEAKER_03]: and so on
[00:53:18] [SPEAKER_03]: so the main resistance we face
[00:53:20] [SPEAKER_03]: isn't from the public
[00:53:21] [SPEAKER_03]: or from the politicians
[00:53:22] [SPEAKER_03]: it's from conventional economists
[00:53:23] [SPEAKER_03]: who don't like people like us
[00:53:25] [SPEAKER_03]: pushing in on their turf
[00:53:26] [SPEAKER_03]: Doyne
[00:53:27] [SPEAKER_03]: yeah I agree with that
[00:53:29] [SPEAKER_04]: all right well that's a good piece
[00:53:31] [SPEAKER_04]: point to leave it on
[00:53:32] [SPEAKER_04]: I feel like we could talk forever
[00:53:33] [SPEAKER_04]: we've got to get you on again soon
[00:53:34] [SPEAKER_04]: Doyne
[00:53:34] [SPEAKER_04]: but thanks for the last half hour
[00:53:36] [SPEAKER_04]: so very happy
[00:53:37] [SPEAKER_04]: thank you very much
[00:53:37] [SPEAKER_04]: for having me on your show
[00:53:39] [SPEAKER_04]: and that's it for this week
[00:53:40] [SPEAKER_04]: don't forget that you can get this podcast early
[00:53:43] [SPEAKER_04]: ad free if you are a supporter of Steve Keane
[00:53:46] [SPEAKER_04]: on Patreon or on Substack
[00:53:47] [SPEAKER_04]: otherwise with the ads in the middle
[00:53:49] [SPEAKER_04]: middle of the week each week
[00:53:50] [SPEAKER_04]: and that's it for this week
[00:53:51] [SPEAKER_04]: catch you again next week
[00:53:52] [SPEAKER_04]: for another edition of
[00:53:53] [SPEAKER_04]: the Debunking Economics podcast
[00:53:55] [SPEAKER_04]: thanks for listening
[00:53:55] [SPEAKER_02]: the Debunking Economics podcast
[00:54:02] [SPEAKER_04]: If you've enjoyed listening to Debunking Economics
[00:54:05] [SPEAKER_04]: even if you haven't
[00:54:06] [SPEAKER_04]: you might also enjoy the Y-curve
[00:54:09] [SPEAKER_04]: each week
[00:54:10] [SPEAKER_04]: Roger Herring and I
[00:54:11] [SPEAKER_04]: talk to a guest about a topic
[00:54:12] [SPEAKER_04]: that is very much in the news that week
[00:54:14] [SPEAKER_04]: it's lively, it's fun, it's informative
[00:54:16] [SPEAKER_04]: what more could you want?
[00:54:18] [SPEAKER_04]: So search the Y-curve
[00:54:20] [SPEAKER_04]: in your favourite podcast app
[00:54:22] [SPEAKER_04]: or go to ycurve.com to listen