Home > Uncategorized > Heckman on where causality resides

Heckman on where causality resides

from Lars Syll

James HeckmanI make two main points that are firmly anchored in the econometric tradition. The first is that causality is a property of a model of hypotheticals. A fully articulated model of the phenomena being studied precisely defines hypothetical or counterfactual states. A definition of causality drops out of a fully articulated model as an automatic by-product. A model is a set of possible counterfactual worlds constructed under some rules. The rules may be the laws of physics, the consequences of utility maximization, or the rules governing social interactions, to take only three of many possible examples. A model is in the mind. As a consequence, causality is in the mind.

James Heckman

So, according to this ‘Nobel prize’ winning econometrician, “causality is in the mind.” But is that a tenable view? Yours truly thinks not. If one as an economist or social scientist would subscribe to that view there would be pretty little reason to be interested in questions of causality at all.  And it sure doesn’t suffice just to say that all science is predicated on assumptions. To most of us, models are seen as ‘vehicles’ or ‘instruments’ by which we represent causal processes and structures that exist and operate in the real world. As we all know, models often do not succeed in representing or explaining these processes and structures, but if we didn’t consider them as anything but figments of our minds, well then maybe we ought to reconsider why we should be in the science business at all …

The world as we know it has limited scope for certainty and perfect knowledge. Its intrinsic and almost unlimited complexity and the interrelatedness of its parts prevent the possibility of treating it as constituted by atoms with discretely distinct, separable and stable causal relations. Our knowledge accordingly has to be of a rather fallible kind. To search for deductive precision and rigour in such a world is self-defeating. The only way to defend such an endeavour is to restrict oneself to prove things in closed model worlds. Why we should care about these and not ask questions of relevance is hard to see. As scientists, we have to get our priorities right. Ontological under-labouring has to precede epistemology.

The value of getting at precise and rigorous conclusions about causality based on ‘tractability’ conditions that are seldom met in real life, is difficult to assess. Testing and constructing models is one thing, but we do also need guidelines for how to evaluate in which situations and contexts they are applicable. Formalism may help us a bit down the road, but we have to make sure it somehow also fits the world if it is going to be really helpful in navigating that world. In all of science, conclusions are never more certain than the assumptions on which they are founded. But most epistemically convenient methods and models that work in ‘well-behaved’ systems do not come with warrants that they will work in other (real-world) contexts.

  1. yoshinorishiozawa
    July 23, 2022 at 8:55 pm

    Heckman is in a special sense right when he says that “causality is in the mind.” In most of econometric models, causality is only assumed to exists behind the phenomenon that the model builders expect to investigate. However, relations imagined to exist in econometric models are only assumed to be causal in the mind of model builders. There is no real causality in their models.

    There is little possibility that we can construct a macro econometric models that reflect real causality because of, as Lars Syll points, complexity, intractability, and heterogeneity of economic agents. This is the reason why we should distinguish algebraic theories (Hayek after J.W.N. Watkins) and econometric models. Econometric models are subdued to constraints that they should use available statistics. Most of causal relations are inevitably fictive.

    Lars Syll has a tendency to identify econometric models and algebraic theories. But, by a good selection, we can construct a theory that shows how the total economy works after a causal processes. See our book Shiozawa, Moiroka, and Taniguchi (2019). See my reply in this blog to MIchael Joffe for a short introduction of our works:
    Please note that our theoretical model is half open, because we have assumed that the final demand for each firm’s products are random variables the time average of which moves slowly.

  2. deshoebox
    July 24, 2022 at 1:33 am

    For a different approach, let’s assume the value of getting precise and rigorous conclusions about causality is zero. This may not accord exactly with the ‘real world’ but it’s certainly close. Then let’s assume that people living in the ‘real world’ need to grow food, take care of their children, be protected from the elements, educate young people appropriately…no, wait! We don’t have to assume those things. They are all true and they are all affected by the economy, how it works, what activities are encouraged or prohibited, how goods and other benefits are distributed. Gosh, that makes it sound as if maybe there is causality in the ‘real world’ and that it is somehow connected to economics, doesn’t it? If that is true – and it certainly seems to be – then maybe a small number of us should be thinking about how an economy can best meet the actual needs of actual people living in the actual world. Let’s assume for the sake of this analysis that extreme inequality in the distribution of income and wealth is a bad thing….no, wait! We don’t have to assume that at all. We know it’s true! [To be continued…]

  3. July 25, 2022 at 10:42 pm

    I mostly agree with what each of you has said (this is very unusual!). Especially where Lars says that a model’s view of causal relations is only valuable *if* it corresponds with something in the real world; deshoebox emphasizes that [we] “should be thinking about how an economy can best meet the actual needs of actual people living in the actual world”; and Yoshinori’s modeling is careful to approximate *real forces in the economy* as closely as possible, which is rare in economics and extremely valuable.

