(And an IS-LM model by itself is probably sufficient to keep us from goin down the MMT "maybe lower interest rates will cause lower inflation" rabbit hole?)
TBF, there's a sensible core to MMT, namely Lerner-style functional finance, which is essentially 1940s Keynesianism. But MMT advocates are never willing to disappoint their fans, who see it as providing money for nothing.
As far as I can tell, I don't think there's anything that's true in MMT that you can't glean from a simple IS-LM model. Maybe I haven't bothered to go far enough down the MMT rabbit hole, but I also don't particularly want to?
Yeah. Pretending that MMT has anything both new and true is pure politics. They're very good at shitpoasting for clicks, so Keynesians sometimes value their politicking and slick public appeal. But they're ultimately just charlatans, so better in the long run to just throw them off the bus.
As I think Brad DeLong pithily put it, nothing that they say that’s true is interesting, and nothing that’s interesting is true.
They are exceptionally good at being annoying though. Like Rohan Grey was a year behind me in law school and had a reputation for being very very annoying but not very bright. And now I think sells himself as an economist, which he… very much isn’t.
Do any modern macro economists ever think about Georgism? I’m just in this journey of discovery and it seems fundamentally so necessary. Almost too easy and obvious and I’ve been seeking counter arguments.
I loved this and how it lays out the issues at play. Now can you do the Laffer Curve so I have something to shut up the next person who tries to cite it?
People have been trying to replicate Shalizi and McDonald conclusions. It seems that their results won’t replicate ( https://twitter.com/joshuabrault3/status/1588861757027938305?s=46&t=YTJt3sPHqOS77LYVNMn8Mw) and might be caused by a coding mistake (they decided to code everything from scratch in R). In fact standard packages like Dynare do recover true estimates pretty well.
Also, the similar exercise with opposite results was done by Iskrev (2010), whom CS does not cite (they also do not cite Canova and Sala and are in general dismissive of the identification in DSGE literature, for reasons I cannot understand).
All that time DSGE people in Econotwitter are writing humble and engaging threads like this:
CS decided not to engage with this critique, unfortunately.
Overall DSGE has a number of problems, but I do not think that strangely coded paper that does not engage with the lit and reaches conclusions opposite to other peer reviewed results moves us to the right direction
That's interesting! I can't say anything about the replication, only that Shalizi is a good guy and they should email him and try to work out what's going on.
But in terms of permuting the series and fitting the model to the nonsense data, that is pretty solidly grounded and is going to turn a lot of heads. Basically it's a null result for finding the kinds of patterns that could be interpreted as structural. A null result doesn't mean they don't exist, but at the very least it means that this data is too noisy for them to shine through.
As Prof Boldea rightly noted in her thread this is not a rigorous check, and when comparing likelihoods instead of MSEs, the likelihood of the original model in the paper is higher than all of the other likelihoods obtained by permutation.
I once again want to note how strange it is that CS claims he is first one to run model on simulated data, and if you know just a bit about modern macro you should know that macro people check models on simulated data ALL THE TIME, and it works quite well and dynare code is always available (and why this work took 14 years?)
Will see what will come out of it of course, but so far it looks to me like strong claims of CS are simply not supported by their analysis
I disagree with Prof Boldea that it's not a rigorous check; any physical model that you run through this kind of check, for example, won't allow you to permute the series without a severe dropoff in model fit. And I do believe the likelihoods they get with some permutations are higher than the original.
Really helpful summary for me. Explains why I detest the entire macro-project beyond really basic things like gross national income trends. I am not surprised that microeconomic models have been much more useful and easily validated. They reduce the chance of not knowing the relevant variables. By reducing the scale of the investigation.
So are many other economists. I prefer the economists that are good at research and reality. Claudia Sahm is an example of this. Unlike Sahm, Summers name will not be attached to any empirically based model that has use in the real world. For the record, he treated Sahm horribly when he was her boss. As well as the other realists who also did research and warned of the danger of a mountain of derivatives, e.g., Brooksley Born, Shelia Bear. Summers and his crowd had nothing but scorn for these researchers. If he’s so good at research, how is it he couldn’t see a financial freight train coming down the tracks?
