Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

WHY CAN'T WE USE THE SAME CURE FOR ALL CRISES?

-AN AGENT BASED MODELLING APPROACHCIUTACU Ileana Ph.D. Student, Economics Department, Bucharest Academy of Economic Studies, Bucharest, Romania, ileana.ciutacu@gmail.com

VNTU Denis Ph.D. Student, Economics Department, Bucharest Academy of Economic Studies, Bucharest, Romania, d.vintu@reading.ac.uk SVULESCU Iulian Ph.D Student, Accounting Department, Bucharest Academy of Economic Studies, Bucharest, Romania, savulescuiulian@yahoo.com

Albert Einstein once said everything is relative and he was right . The results of this paper prove him right even when talking about economics, because in real life when we find ourselves in front of a crisis situation we try to solve or cure it using 'old remedies' or what we know, but what we never seem to learn is that using the same inputs will not give us every time the same results. To reach this conclusion we designed and created a simple economy agent based model that was after that developed using NetLogo. Key words: economy, crisis, netlogo, agent based models, JEL classification: A10, C60, E20, J00, P40.

1. Introduction Albert Einstein once said everything is relative and he was right, not only in what regards Physics, time and space, but also in what regards most Social Sciences, including Economics. In this last few decades, and especially in the last five-six years due to the subprime crisis in the United States of America that lead to the current mondial crisis, Economics and economies have changed considerably, leaving economic policies decidents in great doubt. According to Farmer and Foley (2009), these problems occur because most governments don't use sophisticated quantitative computer models to guide the country's economy out of crisis. Instead, they use two main cathegories of models, the first type are econometric: empirical statistical models that forecast what will happen in the future with data from the past, assuming that things will stay more or less the same. And the second type of models are based on the dynamic stochastic general equilibrium and imply a perfect world that by their very nature rule out economic crisis. Unfortunately, both have many flaws, mainly because there is no or little attempt to assemble the pieces of an economic system and understand it's behavior as a whole. As a solution to this problem, Farmer and Foley present a better way: agent based modeling. This is a computerized simulation of different types of agents (from consumers to policy-makers and countries) which interact using prescribed rules in a 'world', an artificial economy, that can have a structure as diverse and complete as is needed. Instead of expecting the movement of the economy towards a predetermined equilibrium state, in the case of agent based models, agents behave all their life according to given rules, what they have learned from past experiences and the situation they find themselves in that moment. Thanks to the computer, this type of models can keep track ot all the agent interactions, so the newly created model of artificial economy can be seen clearly, at any given time, from the smallest detail to it as a whole complex system. Also, agent based simulations are able to handle nonlinear behaviour in a wider range that conventional equilibrium models can, thus allowing policymakers to simulate and explore the quantitative consequences of different policy scenarios upon an artificial economy.

Agent based models are already used with succes by policy-makers in some areas of science as are epidemiology, traffic control or demographics. There also exist some succesfull models that simulate small portions of the economy, as are the financial market model of Blake LeBaron, the firm dynamics model of Rob Axtell, the credit sector model of Mauro Gallegati's group, presented by Farmer and Foley (2009) or the EURACE project that models the European economy (Langnick 2011). Another agent-based model that simulates very well the economy the one built by Langnick and presented in one of his papers ( Langnick, 2011). This model is a baseline macroeconomic one, that even if simple, can reproduce a lot of the stylized facts of business cycles and can be an adequate response to recent criticism of macroeconomic methodology. This is because, according to Langnick (2011), his model does not depend on the strict assumption of rationality, allows aggregate behavior and absolutely consistent micro foundations and does not depend on equilibrium assumptions that might rule out by definition coordonation failures, instability and crisis. His model is a stylized one that abstracts from the real economy because it contains only a small number of different agent types and interaction rules and can be run on an ordinary desktop PC. Being built upon the model developed by Gaffeo et all. (2008), Langnick's model has only two types of agents: households and firms which use simple behavioral rules in order to choose prices and wages; the economy is allowed to self-organize toward a spontaneous order and a part of the agents are allowed to buy only from a subset of agents. But, instead of analyzing growth as a result of investment in Research and Development, using another type of agents (banks) and indexing time by quarters, Langnick's model measures time in days and months and aims to define a model that is concerned with basic macroeconomic relations in a non-growth environment. Agent-based models can be designed and created by using special computer programs. One of these programs is the one built in 1999 by Wilensky and developed over time at the Center for Connected Learning and Computer-Based Modeling of the Northwestern University, named NetLogo. This computer program belongs to the next generation of multi-agent modeling languages like StarLogo or StarLogoT, and is run as a standalone application on the Java virtual machine of computers that use Mackintosh, Windows, Linux, etc. platforms. When using NetLogo, the creator of the agent based model uses a special programing language that allows him to define an environment for simulating natural and social phenomena by using four type of entities: turtles (mobile agents), patches (static agents), links and the observer. These agents can be part of populations that can number up to thousands, operate independently in an artificial 2D or 3D world using predefined rules and instructions. This caractheristic of NetLogo makes it possible to explore the connection between the micro-level behavior of individuals and the macro-level patterns that emerge from their interactions. 2. The Model By creating a simple economy agent-based model, this paper aims to see the causes for its entering a crisis. Our simple economy model was build based on the NetLogo Wolf Sheep Predation Model (Wilensky, 1997). In this model, Wilensky created a world in which live three types of agents: one type of patches: the patches of grass, and two of turtles: the sheep and the wolf, therfore two prey-predator systems that interact one with the other. The sheep eat the green grass when they encounter it and gain energy, reproduce according to a predetermined probability and are eaten by wolves. The wolves eat the sheep when they encounter them and gain energy, and also reproduce according to a predetermined probability. The grass is eaten by the sheep and grows back according to the time of regrowth set by the model's creator or user. Instead of using two types of turtles and one of patches, the current model has three types or breeds of turtles: firms, men and apples, that live on the surface of a sphere-like world. Their position (firms and men) in the world is randomly chosen by the computer every time the model's used presses the 'setup' button.
1. Programming code for the populations breed [men man] breed [firms firm] breed [apples apple]

