Like Silas Barta, I have come to the view that a lot of macroeconomics is terribly confused. I have an ongoing mission to make sense of macroeconomics. My explanation of how most macroeconomic theories of macroeconomic fluctuations work at their core is here.
I’m not sure I understand your question, so if I’m answering past you, let me know. Anyway, if you came up with a model that predicted macroeconomic variables better than the marginal traders who trade assets strongly affected assets, yes you could make a killing. As I see it macroeconomic modeling is about producing theoretical understanding rather than producing predictions about the path. If we had really good and detailed prediction markets of all important macroeconomic measures, that would generate good predictions but it might not be very enlightening about what policy makers should do differently and why. In other words, modeling answers questions about what forces are at work while prediction about whats going to happen at some specific point in time is about what the net effect of those forces are.
Theoretical macroeconomics can help us understand what kinds of things we should try to measure, what kinds of general rules policy makers should adopt etc. A good predictive mechanism can help us determine what people should do right now (should your grocer expect an increase or decrease in sales), whether policy makers do a good job.
Anyway, if you came up with a model that predicted macroeconomic variables better than the marginal traders who trade assets strongly affected assets, yes you could make a killing. As I see it macroeconomic modeling is about producing theoretical understanding rather than producing predictions about the path.
Such “theoretical understanding” is as if you had a theory of physics that purported to “explain” past observations but was unable to make any predictions about future events. That is not science. At best it’s just empty philosophizing, and at worst pernicious bullshit used to rationalize actions out of touch with reality.
As for the relevant prediction markets, they already exist for all practical purposes. I don’t know much about finance or investment, but I do have at least a rough idea how any correct non-trivial macroeconomic prediction could be translated into a winner investment strategy.
Perhaps I wasn’t clear. I don’t think that macroeconomic theorizing is useful even if it predicts nothing. Obviously that’s useless. I think that macroeconomic theorizing is useful even if they don’t generally outperform market predictions of market forcasted variables because it gives you different kinds of information about the economy.
For example, imagine a world where relevant traders have very good models for predicting the value of the S&P500. However, these models are proprietary, detailed and heavily specialized for predicting the S&P500. Academic models are open, relatively simple and based on ‘first principles’.
In this scenario, how useful are these two types models? Depends on what kind of question you’re asking:
What will the value of the S&P be in 1 year? Trader models useful. Academic models useless.
What kinds of alternative monetary institutions would be better than current ones? Trader models useless. Academic models useful.
Perhaps you could formulate prediction markets to answer the second kind of question well, but I sure don’t know how and in any case is not currently something people do.
There are also plenty of non-trivial macroeconomic predictions that you cannot make money off of. For example, ‘employment will be 1 point higher in 1 year’ is non-trivial, but it could be information already perfectly incorporated into the markets. ‘printing money now increases employment and inflation for 1 year out’ is also nontrivial but may also be fully accounted for. If you don’t have access to markets that predict precisely this information, such forecasts can easily be useful.
Also, all relevant prediction markets do not exist. Examples: there are no very liquid unemployment prediction markets, there are no very liquid real GDP prediction markets etc. What is individually useful to trade is not necessarily the same thing as what is socially valuable to know.
What kinds of alternative monetary institutions would be better than current ones? Trader models useless. Academic models useful.
It may well be that the existing trader models are useless for answering this question, but any novel model that is capable of answering them should ipso facto be able to provide useful investment information. There are endless public controversies over the expected effects of the current monetary policy, which have direct bearing on all sorts of markets, and if you can forecast their effects with more accuracy than any existing model, you should be able to beat the markets.
There are also plenty of non-trivial macroeconomic predictions that you cannot make money off of. For example, ‘employment will be 1 point higher in 1 year’ is non-trivial, but it could be information already perfectly incorporated into the markets.
If you really know that employment will be one point higher in a year, there are straightforward implications on trading. For example, people who bet (possibly as a way of hedging their investments) that there’s going to be a very bad recession within a year are certainly wrong, and you can profit by betting against them. If I, a complete amateur, can easily think of such strategies, then I can only imagine what expert financiers could do with this information! It’s similar with predicting NGDP and all other aggregate variables.