    But my problem with this discussion is that it seems to accept that constructing models based on assumptions is *the* way science operates. First, modeling in science is mostly based on observations, not assumptions. But in most sciences – physics is perhaps an exception, and is often the only science that economists talk about – modeling is not a major activity. Take the germ theory of disease. It’s based on observations, along with clever theorizing that stays close to the evidence. The important and fundamental contribution that this theory made, around 150 years ago, was demonstration of the causal forces that were actually operating in some disease processes, and debunking the existing theories (“miasma”, principally) that had hitherto been dominant. Modeling was not involved at all in the work of Pasteur, Koch, etc – and although Ross did create a mathematical model that quantified the risk of malaria, this was only *after* he had demonstrated (causally) that transmission was via mosquitoes. Modeling of the spread of infection is now well developed, but *it would be impossible without the prior evidence-based work on the causal processes*. This is typical of most scientific work. A similar story can be told about the development of tectonic plate theory.

    That paragraph was a bit long, sorry. But the take-home message is: if you equate the scientific method with modeling, you will not understand how science works.

    One other thing: it is also a mistake to equate “evidence” with statistical analysis. Statistical evidence is obviously very useful and valuable, if done well and on reliable data – but it is only one type of evidence. There are many other sources of good evidence. Also, the evidence needs to withstand replication – for example, epidemiologists only accept a statistical finding if essentially the same result is found in more than one study, preferably many different ones with different datasets and different methods.

  4. July 26, 2022 at 2:00 am

    Well said ev, but I would still add an important point to this discussion.

    A ‘model’ does not have to be mathematical. If you don’t like the terminology of a non-mathematical model, then call it a hypothesis. I prefer that term anyway.

    I can hypothesise that the sun rises in the east about every 24 hours. I can write an equation that expresses that mathematically, but the hypothesis already exists. The equation may let us explore the implications in more detail, but the idea (the hypothesis) comes before the equation.

    Plate tectonics (my subject) was proposed by Tuzo Wilson in 1965 and illustrated with a sketch map of the world. A sketch map – qualitative and simplified. Soon others showed how plate motions can be described and quantified as rotations on a sphere, and others again mapped the plate boundaries with greater accuracy, but the unifying power of Wilson’s hypothesis was already evident in his qualitative account.

    Some of us then spent the next thirty years arguing about what underlying forces drive the plates (i.e. what *causes* the plate motions). The best answer is the weight of the plates themselves – the colder, denser subducting plate sinks under its own weight, pulling the surface plate behind it and driving a mode of convection in the interior of the mantle.

    This ’cause’ of plate motions is not something anyone can see. Seismologists have imaged the sinking plates, and we know they are colder than the surrounding mantle they are sinking into, so the ’cause’ is a pretty good inference – but only an inference.

    Regarding Lars’ concern about where ’causes’ reside, I can presume the cause of moving plates is their cold, sinking portions. I can presume that the idea in my head has a corresponding phenomenon out there in the world. The notion that ’causes’ exist in the world and we can discover them is basic to science, and to making sense of daily life.

    But I cannot *prove* in any logical sense that the cause actually exists out there in the world. For one thing, someone might come up with a better interpretation, as Einstein came up with a better (i.e. more general and more accurate) interpretation of gravity than Newton’s.

    One can avoid bogging down in a debate about what really ‘exists’ by dealing only with *observations* and *hypotheses* (as distinct from ‘facts’, ‘reality’, ‘truth’, ‘proof’ etc.). We can observe the apparent motions of sun, moon and planets in the sky and we can hypothesise Copernicus’ arrangement of them and we can hypothesise Newton’s ’cause’ of the motions, and that allows a great deal of progress in figuring out how the solar system works and how to send a space craft to Neptune. Does Newton’s strange idea of action at a distance really exist? We can leave that debate to metaphysicists and theologians.

    I wish more economists would get on with creating a better science of economics instead of agonising about what such a science might look like, or which philosophical label might apply to it.

    You do have to move beyond the neoclassical perversion that just doing mathematical deductions based on the standard set of absurd assumptions (hypotheses) is science. Science involves the iterative back-and-forth between observations and hypotheses, looking for more *useful* hypotheses.

    My green book ‘Economy, Society, Nature’, in the right-hand column here, has examples. For example, a simple model of a financial boom and crash, something that does not exist in the abstract neoclassical world. It’s because there is too much debt, and neither does debt exist in the neoclassical world.

