This Paul Krugman column helped crystallize the weirdness of the ongoing economists versus epidemiologists spat, perhaps more accurately described as the ‘some economists, especially those with libertarian politics, versus epidemiologists spat.’ Different theories, in turn below the fold.
(1) The theory that economists actually are superior. This is the implicit or explicit argument offered by many of the economists who are criticizing epidemiologists, and who seem to believe that epidemiologists (on average) are far less quantitatively capable than economists (on average), and that therefore their models and arguments should not be taken very seriously. This is obviously the explanation of the dispute that is most flattering to economists. The problem with this explanation, as I see it, is that this estimation of the relative worth of the two disciplines is not shared by statisticians and statistical physicists, who may reasonably be considered to be quantitatively sophisticated too, but who don’t have a dog in this fight. It is likely that my reading is biased, but my best estimation is that the specific animus against epidemiologists is associated with economics in particular, rather than other quantitatively oriented disciplines in general.
Certainly, epidemiologists’ workhorse models have had mixed predictive success – the question is whether other modeling strategies preferred by economists would systematically do better at capturing the dynamics of a highly complex social contagion process, with grossly imperfect data. At best this is an unproven case (economists tend to reduce away these kinds of complexity through a variety of simplifying assumptions, which may possibly be helpful, but also possibly be actively unhelpful), and at worst, again, there’s the contrary argument from Kevin Hassett.
(2) The theory that economists’ superiority is a sociological construct that economists desire to maintain. This could start from a gross simplification of Marion Fourcade et al., boiling down down their argument about how economists “see themselves at or near the top of the disciplinary hierarchy” into a much cruder claim about status maintenance. In a world where economists have been used to policy makers and elite commentators going to them first for quantitative advice about public policy, they might feel disgruntled when policy makers and elite commentators turn to epidemiologists instead. Their resentment could easily be transformed into the desire to reassert the previous status hierarchy, through the systematic disparagement of epidemiology.
This is obviously the explanation that is least flattering to economists. Notably, it could take either a sociological flavor (Bourdieu, Merleau-Ponty and strategies emerging from the interaction of field and shifting self-understanding) or a public choice flavor (deliberate efforts by economists to protect the economic rents accruing to them from an existing set of institutions). However, there is a whole lot that it doesn’t explain – most notably the high degree of variation among economists, not all of whom appear to feel threatened by epidemiologists.
(3) The theory that economists and epidemiologists have different motivations or values. Some of the criticisms I’ve seen expressed boil down to the claim that epidemiologists don’t pay enough attention to the economy. That is, that they advocate solutions that minimize the spread of contagion, without paying attention to the possibly dire economic consequences that these solutions might have. Hence, much of the dispute is actually a dispute between professions over the values that society ought to prefer. This is, plausibly, a real dispute. It seems to me pretty uncontroversial that epidemiologists are more likely to care about public health, and economists about markets and the economy, in both cases through a combination of training and selection effects. However, it doesn’t seem to me to be obvious that economists’ take is innately superior to epidemiologists; rather, both disciplines have likely systematic biases towards one notion of the broader welfare rather than another.
(4) The theory that epidemiology challenges the basic ideological presuppositions of (some) economists. This is what Paul’s column points toward. If you start from a naïve-Hayekian understanding of spontaneous order, or from a naïve-Coaseian account of markets and externalities, you will assume that the best solutions are necessarily the solutions that arise spontaneously from the interactions of individuals, rather than the imposition of order by the state or some other actor. This will lead you to oppose the arguments made by epidemiologists that the coercive power of the state should be deployed e.g. to enforce lockdowns, and perhaps to look for reasons why the people suggesting these kinds of measures are incompetent to make them (of course, there are more sophisticated versions of Coase too, some of them offered by Coase himself, depending on whether the wind was blowing nor-north-west).
The problem with these notions of course is that the course of the pandemic suggests that spontaneous order and private contracting are grossly insufficient to address the problem in the absence of concerted state action (and for that matter, many of the predictions about spontaneous market adaptation – e.g. wage premiums emerging spontaneously for workers who could show that they had survived coronavirus – don’t seem to have much empirical traction). The externalities of people’s actions during coronavirus are extremely high, and the prospect of decentralized solutions for those externalities extremely low.
There may be other possible explanations of this dispute than those I’ve listed, and different assessments of their respective merits, given limited and ambiguous evidence. Personally, I’d emphasize the importance of (3) and (4), with a smidgen of (2). What the third and fourth explanation potentially explain that the second does not is the apparent ideological variation within the economics community. Casual observation would suggest that vehement criticism of epidemiologists is associated with right leaning and libertarian political beliefs. This is plausibly explained if those economists who are most likely to criticize epidemiologists are also those who are most committed to markets and most skeptical of state coercion.
My rank ordering may or may not be right. The broader point however, is that unless (1) is correct – that is, unless economists are indeed greatly intellectually superior to epidemiologists so that the latter need to be put in their place – the spat obscures more than it enlightens. It turns possibly substantive disagreements into a dispute over relative status, which appeals to our primate curiosity for gossip about troop hierarchy and intrigue, rather than practical problem solving. A better discussion of (3) would involve more structured disagreement about the relative importance of public health, the economy, and other highly important desiderata. A better discussion of (4) would focus on the interaction between state action and public norms in affecting social behavior and contagion. In both cases, focusing on the actual disagreements, rather than the relative disciplinary status questions entangled with them, would likely lead to better debate.
{ 64 comments }
Dwight L. Cramer 07.20.20 at 2:33 pm
Respectfully suggest that there are additional perspectives between the economic and public health ones. The first that comes to my mind is a culture-based concern for the impact of shutting down the school system for a year or two. That has dimensions way beyond ‘children as disease vectors’ vs. the ‘economic impact of eliminating childcare for working parents’.
Otherwise, I think you’ve pretty much laid out the battleground. And, I hate to break it to the economists, but the fields that have has been attracting the best and the brightest of the applied mathematicians for the last decade or so are not theirs (or public health) but biostatistics and an amorphous area of interest variously labelled big data/AI/machine learning (which is based more in the private sector than academia).
steven t johnson 07.20.20 at 2:37 pm
“1) The theory that economists actually are superior…Certainly, epidemiologists’ workhorse models have had mixed predictive success…” The predictive power of the economists’ models, or their guidance of government policies that have ended business cycles (for an example) by contrast has had presumably unmixed success? I believe it is misleading to phrase the point this way: The theory is that the laws of economics as expounded in the academy are laws of nature, a description of the anatomy of the final society, the endpoint of human history, if not human evolution. Thus, any unwelcome conclusions from epidemiological models must be rejected.
“(2) The theory that economists’ superiority is a sociological construct that economists desire to maintain.” The word “sociological” may be written here, but the discussion seems to indicate that it actually means, psychological trait. In plain terms, it means the vanity of economists. Insofar as it is a sociological construct, it refers to things like the status ascribed to economists by those who employ them, who fund their institutions, who make sure the right kind of economic education is started in elementary schools and beyond, etc. In this view, critiquing the competence of epidemiologists is merely a tactic in servicing the constitutency that desires a different policy.
The OP has vehemently asserted however such a view of the academy is criminal. And that it is the first priority of all decent people to fight such imitations of thought. The OP is as one with Jason Brennan on this, despite a trifling disagreement over manners or the lack of. Nonetheless I suggest the variation in economists in their distaste for epidemiologists roughly coincides with the variation in what kind of service they’re selling and in the constituency they sell too. A Hanson who sells a radical critique of humanity will tack a different tack than another economist.
“(3) The theory that economists and epidemiologists have different motivations or values… both disciplines have likely systematic biases towards one notion of the broader welfare rather than another.” This says that the economists are driven solely by disinterested commitments to independently chosen values and that particular interests have no role whatsoever. This says it is not even conceivable some aggressive economists are attacking epidemologists’ competence to help sell the pre-determined policy—predetermined by those who will benefit monetarily by it and who have the power to rule not formally guaranteed (and strenuously denied) to make policy independent from mere elections—of reopening the economy no matter what.
