Wednesday, January 14, 2015

What does "structural" mean?

There are a million and one definitions of the word "structural." To a builder or architect, the meaning is straightforward: "of, relating to, or forming part of the structure of a building or other item." But that's just the first dictionary definition. The second is "of or relating to the arrangement of and relations between the parts or elements of a complex whole".

In other words, you can use "structural" to mean just about anything you want, as long as you're talking about a complex system. Here are some examples of how people in different walks of life use the word "structural":

1. "Structural estimation"

Econometricians sometimes use a technique known as "structural estimation." This term is usually used in contrast to "reduced-form estimation." It's very difficult to pin down exact definitions of "structural" and "reduced-form" - even economists often don't know where the boundary lies. In general, "reduced-form" models tend to be either linear, or simple stuff like probit or logit. "Structural" models tend to use more complicated functional forms that represent "underlying" economic relationships. The implication of using a "structural" model is that your model can describe a wider array of situations, but this is just sort of implied. "Reduced-form" is often used as a diss, but that's not always fair - after all, if the assumptions that make your model "structural" are wrong, your model will give you a false sense of confidence. 

2. "Structural Poverty"

Advocates for the poor usually don't like the idea that poor people could be non-poor if they altered their behavior. They claim that poverty is the unavoidable result of the way the economy, society, and politics are set up. Being screwed by the system is what they label "structural poverty".

3. "Structural VARs"

In time-series econometrics, you have "reduced-form vector autoregressions," in which the shocks (random stuff) are things you can observe. Unfortunately, because you can observe anything you like and call it a "shock", the shocks in your VAR model will inevitably be correlated, which limits the applicability of the model. So if you want a VAR that will work in all cases - or at least, a much wider set of cases - you need to estimate the "structural" shocks. Since you can't observe these. you have to make some assumptions about how they interact to form the reduced-form shocks.

4. "Structural Deficits"

This refers to government deficits that are unrelated to the business cycle. It is a term used by deficit hawks to mean a deficit that is very dangerous, because it will just keep getting larger unless there are major political changes. A related concept is structural unemployment, which is unemployment that doesn't change with the business cycle.

5. "Deep Structural Parameters"

This is a word that economists use to mean things that don't change in response to economic policy. For example, followers of Robert Lucas often argue that preferences and technology don't respond when policy changes, and should thus be treated as "deep structural" parameters in economic models. In practice, anything will probably change at least a tiny bit in response to some kind of policy, so the assumption of "deep structuralness" will always be an approximation.

6. "Structural Pluralism" 

This is...OK, I don't understand what this is. It's a term used by sociologists. According to this article, "Structural pluralism is defined as the degree of differentiation in the social system along institutional and specialized interest group lines, in a way that determines the potential sources of organized social power." I think that means that when a society has different types of organizations - e.g. religions, unions, corporations, etc. - that all have some sort of influence, you have structural pluralism. But I'm not sure.

7. "Structural Integration"

This is a kooky-sounding type of physical therapy. But it might work!

8. "Structural Assimilation"

This is when minorities have equal access to public institutions. How this could be observed and verified in practice, I'm not sure...maybe you could use a structural model.

9. "Structural Racism"

This is when "dynamics" exist that discriminate against minorities. For the meaning of "dynamics," you will have to wait for a different blog post.

10. "Structural Realism"

This is a philosophy, possibly part of epistemology or possibly part of ontology (I'm never sure where one ends and the other begins), that, according to the Stanford Encyclopedia of Philosophy, "is considered by many realists and antirealists alike as the most defensible form of scientific realism." Further down in the article, we learn that "Structural realism is often characterised as the view that scientific theories tell us only about the form or structure of the unobservable world and not about its nature." Basically, it seems to mean that there's stuff we can't totally know about, but that science can let us know some stuff about that stuff.

So what does "structural" really mean? Looking at all these examples, it seems to mean "stuff I can't directly show you, but which I'm pretty sure isn't going to go away or change any time soon." So basically, "structural" = "stable" + "unobservable". Structure is the man behind the curtain. Does that sound right?