I don’t see any logical possibility how some macroeconomic prediction could be at the same time: (1) meaningful and non-trivial, (2) more accurate than the state of the art, and (3) useless for investment.
3) sure; my point was that (1) does not imply (2).
1) I’m not clear on how you’re disagreeing with me, so this is mostly going to be a reworking of my previous answer. Let me know if you can clarify your disagreement.
Academic and trader models can be better at answering different questions. Neither one need always dominate the other. Academic models might be better at answering questions which rarely come up, for example ‘what happens when we change monetary institutions?’, so not offer useful investment advice except in those rare occasions but still be useful for deciding what options policy makers should consider.
Academic models can also provide much more understandable and accessible models of the economy (see 2). Non-traders may find these much more useful than trader models (which they may not have access to or be able to understand).
2) No. Remember I stated that this information was already taken into account by the market. NGDP, employment etc. are not random walks (Edit: I mean they are not martingales); their expected future value is not independent of past values. Unemployment is highly non random see here. Some macroeconomic variables are random walks, for example stock market indexes, but most of them are not.
I’m not sure I understand you either. Are you actually saying that if I had an oracle capable of telling me what, say, the rate of unemployment or NGDP growth will be in a year, it would not be possible to make investments with above-market returns using this information?
Moreover, I am at a loss trying to imagine a theory that would enable us to predict with reliable accuracy what would happen if we changed the monetary institutions in a given way, and which wouldn’t also enable us to get reliably accurate information on the uncertain and controversial questions about the consequences of the present monetary policies. This would also constitute valuable investment information. (Or do you think it wouldn’t?)
On the whole, the problem is that I simply cannot imagine any questions that macroeconomic theories purport to handle where a truly reliable and non-trivial information would not be valuable for investment.
Also, you can’t answer these questions by claiming that the relevant information has already been incorporated into the market prices, only in some obscure way that we now seek to disentangle. Many investments are specifically made in order to hedge against uncertainty in macroeconomic trends. If you have a theory that eliminates these uncertainties, or at least provides more accurate probability distributions, there’s a straightforward killing to be made there.
“On the whole, the problem is that I simply cannot imagine any questions that macroeconomic theories purport to handle where a truly reliable and non-trivial information would not be valuable for investment.”
One where the prediction is longer than your investment window.
1) It all depends on whether your predictions are better than or worse than the relevant traders. If the traders already have access to such an oracle, you won’t be able to make any money; if they don’t, you will. Many macroeconomic variables of interest (GDP, employment etc.) are not martingales which means that predicting their movements is not the same thing as predicting them better than relevant traders. Being able to predict an asset price difference (between now and some future time) and acting on that information tends to move the price to eliminate that difference now. Being able to predict say a difference in unemployment (between now and some future time) does not necessarily tend to move unemployment to eliminate that difference right now.
Perhaps the following is a more relevant example: lets that you and everyone else found out a couple of days ago that aliens are going to land on earth in a month. The variable aliens-on-earth (binary) is certainly an important macro variable. Naturally market prices currently reflect the fact that aliens will soon be among us. The current value of aliens-on-earth, however, is False and no amount of trading can change that. Aliens-on-earth is not a martingale; its current value is not equal to its discounted expected value and you can predict it quite well (aliens-on-earth(t) = {False for t < 1month and True for t >= 1 month).
2) I’ll give a concrete but extreme question: what would happen if the US moved to a uranium based commodity money system? Would it be good? bad? Trader based models are likely pretty useless for this because no one thinks this is likely to happen so there are few benefits to developing a model for it. Even if they did, you might not have a way to get at those predictions. However, you could get a rough idea about the consequences by building an economic model for this scenario taking into account the economics of money, estimates about the stock and potential supply of uranium, the costs of avoiding radiation poisoning during transactions etc.