“(4) The theory that epidemiology challenges the basic ideological presuppositions of (some) economists…The externalities of people’s actions during coronavirus are extremely high, and the prospect of decentralized solutions for those externalities extremely low.” This is a plea for the golden mean, maybe? The assumption that sophisticated von Hayek should be taken seriously strikes me as doubtful. (And, no, I’m not too sure about Coase either.)
So it seems to me the problem of why some economists and fellow travelers like Brennan are aggressively impugning the competence of epidemiologists is because it helps promote favored policies of inaction. That is, that it’s not much of a problem. The problem is why such transparent nonsense isn’t dismissed for what it is. The OP’s herculean diplomacy seems to me to deny this is a problem at all.
Tom 07.20.20 at 3:25 pm
“perhaps more accurately described as the ‘some economists, especially those with libertarian politics, versus epidemiologists spat.’” Henry, that sentence does a lot of work here. Yes, Kevin Hassett is an economist, and so is John Cochrane. But there are many economists who are not libertarians, or, even if they believe that markets have some values, do not want to be grouped with Tyler Cowen who has not published an article in an economic journal for years.
As to (1), economists are not superior. But, as an economist, if you look at a standard SIR model, you see the R0 is a function of the contact rate and the contact rate depends on individual choices and cannot be taken as a constant. I cannot say (don’t know the field) whether economists are able to model it better than the adjustments that epidemiologists do in practice but economists spend their lives modeling this type of environments.
As to (3), I think that’s most likely correct: economists and epidemiologists have different motivations or values, and that’s true even for non-libertarian economists. However, most economists, as far as I know, have been in favor of lockdown, masks etc..
Interestingly, through the current crisis, I found out that economists and epidemiologists are very similar in their willingness to intersperse their value judgements into their recommendations. One thing is to say:
1) “policy X is better because it does Y, and so you should do it.”
Another is to say:
2) “policy X has been shown to do Y but we have not necessarily measured all the other factors that are of value to citizens and so it is up to politicians, and citizens, to make their decisions.”
Economists, in their public pronouncements, often tend to favor 1) rather than the more plausible 2): I always found this very annoying. It has been surprising to me to find out that public health experts seem to do the same.
bianca steele 07.20.20 at 3:38 pm
I think it’s obvious that a lot of economists think their particular expertise allows them to look at other fields’ research, even their popularization, and redo the work in what they feel is a more “competent†way. This demonstrates a basic misunderstanding of science on their part. A lot of what I’ve seen online from economists is in this vein.
Many economists appear to think the world needs them to fix other people’s work up, and only needs epidemiologists and other specialists to do low-status grunt work.
I’m predisposed to want to trust epidemiologists on public health issues but I admit I’m not sure how their expertise can be channeled directly into public policy – as a lot of journalists appear to believe it can be. I had understood it to be a descriptive science and largely mathematical, and concerned with a single aspect of the more manifold field of “public health.â€
Certain left-leaning people, at least on the Internet, also seem to believe that since epidemiology deals with statistics and “populations,†it’s more community-oriented than economics or more person- or microbe-centered medical approaches, and thus will produce more moral outcomes.
The Boston Review has published some very good articles on these issues.
bianca steele 07.20.20 at 3:48 pm
Sorry, the numbers fell off my previous comment again. (Reminder: “numeral plus .†at the beginning of a line doesn’t work.)
I might also suggest that this is possibly a dispute between economists who’ve been acting as pundits for some time and epidemiologists who are acting as private citizens with a certain amount of relevant expertise to bring to the table. To some extent they are probably at cross-purposes as to what kind of contribution is appropriate.
Peter Dorman 07.20.20 at 4:46 pm
The premise of the OP is odd. I doubt that anyone has done a representative survey of economists on this issue, but I receive regular digests from NBER, various SSRN classifications, etc. and scan a lot of abstracts, and the vast majority of them are supportive in a general way of public health modeling and policies. In particular, several studies have provided evidence for the view that most of the economic damage during the pandemic has resulted from changes in personal behavior un-mandated by government officials rather than official lockdowns. Many use GSP data to support this.
I agree that economists get usurpatory in other contexts — definitely vis-a-vis climate scientists. It’s a real problem. I haven’t seen it on the Covid front though.
Peter Dorman 07.20.20 at 6:15 pm
Oops — make that GPS data in the previous comment. (Maybe the Fingerfehler was the result having worked in the past on international trade policy….)
Aardvark Cheeselog 07.20.20 at 6:16 pm
I don’t get something about the basic thrust of this post, I suppose because of some narrowminded notions about truth.
It seems to me that epidemiologists are, basically without exception, people who work to produce models that are true no matter what anybody’s opinion is. Economists, at least many of them, are not like that. Some might even view themselves as being in the business of making self-fulfilling predictions. In any event they have theories of knowledge that can encompass questions like “The Earth… basically spherical, or shaped like a giant burrito?” without anybody losing their job.
Andres 07.20.20 at 6:17 pm
Two comments. First, I am not sure that Hayek and others who built on his work (e.g., Lachmann, Shackle) would side with the economists on this “dispute”. Hayek would argue that the decentralized market is an efficient information processing machine but that the dissemination of this information takes time, time that may not be available when dealing with the exponential growth mathematics of pandemics. So he and others might argue that the greater inefficiency of government action (on lockdowns etc.) is an acceptable tradeoff given the time lags involved in the decentralized market adjusting to the pandemic. Of course, the real Hayek’s ideology might well have trumped this analysis, but I think that is the logical conclusion of his work on economics.
Second, I think explanation (4) only scratches the surface of the objections that some economists have to the implied conclusions of epidemiologist models. The real problem is that the epidemiological models challenge the implicit methodology not ideology of the neoclassical approach that dominates the profession. The five quick and dirty methodological tenets of mainstream neoclassical economics are:
(a) Epistemological rationalism: deductive analysis precedes empirical observation; the latter is only used to confirm the former, not to initiate it.
(b) Ontology: methodological individualism. i.e. aggregate quantities are always passive sums of individual decisions and do not have any causal effect in themselves.
(c) Optimization: individuals always rely on some optimization algorithm when making decisions.
(d) Equilibrium: reality is much better described by equilibrium than by out-of-equilibrium situations; the latter are relatively rare.
(e) Ergodicity: all outcomes are probabilistic and can be forecasted given a knowledge of the frequency of past events.
The pandemic and the models used to track it take issue with at least the last four assumptions, possibly the first as well. That is, individuals take current and likely future aggregate outcomes into account when making decisions (e.g., that’s why hoarding takes place, and why some individuals flee from hot spots, which might help accelerate the spread of pandemic). They can’t use optimizing algorithms, since they don’t know how many in their immediate vicinity (i.e. neighborhoods or less) are infected nor even how vulnerable they themselves are to the virus as individuals. The pandemic is a situation of transition from one (pre-virus) equilibrium to another (herd immunity, by vaccine or otherwise), not an equilibrium itself. And probabilities can’t be modeled since outcomes depend on (unpredictable?) government actions; at most, one can say what probabilities there are for each of a narrow range of government actions, but one then has to make assumptions about non-government organisms (e.g., how many people will protest against face masks and shutdowns?).
So in effect, I think the main problem that mainstream economists have with the current situation is its imperviousness to economic imperialism. The ghost of Gary Becker would be laughed out of the conference room if he said “I can model COVID19 using atomistic, optimizing individuals interacting to reach equilibrium outcomes”.
oldster 07.20.20 at 6:21 pm
Two points:
1) Krugman points to the rebarbative Rand Paul’s invocation of Hayek. Isn’t it characteristic of the glibertarian/market-fetish outlook that they rely heavily on arguments that work only if all of the ideal contractors have perfect (or at least equal) information, while meanwhile doing their level damnedest to ensure that false beliefs drive out true among the actual population of consumers?
It’s absurd to talk about people making their individual unconstrained rational choices about the risk of disease, when you are also trying to muzzle the people who can best inform them about the risks of disease. Without good data, what sort of choices does anyone have? None.
And yet, that’s the Randites game. In every arena, not just pandemics. They want us all to be individualists about guns, while passing laws that forbid the government from collecting information about gun crimes and deaths. They want us to make up our minds about pollution, while making sure that corporations can bury the secrets about how and how much they pollute.