    I think a basic problem is that people trained in the neoclassical tradition simply do not recognise actual science even when it is put in front of them. I don’t mean just that they do not deign to acknowledge something different, I mean they simply do not perceive the different approach to be the science they are looking for. Nor do I mean that observation to be a put-down, rather I hope to alert a few more people to expanding their thinking.

    • rsm
      July 27, 2022 at 1:32 am

      Does Wegener’s vilification indicate the extent to which geological science too is really about geologists themselves, not rocks?

      • July 30, 2022 at 1:24 am

        That would be a huge overstatement. Scientists are people too, and there are plenty of big egos and people too attached to one idea. But science has a better bullshit filter than other human activities, and over time the bullshit tends to get filtered out. There is a great deal about the Earth that is well understood, even though there are still big controversies.

        In Wegener’s case it took decades before new kinds of evidence was brought to bear, and then accepted ideas began to shift.

        On the other hand neoclassical economics is attached to one idea that is easily shown to be inconsistent with observations of actual economies – both in the assumptions going into it and the conclusions coming out of it. So all that is left is big egos and attachment to an idea.

      • yoshinorishiozawa
        July 30, 2022 at 3:33 am

        In addition to the conservative nature of people (economists or natural scientists) who want to keep familiar ideas as long as possible, as Geoff Davies pointed out in the above post on July 30, 2022 at 1:24, I would like to add another logical structure that may be named “theoretical necessity“. I use these terms with pejorative sense, because this is the structure that forces people to attach to the actual framework even though they know some anomalies concerning their system of theories. Lars Syll may have called it “theory-induced blindness” in his post on April, 2022. My comment on this post is here.

        A typical example is the demand and supply framework, which is taught at every high schools it the economics is taught there. If you admit that demand and supply functions exist (or can be defined), you are induced to admit that (1) people can maximize their utility given the budget constraint, and (2) firms produces at a decreasing returns. Both of these two “assumptions” are obviously false. (1) People’s capability to gather information and process it is limited (myopic and boundedly-rational capability of human beings). As for the decreasing returns, It is easy to find many counter examples. But, these properties are admitted because these are necessary assumptions to define respectively demand and supply functions.

        Attacking the absurdity of assuming these assumptions is almost ineffective, because neoclassical economists already know that thy are thinking in an unrealistic theoretical system. To counter this kind of theoretical necessities, it is necessary and effective to present a new theory that does not assumes demand and supply equilibrium.

        Our recent contribution (Microfoundnations of Evolutionary Economics, 2019) made it clear that, in the modern industrial economy, the demand and the supply are kept almost equal by producer firms’ quantity adjustment operations and prices play little role in making demand and supply equal. Prices serve more important function than bringing demand and supply equal. In other words, prices work as the guide to choose a better production techniques than the current ones. A short but exact introduction to the above result is given in pages 190-191 in the Second Edition 2022 of his Post-Keynesian Economics. See also my paper A new framework for analyzing technological change (2020) in the Journal of Evolutionary Economics.

      • July 31, 2022 at 4:04 am

        You might be interested to read about Wegener and others in my recent book Stories from the Deep Earth https://link.springer.com/book/10.1007/978-3-030-91359-5

      • Meta Capitalism
        July 31, 2022 at 3:28 pm

        Geoff I am looking forward to reading your new book Stories from the Deep Earth. I love studying the history of continental drift / plate tectonics.

      • rsm
        July 31, 2022 at 8:02 pm

        Geoff Davies said “science has a better bullshit filter than other human activities, and over time the bullshit tends to get filtered out.” But how did that help Wegener, or Aristarchus, or Turing (the medical science was sure homosexuality was a disease), etc.?

      • August 1, 2022 at 1:49 am

        rsm, I said ‘tends to’, not ‘always’. If you’re determined to focus on the negative then I won’t waste my time further. It’s not perfect, but a great deal has been figured out – even the examples you cite, eventually.

        Some day we may even get rid of the neoclassical nonsense from our public policy making.

      • Meta Capitalism
        August 1, 2022 at 5:56 am

        There are important lessons to be learned in how Alfred Wegener’s theory of continental drift was received in different parts of the world. These histories tell us not only about how science is done by “Scientists [as] people” but as an evolving intersubjective social institution (if that is the right word). Wegener was indeed vilified by some; the overstatement is in dismissing this historical fact as Geoff seems to be doing (I could be wrong on this). Geoff is right, though, that the evolving intersubjective social norms of science, over time, sometimes decades as was the case with Wegener, self-correct as old scientists whose philosophical presuppositions blind them to the evidence in front of them die and a younger fresher pair of eyes replace them. New technologies are developed, new lines of evidence become available, and eventually the facts overwhelm the naysayers that are left.

