I'll flip this around and instead elucidate when it does apply if that's alright.
The supply and demand model only applies to analysis of microeconomic systems with perfect competition, namely of systems that lack monopoly and monopsony; that have perfect information; that lack an ability of a single economic actor to affect the price in the market; that lack externalities[0]; that lack transaction costs; that lack unknowable probability distributions of outcomes; that lack economic actors that change their behaviors based on what they expect other actors to do; and that lack non-market rules or structures that restrict prices, quantities, or entry. These assumptions are all-or-nothing, and there are a lot of them. It's also an equilibrium model, and we aren't guaranteed to have equilibrium.
For macroeconomic analysis (i.e., analysis of multiple sectors of an economy or multiple economies), it's problematic because of a theorem from the 1970's due to Sonnenschein, Mantel, and Debreu. I've stated it elsewhere, but it also has the implication that the behaviors of the actors in a microeconomic system do not carry over to the macroeconomic context. This tends to violate many of the assumptions above. (Though, if it isn't obvious how or why, please ask.) Furthermore, in the macro context, I don't think we'll ever have equilibrium; at least, I'm having a hard time thinking of an example when it exists.
Let's bring in a physics analogy. While we can look at mechanics through the lens of elementary algebra, those tools really only give us the ability to look at static force-balance problems for springs and pulleys in equilibrium. To model such a system more closely to how it behaves in the universe, with damping, inertia, and feedback (and whatever other time-dependent behaviors I may be forgetting), we are best served moving to differential equations.
Economics is no different; the supply and demand model is a lot like Hooke's law in its simplicity and its applicability. (In fact, in my modeling, I very heavily used numerical solutions of PDEs.)
--
0: When the retail price of kerosene tanked by 70% between 1860 and 1880, this resulted in millions more kerosene lamps and lamps broadly being lit for longer periods of time. This had externalities in the form of more soot and particulate matter (and probably also respiratory illness, but I couldn't seem to find anything with observations from the time period).
> The supply and demand model only applies to analysis of microeconomic systems with perfect competition
I'm going to stop right there. Perfect competition is unnecessary. Part of perfect competition is perfect information.
No transactions have perfect information. There is always "risk" involved. And risk contributes to determining the price of something. Ever wonder why used cars lose a third of their value when you drive a new one off the lot? It's because of the risk that a person might be selling that almost-new car with a hidden expensive fault in it.
Also, risky investments come with higher returns. Safer investments come with lower returns. The lack of perfect information gets a dollar value for the amount of imperfection.
Risk is absolutely part of the free market, and absolutely a part of the Law of Supply and Demand.
It is definitionally a precondition to applying the supply and demand model for analysis.
> Part of perfect competition is perfect information.
This is mostly correct. It's actually a biconditional: perfect competition iff perfect information.
> No transactions have perfect information.
This is correct.
> There is always "risk" involved. And risk contributes to determining the price of something.
Risk (due to Frank Knight, sometimes called Knightian risk) is a term of art that refers to known probability distributions of market outcomes. My professional training is literally to build pricing models against these.
> Ever wonder why used cars lose a third of their value when you drive a new one off the lot? It's because of the risk that a person might be selling that almost-new car with a hidden expensive fault in it.
No, I can't say I've ever wondered this; I've not been around a lot of new cars. (See also: the whole Mennonite thing.) I know that it's because of the term of art that we call uncertainty (again, due to Frank Knight, sometimes called Knightian uncertainty), which refers to those unknown probability distributions of outcomes that I listed as a precondition to applying this model for analysis. I can't price against those.
A used car market is what we call an Akerlof market, which is a market with a feedback system that has a whole lot of pathologies linked to information asymmetry. This also happens to be the textbook example of adverse selection, which is itself an example of how the supply and demand model fails to predict the market due to that information asymmetry.
> Risk is … absolutely a part …
This is correct with the words written here, but I'll caution that risk as a term of art has been misused.
Ironically you are making his point. Countries that have stronger respect for property rights dump way less or no garbage in the oceans as the countries where there is no such respect.
I'm curious. When does it not apply?