It’s not a terrible two-step, it’s not a motte and bailey, but it deserves a name of its own: fetishizing the pristine epistemic purity of the decentralized crowd, while fighting to make sure that the common clay are uninformed morons.
2) The phenomenon of over-weaning intellectual arrogance is not restricted to economists — lots of prestige disciplines get high on their own supply.
Physicists make themselves look foolish by pontificating on philosophical questions; philosophers show their ignorance when they attempt to lay down the law for science. Niall Fergusson is busy showing us that historians are not immune from belief in their own polymathy, while Engineering is famous for generating cranks (holocaust deniers, 9/11 truthers, etc.)
Not that I want to defend economists — they are among the worst, no doubt. (And rumor has it that Larry Summers used to think himself personally qualified to pronounce on the intellectual merits of every tenure case in every department at Harvard, which for anyone who understands the structure of expertise is laughably absurd.) But the disease is wide-spread among smart people of all sorts.
Thank god that I’m too smart to fall for it!
William Meyer 07.20.20 at 6:33 pm
“IF ECONOMISTS could manage to get themselves thought of as humble, competent people, on a level with dentists, that would be splendid!†said John Maynard Keynes, a British economist.
Jerry Vinokurov 07.20.20 at 8:52 pm
But these are, in fact, ideological tenets too. Actually, every methodological tenet carries with it an implicit ideological commitment!
CHETAN R MURTHY 07.20.20 at 9:30 pm
Henry, it occurs to one that we’ve been here before! Remember good ol’ Bjorn “but think of the poor!” Lomborg? He also claimed primacy over (in that case) climate scientists, on the grounds that he was an economist. [well, not so much that he was an economist, but that economics was the right way to approach the problem]
So obviously I think he was hilariously [narrator: also, tragically] wrong.
But really, I just thought I should note that this “the experts in field X are fools, lemme correct them” isn’t a new schtick. And it’s always done in the service of Capital, right?
It’s all pretty embarrassing, but not unexpected.
Ebenezer Scrooge 07.20.20 at 9:54 pm
Coming from a physical science background, I am relatively immune to physics envy. The economists amuse me when they confuse math skills with situation insight. Every physical scientist knows a few people with comparatively weak mathematical skills and deep physical insight. And they also know a few people with superb mathematical skills and little physical insight.
Andres 07.20.20 at 10:12 pm
Jerry: It is true that every methodological tenet carries an implicit ideological commitment, but this commitment sometimes leads to unexpected results. As I have stressed before, the five methodological tenets of neoclassical economics can lead to widespread ideological stances depending on the initial starting points or assumptions. If neoclassical economics were to ignore externalities, excludable/rival good problems, unequal initial wealth endowments, imperfect competition, and free rider/prisoner dilemma problems, then you would get free market Chicago economics. When neoclassical economics doesn’t ignore these problems and allows for their important presence in reality, neoclassical economics becomes the discourse/analysis of social democracy. So starting point methodological commitments tend to be more subtle than endpoint ideological positions.
rdb 07.20.20 at 11:37 pm
In a similar vein, Jules’ klimaatblog has a series of posts (currently 40)
Links to the .
John Quiggin 07.20.20 at 11:51 pm
The Australian academic economics profession is small enough (about 500 by my last count) to get a pretty good picture. The biggest intervention was a statement by 122 economists (including me) opposing early lifting of restrictions. The signatories covered a wide span of views from CT to sane versions of Chicago.
https://newsroom.unsw.edu.au/news/business-law/open-letter-122-australian-economists-dont-sacrifice-health-economy
Against that were a handful of academic economists and a slightly larger group of business economists (either propertarian or reflexively contrarian) supporting herd immunity, and another smallish group trying and failing to find the sensible centre. Between them, this group might total 20 or 30.
The largest group, as always, were those who never have anything to so about policy and are mainly concerned with getting publications in “good” US journals.
It’s also important to remember that not all epidemiologists favored suppression. Most started with the view that “flattening the curve” was the best we could do, and a few favored herd immunity.
As regards the view that “economists actually are superior”, I’d say this was a natural experiment in testing for Dunning-Kruger syndrome.
faustusnotes 07.21.20 at 1:11 am
I don’t mean to single anyone out, but Tom’s comment at 3 is a sterling example of why epidemiologists get frustrated by this kind of economists’ intervention in our field. Tom writes
So without actually checking our field, Tom presents a very basic part of our most basic model as if it is a huge revelation to us (the contact rate cannot be taken as constant), and suggests economists would be experts at studying this. He doesn’t consider that epidemiologists already know this, even though it’s obvious we do, since our primary policy recommendation – lockdown – explicitly assumes the contact rate can be changed. He also doesn’t check whether perhaps epidemiologists are experts at studying a key component of our models (we do) and doesn’t present any evidence that economists would know how to do this. Economists study human behavior true but do they study the number of people you contact a day from an epidemiological perspective? I doubt it?
I don’t want to accuse Tom specifically but the arrogance required to look at the most superficial model in another field, conclude that experts in that field have missed a basic thing about it, and suggest your own field would be better placed to study it, is infuriating.
This shade being cast on epidemiologists by economists seems to be strongly based on the idea that economists are better at quantitative stuff than other fields, which is annoying coming from physicists but especially frustrating from economists. Looking on the field from the outside, it seems to be completely full of mathematical mumbo jumbo. By this I don’t mean “you can’t use maths to study human behavior”, I mean the maths in economics is wrong. Economists don’t seem to understand units or dimensional analysis, and their basic equations are hideously unphysical with their weird obsession with raising everything to arbitrary powers. It’s pretty routine to find economics papers with no connection between the model used for physical evaluation and the model used for theoretical exposition. Economists don’t seem to know anything about time-to-event models, and it seems pretty clear to me that economics as a field hasn’t moved beyond OLS regression – when economists study rates it is so common to find them using OLS on logged mortality rates, and they use OLS to study proportions (and measure differences in proportions rather than odds ratios) so often. Indeed, it’s only in economics that you still find the horrible words “Linear Probability Model”. Economists also don’t seem to understand standardization (see Case & Deaton’s unfortunate excursion into “deaths of despair”), and often don’t adjust for age or sex. The main contribution of economists to my field has been the incremental cost effectiveness ratio (ICER), which is mathematical nonsense and statistically intractable but sadly has become the norm in economic assessment of medical interventions.
I think economics needs to clean its own house before it starts casting aspersions on the quantitative skills of others. It cannot be a serious science while it has no mathematical basis to its work, and in particular until the unit issue is cleared up there is just no sense to any of the maths that these critics of epidemiology think makes them so special. There are a lot of problems in epidemiology and public health at the moment, and I think a lot of people working in my field aren’t aware of them or willing to address them, but we certainly don’t need advice on how to improve our methods from a field that hasn’t moved beyond OLS regression and can’t balance the two sides of their fundamental equations.
ph 07.21.20 at 1:28 am
“Why Most Published Research Findings Are False” https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124
Hi Henry, nice to see you back. I didn’t think much of the Krugman article, or the linked pieces which seem too ‘localised’ and Trump specific to have much utility. Indeed, the OP seems much more sound.
The linked Ioannidis article from 2005 identifies some of the problems of gathering and assessing medical data. And the footnotes link to a number of other interesting pieces.
Lee A. Arnold 07.21.20 at 2:17 am
The full premise, never stated, is that the unlucky ones must die for the economic order, which cannot think of another way to temporarily arrange things to save their lives.
While we expect this mental vomit from our politicians, any economist whose “value” is the “economy” is a callous idiot.
John Quiggin 07.21.20 at 5:51 am
I’m not known as a fan of the current state of the economics profession, and certainly not of economic imperialism in the social sciences, but just about every statement made by faustusnotes is wrong or meaningless (at least to me). As an example of wrong, OLS was old hat when I first studied econometrics in the 1970s. For meaningless, “weird obsession with raising everything to arbitrary powers” is hard to beat.
faustusnotes 07.21.20 at 6:19 am
John, I can’t access all the papers at NBER without paying, but This week’s Health Bulletin has two freely accessible papers on COVID-19, one of which uses OLS, and I’m pretty sure that the paper on influenza and oxycontin both use OLS, to guess by the (extremely vague) language in the abstract. As to raising everything to arbitrary powers, the Cobb-Douglas function does this explicitly (and I don’t think the units balance even in the Wikipedia summary!). William Barnett has raised the issue of dimensional analysis in economics, and lots of physicists have complained about it. Gelman made clear that Case and Deaton didn’t properly adjust for age (and after standardization their results disappear); NBER is riddled with OLS analyses of logged mortality rates (or voting proportions); even some studies of the National Firearms Agreement used this retrogade method in the 2010s. I’m guessing you’re not familiar with the ICER but I can guarantee you my criticisms of it are accurate.