        We don’t want to create whiggish histories if we really want to benefit from reflecting on the history of continental drift and how Wegener was treated. That is a mistake.

      • August 2, 2022 at 1:52 am

        I certainly agree that Wegener was vilified. (My initial response was to rsm’s suggestion that geology is more about geologists than about the Earth: that is a gross overstatement.)

        Sir Harold Jeffreys rudely dismissed continental drift on the basis that Wegener’s tentative ideas of a mechanism were quite unrealistic, physically. But it is clear from Jeffreys’ language he found the whole idea ridiculous. But Jeffreys was a mathematician who had no appreciation of the geological evidence Wegener was citing, nor of the potential for rocks to deform (even to the end of his long life). His British contemporary Arthur Holmes was a far better scientist who established radiometric dating and proposed mantle convection as a mechanism for continental drift, among many other things.

        Many geologists were not well disposed to physicists opining on their subject (Wegener was a meteorologist) since Lord Kelvin insisted they were wrong about the great age of the Earth. Kelvin did not know about radioactive heating, nor dating. Nor do most geologists know much about fluid flow.

        As well, Wegener was German at a time of great hostilities, so his ideas were not welcomed in the anglophone world. Continental drift was highly disreputable in North America for decades, to the point that advocating it would impede or end your career. Reginald Daly at Harvard was a lonely exception.

        So the intemperate resistance to Wegener’s idea was indeed a sorry and long episode. On the other hand it was a radical idea.

      • Meta Capitalism
        August 2, 2022 at 3:18 am

        Geoff, I am filled with anticipation to read your history of continental drift/plate tectonics. You have a wonderful way of writing. Thank you for taking the time to share. I enjoy it greatly.

    • July 30, 2022 at 8:40 pm

      I very much agree with Geoff Davies’s statement “The notion that ’causes’ exist in the world and we can discover them is basic to science, and to making sense of daily life”. Heckman seems to think that because it is in the mind, it is *only* in the mind. On the contrary, secure knowledge is gained when the conceptual categories in the mind correspond closely with those that occur in real life. “Science and making sense of daily life” is exactly about trying to get one’s mental categories to align with the real-life ones.

      And I strongly agree with “A ‘model’ does not have to be mathematical”. Some of the most important models in science are non-mathematical. Take the periodic table, which is at the core of chemistry. It’s about classification – which shows that when I have stressed the importance of attaining causal knowledge, I have mistakenly ignored the importance of classification – it can greatly help in attaining causal knowledge but is not necessarily itself causal.

      Another such example, fundamental this time to biology, is Linnaeus’s realization that bats and dolphins are mammals, not respectively birds and fish. There was no addition to causal knowledge here, but Darwin’s notion of the “tree of life”, and his and Wallace’s provision of the mechanism of biological evolution 100 years later, would have been impossible if the earlier – wrong – classification had been maintained. I’m not sure I would call Linnaeus’s classification a “model”, and I agree with Geoff again that we should not get hung up on exactly what a model is. The point is that these scientific contributions are enormously valuable whether or not they are models – economists and economic methodologists are much to fixated on models anyway. A small disagreement here though: Geoff proposes using “hypotheses” in place of models, but I think this is too tentative: Linnaeus’s classification and the periodic table have long been established parts of science.

      Another example of a major breakthrough in science is the recent work on AI to discover the 3-dimensional structure (the folding) of proteins when given their amino acid sequence. This is very recent, and an astonishing advance. It is definitely a model, although not perhaps of the kind that most economists would recognize.

  5. yoshinorishiozawa
    August 1, 2022 at 2:50 pm

    Evidencebas (post on July 25, 2022 at 10:42 pm) raised a basic problem on how the science or sciences operate. For the discussion of this theme, I would like to add three keywords in addition to hypotheses and models. They are (1) stylized facts, (2) connecting principles, and (3) visions. All three terms were “invented” within economics and used in it, because they catch some necessary aspects of economics reflections. I will explain each in tern.

