Tom Sargent, Nobel Laureate
The "Nobel Prize in Economics" (it isn't really a Nobel Prize, but a prize in memory of Alfred Nobel) for 2011 was announced this morning, going to Tom Sargent of Hoover and NYU, and Chris Sims of Princeton. To those who have accused the Nobel of being a bit...political (as in, left-leaning), this is anything but. Sims is less outspoken than Sargent, but that's because Sargent is very, very outspoken. For a sense of this, I recommend the interview Sargent did with former Research Director of the Minneapolis Fed, Art Rolnick (and not only because Sargent is nice enough to cite a paper of my own in the course of the interview). The interview has a lot to say about important and current issues in bank regulation, but to get a fast fix, have a look at the section entitled "The 2009 Fiscal Stimulus," which begins on page 32 of the web-posting. Or, for those with a lot more time and energy and some tolerance of math, Sargent's book "The Conquest of American Inflation" (Princeton University Press, 1999) is a small masterpiece that illustrates perfectly what Sargent brings to the intellectual table.
But, if I may, I'd like to see if anyone in Ricochet-land will react to a quote from the Rolnick interview that almost perfectly represents Sargent's methodological mindset. Sargent asks the interviewer (Rolnick) to specify criticisms of the sort of macroeconomics that he (Sargent) pioneered, and Rolnick, in a long list of criticisms, says that "modern macroeconomics makes too much use of sophisticated mathematics to model people and markets." Sargent responds to this particular criticism with "... it is true that modern macroeconomics uses mathematics and statistics to understand behavior in situations where there is uncertainty about how the future will unfold from the past. But a rule of thumb is that the more dynamic, uncertain and ambiguous is the economic environment that you seek to model, the more you are going to have to roll up your sleeves, and learn and use some math. That’s life."
Is that life? Do mathematical models help or hinder our understanding of a world that is increasingly dynamic, uncertain, and ambiguous? (Obviously, I have my own opinion about this, very much aligned with Sargent's. But I'll save the "why" until others weigh in, if anyone does.)
- Comment (18)
- · Quote
- · UnfollowFollow (1)
Comments:
Jun '10
Re: Tom Sargent, Nobel Laureate
Let's adopt a rule: the more recent Nobel laureates in Economics trump older vintages (especially Krugman).
Jan '11
Re: Tom Sargent, Nobel Laureate
Much of the answer depends upon the factors being used and the weighting they are assigned in the regression analysis.
I think that there is an arguement to be made for 'Peer Reviewing' the modeling assumptions before gathering & analysis rather than always arguing about the finished product. This could help in many areas: economics, climate change, epidemiology.
Feb '11
Re: Tom Sargent, Nobel Laureate
Obviously some mathematical models lead to foolishness because the models are bad. But my bachelor's is in engineering. I don't think you can understand anything deeply without math. The things we understand without math are the things we understand least.
May '11
Re: Tom Sargent, Nobel Laureate
It seems counter-intuitive that math is more important in explaining dynamic, ambiguous behavior than in more predictable situations, but I believe it is true. My eyes glaze over and all thought regresses whenever I try to read an article using macroeconomic mathematical models, so I can't explain this well. I wish Bill James were a Ricochet member because he has spent a lifetime using statistical analysis to answer the most basic and important questions of life. The movie, Moneyball, has renewed criticism of James' approach, but I think James would say that the most important process in the development of knowledge is to ask the right question and frame it in such a way that the answer can be "proven" with statistical analysis and results verified independently by objective observers.The more common approach relies on common sense and other forms of mythology.
Dec '10
Re: Tom Sargent, Nobel Laureate
Paging Doctor Asimov; Psychohistory is about to emerge in Thread 1.
(Not that I mind; I'm an Asimov fan)
May '11
Re: Tom Sargent, Nobel Laureate
CJRun: Paging Doctor Asimov; Psychohistory is about to emerge in Thread 1.
(Not that I mind; I'm an Asimov fan) · Oct 10 at 4:21pm
Yes! If only we can define that Prime Radient!
