Am enjoying playing the game casually, and have realised I’m being focused but INEFFICIENT in my policy choices - it’s losing me votes.
Solution - I’m thinking of building a £/Vote metric (quarterly cost vs 1% of overall vote) - in Excel / Power BI, which I can use to guide my policy decision making. A more accurate label would be ‘net direct £/Vote’ - as am only considering direct impacts on voting groups.
Note - I’m not expecting this to be in the game, just something I do on the side / offline.
How - if for each TAX + ECONOMIC policy (plus those on the main expenditures + incomes), I select an optimum level for each policy (urgh) and note down the (1) £b finance gain/loss and (2) the vote impact for each group (liberals, etc), then I should be able to calculate the overall vote impact (I.e. SumProduct of the vote impact for each group x the % of that group within the overall population) … then compare the two values together to give a £/Vote metric.
Example - UK at the start of the game, with default settings - option (a) cutting military spending to the minimum amount saves ~£10b (!), but loses you X% of votes, while option (b) eliminating foreign aid saves only ~1b (9b less), but you lose a similar X% of votes. Therefore simplistically it’s more sensible (simplistically) to cut Policy (a).
Simplification - unfortunately this deliberately ignores some of the complications.
ignores indirect impacts like crime or GDP, which also affects voting
ignores voting group changes over time, like how the capitalists group grows when I have more capitalist policies.
ignores people who don’t vote - so buying a 1% vote increase doesn’t actually get me that 1% result.
ignores events - having a minimum UK military also gives negative Argentina / Russia military events which then affect voting.
ignores other elements - like GDP, crime, health, transport, etc.
BENEFIT:
However theoretically this should give me better (simplistic) guidance on the finance vs vote impact of my policy decisions. It won’t be perfect, but should be ‘better’ (I.e. more £/vote effective) than what I do now.
An interesting idea. Budget is one of the most important resources in this game and definitely needs some consideration before spending decisions. I’d add some of my thoughts.
Some of GDP stimulus policies (or ones that can remove uncompetitive economy or corporate exodus) are effectively tax raisers. While it may be impossible to measure accurate budget/vote for GDP policies since income tax revenue is (slider)(0.5+0.5GDP), They are worth noting when you are trying to maximize budget availability.
(refer to the gif pic below) It seems you are trying to measure voting intention changes after each policy adjustments. I expect you will get fairly reproducible results but the same approval changes can bring different vote gains/losses since a) voting groups will change as you’ve mentioned in your post and b) only the number of electorates at the threshold (roughly at 60~70% approval in 3-party-system) will change their votes. So there’s a chance your model ending up over-representing some groups which are at the threshold at the start.
Nonetheless, I think this can work as a decent first-approximation and assist players with more insights.
Huh. That’s an interesting + useful response, thank you. Some detailed responses below…
I hasn’t thought of including GDP stimulus measures. That’s a nice idea, even if it’s only directionally (not exactly) correct, though am not sure how I’d do it yet. Will ponder that.
Actially I intended to measure “expected” voting intention changes, not “actuals” - however actuals would be interesting to see, I hadn’t thought if it, just depends if I’m willing to put I the time to test that out. Hmmm another to ponder.
You are also right about the over-representation issue. I hadn’t thought of that, but makes complete sense. I guess that’s a weakness I’ll have to accept, unless I use a different tool / method.
Specifically for a UK start - here are most of the tax & economic policies, with their financial impact vs voting impact mapped on a grid.
It shows there are 8 (eight!) policies which BOTH improve finances AND improve expected voting intentions - which is a big surprise.
Note - that some with have extra negative impacts - e.g. cutting CCTV & Intelligence Services directly saves finances AND improves voting intentions, but increases crime etc and so indirectly may harm later voting.
Those tables show you the policies in each of the key 3 quadrants. Notice in 2 quadrants, there’s also my $/vote metric (yellow formatting), to assist / guide on the best trade-offs.
Interesting results. They are confirming some intuitions and presenting potential insights.
Prisons, Police Force, Community Policing, and Science Funding are indeed good investments. I tend to address these policies very early.
It is understandable that Inheritance Tax & Diverted Profits Tax are seen as inefficient ways of raising tax revenue. The former is there to improve Equality and the latter prevents a potentially costly situation.
While the result shows that Military Spending cut is generally good as expected (as long as you can avoid potentially bad events), I didn’t expect Automation Tax to be that good. It might raise even more income since the latest patch introduced negative impact from unemployment to income tax revenue, effectively buffing the automation tax.
Also interesting to know that cutting Agricultural Subsidies, CCTV, Nuclear Weapons, and State Broadcaster might be better. Not going to cut Intelligence Service to prevent some bad situations though.
Some Additional Thoughts
Would be interesting to know whether directly cutting Pension spending is better or raising Retirement Age is. Both can be used to cut spending but has different effects.
I generally find state corporations to be ineffective in terms of approval. Since State Broadcaster has been found to be bad for both money & votes, I’d expect similar to the others.
Oh wow. Big update coming up today or tomorrow with some new visuals / new analysis.
