3 ms·

Due to the too frequent economic crises, during my education I considered getting a Ph.D. in economics. I got some famous texts on economics, read some chapter

by graycat 13y ago

Due to the too frequent economic crises, during my education I considered getting a Ph.D. in economics. I got some famous texts on economics, read some chapters, talked to a famous economics professor, did a big upchuck, concluded that the field was hopeless, and stayed with applied math. Why hopeless? Here are my guesses: Data and Reality. Historically the people studying economies just had far too little data on their subject. By analogy they were trying to understand, repair, or re-engineer a car but had never looked under a car, never popped the hood, didn't know what piston rings were, and still were in doubt if the thing had front wheel or rear wheel drive. In a medical analogy, they had no cadavers to dissect, X-rays, MRIs, blood tests, etc. In an astronomical analogy, they had no telescopes. So, they never did well with what is usually one of the first steps forward for a science -- the descriptive part where we just give a good description of the subject. Astronomy, biology, thermodynamics, chemistry, electricity and magnetism, and more all started with good descriptions of their subjects. Or, for cars, start tinkering, as Henry Ford did, with a lot of time with dirty hands, and only later use finite element analysis to build models of stress and strain in continuum mechanics. Or naval architects had a lot of experience before they moved to towing tanks and the Navier-Stokes equations. Bluntly I had to conclude that the academic economists really just didn't have even a first, good descriptive understanding of a real economy. During WWII for war production planning, Leontief worked on 'input-output' models of the US economy. Good for him. But I was told that the US academic economics community very much did not like his work because it was not 'theoretical' enough as in, say, 'political economy'. So, it looked like academic economics wanted to stay with pomp, pretense, prestige, ignorance, and incompetence. One of my Ph.D. advisors wanted me to take a course in economics so that if I did any work on a committee on a 'public sector' problem, then I could defend myself from attacks by floods of gibberish from academic economists. I've had no desire to do any such 'public sector' work and have not, but I signed up for the suggested course. I wanted to be nice to the professor and not cause trouble. So, during his lecture with a lot of hand waving and free hand curves but no data and nothing convincing, I just took notes and said nothing. Then after class I asked him what he was assuming about his curves -- continuity, uniform continuity, differentiability, continuous differentiability, monotonicity, concavity, pseudo-concavity, quasi-concavity (e.g., in case he was intending to use constraint qualifications for the Kuhn-Tucker conditions for optimality). He was unhappy. Later in the day, I got a message to see my advisor. I was out of the economics course -- the professor claimed that I might disrupt the class. That was not my intention, but good riddance! But that professor's reaction seemed to be a special case of a major 'feature' of academic economics -- have a tightly knit 'club' that wants only true believers and pushes out any skeptics. Or the first rule of Economics Club is never talk about the rules of Economics Club! Net, academic economists know next to nothing important about any real economy. Real economic policy needs much more in data on real economies, insight into reality, good judgment, and real effectiveness with applied math than is common in academic economics.

