Economic Model II

[Mark_Everson]

It's time to start a new Economy thread going, since the old one is getting very close to the 150 post limit. Here is a link to the previous thread Economic Development Model - Opinions Please?. The economy model page on the Web site is pretty much up to date, so I'm not going to write a summary here.

There's just one thing I want to ask everyone's opinions on at this point. I have parts of the Economy model up and running, and hopefully you will see it as part of Demo 5 within a week or two. I have what I think is a flexible and powerful way for the player to handle the economy, and I wanted to run the general idea by people to see if anyone barfs . The main thing I'm worried about is that it has Numbers in it. Essentially there are two factors to the interface, how much you want to invest in a given area when you view it as a good opportunity, and what you define as a good opportunity.

So here is the deal... say you are an emperor that wants to develop the food-producing capacity of your civilization. In Civ2 you would have to push around all these annoying settlers by hand to irrigate lots of land. Apparently the public works concept in CTP works moderately well, although I don't know personally because I never played it. Here is the idea I have right now for Clash:

First I'll give the numerical aspects of the idea, and then we can decide at a later date if we need to work it through an adviser without numbers, or at least give the player the opportunity to do it without numbers. So if you want to push food production, but only in areas where will really make a difference, you call up your Econ command screen at the civ level. This will show all your orders with respect to the civ-wide economy. For an economic area you want to invest in, you'll get the option to order how much money you want dedicated to the area per turn, and at what point you would no longer consider it worth investing that money. Specifically, the numbers you will see in the Demo 5 Econ screen are "percent of local tax revenues to use" and "rate of return". Let's say for the farming sector you have set 10% of local taxes as the investment amount with a rate of return needed to invest of 30%. What the game will do when it gets to the economic orders phase of the turn, is examine your order for every square in your civ, and decide based on your return rate whether to spend up to the amount you allocated (10%).

So let's say in one particular square the taxes you get are 100C. The Econ model will do a simple rate of return calculation, and find out how much you should invest, up to the 10%, while still meeting your needed rate of return of 30%. What the rate of return calculation considers is how much the investment will pay back directly to you the government in the future, and what annual rate of return this is equivalent to. (Right now the rate of return doesn't consider what the people might get out of an investment, only what the govt gets back in the way of increased taxes.) Now 30% is a pretty high rate of return, so your selecting this means that you really only want to invest in farming in areas where you will get pretty rapid payback. So in most places the 10% of local taxes investment that you have called for won't actually be spent. Now if this square has really lush farmland it may be that you will spend the 10% (= 10C) on building up farms. If it isn't a really good deal, nothing will be spent, and you will get the money sent to the treasury instead. If you had wanted to invest more broadly you might pick a rate of return of 5-10% as required.

So why do I think this is so cool? You can switch on or off investments over your whole civilization with just a few orders. By adjusting the rate of return you can either invest any where it will do the least bit of good (1% rate of return), or invest in places it's fairly advantageous to do so (10% or so) or only cherry-pick the very best places to invest (30%-100%). You can also override or augment the civ-wide settings at either the provincial, or individual map square level. It just depends on how interested you are in tweaking the economy! If what you really want is money, just set all the local tax percentages to 0, then you will just get your taxes sent directly to the civ treasury.
When switching from peacetime to wartime I envision we would give the player an order they can give to basically kill all investment orders. Then, in the same way that you put 10% of tax assets into building farms, you could put 80% of income toward building, and/or mobilizing military units. The only problem here is that I don't yet have a clever way to gauge rate of return for military units or several other things. But I figured we should first see if people like the way the system works for things like farming and gold mining, and if it's a winner I'm sure we can come up with some way to do it.

So at least for now you'll get a couple numbers you can set for each area of investment, that will give you widely different responses over the whole civ. The numbers don't bother me personally, and if the idea works in general I think we can come up with words for orders you give to your adviser that are basically code for several different levels. In that way the people that don't want numbers in their game wouldn't have to have them.

What you think?

[Richard Bruns]

I like the model; it should give the player plenty of control without too much effort, and it allows fine control if the player wants that. It looks like it would be a good interface.

