Along with stupid people who just so happened to get a ton of money by no merit of their own, but don't realize this.
In Intellectuals and Society, he notes that many "intellectual elite" have been cheerleaders for some of the worst (economic) disasters and human rights abuses of the 20th century. In 2020, we can see this vividly with the coronahysterics causing untold and immeasurable damage to societies with arbitrary and poorly calculated lockdown rules.
Sowell argues that these intellectuals believe their excellence in one field gives them the legitimacy to talk in depth about other fields in which they're not so excellent - perhaps most notably economics, which is Sowell's own field of expertise. In A Vision of the Annointed, he describes how these intellectuals have concocted a fantasy world view which is inconsistent with actual evidence, and the academies work to insulate themselves away from this evidence when it conflicts with their world view. This is true in economics, where the current leading school of thought (MMT) pretends they're not just kicking the can down the road, and ignores the works of the Austrian school.
When he refers to 'intellectuals,' he doesn't just mean anybody intelligent or accredited with degrees. He is referring to people whose results are just ideas, and not products, inventions or scientific breakthroughs. Many "idea people" ultimately face no consequence upon themselves for their poor ideas; they have no skin in the game and are eager to inflict their world view upon society even if there is collateral damage.
A slightly older of his works, A Conflict of Visions, suggests there are two prevailing visions: an unconstrained and a constrained vision. The unconstrained vision is that held by many intellectual elite and largely ignores some aspects of human nature. It believes that people are innately good and well meaning (The just-world fallacy). They don't believe that humans should have constraints because given the opportunity, they will act in their own or others' best interests. The constrained vision is one in which humans believe that man is not innately good and has some bad inside of him. In this vision, it is necessary to have some constraints (laws, fears, etc) so that they act in a manner which is in the best long term interest of themselves or the wider society.
In the study of economics, it is not possible to ignore human nature because all economic activity is the result of people taking actions in their own self-interest. Praxeology studies how humans behave, without consideration for their thoughts or speech, but merely how they act in the real world. It argues that what someone does tells you far more than what a person says they will do, and even what they tell themselves.
The majority of IYIs are for whatever policy makers put forward because the idea of a benevolent and well-meaning State aligns with their world vision.
And of course, "we're just dismantling the evil capitalist patriarchy" is part of that cartoon world vision.
This describes Sowell and his popular works in a nutshell.
On the role of empiricism in economics, I'll try to express why I think it is mostly witchcraft.
Consider a simple scenario where I'm measuring the effect of petrol and oxygen in an engine. I might measure how explosive a combination of petrol and oxygen is, then increase both the quantity of petrol and oxygen and find that the result is more explosive. I can conclude that, increasing the amount of petrol makes the explosion more vigorous, no?
Of course this isn't necessarily the case. Was it the petrol or oxygen increase that caused the increased explosiveness? We can't know by adjusting both at the same time and deducing the resulting change in output is merely down to the change of only one of the inputs. To conduct the scientific method, we must measure one variable at a time, making every effort we can to keep other potential variables constant. This is Science 101.
In an active economy, there are many variables at play. Millions, perhaps billions or more. These variables are also not controlled by us, the observer making the measurements - because they are the actions of many rational, independent humans, making their own personal choices. So what exactly are we concluding if we take some "empirical measurement" about an economy? Essentially, we're measuring whatever we wanted to measure in the first place based on our preconceived notions. In reality, a stab in the dark.
An economy is not something you can kick start and measure some initial conditions, then determine how each variable influences the an economic effect. We can't pause it to take measurements. It is a continuously running system, almost like a living ecosystem, and you can't possibly repeat a measurement because there is never a time when the initial conditions will be the same as a previous time.
To try and measure the effect of some economic activity in a large economy is like me paddling out 1km from sea on a dinghy, and suggesting that the motion from my oars is causing the waves on the beach. An observer at the beech can monitor the tide, and if he sees waves, it's obvious that the waves are caused by my oars, no? This is the absurdity of 'empirical measurement' in an economy. The observer has a tiny fraction of a percent of the information they need to make any meaningful assessment, and moreover, they cannot repeat their measurement in order to see if other factors are at play, because the economy is never in the same state twice.
If we look at an economy like a living ecosystem, we might compare it to some other science which observes this kind of thing. For example, ethologists study the behaviour of living creatures in their natural habitats. Rule #1 of ethology is that the human observer must not interact with the subject being observed, must make their best efforts to not disturb the subjects when installing monitoring devices, and such. If we were to look at the economy the same way, then any observation we want make about the behaviour of rational actors in the economy must exclude any action from ourselves, the observer. However, in modern economics, the observers are almost always attempting to influence the economy in some ways, such as through printing large amounts of new money out of thin air. It is not possible to deduce what the effect of increasing the money supply has on rational actors in an economy, because the act of increasing the money supply influences the decision making of those actors - you have no idea what the actor may have done without your influence. This is like the ethologist attempting to determine the natural behaviours of captive animals in a lab. The economy you are measuring is not a free market, but a market captive to violent aggression. It is not possible to use this to make predictions about a market which is not captive to violent aggression. This is probably why the intellectuals have such a difficult or impossible time even contemplating an economy without intervention.
