Winner of the New Statesman SPERI Prize in Political Economy 2016


Wednesday 26 October 2016

Being honest about ideological influence in economics

Noah Smith has an article that talks about Paul Romer’s recent critique of macroeconomics. In my view he gets it broadly right, but with one important exception that I want to pursue here. He says the fundamental problem with macroeconomics is lack of data, which is why disputes seem to take so long to resolve. That is not in my view the whole story.

If we look at the rise of Real Business Cycle (RBC) research a few decades ago, that was only made possible because economists chose to ignore evidence about the nature of unemployment in recessions. There is overwhelming evidence that in a recession employment declines because workers are fired rather than choosing not to work, and that the resulting increase in unemployment is involuntary (those fired would have rather retained their job at their previous wage). Both facts are incompatible with the RBC model.

In the RBC model there is no problem with recessions, and no role for policy to attempt to prevent them or bring them to an end. The business cycle fluctuations in employment they generate are entirely voluntary. RBC researchers wanted to build models of business cycles that had nothing to do with sticky prices. Yet here again the evidence was quite clear: for example data on real and nominal exchange rates shows that aggregate prices are slow to adjust. It is true that it took the development of New Keynesian theory to establish robust reasons why prices might be sticky enough to generate business cycles, but normally you do not ignore evidence (that prices are sticky) until you have a good explanation for that evidence.

Why would researchers try to build models of business cycles where these cycles required no policy intervention, and ignore key evidence in doing so? The obvious explanation is ideological. I cannot prove it was ideological, but it is difficult to understand why - in an area which as Noah says suffers from a lack of data - you would choose to develop theories that ignore some of the evidence you have. The fact that, as I argue here, this bias may have expressed itself in the insistence on following a particular methodology at the expense of others does not negate the importance of that bias.

I do not think this is just a problem in macroeconomics. David Card is a very well respected labour economist, who was the first to present detailed empirical evidence that imposing a minimum wage might not reduce employment (as the standard supply and demand model would predict). He gave an interview some time ago (2006), where he said this about the reaction to this work:

“I've subsequently stayed away from the minimum wage literature for a number of reasons. First, it cost me a lot of friends. People that I had known for many years, for instance, some of the ones I met at my first job at the University of Chicago, became very angry or disappointed. They thought that in publishing our work we were being traitors to the cause of economics as a whole.”

As Card points out in the interview his research involved no advocacy, but was simply about examining empirical evidence. So the friends that he lost objected not to the policy position he was taking, but to him uncovering and publishing evidence. Suppressing or distorting evidence because it does not give the answer you want is almost a definition of an illegitimate science.

These ex-friends of David Card are not typical of academic economists. After all, his research was published and became seminal in subsequent work. Theory has evolved (see again his interview) to make sense of his findings, but unlike the case of macro the findings were not ignored until this happened. Even in the case of macro, as Noah says, it was New Keynesian theory that became the consensus theory of business cycles rather than RBC models.

Yet I suspect there is a reluctance among the majority of economists to admit that some among them may not be following the scientific method but may instead be making choices on ideological grounds. This is the essence of Romer’s critique, first in his own area of growth economics and then for business cycle analysis. Denying or marginalising the problem simply invites critics to apply to the whole profession a criticism that only applies to a minority.



49 comments:

  1. Possibly your most important blog ever.

    I invite you (as I invited Mark Thoma over 10 years ago) to consider that if the majority want to be free of criticism that only applies to a minority, they need to actually, loudly, clearly and forcefully disown that minority in a timely manner.

    ReplyDelete
    Replies
    1. "Metatone26 October 2016 at 02:42
      Possibly your most important blog ever."

      Leave us not exaggerate.

      Delete
    2. Karl Marx would surely have disagreed.

      Delete
  2. Thomas Kuhn wrote the Nature of Scientific Revolutions over 50 years ago. In it he wrote that paradigm shifts occur when the 'Old Guard" retire or die. In other words people get locked into their belief frameworks and are not usually persuaded to change them by reason and evidence. It is the exceptional few who do this.

