WEBVTT NOTE The Origins of Money: Computer Simulations for Austrian Insights 1 00:00:00.000 --> 00:00:30.000 So my paper is about computer simulations for Austrian insights and there is, I before I start I want to make just a fast question for you because every time I try to present a paper I really have no idea on how many people in the room are confident with these agent based simulations so is there anybody who knows what they are? 2 00:00:30.000 --> 00:00:33.760 Yes, it's pretty good, okay. 3 00:00:33.760 --> 00:00:38.760 So, and recently there's been quite a vivid debate 4 00:00:41.600 --> 00:00:46.600 if evolutionary game theory and game theory 5 00:00:47.300 --> 00:00:50.960 and these kind of simulations fit 6 00:00:50.960 --> 00:00:55.960 with the methodological principles of Austrian economics. 7 00:00:56.080 --> 00:00:58.960 Although the default answer, of course, 8 00:00:58.960 --> 00:01:03.360 is no for, I'm sure you know most of the reasons, 9 00:01:03.360 --> 00:01:06.080 but we can still give it a try, I think. 10 00:01:06.080 --> 00:01:10.720 And despite all this debate, I think there, 11 00:01:10.720 --> 00:01:13.240 or maybe there are, but I don't know them, 12 00:01:13.240 --> 00:01:15.880 there are no actual agent-based simulations 13 00:01:15.880 --> 00:01:18.800 of made just for the ostriches, 14 00:01:20.080 --> 00:01:22.760 like some kind of custom simulations for them. 15 00:01:22.760 --> 00:01:27.040 So I tried to make one, and I chose to model 16 00:01:27.040 --> 00:01:32.040 and the other theory of Carl Menger on the origins of money, 17 00:01:32.840 --> 00:01:36.040 because I think it's quite simple in a way 18 00:01:36.040 --> 00:01:39.840 and it's really, and it's interesting and fun. 19 00:01:39.840 --> 00:01:43.460 So first, what are these, what are these models? 20 00:01:43.460 --> 00:01:46.800 And they're also known as artificial societies 21 00:01:46.800 --> 00:01:49.240 or bottom-up models. 22 00:01:49.240 --> 00:01:53.240 And they, basically they, what you do, 23 00:01:53.240 --> 00:01:58.160 So you simulate the interaction of autonomous individuals. 24 00:01:58.160 --> 00:02:02.160 They are used both in the natural sciences 25 00:02:02.160 --> 00:02:05.840 and in the social sciences, like in biology, 26 00:02:05.840 --> 00:02:10.480 but also in physics, and of course in economics. 27 00:02:10.480 --> 00:02:15.480 And so they're interesting when you want to study complexity. 28 00:02:16.120 --> 00:02:20.560 So if we think, this is, I don't know if you can see them, 29 00:02:20.560 --> 00:02:25.560 and the schooling of fish, which is complex behavior, 30 00:02:29.100 --> 00:02:34.160 which emerges from, of course, fishes are not that smart 31 00:02:34.160 --> 00:02:38.200 and there is no benevolent dictator 32 00:02:38.200 --> 00:02:41.120 who makes the fish swim like that. 33 00:02:41.120 --> 00:02:45.860 And of course, they do it and it's an advantage for them 34 00:02:45.860 --> 00:02:47.840 to defend against predators. 35 00:02:47.840 --> 00:03:04.840 The fishes are not aware of what they're doing, they just swim around according to some very simple behavior rule, and what emerges is a complex behavior, like schooling. 36 00:03:04.840 --> 00:03:15.840 And you can use a simulation to study all these kind of complex things like traffic jams or how a crowd will behave. 37 00:03:15.840 --> 00:03:20.840 So whenever you have a lot of individuals, 38 00:03:20.960 --> 00:03:25.800 and you want to see what emerges from their interaction. 39 00:03:25.800 --> 00:03:29.400 So how do these models work? 40 00:03:29.400 --> 00:03:33.840 And well, basically, what they are, according to me, 41 00:03:33.840 --> 00:03:37.720 if I have to tell you in one minute, 42 00:03:37.720 --> 00:03:42.380 there are a way of doing thought experiments. 