Re: Tom Sargent, Nobel Laureate
First a pre-comment: Thanks for joining us, David. (For those of you who may not know, David Kreps, former academic dean at Stanford business school and the winner of the John Bates Clark Medal, is a major economist in his own right.) You have a reputation for demanding rigor in the lecture hall. I'm on my toes. Which leads to:
My comment: I'm sorry to drag global warming into all this, but Freeman Dyson's critique of the current science, to the extent that it rises to the level of science, strikes me as relevant here: Dyson argues that there is much too much modeling going on and much too little actual observation of the physical environment. Too much speculative mathematics, in other words, and too little reality. (If you've never seen it, by the way, this clip of Dyson is well worth the watching: serene, analytical...and devastating.)
Now, setting Tom Sargent's work to one side for a moment, may I just ask a general question? How does the discipline of macroeconomics--or, for that matter, of economics of any stripe--protect itself against this sort of tendency?
Edited on October 11, 2011 at 1:47amFeb '11
Re: Tom Sargent, Nobel Laureate
I like the way Southern Pessimist makes the point. The problem isn't models and math, it's method. Statistics without explanatory models are wrong (e.g., if increases in the minimum wage aren't correlated with increases in unemployment, then minimum wages somehow don't reduce employment opportunities). Models without adequate statistical tests are wrong as well (e.g., the model says that $1 of stimulus will produce $2 of growth, so the failure of the economy to grow must mean that it would have shrunk more without the stimulus).
To the extent that it is possible to theorize about macroeconomic behavior (and I'm skeptical), it has to start with a behavior model that can be tested. Ideally, there are alternative models that can be tested in the same manner, and all the testing will accomplish is to tell you which model best explains the statistics. This is much hard to do for economic activity than it is to measure the impact of fielding in baseball.
May '11
Re: Tom Sargent, Nobel Laureate
James has found it to be very easy to quantify the value of hitting and pitching in his analyses but has been severely frustrated with developing a predictive model for the value of fielding. It is very difficullt to measure because poor players never get to a ball that a good player might make an error on.
Sep '10
Re: Tom Sargent, Nobel Laureate
In an interview last week Greenspan said that all of the models missed the 07-08 down turn. Leading up to the crisis you could turn on cable TV and watch 3 or 4 shows on house flipping. I imagine there was a similar scenario that played out with tulips a few centuries ago. I know there was one with tech stocks a decade earlier. In answer to your question it depends on the models and how you use them.
Jul '10
Re: Tom Sargent, Nobel Laureate
Great, first reviews by David Berlinski and now David Kreps.
Models not only require an understanding of the mathematics, they also require an understanding of what the math is trying to capture and the interplay of various formulas. Dynamism, uncertainty, and ambiguity will make fools of us all, but they have to be engaged to be understood; the more intensely they are engaged, the better the insight and the better the necessary revisions will be.
That's life; suck it up.
May '10
Re: Tom Sargent, Nobel Laureate
One model is all you need:
The race is not to the swift, nor the battle to the strong, but that is the way to bet
-- Damon Runyon.
May '11
Re: Tom Sargent, Nobel Laureate
Global warming seems to me to be the perfect example of fuzzy math applied to political ideology. My sister is a long-tenured professor at a major university in the field of crop science. She hasn't taught an actual class since she was a graduate student but she has had a sucessful career based on developing computer models predicting crop yields. A few years ago she was describing a program she was developing which was related to the effects of rotating pesticides or some such factor and I asked her what the effect was of El Nino on her crop yields. Her IQ is at least 30 points above mine; she stared at me a moment, blinked a few times and said "I can control for pesticide use in my models but not weather".
Sep '10
Re: Tom Sargent, Nobel Laureate
I'm going to assume this is too obvious to mention; so I'll mention it. Non linear problems are notoriously tough to model, require incredible levels of computational power and often require Partial Differential Equations and their associated Finite Difference methods to even begin to approach the problem rationally, never mind accurately.