Now onto the details:

What provision is there for redistribution of income; giving poor provinces money from the treasury if investment would be worthwhile? Suppose that a newly colonized tile has really great land, but since their economy is so small, there isn't nearly enough tax revenue to develop the land well. Is this good investment ignored until the square can pay for it itself?

Also, how is the rate of return calculated? How do you calculate the rate of return for investments like better sewer systems or schools?

Here is something I posted in the ecology thread; I think it might apply to this situation as well, since it could be a method for calculating the benefit from intangibles:

-----

The cost of transforming a province or square would have to be defined in the ecology model. There are two costs for clearing: The actual cost of the operation and the opportunity costs of the lost ecosystem. The operation cost is easy to define: The vegetation characteristics could include such a number. The opportunity costs, however, are more interesting and complicated.

Some oppportunity costs can be easily defined. If you destroy a forest you lose the ability to get wood from that square or province. The loss of one type of economic activity is the tangible opportunity cost of altering the terrain. However, there are also non-economic, intangible opportunity costs of destroying natural environments.

We have only recently learned the value of certain natural ecosystems. Wetlands provide a lot of benefits such as water purification and flood control that people didn't know about a hundred years ago. Thus, the percieved opportunity cost of destroying wetlands has risen. This is why the government no longer supports wetland drainage like it used to.

So, in Clash terms, the tech level determines the percieved intangible opportunity cost (PIOC) of destroying the natural terrain. A civ with a low tech level would see zero opportunity cost for draining a big smelly swamp, while a more advanced civ would see a very big opportunity cost in destroying a wetland ecosystem. And to further complicate things, the culture and social conditions of the civ will also impact this percieved opportunity cost.

Regardless of what the civ knows about, there will be a real opportunity cost to terrain alterations, in the form of flooding, erosion, loss of groundwater, etc. These changes will affect the civ over time, as the ecology model calculates the consequences of human actions.

It is impossible to predict exactly what will happen; the ecology model is a chaotic dynamical system with a lot of randomness. We will have to define an average real intangible opportunity cost for destroying natural ecosystem based on what will most likely happen. This will mean putting a cash value on things that are not economic, but we should be able to do it.

To further complicate things, the real opportunity cost of destroying an ecosystem will change based on how many tiles of that ecosystem are on the map. The marginal opportunity cost of losing a terrain type will increase as that terrain becomes less common.

So the PIOC is based on technology, society, and the average marginal real intangible oportunity cost (AMRIOC) of altering the terrain. The exact calculations for AMRIOC and PIOC are a problem best dealt with later.

The cost of the terrain alteration that is entered into the economic calculations is the operation cost plus the tangible opportunity cost plus the PIOC.

The benefits of terrain alteration are also divided into tangible and intangible benefits. The main tangible benefit is an increased number of farm sites. This can easily defined, and the economy model can put a cash value on the farm sites.

There can also be some intangible benefits to altering the terrain. Swamps do breed mosquitoes, and clearing them can reduce disease. These intangible benefits would be calculated in much the same way as the intangible costs of terrain alteration. A civ that did not know how malaria spreads would see little intangible benefit in draining swamps, while the perciened intangible benefit (PIB) of swamp draining would be much higher for a civ that knew about the role of mosquitoes in spreading disease. In some cases, the PIB alone could cause people to alter the terrain.

It may seem like these PIOC and PIB calculations would take a lot of time, but they will not be run for every square. There will be a few dozen terrain types, and all of the intangibles will be the same for every tile of that terrain. Unlike the tangible costs and benefits, they do not depend on the Land and Water values of the terrain. These PIOCs and PIBs only have to be calculated once for the entire civilization and then stored as variables to be accessed when the economic cost/benefit calculations are run. If time is still a problem, they do not have to be recalculated every turn. Some fraction of the intangibles calculations could be run every turn, so the variables for each terrain could be updated every four turns rather than every turn.

The concepts of PIOC and PIB can be used for things other than terrain alterations. Power plants, factories, mills, roads, and farms all have non-economic costs and benefits, as do fertilization and irrigation of farming terrain. I think that some version of this system would make Clash a lot more realistic, dynamic, and fun.

