I'm glad to hear that Noah Smith, a grad student at Michigan is interested in frictions and financial economics, with a little history thrown in. We would be glad to have him here at Washington University, provided he can pass our prelims.
I taught a course in the fall of last year in advanced monetary economics and macroeconomics. This is for students who have been through the first year of the PhD program, and the students who took it were mostly second-year PhD students. As you can see from the syllabus, this is pretty friction-heavy. It's full of non-neutralities of money, private information, financial contracts, financial intermediation, and financial crisis-related papers, including my own work. Where it's useful, I throw in some history to motivate the ideas. Most of the monetary history I know came my way informally. I pick it up when I need it, and I learned a lot from Bruce Smith, Warren Weber, Art Rolnick, and other people.
Now, some people, including Brad, Paul, and Larry, on the sorry state of macroeconomics and how no one is teaching what is relevant. It's unfortunate that some people who like to spout off are so ill-informed. When the financial crisis hit, there was plenty of off-the-shelf economic theory of financial contracts, financial intermedition, and monetary theory that could be, and was, brought to bear on understanding the important issues. Pay attention, you guys!
You learn it when you need it? How do you know you need it if you don't know it? History is not a tool, it's series of events (and interpretations of those events) about which you can be aware or unaware. If you are unaware then I don't see how you can decide if you need it or not.ReplyDelete
Exactly. History is not a special tool. I learn that maybe it is important to know about some events, so I learn about them. What could be wrong with that?ReplyDelete
Yes, it seems Noah Smith would have liked to have seen some more frictions (and other stuff) in his class. But he reserves his scorn for 'moment matching' as a commonly used standard for checking macro models against data.
guy seems very confused. Why the Hayek bashing? Don't most economist's at least know that up until the 40's the austrian's, and Hayek, were still within the mainstream of the the economic profession. The man was hired LSE for cripes sake.ReplyDelete
Second, if I'm not incorrect Hayek was one of the first people to begin using time series when he empirically studied the BC. He worked at a research institute in Austria for some time studying the BC. Of course he disagreed with others about what kind of predictions and data would be needed to validate his theory, but nevertheless.
I'm graduating with an econ degree soon, and I'm still pretty confused about what most economists think there going to get as empirical "proof" of their theories.
Yes, he certainly seems to be confused. Economics is not physics. I think we pay more attention to, and reward, progress in theory, partly because the empirical work is typically indecisive. In macro especially, there are no clean experiments, there are measurement problems, and the data we have are confounded by endogenous policy and economic agents adapting to those policies.ReplyDelete
That's true. You have to advance theory based on certain economics rules that we take as true to begin with, i.e. marginal utility, profit seeking. And these theories have to incorporate certain assumptions that are never going to be fully true. I mean, if you assume that people act out Ricardian equivalence, then your wrong. If you assume that they act out some other assumption, your wrong too. there's a mix. There are certain rules we take as outside of time and place, and assumptions that are very much not. That's just the way it is. No theory will ever take into account the differences between every individual or business in society.ReplyDelete
I think most's economist's just wish econ was easier, especially macro.
That makes it interesting though. Easy is boring.ReplyDelete
Noah Smith (and Summers and others too) is in the first place very confused about what the first year macro sequence is supposed to be about.ReplyDelete
It is not supposed to teach you how to predict unemployment for the next month. It is not supposed to give you a broad survey of what other people think about important issues. It is supposed to teach you the tools that will allow you to think about the important issues by yourself. If you want to see the details on any specific topic, head to the second year macro sequence!
What is to be gained from reading Kindleberger or Minsky ? We learn that there are financial crises. Big deal. The economists job is to model the forces causing crises, and nest these in forecasting or policy models.ReplyDelete
Arguably the problem is not lack of history knowledge but US-centrism and complacency among macroeconomists. I am sure economists with a focus on international macro like Mendoza were building dsge models of financial crises long before the crisis broke.
Yes, the first-year PhD program is quite intense, and some people can lose sight of why they are learning what they are learning. If you have some background in economics - an undergraduate program, an MA program, some experience working in a central bank or in government, for example - the things you learn in the first year could be a revelation. You can project forward and see that these are powerful tools that you can use to solve some interesting problems. It's easy to see why Noah is confused, as he went into an economics PhD program cold.