dnautics 13y ago
Lest we forget though, there is Hayek's "information problem". Econometric data are by their very nature aggregated, and that fundamentally introduces bias. How do you choose how to aggregate your data? How do you choose what relevance that particular aggregation means (from a policy standpoint, or from a fundamental standpoint)? A number such as "unemployment", for example, is loaded. Are parents that could have a job but choose to stay home to take care of children "unemployed"?
- graycat 13y ago
  Common problem. E.g., how to "aggregate" data in astronomy? For a simple first cut answer, work hard, get a lot more data, work hard, have some good ideas, get a lot more data, work hard, etc. So, for astronomy, use spectroscopy to identify the elements in the star. Use the brightness of the star and the red shift. Get empirical distributions of various characteristics of the stars and begin to guess that stars go through stages and where the distributions are low the stages are short. So, start by observing, getting a lot of data, and being just descriptive; economics didn't and still doesn't do nearly enough here. Then look a little deeper and start to make some sense of it; economics has done next to nothing here, necessarily so since they didn't get a good grade in the first grade with the descriptive work. For something with a little promise, e.g., Leontief's work , they don't like that and regard it as not acceptable in Economics Club. So, with such work, in astronomy get Hertzsprung–Russell diagrams. Nice progress. Now, continue: Take what is known about weights in the periodic table and observe that if press together two deuterium atoms to make one helium 4 atom, then lose some mass and, thus, get some energy. Keep this up across the periodic table and see that iron has the least such energy. Observe that can keep pressing light atoms together to get heavier ones, e.g., carbon, oxygen, ..., iron and get off energy. Do a lot of model building of the reactions in the centers of distant stars. Check with the data, e.g., the Hertzsprung–Russell diagrams. Eventually come up with a well tested, apparently quite solid, model of stars. A lot work, bright ideas, spectroscopy, expansion rate of the universe and red shift, nuclear physics, etc. Didn't say it was easy. Much the same in other fields of science that have been successful. So, if economics wants to copy that methodology, then they need to get their hands on more data. Then, say, they need to look at the flows that, intuitively, appear to be generating the data. I know; I know; too soon want a lot of curves on propensities, and will have a tough time there. Not guaranteed to be easy. But neither were the other sciences that were successful. My view of academic economics is that they sit in small, dark rooms with the doors closed, hope for a single stroke success comparable with E = mc^2 in physics, dream up models, essentially of imaginary economies, hope, but not get very far. So, they don't want to do the first, observational, low level grunt work of data collection and basic description that astronomy, chemistry, biology, etc. did. There have been suggestions that the Princeton econ department has two halves with one half, then with Bernanke, the more empirical. If so, good for the Fed and the US since we will be less likely to be stuck with crack pot ideas from "defunct economists". A danger here: So, with such work with imaginary economies, some to be defunct economist makes a lot of absurd assumptions and finds some optimal solution -- maybe he studied some linear programming (Kantorovich, Solow), Kuhn-Tucker theory (Arrow), Lagrange multiplier theory (Debreu), or dynamic programming (Samuelson). Then, seeing a real economy and an opportunity for fame, status, prestige, tenure, lots of undergraduate co-eds for secretaries, he takes the solution from his imaginary economy and says that that is what the real economy should do. Bummer. Good way to kill tens of thousands of people from unemployment, busted homes, street crime, infant mortality, drug abuse, suicide, etc. Incompetence is a bad thing; incompetent economists are really ugly things. Your point about unemployment is an example: Early on I looked at it and asked, where the heck is the definition, that is, the empirical or operational definition? Next, one level deeper, what the heck are we really counting? E.g., might we be partitioning the data in some way that would be better? So here what do we want for better? Sure, we want something that can help us with reductionism, that is, clear causality. But the actual unemployment data is so poorly defined and so "aggregated" that our hope for anything causal is just smoking funny stuff, like the astronomers calling stars just points of light and not looking deeper with spectroscopy, the periodic table, nuclear physics, etc. Finally I understood what the empirical economists intended to do with data such as unemployment: They just said, it's a measurement; hopefully it is done in a consistent way month by month; we agree we don't have a good definition and that the data is too aggregated and has basic problems in measurement, but we just take the data as it is; with the data, we will do just empirical statistical modeling. Then, recognized or not, they have essentially given up making unemployment data an input to anything reductionist, causal, or scientific. Suckage. Meanwhile academic economists are driving late model cars, and millions of people are suffering from unemployment likely for no very good reasons. Suckage.

by graycat 13y ago

dnautics 13y ago

Lest we forget though, there is Hayek's "information problem". Econometric data are by their very nature aggregated, and that fundamentally introduces bias. How do you choose how to aggregate your data? How do you choose what relevance that particular aggregation means (from a policy standpoint, or from a fundamental standpoint)? A number such as "unemployment", for example, is loaded. Are parents that could have a job but choose to stay home to take care of children "unemployed"?