Obviously this question is extreme, but questions like it are valuable: should we move to a competitive currency system (free banking)? what would that look like? How does an optimal central bank behave (for various definitions of optimal)? Are countercyclical unemployment insurance extensions welfare improving? Is countercyclical government spending welfare improving?
I have some opinions on the preceding questions including that some of them are obviously of little interest, but I only came to them because I learned some macroeconomics.
4) ‘There’s potential money to be made by selling rough macro models to say industrial producers’ is a fair point. However, it’s a lot of work to turn a model that says ‘keeping money supply constant, an increase in people’s desire to hold money produces recessions’ to something that might be useful to producers for planning. It’s the job of macroeconomists to come up with the former not the latter in the same way that it’s the job of government funded battery researcher to come up with basic theory related to batteries, or a computer scientist to come up with interesting and potentially useful classes of algorithms. If one of these researchers comes up with something really big that is immediately implementable, perhaps they will go off and start a company to take advantage of the idea, but generally they stick to basic theoretical research because thats whats easy and comfortable and where their advantage is.
(1) I know what martingale variables are, but I don’t see why the non-martingale nature of the macroeconomic variables is relevant. Clearly, if you have figured out a novel way to predict the coming of the aliens ahead of others (or even just to predict its timing and other details more accurately), you can get rich by figuring out how their coming will affect the markets. This is perfectly analogous to a theory that will predict various macroeconomic variables more accurately than the state of the art, since these variables have predictable effect on asset prices. (In fact, once you have this information, they are no longer martingales for you, since e.g. if you know a recession is coming withing a year, the expected trend for countercyclical assets is upward.)
Or to put it differently, from all that you have written thus far, I still don’t see a concrete example (either actual or hypothetical) of the thing whose existence you assume: macroeconomic predictions that are interesting, novel, accurate, and at the same time useless for beating the markets.
(2) I understand that there are hypothetical questions about monetary systems where an accurate answer would have no practical implications by itself. However, presently we are in a situation where there are deep and bitter disagreements even about the predicted consequences of the ordinary and standard monetary policy options. What I find implausible is that one could obtain correct answers of the former sort without a theory that would at the same time be able to give more accurate answers to questions of the latter sort (which would again translate into investment information in a straightforward way). It would be as if you had a theory of mechanics capable of predicting the motions of hypothetical planetary systems but of no use for practical technical problems.
(3) Regarding your point about theoretical vs. applied research in other areas, the same heuristic actually is widely applicable. Whenever you see people doing research into something that should have straightforward practical applications, but you don’t see them running to monetize the results, something fishy is likely going on.
Of course, sometimes there is real insight that can’t be monetized in any obvious way, like for example fundamental theoretical physics. However, there is an essential difference here. A physical theory can make predictions only about things that are of no business interest, so in fact you have to spend money to contrive experimental setups to test it. In contrast, anything that a macroeconomic theory might be making predictions about and that might actually occur in the real world is inherently of business interest. (And again, if you have a counterexample, I’d be curious to hear it.)
For some reason I feel compelled to return to this topic:
My point is not that macroeconomics is a great field filled with great insights (it’s not and most macro theorists are terribly confused) but that it’s not as ridiculous as you seem to imagine it that some economists have novel, interesting and true things to say about inflation, unemployment, GDP etc and are not themselves fabulously wealthy.
(1) For example, some macroeconomic theories predict behaviors like the parable of the babysitting co-op. You can also run experimental economies (like this) and make predictions about the behavior of the economy.
I might set up a play economy where different people produce different goods and trade and consume them. Money is traded but not produced. We let this economy do its thing for a while and then suddenly (and without announcing in advance) give everyone 20% more money. Using my favorite macro theory I could make a number of interesting and novel predictions about what will happen (after a long time, prices will be 20% higher; in the short run people will devote more resources to producing traded goods instead of traded goods). Because this is basically irrelevant to the workings of the real economy, my predictions will be both more accurate than market predictions as well as useless for making money in financial markets.
Such theories would also make predictions about how good of an idea it would be to transition to a different monetary regime (say a competitive currency regime).
