Supply and Demand Usually Works: A Very Long List of 50 Cases
Since you might want them...
Are there really laws in economics? I previously wrote in agreement with John T. Roberts that there aren’t laws, but there are good generalizations we can make. Supply and demand is the best one.
But oftentimes, I want to claim that the model works in real life, and I don’t have a good way to show this. After all, it’s very hard to make the case that something holds in general when there are so many different cases to check. Does it work in the markets for apples, cars, homes, and haircuts? How about in the 1950s, as well as the 1750s and today?
Here is a very long list of cases explained by the model. If you know it well enough, it should be clear how they’re explained and predicted by it. For example, in the first case with Minneapolis and other cities, we expect demand growth to be similar across Midwestern cities, and if Minneapolis is building more, that suggests the supply curve is more elastic, allowing quantity to grow while rents don’t rise as much. But if you somehow don’t know the model well, I strongly recommend this fun series of videos.
The Michael Burry example towards the end is my favorite. It shows how supply and demand can be used to predict prices, not just explain them, even if the predictions feel obvious. Other people might have predicted the changes that occurred in the other examples (would it have been so hard to predict if you had been the one to find the oil in Alaska?). I don’t want to spend time tracking them down, because it’s a lot of work. Many examples come from the same source, so if no source is given, click the previous link.
Here’s the list:
Upzoning in Auckland, which allows more housing units to be built in one place, caused rents to stabilize after adjusting for inflation and eliminated the gap in rents with Wellington.
Congestion pricing was introduced in lower Manhattan, charging anyone driving in the zone. This reduced the quantity of cars there and reduced the average travel time.
More men per woman significantly increased the likelihood that a woman was married in the US.
The pandemic simultaneously disrupted supply chains and increased demand for computers due to the rise of at-home work. The price of semiconductors began to rise in late 2020 after a long decline beginning in the 1990s.
Similarly, the price of used cars suddenly began to rise in 2020 because the chips necessary to produce new cars couldn’t be made in sufficient quantities.
The prices of ice, generators, and chainsaws rise a lot after a hurricane. (Cowen and Tabarrok’s Principles, page 120.)
A price control for oil was introduced by Richard Nixon and eliminated by Ronald Reagan, ending the shortage immediately. (Ibid., 142.)
OPEC deliberately reduced the supply of oil in the 1970s, resulting in a price spike and high inflation around the world. (You can just Google this one. This implicitly provides lots of other examples you can check because oil is an important input in the production of many things, like electricity.)1
Rent control reduces the maximum price that housing suppliers can charge. This fairly consistently creates a shortage or some other adverse consequence, depending on the design of the policy.2
There’s strong demand for Uber rides after midnight on New Year’s Eve. When Uber’s surge pricing system temporarily broke down, wait times more than doubled, and the completion rate fell from 100% to below 20%.
Mansa Musa, king of the Mali Empire, donated gold all over Africa, depressing its price across the continent. It fell by more than 12% in Egypt and did not return to its previous price for at least twelve years.
Every year, various companies cut their prices for Black Friday sales. This results in a huge increase in sales figures.3
The discovery of oil in Alaska in 1968 and the construction of an oil pipeline there dramatically increased demand for labor. Both employment and wages doubled in Alaska. When the boom cooled off, wages fell, and employment growth returned to its long-run trend. (Borjas’ Labor Economics, page 7.)
The pandemic led to a huge increase in demand for lumber, since people were staying at home and wanted to do home improvement.4 This was followed by a huge surge in lumber prices.
The typical elasticity of the minimum wage with respect to employment is small, but negative.
The typical effect of immigration on employment is zero.5
I quickly discovered that ChatGPT is good at generating additional examples. And I, being a human with internet access, am good at finding sources to check to make sure they’re right. Pretty good combination, right? Not all of the following examples are AI-generated, but something like 90% of them are.
Bad coffee bean harvests in Brazil. They faced a severe drought and heatwave in the second half of 2024, leading to higher prices. (Brazil produces about a third of all coffee.)
Canada imposes various restrictions on imported dairy. Dairy costs ~29% more there, according to a free market think tank, and random Redditors living in Windsor (right across the border from Detroit) report the same difference.
The US government subsidizes ethanol, which is made with corn. This drives up the price of corn.
There appears to be seasonality in the price of pumpkin, as people buy more during Halloween. (Data is scant.)
Adverse weather and crop conditions limited Ivory Coast cocoa production. Similar events in other cocoa-producing countries have triggered a massive spike in cocoa prices.
Usually, new devices are launched at a price that appears to be below the market-clearing price, as they are quickly sold off in stores. Evidence for this can be seen in high resale prices charged by scalpers.6
Sales of candy before Halloween tend to lead to a surplus after it’s over, forcing producers to cut prices to unusually low levels.7
The death of an author should draw attention to them and increase demand for their books. In practice, this is exactly what happens, with the probability of getting on the bestseller list more than doubling on death.