43 00:03:42.380 --> 00:03:45.940 There is nothing mysterious about them. 44 00:03:45.940 --> 00:03:48.320 And I think that thought experiments 45 00:03:48.320 --> 00:03:51.460 are one of the main tools in economics, 46 00:03:51.460 --> 00:03:53.300 and especially they're very precious 47 00:03:53.300 --> 00:03:54.960 for the Austrian economics. 48 00:03:54.960 --> 00:03:59.960 And I think, for example, even the rotating economy of Mises, 49 00:04:00.460 --> 00:04:01.660 that is a thought experiment. 50 00:04:01.660 --> 00:04:06.660 See, ask us to imagine what happens in such a kind of economy. 51 00:04:08.620 --> 00:04:10.840 So if we go to an easier example, 52 00:04:10.840 --> 00:04:13.780 and we go to Adam Smith, 53 00:04:14.960 --> 00:04:19.960 how the division of labor came into existence. 54 00:04:21.200 --> 00:04:24.080 He tells us to imagine a small village 55 00:04:24.080 --> 00:04:27.800 in which people have different skills. 56 00:04:27.800 --> 00:04:30.680 So when we read this passage, 57 00:04:30.680 --> 00:04:33.000 we really try to imagine this village, I think, 58 00:04:33.000 --> 00:04:35.360 and one guy is better at fishing, 59 00:04:35.360 --> 00:04:37.720 another guy is better at hunting, 60 00:04:37.720 --> 00:04:41.800 but we don't really imagine a big city, yeah? 61 00:04:41.800 --> 00:04:45.600 We can't, we can't say, okay, now I imagine 10,000 people 62 00:04:45.600 --> 00:04:50.360 or 100,000 and each of them is better doing something else 63 00:04:50.360 --> 00:04:55.360 and then we can't deduce the consequences in our mind. 64 00:04:55.680 --> 00:05:00.680 So maybe, what do the agent-based models 65 00:05:02.280 --> 00:05:05.840 is just to help our thought experiment 66 00:05:05.840 --> 00:05:07.920 with a computer. 67 00:05:07.920 --> 00:05:11.040 We could do them with paper and pencil 68 00:05:11.040 --> 00:05:14.880 and that's how Thomas Schelling did it, in fact, 69 00:05:14.880 --> 00:05:16.280 back in the 70s. 70 00:05:16.280 --> 00:05:19.880 So some appealing features of these kind of models 71 00:05:19.880 --> 00:05:24.520 are that agents are heterogeneous. 72 00:05:24.520 --> 00:05:29.440 Okay, so we can make each agent behave differently 73 00:05:29.440 --> 00:05:32.080 and have different characteristics. 74 00:05:32.080 --> 00:05:36.880 and then agents are not olympically rational 75 00:05:36.880 --> 00:05:41.200 and it's not interesting if we make a simulation 76 00:05:41.200 --> 00:05:45.600 which agents are gods and like ants or fish 77 00:05:45.600 --> 00:05:47.660 and it just doesn't make sense 78 00:05:47.660 --> 00:05:50.000 and it's probably very hard if not impossible 79 00:05:50.000 --> 00:05:55.000 for the computer to manage all these god-like agents 80 00:05:56.040 --> 00:06:01.040 and then the interaction and knowledge of agents are local 81 00:06:02.080 --> 00:06:04.560 They only know what's going on around them. 82 00:06:05.480 --> 00:06:08.060 And what is also very interesting, I think, 83 00:06:08.060 --> 00:06:12.320 for Austin Economics is that the causal mechanism 84 00:06:13.640 --> 00:06:15.200 can be followed step by step, 85 00:06:15.200 --> 00:06:17.740 meaning that the focus is on the process 86 00:06:17.740 --> 00:06:20.600 rather than on the final equilibrium. 87 00:06:20.600 --> 00:06:23.280 So for example, here we have ants, and as you know, 88 00:06:23.280 --> 00:06:28.280 ants find out the shortest path to the food, 89 00:06:28.360 --> 00:06:31.400 and they don't do it because they are very smart 90 00:06:31.400 --> 00:06:34.560 or because they are aware of what is the shortest 91 00:06:34.560 --> 00:06:38.920 but they just, it's an emerging pattern from there. 92 00:06:38.920 --> 00:06:41.580 So it's like a sort of swarm intelligence. 93 00:06:42.600 --> 00:06:45.160 So all these features I've been talking about 94 00:06:45.160 --> 00:06:48.880 should be, I think, appealing for Austrian economics 95 00:06:48.880 --> 00:06:53.880 and I think they really fit into 96 00:06:54.340 --> 00:06:58.060 at least some of von Hayek insights. 