And here I'm only speaking of the mathematical modelling of derivatives by quants at financial institutions, never mind whole markets.
Apr '11
Re: Tom Sargent, Nobel Laureate
Peter Robinson: My comment: I'm sorry to drag global warming into all this, but Freeman Dyson's critique of the current science, to the extent that it rises to the level of science, strikes me as relevant here: Dyson argues that there is much too much modeling going on and much too little actual observation of the physical environment. Too much speculative mathematics, in other words, and too little reality.
How does the discipline of macroeconomics--or, for that matter, of economics of any stripe--protect itself against this sort of tendency? · Oct 10 at 4:46pm
Edited on Oct 10 at 04:47 pm
The problem with global warming is that the "scientists" who designed the models had a pre-conceived goal in mind and created systems that fit their political ideologies. I understand that the same ideological temptation is there for economists of all stripes as well, but this potential corruption shouldn't completely discount the importance of mathematical modeling which should be used as one tool among many. I'll chanel Churchill on this, "mathematical modeling is the worst method for predicting complex, dynamic systems except all the others that have been tried".
Aug '10
Re: Tom Sargent, Nobel Laureate
I'm a software engineer first, hobbyist economist second. I write code for a living. And I can assure you, as you try to cram more of the real world into a model, the complexity increases exponentially to the point where you wind up with a fragile, complex model that tends less and less to resemble reality.
I am less a fan of mathematical models, and more a fan of Austrian economics' methodology. It starts from basic, provable statements about human interaction and economic activity; crafts a set of observable relationships; and builds long term predictions from there.
Austrians dis-aggregate the large "blob" numbers that macroeconomics relies on, because there is no identifiable actor that exercises "demand" or "supply" on its own. This keeps macroeconomic discussions in the realm of human action, instead of the abstracted realm of economic theory. It also prevents the fallacious suggestion that aggregates can be directly manipulated--the government can not flail about trying to "stimulate aggregate demand" if there is no such aggregate.
It won't let you make headline-grabbing short term predictions, but it seems more solid than 95% of the math models which failed so spectacularly in 2008.
Edited on October 11, 2011 at 8:53pmAug '10
Re: Tom Sargent, Nobel Laureate
Re: climate change: remember the big University of East Anglia scandal a few years ago? The media focused on the e-mails where scientists discussed plans to exaggerate global warming material... and completely missed the real treasure hackers uncovered: the FORTRAN source code used to analyze tree ring data and come up with global warming numbers.
The code was, to put it politely, complete trash. But more damning was a detailed journal by the university's main coder, who inherited the computer model and was charged with maintaining it. The poor guy carefully documented his 3 year long quest to understand the code, marshal the input data, and reproduce the final warming numbers.
In the end he gave up. UEA's own coder concluded there was no way he could understand the computer model being used as justification for strong scientific claims.
Moral of the story: introducing models adds a breakable component to any theory. Models amplify bad assumptions, while simultaneously hiding them; they require the creation of additional assumptions, to cover shortcomings in modeling the real world; and their implementation adds an element of technical error, wholly separate from the "pure" scientific theory it is testing.
Aug '10
Re: Tom Sargent, Nobel Laureate
As an engineer, I'm a big fan of math. However, I'm also very cautious regarding the predictive power of mathematical models, especially when applied to complex systems, and most especially when applied to complex adaptive systems.
In mechanical engineering, structural models are modelling things that have very few variables and which have well-understood, repeatable behaviors. For example, modeling the stresses on a wing can be done with finite element models. But even these models get very complex, and it's easy to go wrong with a model.
I'm highly skeptical of models that claim to be able to predict in detail the response of complex adaptive systems like the national economy or the global climate. These are systems that have millions of variables that interact in ways we still don't clearly understand, and which may not respond to the same input in the same way from day to day because of the adaptive nature of the system.
For example, it seems rather obvious to me that if a fiscal stimulus fails, actors in the economy will take that information into account, and it will change their response to a second stimulus - probably for the worse.