Am shocked at how “payroll tax” is even more £/vote effective - even more than automation tax! - where you can raise significant income (20b) with only losing a little % voting intentions …
… though of course there are other (non-£ and non-vote) negative effects - like GDP.
Yeah Payroll Tax is a really good tax raiser. It’s not even that hard to maintain high GDP with high payroll tax rate. I sometimes think that maybe payroll tax revenue should be capped or aligned with state welfare expenditures.
Top Right Quadrant - easy to gain BOTH money + votes, through frequent flier tax, mansion tax (not at max!), reducing foreign investor tax breaks and cancelling state broadcaster (bye bye BBC).
Bottom Right Quadrant - reducing Inheritance, Property and Diverted Profit taxes are all good vote winners.
Top Left Quadrant - there are 4 great taxes which raise +£50b/quarter in total, while losing only <8% of everyone’s expected votes (payroll, automation, sales upto £10b extra, and graduate tax). You can counter-act that vote loss through the +18% gain of everyone’s expected votes, via spending +£1.5b more maxing out the police force and prisons.
Corporate Exodus - note that the combination of higher payroll + new automation taxes may cause a ‘corporate exodus’ (not fun for GDP).
Terrible Taxes - some of the other taxes in the top left quadrant are shockingly terrible, inc. petrol tax, property tax, high/max sales tax, tobacco tax, max. mansion tax, and surprisingly crypto. tax.
Military Spending - cutting this to the minimum saves £10b/quarter, while only losing 6% of the expected vote (from patriots). The other 4 great taxes are a better first choice, but this remains a good backup fundraising option if you can spare the 6% voting intentions.
Security - cutting / cancelling some security policies are great vote winners, but perhaps the crime-related impacts are too scary for me to consider - i.e. CCTV, intelligence services. Also cancelling tobacco tax is also a good vote winner (+6% in expected votes), while only losing £1.5b - but the health impacts feel tough to accept.
Hope this helps, and is interesting for some of you!
Another thing to add) Tobacco tax might have been estimated to be worse than actual effects. Since the approval effects of tobacco tax is linked to the Tobacco Usage simulation, raising tobacco tax will actually cause less disapproval in a mid-long term as less people will keep smoking. This is why tobacco tax can be used instead of smoking ban by maximizing the tax rate and ending up reducing tobacco usage down to zero, causing no approval drop eventually.
Interestingly … alongside this £/vote graphic + metric … I’m also enjoying using an Influence/Vote graphic + metric, which is particularly useful close to an election where you have limited influence to pick up as many votes as possible!
Stuff like this is super interesting, but getting a really helpful model is almost catastrophically hard because of many factors not least:
Each country is different in initial voter group memberships
Some policies have circular effects: Corp tax income depends on GDP but high corp tax can indirectly lower GDP, leading to a medium term equilibrium that differs from the immediate impact.
There are other factors, such as policies that can lead to extremism, terrorism and potential assassination, which is game ending, and this needs to be factored in.
The value of money in the game is not linear. When you have high debt, and this is causing problems, added govt funds are invaluable. When you have a budget surplus extra funds have less real value.
Its a nightmare!
It would be awesome at some point to write some amazeballs neural network AI that played 50,000 games with different strategies and worked out what strategies were overpowered in general
This would be a nightmare to get accurate + to put into the game itself - for all the reasons you said and more!
However as something on the side / out of the game, which just gives a little more direction (knowing it’s many weaknesses).
Specific to the UK, it’s been really useful to guide my first two terms (10 years) - to pay off the UK’s debt (not needed, but a fun challenge) + still getting a decent vote majority.
After that, I’ve had a lot of fun changing the “outputs” on the y-axis - to economic growth, education, crime, etc - to help me select a more vote / influence efficient policy to solve that problem (e.g. obesity). Not perfect, but useful guidance.
SUGGESTION - if you wanted that neural network model of 50,000 games, why not approach a university’s master or PHD programmes on analytics? The students there are begging for real datasets (or in this case, models of reality) - it would be free for you, and probably you could get 3-5 students would be delighted to build their own neural network, process 50-100k games, and then each summarise the results back to you. Little effort on your part, but you get the end-results you liked the idea of. Just an idea.
Yes absolutely. Funnily enough I used to work with demis hassabis and dave silver from deep mind, the guys who made alphago. (I was at elixir studios with them). I bet alphago would enjoy the game
You are not immune to political propaganda. Very few people are, these are only stoics who are completely disinterested in it, and people who have extensive experience in the field. Of the latter, some of those people are actually the most affected.
Your political ideals, if you are unfortunate enough to have any, will suffer if this game is an accurate model of reality and if the model is then solved. Development of this game should remain politically neutral, and the best way to retain neutrality is to avoid exploring the solution.
In addition to retaining your own neutrality, it will also inevitably alienate a large percentage of the public towards the model itself, if the solution becomes known. The ideals from which the public suffer must be deconstructed only at a pace which is comfortable for them. It is better for them to explore the model at their own pace, and to fail to understand its implications to the desired extent. They can’t handle the truth.