There’s a lot of mathematical mumbo jumbo in economics, at least the economics I see published on epidemiological topics. I don’t think it’s unreasonable to point that out in response to people like Tyler Cowen suggesting epidemiologists don’t know how to analyze our own data.
John Quiggin 07.21.20 at 7:00 am
The powers in the Cobb-Douglas function aren’t arbitrary or the product of weird obsessions – they represent the shares of labor and capital in a very simple model. And, whether or not you can make the units balance in the Wikipedia summary, they do. All this leaves me doubtful about your claim that (what you say has become) a standard measure in your own field is wrong.
On OLS, you seem to have shifted from “the field hasn’t gone beyond OLS” to “any use of OLS is wrong”. On that basis, I could condemn any field where papers are “riddled with” partial correlations (closely related to the very simplest case of OLS). There’s nothing wrong with OLS when the standard (restrictive) assumptions are satisfied. There’s a vast range of GLS methods for cases where the error structure is different, as well as limited dependent variable methods for probability models, which have been in regular use for decades. I’m not really up with more modern (that is, post-1990 or so) approaches like Generalized Method of Moments, but it doesn’t seem as if you are either.
As regards the papers you mention, one makes no use of OLS, while the other estimates one key parameter by OLS (quite legitimately AFAICT), but is primarily concerned with other issues.
On Tyler Cowen: pot, meet kettle.
faustusnotes 07.21.20 at 10:35 am
John if the rhs of the equation has labour raised to a power and capital raised to a power, and the lhs is output, then the TFP has to have constantly changing units in order to balance lhs and rhs. This TFP has to have a weird unphysical set of units (output divided by the weird powers of hours worked and currency) to do this, which is particularly hard to square with a basic definition of TFP as output divided by weighted mean of labour and capital (which is itself a weird and unphysical definition). I think this runs into further problems when further down the mathematical exposition section the researcher introduces the firm or family consumption function which further contorts the L and K from the rhs of the Cobb-Douglas function. It’s a mess in the simplest models.
I didn’t say the field hasn’t gone beyond OLS, there was a hyphenated condition – the field still considers it acceptable to use OLS for rates and proportions. Here, for example, is an analysis of domestic violence reporting during COVID-19, that uses OLS to study the case numbers with month-year fixed effects (wtf) and a variable coefficient of population. This is disastrously bad statistics, but it was on the front page of the NBER website today. This should obviously use a Poisson regression with population as an offset (to ensure rates are being modeled). Shocking. Yet this happens all the time with NBER research. I can’t conclude that economists understand how to analyze epidemiological data while methods like this remain mainstream.
I regularly see this kind of statistical and mathematical work reported from economists and I can’t conclude that it is not part of the mainstream. As far as I can tell, economics as a field thinks this stuff is okay. Do you agree that the front page of the NBER is a good representative of the field?
Tom 07.21.20 at 10:45 am
@Faustusnote, thanks for your answer. I am aware that the contact rate can be modeled as time-variant. The point is whether economists have anything to contribute to this modeling. For a recent paper that I like, you can check out “Socioeconomic Network Heterogeneity and Pandemic Policy Response” by Akbarpour et al. It uses network theory to study how different social distancing policies affect contacts, health outcomes and employment. If you have time and willingness (and I understand if not!), since you are an epidemiologist, I would appreciate your opinion on what Paul Romer has been advocating (see many recent posts on his website https://paulromer.net/ or maybe this tw thread: https://twitter.com/paulmromer/status/1279067975522418688). Basically testing everybody often but with pooled testing to keep the costs low. He claims this is the approach taken by the Stanford’s medical center and what Cornell plans to do in the Fall.
As to your point, “the arrogance required to look at the most superficial model in another field, conclude that experts in that field have missed a basic thing about it, and suggest your own field would be better placed to study it, is infuriating”, I can relate because I feel the same about my field when I hear complaints about my field by people who clearly are barely familiar with it. This is especially annoying because I believe that my field has several limitations but that often the attacks to it are misguided (on the other hand, I have read some of the best attacks to my field here on CT and that is why I keep coming back to it).
As to your other points, they are interesting. I agree about Case & Deaton and Gelman’s critique. As to the aversion about LP models, this is definitely a division across fields: in my experience, many in the medical sciences prefer odds ratios while in economics Angrist & Pischke recently pushed that, in the end, OLS is a more transparent method than assuming a certain distributional form for the error terms as logit and probit models do. Some economists disagree. I usually use LP models as I find easier to interpret the magnitudes, especially when there are interaction terms: but I run logit and probit as robustness checks and almost always I get the same results. I do not know about ICER. The point about OLS is not correct though: if anything, as a field, it is not that we are still stuck with OLS but rather we have moved back to it (instrumental variables more than OLS frankly) because many felt that more complicated structural models would obscure identification of the relevant parameters. The debate about this issue was, and remains, quite intense as many feel this is a step back and would prefer the more complicated models.
oldster 07.21.20 at 12:47 pm
I think you’re all missing the fundamental problem with economists’ claim to be scientists:
they routinely violate a well-known law of nature, that the independent variable always goes on the horizontal axis.
I mean, if you’re wrong on the basics like that, how can any of the rest of it have any validity?
Jerry Vinokurov 07.21.20 at 2:11 pm
Now this is a crime!
roger gathmann 07.21.20 at 3:09 pm
I’m thinking that there is one other major point – guilt.
Economists in the U.S. and other neolib countries have long advocated disinvesting in public goods, leaving them to the market. That disinvestment, in healthcare, has led to quicktime HMO oriented medical policies long on shaving off costs in order to increase profits for healthcare corporations.
You can look at the U.S. – or you can look at France, which is supposedly different – and you see the catastrophic results of urging public disinvestment. It is like a pre-play for climate change. Economists, in my cynical view, don’t drive economic policy as much as operate as theologians for it – although I suppose you could argue that the Larry Summers type does have strong governance positions on his C.V. However, the most known names in economics, with few exceptions – the ones who get referenced as experts in the thumbsucking articles – have been down with the notion that most of what goverment did in the high tide of 20th century liberalism – from operating power plants to public housing to public clinics – should be done (more efficiently) by the market. The market, of course, names a blur, or an emptiness – I’m rather with James Galbraith when he writers that markets are various, they don’t have uniform laws, and the “market” in the economics sense is vacuous – but if we substitute various healthcare corporations and businesses for market, we are talking more concretely about what happened. The epidemic has simply speeded up the abysmal crackup of what used to be public healthcare.
The economist mainstream, or at least the experts, from Summers to George Mason U. peeps, fear disasters like Covid-19 might ruin all of that work – and worse, that people might actually ask who killed public healthcare.
Just my two cents.
Andres 07.21.20 at 4:42 pm
Oldster, speaking for the other economists here I will say nice trolling, but no cigar.
not feeling very nymous 07.21.20 at 6:37 pm
I’m second to none in my contempt for the mathematical and statistical level of economists, but I find the dimensional-analysis complaint Faustusnotes is making against Cobb-Douglas sort of bizarre. It’s true that writing Y=A L^t K^(1-t) looks dimensionally weird, because the units are [money] on the LHS, and we usually measure L in [labor time] and K in (again) [money], so it looks like total factor productivity A has to have units of [money/time]^(t), i.e., units which change with the estimated parameter t. But to make this all go away all we have to do is introduce scale factors, say L0 and K0, and write the equation as Y=A (L/L0)^t (K/K0)^(1-t). Now we’re raising dimensionless ratios to weird powers, and that’s fine. Since neither L0 nor K0 would be identifiable on their own, we can even pick their values by convention to be 1. This really isn’t a problem.