The idea that currency-issuing governments can issue debt free money, without taxing or borrowing from the private sector, on behalf of public sector spending - who cares if it's a theory or not.
The important thing is the necessity for the total of public and private spending to not exceed the nation's productive capacity., to avoid inflation.
As for Turkey: in MMT a ZIRP is recommended, along with a JG to act as a price anchor.
Nothing to do with Turkey.
And re "Biden's inflation": in a pandemic, money shouldn't be given to people who don't need it....
"That doesn’t mean macro is a “science” yet — I’d say it’s more a proto-science, maybe a bit like medicine in the 1600s. " Maybe more like political science, still highly an art with some science.
Whatever happened to good old Keynesean Economics. In bad times, one spent and built up debt. In good times, one increased taxes, controlled spending and paid off debt?
Noah, in becoming a science, aren't the first methods, methods of measurement? (i.e., going from qualities of motion, getting to quantities; going from qualities of force to quantities of force; going from feelings of temperature to repeatable effects of temperature).
The desire for the micro-to-macro transmission to be elegantly tractable seems to be where the importance of the above gets stripped out... premature mathematization.
EXACTLY! In science we measure things. Look up the definition of one metre : it is precise and standardised. So what do we use to measure things in economics? Money. Now look up the definition of money!!!!!!!
So the basic questions are:
What is money?
How much of it is there?
How is it created or destroyed?
If you are going to correct monetary measures for inflation then what is inflation?
Economics has no hope of producing anything scientific unless it can agree on sensible definitions of the above.
True to form, make sure you cite Larry Summers in any post in re economics. But maybe this is apt for the universe of economic modeling; because in the real world of economic policy-making, few economists have done as much damage as Larry Summers. I get the feeling that writing columns about economics is a poker game and Larry Summers is the ante. Personally, I’d rather read Carlota Perez or Ray Dalio on the 50-year business cycle. Keep up the great work but consider going on a serious Larry Summers diet. Surely, you can find other economists with the same ideas.
I'm a little puzzled by this post. You seem to be praising empirical macro people who go out and measure a bunch of micro-elasticities, while at the same time you're complaining about general equilibrium modeling.
As a macroeconomist, what's the point of measuring the micro-elasticities if you're not going to plug them into a GE model? In fact, measuring micro-elasticities to plug them into medium-to-large-scale GE models is precisely what Nakamura and Steinsson advocate.
If you think small-scale DSGE models are dubious, then the Nakamura-Steinsson agenda is completely ludicrous! The pretense of knowledge embodied in big DSGE models is an order of magnitude greater than the list of ridiculous assumptions in small DSGE models.
Understanding, say, the consumption response to the end of extended unemployment insurance is going to be a valuable thing to know whether or not you eventually managed to get something by plugging it into a DSGE model. If you can plug it in, great, if you can't you still learn something about the economy and about policy.
Right, you learn something about how individuals respond to the end of unemployment insurance. But you don't necessarily learn very much about how that policy affects aggregates that have traditionally been of interest to macroeconomists (say, output and employment).
To do that, you *must* be taking a stance on what's happening in general equilibrium. If you're making a prediction about aggregates, then you're specifying some set of explicit or implicit assumptions about the macroeconomy's behavior. This is true regardless of whether you explicitly write down a model with a million parameters or whether you just say "DSGE models don't make sense, here's my well-informed prediction."
You probably do learn a lot that's useful about how the policy will effect aggregates. Knowing the first order effects that are large is more useful than knowing second and third order effects (and so on) that are smaller.
This is the reason I subscribed to your blog: to try to understand more about how macroeconomists think and why they don't understand stuff. I love your realism and honesty about the state of the field.
