I was drawn to economics as a young man by the promise of doing good through macro economic policy implemented by government agents following prescriptions of Keynesian macro economic theory. I was lied to by my professors who assured me that the Fed used and expansive mathematical models that accurately predicted macro economic outcomes that result from flicking a switch here, pulling lever there.
Today, I am an old economist who long ago became enamored with Austrian economic thinking and public choice theory, after spending 10 years consulting and conducting contract research for federal agencies in DC. My personal belief, based on 45 years of experience paying attention to the political economy is that macro economic modeling is delusional, mystical, and postulated on nonexistent epiphenomenon.
Econometric modeling has proved to be useless for what I take to be obvious reasons that are roundly ignored by macro economists. The fundamental issue is that there are no economic constants to be estimated using advanced statistical methods. That fact does not deter macro economists, surprisingly enough.
In other words, Arnold, bravo and a HT to you for writing this thoughtful essay.
But….but…..True economic modeling has never been tried! 😊
Seriously, I find modeling useful for estimating a range of potential impacts from policy changes or exogenous events. Because nothing else is held constant in the real world, these estimates don’t make for accurate predictions though are usually directionally accurate, at least.
An issue is that even policy markers seem willfully blind as to inputs. Did Powell ever criticize 6 percent fiscal deficits at full employment while the Fed was in the midst of trying to slow the economy and inflation during the latest tightening cycle?
Modelers pretend that human choice and actions are governed by constants that can be measured. Yes, the signs of relationships can be predicted. Other things unchanged, as price per unit rises, quantity demanded falls. But how much is unknowable, not governed by some knowable, reliable elasticity.
Sometimes, predicting within the correct order of magnitude may be useful. Sensitivity analysis can sometimes be instructive. But I am unaware of any durable, reliable macro economic relationship that sufficiently stable to guide policy decisions exercised by the Fed. If there are any, I am perfectly willing to learn of them.
I'm sympathetic, but fear that skepticism abut explicit modeling just leave one open to implicit modeling that not guaranteed to be better. The fact of the matter is that what the Fed does, what Congress does and does not do has effects. It makes sense to ty to think systematically about those effects
You say that "the fact of the matter is that what the Fed does, what Congress does and does not do has effect." Let's suppose you are right to say this is a fact, although it really isn't. The real question is this: does anyone know what the effects will be? The answer is decidedly no. Explicit modeling is fine, except when the explicit model doesn't work, which is all the time when it comes to macro economic modeling.
I find this discussion amusingly parallel to the thread "The Seminar Continues."
In both cases, the discussion begins by trying to analyze and thereby predict certain people's behavior, but it seems to me that both the process and the set of data being analyzed are so open-ended that the process never reaches a conclusion and thus is never useful. Am I imagining this or are we just wanking?
Depends on the time horizon. For the prehistory of human species as a whole, economically relevant constanst would be things like how much annual sunlight hits earth, pretty consistently at 3.85m exajoules, geographical constraints, and the energy requirements per human to prevent starvation. Bring it down to just historical records and stuff like climate becomes mostly constant. On short time horizons like a few years, the annual production of many commodities isn't constant but is extremely predictable. So I'm confused about what's meant by constants firstly, and Secondly how that impacts modeling since there's lots of engineering design systems for feedback loops and dynamic systems to the extent that there's an entire field called Cybernetics for the topic. Even your household thermostat works because of modeling dynamic systems. I also know people that work at companies like Samsung that build dynamic demand models for products which are in part dependent on macro factors.
I agree with your position but I'm uncertain about the logic.
Your first example is not an economic constant. It's a physical constant. Economics is about human choice. There are no constants in human choice and action. I'm talking about constants in a mathematical modeling context.
Almost constant is not constant. Avogadro's number is a constant. The speed of light is a constant.
Dynamic demand models would involve relating quantity demanded per unit time to price. The relationship would require the use of constants. But there are no constants in economics that are durable, reliable, and measurable.
I've been a modeler all my career. I've built all kinds of models, including dynamic and stochastic models. Models are fabulous, until they are not. And when they are not, it's because there are not constants in human behavior.
Your footnote 2 critique of Sumner's view is spot on.