What’s rather more serious is that the appearance that Cobb-Douglas fits the data is entirely due to the near-constancy of the labor and capital shares in national income [http://www.jstor.org/stable/4032642], as people have been pointing out for decades [http://doi.library.cmu.edu/10.1184/pmc/simon/box00064/fld04934/bdl0001/doc0001]. This isn’t even something economists hold on to because of its ideological valence, it’s just pure dysfunction as a scientific community. (And, no, it’s not a known-to-be-wrong idea which is kept around for pedagogical purposes or as a merely-theoretical example, see first link.)
Joseph 07.21.20 at 6:45 pm
Dear God! Are their actually non-economists out there who believe that economist belong higher in the intellectual & rigor hierarchies than physicists, mathematicians, and any of the hard sciences? You must be kidding!
bianca steele 07.21.20 at 8:00 pm
“Economists, in my cynical view, don’t drive economic policy as much as operate as theologians for it ”
I thought this was interesting but doesn’t it depend on what “theologians” means? If they’re the kind of theologians that operate in the modern secular university, then they’re kind of like anthropologists, describing what ordinary people mean.
If they’re the kind of theologians that operated in the medieval university, they’re more like legislators and legal scholars, and then they do drive economic policy. I think it’s obvious this is the case. One point of evidence in favor is that it’s impossible to imagine anyone being hired to do economic policy who hadn’t learned what the economists already think policy experts are and should be doing. There’s no naive “practice of economic policy” the way there might be a way for uneducated people to practice their religion.
steven t johnson 07.21.20 at 8:35 pm
oldster@26 “they routinely violate a well-known law of nature, that the independent variable always goes on the horizontal axis…”
This is gorgeous. As nuts as it is, there’s just enough truth to make this a belly laugh.
And a sharp irony cuts both ways, which is double the fun.
marcel proust 07.21.20 at 8:55 pm
@Joseph wrote:
Dear God! Are their actually non-economists out there who believe that economist belong higher in the intellectual & rigor hierarchies than physicists, mathematicians, and any of the hard sciences? You must be kidding!
Not sure how to read this. Should I infer that physics is not one of the hard sciences? Certainly, mathematics at least since Bourbaki (if not Gödel) is closer to the humanities than to the hard sciences, even if it is uncannily useful for them.
Kiwanda 07.21.20 at 10:27 pm
I have the impression that both economics and epidemiology work with social networks, agent-based models or other simulations, continuous approximations of discrete systems, intervention assessment, and I’m sure more. Also, apparently the analogy betweenfinancial and biological contagion has lead to some cross-fertilization. Since their tools and settings have some overlap, and since the pandemic and economy interact, it seems like economists and epidemiologists would both try to have something to say there.
I don’t have any first-hand experience with economists and their arrogance, but the prize for the widest gap between how smart they think are, and how smart they actually are, must surely go to physicists. They also have the remarkable ability to go into another field and ignore all prior work there, preferring instead the profound interest of whatever they happen to pull out of their asses.
Richard melvin 07.22.20 at 12:42 am
I think the claim is not that they are ‘better’, more accurate, deterministic or mature, but that they are ‘downstream’ of other sciences. So they can take, say the prediction of an astronomer than a planet-busting asteroid will impact in 20 years, and ask ‘how much money should we spend on deflecting it?’. With the expectation that they will be able to come to a better and more trustworthy conclusion than a non-economist would.
Of course, all-too commonly what they actually do is add an additional source of noise and uncertainty. Which leads to the well-known situation where, for any wrong cause of action. the least-obviously-wrong argument for it will be made by an economist.
faustusnotes 07.22.20 at 1:16 am
Not very nymous, maybe I’m misunderstanding you but I don’t think your approach is correct. First of all I’m not convinced the lhs of this equation is always money – it can be units of output in industry-specific models (though it’s hard to say because typically the researcher never specifies the units). Secondly you can’t just divide the rhs of the equation by two scale factors – you need to divide the lhs as well. Otherwise the equation becomes a different equation describing a different thing, in fact it becomes a relationship between two forms of total factor productivity. But if you do divide the LHS as well (to preserve dimensions, which is important!) you now are no longer talking about output, but some borked measure of production efficiency (which should be obvious since you now have an equation relating two different forms of total factor productivity). Thirdly, you have specified a special case where the elasticities sum to one, but in general you will be dividing by arbitrary powers of these two scale factors so then the units of the lhs will vary according to the problem. This solution also doesn’t change the problem with the definition of TFP being weird and unphysical in any case. And what exactly do these scale factors represent? How does the interpretation of the equation change once it is in terms of ratios? I don’t think you’ve fixed anything here.
Barnett discusses this problem in the Austrian Economics Review with examples, and also with some shocking examples of peer review comments on the issue from major journals, which show a real lack of understanding of basic physics. Texocotitla describes the problem in a preprint here. But the problem doesn’t just stop with the production function. There are many situations where the units of terms in the production function cannot be consistent with the use of the same terms elsewhere in the work. For example, Duarte in the Quarterly Journal of Economics, 2010 (which I am familiar with because I had to read it for a work project) has terms that simply cannot be dimensionally consistent across the production, consumption and utility functions. The equations in this paper are, essentially, made-up nonsense. And this seems to happen a lot, although it can be very difficult to figure out because I have noticed economists are very bad at actually stating the units of the terms they’re using, another sure sign that they don’t really know what their equations do or represent, or are using them more as a kind of symbolic logic than as a form of inquiry.
This use of equations as symbolic logic also explains the weird lack of connection between the mathematical exposition that is usually found at the beginning of a paper, and the OLS model that follows it in the data analysis section. Typically for example a production function with weird powers is introduced and partial derivatives or some similar analysis performed in the mathematical exposition; then in the data analysis a simple linear OLS is performed and the general direction and significance of terms used to confirm the theory in the exposition. There is no direct connection between these chapters at all, in many cases.
Tom, I can’t make a lot of judgments about Romer’s approach based on your links, but in the twitter thread he appears to be advocating the porn industry approach to infectious disease (extremely regular testing of high risk people). This is why Trump hasn’t got it while Lance Bottoms (Atlanta mayor) has despite wearing a mask. It might work! But his comments on contact tracing seem poor.
Regarding LPMs, I saw that discussion about whether to use LPMs that you mention (by Angrist and Pischke). It seems very shallow to me. The focus on non-linearity (not the only issue!) and the obsession with marginal effects, etc. It’s a bad sign for your field that this stuff is under discussion. Really, I don’t know what else to say to people who think that estimating negative probabilities with inconsistent estimators based on completely the wrong distributional assumptions is okay. It’s not.
I think roger gathman gets at the real point here, which is that this pandemic (like global warming) has destroyed almost all the assumptions underlying the classical models of economics and capitalism, and it sends the high priests of that theory into connuptions. Public health (through the WHO) has been pushing for universal health coverage for years and that goes against all of American economics’ leading lights. If they admit that we’re right about this their entire system falls apart, which is also why economists have led the charge to undermine global warming science.
John Quiggin 07.22.20 at 2:57 am
“Public health (through the WHO) has been pushing for universal health coverage for years and that goes against all of American economics’ leading lights. ”
There is actually empirical evidence on this. A survey of health economists shows overwhelming support for Obamacare and opposition to various rightwing proposals for cuts. They didn’t report asking about single-payer or Medicare for all, but there are plenty of “leading lights” who support one or both.
https://www.nytimes.com/2020/02/17/upshot/health-system-economists-survey.html
“which is also why economists have led the charge to undermine global warming science.”
Nonsense – name some names. Even rightwing economists like Mankiw support carbon pricing. Given the size of the profession, there must be some loonies, but I don’t know of one with any professional credibility.
Again, there’s plenty to criticise about economics without making stuff up.
John Quiggin 07.22.20 at 3:03 am
@Faustusnotes It’s also strange that you are quoting a member the lunatic Austrian fringe Mises Institute as proof that the production function is incoherent – that’s the first place I’d look for climate science deniers and lukewarmers.