    ReplyDelete
  3. Such a nice and polite critique. Only an Oxford Professor would do such. Here on my side of the pond, amid a rather comical election cycle, we see "economists" not only disregarding any semblance of common sense but coming forward with "facts" that prove black is white, up is down and forward is backward. It seems that these chaps will spew any nonsense that will get their party or candidate elected, or that will get them a check in their bank account. There is no attempt to be scientific or analytical, but only to appeal to an uneducated electorate. They do great disservice to us all.

    ReplyDelete
    Replies
    1. How many academic economists from better universities support Trump?

      Delete
    2. see http://thehill.com/blogs/pundits-blog/presidential-campaign/297719-economists-have-a-message-clintons-policies-are

      *not* all of the usual suspects but not all marginal either. Some might ask if all are economists (I checked they are profs)

      Eugene Fama, University of Chicago
      Michael J. Boskin, Stanford University
      Christopher Flinn, New York University

      Gordon Alexander, University of Minnesota
      Frank Murray, University of Minnesota
      Stephen Parente, University of Minnesota

      Mario Rizzo, New York University

      Ray Ball, University of Chicago
      Charles Calomiris, Columbia University
      Gregory C. Chow, Princeton University
      Jerry Zimmerman, Univeristy of Rochester

      I'd say the profession comes out looking OK actually. Trump is a bridge too far many extremely conservative economists.

      Delete
    3. Do these count?

      http://www.ssc.wisc.edu/~mchinn/Statement_Economists_Concerned.pdf

      As Brad DeLong noted, the mailing address at the end is the Trump Tower in New York.

      Ok, perhaps it's a bit unfair as they are apparently reluctant to openly support Trump's policies, but they are certainly willing to act as his attack surrogates.

      Delete
  4. Economists choose economic models primarily based on whether adopting them gives a better chance to win tenure at prestigious institutions and makes more likely to get $1,200/hour consulting fees as for G Hubbard.
    There is a market for ideas, and ideas that reduce the chances of tenure or getting $1,200/hour consultancy fees don’t sell too well in that market, regardless of their other qualities, as to “real-world economics”.

    Thus models that “prove” the central truthiness of Economics that the income distribution is optimally and uniquely determined by productivity absent government intervention sell a lot better in the market for ideas, regardless of whichever assumption need to be adopted to “prove” that truthiness.

    If endowed chairs at Harvard and $1,200/hours consultancy fees were funded by minimum wage workers and the unemployed, or at least by the non-existent All-American Union of Walmart Shelf Restockers, perhaps “heterodox” political economy would become more attractive in the market for ideas.

    That is, probably “nearly everybody” realizes how unrealistic and contradictory are the assumptions behind “orthodox” models, but very few care, because those models sell well in the market for ideas, and it is largely pointless to go on about it; the lesson of the Cambridge Capital Controversy whose outcome was judged a “waste of time” by Economists benefiting from tenure at prestigious institutions and $1,200/hour consultancy fees, should be relevant here.

    ReplyDelete
  5. Prof. Wren-Lewis:

    "Yet I suspect there is a reluctance among the majority of economists to admit that some among them may not be following the scientific method but may instead be making choices on ideological grounds."

    That seems to me to be an exact description of what you do.

    "I cannot prove it was ideological" - which means that it is always deniable. And you will no doubt deny it.

    ReplyDelete
    Replies
    1. But you could follow my post and say that I ignore key pieces of evidence when it is inconvenient for my ideology or politics. I look forward to your response.

      Delete
    2. The question was whether YOU admit to being aware of ideological biases in your own work not whether someone else is capable of proving them.

      Delete
    3. And if I reply as I do to the Q below, that just shows I'm denying my bias? I think they did something similar to those accused of witchcraft. Evidence is crucial!

      Delete
    4. Thank you. We have your answer.

      Delete
  6. Can you be honest about your own ideological biases, Simon? Please list them in reply to this comment.