97 00:06:58.060 --> 00:07:00.200 Well, first we have a very strong commitment 98 00:07:00.200 --> 00:07:02.420 to Methodological Individualism. 99 00:07:02.420 --> 00:07:07.420 So we can, there is nothing but the individuals 100 00:07:07.740 --> 00:07:09.340 in these kind of models. 101 00:07:09.340 --> 00:07:14.340 And when we run a model, we usually aim at some kind 102 00:07:16.460 --> 00:07:19.300 of rational reconstruction, which will be 103 00:07:19.300 --> 00:07:24.300 some conjectural history, or principle explanation. 104 00:07:25.000 --> 00:07:29.660 We don't end up with a quantitative result. 105 00:07:29.660 --> 00:07:34.660 We will not say that we need 115 ants, 106 00:07:35.300 --> 00:07:40.300 which go after two bagels to have something interesting. 107 00:07:40.820 --> 00:07:43.820 Well, we can say that we have some interesting emerging 108 00:07:43.820 --> 00:07:47.980 pattern given this simple behavior rule, for example. 109 00:07:47.980 --> 00:07:52.820 And they are deductive, yes. 110 00:07:52.820 --> 00:07:56.060 There is nothing but what we put inside the model. 111 00:07:56.060 --> 00:08:01.060 The point is as we're not, they are interesting 112 00:08:01.220 --> 00:08:05.740 because we put what we want inside, 113 00:08:05.740 --> 00:08:08.620 but it's very hard for us to predict the outcome 114 00:08:08.620 --> 00:08:11.940 because we have a lot of agents interacting. 115 00:08:11.940 --> 00:08:15.060 And so they are strictly deductive 116 00:08:15.060 --> 00:08:19.260 and I think that fits with Austrian economics. 117 00:08:19.260 --> 00:08:24.260 And if we take a step forward 118 00:08:24.260 --> 00:08:29.260 forward and we imagine how a system like that evolves. 119 00:08:31.060 --> 00:08:34.260 I think that also fits with the Austrian economics 120 00:08:34.260 --> 00:08:36.740 and the thought of von Hayek, 121 00:08:36.740 --> 00:08:41.420 as we don't have ultra-rational individuals, 122 00:08:41.420 --> 00:08:44.020 individuals act conforming to norms, 123 00:08:44.020 --> 00:08:46.100 for example, social norms, 124 00:08:46.100 --> 00:08:49.220 and these norms are the result of 125 00:08:49.220 --> 00:08:51.980 a complex evolutionary process. 126 00:08:51.980 --> 00:08:56.980 and we can model that easily in a simulation. 127 00:08:57.420 --> 00:09:02.420 So I chose to give it a try with Menger's Origins of Money 128 00:09:03.700 --> 00:09:07.340 and it's good that we talked about it yesterday. 129 00:09:07.340 --> 00:09:10.040 Makes the talk easier. 130 00:09:10.040 --> 00:09:15.040 So Menger thought that money emerges spontaneously 131 00:09:16.520 --> 00:09:19.900 in society and it was not created by the state. 132 00:09:19.900 --> 00:09:29.900 And his explanation goes on like this, first he notices how different goods are saleable in different degrees. 133 00:09:29.900 --> 00:09:36.900 And suppose that you go to the market and you want to sell one good, which is not highly saleable. 134 00:09:36.900 --> 00:09:44.900 So what is reasonable to suppose you do? First you try to sell it and to get what you want. 135 00:09:44.900 --> 00:10:07.900 But as your good is not highly saleable, maybe you can't, and if you can't, you might look for a good which is more saleable than yours, so you can exchange it for it, and then once you have this highly saleable good, you can look for what you want, because it's very likely that the guy who has what you want, wants what you have. 136 00:10:07.900 --> 00:10:16.420 so it makes the double coincident of interest much easier and so if you if 137 00:10:16.420 --> 00:10:25.420 this is a plausible strategy and I think it is then we can we can imagine a chain 138 00:10:25.420 --> 00:10:35.380 reaction which is triggered because as is it clear so far yes okay so as more 139 00:10:35.380 --> 00:10:43.720 Whenever individuals perceive a good highly saleable, this good is going to be more traded, 140 00:10:43.720 --> 00:10:49.420 but this enhances the saleability of the good. 141 00:10:49.420 --> 00:10:56.100 So Menger says after a while, one or more goods are going to be universally accepted 142 00:10:56.100 --> 00:11:00.520 and they're going to be money, they're going to be a medium of exchange. 