Assumptions for the simulation

[Fabio]

This post is intended to discuss the assumptions of the simulation in an open way.

It is incredibly complex to simulate human behaviour. And of course, uncountable feedback loops emerge in a complex system. Therefore, we need to make assumptions for the simulation. This means of course that the simulation will not reflect reality. Depending on the assumptions, we'll get different results.

My opinion: The goal should not be to approach as close as possible reality, but to get a QUALITATIVE feeling on the dynamics of the system.

Assumptions so far:
- 3 million accounts (a sim could use much less, which makes it run faster; of course possible feedback loops are then reduced)
- 1000Q initial float, 3bnQ total inital quid balance
- Adjustments of 5% of account balance per "week" on crossing of upper resp. lower velocity theshold (this can be parameterised so that we can play with different percentages; maybe it should be adjustable "over time" too)

The more difficult part is the simulation of human behaviour. After every step, a certain amount of accounts would trade. How many, and by how much? Will they be the same on every step (e.g. will the "growing" accounts always grow, etc.)? Of course, it boils down to align accuracy with simplicity. The more rules, the more complex the sim, the more time and effort (programming work) needs to go into the sim.

We discussed simulating resource flows too. Maybe it is important to include them in the sim, as money is linked to real resources. A first "rough-and-ready" simulation code I got from Joerg included bread, wheat as well as time ("I can offer work") elements. What are peoples thoughts on this?

[Caroline Whyte]

If I understand correctly, quid are designed to be used in a manner similar to stamp scrip; there's a built-in incentive for people to keep them in circulation as much as possible, rather than hoarding them. So could we examine historical examples of stamp scrip to get an idea of the possible velocity and volume of quid transactions?

[Steve Allin]

Hi there
Reading the latset posts I wouild like to add a few thoughts . The difference between Banks and Credit Unions was that people tended to trust the banks as they were bigger but with what is unravelling now that may no longer be the case.
Some time ago Richard suggested that local authorities ie County councils may back the system, now they may not have the electronic systems in place to manage the transaction at present but if the software team can get that up and running and simple to use, then they may be trusted more than others such as the banks etc.
My other thought is that one of the problems with this system is the large amount of money/value that is spent outside the areas we are imagining using this system for instance in buying goods ; food /materials etc that in the main this is going to be initially at least used mainly for payment of services/labour and the purchase of locally produced food if there is enough of it. I like the idea that local authorities might have a part to play as the quids could be used to pay rates , Road tax,and as maybe part
payment of their wages to spend into the local economies.
Steve Allin

[rdouthwaite]

We can explore Steve's suggestions for non-bank ways to launch the system - such as local authorities or credit unions - if the bank road proves to be blocked. (Don't forget that we don't have to work with the Irish banks - a foreign bank or an organisation like PayPal could offer quid accounts to Irish customers). So let's work, for the moment, on the assumption that the banks are going to agree to open quid accounts for their customers in addition to their euro ones.

This should be built into the simulation. People will tend to prefer to spend quid and hang on to their euros, reserving them for servicing their debts, spending abroad, and buying goods whose suppliers insist on getting part payment, or 100% payment, in euros. So it may not be enough to model quids alone. The interaction between the two currencies will be important.

On the other hand, I don't think we need to look at resource constraints at this stage. The Irish and the world economies are depressed and the introduction of the quid will not push demand to anywhere near the limits of supply. Moreover, the restoration of the equivalent of three billion euro to the Irish money supply will only compensate for around a quarter of the fall since the supply peaked.

[jhb]

This post is intended to discuss the assumptions of the simulation in an open way.

include them in the sim, as money is linked to real resources. A first "rough-and-ready" simulation code I got from Joerg included bread, wheat as well as time ("I can offer work") elements. What are peoples thoughts on this?

That was just code to see if my approach for the structure would work. In the end one needs something (so was my thinking) to get money flowing around. As far as I understand there is a money flow, and a counter flow of good, services and property rights. One might not make too much sense without the other.

Not sure though if this is needed at all.

[Fabio]

This is Piotr Magnuszewski's contribution
for the setting up of a simulation.
-------------------------------------------------
Dear Fabio,

I am following on my previous message.
As I said I recommend an agent-based approach (I know it is not too precise)
Below I give you links to some very popular agent-based software. They are more like libraries and you need to program on top of them

Swarm:
http://www.swarm.org/index.php/Main_Page

RePast:
http://repast.sourceforge.net/

I can look further and find some agent-based model of artificial economies. But as you are aware there is a need to be cautious with complexity.
The simulation of this kind could easily be a subject of a 3 year research project.
So the design of the simplified environment (virtual world) will be critical here - it should be simple enough to be doable in relatively short time but we cannot lose a link with a real world. Conslusions from a virtual world need to be applied to real situation.