Yes, economics is not physics, there are lots of data and measurements problems in macro data, etc. But the 'calibration' approach seems to be a bit like 'having your cake and eating it too.' It's done, as Sargent explains in his 2005 'Macroeconomic Dynamics' interview, because some parts of models are trusted more than others (such that a full likelihood-based econometric approach would 'reject too many good models'). But where\how is the dividing line set between what is trusted and distrusted in models?
Phil Rothman: I thought that calibration stuff was so 1990s. Surely the dsge econometricians have moved on a long way, to estimating models not calibrating. That's what I took away from the Jesus Villaverde work, interviewed in Economic dynamics.ReplyDelete
Another reason why Noel Smith needs to get away from Michigan, to somewhere where they have up-to-date macro econometricians!
Actually, I think a modern quantitative macroeconomist is comfortable with both calibrating and estimating.ReplyDelete
"But where\how is the dividing line set between what is trusted and distrusted in models?"ReplyDelete
There are no rules about this. You have to use your judgement. Another way to think about this is that, by virtue of the fact that we are working with a model, it has to miss some features of the data. If it did not, it would be far to complicated, and therefore would not help us understand anything. Maybe we can replicate the world, but the world is too hard to understand, which is what we are using the model for. Thus, to be useful for policy purposes, the model has to be simple enough that it is wrong on some dimensions that we don't care about.
Anonymous: 'I thought that calibration stuff was so 1990s.' Yes, the past decade saw lots of excellent work done on estimating DSGE models, and that literature continues. But as Steve notes in a subsequent comment, it's not accurate to claim that calibration is a fossil from a foregone era.ReplyDelete
Calibration is theory with numbers. You do theory when you want to explain a mechanism. However, if you are smart enough you can always come up with a theory to explain some qualitative relationship you might be interested in. Calibration increases the burden on the theorist by demanding that you also show that your mechanism can generate relations of the appropriate order of magnitude.ReplyDelete
Whomever has done calibration for real knows that it can be incredibly hard and frustrating. No cake eating here...
The cake eating metaphor was used to represent the dividing line between what's trusted (and whose moments are attempted to be matched) and those which aren't (and whose moments are ignored), i.e., it may strike some that there's potential for sample-selection bias in the setting of this dividing line. No assertion was implied about whether such an exercise is 'hard and frustrating,' because, among other things, this particular aspect of calibration doesn't seem quite germane.
I agree with RV. Now, to be more specific about the estimation/calibration issue, note what happens when estimation is carried to the extreme. Models like Smets/Wouters and Christiano/Eichenbaum/Evans fit the data well. But fitting the data well means throwing in seventeen stochastic processes, and filling the model with fixed-this, fixed-that, adjustment-cost-this, adjustment-cost-that, this decision fixed-one-period-ahead, etc. Ultimately those models are such a mish-mash of mutually inconsistent theory that you might as well have just fit a VAR to the time series, for all you are going to learn from it. I certainly would not use it as an input into a policy decision. Now, the basic RBC model, in spite of its faults, can actually teach you something, which is why the thing is a standard part of the macro PhD core. Students can see how the calibration is done, and they learn something about some standard elements of macro theory. Anyone who is teaching the students properly will get the idea across that this thing is just a building block.ReplyDelete
Exactly. We teach frictionless models in the first year not because we'd be simpletons ignoring tons of interesting stuff out there, but because they are in the core of all more complicated models. The more economics you do, the more you see how important this is.
If there is any failure in Noah Smith's first year Macro education, then it is not in the content of the class, it is in that he failed to get this point.
Can you give some example of "mutually inconsistent theory" in models a la Smets & Wouters or CEE?
I wonder why we spend any time discussing the opinions of a first-year grad student. They know so little it is somewhat amazing they don't fall down more often.ReplyDelete
Start with the decisions made by the firm. Making these mutually consistent would require integrating the firm's decisions about setting prices and wages with its decisions concerning investment and employment. In the model I saw, the investment decisions were actually at the household level, so total divorced from the pricing, and of course with the ad-hoc costs-of-adjustment thrown in.