-----

Now, I think that this system would work well with the new economy model, after a little tweaking. For example, sewers have many intangible (non-monetary) benefits, such as disease prevention, ecological protection, and quality of life improvement. It isn't the same in each case and, given the randomness of the disease model, we can't predict the efforts exactly, so we have to calculate averages and assign them monetary values like we do for the enviromnental AMRIOC values. Also, the first sewer system in a tile will be more valuable than the following ones, so the marginal benefit of more investment in sewers changes based on how many sewers there are. So the average marginal real intangible benefit, or AMRIB, of an improvement would serve as the basis of calculations.

The PIB of sewers would change based on the AMRIB and tech advances like germ theory and ecological awareness. Note that unlike PIOC, PIB is a continuous thing; benefits from infrastructure come in every turn. This monetarized value is an indication of how much people think they would benefit from the sewers. So early on, the PIB would be low, so your people would only invest in the simplest system, a pipe leading into the lake. The AMRIB is unchanged, however, so when tech advances, the PIB would increase as people become more aware of things. Now they would want to invest in more and better sewers, since the percieved benefit is now greater than the cost.

To fit this into the economy model, all we have to is turn PIB into a monetary value. This value then serves as the "Return on investment" that the economic model demands in its calculations. So assume the player allocates up to 20% of tax revenue to infrastructure, with a cutoff of a 20% return on investment. If the PIB for more sewer investment is greater than a fifth of the cost of building that sewer and the tile can build it with that 20% of taxes, it will get built.

In the case of multiple needs, the one with the highest PIB to cost ratio would get built first. Suppose the above tile generated 100C in taxes and the models calculated the following PIBs and costs for various improvements:

Sewers: 10 C a unit, PIB per unit: 25, 12, 6, 2
Water Supplies: 5 C a unit, PIB per unit: 20, 10, 5, 1

The province would build one sewer unit and 2 water units, even though the model says it is worth it to build all 4 sewer units and all 4 water units, since they would still return 20% of the investment value each turn. So the province wants to spend 60C to invest in things that would return up to 20% of the benefit.

Which brings us back to the question of redistribution. This square obviously has serious needs that they can't support themselves. Perhaps we could add in a third value; treasury support. The player could say that if the return on investment exceeds a certain value, like 40%, then the governnemt would pay for that investment. So then, the civ's treasury would pay for the second and third sewer units and the third water supply unit.

Questions? Comments? Cuss Words?
[This message has been edited by Richard Bruns (edited February 01, 2001).]

[Mark_Everson]

Hi Richard, I'm glad you like the general idea. Thanks for the extensive comments. I just have a few moments, so I'll be brief.

Yes, there definitely should be some sort of treasury support available for fast-starting provinces. But in line with our new streamlined development philosophy I want to see what people think of the general ideas and practicality behind the proposal before embellishing it. There will also be straight lump-sum purchases of things that don't depend at all on local tax revenue.

I agree in principle with pretty much all of your comments. We'll see how far we can go with them! One wrinkle in what you said is if the player knows sewers pay off, even if the return rate formula does not give the proper benefit, it just gives the player the incentive to give orders that are skewed by that knowledge. So I'm not quite happy with that part of your argument. It may be simpler just to make percieved benefits always match the reality. But we've a Long way to go before we need to make that decision...

[LDiCesare]

I rather like the return rate.
I think the player will want it. For instance, when you build a bank in civ, you oft wait till you know it will bring you more money than it costs.
It is difficult to show everything in terms of cash however. The public works in CtP work as part of production (not cash) spent on building stuff. To me building a mine means I have a return in terms of production, if you can translate that into cash it is fine. It could be interesting to define several RoI figures, like roi in terms of cash or of science or pop growth or production. That could allow to have an econ policy based on non-economic priorities (which may mean bankrupt policies if you never get youe cash back by the way).

I also think there are times when things are not percentile but quantic: e.g. Build 30% of military units until you have at least one army. In civ1 I also always built/bought a granary in every city no matter what. Additional improvements may be percentile based, but I would like to have minimum buildings or at least units specifiable. Note that with the building-less approach, this may be relevant only for armies.

