Hello all,
Lacking a more profound technical background yet hopefully in possession of a basic understanding of linear programming, a question that concerned me for a long time now is how an optimization process could account for a gradual priority of needs. If we optimize for plan fulfillment, is it possible to hierarchize different consumer goods in the optimisation process? My understanding is that once we optimise the economy in regards to the plan target, all those needs expressed in the plan target are on a flat plane.
Might there be a bias toward the fulfillment of smaller goods then? Take a simple demand of 5 airplanes to 50.000 cars. If not all can be produced will a car unit be equal to an airplane unit in the eyes of the optimisation solver? Can we weight them in any way?
The other question I have is how can we prioritize different goods, let's say Covid vaccines over footballs for instance. The only way, correct me if I am wrong, my mathematical limited brain can come up with to rank them would be to introduce the prioritized item as a constraint so that an optimal solution is restricted to those where the number of Covid vaccines is higher than value x. This however only allows a very crude binary prioritsation and nothing fine-graded. Either the item appears as a constraint or not.
Did anyone give this a deeper thought? Or do you even think it is problematic at all?
Best,
Max
If I remember right Kantorovich imagined that a fixed ratio would be established between goods, speaking of tables and chairs. So concerning the planes and cars example, there would be a ratio of 1/10.000. So it would never produce 30.000 cars and no plane.
But it will likely produce 9999 cars before producing the first plane right?
After watching Paul’s harmony function again, I don’t find a solution there either.
I understand that Paul wants to optimize the economy by shifting productive capacities to those industries that are punishing the function the most (underproduction). But imagine a highly simplified toy economy of food and luxury goods. The harmony algorithm is blind to the marginal utility these use values might bring. All it knows is the projected demand for these goods. But it would be way more important to meet demand for food and stomach the severe underproduction of luxury items, even though the harmony function might provide a different or more balanced solution. If there is scarcity and the demand for both cannot be fulfilled, the harmony function cannot tell you which of the two to prioritize. It just maximizes the harmony function of industries regardless their utility and importance.
An idea of course would be to introduce weights to each good to increase the punishment of the function for crucial goods like vaccines for instance. In the sense of a hierarchy of needs to ensure that food and housing might be prioritized over game consoles and champagne. But how would these weights be asigned? By planners or consumers?
[...] If we optimize for plan fulfillment, is it possible to hierarchize different consumer goods in the optimisation process?
You can try to maximize production of good A and after that maximize B if there's any "space" left for that, but that is unlikely to be the case. This is why me and Dave suggest turning demand into constraints instead. That way you can say "increase production of A by 10% and B by 10%" rather than "increase production of A+B by 10%" which would result in very different plans.
My understanding is that once we optimise the economy in regards to the plan target, all those needs expressed in the plan target are on a flat plane.
That is correct.
Might there be a bias toward the fulfillment of smaller goods then? Take a simple demand of 5 airplanes to 50.000 cars. If not all can be produced will a car unit be equal to an airplane unit in the eyes of the optimisation solver? Can we weight them in any way?
You can use the value (SNLT) of each good as that good's cost.
The other question I have is how can we prioritize different goods, let's say Covid vaccines over footballs for instance. The only way, correct me if I am wrong, my mathematical limited brain can come up with to rank them would be to introduce the prioritized item as a constraint so that an optimal solution is restricted to those where the number of Covid vaccines is higher than value x. This however only allows a very crude binary prioritsation and nothing fine-graded. Either the item appears as a constraint or not.
Looks like you're already on the right track 🙂 Turning all demand into constraints is much easier in my opinion. Kantorovich's plan rays were a mistake.
But it will likely produce 9999 cars before producing the first plane right?
Planes are capital investments, so you'd likely fix the number of planes to build and let the amount of consumer goods be whatever there is space for.
But imagine a highly simplified toy economy of food and luxury goods. The harmony algorithm is blind to the marginal utility these use values might bring.
It's not so much the algorithm as the use of plan rays that is the issue. We can easily imagine a situation where some luxury good is overvalued, causing the plan solver to allocate all resources to that good and zero resources to food production. I've been meaning to write a text how plan rays have a latent bourgeois character partly due to cases like this.