My training is in science and biology and I never thought much about money until my late 20s. When I started reading the business pages for the first time I was absolutely stunned to discover that there were vastly differing opinions in the field of economics. If humans are so intelligent and have managed to learn so much about the physical world around them that was created by Who Knows What, then how could they not understand and agree on something created by humans!!
There are a couple of possible explanations. One is that economics is about human interaction and the reason we don't understand it is that we have a very poor understanding of ourselves. The other is that we don't want to understand it. I would go for a mixture of both. What do you think?
The discussions around the science of Covid and the climate demonstrate a vitally important lesson. When we talk about something really important and the conclusion affects large numbers of people, this thing called politics becomes involved. I believe that in fact early discussions about the subject were called political economy.
I don't follow you that closely because I believe that macroeconomics is impossible. You have a massively complex, chaotic system that is self-referential, so the best any of us have are stabs at history and philosophy etc., but never a science.
I guess that's why I've never read your essential humility about all of this, and I applaud it. (As far as I'm concerned, this is the best thing you've written here.)
Go out and preach this to your fellows so they'll get out of the business of trying to affect the world.
IS-LM macro still works well enough. Macro was in a better state in 1958 than it now. https://johnquiggin.com/2013/01/05/the-state-of-macroeconomics-it-all-went-wrong-in-1958/
As for DSGE, you know my view from Zombie Economics.
(And an IS-LM model by itself is probably sufficient to keep us from goin down the MMT "maybe lower interest rates will cause lower inflation" rabbit hole?)
TBF, there's a sensible core to MMT, namely Lerner-style functional finance, which is essentially 1940s Keynesianism. But MMT advocates are never willing to disappoint their fans, who see it as providing money for nothing.
As far as I can tell, I don't think there's anything that's true in MMT that you can't glean from a simple IS-LM model. Maybe I haven't bothered to go far enough down the MMT rabbit hole, but I also don't particularly want to?
Yeah. Pretending that MMT has anything both new and true is pure politics. They're very good at shitpoasting for clicks, so Keynesians sometimes value their politicking and slick public appeal. But they're ultimately just charlatans, so better in the long run to just throw them off the bus.
As I think Brad DeLong pithily put it, nothing that they say that’s true is interesting, and nothing that’s interesting is true.
They are exceptionally good at being annoying though. Like Rohan Grey was a year behind me in law school and had a reputation for being very very annoying but not very bright. And now I think sells himself as an economist, which he… very much isn’t.
With the one exception of the empirical turn. Bringing causal identification to macro has improved macro (as opposed to DSGE modelling).
Keynes is the father of macro. Going back to micro foundations and exchange never breaks down was a regression towards the pre Keynesian era.
Do any modern macro economists ever think about Georgism? I’m just in this journey of discovery and it seems fundamentally so necessary. Almost too easy and obvious and I’ve been seeking counter arguments.
They definitely don't think about Georgism, at least not during work hours
Why not? Too socialist? It seems reasonable to me to tax Rentiers.
I loved this and how it lays out the issues at play. Now can you do the Laffer Curve so I have something to shut up the next person who tries to cite it?
Haha ok!
I look to Harry Seldon.
Which I probably read a few years before that real economist. 🤗
On the other hand, micro isn't in the clear. In the absence of full employment, general equilibrium doesn't work, and that undermines lots of micro. You get a mention in this post https://johnquiggin.com/2013/10/25/the-macro-foundations-of-micro-crossposted-at-crooked-timber/
People have been trying to replicate Shalizi and McDonald conclusions. It seems that their results won’t replicate ( https://twitter.com/joshuabrault3/status/1588861757027938305?s=46&t=YTJt3sPHqOS77LYVNMn8Mw) and might be caused by a coding mistake (they decided to code everything from scratch in R). In fact standard packages like Dynare do recover true estimates pretty well.
Also, the similar exercise with opposite results was done by Iskrev (2010), whom CS does not cite (they also do not cite Canova and Sala and are in general dismissive of the identification in DSGE literature, for reasons I cannot understand).