[F_Smith]

More thinking outloud . . .

Similar to the combat orders, I wonder if the player should be limited to 'strategic' econ orders, and no 'tactical' ones? In other words, 'provinces' should be given 'missions', with no specifics on how to carry that out.

Orders like:

[*]Maximize tax revenue -- safe, moderate or agressive. (Relative to the level of public unhappiness to tolerate).[*]Maximize growth in industry x. Or x and y.[*]Maximize public happiness.[*]Maximize military preperations.[/list]

An emperor wouldn't be involved in telling specific provinces how much to spend on sewers, would he? He'd just say 'improve your sewer system immediately'. The province would then do the rest themselves.

The tactical/micromanagement of provinces can still be a game option, without a doubt. But I don't know if if fits with the 'default' game we're considering.

Am I daft?

Well, of course I am, but . . .

[Mark_Everson]

Hi All, thanks for all the good comments and criticism.

Rüdiger, I had actually meant that the player should always know how much money their orders are costing, but just forgot to put it in the writeup . Your approach certainly handles in a graceful way the problems Richard brought up, that sometimes you want to rapidly accelerate the growth of small economies. It probably is best not to use the local investment idea I originally had too rigorously, since the player is generally interested in the best overall strategy.

I think a combination of our two approaches would probably work best. The problem with yours as stated IMO is that without a rate of return number shown to him the player doesn't know if his money is being spent wisely or stupidly, only that a given amount is being spent the best way possible. What I would propose is that the player should get two different places to add detailed information. Specifically, what is the rate of return, and what is the desired amount of money to spend (or perhaps fraction of income). The player can enter what they want in either one of these places, and then see what the result is in the other place. Whichever way works best for the player could then be used. However, with this proposal it's not clear to me now exactly how I would work micromanagement at a lower level such as the provincial or MapSquare level. I guess if we use amount of money or percentage of income it would work okay, but percentage of civ income could be extremely small when you get down to the MapSquare level. Anyway, I'm sure we can figure out some way to do this.

Laurent, glad you like it! I agree with all your points, but I think we need to see if the base system works before we worry about too many details. If people generally like it in demo 5, then we will certainly have opportunities to give cash equivalents to various things like population growth etc. And certainly sometimes people want to limit quantities, but I think if we set the cash values right for non-cash things that problem can be solved also.

F. Smith, I think a lot of players will want what you describe, and I personally am hoping to be able to run things just at the civ level. But I believe that without the detailed tools that we are talking about here it will be almost impossible to give the player what you want. It's easy in these games to say "maximize x", but it is very rarely that simple. This kind of game has complex interactions between myriad areas, and it's only by adding a lot of human judgment that I think we can determine what "maximize taxes" really entails. Once a lot of clever people have played it, will have a really good idea what settings to tell the AI to use when the player says to maximize taxes. So I think you should look at what we're talking about as a necessary toolbox for providing the kind of hands-off economic handling that you're looking for. But personally, I would rather set some commands at the civ level rather than hand tailor a bunch of provinces with different orders. Different strokes for different folks...

[Rüdiger]

Hi!
I think the model is a bit too complicated for the player, especially the rate of return setting. The player has really no clue how much money will be invested at a given setting of rate of return. In my opinion, the system should look a little different:
The player sets the percentage of taxes used to improve infrastructure. The system will then determine the amount of taxes available in the whole civ, and will take the desired percentage thereof to improve the specified economic area, eg. farming. Where the money is spend is determined by the rate of return.
I think this system has several advantages:
- The player does always know how much money (or what percentage of taxes) is spend.
- He can be certain, that the investments are used in a profitable way.
- The player is presented a simpler interface, he has to control only one variable, one not so difficult to use reasonably as "rate of return".
This system is using the taxes generated by the civ as a whole, instead of per province or per map tile, what might not be what you intended. But I think this is better, because it doesn't add to the differences between more and less developed provinces. And isn't it, in theory, the way modern goverments distribute subsidies? Which is, what this system is all about, or not?
Besides, the ideas, how to calculate the rate of return, sound quite good to me.
[This message has been edited by Rüdiger (edited February 02, 2001).]

