> Leontief matrices

These are commonly called IO models in econometrics (for input-output). There are a number of successful IO models. Characterising these as "competitors" to supply-demand curve drawing is silly since they share common themes and complement each other. IO models need lots of data and underperform in situations where the market computes the table for you, i.e. where prices are clear, e.g. for commodities or liquid financial assets.

> IO models need lots of data

To draw supply and demand curves to just find the equilibrium also requires quite a lot of data. And if you aggregate, then it's not even clear what these curves are and what is the equilibrium.

> where the market computes the table for you

I am not sure what you mean by that. How is supply-demand model better? What it gives you in those situations?

Also, I should add: Blatt's IO model (if you want to call it that) is not an econometric (meaning statistical) one. I am not familiar with IO econometric models, so I cannot say how it is different. And it's kinda underdeveloped, too, unfortunately though understandably.

> I am not sure what you mean by that. How is supply-demand model better? What it gives you in those situations?

It yields better predictions with fewer inputs. For simple calculations of optimal pricing or optimal production quantities given input costs, a simple table of volume versus price beats a matrix. IO models become interesting when observing entire industries. (That is why the DoJ uses the HH metric [1], an IO-derived metric, to measure competitiveness.) But they're macroeconomic, not microeconomic, models. Macro has a worse track record than micro, which starts with simple regressions of volume and price that we call supply and demand.

These kinds of matrix models are deeply embedded into the IMF's approach, by the way. Many are critical of the IMF's over reliance on such difficult-to-intuit models.

[1] https://en.m.wikipedia.org/wiki/Herfindahl_index

Again, I think you're talking about statistical model, which I don't object to. I object to supply-demand model as a phenomenological model, as taught in the economics textbooks. What they are doing is not regression.

Basic rules like "if you decrease supply the price will go up" are true in more cases than they aren't. I agree, however, that entry-level students tend to overfit their models. If they never receive later training, they go into the world with a faulty view. Those same people then find those models failing in places we've known, for decades, they fail (in some cases, in a predictable way) and then come back to blame the models.

In particular, one big issue I personally have with supply/demand model is that I don't know how to take two markets and their respective supply/demand curves and equilibria and combine them into supply/demand curves and equilibrium for the common market, under whatever additional assumptions you want to have. And I tried, it was one of the first things I tried to do when learning economics (it's tricky, because the equilibrium has to match).

And I have never seen an economic textbook do that, either. But they all magically jump to aggregate demand and supply, where you have millions of products. They cannot even aggregate two!

Someone mentioned physics here. So that's what you do in physics, you analyze one thing first, then two things, then millions of things in aggregate. With supply and demand model, the 2nd step is simply not done, and for a good reason - the model will fall apart (as far as I can tell) at that point.

That's because of the nontrivial math. It will be found in general equilibrium section toward the end of introductory microeconomics, or at the beginning of a graduate course in macroeconomics.

Page 182 of this excerpt from Acemoglu's Economic Growth discusses aggregation and the possibility of a representative household. http://users.econ.umn.edu/~guvenen/DaronBook1.pdf

Aggregation of firms (and production functions) works quite nicely theoretically. Unfortunately, the same is not true for household / demand aggregation. The Debreu-Mantel-Sonnenschein Theorem aka. the "anything-goes" theorem gives a stark negative result. Utility functions are just too idiosyncratic to aggregate in every possible case.

However macroeconomists still do this for purposes of tractability. We know that sometimes aggregation is possible, if utility functions take certain forms. This is only one of many dimensions modern macroeconomic models are far from realistic. This is no secret to anyone.

This is a big problem, which is enough to kill the model, in my view, but not the only problem, and not the one I alluded to. (And IMHO it affects supply as well.) The problem I alluded to is that how you aggregate supply and demand curves so that the equilibriums (the points where they intersect) match in some reasonable way.

Supply and demand curves are such an unsalvagable mess, in my opinion. Economists should just stop drawing them.

If the supply curves of both markets are expressed as functions that accept Price and return Quantity, then the supply curve of the combined market is the sum of these two functions. I.e. for each price the combined quantity of supply is the sum of the two supplies. Same with demand. Then the equilibrium of the combined market is the intersection of combined supply and combined demand. Am I missing something?

To be honest, my memory is really hazy about this, but the problem is, how do you interpret the aggregate price in aggregate equilibrium, then? Is it what you would expect it to be?

I just did a quick calculation - based on your suggestion - with linear demand/supply curves and what I got as the aggregate price is a weird combination of the two prices. It's not any kind of (weighted) average of the two, as far I can tell. In particular, it seems to me that the aggregate price will differ if, for one of the markets, we change either demand curve or supply curve but keep their intersection at the same point. That's wildly counter-intuitive, I don't know why should aggregate price behave this way.

Maybe I am wrong, and you can make it work. If you do, write it up! It really should be in textbooks, it's embarrassing they don't discuss this case.

And this is a case of two (presumably) completely independent markets with linear curves. It gets much more hazier when you have a relationship between the two markets (for example, one is a substitute for the other).

I don't think you can get an exact aggregate supply/demand model if you are aggregating over non-interchangeable products. To combine the models for different products, you'd need vector-valued price, supply, and demand. That could lead to strange interactions, which wouldn't be captured by lower-dimensional models.

Disclaimer: I've never even taken Econ 101, so I might completely misunderstand something. If that's the case, please downvote and tell me where I'm wrong.

This is kinda the conclusion that Steve Keen's Debunking Economics arrives at too, although amazingly enough, through a different path (avoiding breaking the monotonicity assumptions on supply and demand curves). Basically, any aggregate supply/demand model has a single consumer and a single product. Which makes them pretty much useless, and they just need to go.