Except for structural integration, of course. I don't know how to fit that one in.


  1. Isn't the "man behind the curtain" the Wizard of Oz?

  2. Take "Structural" out and then discuss each of the ten. See how long can you write on each without using "structural"? Also, there is an etymology site, look up "structural", it might help.

  3. Advocates for the poor usually don't like the idea that poor people could be non-poor if they altered their behavior. They claim that poverty is the unavoidable result of the way the economy, society, and politics are set up. Being screwed by the system is what they label "structural poverty".

    The left also doesn't like it if you say they are poor because of biological factors, such low IQs. The only acceptable explanation is to blame the rich

    1. The first paragraph is one that Noah wrote to represent the view that some believe the poor could better themselves if they changed their behavior. Advocates for the poor claim that poverty has untoward effects on the poor, such as lack of opportunity, stress that alters brain chemistry and not for the better, and malnutrition with similar ill effects. These structural aspects of poverty are well studied, well understood, yet generally discarded by moralists.

      I learned of these effects of poverty by teaching in high-poverty areas; I even lived in high-poverty areas and saw the stress, malnutrition, and the trap of poverty first-hand. I lived in a trailer park for 16 years and studied my students' environment close up. The studies are true; moralizing and bible thumping doesn't repair the broken brains and broken wills of the dispossessed. Education does some good We had our successes. Many of my colleagues and I worked hard. Some of the poor are too beaten down to make it back. It is, quite observable, generational in its reach.

      Your final paragraph is apparently your only original work. It contains no truth, not one, only snark. It adds nothing to the debate. Come back to the discussion when you've purged yourself of your predispositional bile.

    2. "The left also doesn't like it if you say they are poor because of biological factors, such low IQs. The only acceptable explanation is to blame the rich"

      This is rather confusing. Both the right and the left recognize that many developmentally disabled are poor because of physical disabilities, and I don't know of anyone who would deny this. The difference between the right and the left is not one of being able to or unable to identify the physical causes of poverty, but of being able to develop compassion for the effects of poverty.

      When you have compassion you try to do something about the effects whatever the causes may be. When you lack compassion you try to not do anything about the effects. As a result, those with compassion do not have to "blame the rich" to provide a rationalization for trying to do something about poverty, whereas, those without compassion have a tendency to try to rationalize the problem away by blaming the poor for the situation in which they find themselves even when that situation is totally beyond the control of those who are poor whether it arises from a physical disability or from the circumstances in which they were born.

      It seems quite clear to me that compassion marks the difference between the right and left in today's world, not the ability or lack thereof to identify physical causes for poverty. And while the right does find it necessary to blame the poor for the circumstances in which they find themselves, however those circumstances may have arisen, the left does not find it necessary to blame the rich those circumstances.

    3. Anonymous9:26 AM

      "Structural" may be the word du jour, but "compassion" as used in this post could use some scrutiny. Specifically, is there much connection between whatever effort is sufficient to gratify or satisfy the feelings of the compassionate, and actions that so alter the circumstances of the recipient of the compassion, that they are no longer require it? If you are content with ineffective remedies, how compassionate are you?
      Even if the world is as Manichean as described, this question is, for some of us, vexing and unsettled.

    4. "The left also doesn't like it if you say they are poor because of biological factors, such low IQs. The only acceptable explanation is to blame the rich"

      Google ideas such as 'being poor is expensive'.

  4. "Structural" pops up in Physics all the time too. e.g.

    I suppose it means effects associated with the position of individual atoms, but its a bit vague.

  5. In philosophy, one meaning of structural, the one you refer to, holds that language, including formal language, can only describe states of affairs, that is, how things stand and permute, and it can assert that things do so stand in particular cases, These cases then become a testable hypotheses to the degree that what is asserted as fact is mapped specifically enough for observation. That is, those conducting an experiment know exactly what qualifies as a result.

    However, language is not capable of either confirming what is assert outside of observation, which is what distinguishes science from philosophical speculation. Language can assert that a possible state of affairs is actually factual, but it cannot provide an empirical warrant for the truth of the fact. The best language can do is provide a logical pedigree, for example, in the form of a theorem.