A better way to do this could be to deem certain goods "essential" and insert constraints one after the other and checking that the system is still viable. Only when all essential goods and services can be produced do you allow for the production of luxury goods.
But how would these weights be asigned? By planners or consumers?
Likely by a combination of political decisions and science.
Turning all demand into constraints is much easier in my opinion. Kantorovich's plan rays were a mistake.
A better way to do this could be to deem certain goods "essential" and insert constraints one after the other and checking that the system is still viable. Only when all essential goods and services can be produced do you allow for the production of luxury goods.
Hey Tomas, thanks for your thoughts, don't you see a problem that we would just achieve a binary hierarchy with this? Either goods are classified as "essential" and introduced as constraints or they are not. I would argue that it would require a finer-graded hierarchy between consumer goods. The scarcity of some goods hurts more than others. The same problem also applies to capital goods and infrastructure projects by the way, which will also have to be ranked by importance. Concerning consumer goods, we might want to ensure that there is enough food, before producing painkillers, and only when this demand is met is when we begin to produce gaming consoles. Then we are also struggling with non-linearities. In theory, such a hierarchy is established in an idealized market through the price mechanism and subjective judgment of marginal utility.
How can we give justice to these subjective judgments in the planning process? Although I am highly skeptical of defining something like a hierarchy of needs (who does the classification? Even a democratic procedure would result in a tyranny of the majority), but without it all these needs would be on a flat plane in the optimisation process, which is even more problematic in my opinion. I am pretty certain that such a hierarchy could be established with weights as proposed for the harmony algorithm. The challenge might be more the problem of classification rather than the technical implementation, as we will inevitably run into problems when differentiating between basic food and delicatesses for instance. Because the marginal utility for these items is not assessed directly by the subject but a democratic rating process, it will lead to situations where many individuals who might regard a product as essential will have to pay a higher price for it due to a scarcity that is the result of a lower social evaluation, determined either by planners or the social body by vote. And this is a problem any economic planning proposal faces, not just one with an optimisation solver at its core.
As far as I am concernced, something like a ten priority scale would practically suffice maybe to further differentiate the importance of plan fulfillment for individual consumer goods. With food taking up the highest priority for instance and luxury goods the lower ones. The priority of different product groups on a really aggregate level could be put up for vote. However, such a rating process will only result in the most dissatisfactory hierarchy and I fear this is the best we can do, better than not to hierarchize at all though. Here is a related excerpt of Ernest Mandel's In Defence of Economic Planning on the hierarchy of needs:
In responding to Nove’s critique of the Marxian heritage, we have introduced the concept of ‘relative intensity of needs’. This notion has several important implications for a discussion of socialist planning, to which we should now turn. In the West today the variable intensity of needs finds expression in differential consumer behaviour towards ‘priced’ goods and services (if also ‘unpriced’ ones). But it does not have to be measured indirectly in money. It can be empirically ascertained by, for example, studying changes in physical consumption patterns when income suddenly declines (as it has done for vast numbers of people during the present depression). Certain widespread features will then clearly stand out. For some expenditures will be cut before others. Certain varieties of goods within each major category of consumption will be reduced, while others will be increased (more pork and less lean beef will be consumed). Expenditures on health will prove more rigid than on items of toiletry. These are not random preferences. One of the most important advances in knowledge brought about by capitalism – in a sense, it is a compliment to capital – is that because of the rise in the living standards first of the middle classes, and then of wider layers of the population, there are now a great deal of empirical, statistical data on consumption patterns that are remarkably similar across a large number of countries. These reveal an objective order of priorities common to hundreds of millions of people, over many decades. All responsible enquiry into human needs should start from that evidence. What emerges from any such enquiry is a pattern which the Prussian statistician Engel already noticed one hundred and fifty years ago. Once needs become diversified with economic growth, a definite hierarchy can be discerned among them. There are fundamental needs. There are secondary needs. There are also luxury, or marginal, needs. Roughly speaking – and here we stand ready to be corrected by empirical data, not by metaphysical speculation – we would put in the first category: basic food and drink, clothing, shelter and standard comforts linked to it (heating, electricity, running water, sanitation, furniture); education and health provision; guaranteed transport to and from the workplace; and the minimum of recreation and leisure indispensable to the reconstitution of labour power at a given level of work pace and stress. These are the needs which for Marx must be satisfied if an average wage-earner is to continue working at a given level of effort. They can be subdivided into a physiological minimum and a historical-moral supplement. They vary across space and over time. Their fluctuations do not depend only upon major changes in the average productivity of labour. They are also a function of the great shifts in the balance of historical forces between contending social classes. But at any given moment, in any given country, they are objective data – which are also clearly present to the consciousness of the great majority of the population. They cannot be arbitrarily altered (including by the operation of ‘market forces’) without violent disturbances in the social and economic fabric.
