Supply and Demand Is Overdone

Many, many, post keynesians and MMTers have criticized supply and demand curves, as reductive, misleading, and hard to observe.

Here is a video clip of Randall Wray doing precisely that. Watch from about 9:40 to 11:40

Randall Wray on Observing Supply and Demand Curves

Steven Keen, as well, frequently criticizes supply and demand curves, especially in terms of an aggregation problem. (Varian, Microeconomic analysis) It is often difficult or impractical to mathematically create nice aggregate supply or demand curves, from individual curves.  His book "debunking economics" apparently addresses this.  In both cases, Keen and Wray reference specific research or mathematical results done by others, and use that to make a broader argument about the (non)practicality of supply and demand.

I want to criticize the use of supply and demand models as well, but less of in a machine gun fire fashion, throwing out random criticisms, and instead delineate precisely when and why it is useful, and where it often fails.

As an empirical model, supply and demand curves are often especially problematic, precisely because bidding is an adversarial process.  If you know much about price negotiations, each party wants to keep as much information as they can private, which gives them an advantage in negotiation.  Furthermore, market prices are observable

Two Common Price Practices

1. Fixed prices with floating inventory
2. Fixed inventory with floating prices

Before we get into these, I wish to talk about the markets that do behave more as a conventional supply and demand model would suggest, in which prices and inventory both float on roughly the same timescale.  These two markets are oil, and housing.  The reason why these markets have realtime fluctuations in both price and inventory, is because both production and consumption have volatility.  For housing/real estate, this is because the markets are highly local, and also involve large assets with complex financing.  So even though the assets are a local, the financing market is global, which makes for a particularly complex marketplace.

When it comes to oil, oil is produced and used constantly, and is a critical resource.  Unlike other consumption commodities where businesses can smooth out underlying costs and volatility, the oil market is particularly difficult to predict.  But outside these two markets, most markets will roughly follow one of the two above pricing practices.

In the wild, there are basically two common ways in which prices are managed.  Either way allows businesses to interact effectively with markets.  Scenario one is where the price is fixed, but the amount of inventory can float. This is especially common in consumer markets.  Businesses strive to offer stable predictable prices, so that customers can act out of habit.  Whether video game consoles, or new vehicles, companies will often try to make the prices of their products as stable and consistent as possible.  They turn a profit by keeping costs down and sales up.

The second scenario, is where the inventory is basically fixed, and then prices float in real time.  This is common in the stock market.  Capital raises and stock buybacks are how companies issue or delete shares, but this process happens rarely, is announced beforehand, and

These pricing practices do not mean that producers and consumers are not making decisions based on negative tradeoffs, it simply means that it is hard to empirically separate one curve from the other.

For assets that trade in realtime with fixed inventory, you frequently will have an exchange with an "order book", which can be visualized as depth charts.  These order books look deceptively similar to supply and demand curves, but they are not the full picture.  Trading platforms may allow people to place stop losses or other "limit" orders privately, without posting actual orders until the price gets within a range.

Realtime Price Feedback Systems

In any case, for such real time prices, the order books are constantly shifting based on information, and then agents use that information to adjust their future bids.  So this is explicitly a feedback process.  I took one course in feedback and control theory from the computer science department at BYU, under professor Sean Warnick about  2010 or 2011.  We used the textbook "Feedback Systems: An introduction for scientists and engineers".  I consider this to be one of the best textbooks ever published.  Here is what Astrom and Murray have to say about feedback: 

Simple causal reasoning about a feedback system is difficult, because the first system influences the second and the seconds system influences the first, leading to a circular argument.  This makes reasoning based on cause and effect tricky, and it is necessary to analyze the system as a whole.  A consequence of this, is that the behavior of feedback systems is often counterintuitive and it is therefore necessary to resort to formal methods to understand them.  

-Astrom and Murray Feedback Systems: An Introduction for Scientists and Engineers

So because agents use real time information to adjust their bids, it is difficult to evaluate supply and demand separately.  Furthermore, the economy is in feedback in an additional way: the price of one item affects the costs faced by others.  For these reasons, it is often simpler to assess directly effects on the market or clearing price, and the quantity cleared, rather than constantly reconstructing independent price response curves, which could be just as volatile as the market price, and more difficult to observe.

Supply and Demand as a Conceptual or Theoretical Model

So even if we have reservations about exclusively relying on supply and demand for empirical studies of prices, it could still be a useful conceptual model for understanding how prices and the economy work in general.  In particular, the concept of "surplus" in the context of the supply and demand model is a useful idea.