All that time DSGE people in Econotwitter are writing humble and engaging threads like this:
https://twitter.com/otiliaboldea/status/1588900104735719425?s=46&t=YTJt3sPHqOS77LYVNMn8Mw
https://twitter.com/hpfilter/status/1589319404478951424?s=46&t=YTJt3sPHqOS77LYVNMn8Mw
https://twitter.com/javiergc14/status/1588191441913876483?s=46&t=YTJt3sPHqOS77LYVNMn8Mw
CS decided not to engage with this critique, unfortunately.
Overall DSGE has a number of problems, but I do not think that strangely coded paper that does not engage with the lit and reaches conclusions opposite to other peer reviewed results moves us to the right direction
That's interesting! I can't say anything about the replication, only that Shalizi is a good guy and they should email him and try to work out what's going on.
But in terms of permuting the series and fitting the model to the nonsense data, that is pretty solidly grounded and is going to turn a lot of heads. Basically it's a null result for finding the kinds of patterns that could be interpreted as structural. A null result doesn't mean they don't exist, but at the very least it means that this data is too noisy for them to shine through.
As Prof Boldea rightly noted in her thread this is not a rigorous check, and when comparing likelihoods instead of MSEs, the likelihood of the original model in the paper is higher than all of the other likelihoods obtained by permutation.
I once again want to note how strange it is that CS claims he is first one to run model on simulated data, and if you know just a bit about modern macro you should know that macro people check models on simulated data ALL THE TIME, and it works quite well and dynare code is always available (and why this work took 14 years?)
Will see what will come out of it of course, but so far it looks to me like strong claims of CS are simply not supported by their analysis
I disagree with Prof Boldea that it's not a rigorous check; any physical model that you run through this kind of check, for example, won't allow you to permute the series without a severe dropoff in model fit. And I do believe the likelihoods they get with some permutations are higher than the original.
You may disagree. But please check the paper: the likelihood in sample is the highest compared to permuted models. Out of sample it is in top 8%
Really helpful summary for me. Explains why I detest the entire macro-project beyond really basic things like gross national income trends. I am not surprised that microeconomic models have been much more useful and easily validated. They reduce the chance of not knowing the relevant variables. By reducing the scale of the investigation.
Great writeup. The 14 other likes this post has received so far agree with my sentiments.
So are many other economists. I prefer the economists that are good at research and reality. Claudia Sahm is an example of this. Unlike Sahm, Summers name will not be attached to any empirically based model that has use in the real world. For the record, he treated Sahm horribly when he was her boss. As well as the other realists who also did research and warned of the danger of a mountain of derivatives, e.g., Brooksley Born, Shelia Bear. Summers and his crowd had nothing but scorn for these researchers. If he’s so good at research, how is it he couldn’t see a financial freight train coming down the tracks?
The idea that currency-issuing governments can issue debt free money, without taxing or borrowing from the private sector, on behalf of public sector spending - who cares if it's a theory or not.
The important thing is the necessity for the total of public and private spending to not exceed the nation's productive capacity., to avoid inflation.
As for Turkey: in MMT a ZIRP is recommended, along with a JG to act as a price anchor.
Nothing to do with Turkey.
And re "Biden's inflation": in a pandemic, money shouldn't be given to people who don't need it....
"That doesn’t mean macro is a “science” yet — I’d say it’s more a proto-science, maybe a bit like medicine in the 1600s. " Maybe more like political science, still highly an art with some science.
Whatever happened to good old Keynesean Economics. In bad times, one spent and built up debt. In good times, one increased taxes, controlled spending and paid off debt?
Generally probably a good idea?
But little practiced, sigh.
Noah, in becoming a science, aren't the first methods, methods of measurement? (i.e., going from qualities of motion, getting to quantities; going from qualities of force to quantities of force; going from feelings of temperature to repeatable effects of temperature).