The world is a 61*61 units torus, which means that this world does not have borders. Our world's time is measured in ticks by a tick counter (see Figure 1.) and lasts as long as men live in it. When there are no more men, the tick counter stops showing the length in time of that world. The firms and men appear first on the surface of the world in populations of 0 to 100 firms and 0 to 500 men. The currency of this world is energy and can be measured by rational numbers.
Figure 1. The world in the NetLogo interface before pushing the setup button

2. Programming code for the economy's 'currency' turtles-own [energy]

The firms are born with an initial energy of 250 units, receive 20 units of energy when hiring a man and get their energy cut in half if the odds, determined by a probability and a rule, tell them that they can produce apples.
3. Programming code for the production of goods (apples) to produce-goods ask firms [if random-float 10 > firms-produce and energy >= 75 [ set energy (energy / 2) hatch-apples firms-no-of-products [ ] end

The apples are linked to the firms, because, even if turtles, apples can not reproduce themselves, but are produced by firms and can live in the world only for 0 to 10 ticks (see the apples-perisability slider in Figure 1.). As in the case of the firms, the apples' position is fixed during the running of a simulation of our simple economy model (or world's time) and is randomly chosen by the computer when the firms that produce them are allowed by the odds and have enough energy for producing apples.
4. Programming code for the setup button to setup clear-all setup-world end

Even if NetLogo turtles have the ablity to move in the world, in this simple economy only the men can change their position during a simulation. By moving around, one step at a tick, men receive energy by getting a one-time tick job, when encountering a firm, or by consuming an apple, when encountering one. The job cand bring to a man a fix wage of 15 energy units, while consuming an apple can bring the man up to 20 energy units, depending on what the user chooses the men-gain-from-food slider to be before pushing the setup button (Figure 1.). Of course, men can also reproduce, and, just like in the case of firms producing apples, this happens only if they have enough energy and if the odds allow them.
5. Programing code for the men's reproduction process to reproduce ;; if a man has more than 50 energy and the right probability, it can reproduce ask men [ if random-float 10 > men-reproduce and energy > 50 [ set energy (energy / 2) end

Men die or get out of our simple economy model's market if they have no more energy left. In that moment, the go button, that allows us to run the simulations, stops.
6. Programing code for the men's 'death' of leaving the simple economy's market to death ask men [ if energy < 0 [die] ] end 7. Programming code for the go button to go if not any? men [stop] move work . . . tick end

In order to see if our simple economy model can get into a crisis, we will set 3 scenarios in which only some variables will change. These variables can be easily changed by modifying the sliders on the interface screen of the NetLogo application (see Figure 1.). We will run each sceario 3 times, until the go button stops, to see what happens and why this happens: a) First scenario: initial-number-men: 150 initial-number-firms: 10 firms-no-of-products: 4 men-reproduce: 4 firms-produce: 4 apples-perisability: 5 men-gain-from-food: 15 b) Second scenario: initial-number-men: 250 initial-number-firms: 20 firms-no-of-products: 10 men-reproduce: 5

firms-produce: 5 apples-perisability: 3 men-gain-from-food: 15 c) Third scenario: initial-number-men: 450 initial-number-firms: 40 firms-no-of-products: 20 men-reproduce: 7 firms-produce: 2 apples-perisability: 2 men-gain-from-food: 10
Figure 2. First scenario results