And, bingo https://mises.org/wire/what-universities-wont-teach-college-students-about-economics-climate-change
faustusnotes 07.22.20 at 3:58 am
John, I’m not surprised that you jumped on the journal name to criticize that point. In fact the article I’m referencing was rejected from two major journals, and Barnett cites the peer reviews he received, which show that the peer reviewers didn’t understand basic physics. Barnett isn’t alone in this criticism – lots of physicists have been complaining about this over years, which is why I cited another paper on the same topic.
It’s a bit strange that you would complain about my misrepresentation of economists’ position on climate change, then immediately following cite economists … opposing climate change. I hardly need to point to the long list of economic opposition to the theory: there is one, initials RT, who I can’t name here for fear he’ll appear and derail everything. There are the Worstalls with their luke-warmer do-nothingism, and then there are the various free market solutionists who obstinately refuse to understand the issue, and then there are the rogues gallery at the Heritage Foundation, AEI etc who vehemently oppose it. Organizations like IEA, GWPF etc are stacked with economists. I think everyone here can remember the response of the economics profession to the Stern Review, where they spent months quibbling over discount rates, or the damage models of that time which basically dismissed basic results we now all understand to be true (and Nordhaus is lauded for them!) It is economists who have led the charge against mitigation by claiming that we will be richer in the future. It is economists who have taken tobacco company money on this issue.
We were all there in the 1990s and 2000s, and we can all remember which institutes and which disciplines were speaking out most strongly to obfuscate, delay and defuse. Here, for example is The Economist talking up the pause in 2014, when everyone else already understood that global warming was going to be catastrophic. We all know that the “pause” was an obfuscatory tactic, promoted by all the big deniers. Yes, the discipline as a whole has finally come to grips with global warming, but only after it’s too late.
Fake Dave 07.22.20 at 9:32 am
Part of me has always suspected that the “everything is really economics” mentality is just people overcompensating for the deeply square major they picked. Like the epidemiologists were out smoking reefer and casual sexing while the elite econ kids were picking the right bowtie for going straight from Young Republicans to the Associated Students meeting. Now those long-haired bleeding heart liberal goofballs are acting like experts on public policy when they haven’t even read Smith, much less Ricardo and Hayek. Everyone knows you’re supposed to spend your formative years specializing in something so narrow, dry, and tedious that everyone knows you must be “serious.” Only then do you have that “right” to cash in.
If you owe your own position mostly to department politics and savvy self-promotion, you obviously don’t want someone guileless and apolitical yet actually competent around to make you look bad. Econ’s dominance over other sciences has an obvious political explanation, but it’s embarrassing to admit that the sciences follow intellectual fads and reward people who “game” the system. So we get to live with the pretence that a few hundred mostly rich old white guys from a handful of elite economics departments actually represents the cutting edge of social theory.
These guys might all belong to their little tribes of Austrians and Keyensians and so on, but it’s amazing how quick they are to close ranks and defend each others’ rights to wade into other fields and tell the natives how to run things. Paul Krugman does this all the time with public health, climate, and (especially) electoral politics and it’s done more to sully his own reputation for competence and consistency than to genuinely uplift those fields. The confused hash he made of the single payer debate alone should have been enough to provoke “stay in your lane” responses (and did from some of the actual health policy people), but he almost always gets a pass from the liberal press because he’s a high status academic and, in fact, one of the intrinsically respectable “very serious people” he always complains about.
The tendency of mainstream econ (and perhaps certain CT posters) to give the consistent benefit of the doubt to silly and self-serving neo-classical idiots is arguably another example of closing ranks. The other social sciences have spent decades grappling with the toxic legacy of Behaviorism, Rational Choice theory and other “economic” approaches. Our fields have actually moved on since the debates of the 1950s, but the Chicago zealots and the “wonk” set still insist we’re the ones who need to learn from them. It’s exactly the sort of rank ignorance common to people who take their elite educations for granted and it’s pretty embarrassing to watch, to be honest.
Tom 07.22.20 at 6:38 pm
@Faustusnotes, thanks for your answer. I think the discussion is going in too many ways, and I guess my first comment is partially responsible for this. Let’s see.
1) I clicked on your website and read some of your Covid-19 posts and really enjoyed them, especially the one about the comparison with the HIV. I reiterate that I do not think economists are better than epidemiologists, or physicists, sociologists, or whomever: I find these comparisons extremely annoying. Covid-19 is affecting the economy, clearly, and so economists have been trying to get up to speed with SIR models etc. There may be, hopefully, potential cross-pollination across fields, especially in evaluating the effect of different economic policies on health and economic outcomes. Here is a better link to Romer’s proposal in case you are interested:
https://roadmap.paulromer.net/paulromer-roadmap-report.pdf
2) Regarding Y = F(K,L). F is a function: there is nothing in calculus that says that K, L and Y should have the same units. You need an oven, some labor effort and some ingredients (which would be another argument in the function) and you get a pizza. We usually do not observe the entire production possibilities of a firm, let alone of an economy, and so we use a functional form to help us represent the relationship between the production factors and the output. I have never been a fan of the aggregate production function and so if somebody comes up with a way of doing macro without it, I am all for it. But economists have found this tool useful for certain purposes: in many cases people outside the field see the choice we make and raise criticisms about them. That’s perfectly fine but these criticisms may miss why certain tools were developed in the first place. If one is in favor of, say, Krugman’s criticisms of austerity and believe that aggregate demand was very low in 2008, well, he based his reasoning on common macro tools. One can’t have both: jettison the field but keep the policy recommendations that s/he likes. In general, economics is much more flexible than the policy recommendations of free market economists would let you think. It is a fact that the ideas of the latter have carried a lot of weight in policy recently. But this need not be the case. If you throw away the entire field, you may be able to replace it with something better; or you can end up listening to cranks.
Regarding LPM, I agree to disagree. Regarding your criticism of the domestic violence paper, again, I see your point but I am not sure you see why they used that model and which advantages it has that Poisson does not have. If I have time, I will respond to your specific criticism of the paper more extensively. That will be my last one about defending economic methodology because I do not want to derail the thread further on this. But thanks for engaging and hopefully there will be another occasion.
bianca steele 07.22.20 at 8:11 pm
Riffing off @41, especially the last paragraph,
I think it wasn’t totally surprising that things like rational choice theory and freakonomics might seem like promising ways to start social science and parts of philosophy from scratch on ideologically correct grounds, given certain assumptions about capitalism being an ideological society of a certain kind. If you’re over in the humanities looking for a new approach to everything, maybe it makes sense to assume the emerging creativity is going to be found among the guys with both pocket protectors and country club memberships. It’s also possible to imagine a liberal back in the 60s thinking public choice and so on could go more than one way. In retrospect it should be obvious why that didn’t happen.
Ronan(rf) 07.22.20 at 10:10 pm
The overwhelming majority of economists, including GMU and libertarian types, supported the lockdown, so it’s unlikely ideological. Epidemiologists have been quite tribal protecting their turf, so it seems to just be standard arguing across disciplines.
J-D 07.22.20 at 10:49 pm
Fake Dave
I do not think that word means what you think it means.
faustusnotes 07.23.20 at 4:00 am
Tom, the Paul Romer roadmap looks interesting, but I should say that it seems like a classic example of an economist jumping into a field to state the obvious. All he’s doing there is spending a lot of verbiage repeating what the WHO recommended in January, and what sensible Asian countries did from the start.
Regarding the paper I offered as an example of bad use of OLS, it appears you seem to think it’s not? For starters, let’s consider this: it is likely that the coefficient of population in that paper will be very small. There will be other coefficients that will be negative. This means that in the OLS framework it is perfectly possible to have a predicted estimate of number of domestic violence cases in a particular combination of low population, high-elderly districts (perhaps in the red zone) that is negative. A model that predicts – or can potentially predict – negative cases, or that has 95% CI (or any other plausible CI) that includes negative cases, is obviously silly on its face. This is without considering the additional problems in its fundamental conception (it treats changes in cases as having the same meaning no matter the scale they occur at, and it doesn’t properly model population rates, and has the wrong probability distribution) and the problems with analyzing model diagnostics (they’re all wrong because the probability distribution is mis-specified). What does this model offer to make up for all of that?