The desire for the micro-to-macro transmission to be elegantly tractable seems to be where the importance of the above gets stripped out... premature mathematization.
EXACTLY! In science we measure things. Look up the definition of one metre : it is precise and standardised. So what do we use to measure things in economics? Money. Now look up the definition of money!!!!!!!
So the basic questions are:
What is money?
How much of it is there?
How is it created or destroyed?
If you are going to correct monetary measures for inflation then what is inflation?
Economics has no hope of producing anything scientific unless it can agree on sensible definitions of the above.
True to form, make sure you cite Larry Summers in any post in re economics. But maybe this is apt for the universe of economic modeling; because in the real world of economic policy-making, few economists have done as much damage as Larry Summers. I get the feeling that writing columns about economics is a poker game and Larry Summers is the ante. Personally, I’d rather read Carlota Perez or Ray Dalio on the 50-year business cycle. Keep up the great work but consider going on a serious Larry Summers diet. Surely, you can find other economists with the same ideas.
Summers is quite good at research!
I'm a little puzzled by this post. You seem to be praising empirical macro people who go out and measure a bunch of micro-elasticities, while at the same time you're complaining about general equilibrium modeling.
As a macroeconomist, what's the point of measuring the micro-elasticities if you're not going to plug them into a GE model? In fact, measuring micro-elasticities to plug them into medium-to-large-scale GE models is precisely what Nakamura and Steinsson advocate.
If you think small-scale DSGE models are dubious, then the Nakamura-Steinsson agenda is completely ludicrous! The pretense of knowledge embodied in big DSGE models is an order of magnitude greater than the list of ridiculous assumptions in small DSGE models.
Understanding, say, the consumption response to the end of extended unemployment insurance is going to be a valuable thing to know whether or not you eventually managed to get something by plugging it into a DSGE model. If you can plug it in, great, if you can't you still learn something about the economy and about policy.
Right, you learn something about how individuals respond to the end of unemployment insurance. But you don't necessarily learn very much about how that policy affects aggregates that have traditionally been of interest to macroeconomists (say, output and employment).
To do that, you *must* be taking a stance on what's happening in general equilibrium. If you're making a prediction about aggregates, then you're specifying some set of explicit or implicit assumptions about the macroeconomy's behavior. This is true regardless of whether you explicitly write down a model with a million parameters or whether you just say "DSGE models don't make sense, here's my well-informed prediction."
That's right, you can't necessarily aggregate up just by knowing those individual responses. But they're still quite valuable to know!
You probably do learn a lot that's useful about how the policy will effect aggregates. Knowing the first order effects that are large is more useful than knowing second and third order effects (and so on) that are smaller.
This is the reason I subscribed to your blog: to try to understand more about how macroeconomists think and why they don't understand stuff. I love your realism and honesty about the state of the field.
My training is in science and biology and I never thought much about money until my late 20s. When I started reading the business pages for the first time I was absolutely stunned to discover that there were vastly differing opinions in the field of economics. If humans are so intelligent and have managed to learn so much about the physical world around them that was created by Who Knows What, then how could they not understand and agree on something created by humans!!
There are a couple of possible explanations. One is that economics is about human interaction and the reason we don't understand it is that we have a very poor understanding of ourselves. The other is that we don't want to understand it. I would go for a mixture of both. What do you think?
The discussions around the science of Covid and the climate demonstrate a vitally important lesson. When we talk about something really important and the conclusion affects large numbers of people, this thing called politics becomes involved. I believe that in fact early discussions about the subject were called political economy.
I don't follow you that closely because I believe that macroeconomics is impossible. You have a massively complex, chaotic system that is self-referential, so the best any of us have are stabs at history and philosophy etc., but never a science.
I guess that's why I've never read your essential humility about all of this, and I applaud it. (As far as I'm concerned, this is the best thing you've written here.)
Go out and preach this to your fellows so they'll get out of the business of trying to affect the world.
Leave that to poets and warriors.