(2) As I said before, if traders approximate the parts of your model that are directly applicable then you don’t have any useful information advantage.
(3) Sure, and there’s plenty to be skeptical of in mainstream macro, but that doesn’t imply unlimited skepticism.
Like Silas Barta, I have come to the view that a lot of macroeconomics is terribly confused. I have an ongoing mission to make sense of macroeconomics. My explanation of how most macroeconomic theories of macroeconomic fluctuations work at their core is here.
I’m not sure I understand your question, so if I’m answering past you, let me know. Anyway, if you came up with a model that predicted macroeconomic variables better than the marginal traders who trade assets strongly affected assets, yes you could make a killing. As I see it macroeconomic modeling is about producing theoretical understanding rather than producing predictions about the path. If we had really good and detailed prediction markets of all important macroeconomic measures, that would generate good predictions but it might not be very enlightening about what policy makers should do differently and why. In other words, modeling answers questions about what forces are at work while prediction about whats going to happen at some specific point in time is about what the net effect of those forces are.
Theoretical macroeconomics can help us understand what kinds of things we should try to measure, what kinds of general rules policy makers should adopt etc. A good predictive mechanism can help us determine what people should do right now (should your grocer expect an increase or decrease in sales), whether policy makers do a good job.
Does that clarify things?
Such “theoretical understanding” is as if you had a theory of physics that purported to “explain” past observations but was unable to make any predictions about future events. That is not science. At best it’s just empty philosophizing, and at worst pernicious bullshit used to rationalize actions out of touch with reality.
As for the relevant prediction markets, they already exist for all practical purposes. I don’t know much about finance or investment, but I do have at least a rough idea how any correct non-trivial macroeconomic prediction could be translated into a winner investment strategy.
Perhaps I wasn’t clear. I don’t think that macroeconomic theorizing is useful even if it predicts nothing. Obviously that’s useless. I think that macroeconomic theorizing is useful even if they don’t generally outperform market predictions of market forcasted variables because it gives you different kinds of information about the economy.
For example, imagine a world where relevant traders have very good models for predicting the value of the S&P500. However, these models are proprietary, detailed and heavily specialized for predicting the S&P500. Academic models are open, relatively simple and based on ‘first principles’.
In this scenario, how useful are these two types models? Depends on what kind of question you’re asking:
What will the value of the S&P be in 1 year? Trader models useful. Academic models useless.
What kinds of alternative monetary institutions would be better than current ones? Trader models useless. Academic models useful.
Perhaps you could formulate prediction markets to answer the second kind of question well, but I sure don’t know how and in any case is not currently something people do.
There are also plenty of non-trivial macroeconomic predictions that you cannot make money off of. For example, ‘employment will be 1 point higher in 1 year’ is non-trivial, but it could be information already perfectly incorporated into the markets. ‘printing money now increases employment and inflation for 1 year out’ is also nontrivial but may also be fully accounted for. If you don’t have access to markets that predict precisely this information, such forecasts can easily be useful.
Also, all relevant prediction markets do not exist. Examples: there are no very liquid unemployment prediction markets, there are no very liquid real GDP prediction markets etc. What is individually useful to trade is not necessarily the same thing as what is socially valuable to know.
It may well be that the existing trader models are useless for answering this question, but any novel model that is capable of answering them should ipso facto be able to provide useful investment information. There are endless public controversies over the expected effects of the current monetary policy, which have direct bearing on all sorts of markets, and if you can forecast their effects with more accuracy than any existing model, you should be able to beat the markets.
If you really know that employment will be one point higher in a year, there are straightforward implications on trading. For example, people who bet (possibly as a way of hedging their investments) that there’s going to be a very bad recession within a year are certainly wrong, and you can profit by betting against them. If I, a complete amateur, can easily think of such strategies, then I can only imagine what expert financiers could do with this information! It’s similar with predicting NGDP and all other aggregate variables.
I don’t see any logical possibility how some macroeconomic prediction could be at the same time: (1) meaningful and non-trivial, (2) more accurate than the state of the art, and (3) useless for investment.