I am ready to discuss it further, I would need to know how do you plan the process. I can talk on Skype (*****).

Best,
Piotr

[Fabio]

I think using such a framework like swarm could
be very beneficial. We'd still have to develop the
assumptions (the "schedule", the "swarms" and the
"agents" in swarm terms), but it would provide us
a solid framework in which to "plug in".

However, it imposes restrictions (like java or object C
as prog lang).

what do others think?

[Graham Barnes]

Fabio

More learning curve! It seems to me that agent based modelling is exactly what we need (=autonomous agents interacting in a shared environment with an observed aggregate, emergent outcome [Wikipedia definition]).

But which tool to use I think must depend on (a) ease of start-up plus especially how accessible is the definition of the individual agents behaviours, so that the outside world (represented by Group3) can see, understand and suggest changes to the documented assumptions and (b) the technology match of the tool to our skills (Java, whatever) - which is something you Joerg and Trea would have views on.

Whether application into economics might give us a leg-up is difficult (for me) to assess e.g:
http://en.wikipedia.org/wiki/Agent-Base ... _Economics

I suspect it might lead into over-complex implementations when as Piotr points out we dont have three years to play with.

[Fabio]

We need to make progress, or the simulation won't have any value.
I apologise if what follows looks dilettant and/or simplistic / whatever to someone, but I'd like to ignite action. Ideally, this should be modeled by economists.

Taking an agent approach, we'd have, to make it very simple,
the following agents (good granularity? can be refined, e.g. high-income individuals, small-mid-big-biz, local-regional-national govt):
- Individuals
- Businesses
- Government
- The Trust

Assumption #A1: If people get 1000Q for free, they'd like to spend them all, as they loose value.

Leads to problem #P1: To spend them, businesses need to accept them.
Problem #P2: Government needs to accept payments(taxes?).

Question: Are there other inhibitors for businesses to accept them other than trust?

Assumption #A2: If businesses accept them, they are likely to want to spend them all too, or they loose value.
Assumption #A3: If govt. accepts them, it/they are likely....

Both A2 and A3 go back to problem #P1

Variable #V1:Acceptance. So the first variable of the model would be the acceptance of the quid. This could be modeled on the basis of the transactions. A transaction is initiated by an individual. A transaction either succeeds or it fails. So acceptance could be modeled by making the percentage of succeding transactions variable. This variable could also change over time (ie. low acceptance in the beginning, grows over time).

Initially, we have only individuals, later also businesses and govt. initiate transactions.

Assumption #A4: Every transaction "adds" participating agents (biz, govt.) to a pool. In the beginning, high % of new participating agents are newly "created businesses, later, higher % of transactions happen with "established" businesses.
Variable #V2: % of newly created businesses on transactions, 60% let's say the first 1000 (?) transactions create 60% (?) new biz members, decreases over time (how much?)

Question: What are individuals wanting to buy? -->
Assumption #A4: Transactions are likely to happen on small amounts first (e.g. most between 10Q and 100Q). Every transaction gets assigned a random value between 10 and 100Q.
Variable #V3: Lower and upper value of range per transaction

Assumption #A5: a time unit is a day.
Question: How many transactions take place per time unit? -->
Assumption #A6: Randomly generated per day and week
Variable #V4: Lower and Upper range of transactions per day. E.g. lower 0, upper 10 (?)

Assumption #A7: The system checks velocity every week
Variable #V5: Upper velocity limit where quids get added (suggestion?)
Variable #V6: Lower velocity limit where quids get removed (suggestion?)
Variable #V7: Amount of quids changed (Assumption #A8: 5%)

Variable #V8: Initial Quids amount (Assumption #A9: 1000Q)

Assumption #10: Every transaction pays a fee: begins at 0.0125% (don't hang me for this, just a number :-)
Variable #V9: Transaction fee

Question: How are transactions taxed????
Assumption #11 (probably wrong): Every receiver of a transaction pays a tax

This is just a start, no idea of the validity of all this. Also, I have no clue how far we need to go in modeling things.