I've tried to reconcile this with the argument that under Scott's ideas the central bank can control expectations of NGDP, which enables it to control trend growth of realized NGDP but it doesn't control actual, realized NGDP.
So basically the Fed got surprised by weak NGDP in H2 2008, but then allowed NGDP to recover at a slower trend.
That results in the question: if the Fed really control expectations of NGDP and trend growth of it, then why did they allow the slower trend growth in the 2010s? One potential answer is that in the aftermath of a financial crisis, the usual channels of monetary policy were disrupted such that this was not possible. Another is that the Fed simply preferred a slower pace of trend NGDP growth. I don't know if we can know which of these is right.
Macro is claimed to work because, excluding the Great Depression, the frequency and severity of downturn decreased over the 20th and 21st century. Never mind that the economy completely changed, as you say in this excellent post, over that same period. It's entirely possible, and for some reasons that get a bit too deep for a comment, that the Fed actually made recessions worse than they would have been otherwise; just as an example, advances in technology and banking such as global diversification may have been the driver instead.
Physics rests on significant idealizations as well. To speak of the "temperature" of an object that isn't totally isolated from its environment is a useful fiction, for example.
The question is whether the idealizations of macro are close enough to the ballpark that they work ok as models. I'd love to see that question foregrounded when you make this sort of criticism. For an outsider it's hard to tell how much worse things really are than the situation in other well functioning areas of science.
“I take the view that as consumers and businesses we wish to issue risky, long-term liabilities while holding riskless, short-term assets. Banks allow us to do this by taking the reverse position, holding long-term loans as assets and issuing liquid deposits as liabilities.” You’ve expanded the borrow-short-lend-long explanation of banks usefully here. Have you ever expanded on this? After 25 years I’m going back to teaching money and banking this fall and I would like to include this in an introduction.
I would note that we also must consider who owns the banks. Household balance sheets must include bank equity as well as bank deposits. If you think of a bank as making mortgage loans, indirectly the households that have bank stocks are holding mortgage loans. What banks are doing as intermediaries is carving up the mortgage loans into a portion that backs deposits and a portion that goes to shareholders. Banks bring to the process their expertise in mortgage underwriting, the bank's effort to monitor borrowers (Doug Diamond has a paper on "delegated monitoring") and the risk-pooling advantages of having many deposits (so that the chances of everyone wanting their money at the same time are low) and risk-pooling of mortgages (a diversified portfolio of loans).
“Macroeconomics treats the economy as if it were a simple process that can be centrally managed.“ Arnold, if you were President of the Executive Branch and thus Commander in Chief of the armed forces, what macroeconomic measures and tools would you want in order to carry out your duties? How should the Commander in Chief view the economy?
Excellent framing. At IntelliSell, we’re seeing firsthand how macro blind spots—especially around production, trade frictions, and capital formation—impact real-world decisions in manufacturing and supply chains. Models too often assume away the complexity that operators live daily. Subscribed and looking forward to more.
I thought that Sumner’s view was that the downturn was much worse because of Fed policy, not that it was *caused* by Fed policy.
But you are saying that you believe in his counterfactual the results would have been little different then what occurred. Can you explain why you think this?
"If we could re-run history with the Fed making more open-market purchases of Treasury bills in 2006-2008, we would find that the economic results would not be much different."
A related question: if we could re-run history with the Fed using NGDPLT, or a directionally similar policy regime, would the result have been different?
Another interesting post for a layman like myself, and the argument that macroeconomics is historical makes sense. Checking with ChatGPT for reading that I could digest on the interplay of banking and economics produced a short list and a couple of papers. Do you have any recommendations?
I was drawn to economics as a young man by the promise of doing good through macro economic policy implemented by government agents following prescriptions of Keynesian macro economic theory. I was lied to by my professors who assured me that the Fed used and expansive mathematical models that accurately predicted macro economic outcomes that result from flicking a switch here, pulling lever there.
Today, I am an old economist who long ago became enamored with Austrian economic thinking and public choice theory, after spending 10 years consulting and conducting contract research for federal agencies in DC. My personal belief, based on 45 years of experience paying attention to the political economy is that macro economic modeling is delusional, mystical, and postulated on nonexistent epiphenomenon.