[LDiCesare]

Thinking aloud too...
"Maximizing" is a very difficult task, because there are so many parameters. But then, Mark's idea about AI was to use genetic algorithms and GAs are very good at optimizing things. So maybe in some far-fetched future you could use the GAs to determine which production seems best.

[Rüdiger]

I really like the idea, that the player can adjust either "rate of return" or "money spent", and is presented an estimate of the other one.

I will now talk about province level, but all this can be applied to map square level, too.

Maybe we should introduce another setting for the provinces: A setting for how much of the total taxes of the civ should be spent in that particular province. Per default this would represent the amount of taxes generated by the province. So the player can boost individual provinces, if he likes, for example to boost production in provinces far away from an agressive neighbor, or to boost very poor provinces.

But the player should have the possibility to disable or enable the setting individually for each province. The default should be disabled. That way, the AI can distribute the money in dependance of rate of return, which could imply, that all the money is spent in one province.

Only if enabled, the setting would force the AI to spend the specified amount in that province, and the other provinces would be handled normally.

Ok, that are my ideas so far.

[axi]

This is an issue that greatly interests me and is really relevant with what I have contributed to this project so far. So I should have posted earlier, but I was really up to my neck in other oblligations...

Now, let's give this discussion some grounding:

----------- Facts --------------------

Fact 1: Given Mark's confirmation some time ago, infrastructure will be divided in private/public sector.

Fact 2: Private sector preferences will be decided:
a) EP% by the social classes (UC, MC, LC, RC, WC, BE, RU) investment preferences according to their political power (where EP is the Econ Planning variable).
b) The remaining (1-EP)% by the socio-economic classes (UC, MC, LC, XC) investment preferences according to their income.
Public sector preferences will be decided:
i) Rpp% directly by the player (where Ppp is ruler's political power)
ii) (1-Rpp)% by the economic AI.

Fact 3: It is generally accepted that apart from the direct intervention into private investment included in fact 2a, the ruler will posess the indirect method of subsidies/tariffs to affect prices and consequently investment. It is obvious that the taxes/subsidies interface is similar in essence to what is described in issue 2b, below.

Fact 4: Given our firm decision upon using a no-buildings inrastructure model, the definition of an investment, that is of a job that has to be performed by the investment model, is "the addition/substraction of x units of infraclass y in province z". With the exception perhaps of military units and Wonders, these jobs do not need to fulfill specific xyz requirements to result to usable infrastructure. This is in contrast with civ2, where city improvements (specific #s of shields) and not shields were the valid infrastructure jobs.

Fact 5: As it was discovered during a discussion in the tech thread and later confirmed in the infra thread, 1 unit of infrastructure should mean the same thing for all civs and all times, regardless of tech advancement.

--------------- Pending Issues -------------

Issue 1: We have not yet decided which way these preferences should affect the actual investments. This doesn't actually effect the way investments are made, but the way preferences have to be computed. Two alternatives:
a) As in my infrastructure model, the preference vector gets copied into the budget vector.
b) As in Mark's proposal, the preference vector gets translated into the IV (Ideal Value) vector, which is compared with the current # of units vector and the budget is derived from the difference.

Issue 2: We have not yet decided what is the best way for the player to interface with the model. There are two fundamental types of interface:
a) As in the other civ-type games, the player is setting, within specific limits (determined by his power of intervention), what is the exact outcome he requires from the investment machine. The judgement-by-return method discussed here falls under this category, since it is referring to the question if it is worth performing a certain job.
b) As in my infrastructure model spreadsheet, the player gives preferences, which act as general directives to the investment machine. This is analogous to the way civ2 is handling Tax-Sci-Lux.