    ReplyDelete
    Replies
    1. I can be honest and say that I don't let my politics influence my economics. If you want to claim otherwise, see my reply immediately above.

      Delete
    2. Isn't that what they all say?

      Delete
    3. I suppose you might claim that your politics and economics tend to be aligned because the causality runs from the latter to the former. As it should, and perhaps does.

      Delete
    4. Anonymouse29 October 2016 at 13:44

      How do we know that doesn't apply to the others, too?

      Delete
  7. In the Greek Myths Procrustes offered a unique proposition to the weary traveller. He guaranteed that whoever you were he had a bed that would fit you precisely.

    Of course once the deal had been made, the reality emerged. There was just one size of bed. If you were too short you were put on the rack and stretched to fit. Too long and your legs were shortened with an axe.

    So it has been with the economies across the world for decades now. They have been promised riches by the snake oil salesmen who try to pass themselves off as scientists and then rammed into a 'one size fits all' ideology, based around the false idea of equilibrium.

    All the world economies have been forced to fit this peculiar bed, and have had their policies tailored to the idealised model, with the actual underlying people rammed into the structure regardless.

    The consequence of this should be obvious, but seems to pass people by. The data collected from such economies will resemble the data you would expect from the model.

    It really can't do much else - since the policies are there to prevent deviation from the norm. Therefore when you do empirical studies what you are actually using as data is the output from a system rammed into an inappropriate model. The data is tainted and what the taint means has to be understood.

    It becomes part of the Orwellian self-referential structure that reinforces the norm - academic journals where the peers doing the reviewing 'believe in the concept' and therefore reject anything novel out of hand, awards created by people who 'believe in the concept' and awarded to other people based on how hard they also 'believe in the concept', and data used to try and validate policies that are derived from structures that are already moulded into the form of the policies.

    If you observe the movement data from a man in a straitjacket then you will form a particular view of the movement capabilities of a man. The problem is that you are neglecting to notice the straitjacket or the possibility that it could be removed.

    Moreover even if you do notice the straitjacket, it is impossible to find an instance of a man without one, or to get into a position where you could remove it and see what happens. To move forward we really need a world where the restrictions can be removed. The mainstream may have locked out the physical world with their politics and policies, but these days people have alternatives to the physical world that they can immerse themselves in.

    You don't just run agent based models, you actually invite people in to play them. Multi-player games involving hundreds of people allow you to see how real people respond to changes in policy.

    You get much more accuracy if your non-player characters watch actual humans playing the game and train their neural nets based upon those responses.

    ReplyDelete
  8. Thanks!

    But note that what I was saying was that even if the other problems in macro are fixed, uninformative data will still be a huge obstacle to getting theories that work. There's a lot of progress to be made in discarding stuff like RBC, but that wall is still looming up ahead...

    ReplyDelete
  9. Yes. Tell me, when will the academic left admit that most public spending, if it's to be assessed on the basis that it more than pays for itself in employability and reduced claims on the state, is in fact money down the drain? "New Labour, after all, doubled health and education spending in real terms, yet productivity went nowhere and we're sicker than we ever were. We're sending half the population into higher education. But the universities are structured so as to minimise the contact between the staff and the students, there's little assessed work, there's little structured feedback and little in the way of skill development. The producer capture is almost total. Yet we hear not a word about this. When will the dogged adherence of the left to wrong-headed and self-serving beliefs be addressed?

    ReplyDelete
  10. "Why would researchers try to build models of business cycles where these cycles required no policy intervention, and ignore key evidence in doing so?"

    As you say, the reason lies in ideology. I cannot see how anyone could question the motives of those that push these models. The motives are clear. How can anyone argue that involuntary unemployment does not exist? Who needs to collect data to evidence the fact? If there is anything that undermines the New Classical approach it is this one fact alone. And sticky prices don't save New Keynesianism either. Mass unemployment results when businesses fail en masse. Businesses fail en masse or are no longer viable at a particular level of output for a range of reasons that have little to do with sticky prices.