Another caveat of the Smets Wouters approach is the fact that some of the shocks they introduce are not structural. Structural shocks are those which are (1) invariant with respect to policy interventions and (2) must be interpretable in such a way that we can determine whether they are "good shocks" that policy must accommodate or "bad shocks" that policy should offset. For instance, Smets and Wouters have some type of labor wedge shock (I think they call it wage markup) that is consistent with both a model with taste shocks on leisure and a model with labor-union type frictions. The problem, then, is that depending on which interpretation you give to the shock, you get different welfare consequences (e.g. if taste shocks than it is optimal not to do anything and if labor-union then offset). This point is made clear on a paper by Chari, Kehoe and McGrattan called "New Keynesian Models: Not Yet Useful For Policy Analysis".ReplyDelete
Thanks. I think the bottom line is that these models have evolved to the point where they fit the data in the same sense as the old 1960s/1970s large macroeconometric models (some of which are still in use, for example at the Board), but have the same faults:
Fischer Black in "Exploring General Equilibrium" argued that he could explain the world to his satisfaction by using general equilibrium and saw no reason to add arbitrary assumptions of the type favored by theorists such as Lucas and Prescott. If you add even more arbitrary restrictions like Smets and Wouters you carry the model ever further away from being useful. Black thought that these arbitrary assumptions caused "puzzles" such as the equity premium puzzle that were artifacts of unnecessary restrictions, not reality.ReplyDelete
I hope you did not spend a lot of time poring over Fischer's book. Black was lucky. He was in the right place at the right time, knew how to solve a partial differential equation, and got his name on one of the classic contributions in finance. He was trained in physics and math, but knew little about economics, and I don't think you can say that he actually practiced serious economic research. The book you mention came out of Fischer's touring around economics conferences, while he was a partner at Goldman Sachs. Fischer would sit with his notes spread out in front of him, and spend his time at the conference jotting things down. He never had much to say that was very interesting, I thought.ReplyDelete
General equilibrium on its own has nothing to say. It's the restrictions that make it interesting. That puts discipline on what you are doing.
Actually, as I recall the story he derived the pde but actually didn't know how to solve it. (Very odd that an applied math Phd wouldn't recoginze the heat equation!)ReplyDelete
Scholes' entire contribution was to take the equation over the MIT's math department and ask if anyone know how to solve (which absolutely every one of them would).
Interesting tidbits on Black. I think his view that business cycles are the result of sectoral mismatches resonates. Of course, Long and Plosser had already modeled along those lines, but somehow the inferior product of Prescott won out.ReplyDelete
Yes, one conference where I remember seeing him was a group NBER meeting at Stanford, a good part of which was specifically about sectoral reallocation. This was maybe 1990 or thereabouts. I remember Steve Davis was there, Kevin Murphy, Russ Cooper, Stiglitz, and Andy Weiss, and I don't remember who else. As I said, Black sat taking notes throughout, did not present a paper, and did not say much. The funny part was that Stiglitz was presenting a paper by himself and Andy Weiss, and the two of them got in an argument over what was actually in the paper. Here's another funny thing. This is the paper I presented:ReplyDelete
It gets 7 citations in Google Scholar, which is pathetic, and I had actually forgotten until now that I wrote it.
Thanks for straightening out the pde story. I knew the pde and its solution played in important role, but obviously I forgot the details.
Lots of interesting Fischer Black details can be found in Perry Mehrling's biography:ReplyDelete
Here's another story, told to me by Gary Gorton. Again, you may have to check with Gary, as my memory is obviously not infallible. Gary is working at Wharton at the time this takes place. We'll say this is early 1990s. He gets a call from Fischer Black's secretary at Goldman Sachs saying that Fischer wants to talk to Gary and his coauthor about a paper he has written. Gary and coauthor take the train to Manhattan, go to Goldman Sachs, and are led to Fischer Black's office. They sit with Fischer, and talk about the paper for an hour or two. Fischer says thank-you, and they leave. That's it.ReplyDelete
My whole family lives in St. Louis, Stephen. I go there all the time. And that is how I know that I have absolutely NO desire to live in St. Louis at any point in the future. Sorry to disappoint you. ;)ReplyDelete
Great to hear that your syllabus is full of financial and other frictions. Sounds cool to me. Now why aren't most universities offering similar classes?