graycat 13y ago
Common problem. E.g., how to "aggregate" data in astronomy? For a simple first cut answer, work hard, get a lot more data, work hard, have some good ideas, get a lot more data, work hard, etc. So, for astronomy, use spectroscopy to identify the elements in the star. Use the brightness of the star and the red shift. Get empirical distributions of various characteristics of the stars and begin to guess that stars go through stages and where the distributions are low the stages are short. So, start by observing, getting a lot of data, and being just descriptive; economics didn't and still doesn't do nearly enough here. Then look a little deeper and start to make some sense of it; economics has done next to nothing here, necessarily so since they didn't get a good grade in the first grade with the descriptive work. For something with a little promise, e.g., Leontief's work , they don't like that and regard it as not acceptable in Economics Club. So, with such work, in astronomy get Hertzsprung–Russell diagrams. Nice progress. Now, continue: Take what is known about weights in the periodic table and observe that if press together two deuterium atoms to make one helium 4 atom, then lose some mass and, thus, get some energy. Keep this up across the periodic table and see that iron has the least such energy. Observe that can keep pressing light atoms together to get heavier ones, e.g., carbon, oxygen, ..., iron and get off energy. Do a lot of model building of the reactions in the centers of distant stars. Check with the data, e.g., the Hertzsprung–Russell diagrams. Eventually come up with a well tested, apparently quite solid, model of stars. A lot work, bright ideas, spectroscopy, expansion rate of the universe and red shift, nuclear physics, etc. Didn't say it was easy. Much the same in other fields of science that have been successful. So, if economics wants to copy that methodology, then they need to get their hands on more data. Then, say, they need to look at the flows that, intuitively, appear to be generating the data. I know; I know; too soon want a lot of curves on propensities, and will have a tough time there. Not guaranteed to be easy. But neither were the other sciences that were successful. My view of academic economics is that they sit in small, dark rooms with the doors closed, hope for a single stroke success comparable with E = mc^2 in physics, dream up models, essentially of imaginary economies, hope, but not get very far. So, they don't want to do the first, observational, low level grunt work of data collection and basic description that astronomy, chemistry, biology, etc. did. There have been suggestions that the Princeton econ department has two halves with one half, then with Bernanke, the more empirical. If so, good for the Fed and the US since we will be less likely to be stuck with crack pot ideas from "defunct economists". A danger here: So, with such work with imaginary economies, some to be defunct economist makes a lot of absurd assumptions and finds some optimal solution -- maybe he studied some linear programming (Kantorovich, Solow), Kuhn-Tucker theory (Arrow), Lagrange multiplier theory (Debreu), or dynamic programming (Samuelson). Then, seeing a real economy and an opportunity for fame, status, prestige, tenure, lots of undergraduate co-eds for secretaries, he takes the solution from his imaginary economy and says that that is what the real economy should do. Bummer. Good way to kill tens of thousands of people from unemployment, busted homes, street crime, infant mortality, drug abuse, suicide, etc. Incompetence is a bad thing; incompetent economists are really ugly things. Your point about unemployment is an example: Early on I looked at it and asked, where the heck is the definition, that is, the empirical or operational definition? Next, one level deeper, what the heck are we really counting? E.g., might we be partitioning the data in some way that would be better? So here what do we want for better? Sure, we want something that can help us with reductionism, that is, clear causality. But the actual unemployment data is so poorly defined and so "aggregated" that our hope for anything causal is just smoking funny stuff, like the astronomers calling stars just points of light and not looking deeper with spectroscopy, the periodic table, nuclear physics, etc. Finally I understood what the empirical economists intended to do with data such as unemployment: They just said, it's a measurement; hopefully it is done in a consistent way month by month; we agree we don't have a good definition and that the data is too aggregated and has basic problems in measurement, but we just take the data as it is; with the data, we will do just empirical statistical modeling. Then, recognized or not, they have essentially given up making unemployment data an input to anything reductionist, causal, or scientific. Suckage. Meanwhile academic economists are driving late model cars, and millions of people are suffering from unemployment likely for no very good reasons. Suckage.