143 00:11:00.520 --> 00:11:05.640 These people are going to accept this good in exchange for their goods, despite the fact 144 00:11:05.640 --> 00:11:11.960 that they like it or not, and it is a matter of fact that money emerges spontaneously in 145 00:11:11.960 --> 00:11:13.560 many occasions. 146 00:11:13.560 --> 00:11:21.440 For example, we all know that in prison camps cigarettes have been money, and even in post-World 147 00:11:21.440 --> 00:11:26.280 War II Germany, cigarettes, chocolate and tobacco were money on the black market for 148 00:11:26.280 --> 00:11:28.240 years. 149 00:11:28.240 --> 00:11:33.400 and so this is some kind of empirical validation of Menger's insight. 150 00:11:33.400 --> 00:11:35.560 And okay, so how does simulation work? 151 00:11:35.560 --> 00:11:38.120 Very fast. 152 00:11:38.120 --> 00:11:42.380 We have a bunch of agents, they're all different. 153 00:11:42.380 --> 00:11:50.680 They bring to the market some goods and they have different preferences over them. 154 00:11:50.680 --> 00:11:52.920 And there are no utility functions. 155 00:11:52.920 --> 00:11:54.720 That is why the simulation runs so slow. 156 00:11:54.720 --> 00:12:00.080 So, there are no functions, they only have a string of preferences in their mind. 157 00:12:00.080 --> 00:12:04.480 And they try to exchange them, and we will see what happens. 158 00:12:04.480 --> 00:12:10.880 So during each round of simulations, agents pick a partner and propose a deal. 159 00:12:10.880 --> 00:12:18.920 So they will first try to sell what they value least in their stock, and they will ask for 160 00:12:18.920 --> 00:12:21.000 for the goods they want most. 161 00:12:21.960 --> 00:12:26.960 And, well, if their partner has what they want 162 00:12:27.200 --> 00:12:31.520 and wants what they have, the exchange takes place, 163 00:12:31.520 --> 00:12:35.720 no problem, but what happens if it doesn't occur? 164 00:12:35.720 --> 00:12:37.920 For example, the other agent might not have 165 00:12:37.920 --> 00:12:42.920 the good we want, then if we employ Menger's strategy, 166 00:12:43.200 --> 00:12:47.960 we will ask, instead of what we want, 167 00:12:47.960 --> 00:12:50.000 a good which is highly saleable. 168 00:12:50.000 --> 00:12:52.600 And in the simulation, I modeled that as the good 169 00:12:52.600 --> 00:12:56.480 which is most traded in agent sites. 170 00:12:56.480 --> 00:13:00.240 So if agent site is two squares, they're going to, 171 00:13:00.240 --> 00:13:04.640 the agents keep track of all the exchanges that take place 172 00:13:04.640 --> 00:13:08.520 and they're going to ask for the most traded good 173 00:13:08.520 --> 00:13:11.560 locally, okay, not on the overall population. 174 00:13:12.400 --> 00:13:15.420 And okay, so we might have a look at the simulation 175 00:13:15.420 --> 00:13:16.260 very fast. 176 00:13:16.260 --> 00:13:29.860 So, first, we might see what happens if we don't employ mangers inside. 177 00:13:29.860 --> 00:13:36.480 We just look for what we want most and try to get rid of what we don't want. 178 00:13:36.480 --> 00:13:42.860 So agents cannot trade in the whole population. 179 00:13:42.860 --> 00:13:44.620 They can just trade with their neighbors. 180 00:13:44.620 --> 00:14:07.620 Okay, so now we start, and we see now there are 121 agents I think, because otherwise we wouldn't have time to see what happens really, and we see that they can't really trade, they don't, the double coincidence of interest is not met, so they're pretty sad, huh? 181 00:14:07.620 --> 00:14:36.620 So what happens now if we trigger on what Menger suggests us, so now the color of the patches in the black square which, okay, colors indicate what agents think is the most traded goods, so sameness of color means the sameness of perceived most traded goods. 