    In addition, language cannot say what actually is "in itself" or in reality. For example, physics provides a general description of what various forms of energy do. But physics doesn't tell us what energy is, other than the rather unenlightening "power to do work." How "metaphysical" is that?

    In other words, even the most precise language of science does not and cannot answer the really interesting questions in ontology and epistemology, such as what is there and what can be know about what really exists. This issue became paramount when science discovered how perception and conception take place.

    Observable phenomena are produced as effects of light, sound, etc, on a nervous system. Phenomena are structured cognitively -affectively through brain function that is species determined. Kant attempted to answer this through cognitive structures that output human experience and knowledge, but this does not explain the "things in the themselves" assumed to be be the cause of sense data.

    Very generally speaking, realists hold that "reality" is experienced directly and that existence is known through intuition. Idealists hold that experience is reality and that being and knowledge are one. Skeptics hold that all we can know is our own experience and nothing that lies outside of it, "causing" it. While realists and idealists hold that knowledge is substantial, skeptics hold that knowledge is only structural in the above sense rather than substantial.

    The structure of language is assumed to approximate the structure of phenomena closely enough to enable description. However, what is described is simply relationships of data and not the "reality" of what may be causing the data. Humans don't know directly the objects that reflect the light through which objects are seen, nor do they know what light actually is, only its effect on the nervous system.

    A bit terse, but I hope you get the drift.

  6. Britonomist7:52 PM

    How about this: a relationship between two variables is structural if this relationship is invariant to external shocks. This seems like it could broadly apply to the cases used in econ.

  7. How about structural linguistics and structural anthropology. There was "structuralism" in Paris before "deconstruction" which came before "postmodernism".

    I think that 1 structural estimation and 5 deep structural parameters are the same or at least very close. Structural estimation is estimation of structural parameters. The "deep" was a change based on the assertion that tastes and technology are structural while other things which had been treated as structural weren't.

    The pre-lucas definition of structural parameters was of parameters which describe causation not just correlation. the difference between reduced form and structural is that structural models are "identified" which means the direction of causation can be determined. Nowadays, causation is usually studied either with natural experiments or with real experiments. Oddly, if the models with convincing identifying assumptions don't include utility maximization under constraint, then the people who estimate them say they aren't structural. this is a genuine change in the usage of the word which bothers me,

    My usual diatribe against Lucas below:

    You know the idea that technology is constant is a bit odd and rather hard to reconcile with technological progress. What we have here is a process which is too complicated to model and so exogenous to the model being treated as exogenous to the economy, that is, policy invariant. The logic is that if we don't understand something at all, we must make strong assumptions about it and then should rely on those strong assumptions. Tastes, of course, are much more mysterious than technology.

    1. Oooh, that brings in those Frenchies, like Claude Levi-Strauss. But, lots of them touted "post-structualism," so they just lke to mess up what is already confused.

      Barkley Rosser

  8. Noah sez: [Structural] seems to mean "stuff I can't directly show you, but which I'm pretty sure isn't going to go away or change any time soon."

    Structural poverty has been easy to show at least since Malthus. I like to illustrate it with the game of musical chairs: at the end, no matter how much training and competition are added, there is still only one person seated and the rest are all losers.

    Perhaps Noah should have said that "structural means that you need a model (rather than direct observation) to explain an idea."

  9. Phil Koop8:12 AM

    In general, "reduced-form" models tend to be either linear, or simple stuff like probit or logit.

    LOL! What does "in general" mean? Here it means "mostly true, but with exceptions." But a mathematician would use it to mean "true in all cases." E.g., "in general we cannot assume that a reduced-form default intensity process is tractable because we may wish to incorporate stochastic jumps, parameter regime switches, path-dependent drift terms, or other features that require numerical solution."

    I recall that Borges was fascinated by words which are their own antonyms, which he viewed as particularly common in English. "Cleave" was a favourite of his.

  10. Might also want to look at structuralism (and later, hyperstructuralism) in architecture

    General idea is that the form of the building should be evocative of the structure that holds it up but that no one can actually see.