Oh, and I would CRUSH your prelims. :)
Noah, professors teach what they think is important. Believe it or not, many people sincerely think that frictions are second order stuff that doesn't really matter for the big picture. And you know what? They have good reasons to. Making a case that they do is far from trivial.ReplyDelete
In fact, this is a good example of why calibration is important exercise. It is a major stumbling block for people who would like to argue that financial frictions matter for the business cycle. What you find when you put these frictions inside a business cycle model is that unless you couple them with some other friction such as rigid prices or cash in advance constraints, it is really hard to get a lot of amplification.
Does this mean financial frictions don't matter? No. But the calibration exercise teaches us something important: You need to believe in more than that financial frictions exist if you want to argue that you should pay attention to them when discussing aggregate fluctuations.
Keep your cool. Right now you are learning how to do stuff. In time you should get to how to use it to understand things that you actually care about.
I think RV makes important points. To the non- trained economists who make up a lot of the blog comments on sites such as Mark Thoma's, everything looks so obvious, which is why they tend to love Krugman who keeps it simple for them. Williamson has tried to venture onto Thoma's blog to try to educate some of them, but he's wasting his time. They are not interested in models or thinking, they prefer their ideology.ReplyDelete
Its so obvious that increasing G must make us better off, you have to be stupid to think otherwise. Its so obvious that financial frictions matter. Its so obvious the sun goes round the earth.
But once you actually have to discipline your thinking by writing down a model that is well-founded in theory and with the data, they may not be so obvious after all. This was the point that Kartik Athreya was making I believe last year.
I guess you either don't like your family much, or the city gets on your nerves. I actually like it here.
"Now why aren't most universities offering similar classes?"
I think if you looked up the classes taught by people like Nobu Kiyotaki (Princeton), Mark Gertler (NYU), Ricardo Lagos (NYU), Pat Kehoe (Princeton), V.V. Chari (Minnesota), for example, you would find some interesting things. Remember, you're just looking at your own first-year PhD program. You can't infer from what you see there that the whole world looks like that, or that the second year of your program would disappoint you. In retrospect, you might come to appreciate what Chris House taught you in the first year.
Kartik's main point was that serious progress in economic science is going to be accomplished in the standard way: hard slogging with the computer and pencil and paper, peer review, and a lot of discussion about ideas in seminars and at conferences. I don't think that trying to educate people in the blogosphere is a waste of time though. There is no point in ceding ground to bad ideas in the policy debates.
I do like my family, but St. Louis is...to be brutally honest...a bit of a wasteland.
I do know that curricula vary wildly from place to place and person to person. I also know that this is not true in most sciences, underscoring the degree to which macro remains undeveloped. Macroeconomists have made a near-infinite cornucopia of models telling every possible kind of "story," but without a way to tell which theories are right and which are wrong, no one really knows which to use or which to teach.
I absolutely do appreciate what Chris House taught me. But I appreciate it because it allowed me to understand what's going on in macro.
"I don't think that trying to educate people in the blogosphere is a waste of time though. There is no point in ceding ground to bad ideas in the policy debates. "ReplyDelete
I agree in general. I was referring mainly to the Thoma website where the typical commenter believes everything Krugman tells them. He is a brilliant polemicist whatever his merits as a macroeconomist (we know he is a master of international trade theory)
Kartik here: since my views came up, let me say: I was trying to make both the points that " anonymous" nd Steve noted. I was too broad though in my skepticism. The "industry " blogs I read are often excellent. For example, I learned quite a bit of vital info on the workings and important forces in the auction rate security market from industry insider blogs. Things that no outsiders, macroeconomist or otherwise, could have known.ReplyDelete
I don't see a wasteland here at all. St. Louis has one if the nicest urban parks in the country, a world-class botanical garden, a world-class university (ahem), a world-class symphony orchestra, some other decent music, and some good restaurants. It's also very easy to live here. I live just off campus in a very nice neighborhood and walk 10 minutes to school. There is urban blight alright, in the city proper, but you could find much worse blight if you get in the car in Ann Arbor and travel a short distance to the east.