182 00:14:36.620 --> 00:14:44.620 So, pink agents all think that pink wood is the most traded and are going to accept it in exchanges. 183 00:14:44.620 --> 00:15:00.620 So, as the simulation goes on, we see we have a lot of exchanges going on and locally we see areas growing with agents thinking pretty much the same thing. 184 00:15:00.620 --> 00:15:05.620 It obviously, money is a good thing for making business, 185 00:15:06.760 --> 00:15:10.700 but we have to coordinate and agree on which good is money, 186 00:15:10.700 --> 00:15:12.560 otherwise it doesn't work. 187 00:15:12.560 --> 00:15:14.600 If we all disagree on what is money, 188 00:15:15.920 --> 00:15:17.400 then simply doesn't work. 189 00:15:17.400 --> 00:15:22.400 So agents now can only trade with their immediate neighbors 190 00:15:22.400 --> 00:15:23.920 and it wouldn't make sense for them 191 00:15:23.920 --> 00:15:27.560 to coordinate all on one good, 192 00:15:27.560 --> 00:15:31.760 simply because maybe in the right hand corner, 193 00:15:31.760 --> 00:15:33.080 there is no such good. 194 00:15:33.080 --> 00:15:38.080 So if I think that dollars are the medium of exchange, 195 00:15:38.160 --> 00:15:40.360 and there are no dollars where I live, 196 00:15:40.360 --> 00:15:43.000 and I keep asking for dollars in exchange of my goods, 197 00:15:43.000 --> 00:15:46.440 I don't make very much business, okay? 198 00:15:46.440 --> 00:15:50.800 So now, after a few ticks, well, they did pretty good. 199 00:15:52.000 --> 00:15:55.460 Almost 5,000 exchanges took place. 200 00:15:55.460 --> 00:16:12.460 Okay, now we, the other insight of Menger was about how this behavior will evolve, so social norm should outperform the basic mechanism asking what you want. 201 00:16:12.460 --> 00:16:26.980 So now, red patches will be people behaving in a direct barter way, and the blue one will 202 00:16:26.980 --> 00:16:31.300 be agents who employ Menger's strategy. 203 00:16:31.300 --> 00:16:37.260 So looking for the most traded goods and then exchanging it afterwards. 204 00:16:37.260 --> 00:16:41.420 And we see that pretty fast, they take over. 205 00:16:41.420 --> 00:16:47.260 So agents change behavior, they look around them, and if the agent next to them is doing 206 00:16:47.260 --> 00:16:54.620 better than them, they change their mind, they imitate his strategy. 207 00:16:54.620 --> 00:17:04.980 And very last thing, now we allow agents to trade not only with their immediate neighbor, 208 00:17:04.980 --> 00:17:08.780 we just extend their range of trade a little bit, and we see what happens. 209 00:17:08.780 --> 00:17:20.780 and now they're all in play in Menger's strategy and we see that very fast, we have big areas where agents use the same good as money, okay? 210 00:17:20.780 --> 00:17:36.780 And one of my hopes coming here was, I think, why should we do this kind of stuff with the agents on Austrian economics? 211 00:17:36.780 --> 00:17:41.100 I think first it's fun, and yeah, I have to be honest. 212 00:17:41.100 --> 00:17:42.060 They're fun. 213 00:17:42.060 --> 00:17:47.060 And I think they add some strength to the arguments. 214 00:17:47.500 --> 00:17:51.260 Like thought experiments typically, it is deductive, 215 00:17:51.260 --> 00:17:55.420 but we're still not sure what is the outcome 216 00:17:55.420 --> 00:17:57.220 of a thought experiment. 217 00:17:57.220 --> 00:18:01.580 And then I think they can help discussion among scholars. 218 00:18:01.580 --> 00:18:05.620 Because if this is a neutral language, we can say, 219 00:18:05.620 --> 00:18:10.020 that can be employed by mainstream economics and if it also fits with Austrian economics 220 00:18:10.020 --> 00:18:13.900 then we can really discuss and confront theories, okay? 221 00:18:13.900 --> 00:18:14.900 Thank you for your attention.