Improving accuracy of state employee & unemployment effects in Democracy 4 May 1, 2021 cliffski One of the problems with making a vidoe game with one person doing the development, where you try to model the entire world in inter-connecting detail…is that its impossible, so you end up doing a lot og guessing and thinking ‘yeah that looks ok to me’. Then eventually you find the free time (ha! its the weekend, who am I kidding?) to go back and check that the wild ass guesses you made were just wrong, and not OMG emebarrasingly badly wrong. Its a low bar, but i’m determiend to hit it. My game Democracy 4 has a lot of policies that affect two important variables in the game – The membership of the ‘State Employees’ group, and the level of unemployment. For hopefully obvious reasons, these numbers are important in a government/ politics sim. They have to make some vague sense. Until now, the numbers in the game have kind of been guesswork, and result in equations like this: StateEmployees_freq,0.02+(0.05*x) That means that the effect of that policy will vary between a 2 percent and 7 percent boost to the number of voters who identify as state employees, depending how the policy slider is set. In other words, this policy assumes that at max capacity, this policy represents a seven percent higher chance of any voter joining that voting group, although in practice its much more complex than this, due to internal algebra that I wont bore everyone with… The problem for me is that although I do not mind that seven percent figure being possibly inaccurate, I DO want the games model of this stuff to be internally consistent. To put this another way, if in the real world, a state health service employs 10x the people of a state postal service, the game should attempt to get that ratio correct, at the very least. With this in mind I have done some research, using the USA as my base case for the policies most impacting state employees: These figures are NOT 100% accurate. This is mostly because the US does not have a state health service in the same way the UK does, so I had to take NHS figures and then adjust for population. I also had trouble getting energy figures, so I extrapolated from the top 10 companies employees and adjusted on a per-household basis. The point is that although the figures (like all figures) are a bit wrong (probably) they are massively less wrong than my guesses! So now to make the game values make sense, I need to work out how to adjust those values in the second and third columns, which are my current effects (at max slider) on state employees and unemployment. Given that I do not want top massively unbalance the game, I thought it was prudent to keep the total combined effects of all of these policies the same (93% and 126% respectively) and just adjust the figures internally to fix the relative impact, The way I’ve done this is to use the employment percentage (the actual percentage as a portion of total employment of all these policies) and multiply that by the total current in-game effect (93 or 126) to give me the new adjusted effects. That looks like this: In some cases its not too big a change but in others its hilariously different. Currently the game is giving a HUGE (18%!) boost to state employees from armed police when it reality it should be about 3%. On the other hand state schools were set to have a 10% boost to employees, and should be having a 29% boost instead! So many teachers! Its also evident that in the grand scheme of things, prisons and a state broadcaster employ virtually nobody. (I scaled up the BBC employment figures to USA size to get that data too). I have just done the data so far, but later today I’ll go through all the policies and adjust the values in each equation. I guess the takeaway from all this as a player is that if you really want to cut unemployment when you spend money on public services, you want to splurge cash on the health service and schools/universities. Everything else is trivial. Keen-eyed economists might note that the REAL employment impact would be different. For example, the direct employment from the US military may be 1.3 million, but defense contractors etc will employ many more, and the knock-on effects from the contractors CEOs buying new ferraris is even higher. I agree, but that is best dealt with through a GDP boost I think.