Replying by paragraph:
3) sure; my point was that (1) does not imply (2).
1) I’m not clear on how you’re disagreeing with me, so this is mostly going to be a reworking of my previous answer. Let me know if you can clarify your disagreement.
Academic and trader models can be better at answering different questions. Neither one need always dominate the other. Academic models might be better at answering questions which rarely come up, for example ‘what happens when we change monetary institutions?’, so not offer useful investment advice except in those rare occasions but still be useful for deciding what options policy makers should consider.
Academic models can also provide much more understandable and accessible models of the economy (see 2). Non-traders may find these much more useful than trader models (which they may not have access to or be able to understand).
2) No. Remember I stated that this information was already taken into account by the market. NGDP, employment etc. are not random walks (Edit: I mean they are not martingales); their expected future value is not independent of past values. Unemployment is highly non random see here. Some macroeconomic variables are random walks, for example stock market indexes, but most of them are not.
I’m not sure I understand you either. Are you actually saying that if I had an oracle capable of telling me what, say, the rate of unemployment or NGDP growth will be in a year, it would not be possible to make investments with above-market returns using this information?
Moreover, I am at a loss trying to imagine a theory that would enable us to predict with reliable accuracy what would happen if we changed the monetary institutions in a given way, and which wouldn’t also enable us to get reliably accurate information on the uncertain and controversial questions about the consequences of the present monetary policies. This would also constitute valuable investment information. (Or do you think it wouldn’t?)
On the whole, the problem is that I simply cannot imagine any questions that macroeconomic theories purport to handle where a truly reliable and non-trivial information would not be valuable for investment.
Also, you can’t answer these questions by claiming that the relevant information has already been incorporated into the market prices, only in some obscure way that we now seek to disentangle. Many investments are specifically made in order to hedge against uncertainty in macroeconomic trends. If you have a theory that eliminates these uncertainties, or at least provides more accurate probability distributions, there’s a straightforward killing to be made there.
“On the whole, the problem is that I simply cannot imagine any questions that macroeconomic theories purport to handle where a truly reliable and non-trivial information would not be valuable for investment.”
One where the prediction is longer than your investment window.
1) It all depends on whether your predictions are better than or worse than the relevant traders. If the traders already have access to such an oracle, you won’t be able to make any money; if they don’t, you will. Many macroeconomic variables of interest (GDP, employment etc.) are not martingales which means that predicting their movements is not the same thing as predicting them better than relevant traders. Being able to predict an asset price difference (between now and some future time) and acting on that information tends to move the price to eliminate that difference now. Being able to predict say a difference in unemployment (between now and some future time) does not necessarily tend to move unemployment to eliminate that difference right now.
Perhaps the following is a more relevant example: lets that you and everyone else found out a couple of days ago that aliens are going to land on earth in a month. The variable aliens-on-earth (binary) is certainly an important macro variable. Naturally market prices currently reflect the fact that aliens will soon be among us. The current value of aliens-on-earth, however, is False and no amount of trading can change that. Aliens-on-earth is not a martingale; its current value is not equal to its discounted expected value and you can predict it quite well (aliens-on-earth(t) = {False for t < 1month and True for t >= 1 month).
2) I’ll give a concrete but extreme question: what would happen if the US moved to a uranium based commodity money system? Would it be good? bad? Trader based models are likely pretty useless for this because no one thinks this is likely to happen so there are few benefits to developing a model for it. Even if they did, you might not have a way to get at those predictions. However, you could get a rough idea about the consequences by building an economic model for this scenario taking into account the economics of money, estimates about the stock and potential supply of uranium, the costs of avoiding radiation poisoning during transactions etc.
Obviously this question is extreme, but questions like it are valuable: should we move to a competitive currency system (free banking)? what would that look like? How does an optimal central bank behave (for various definitions of optimal)? Are countercyclical unemployment insurance extensions welfare improving? Is countercyclical government spending welfare improving?