Russia’s invasion of Ukraine in February 2022 disrupted wheat production there, resulting in a large spike in the global price of wheat at exactly the same time.
Targeted had a “bloated inventory” of TVs after the decline of pandemic-era TV demand, forcing it to mark down prices.8
The reverse problem occurred with toilet paper during the pandemic, resulting in shortages and higher prices.
Climate change appears to be increasing the rate and intensity of extreme weather events, damaging properties in places like Florida, Louisiana, and Texas. This drives up the price of insurance as supply shifts backward. Other areas are experiencing strong, but lesser increases due to higher prices for construction materials and skilled labor used to fix homes.
San Francisco imposed rent control on buildings constructed before 1980. With the new price below the minimum price at which landlords were willing to maintain the buildings, they converted these buildings to condos and redeveloped them to exempt them from rent control.9
Cosmetologists and morticians have similar skillsets, but morticians are paid more.
Germany introduced smoking bans in restaurants. Coincidentally, wages fell for hospitality workers when these bans took effect. (Labor supply shifts forward when working conditions become nicer.)
Interest in road trips tends to spike around Memorial Day, and gas prices tend to go up around that time as well.
There is similar seasonality in the price of rental cars, though you have to be careful to see it in this graph.
New York City severely restrained the supply of taxis by providing a limited number of taxi medallions. As the population grew and demand grew more and more for transport in the city, the price of a medallion grew higher and higher, peaking around 2014 at over $1 million. Prices crashed as soon as demand collapsed once ridesharing, a substitute, was introduced.
The supply of parking spaces is highly inelastic. So as populations grow in major cities, the price of parking rises a lot.
My twin sister and I went to Disneyland California in March. When I was searching for a hotel, prices tended to be higher near the park.
US Airlines [sic] were part of a government-enforced cartel until deregulation in 1978. After this occurred, prices fell dramatically, albeit unevenly.
The 2008 crash was triggered by the proliferation of mortgage-backed securities, packages of many loans that can be bought and sold. Their value was dependent on creditors—i.e., homeowners—paying back the loans they received, but many were far less likely to do so than MBS owners knew. These mortgages were thus “subprime”. Michael Burry discovered this before the crash, and shorted the market before everyone else figured it out, making lots of money from his foreknowledge of the change in demand—and what this does to prices.
The supply of doctors has been deliberately constrained by the American Medical Association and its lobbyists through licensure rules and residency caps. This raises wages for doctors, with American doctors paid more than doctors anywhere else on Earth with the exception of Switzerland.
Licensed workers in general earn more, even after controlling for observable characteristics like gender and years of education. (This form of analysis annoys me because it suffers from an endogeneity problem: you shouldn’t control for an outcome variable like education. One way licensing boosts wages is by requiring more education, so this is kind of like saying “the level of pressure at the bottom of the ocean isn’t so bad once you control for depth”. Anyway, this probably raises quality, but part of the problem is that you’re forcing people to pay for more quality even if they don’t want to. Some people are fine with a middle-of-the-road cosmetologist.)
The cost of providing an internet domain is the same regardless of the name. But demand should vary depending on the length and desirability of the name. “jackwhitcomb.com” can be purchased for $22/year on GoDaddy, while “artificialize.io” is $78/year and “artificialize.com” is $16,999.
Lab-grown diamonds (obviously) have greater supply and lower demand. They cost 60 to 85% less.
The instrumental variables approach to identifying supply and demand curves was first applied in 1926 by either Philip G. Wright or his son, Sewall. The elasticities do not seem to appear in that citation, but doubtless they conformed to expectations, otherwise the author would have immediately noted that they didn’t.
And of course, we have the experiments of Vernon Smith, where the predictions of the competitive model were almost perfect. This was replicated thousands of times.
A potential problem pointed out by Brian Albrecht is that seemingly anything verifies supply and demand. If the price and quantity change in some way, you can always tell a story about what happened.
I don’t think this is too bad a problem. Looking back at the earlier examples, imagine if instead, Minneapolis had the highest rent growth, egg prices fell, and traffic in Manhattan went up. Would we still be able to say the model of supply and demand works?
The answer is that it depends. For the Minneapolis case, we could say that the variation is explained by demand rather than supply: “Minneapolis had the highest rent growth and construction rate because demand increased most strongly there.” But the model wouldn’t work if we checked and found that demand growth really was similar across Midwestern cities, and they each had the same supply elasticity. In truth, this case is explained by the model and suggests it, but it doesn’t imply that the model is true.
A similar problem seems to arise with the egg price case. If prices had fallen, we could just say “Clearly, demand for eggs went down more than supply” and call it a day. But this feels much less plausible than the alternative, because we have no reason to expect such a strong change in the demand for eggs. (When was the last time you heard of an egg craze, anyway?) That’s why it strongly suggests the model works.
The congestion pricing case is the most obvious verification of the law of demand. There’s no other explanation that works. We might want to say that traffic fell because fewer people wanted to be in downtown Manhattan, not because of the introduction of a price on traffic, but there’s absolutely nothing to suggest demand changed like that, and at the same time as the price was introduced, just by coincidence. We might want to say that the supply of road space increased significantly, but that’s physically impossible and obviously didn’t happen. The law of demand is true: if the price rises, fewer people want to buy something.
Iran’s market for kidneys provides an analogous case for the law of supply. You might try to counter that Iranians only give away so many kidneys because of their unique culture—after all, they’re one of the few Shia muslim-majority countries in the world. But neighboring Iraq and Lebanon don’t have a market for kidneys, and despite their Shia majority, they have a kidney shortage, like every other country on Earth with the exception of Iran. Institutions matter.
How confident should we be in the model?
Whenever we look at evidence, we’re essentially drawing from an incredibly large pool of events that could point towards or away from a conclusion. We want to know how confident we can be in the model if we’ve looked at an unbiased draw of 50 events from this pool. (Really, I included more, but the math is qualitatively the same.)
What we’re trying to do is distinguish between two different pools of evidence we might be in: the pool where supply and demand (SnD) is true, and the one where it isn’t. In the SnD is true pool, 90% of events conform to the model. In the SnD isn’t true pool, 50% of events conform to the model, so its predictions are no better than guessing. It’s obvious that in the first case, the probability of drawing 50 conforming events in a row is 0.9^50 = ~0.5%, while in the second, the probability is 0.5^50 = 0.000000000000088817842%. So it seems that if we have this much evidence, it’s much more likely that supply and demand is true.
But what we really want is the actual probability that the model is true, and we can get that probability with Bayes’ theorem. (Try this post for an introduction, but you don’t really need it if you trust that it works.) For our prior, let’s say we’re a skeptic who thinks the probability that SnD is true is 0.1. In our first case, you are exposed to one observation, a bit of evidence, that conforms to supply and demand.
Acting rationally, we’ve changed from being 90% confident SnD isn’t true to ~83.33% confident it isn’t true. What about 50 observations?
Under our assumptions, this much evidence should make you 99.99% confident that supply and demand is true, even if you were a skeptic beforehand.
But there’s a problem: an economics fan gathered all the evidence! Can you be confident that I would even show you a counterexample if it existed? Probably not. I stuffed those under the couch cushions when you weren’t looking. And if neither the pro side nor the con side on this issue will give you an unbiased sample of evidence from the whole picture, neither one is informative.
I don’t think I can solve this problem. But I can give you a counterexample: “Supply Constraints Do Not Explain House Price and Quantity Growth Across U.S. Cities” by Louie et al. Contra what almost all economists believe, “from 2000 to 2020, we find that higher income growth predicts the same growth in house prices, housing quantity, and population regardless of a city’s estimated housing supply elasticity [i.e., how much housing supply can grow in response to growing demand].”
I could go into the multiple academic responses this paper received and the serious problems it has, but here is my honest reaction to this information:
If you’re really skeptical and you want to get an unbiased picture of whether supply and demand describe how prices and quantities are determined, I simply think you should look around.
Clip taken from College Humor. I’d also like to thank Twitter/X users Norvid, Sneedle, and Yah5us for their help finding examples. The latter has an economics blog.
This is more macroeconomics than microeconomics, but it still points toward supply and demand working (why would this happen if the law of demand wasn’t true?)
Talked about in EIAS pt 3. Also talked about in Cowen and Tabarrok’s Principles, page 143.
Couldn’t find data on this, but, Christ, do I really need to put in the effort for this one?
Just check the search interest: https://trends.google.com/trends/explore?date=all&geo=US&q=Home%20improvement&hl=en
This one is always counterintuitive to people. But if you add people to an economy, labor demand will go up by about as much as labor supply.
The situation most people imagine is something like a bunch of plumbers immigrating to a city. That would certainly increase the supply of plumbers more than demand—they can do their own plumbing, obviously, and even if they had to pay someone, just one plumber could service all the other plumbers. Immigration is usually more like when people have babies.
https://www.reddit.com/r/badeconomics/comments/1i5vd8c/im_here_to_preach_to_the_choir_mass_deportation/
This link eventually won’t work, but you can see this happening right now with the Nintendo Switch 2, which is sold in stores at an MSRP of $450: https://www.ebay.com/sch/i.html?_nkw=Nintendo+Switch+2&LH_Complete=1&LH_Sold=1&LH_ItemCondition=1000
Couldn’t find a broader source, but there is this lady’s anecdote.
Anne D’Innocenzio, “Facing Huge Inventory, Target Cuts Vendor Orders, Prices,” Associated Press (June 7, 2022)
HT: Professor Daniel Lin.
The Effects of Rent Control Expansion on Tenants, Landlords, and Inequality:
Evidence From San Francisco,” American Economic Review (September 2019)