Econometric modeling has proved to be useless for what I take to be obvious reasons that are roundly ignored by macro economists. The fundamental issue is that there are no economic constants to be estimated using advanced statistical methods. That fact does not deter macro economists, surprisingly enough.
In other words, Arnold, bravo and a HT to you for writing this thoughtful essay.
But….but…..True economic modeling has never been tried! 😊
Seriously, I find modeling useful for estimating a range of potential impacts from policy changes or exogenous events. Because nothing else is held constant in the real world, these estimates don’t make for accurate predictions though are usually directionally accurate, at least.
An issue is that even policy markers seem willfully blind as to inputs. Did Powell ever criticize 6 percent fiscal deficits at full employment while the Fed was in the midst of trying to slow the economy and inflation during the latest tightening cycle?
Modelers pretend that human choice and actions are governed by constants that can be measured. Yes, the signs of relationships can be predicted. Other things unchanged, as price per unit rises, quantity demanded falls. But how much is unknowable, not governed by some knowable, reliable elasticity.
Sometimes, predicting within the correct order of magnitude may be useful. Sensitivity analysis can sometimes be instructive. But I am unaware of any durable, reliable macro economic relationship that sufficiently stable to guide policy decisions exercised by the Fed. If there are any, I am perfectly willing to learn of them.
Arnold Kling has it right, in my opinion.
I'm sympathetic, but fear that skepticism abut explicit modeling just leave one open to implicit modeling that not guaranteed to be better. The fact of the matter is that what the Fed does, what Congress does and does not do has effects. It makes sense to ty to think systematically about those effects
You say that "the fact of the matter is that what the Fed does, what Congress does and does not do has effect." Let's suppose you are right to say this is a fact, although it really isn't. The real question is this: does anyone know what the effects will be? The answer is decidedly no. Explicit modeling is fine, except when the explicit model doesn't work, which is all the time when it comes to macro economic modeling.
So what’s the alternative. Polling us folks writing Substacks? :)
😂 Believe it or not, we would probably get better results. The future she is uncertain. Predicting human behavior is especially difficult.
Well my opinions on monetary policy are there for the taking if the Fed is interested. :)
I find this discussion amusingly parallel to the thread "The Seminar Continues."
In both cases, the discussion begins by trying to analyze and thereby predict certain people's behavior, but it seems to me that both the process and the set of data being analyzed are so open-ended that the process never reaches a conclusion and thus is never useful. Am I imagining this or are we just wanking?
Are you confident that it is “all the time” as opposed to a large majority of the time that it doesn’t work?
I read this piece as saying it works well enough for long stretches… until it doesn’t.
Which admittedly is the most important point.
But still different from your claim.
Are you aware of some macro economic model that delivers even approximately accurate predictions any time? I am not.
I agree, but can you explain in a bit more detail about how there are no economic constants to be estimated? I didn't understand that sentence.
Tell me some economic constant that you are aware of.
Depends on the time horizon. For the prehistory of human species as a whole, economically relevant constanst would be things like how much annual sunlight hits earth, pretty consistently at 3.85m exajoules, geographical constraints, and the energy requirements per human to prevent starvation. Bring it down to just historical records and stuff like climate becomes mostly constant. On short time horizons like a few years, the annual production of many commodities isn't constant but is extremely predictable. So I'm confused about what's meant by constants firstly, and Secondly how that impacts modeling since there's lots of engineering design systems for feedback loops and dynamic systems to the extent that there's an entire field called Cybernetics for the topic. Even your household thermostat works because of modeling dynamic systems. I also know people that work at companies like Samsung that build dynamic demand models for products which are in part dependent on macro factors.
I agree with your position but I'm uncertain about the logic.
Your first example is not an economic constant. It's a physical constant. Economics is about human choice. There are no constants in human choice and action. I'm talking about constants in a mathematical modeling context.
Almost constant is not constant. Avogadro's number is a constant. The speed of light is a constant.
Dynamic demand models would involve relating quantity demanded per unit time to price. The relationship would require the use of constants. But there are no constants in economics that are durable, reliable, and measurable.