    Stylized facts were used by Nicholas Kaldor (Economics without Equilibrium 1985) and Arthur M. Okun. Kaldor’s small book is the first lecture in memory of Arthur Okun. Here are some excerpts form his book:

    I particularly valued in Okun what I once called the method of proceeding by collecting “stylized facts” and then constructing a hypothesis that fits them. … I called them “stylized facts,” a term used also by Okun, because in the social sciences, unlike the natural sciences, it is impossible to establish facts that are precise and at the same time suggestive and intriguing in their implications, and that admit no exceptions. (Kaldor 1985 pp.8-9)

    Connecting principles are the term presented by Brian J. Loasby to explore a new synthesis between Equilibrium and Evolution (1991) but he emphasizes that this term was invented by Adam Smith. This term appears only once in his Wealth of Nations, but was a main theme of his so-called History of Astronomy that was not published when Smith was alive. The full title is “The principles that lead and direct philosophical enquiries: illustrated by the history of astronomy”. I am against the equilibrium framework in formulating any economic processes but what Loasby presents contains a deep insight and is quite indicative if we understand equilibrium as another name of circular flow (Kreislauf) in classical economics.

    I suggested earlier that it is sensible to treat knowledge as reliable when it is supported by evidence. I can now add that compatibility with other knowledge is another very important criterion of reliability. … Network, or cluster, of knowledge are particularly useful on coping with phenomena which threaten to escape far beyond the bounds of human rationality … which is characteristic of economics.

    People prefer not to have to think, but what they like even less is the feeling that they do not understand, and in such a situation, they are driven to seek an explanation. A satisfactory explanation is one that will somehow associate the disturbing phenomenon with what is already familiar, and thus restore a pattern of coherence. The motivation of science, therefore, according to Smith, is the psychological need to invent a set of connection principles which will make sense of experience, and leave the brain in peace. (Loasby 1991 pp.6-7)

    We live in the economy. We observe everyday so many disparate facts about the economy and everybody possesses enormous sum of knowledge concerning various aspects of the economy. Even if economists give no theses on how economy works, people possess spontaneously some naive “theories” on the economy. The economics is partly supported by this naive set of experience. If no one gives any connecting principles, people would make their own. In this point, economics is a bit different from medicine and geology. People know that the medicine (or medical science), although it is very familiar, requires expertise and specialist knowledge. In the case of economics, people are inclined to believe to possess enough knowledge on the economy, because they are experiencing a life in the economy and have accumulated enormous knowledge within his or her life time, although in fact they are “usually the slaves of some defunct economist” (in the last paragraph of Keynes’s .General Theory 1936).

    The connecting principles may be near and must be related to Kuhn’s paradigm.

    The last term Vision is used by Schumpeter in order to indicate a preanalytic picture which plays an important role in economics arguments, because major differences of economic schools are those of visions.

    Here is an example in which Schumpeter explained his idea in his History of Economic Analysis (1954):

    [S]suppose we did start from scratch, what are the steps we should have to take? Obviously, in order to be able to posit to ourselves any problems at all, we should first have to visualize a distinct set of coherent phenomena as a worth while object of our analytic efforts. In other words, analytic effort is of necessity preceded by a preanalytic cognitive act that supplies the raw material for the analytic effort. In this book, this preanalytic cognitive act will be called Vision. It is interesting to note that vision of this kind not only must precede historically the emergence of analytic effort in any field but also may re-enter the history of every established science each time somebody teaches us to see things in a light of which the source is not to be found in the facts, methods, and results of the preexisting state of the science. (Schumpeter 1954 p.38-39)

    What happens after analytic studies begin may be also important. Schumpeter explains us on this phase as follows:

    Analytic effort starts when we have conceived our vision of the set of phenomena that caught our interest, no matter whether this set lies in virgin soil or in land that had been cultivated before. The first task is to verbalize the vision or to conceptualize it in such a way that its elements take their places, with names attached to them that facilitate recognition and manipulation, in a more or less orderly schema or picture. But in doing so we almost automatically perform two other tasks. On the one hand, we assemble further facts in addition to those perceived already, and learn to distrust others that figured in the original vision; on the other hand, the very work of constructing the schema or picture will add further relations and concepts to, and in general also eliminate others from, the original stock. Factual work and ‘theoretical’ work, in an endless relation of give and take, naturally testing one another and setting new tasks for each other, will eventually produce scientific models, the provisional joint products of their interaction with the surviving elements of the original vision, to which increasingly more rigorous standards of consistency and adequacy will be applied. This is indeed a primitive but not, I think, misleading statement of the process by which we grind out what we call scientific propositions. (ibidem, pp.39-40)

    Although these terms are rare to be employed in philosophy of science (or sciences), they represent some specific aspect of economics and I believe they are useful in the arguments in search of a new alternative economics. Major differences between mainstream and heterodox economics are not on the technical levels but a conflict of visions on how the economy works. The efforts to gave birth evidence-based economics is precious, and I hope they will open new scenes in the development of economics, the connecting principles are necessary forces that bind and move economics.

  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: