Considering some alternatives to having the workers mindlessly spend all their money on food because of the problems of balance.

The problem with the economic model right now is that the workers spend all their available money on food, which leads to an infinitely high, inelastic demand cruve, making sense for the firms to continuously raise the prices. I want to avoid having a separate utility function for choosing where to shop and choosing the quantity to buy.

Maybe some utility function for the value of money based on income rate, which can be put into the shopping location utility function and give the quantity. Also a utility function for work, if there's value in work itself.

But I don't want anything too complicated, and marginal utility would be heavier on the CPU. What's the most direct way to do this, from real life?

You are not giving much information about the game?

Anyway, just have a table with the consumption of 1 person per day for all necessary goods. Then calculate where these demands can be satisfied most cheaply. Possibly check everyone got enough money to actually buy. Maybe check if people are gonna starve from not being able to afford enough.

If you want, you can have different tables for different kinds of people, like rich people needing some luxury goods and eating different types of food.

You could try this optimization.

1) Make complex calculations for budget spending like:

A: if the worker earns <9 spending is 100% food

B: if the worker earns 9-15 spending is 70% food, 30% luxuries

C: if the worker earns 16+ spending is 40% food, 60% luxuries

Update this when needed and take into account market prices, trends, new goods introduced, etc, anything you want.

2) For each worker compare the salary and assign the formula grade (A-C), so you don't need to do those complex calculations per worker but only once.

A very simple model would be that if the population needs N calories, the cost per calorie is the marginal cost of Nth cheapest calorie. You could include a fixed 'inefficiency' multiple to simulate that a real economy isn't that efficient. The assumption is that although demand for calories are inelastic, competition is extremely high so it still moves towards the marginal production cost.

The big issue there is everyone will be eating whatever rice paste is cheapest to mass produce. The solution I'd look to is tiers of food quality as substitute goods. Treat better quality food however you treat luxury goods, and subtract those calories from the subsistence food category.

If you arbitrarily set a minimum wage you can look at typical real life budget guidelines to determine how much money a worker spends on food, rent, entertainment, or whatever you want to include in your game. Assuming that a minimum wage would allow a worker to survive with at least the lowest quality products for each need, you could then say that all workers adhere strictly to this guideline and will always spend the appropriate amount of money for each need. The more money a worker has available the better the quality of products they automatically purchase. Or if you don't plan on different quality of goods maybe they'd get a better morale boost or bonus of some kind.

But I don't want anything too complicated, and marginal utility would be heavier on the CPU. What's the most direct way to do this, from real life?

This reeks of premature optimization to me. How do you know at this point that calculations involving marginal utility would be too slow?

People need to spend at least for the very base of Maslow pyramid ( food, water, clothing, shelter) , if all three is provided by employer as it was during early Industrial Revolution, wage becomes a function of food therefore having a Giffen product curve as you mentioned.

Combining this with Harris-Todaro model which involves formal and informal sectors where formal sectors being scarce with benefits leads to an informal sector at a wage function of cost of living in the city , I think you can have calculated amounts for each step in Maslow pyramid specifying where people spend money for. You may prefer to spice things up by incorporating irrational non homoeconomicus choices as well.

I suspect the solution is in not using curves at all, but making agents decide when they're hungry and choose what to buy. Why would a worker stuff themselves with extra food just because it's cheaper? I think this only makes sense for certain substitutes who the worker wants more but can't afford yet/as much. The curve for inferior goods would be negative to income as the workers can buy others with money saved. The point I think is I need several substitutes. Having many competing firms is not enough. I thought this improvement was enough. This effects not only food then but all commodities that can have subtitutes. Commodities that are needed at a certain quantity but are not elastic to price will ultimately fall on the final consumer and be elastic that way. I thought that the substitution effect at the level of different places to shop would be equivalent to that of substitute goods. But I think the main point is that the workers don't mindlessly spend all their available money on food and make sense for firms to lower prices. The price of inferior goods, being the last resort and having a slightly positive curve in proportion to how much of the income is spent on it (greater proportion of income leaves less to be spent on prefered luxury good), might rise to the price of luxury goods if they are indeed positively curved, making the inferior good the one that costs less. I think that's a real concern here. But as long as workers are able to leave some income in reserve and are able to afford enough food to cut back on further purchases, it makes sense for firms to lower prices reasonably, and the inferior good will not have a positive demand to price. So having several shopping places might be enough of a substitution effect if the workers are able to vary their spending based on price/income, and start families/divide when they are able to afford more.