I'm not sure why you find the diversity in modern macroeconomics a shortcoming. We don't want some universal model that does everything. That would be a bad model. The diversity I find interesting and stimulating.
Re: St. Louis: You're right, it's not hopeless. I'm just very picky. The urban blight definitely bothers me. Here in Ann Arbor I'm insulated from the urban blight of Detroit, but I still long for a cool cosmopolitan city, and the closest one is 4 hours way. I really want to go back to the West Coast (where I did undergrad).ReplyDelete
Re: macro diversity: Diversity is well and good. But like I've been saying, these models are mostly "storytelling" models, whose function is to demonstrate an idea about one effect that's going on the economy. But there's no way to tell in advance which one to use, whether for forecasting or for policy analysis. What ends up happening is that some event happens (like the crisis of 2008), and then we say "Oh yeah, this is just like (such and such model) that everyone mostly ignored before now!" Then we start emphasizing that model and making a million variations of it. Then some different event happens, and we say "Oh, well THIS is just like (another model)!" We have all the models we need, and yet most of them are mostly useless in practice. The diversity is interesting and stimulating, but so is Battlestar Galactica. The problem IMO is not that we make too many models, it's that we reject too few.
"The problem IMO is not that we make too many models, it's that we reject too few."ReplyDelete
You wrote earlier "without a way to tell which theories are right and which are wrong, no one really knows which to use or which to teach." So how do we know which models to reject?
1. If you are shooting for an academic job, locational choice is going to be limited. Even the people at the very top of the profession typically can't literally choose where to live, though some get lucky. However, you can find interesting things going on in surprising places, and you get to travel more than the average person does.
2. This is the nature of the game. Economics is hard. All of these models can easily be rejected, and we're then left with debates over how useful they are. You're too cynical though. I've actually seen firsthand how people use these models productively to make better policy, and am much more positive about the benefits of macroeconomic research, particularly what has been done over the last 40 years.
See also: http://www.phdcomics.com/darkmatter/index.php?page=1ReplyDelete
Noah seems confused -- he wants to reject models he claims cannot be rejected because we don't know how. Noah, here is a suggestion from a tenured professor: shut up and do your work. Leave speaking publicly to those with a little more perspective.ReplyDelete
Actually, Noah seems like a smart guy with healthy skepticism about the models he's seen. (Confession:) I'm a mathematician with some very minor knowledge of physics models and even more minor knowledge of economics models (including one influential but very bad one). But having been reading some of the blogs (not limited to Krugman's) I think Noah is implying something very important about economics that is distinctly not true in physics or other sciences: that there's not enough established body of theory in economics. There are theories and equations but how much of this is established with something like the degree of confidence in physics, chemistry, or biology? Please take that as a question.ReplyDelete
P.S. Not everyone in the MIT math dept. would be able to solve those equations. Some do other kinds of math. There are other kinds of math! but in every university we expect much the same basics to be learned by every first-year grad student. It's not professor's choice.
"Noah, here is a suggestion from a tenured professor: shut up and do your work. Leave speaking publicly to those with a little more perspective."
Pulling rank on a graduate student is a fairly lame way to try to win an argument. Even if one is in fact a tenured professor, dissing a grad student for questioning one's oracular view of economics is simply not in keeping with the academic tradition, nor with the concept of mentorship as it should be embraced by tenured faculty.
Agreed. There is way too much condescension for my liking. Models are cool, but if one is looking for insights, the ultimate test is whether or not the model you have (either formal or informal) actually helps you in the real world--you know, perhaps helps you predict something of what is going to happen... for example. So, for all of the amazing intellectual talent on tap here, I am curious--how many here predicted the meltdown? What did your models tell you about that?ReplyDelete
Anonymous: Pulling rank on a graduate student is a fairly lame way to try to win an argument.ReplyDelete
Yes, and you put it kindly. It's strongly reprehensible.