    Henry

    ReplyDelete
  11. I definitely do not like your parenthesis. First, "unemployment is involuntary (those fired would have rather retained their job at their previous wage)": that "previous wage" comment is akin to me saying my not buying a 5-dollar-hotel-room Snickers bar was involuntary. Involuntary unemployment is a disequilibrium argument, where the equilibrium price is unattainable.

    Second, "evidence (that prices are sticky)": the evidence is that prices are persistent. Stickyness is a theory to explain the persistence. Could be there is an indirect cause.

    Maybe you are being ideological ;) Anyway, all in good fun.

    ReplyDelete
  12. And the reason mainstream economics has become so unbelievably corrupt is because it's the only "social science" that never underwent the postmodern revolution; there's no study of the field's history, no situating theory in its historical context, active discouragement of critical engagement among students, and most of all no concept whatsoever of the role of power in building and replicating the discipline.

    No wonder tenured profs can be bought off by the Republicans with the promise of increased textbook sales a la Greg Mankiw, while desperate adjuncts feel they have to spout off Randroid goldbug nonsense at their first and second year classes to get rehired.

    ReplyDelete
    Replies
    1. Sadly, the field has been examined as you would have it; the critique and alternative is around yet today, among the Veblen/Dewey/Ayres Institutionalists. See the Journal of Economic Issues for their work.

      Delete
    2. I fear it has been examined but not taught. A philosophy of economics should be the first things students learn, as it should for all subjects.

      Delete
  13. Identification? I wish you'd address the issue.

    I think it's possible that the business cycle is explained by waves of laziness. In 2009 there was a surge of laziness possibly reflecting millennial educational norms.

    ReplyDelete
  14. I don't think Romer is satisfied with the sticky-price New Keynesian model either. I think he has a problem with the overall intellectual dishonesty of the macro-profession which wants to standardise analysis according to a model based on questionable neo-classical foundations. Adding sticky price frictions to a model which is not ontologically sound is not being honest. Firstly, sticky prices in many cases are not the key friction. Secondly who said that we should talk in terms of 'market failure' and 'frictions' anyway. Answer: someone who said that you should look at an economy as something that 'should' operate like a neo-classical market. But is there any reason you should?

    Secondly, it is not only quantitative data you need to think about. The real reasons why you get liquidity traps or so on are to be found in non-quantitative analysis - eg looking at why banks and others made the decisions they did by looking at the records of their decisions.

    Economics has been the oddest of the social sciences from the beginning. Perhaps the chickens are now coming home to roost. But don't hold your breath.

    NK.

    ReplyDelete
  15. I don't understand why it is ideological to want to build a model that is transparent about behavior and is internally consistent. From here you can build better models (e.g. NK models) that can be useful in policy formulation.

    What would be ideological would be to take the basic RBC model and then argue that it describes reality.

    But in methodological terms NK models owe RBC a lot.

    It seems weird to want to airbrush RBC models away - methodologically - when NK models essentially use the entirety of its apparatus. NK models rightly include some heterogeneity in the goods markets and also (partly therefore are able to) bolt on some price stickiness.

    The size of the minority that believes RBC models describe reality must be very small?

    ReplyDelete
    Replies
    1. One thing New Classicals did not think they were doing was microfounding Keynesian models. They did not say we want to rebuild Keynesian models in a way that avoids the Lucas critique. I wish they had. I have no problem seeing the RBC model as a flex-price NK model, but for a long time that was not how it was seen by academic economists.

      Delete
    2. Currently standard NK models are transparent about behavior only because they make assertions about behavior which are transparently false.

      Or in other words how much did you get when you sold the asset which requires you to provide the buyer with one unit of good the next time your wage adjustment calvo alarm clock rings ?

      The claim is that aggregates behave as if people faced the problems agents face in NK models and used the techiques economist use to solve them.

      That's not being transparent about behavior in any way consistent with the ordinary English meanings of the words "trasparent" and "behavior".