Figure 3. Second scenario results

Figure 4. Third scenario results

3. Conclusions As it can be seen in Figure 2, the three runs of our simple economy model's first scenario, even if have the same inputs, give different results of time and population numbers. Thus, the first run lasted 54 ticks or 58.8 secconds and had a maximum of 183 men and about 80 apples. The second run lasted 64 ticks or 73.8 seconds and had a maximum of 184 men and about 70 apples. And the third run lasted for 48 ticks or 58.8 seconds and had a maximum of 164 men and about 55 apples. The number of firms remained constant during all these runs, because the firms don't have a mechanism of getting out of the market. In this scenario, both men and firms have a big probabilityof reproduce and to produce, to be more precise 60%. When allowed to produce, the firms produce 4 apples that, if are not encountered by a man, have a lifetime of 5 ticks. In the case of this scenario, even if the men and firms had good chances to reproduce and to produce, due to the fact that men did not know where the apples or firms were in order to go towards them when they did not have anymore energy there were only a few men that ate apples and got jobs. This sent the world directly into a crisis that ment that the men died, when they were left out of energy left, that only a small number of men were able to reproduce and that many of the apples were disused, because they simply dissapeared from the face of the world before being eaten. The same thing happened for the second scenario too. As it can be seen in Figure 3, the three runs gave different results of time and population numbers even if they had the simillar inputs. Thus, the first run lasted 63 ticks or 73.8 secconds and had a maximum of 308 men and about 287 apples. The second run lasted 49 ticks or 58.8 seconds and had a maximum of 308 men and about 295 apples. And the third run lasted for 49 ticks or 58.8 seconds and had a maximum of 317 men and about 270 apples. The number of firms remained as well constant, but what is interesting is that during the first and second run, in the moment of the forth or fifth tick, the number of apples was bigger than the one of the men. In this scenario both the men and firms had good chances to reproduce and to produce (50%), the number of products was a high one (10 apples) with a rather small lifetime (3 ticks). Compared to the first scenario, the second one had a bigger number of men that reproduced themselves (aproximative 50 men), but because of the same problems men do not know where the firms and apples are in order to get more energy when they don't have anymore left the world entered a crisis and the number of men soon decreased until there was no man left and the world stoped.

As it can be seen in Figure 4, a similar thing happened with the third scenario too. Only that here, another interesting thing happen: for all the three runs, the number of apples almost tripled compared to the number of men for exactly 3 ticks. Thus, the first run lasted 46.3 seconds od 45 ticks and had a maximum of 575 men an 1420 apples. The second run lasted 92.5 seconds or 80 ticks and had a maximum of 1440 apples and 504 men. And the third run lasted for 58.8 seconds or 55 ticks and had a maximum of 1270 apples and 500 men. In the case of the third scenario, the number of men, firms and apples is the highest from all the three scenarios, but the chances a man has to repoduce are of only 30% and the one of a firm to reproduce are very high (80%) and the lifetime of an apple is quite small (only 2 ticks). This lead to a crisis of overproduction of apples and only to a small number of men. The men were able to reproduce themselves (approximatively 50 to 125 men), but soon died because they did not find a job or eaten apple in order to receive energy. The simmilar results of the three scenarios demonstrate that our simple economy model can be improved by adding a way by which the firms can be paied with energy when an apple is eaten and another one by which firms can get out of the market. The different results in population numbers of the three scenarios of our simple economy model, remind us that in fact everything is relative and that even if we give the same or similar inputs we can obtain very different results in time. The only thing we can do is adapt our solutions very well to the problem or crisis we encounter.

This article was written as part of the project Doctorat n economie la standardele Europei Cunoaterii (DoEsEc) POSDRU/88/1.5./S/55287. Project co-financed by The European Social fund through POS-DRU 2007-2013 and implemented by the Academy of Economic Studies in partnership with the Western University of Timioara. 4 References Ciutacu, Ileana, Dospinescu, Andrei (2011), The cost of exiting a system with connected benefits , article presented within The Sixth International Conference on Economic Cybernetic Analysis: Global Crisis Effects and the Patterns of Economic Recovery GCER 2011, 20-21 May, Bucharest, Romania Farmer, J., Doyne, Foley, Duncan (2009) The economy needs agent-based modelling , Nature, 406, p. 685-686. Gaffeo, E., Gatti, D. D., Desiderio, S. & Gallegati, M. (2008), Adaptive Microfoundations for Emergent Macroeconomics, Eastern Economic Journal , 34(4), p. 441-463. Lengnick, Mattias (2011) Agent-based macroeconomics - a baseline model, Christian-AlbrechtsUniversitt zu Kiel Economics Working Paper , 04, p. 1-28. Wilensky, U. (1997), NetLogo Wolf Sheep Predation model. http://ccl.northwestern.edu/netlogo/models/WolfSheepPredation. Center for Connected Learning and Computer-Based Modeling. Northwestern University, Evanson, IL. Wilensky, U. (1999). NetLogo. http://ccl.northwestern.edu/netlogo/. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

You might also like