Surplus is not excess inventory, it is the difference between a limiting price and the actual price for a particular agent.  So importantly, one side or the other of a transaction can end up with greater "surplus".  Again, this comes down to an informational game.  This concept of surplus can be useful in understanding labor market negotiations, regulations like minimum wage, and even business practices like coupons.

If prices only ever reflected the costs for one side, then there would be no discretion or strategy to be had.  While labor regulations like minimum wage could potentially lead to less clearing of the market, if this "bidding surplus" gets stacked toward one side or the other, then we have problems.

In particular, with labor markets, you will have some distribution of income and wages.  This tends to become self-reproducing.  People who have lower wages get used to being more frugal, and they generally will take on fewer costs, and also demand a lower wage.  But what could begin as a mere random chance outcome, can get reinforced over time, and embed itself into further inequality.  Labor regulations like minimum wage are not an interference with bidding processes, they are merely a collective act of bidding, which prevents the labor market from diverging too widely.  If we keep track of unemployment, and have other policies to support full employment, then the potential negative employment effects of minimum wage are avoidable, and this "bidding divergence" can be dealt with.

But I think where supply and demand fails us as a conceptual model, is when we use it "post-hoc" to justify any outcome as merely an expression of market determinism.  Markets can be assessed both in terms of the integrity of market processes and the quality of outcomes, how balanced the "surplus" is on one side or the other of a bidding process.  Importantly, not every market needs balanced surplus, or even needs to clear at all.  Labor markets are particularly an issue, as labor is perishable, non-uniform, and the primary way people secure necessary income.  If a labor market doesn't operate well(fair in terms of both process and outcomes), that is much more problematic than if beany babies or dogecoins are manipulated.

Finally, supply and demand often leads us to think that costs scale smoothly, and monotonically.  The common expression is that "supply slopes downward", but anyone looking to start a business will be able to tell you that a lower ask for a service does not always inspire confidence.  Giffen goods and Veblen goods may been seen as "exceptions" to the rule.  But maybe these "rules" or rather rough heuristics about price, are not so helpful.

In particular, the concern with running a business is often dealing with sales versus fixed costs.  So not only may demand slope downward, but supply may slope downward as well.  The first instance produced of a product is what we call an "invention" and on a unit basis, it is often the most difficult and costly to produce.

If supply and demand both slope downward, then we have a scenario very unlikely the curves that intersect at one point, where what is possible may be dictated by simple social momentum.  What's stopping us from producing solar powered highly efficient cars?  Often, these are issues of consumer expectations, rather than a neutral cost benefit analysis.

In general, we should view markets as an adversarial and collaborative process which both negotiates allocations of scarce resources, and coordinates cooperation in development execution, and even bidding.  A group of consumers who work together can be more powerful and influential than they would be acting alone, and could potentially get companies to offer products they might not have considered.

Summary

Supply and demand is a useful tool, but it can be misused on a mathematical level, a theory level, and a concept level.  In particular, I see seven potential ways in which supply and demand is often applied in a misleading way.

 1. Ignoring feedback processes and the uncertainty based bidding practices of traders/speculators.

 2. Assuming smooth monotonic price responses

 3. Supposing that price is the dominant factor in business operations. Assuming that markets clear if the price is right.

 4. Using post-hoc reasoning to justify a particular market outcome as desireable, inevitable, or fair.

 5. Ignoring theoretically optimal allocations.

 6. Conflating or oversimplifying political issues, such as ownership and distribution of resources, from issues of costs and engineering.

 7.  Treating the supply curve and demand curve as requiring similar analysis, when each are typically driven by very different processes.  Supply curves are often easier to analyze objectively, as they may be driven by costs or ownership distribution.

Optimality

Optimality is a complex issue because people want different things, and while pareto optimality may sound like an escape hatch, it is rare indeed when our decisions don't involve negative tradeoffs.

As I mentioned before, I feel it's important to study mathematical tools like linear programming, and more general mathematical optimization techniques, because they provide a way to evaluate resource allocation effectiveness independent of the political mechanisms involved, like markets and regulation.

Similarly, many tools from computer science, provide a way to evaluate resource allocation problems, independent of political issues.  For example, simple optimization problems such as the knapsack problem are known be be in a computational complexity class called "NP complete" which suggests that simple search.

Asymptotic analysis helps us understand how costs scale from small problems to large problems, and how this is not always linear or even predictable. Sometimes worst case, average case, and best case performance can vary widely.  In contrast with assuming smooth supply curves, computational theory allows us to get specific about the absolute limits of optimality, and how costs scale as you near a particular limitation.