      Delete
    3. Second comment on Magnus. It is odd to airbrush RBC models away, but odder to describe a post which discusses RBC models as an effort to airbrush them away. It is certainly true that the RBC literature had a huge influence on current macroeconomics. This means that the suspicion that it was based on ideology has current relevance.

      NK economics owes a lot to RBC. This should be a source of concern for NK economists who think that the RBC research program was crazy. As noted by (among others) Prof Wren-Lewis, the conventions which govern which convenient assumptions are allowed and which aren't are very important. I'd go a bit further and argue that they are the cause of all of the apparent content of economic theory. They are definitely not based on evidence. They reflect the judgement that some models are interesting and worth exploring.

      Those decisions based on the intuition of a very few economists are central to current NK macroeconomics. It is hard to argue that someone had no common sense at all when thinking about unemployment, but should be trusted when thinking about saving and investment. Almost all current macroecnomic research relies on that argument which is hard to make.

      Exactly because current NK macroeconomics owes a lot to RBC, NK macroeconomists should worry about having no respect for the judgment of people on whose judgment they rely.

      Delete
  16. have followed this debate with a lot of interest, both as someone with both a serious amateur interest in political economy, and a serious professional interest in epistemology (my day job is in advanced statistics, machine learning and modelling). It seems to me that economics as a a subject (and macroeconomics in particular) lives in the gap between those activities that are naturally idealist, and those that are naturally empiricist. I don't think there is ever enough information out there in the world to confirm or refute economic models decisively (esp. given the problem of Knightian uncertainty which seems to me to be relevant) - all there ever is is enough information to tighten up whatever priors you happen to have.

    I also think that this means that macroeconomics is a discipline where it is particularly easy for people to get away with bad faith arguments (or unconscious 'semi-bad-faith arguments, which amount to economic scientism), and lots of people visibly do.

    As a result, I have learned to be very wary around strong claims by macroeconomists, including the claims of macroeconomists with whom I am vey sympathetic unless they bring a lot to the table.

    ReplyDelete
  17. Noah has a propensity to seek a safe, inoffensive position in his discourse. I prefer economic discussions where, if the data confirms the baby is ugly, then the baby is called ugly - not some confrontation-avoiding assessment such as, "those whose visual acuity appreciates unconventional appearances enjoy the way the baby looks."

    ReplyDelete
  18. How is it incompatible with RBC that fired workers would "rather retained their job at their previous wage"?
    For an RBC theorist, a recession results from a negative shock to productivity. Of necessity, such a shock means the marginal productivity of some category of workers falls. Those workers will not be employed at their previous wage. Their unemployment is "voluntary" in the sense that they do not accept a job at a lower wage, at least not right away. Instead they spend some time in search, in a rational attempt to minimize the loss of their income.
    I think that is pretty much standard RBC reasoning, although in many stripped-down model descriptions the job-search sounds like a vacation. I don't claim it is especially compelling as a description of actual business cycles. But it does seem perfectly compatible with workers being fired when they would rather have continued to work at their previous wages .

    ReplyDelete
    Replies
    1. If firms were doing what you say, we would see them reduce wages after a negative productivity shock, and then workers quitting because they did not want to work at that wage. That is not what we see.

      Delete
    2. What?? The headlines in the business section aren't 'Acme Cuts Wages 20% and Hopes Nobody Quits'? Or 'Acme Cuts Wages 20% and Hopes Only the Least Productive Employees Quit'? Or 'Acme Management: We Haven't a Clue Who to Keep So We Cut Wages 20% And Crossed Our Fingers!'.

      Delete
    3. Agreed; in fact, it's the opposite of what we observe. In actual recessions, involuntary job separations (i.e. layoffs) rise, and voluntary separations fall (by a comparable amount, curiously enough).

      Delete
  19. In a few applications, even in microeconomics where we benefit from wealthier datasets, despite best efforts to provide a sensitivity analysis, there is still some space left to force point identification your way. Maybe you'll pick a type of treatment effect that just happens to rely on rather heroic hypotheses for your case; maybe you'll just rely on Wald statistics to do some of your crucial tests (we know this will lead to more rejections of the null than, say, using the score).