I have some opinions on the preceding questions including that some of them are obviously of little interest, but I only came to them because I learned some macroeconomics.
4) ‘There’s potential money to be made by selling rough macro models to say industrial producers’ is a fair point. However, it’s a lot of work to turn a model that says ‘keeping money supply constant, an increase in people’s desire to hold money produces recessions’ to something that might be useful to producers for planning. It’s the job of macroeconomists to come up with the former not the latter in the same way that it’s the job of government funded battery researcher to come up with basic theory related to batteries, or a computer scientist to come up with interesting and potentially useful classes of algorithms. If one of these researchers comes up with something really big that is immediately implementable, perhaps they will go off and start a company to take advantage of the idea, but generally they stick to basic theoretical research because thats whats easy and comfortable and where their advantage is.
(1) I know what martingale variables are, but I don’t see why the non-martingale nature of the macroeconomic variables is relevant. Clearly, if you have figured out a novel way to predict the coming of the aliens ahead of others (or even just to predict its timing and other details more accurately), you can get rich by figuring out how their coming will affect the markets. This is perfectly analogous to a theory that will predict various macroeconomic variables more accurately than the state of the art, since these variables have predictable effect on asset prices. (In fact, once you have this information, they are no longer martingales for you, since e.g. if you know a recession is coming withing a year, the expected trend for countercyclical assets is upward.)
Or to put it differently, from all that you have written thus far, I still don’t see a concrete example (either actual or hypothetical) of the thing whose existence you assume: macroeconomic predictions that are interesting, novel, accurate, and at the same time useless for beating the markets.
(2) I understand that there are hypothetical questions about monetary systems where an accurate answer would have no practical implications by itself. However, presently we are in a situation where there are deep and bitter disagreements even about the predicted consequences of the ordinary and standard monetary policy options. What I find implausible is that one could obtain correct answers of the former sort without a theory that would at the same time be able to give more accurate answers to questions of the latter sort (which would again translate into investment information in a straightforward way). It would be as if you had a theory of mechanics capable of predicting the motions of hypothetical planetary systems but of no use for practical technical problems.
(3) Regarding your point about theoretical vs. applied research in other areas, the same heuristic actually is widely applicable. Whenever you see people doing research into something that should have straightforward practical applications, but you don’t see them running to monetize the results, something fishy is likely going on.
Of course, sometimes there is real insight that can’t be monetized in any obvious way, like for example fundamental theoretical physics. However, there is an essential difference here. A physical theory can make predictions only about things that are of no business interest, so in fact you have to spend money to contrive experimental setups to test it. In contrast, anything that a macroeconomic theory might be making predictions about and that might actually occur in the real world is inherently of business interest. (And again, if you have a counterexample, I’d be curious to hear it.)
For some reason I feel compelled to return to this topic:
My point is not that macroeconomics is a great field filled with great insights (it’s not and most macro theorists are terribly confused) but that it’s not as ridiculous as you seem to imagine it that some economists have novel, interesting and true things to say about inflation, unemployment, GDP etc and are not themselves fabulously wealthy.
(1) For example, some macroeconomic theories predict behaviors like the parable of the babysitting co-op. You can also run experimental economies (like this) and make predictions about the behavior of the economy.
I might set up a play economy where different people produce different goods and trade and consume them. Money is traded but not produced. We let this economy do its thing for a while and then suddenly (and without announcing in advance) give everyone 20% more money. Using my favorite macro theory I could make a number of interesting and novel predictions about what will happen (after a long time, prices will be 20% higher; in the short run people will devote more resources to producing traded goods instead of traded goods). Because this is basically irrelevant to the workings of the real economy, my predictions will be both more accurate than market predictions as well as useless for making money in financial markets.
Such theories would also make predictions about how good of an idea it would be to transition to a different monetary regime (say a competitive currency regime).
(2) As I said before, if traders approximate the parts of your model that are directly applicable then you don’t have any useful information advantage.
(3) Sure, and there’s plenty to be skeptical of in mainstream macro, but that doesn’t imply unlimited skepticism.