In the second category of goods and services we would classify most of the more sophisticated foods, drinks, clothes and household appliances (excepting the fanciest ones), the more elaborate ‘cultural’ and ‘leisure’ goods and services, and private motor vehicles (as distinct from public transport). All other consumer goods and services would go into the third category of luxury expenditures. Of course, the precise frontiers between these three categories of needs are difficult to draw. The first is the easiest to delimit. The gradual passage of needs (and of goods and services fulfilling these needs) from the second to the first category is a function of economic growth and of social progress (in particular of the results of proletarian class struggle). Paid holidays for all are a recent conquest of the working class, dating from the great wave of factory occupations of 1936-37 and its later fall-out across the industrialized world. The distinction between the third and the second category is more a matter of socio-cultural preferences than an observable mass phenomenon.
But while all these points merit emphasis, the general pattern which emerges remains fairly clear. The hierarchy of human needs manifestly has both a physiological and a historical social basis. It is neither arbitrary nor subjective. It can be encountered on all continents, under the most diverse circumstances, albeit in non-synchronous fashion because of the uneven and combined development of economic growth and social progress. This hierarchy of needs is not the result of any diktat, either by market forces or by despotic bureaucracies or enlightened experts. It finds expression in spontaneous or semi-spontaneous consumer behaviour itself. The only ‘despotism’ involved is that of the large majority. ‘Eccentric’ minorities – who for the most part are not so few in absolute numbers – will not fit into the general pattern: teetotallers as against consumers of alcoholic beverages; smokers as distinct from non-smokers; vegetarians as against meat-eaters; people who refuse to look at television or cannot or will not read newspapers or books; others who decline ever to see a doctor or go into a hospital on principle. Nevertheless, given the fact that a very large number of people are concerned – hundreds of millions – the law of averages tends to balance out these exceptions and to maintain across time and space an emergent pattern that testifies to a definite hierarchy of needs among the overwhelming majority of consumers.
This hierarchy has one even more important aspect. Not only does elasticity of demand tend towards zero and into the negative from the top of the priority list downwards, item per item, with each successive stage of economic growth. It also tends to do so by major categories of needs. Per capita consumption of staple foods (bread, potatoes, rice and so on) in the richest industrialized countries is today definitely dropping both in absolute physical quantities and in percentage of national expenditure in monetary terms. So is consumption of native fruit and vegetables and, at least in money values, of basic underwear and socks, as well as elementary items of furniture. Statistics also show that, in spite of growing differentiation of tastes and goods (many varieties of bread and cake, a much greater range of food and clothing generally), the overall consumption of food and clothing and footwear tends to become saturated and even starts to decline, measured in terms of calorie-intakes, square metres of cloth and pairs of shoes.
Either goods are classified as "essential" and introduced as constraints or they are not.
No, all goods are constraints in the model that me and Dave Zachariah are proposing. Optimization is optional and secondary.
What you have to deal with is the case where demand is high enough that there is no feasible plan. If this happens then you either need to slacken one or more demand constraints or introduce more efficient means of production. There is an uncountable many schemes that can be used to ensure feasibility. For the most part they fall under the umbrella of the remuneration problem.
without it all these needs would be on a flat plane in the optimisation process
Not true. In the example earlier good A >= demandA and good B >= demandB form two half-planes and a ridge at the intersection. This is *not* the same as forming one half-plane for A+B >= GDP. It is very important to understand this. Combining lists with weights into a single constraint is not the same as forming a constraint for each list in turn. Two constraints are more than one.
It should also be kept in mind that things would be set up such that the plan is *always* feasible. This is a point that Jan Dapprich has raised. The system will simply not allow you to introduce constraints that take the system from feasibility to infeasibility.
As far as I am concernced, something like a ten priority scale would practically suffice maybe to further differentiate the importance of plan fulfillment for individual consumer goods.
Yes, something like this could work. A comparison could be made with different commodities having different VAT rates, at least here in Sweden: 25%, 12% and 6%. Therefore having different "buckets" of goods and services is evidently not an insurmountable problem. This is also why democratic participation in the system is important, why we want *everyone* to plan. Statistics also plays a part.
Combining lists with weights into a single constraint is not the same as forming a constraint for each list in turn. Two constraints are more than one.
Much appreciated Tomas, I will sit down next week and go through your model in detail. I still wonder how this would be solved though:
What you have to deal with is the case where demand is high enough that there is no feasible plan. If this happens then you either need to slacken one or more demand constraints or introduce more efficient means of production.
In most cases, I would argue, we will not be able to meet all demands. The plan target will not necessarily result directly in a feasible plan, as (until unlearned) human desire will outrun our productive capacities and the carrying capacity of our planet. Thus, we would constantly, arguably more in the beginning, have to lower demand constraints for specific goods to ensure that
things would be set up such that the plan is *always* feasible
Regulating consumption with some form of credits/tokens would help of course.
However, there again arise the questions of who will alter the constraints and how we know which demand constraints to lower. How do you envision the political procedure to rank the myriad of demand targets? How fine-graned will it be? Ranking all existing goods will be infeasible from a complexity standpoint but will the ranking on an aggregate level, for instance in product groups, really suffice?
You would have to enforce what Mises called value judgments (rankings of competing ends) on a collective level. Whether these valuations are applied by adding weights or lowering individual demand constraints is really only a question of implementation - an important one nevertheless. But I am unsure how we interface man and machine here. In capitalist production, we have the price system, which crudely transmits such information in a very limited form. Maybe the bar is not too high and we just have to do as good, or better.
What you have to deal with is the case where demand is high enough that there is no feasible plan. If this happens then you either need to slacken one or more demand constraints or introduce more efficient means of production.
In most cases, I would argue, we will not be able to meet all demands. The plan target will not necessarily result directly in a feasible plan, as (until unlearned) human desire will outrun our productive capacities and the carrying capacity of our planet. Thus, we would constantly, arguably more in the beginning, have to lower demand constraints for specific goods to ensure that
It might be useful to separate aspirations from the concrete continuously changing plan.
However, there again arise the questions of who will alter the constraints and how we know which demand constraints to lower. How do you envision the political procedure to rank the myriad of demand targets? How fine-graned will it be? Ranking all existing goods will be infeasible from a complexity standpoint but will the ranking on an aggregate level, for instance in product groups, really suffice?
This is the remuneration problem. I suspect two categories would be enough: those goods and services for which payment is not demanded, and those goods and services for which payment *is* demanded. The former could be health care, potatoes and housing. The latter beer, beef and jewelry. As for how to go about it, start out with everything in the latter category. We call that state of affairs socialism or lower-phase communism. As things are moved from the latter to the former category we transition from lower to higher phases of communism. When *all* goods and services are in the former category, then we are in capital-C Communism.
haven't had time to read the entire thread so apologies if this has been said before.
i've been trying to rewrite NewHarmony for R and in doing so I've gotten to the section where the "fit" for each possible plan is calculated. one way is to take the average fulfillment across all goods for a plan and pick the one with the highest, another is to take the lowest performing good and have that be the overall score for the plan.
BUT there *is* another option: weighted average. If one wanted to, one could apply a steep weighting to particular goods and then do the averaging method for plan scoring.
The upside to this method is that it avoid unnecessary stockpiling. Stockpiling and extra production are incredibly important (the problems of Just-In-Time Supply Chains are now very obvious in the post lockdown economy) but one might be tempted to use them as a fail-safe for making sure key goods are produced. (how can we assure we get the 100 of good X we need? set the target to 200)
The weighting can help avoid this. Given the political space surrounding how NewHarmony could be used, this method can also avoid the pitfalls of other similar planning endeavors. in WWI, the US had a problem where non-central planners could set priority for their goods. everyone set it as A, the highest. if we adjust the weighting (which is really just a way of setting priority independent of amount), we can avoid this altogether by only setting priority for final, end-user goods rather than anything intermediate. the intermediate amounts needed will follow naturally.