Ronan(rf), do you think maybe that when famous economists publish posts demanding to know epidemiologists GRE scores, it might not be the epidemiologists who are being tribal?
faustusnotes 07.23.20 at 4:08 am
John, that survey of economists you say shows they overwhelmingly support universal health coverage is actually quite damning. First of all, it’s about Obamacare, not UHC. Secondly, it’s of health economists, not economists. Thirdly, 11% of them supported repeal of Obamacare. In contrast, all UN member states supported the UN Declaration committing to achieve UHC by 2030. So these 200 health economists are well behind the global consensus on UHC. And that’s the health economists! Where do the rest of the economics profession line up?
The survey also has some rather disturbing findings about how many health economists support higher costs for pre-existing conditions (even genetic defects), making people with unhealthy habits pay more, or eliminating the individual mandate in the ACA. And these are the health economists!
Fake Dave 07.23.20 at 5:12 am
@45
OK so maybe Krugman has always been something of a blowhard, but at least he’s not yet another goddamn robot from the Chicago factory. When I was younger, during the bad old Bush years and then the recession, he seemed to be the rare Ivy League economist who actually respected the lives and perspectives of working people. Years of anti-populist “Bernie Bro” abuse and his cowardly, purely political, opposition to medicare-for-all has disabused me of that notion. Kill your heros, etc. Now I just assume that the every “prominent” econ pundit to the right of Robert Reich (hell, maybe him too) is a brainwashed idiot or some kind of sociopath and I’m much less prone to disappointment.
reason 07.23.20 at 6:29 am
I like Andres comment @5. Lots of ideas that were vaguely in the back of my head, succinctly expressed. Would Andres allow me to quote this comment elsewhere? (Tagline the usually excellent commenter Andres from Crooked Timber)?
reason 07.23.20 at 6:33 am
As for the argument that Economists are methodologically better than other professions it seems from my perspective quite funny. I’m just a couple of years older than John, but when I studied at Sydney University, having been invited to the AAMT summer school the year before, I found it was quite difficult timetable wise to combine mathematics courses with economics courses, the economics courses having been designed to be taken combined with accountaancy. In the third year I was unable to continue with the mathematics I had taken in the first two years.
But anyway this missed the point. The problems with the economics methodology come not from the techniques but from the underlying framework, see Andres @5.
rogergathmann 07.23.20 at 7:30 am
John, I’m surprised by your last sentences about loonies out there re climate change, since you have talked about a loonie, Richard Tol, who was influential enough that his claim that climate change would be good for the economy was included on the Intergovernmental Panel on Climate Change in the UK, and then had to be withdrawn. https://www.theguardian.com/environment/2014/oct/17/ipcc-corrects-claim-suggesting-climate-change-would-be-good-for-the-economy
I suppose my own sense is: it is not what the mass of economists think, as per a poll, but what the most promoted and prominent economists think. That Tol was anywhere near IPCC is about the politics of economics departments and their interactions with a panoply of business interests. Every sensible economist, I imagine, knew that Obama’s stimulus was just too small, but Reinhart and Rogoff write a nonsense book about deficits and it gets spread all over the mediasphere. Or, to front a more recent case, Yale’s Fiona Scott Morton, fresh from her work in the Obama administration and a supposed go-to person for the question of big tech monopoly, calmly accepts money as an advisor for Amazon and Apple. https://www.bloombergquint.com/politics/amazon-apple-hire-one-of-big-tech-s-most-prominent-u-s-critics
This is why the argument that (many and prominent) economists are, in fact, longtime pushers of public disinvestment and leaving production and services largely in the ands of the “market” is not countered by polls of economists. It is countered by, say, some economist of the stature of Larry Summers – a convenient proxy – arguing in the Clinton or Obama administration for vastly increasing the role of government in an economy in which the countervailing power of labor is in terminal decline. For instance. And, in particular, for investments in public health that are proportionate to the crises we are facing, which were not hard to see – Mike Davis pretty accurately predicted what has happened in Monster at our Door in 2005, including the structure of the epidemic inoculating corporate food chain, from wild markets to vast hogatoriums.
J-D 07.23.20 at 11:31 am
Fake Dave
If my intended meaning had been ‘I think you are wrong about Paul Krugman’, then what I would have written would have been ‘I think you are wrong about Paul Krugman’. I didn’t write that because it’s not what I meant.
When I wrote that I do not think the word ‘sully’ means what you think it means, what I meant was–
–that I do not think the word means what you think the word means.
But maybe I was wrong. What do you think the word ‘sully’ means?
Lance X 07.23.20 at 3:19 pm
18 wrote:
“Looking on the field from the outside, it seems to be completely full of mathematical mumbo jumbo. By this I don’t mean “you can’t use maths to study human behaviorâ€, I mean the maths in economics is wrong. Economists don’t seem to understand units or dimensional analysis, and their basic equations are hideously unphysical with their weird obsession with raising everything to arbitrary powers. It’s pretty routine to find economics papers with no connection between the model used for physical evaluation and the model used for theoretical exposition. Economists don’t seem to know anything about time-to-event models, and it seems pretty clear to me that economics as a field hasn’t moved beyond OLS regression – when economists study rates it is so common to find them using OLS on logged mortality rates, and they use OLS to study proportions (and measure differences in proportions rather than odds ratios) so often. Indeed, it’s only in economics that you still find the horrible words “Linear Probability Modelâ€. Economists also don’t seem to understand standardization (see Case & Deaton’s unfortunate excursion into “deaths of despairâ€), and often don’t adjust for age or sex. The main contribution of economists to my field has been the incremental cost effectiveness ratio (ICER), which is mathematical nonsense and statistically intractable but sadly has become the norm in economic assessment of medical interventions.”
Wow…you just seem to commit the very mistake you accused Tom of doing.
Tom 07.23.20 at 4:19 pm
@Faustusnotes #46. There are many points in both of your comments:
1) Why number of claims as dep. var. rather than population rates? I do not have strong priors on what to use. Clearly, number of claims can vary across districts because districts have different populations: they include a control for population and, most importantly, they include district fixed effects which soak up any (time-invariant) population differences across districts. I looked more in details to the paper and, in Table B.5 they run the regressions using population rates: they claim that “Despite some dilution of the signal and increase in standard errors, we find qualitatively similar results”. I actually do not see this and so, in practice, the distinction seems to matter here. I have a hard time see why it matters given that they use a relatively short time frame and I find it implausible that districts in lockdown saw their population increase so much so that the #of claims went up: if anything, usually you see some people moving away from areas in lockdowns. But anyway, your point seemed to be that, in general, not using population rates is wrong. I am afraid I do not see this if one also controls for population among the regressors.
2) Economists, at least in this type of applied policy papers, focus a lot on getting consistent estimates of the coefficients of interest and not so much on predictions. They can live with negative predictions because they occur for observations that have such extreme values of the regressors that you would expect a model to perform badly there anyway. Having said that, I agree that we could pay a bit more attention to predictions, just to see if the model has a decent fit or not. In other words, I agree with your point about regression diagnostics.
3) Now, onto the more fundamental objection, that they use the wrong probability distribution. Using OLS allows them to use districts FE and month*year FE. Now, in this specific case, you could do this with Poisson too. But you would be imposing strong distributional assumptions, not only on the relationship b/w y and the x’s but also between the y and the fixed effects. In the world I live in, data on my dependent variable of interest does not come with a label that tells me its exact data generating process (I wish!). As I am sure you know, count data often display excess zeros so that in practice Poisson models are often estimated allowing for over-dispersion. But then you do not have a Poisson distribution anymore (indeed, we call this approach pseudo-maximum likelihood). Or one can use another count model such as the negative binomial. Economists often just say: well, you know what, let’s just run OLS and, whatever the correct probability distribution is, if we have a large enough sample size (# of districts in this case), we will let the CLT work its wonders and give us consistent estimates and SE of the coefficient of interests. Again, I think there is scope for reasonable disagreement here: but I dispute your claim that this approach is “shocking”.
ps: thanks for the thoughts on Romer.
SusanC 07.23.20 at 4:31 pm
There’s at least two aspects to this:
A) the reputation for honesty and competence (or lack thereof) of the practitioners of different disciplines
B) what to include and what to leave out of the models the government uses for plannng.
As to (a), whatever economists might think of themselves, their reputation outside their own field is pretty low, to put it mildly.
