This post is inspired by two observations:
- Phil Pilkington is onto something important in recent posts regarding the utility of macro-economic modeling. Specifically, he observes that macro-economics is an open system in which precise experiments and firm conclusions are impossible.
- Personally, I have observed good economic discussions devolve into overly wordy, time-wasting theoretical slogs. In my experience, this happens when either:
--Tangential economic models are introduced, and the frame of reference for the discussion is changed inappropriately.
OR
--The discussion is based upon unrealistic assumptions.
Inappropriate Use of Math and Statistics
Pilkington's posts:
- Abstraction, Language and Modelling in Economics
- Econometrics and Numerical Prediction
- A Quick Note on Michael Emmett Brady’s Paper on Keynes and Probability
- Disturbing Distributions in Economic Statistics
- On DSGE and the art of using absolutely ridiculous modeling assumptions
and a few quotes (color changes are used to distinguish separate quotes from various posts):
This is where economics has erred since at least the turn of the 19th century. The early marginalists occupied two groups. One were the Walrasians who, following Leon Walras, were perfectly content to confine themselves to barren speculation of unrealistic nonsense provided it was done in a nice, formal mathematical manner. The other group were the Marshallians who tried to bring such abstract speculation down to earth...
A good example of a closed system is a controlled scientific experiment. By setting the experiment up so that it is continuous through time (ergodic) and is not interfered with by outside forces, the experimenter ‘closes’ the system upon itself. For realists, any data then generated by this experiment can reliably be used to make inferences about the future.
An open system, on the other hand, is open to change, fluctuation and new trends emerging. It is also not closed to outside forces interfering. The realists think that open systems are what we generally deal with in the social sciences, including economics. We cannot reliably use data generated in such open systems to make predictions about the future because, for example, although inflation and wages may be strongly correlated over a certain time period they may not be in the next time period...
Unfortunately, reality is staring me in the face, and it’s telling me that we don’t need more complicated models.
If I go to the trouble of fixing up a model, say by adding counterparty risk considerations, then I’m implicitly assuming the problem with the existing models is that they’re being used honestly but aren’t mathematically up to the task.
If we replace okay models with more complicated models, as many people are suggesting we do, without first addressing the lying problem, it will only allow people to lie even more. This is because the complexity of a model itself is an obstacle to understanding its results, and more complex models allow more manipulation …
Timewasting Discussions
There are two specific topics which drive me crazy:
- Any discussion about the IS/LM model. Krugman is a big proponent of this, but he generally labels his columns discussing IS/LM as "nerdy". I have concluded that this is because the IS/LM model only makes things much more complicated than they need to be.
- Any discussion with Market Monetarists. These inevitably start with the unrealistic assumption that the "Fed" can do whatever it wants in terms of controlling the economy. Managing expectations by making pronouncements is generally the means by which they can exercise this power. Any conversation discussing such a mythical economy goes in circles.
Here are some examples:
No: Saving Does Not Increase Savings
Here, Asymptosis attempts to clarify the various senses in which the terms saving and savings are used in macroeconomics. All goes relatively well until he throws in this:The IS/LM model seems to be inescapably based on the misconception detailed above — that more saving results in more savings hence, because of supply and demand for loanable funds, lower interest rates. But: if Krugman’s constantly repeated assertions are correct, that model seems to perform very well. Why is this true? What am I not understanding?At this point, the focus of the discussion shifts from What do the terms saving and savings represent in macroeconomics? to How does the IS/LM model work? The latter question is perhaps worthy of a separate post, but only detracts from the main topic of the current post.
Terminal Demographics
Here, Interfluidity attempts to discuss the causes of inflation in the 1970s. He engages with several Market Monetarists, and the results are an extremely long discussion that will make your head spin. The problem is that the Market Monetarism uses 100 sentences where 1 will do. Here is an exchange I had with a Market Monetarist in the comments of the referenced post:
Me: To say that the macro-economy can or should be managed through this one tool (interest rates) is not reasonable. We could just as well say that the inflation of the 1970s could have been prevented by raising taxes, or decreasing public expenditures, or wage and price controls, or breaking OPEC, etc…
Market Monetarist: True, fiscal policy was expansionary in fiscal years 1964 through 1968. The cyclically adjusted Federal budget balance was reduced from (-0.5%) of potential GDP in fiscal year 1963 to (-4.4%) in fiscal year 1968, and the deficit was increased annually during that time:http://www.cbo.gov/sites/default/files/cbofiles/attachments/43977_AutomaticStablilizers3-2013.pdfBut a 10% income surtax was enacted in 1968 and remained effective through 1970. The cyclically adjusted budget balance rose to (-1.1%) by fiscal year 1970 and remained in the relatively narrow range of (-2.7%) to (-1.3%) from fiscal year 1971 through 1982. In fact despite the image of a deficit prone decade the 1970s were one of the most fiscally responsible decades on record with gross Federal debt setting a post WW II record low of 32.5% of GDP in fiscal year 1981 (President Carter’s last budget).
“…or wage and price controls,…”
Wage and price controls were in effect from 1971Q3 through 1974Q1, and core inflation did fall from an average of 5.0% in the year before they were implemented to 3.1% during Phases 1 and 2. But as they were relaxed it bounced back up. During Phase 3 and 4 it reached 4.2% and 6.1% respectively. And in the year after they were ended core inflation averaged 10.1%. Wage and price controls interfere with relative price adjustments ensuring they will be abandoned and that aggregate inflation will return with some catch up inflation to boot.
“…or breaking OPEC, etc.”
Total energy expenditures as a percent of GDP rose from 8.0% in 1970 to 13.7% in 1980, a change of 5.7 points.
http://www.eia.gov/totalenergy/data/annual/pdf/sec1_13.pdfThe EIA doesn’t have total energy expenditure data from before 1970 but the price of crude petroleum in 2005 dollars was $4.46 per barrel according to the World Bank dataset and this was less than in any year in 1960-69 and was down from 26.6% from its price of $6.08 a barrel in 1965:
http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTDECPROSPECTS/0,,contentMDK:21574907~menuPK:7859231~pagePK:64165401~piPK:64165026~theSitePK:476883,00.html
So chances are very good that total energy expenditures in 1970 as a percent of GDP had fallen from the level they had been in 1965 and yet core inflation had had already risen from 1.3% in 1965 to 4.7% in 1970, and continued to accelerate reaching 9.2% in 1980:
http://research.stlouisfed.org/fred2/graph/?graph_id=109579&category_id=0In contrast total energy prices as a percent of GDP rose from 5.9% in 1999 to 9.9% in 2008, an increase of 4.0 points, and yet core inflation only rose from 1.3% to 2.3%.
So in the Great Inflation a change in total energy expenditures of 5.7 points resulted in an increase in core inflation of 7.9 points and in the 2000s a change in total energy expenditures of 4.0 points resulted in an increase in core inflation of 1.0 points.
This shouldn’t be surprising because research shows that commodity price increases are not an important causal factor in long-term inflation:
http://www.bostonfed.org/economic/ppb/2011/ppb111.pdfDo Commodity Price Spikes Cause Long-Term Inflation?
Geoffrey M. B. Tootell
May 2011
Abstract:
“This public policy brief examines the relationship between trend inflation and commodity price increases and finds that evidence from recent decades supports the notion that commodity price changes do not affect the long-run inflation rate. Evidence from earlier decades suggests that effects on inflation expectations and wages played a key role in whether commodity price movements altered trend inflation. This brief is based on a memo to the president of the Federal Reserve Bank of Boston as background to a meeting of the Federal Open Market Committee.”…
This is not a productive discussion. It's more like a filibuster, and fits a dysfunctional pattern I've observed.
26 comments:
Scott Sumner’s reason for believing in the ineffectiveness of fiscal policy is completely mad. His reason is what he calls “monetary offset” which he sets out in this article:
http://mercatus.org/publication/why-fiscal-multiplier-roughly-zero-0
His argument is as follows. The Fed targets a particular level of inflation, unemployment, NGDP growth, or whatever. Thus if any fiscal stimulus is implemented, the Fed will take note and negate that stimulus. Ergo, fiscal policy is useless.
Well in the recent recession we had a glaring example of Sumner’s “monetary offset” falling flat on its face. That is, the recession arrived, and some fiscal stimulus was implemented. But the Fed, far from NEGATING that stimulus via interest rate increases or whatever, did the opposite: that is the Fed SUPPORTED that fiscal boost with interest rate cuts and QE.
To add insult to injury, there is another fundamental flaw in Sumners “market monetarism” which I’ll put on my blog in a day or two. But I’m tired at the moment having just spent the day at a meeting promoting full reserve banking (readers will be horrified / pleased to hear – take your pick).
Dan, it's true that the IS-LM is often taught quickly without regard for real world institutional realities. But many have adapted the model to account for these realities. For instance, David Romer has shown you can adapt the model to account for the fact that the central back targets an interest rate, not the money supply. Not that this was necessary. I'm aware of a few articles dating back to the 70s that use IS-LM with an endogenous money supply.
I'm curious to know more about your problem with the model
Ha! I just noticed that the topic of IS-LM is under the heading 'timewasting discussions'.
I'm curious to know more about your problem with the model
How do you get a model to work that says quantity demanded for money is inversely propositional to the price of money when it's directly proportional across the entire cycle, quantity demanded low at ZIRP and high when the economy is booming.
To justify it, you have to limit the data to that which conforms to the theory, which, of course, is a special case. But how does one then distinguish when the data no longer fits and another model has be to adduced.
Of course, it can be explained away with negative rates not available at the lower bound and citing Volcker jacking the rate into the high double digits to quell inflation by tanking the economy. Or claiming that such cases should be ignored as too far out in the tail to be concerned with and ISLM "usually" works.
This would seem to indicate that interest rates are much less relevant than effective demand and not the all-powerful tool that ISLM suggests.
Krugman prefers ISLM but his cross is actually a more accurate representation according to Parenteau, Fullwiler and Kelton.
What's at stake is monetarism v. fiscalism. Krugman is only willing to admit the effectiveness of fiscalism at ZIRP. Otherwise, it's monetarism all the way.
Tom, all you need to do to make it a fiscalist model is to make the IS curve vertical. This would be consistent with MMT. One policy implication is that govt spending (shift to the right) need not increase interest rates. Another is that the central bank can't affect aggregate demand.
Circuit,
but it you have a Rube_Goldberg machine, why bother adding hoops to make it work?
[in Australia, you may be more familiar with Heath Robinson than Rube Goldberg]
Why not just simplify to the arbitrary impositions of reality, as MMT does?
We can make this difficult & abstract, or make it lean & real.
Occams Razor should apply?
I suppose all I am saying is that the ISLM is a pretty effective way to communicate the message. The basic model has only two equations, which is unheard these days in economics. And I think you can get some pretty decent insight. I actually like the version with a vertical IS and horizontal LM. It summarizes some key elements. The variant by Romer is also pretty good because it incorporates the central bank's reaction function (with endogenous money).
I've never had any luck whatsoever going into a WalMart and shouting "ISLM"
Something more linked to context does work, however.
Since economics studies actually existing societies that can't be manipulated into an experimental setup, the empirical confirmation of the hypotheses economists consider is much more challenging than in some other sciences. Also, many economic phenomena depend on continuously evolving, high-level, complex social institutions and human agents, not fundamental particles that have only a few degrees of freedom and do not change their properties over the time scales of interest.
So economics shouldn't be regarded as a search for deep, unchanging laws of economic nature, but as the study of prevailing patterns and regularities in systems closely approximating the ones we currently live in. As institutional structures evolve, our economics needs to evolve along with it.
That said, there is ample room within such a field of inquiry for sophisticated data collection and data analysis, and for the use of mathematical techniques in teasing out patterns of causal dependency.
And again: everybody uses some kind of model. One cannot make any prediction or make any kind of sensible policy recommendation without employing some kind of model of how the economy works.
Oops...Just to correct my comment at 5:27 above. The horizontal LM curve allows for govt spending w/o rising interest rates. The vertical IS curve means the CB can't affect aggregate demand.
Good point Dan K about models.
Oops...Just to correct my comment at 5:27 above. The horizontal LM curve allows for govt spending w/o rising interest rates. The vertical IS curve means the CB can't affect aggregate demand.
Good point Dan K about models.
I agree with the Dan Kervick comment. Take for example, fiscal stimulus during the crisis. Let's say someone suggested $10 million dollars worth of stimulus was enough to pull the US out of recession, since $10 million is a lot of money (cue Dr. Evil jokes).
Well, it is obviously not enough. Well, why not? You need to invoke some form of mathematical model to say that $10 million is not enough, even if is just hand-waving about the size of GDP.
Right Brian, that's the kind of thing I was thinking about. Note that if one goes all the way into radical skepticism about the future resembling the past in predictable ways, rational policy choice of any kind becomes pointless.
The charge of "no model" leveled at heterodox economists is equivalent to no model that we specify. Of course, heterodox economists have models, at minimum conceptual models, and monetary economics has Godley SFC accounting-based models.
The so-called Keynesians are trying to "perfect" Keynes with econometric models that Keynes presumably was not able to write, even though he was trained as a mathematician.
Keynes gave very specific reason why he rejected the econometric approach that the mainstream requires to be considered a VSP (very serious person). The so-called Keynesians like Krugman have ignored this.
BTW, see Roger Farmer, Faust, Keynes and the DSGE Approach to Macroeconomics
I don't think that radical skepticism is required. Emergence and reflexivity are sufficient in the case of complex adaptive system. Econometrics doesn't model muti-agent complex adaptive systems subject to emergence and reflexivity that are characterized by ontological variability and epistemic uncertainty.
See agent-based model at Wikipedia and compare with DSGE.
Detroit Dan good point that Philip brings up about this being an open system... this will definitely effect my thinking going forward...
Roger,
You write: "impose reality" and yes we need to do that (probably first imo) to create the "conditions" needed...
But THEN people have to actually make something happen, and humans use models often to plan a course of action...
The establishment of the "conditions" part is limit of the "Darwinian" or "evolutionary" part of this iow people have to know what ALL of their "options" are...
BUT THEN, the "intelligent designers" have to go to work to "create" and regulate the system, this part does not "happen by itself" , somebody or a small group of policy makers/regulators have to make it happen ... models are usually employed by humans in this "creative" process...
Dan K has been trying to get us thinking about such a model... iow if we ever get the "conditions" established, then what do we do? And btw please BE SPECIFIC and have a decent rationale type of thing...
How much fiscal is enough? Congressman: "I need a number, etc... "Give me a number, etc...."
Can we adequately answer these questions?
Right now we are going to school on a huge one month fiscal "helicopter drop" of over 100B coming in from the govt to the non-govt albeit in the form of Tax Refunds and are assessing the real impacts of this substantial impulse (+25%/month) in fiscal wrt what happens in the economy wrt gdp/employment/corporate sales/earnings/price stability / etc.... we are trying to develop a rudimentary model ... with all of the $bazillions in research funding we have ;)
With this "experiment" and quantifiable "observation" behind us, we are in a better position to make a specific quantitative fiscal policy recommendation, etc...
So looks like a two step process to me anyway.... "conditions" first, then be ready with a specific regulatory plan of action which would as usual be based on some sort of "model"...
rsp,
One can build a mathematical model of a complex adaptive system. One can also build a mathematical model of a complex, adaptive system in which some agents issue predictions about the future course of the system, and those predictions are reflexively fed back into the system to alter the behavior of the system. And one can build mathematical models of complex, adaptive, reflexive economic systems in which some agents are afflicted by radical uncertainty about key parameters in the system, and in which they respond to that uncertainty in varying ways. All of this is a wide-open field for economic investigation assisted by mathematics. Developing such technique might give us new insights into the interpretation of various episodes in economic history, and help us understand why some phenomena are relatively stable over long periods of time while others exhibit great instability.
My view on DSGE models is that they are doing the math wrong. From where I sit, if the math is done wrong, that's the end of the discussion for a mathematical model.
The issue of complex models, like an agent-based model, is their complexity. This may be more realistic, and if I were to build a video game which has an economy in it, I would probably have a model like that.
But for the purposes of policy analysis, there is a big problem. A clever person can tweak a complex model to get whatever response they want (within limits). Since the model will have hundreds of unobservable parameters, it is nearly impossible to find which is the best model within the class of models you are using.
I have my doubts about predicting the economy with a mathematical model (probably for the reasons identified by Keynes, although I have not read his explanation). The trajectory of the economy is driven by unobservable "animal spirits". But some simpler models should give us some useful insights, even if they have to embed assumptions about some of these unpredictable factors.
I agree with Brian. Yes, all evidence suggests that all humans reason by analogy, i.e., by models.
Yet "reasoning," models & analogies are simply tools, not reality.
There is a prior, functional history that precedes all the tools, practices and methods ever invented.
That method is to simply titrate all models/analogies to reality as reported by accumulating feedback.
Most of our problems seem to be the frictions involved in over-testing models instead of simply following the "moment" of all available feedback.
The adjustment of all models eventually occurs as new levels of feedback accumulates, from new levels of sensory instrumentation, and new levels of analysis, testing and assessment.
Never forget that it's the models that have to adjust to reality, and the SOONER the better.
Biggest failure of all infatuation with modeling? It slows tempo.
If people were willing to adjust their models in real-time, titrated to every subtle change in total feedback, this issue would go away.
It's tempo that's killing us.
As always, Adaptive RATE is the issue.
If data, outcomes & models all adjusted at the same (instantaneous) rate, we'd never have to worry about Democracy, right? :)
By definition only an omniscience being could accurately model (explain with predictive accuracy) a complex adaptive system, because of emergence, which is not foreseeable from the data.
If you've been in military operations or studied it, you realize that almost all strategic planning is contingency planning based on hypothetical scenarios dependent on enemy action that is unknowable before hand.
Surprise gives a huge competitive advantage — think D-Day at Normandy. Once the Allies gained not only air superiority but also a beachhead, it was essentially game over. The fact that economists were surprised by the global financial crisis and have developed no counter-strategy capable of meeting the challenge effectively speaks volumes about economics as a purported "science."
It's only in the natural sciences that robust predictive models can be constructed that predict future change necessary to put a man on the moon, for instance. It's a huge challenge but doable.
When we move to explanation in the life sciences, emergence emerges and takes center stage in evolutionary biology. And when we move to the social sciences reflexivity also emerges along with non-ratonal aspects of individual and group behavior (groups influence individuals, as in group-think). No biologist thinks we can write a set of equations that allow us to explain emergence of variation in evolution, and no social scientist thinks we can model reality accurately enough in the social sciences to overcomes what we philosophy calls "human freedom" as rubric for behavioral variability.
Clearly, models that resemble those of the natural sciences will be of limited use in the life sciences, at least until humans gain a whole lot more knowledge of molecular biology, but even then they can only specify what will happen contingently given certain conditions coming together randomly.
When we move on to the social sciences, the data becomes much more complex owing to the flexibility and reflexivity of human consciousness and consequent variability of human behavior that we call freedom. That is to say, in the engineering sense of freedom humans have an almost unlimited degree of freedom potentially, and companies like Google are putting their money where they think that opportunities that this implies lie, e.g., by hiring Ray Kurzweil as chief scientist in order to explore the frontier of knowledge and technology.
It looks like to me the world is moving forward but conventional economics is still stuck largely in the 19th century and early 20th at best with their thinking and modeling. At least this is ahead of the 17th and 18th century thinking of most politicians. And when one examines the polling about what a significant portion of American public believes, it's more or less medieval in many respects, especially wrt to science. 25% of the people still think that the heavens revolve around the earth, as it appears to them in the sky. Etc.
Wrt to conventional economics, physicist Mark Buchanan recently commented on the dispute over DSGE that Noah Smith initiated by asking why firms don't use DSGE modeling if it is so great:
Knowledge really is power. I know of at least one financial firm in London that has a team of meteorologists running a bank of supercomputers to gain a small edge over others in identifying emerging weather patterns. Their models help them make good profits in the commodities markets. If economists' DSGE models offered any insight into how economies work, they would be used in the same way. That they are not speaks volumes.
Thanks for all the excellent feedback. There were many good points in the comments. This one by Brian stands out for me:
But for the purposes of policy analysis, there is a big problem. A clever person can tweak a complex model to get whatever response they want (within limits). Since the model will have hundreds of unobservable parameters, it is nearly impossible to find which is the best model within the class of models you are using... But some simpler models should give us some useful insights, even if they have to embed assumptions about some of these unpredictable factors.
In conversations with people who have differing world views, the models often get in the way. With Austrians and Market Monetarists, for example, I have very basic disagreements, and in such cases it might help to go back to some common basis of understanding...
To paraphrase Ronald Coase, If you torture the data enough, you can get it to confess to anything.
Aren't there ANY humans left that have NO a priori models .... and instead rely ONLY on unpredictable reality?
??
Post a Comment