So how might one go about testing an economic hypothesis where they are in control of the variables? One idea, perhaps, is to use a simulation, or computer model. Modern "science" loves these, they get peer reviewed, but often you can never find the code or reproduce it even if made available. They often contain plenty of magic numbers. I shouldn't need to say it, but a computer model is not the real world, and its inputs and variables are selected by its designer. Models can't tell us about the real world - they can only tell us about the model.
If we selectively take some inputs in some complex system, then it is very easy to cook up an outcome which is consistent with our preconceived hypothesis. Anything else, we merely have statistics, and as we should know - correlation does not imply causation.
You may remember not too long ago, a certain Mr Neil Ferguson produced a fraudulent computer model predicting a grossly exaggerated mortality for COVID-19 spreading through the UK, which was used as the bases for policy-making. The situation turned out a lot less damaging than his expectations (for which he has a history of getting very wrong). However, the policymakers have already implemented their lockdown, and since the mortality from COVID-19 in reality is less than what was predicted by Mr Ferguson's models, we must conclude that lockdowns have had a dramatic effect on curbing the virus!
Such conclusion, of course, is unverifiable and unfalsifiable, because we can never go back to the initial state and trial it again with different conditions. We merely have to accept this pseudo-science as "empirical evidence" that lockdowns and masks were effective. For what we know, the lockdowns could have had zero effect, or could have actually worsened the mortality from COVID-19. The only thing we can compare one country to is that of another country which attempted something different, but even then, there are too many different factors at play - population density, previous exposure, climate, you name it. There is no way you can deduce an accurate truth about the effect of a lockdown on the effects of the virus. It is 100% junk science. Mr Ferguson's model not only tells us nothing - it has only distorted the minds of "scientists," who are now unable to conceive of a world where, prior to lockdown, the virus may have not been as deadly or contagious as was initially guessed.
So if I tell you that lockdowns had almost zero effect, why is my guess any less plausible than that of Mr Ferguson and his pals? Is it because they have credentials? I thought we were interested in science, not appeal to authority? It is my observation that reactions to COVID-19 were a global mass hysteria event.
There are some things we can measure that are a consequence of lockdown rules, but are wrongly labeled a consequence of COVID-19. For example, we can measure approximately how many patients have gone undiagnosed or untreated for more serious illnesses than COVID-19 compared with previous years, such as cancers, and you'd be right to be upset that so many have suffered due to mis-allocation of health resources by the intellectual elite. You might also measure how suicides have increased upon previous years, and note how this is likely a consequence of loneliness due to isolation enforced via violence and threats. These are just the immediately noticeable effects. Poverty in the developing world will increase due to economic decline, after decades of decreasing. The true effects of these lockdowns will last for years, and will be completely immeasurable - but the policymakers will forever blame COVID-19, and never accept responsibility for the real damage which was caused not by a virus, but by their own authoritarian hands. Their intellectual advisers will pretend nothing ever happened, or will continue to double down on their junk science.
Despite any hard evidence on the constructiveness of a lockdown, policymakers feel forced to continue them because they are sufferers of the sunk-cost fallacy. If they have caused so much turmoil with heavy handed overreactions, then their lack of willingness to admit fault causes them to double down and drag out their bad ideas. They can't accept the guilt and responsibility that they were simply wrong, and so they continue to make the situation even worse (or they have another agenda).
Much of what is thrown around as "science" today is nothing of the such. It has more to do with dogmatic religion than the scientific method. Many of its advocates don't even think to question whether a published work could be flawed, but will accept works of "science" as long as they're written and reviewed by members of the same religious faction. The journals are its scripture, and the editors are its pastors. If you conduct the scientific method but are an apostate, the journals will not accept your work into scripture, and your work will not receive any mainstream acknowledgement.
I am a big proponent of the scientific method, and there is no such thing as "The Science (TM)," because science is not a thing - it is a process. If you don't conduct the process correctly, then you don't have science, but scientism.
If you are interested in learning about Praxeology to make up your own mind, there is a good introductory series of short lectures here: https://www.youtube.com/watch?v=QqGqx6fBts0
Praxeology isn't useful at all. The real behaviour of humans cannot be reduced to three or four inalienable axioms. Even moreso, these axioms are by definition unjustifiable, which makes any system based on them unjustifiable as well. Any attempt to reduce human behaviour to an axiomatic system is necessarily a total failure.
The attempt to criticize empiricism by praxeologists is an exercise in obfuscated hypocrisy and a total contradiction. You cannot at once claim that only one's actual behaviour matters, and at the same time arguing that some unknown agent makes that behaviour unreproducible. It's completely absurd at every level.
This is incidentally why Austrian economics fails so hard at actually making any good predictions. It rejects in its totality the scientific method. It's closer to a theology than anything else.
I agree calling it stupid may not be accurate.
Let's say I want to accumulate wealth, and I play the lottery and win, then, by your definition, I'm smart.
I'm not against playing the lottery, but it's objectively proven that it's not a good way to become rich, and a persons smartness has no bearing over whether they actually do win (maybe a negative one, smarter people will gamble less).