For (b), I was a little surprised to hear the U.K. government wasn’t including economic factors in the modelling SAGE was doing to plan the covid19 strategy.
A possible justification: the predicted death rate of “do nothing†under a model that ignores economic factors was, if I recall correctly, on the order of half a million people.
So you can say to the economists: is this additional economic factor going to kill on the order of half a million people in the short term. If the answer is no, then we can declare the effect small and decide to leave it out. (Note that this argument does not work for modelling normal seasonal flu epidemics, because the death rate from ordinary flu is not big enough to render other factors relatively small).
Tom 07.23.20 at 5:05 pm
I wrote in haste and conflated different things. In 3) above (or maybe below if this message appears first) I meant to say “well, you know what, let’s just run OLS so that, provided the exogeneity conditions are met (error terms are uncorrelated with the regressors), at least the coefficients have a natural interpretation as average causal effects. On top of that, whatever the correct probability distribution is, if we have a large enough sample size (# of districts in this case), we will let the Central Limit Theorem work its wonders and give us the (asymptotically) correct standard errors of the coefficient of interests.”
Andres 07.23.20 at 5:37 pm
reason @49: If you need to quote me that’s fine, but without the adjectives please. I have said enough incorrect things on occasion, in and outside of CT, that “usually excellent” would be an overstatement (but so would “usually stupid” which some critics would prefer). Besides, having to abide by a “usually excellent” QC standard would make me post less frequently…
Jay 07.23.20 at 11:13 pm
I definitely agree with (3) and (4). I agree with (2) insofar as this is an epidemic and it would be really surprising if the people who’ve studied many epidemics over decades don’t know more about them than economists who are mathematically literate but largely ignorant of the subject matter. For example, epidemiologists know that viruses evolve, so letting them run rampant has a good chance of making them much more dangerous. This has already happened for the novel coronavirus.
I’d agree with (1) if the topic was, say, interest rates.
Mircea Popescu 07.24.20 at 1:42 pm
Or rather it could be observed that worldwide socialism having meanwhile managed to destroy money in its realms, the point of even having economists in English speaking lands is altogether dubious, so the priesthood is reforming itself.
Fake Dave 07.24.20 at 2:04 pm
@52
Sorry, I thought you were engaging me in a discussion when you were really just ineptly trolling me in the guise of playing language police. My mistake.
Andres 07.24.20 at 5:54 pm
One last thought. While the methodological dimension I described in @9 is important, there is the additional motivation that some mainstream economists are reacting negatively to epidemiological models and the current spotlight on epidemiologists because the entire pandemic and crisis has highlighted as never before that there are things more important than the economy.
Economists always start 101 courses (and practically every other course until you get to judgment-free econometrics and judgment-focused applied courses on economics and the environment, law and economics, etc.) by focusing on market exchange and the normative goal of maintaining or increasing the upward trend in the volume of real market exchanges. The pandemic has now brought the exclusive focus on markets and growth into question and has also put the spotlight on what one dissident wing of the profession—ecological economics—has been pointing out for some time, which is that economic growth is well on the path to being unsustainable and is starting to have unpleasant Malthusian side effects.
This negative reaction actually transcends ideological boundaries. Liberal economics bloggers such as Peter Dorman and Pro-Growth Liberal are obviously not attacking epidemiologists, but they have been on the warpath for some time now against the de-growth implications of ecological economics. The negative reaction of some economists against epidemiologist models is, I venture to say, a related phenomenon.
J-D 07.24.20 at 11:01 pm
You are correct this far: I was not seeking to engage you in a discussion. It was my intention to hint that you might possibly benefit from looking up the word ‘sully’ in a dictionary. Personally I enjoy consulting dictionaries, but if your preferences are different I have no more to say about that.
I was also genuinely curious to know what it was that you thought ‘sully’ meant, but if you won’t tell me I will just have to satisfy myself as best I can with my own list of guesses, currently headed by ‘flaunt’ (a word which I have just looked up with pleasure in several online dictionaries to find out how, if at all, they treat its increasingly common although originally erroneous usage as a synonym for ‘flout’).
faustusnotes 07.25.20 at 2:04 am
Tom, you really make my point for me when you confidently write these responses on the assumption that what you learnt in your field is correct. Let me go briefly through your three points.
1) Not using population rates is wrong because there is a specific relationship between disease cases and underlying rates. That is, the disease occurs at a rate in a population and thus as the population increases the disease should generate a specific number of cases from the rate. We model the effect of confounders or risk factors by how they modify the rate. This study allows the rate itself to be raised to the power of the coefficient. You can see how this works if you look at the whole thing in log space. This is about the basic way that Poisson processes work. What you are proposing, if applied in physics, would mean that basic theories of radioactive decay didn’t work. It’s shocking. Doing this is a sign that economists don’t understand Poisson processes, rates, or diseases and social phenomena (like violence) that depend on them. Terrible.
2) Here you reveal the real underlying problem of economists’ mathematics when you say economists can “live with” negative cases. This is an unphysical model. No physicist or engineer could possibly get away with a model that fails to predict the physical world in any meaningful sense. You also infer a little much when you say you don’t mind if they occur “in extreme cases”. There is no restriction on these models to make them happen in only extreme cases. With a downward time trend it’s possible you could observe negative cases with a one step ahead prediction (i.e. predicting next year’s events from the trend); it is mathematically guaranteed that if you go to a small enough scale (village, postcode, whatever) you will see negative cases in all the confidence intervals. This means if you ran a Bayesian version of this nonsense model you would be able to calculate the probability of observing negative cases. Epidemiologists have to present our work to policy-makers. What do you think those policy-makers would do with our work if we were predicting negative deaths? start preparing for the zombie apocalypse? It’s just insane. Also, your estimates won’t be efficient.
3) It’s interesting that in the same thread where JQ says economists move beyond OLS you say that a Poisson distribution imposes strong distributional restrictions on the model. Why do you think the assumption of conditional normality is not a “strong distributional restriction”? Why do you privilege the normal distribution? This is the essence of my case here: economists privilege the normal distribution over distributions that actually describe the physical process, because they have an (erroneous) idea that the estimates from these models are efficient (they’re not – they’re only efficient if the response meets the conditions of conditional normality, which it cannot). The assumption of conditional normality is a much worse assumption for a physical process like mortality or disease cases because it does not reflect the fundamental physical process. It has all sorts of bad consequences, which economists seem to be all too willing to ignore. (Also incidentally the idea that you can’t have interactions in a Poisson model is a super weird fiction that economists have invented for themselves).
This is all particularly weird in the modern era because the difference between doing a model with the correct distributional assumptions vs. doing the wrong model is literally just a word in Stata – do I type regress or poisson – and it takes no extra time from anyone’s day to do it correctly from the beginning (okay I grant you you may need an extra line to calculate the log of the population and you do have to type the word offset, two brackets and a variable name). It’s not like 30 years ago where you might have to do these calculations by hand, or program the likelihood maximizer yourself.
There’s a bunch of other things wrong with applying OLS to rates that I haven’t mentioned here. And yet still, in 2020, economists defend this practice, even have blog posts about it, and still this kind of ugliness gets published on the front page of the NBER, or in major economics journals. And then, of course, there is the problem of dimensional analysis. What you’re doing is not mathematics, and it’s not good statistics.
ph 07.25.20 at 2:40 am
@ 41 “Paul Krugman does this all the time with public health, climate, and (especially) electoral politics and it’s done more to sully his own reputation for competence and consistency than to genuinely uplift those fields.”
Krugman’s not alone. In one sense, it’s odd that we’d expect expertise in areas outside any expert’s specialisation, or discipline. My view is that hubris plays a key role, and perhaps a need/desire to ‘help’ in lending one’s name to a worthy cause. I agree, btw, Krugman’s forays outside economics have damaged his own reputation and lay confidence in economics, while adding practically nothing to uplift discussions on non-economics topics.
And while we’re on the topic of adding nothing, you’re not being ‘trolled’ regarding your use of the verb sully. How anyone who boasts of ‘reading dictionaries’ could manage to misunderstand how ‘sully’ operates in your sentence is a lesson of a kind in itself. https://www.google.com/search?client=firefox-b-e&q=sully+etymology
Comments on this entry are closed.