    A good solution is to see what happens when you provide very little structure to the data and include that in your sensitivity analysis. It's typically very convincing to infer an interval around a treatment effect and use non parametric estimators to point identify your bounds. The benefit is that this informs you about how, exactly, your modelling choices influence outcomes -- sometimes, this rules out a few extreme results that would otherwise be published and, other times, it tells you you can't say anything conclusive without committing yourself to very strong hypotheses. For instance, an instrument may not satisfy the orthogonality criterion -- your IV point estimate would thus be biased. However, you could still use it to formally bound your LATE. Ideally, you can at least put a sign on it.

    There is a drawback though: some of those bounds involve optimum operators and those aren't nice smooth functions you can easily use to derivate asymptotic behavior. It's frontier work in research.

    I am sure that if we applied that to business cycle issues, we'd find out what is the cost in terms of hypotheses of moving from one type of model to another. You likely have cases where, as Romer says, you can't be conclusive due to a lack of data -- and we could also see what it takes to be conclusive.

    ReplyDelete
    Replies
    1. But what I am talking about is simply ignoring key bits of data because they are inconvenient. I can't think of any other areas of economics that do that.

      Delete
  20. I am still confused why economics appears to rest on assumptions that while sounding logical do appear to reflect the nature of economic agents (ie people). These assumptions are not consistently applied. An ideological basis would help explain that. I think it also goes someway to the explain use of theoretical arguments dispute the existence of a large and robust set of empirical data. Still it might not explain it all.

    ReplyDelete
  21. But who is using RBC models these days?

    Also to get it straight, in RBC model people choose not to work at the market clearing wage. That's not such a ridiculous conclusion. Even among my acquaintances are some who says that would work if they could earn a "fair" wage (which is much higher then the wage they have been offered). RBC models are not perfect and there is a lot of reasons to criticize those models, but one needs to be fair in the criticism.

    ReplyDelete
  22. I agree entirely. It is true that there is very little macroeconomic evidence, but that doesn't explain why much of it is ignored.

    I do have one objection. I don't get "It is true that it took the development of New Keynesian theory to establish robust reasons why prices might be sticky enough to generate business cycles, but normally you do not ignore evidence (that prices are sticky) until you have a good explanation for that evidence."

    The current standard model of sticky prices is Calvo pricing. I don't think anyone thinks that's an explanation let alone a good explanation. Nor to I see any way in which it could be called "robust". I'd say that macroeconomists just decided to accept the fact that prices are sticky even though we don't have an explanation which is both plausible and tractable.

    I don't think that anyone who might be inclined to argue that new Keynesian economics is an improvement in any way over old Keynesian economics would base his case on the claim that price stickiness is explained by the Calvo fairy.

    ReplyDelete
  23. 'The scientific method' even in the hard sciences involves making lots of choices about what to research and which features to isolate for consideration. Rarely in the hard sciences, it is true, do we find the kind of obvious ideological bias that is often found in the social sciences, including economics, but that, surely, is only because much less politically is at stake. Because economics is all about who gets what and how much we should expect it always to be a battleground as much as it is a theatre of disinterested enquiry.

    ReplyDelete
  24. How does RBC explain the rise in unemployment over the great recession? If you are choosing leisure then you wouldn't be in the labor force. Unemployed people by definition must be actively looking.

    ReplyDelete
    Replies
    1. Oh don't worry about silly little things like that! (sarcasm)

      Delete
  25. Card's friends presumably fell into this camp:

    "When the truth is replaced by silence," said the Soviet dissident Yevtushenko, "the silence is a lie."

    Very important and admirable piece, Simon. Ideology is inescapable, but it need not prevent good work. I fear, however, that all too often, it does.

    On a side note, The University of Chicago must have been a very odd place indeed: how else to explain a site that employed Saul Bellow, Erving Goffman & Friedrich von Hayek as well as Card's erstwhile friends? Right wing economics, left wing sociology. Tolerant, I guess. Pluralistic too.

    ReplyDelete

Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.