Issue 3: We do not have a clear picture for the exact way to compute everybody's preferences. This depends alot on the above two issues. Two approaches stand out so far:
a) As in the preferences machine in my infrastructure model, we use game variables from different models and, for each such variable a set of constants that represent our experience on the way that variable affects demands in each infraclass. Their combination provides us with a set of modification vectors, a combination of which is applied to a fundamental preferences vector.
b) As in the post by Richard above, we compute the Percieved Total Benefit (what we believe that the total, tangible and intangible benefit to us from performing a certain job will be), the Percieved Total Opportunity Cost (what we believe that the total, tangible and intangible benefit to us from performing some other, reference job, instead of this one, would be), subtract them and divide with the job's cost, to derive the Percieved Total Return, which will be a positive of negative percentage. By comparison of these returns we can pick the preferred jobs. This computation must be made once per agent per infraclass and all that once per turn in most cases. The PTB/PTOC figures can be found:
i) From pre-set formulae that come out of our, the game designer's economic analysis. This comes really close to case 3.a
ii) From actually calculating ahead for both the proposed and the reference investment and making a comparison, as the chess programs are doing.

--------------- Speculations ------------------

I haven't completed my speculations on this subject. I will try to inmplement some of this stuff in spreadsheet to see how things work. For now I'll just say that I'm supporting:
- Issue 1.b which I am planning to flesh out the specifics for.
- Issue 2a for the public sector and 2.b for the private sector.
- Issue 3.b.i, although I would like to see it integrated in the prefmach.

I'll return when I have time and something more solid on my hands


------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell

[Mark_Everson]

Hi Axi:

Thanks for the detailed recap. I agree with pretty much everything you say. I can't think of anything to add at the moment, though I'm focused elsewhere. The directions in your 'speculations' section are exactly what I would have put also . On 3, yes we clearly need to find a way to integrate preferences and return on investment. In a sense the preferences already (should) have the values embedded into it... I need to think about that one a bit. Looking forward to hearing suggestions on tradeoffs between preferences and ROI and how to integrate them!

[axi]

You know what? I was trying to include provisions for minimum output in the private sector, in front of the calculations with the Ideal Value. In the most desperate of the cases, when the people are virtually starving, there is need for a priority setting system, pretty much like the ROI thing. The code will be probably the same at any instance where issue 2.a is the case (with either a set #of_units_next_turn or a set output). When people have to choose between starving and freezing to death it is about the same ROI question as when the state wants to choose between investing on transportation or industry. I may sound incoherent right now, but we'll see...

It may take a while though. I am studing Automatic Control right now and from Monday till Thursday I'll switch to Operational Research. Perhaps I'll be more savvy and available after that!

------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell

[axi]

- Show how many FPRS are produced each turn.

- Give some graphical representation of A (effective tech level of industry)

- Do you have any income numbers you can show? It could be a measure of people's econ happiness.

- Ah yes, the % of product that goes to capital vs product that goes to wages, that would be nice to have, but...

...But I'd really have to see what you've got to tell you what more should be shown. I need to have an idea of the degree of progress before requesting stuff.

Can't you issue an online pre-demo which will be able to get polished after some days of our testing?




------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell

[Mark_Everson]

LGJ:

Currently I don't have any plans for modeling inflation at all. Do you really think its essential? Unemployment will definitely be in there eventually.

Axi:

Good suggestions, but not all ready to be implemented (A for example, since the tech model isn't hooked up yet so I don't much worry about A). What is currently 'in' is covered in the Economy Model OO Planning thread. But I still haven't gotten to trade and merchants. I'll talk to F_Smith about us getting out a pre-demo. But it could be more frustrating than useful for you!

[Mark_Everson]

Hi All!

My question for today, is what is Essential info for the player to be able to get out of the econ interface. This is just throwing together something for demo 5, so don't worry about getting everything right. And don't worry about info on trade, since it doesn't even exist yet.

What I have so far is, for any level of the economy:

You can see treasury available, tax rate, and how much income and net income the govt has.

You can issue orders for any type of infrastructure, by ROI and % of income to spend (for starting economies you could have a setting like ROI required = 100%, and %income = 1000%, which would bring in govt funds when there's a Big payoff)

What I plan for the near future is to be able to show growth rate in the tax base, and in production for whatever the economy produces.

Past that there are a zillion things that could be shown, that I don't intend to show much of... (things like the wages or return on kapital in the services sector)

What am I missing that is Most important to you, given the limited state of things so far?

I guess people's economic happiness and population growth rate might be other things...