I've been a modeler all my career. I've built all kinds of models, including dynamic and stochastic models. Models are fabulous, until they are not. And when they are not, it's because there are not constants in human behavior.
Your footnote 2 critique of Sumner's view is spot on.
I've tried to reconcile this with the argument that under Scott's ideas the central bank can control expectations of NGDP, which enables it to control trend growth of realized NGDP but it doesn't control actual, realized NGDP.
So basically the Fed got surprised by weak NGDP in H2 2008, but then allowed NGDP to recover at a slower trend.
That results in the question: if the Fed really control expectations of NGDP and trend growth of it, then why did they allow the slower trend growth in the 2010s? One potential answer is that in the aftermath of a financial crisis, the usual channels of monetary policy were disrupted such that this was not possible. Another is that the Fed simply preferred a slower pace of trend NGDP growth. I don't know if we can know which of these is right.
Macro is claimed to work because, excluding the Great Depression, the frequency and severity of downturn decreased over the 20th and 21st century. Never mind that the economy completely changed, as you say in this excellent post, over that same period. It's entirely possible, and for some reasons that get a bit too deep for a comment, that the Fed actually made recessions worse than they would have been otherwise; just as an example, advances in technology and banking such as global diversification may have been the driver instead.
The fact that this argument isn't made more...
Physics rests on significant idealizations as well. To speak of the "temperature" of an object that isn't totally isolated from its environment is a useful fiction, for example.
The question is whether the idealizations of macro are close enough to the ballpark that they work ok as models. I'd love to see that question foregrounded when you make this sort of criticism. For an outsider it's hard to tell how much worse things really are than the situation in other well functioning areas of science.
“I take the view that as consumers and businesses we wish to issue risky, long-term liabilities while holding riskless, short-term assets. Banks allow us to do this by taking the reverse position, holding long-term loans as assets and issuing liquid deposits as liabilities.” You’ve expanded the borrow-short-lend-long explanation of banks usefully here. Have you ever expanded on this? After 25 years I’m going back to teaching money and banking this fall and I would like to include this in an introduction.
I would note that we also must consider who owns the banks. Household balance sheets must include bank equity as well as bank deposits. If you think of a bank as making mortgage loans, indirectly the households that have bank stocks are holding mortgage loans. What banks are doing as intermediaries is carving up the mortgage loans into a portion that backs deposits and a portion that goes to shareholders. Banks bring to the process their expertise in mortgage underwriting, the bank's effort to monitor borrowers (Doug Diamond has a paper on "delegated monitoring") and the risk-pooling advantages of having many deposits (so that the chances of everyone wanting their money at the same time are low) and risk-pooling of mortgages (a diversified portfolio of loans).
Hello. This might be helpful
https://charles72f.substack.com/p/basel-faulty-the-financial-crisis
“Macroeconomics treats the economy as if it were a simple process that can be centrally managed.“ Arnold, if you were President of the Executive Branch and thus Commander in Chief of the armed forces, what macroeconomic measures and tools would you want in order to carry out your duties? How should the Commander in Chief view the economy?
Typo? “ours of work”
Excellent framing. At IntelliSell, we’re seeing firsthand how macro blind spots—especially around production, trade frictions, and capital formation—impact real-world decisions in manufacturing and supply chains. Models too often assume away the complexity that operators live daily. Subscribed and looking forward to more.
We cover these themes weekly here: https://open.substack.com/pub/intellisell
I thought that Sumner’s view was that the downturn was much worse because of Fed policy, not that it was *caused* by Fed policy.
But you are saying that you believe in his counterfactual the results would have been little different then what occurred. Can you explain why you think this?
"If we could re-run history with the Fed making more open-market purchases of Treasury bills in 2006-2008, we would find that the economic results would not be much different."
A related question: if we could re-run history with the Fed using NGDPLT, or a directionally similar policy regime, would the result have been different?
I don't think that the result would have been different. I don't think of the Fed as possessing a gas pedal.
Another interesting post for a layman like myself, and the argument that macroeconomics is historical makes sense. Checking with ChatGPT for reading that I could digest on the interplay of banking and economics produced a short list and a couple of papers. Do you have any recommendations?