  11. To me, "structural" refers to those elements of a complex whole which are themselves not changeable or subject to deformation, and which govern the gross form and capabilities of that whole.

    So, a birds wings are a form of the forelimbs of the tetrapods structural template, and could theoretically revert to arms, but that bird isn't going to grow another set of limbs or an extra head, or assume the underlying structure of a gastropod or an arthropod.

    In economics, structural elements seem to be often assigned by the theorist, without scrutiny to determine if they are really structural rather than just privileged.

    The truly structural elements would, I think, be determined by looking at what happens if they are reduced, deformed or removed.


  12. In Project Management, it is a key concept in connection with the approach of “divide et impera”

  13. Anonymous11:24 AM

    Structural largely means "contextual". It's doesn't necessarily mean "static" (or not dynamic), as it's used in economics.

    In computer science - my PhD field - contexts are highly dynamic in some senses, static in others, but whether or not structure is static or dynamic depends on what aspects one is considering. We use it most to describe data - in which case it's necessarily dynamic - and code - which is necessarily typically static.

    What I object to about how economics uses "structural" is that it's used as an antonym for "cyclical", which I find to be linguistic nonsense. There are characteristics of dynamism which are anything but cyclical - randomness, for one, which frankly characterizes something as arbitrary as economic phenomenon. I would in fact challenge the economics professional to demonstrate how what they call "cyclical" is truly cyclical in any objectively measurable (i.e., scientific) sense, instead of just random or simply immeasurable.

    1. Actually in Economics there is some debate as to whether "Business Cycles" actually exist. As opposed to sometimes things are up and sometimes things are down.

    2. Anonymous4:23 PM

      Yeah - you're making my point...

      Economics need different terms. A "business cycle" - it such exists - isn't by definition the _absence_ of structure, but the _presence_ of it. A cycle is a structure - a dynamic one; in the best case a sinusoidal one.

      What Noah is I think talking about it monolithic vs. composite structure, that latter being structures whose components, factors, aspects, elements, etc. can be fairly readily identified, as least for sake of discussion, and the whole of which comprises _context_ with reference to which such things relate, both in toto and relative to each other. Dynamism (more more generally, time-dependent behavior) is one such factor or aspect.

      Linguistically, the economic notions of structural and cyclical are simply orthogonal, and thus it makes little sense to treat them as antonyms.

  14. There is also "structural reforms", which seem to mean "serious reforms", which means that the person talking about structural reforms is very serious.

    It usually implies that the person wanting more structural reforms believes all unemployment is structural and all deficits are structural as well. Which is what serious people know.

  15. Anonymous2:42 PM

    In economics is so easy, structural = stocks

  16. Anonymous12:15 AM

    On the policy side, you forgot "structural reforms", a change in some policy stance meant to alter the functioning of the economy in some way. Note that policy change="unstable" & and policy implementation="observable".

  17. Anonymous8:00 AM

    To me, structural vs. reduced form is about how you achieve causal identification. If you achieve it by building a model, then that's structural. If you use IV, diff-n-diff, (natural) experiments, etc. then it's reduced form.

  18. is it possible for two objects to be in thermal equilibrium if they are not in thermal contact with each other. Explain

  19. Noah, you forgot 'structural unemployment', which in use means 'we had a recession, and unemployment went up, but I'll lie and declare it to be Punishment from Gawd, for failing to carry out my desired changes'.

  20. it is used often in systems science - structure of the system is the design of the system and it sure is invisible but footprints of it are visible everywhere

  21. Anonymous12:49 AM

    Is Noah Smith a misogynist? His twitter header photo features a white man with his zipper open while subjugating a woman

  22. Anonymous1:08 AM

    Without reading your post, I'll tell you what this term means to me.

    From an economic standpoint, as a fairly well informed person, structural unemployment implies a major problem in skills vs. employees.

    But in a broader sense, structural means that the result is out of the control of government.

  23. Also Structural unemployment - the quantity of labor supplied exceeds the quantity of labor demanded - the huge problem of Europe