Tuesday, May 1, 2018

Making friends: David Orrell

On occasion when I mention things about econ twitter or my blog, my wife responds sarcastically with "Oh no. Are you making friends again?" with the obvious connotation (because she knows me well) that I am starting fights. I'm not sure what sets me off, but one ingredient is usually some reference to physics in the context of economics. The latest version got a response from the subject David Orrell in comments. This is actually a two-part blog post, the other having been scheduled earlier and contains much more detail about the numerous fabricated and out of context quotes in an article by Dirk Bezemer about claims to have predicted the global financial crisis or global recession.

Below, I have responded to David Orrell's comments on my post. Given how blunt I was, David was relatively nice. But this is exactly the kind of discussion I'd like to be having: What does it mean to make macro- or micro-economics more scientific? My general feeling is that what comes under the broad heading of alternative or "heterodox" approaches is strangely exactly like "mainstream" economics, just with different jargon or mathematical decorations. Neither appear to value empirical data. Neither appear to make accurate forecasts. Neither appear to reject models or give up models that so complicated relative to the data they can never be rejected.

Here is my response:

Feynman

In response to my claim that he is taking Feynman's quote out of context, David writes:
There are two contexts, the local one (the specific paragraph) and the larger one. Here I would say Feynman is saying something that he felt to be generally true, namely that prediction serves as a way of testing models.
David seems to cede that the quote is taken out of context (one in which Feynman himself said models could probably not predict the existence of thunderstorms without observational data from the Earth). The "larger" context that prediction is one way to test models is fine, but that then cedes David's entire point that being unable to predict the global financial crisis somehow should lead us to debate the "scientific legitimacy" of macroeconomics.

Forecasting is one way to validate models. "Retro-diction" is another fine way (Einstein gave us a retro-diction of Mercury's precession). Being able to make post hoc descriptions of the data without altering the model in order to encompass new data is another. Some sciences can't forecast particular phenomena — geologists cannot predict the timing and magnitude of earthquakes. That doesn't make us question the scientific legitimacy of geology.

Bezemer

In reponse to my claim that the Bezemer article appears to have fabricated its examples of predictions of the global financial crisis, David says:
That is a serious charge. Have you checked with the author to get his side? Also, your critique seems to focus on one example of several given, but your statement as written implies they were all fabricated, which is certainly not true.
Yes, it's serious. But no one seems to care.

In the original link I gave, I actually wrote about 2 quotes from Godley and 1 from Keen (not just one example). Given the sample of three quotes were 100% fabricated, I hadn't gone any further. However, today on my lunch break I checked up on another 3 of the 12 listed in a table at the end of the article here, for a total of 6 out of the 12 claims of predictions of the global financial crisis. The same fabrication occurs in each of the six. Only one of those (Sorensen) appears to mention a global crisis and even his quote is taken out of context altering its meaning.

However, when presented with this kind of evidence, Feynman's 'leaning over backwards' approach would not be to claim (apparently without evidence) that it is "certainly not true" all were fabricated but rather check the quotes yourself. I took Bezemer at his word initially as well. But after learning that Keen's claim of predicting the global financial crisis was at best on shaky ground, when I again encountered Bezemer's paper I decided to check for myself.

Just to be clear: I am not blaming David for citing the paper. I originally thought it was legitimate myself. It was cited in the New York Times. Other people I know and trust have cited it. Most of us come to the table with a certain amount of trust that other academics are operating in good faith.

The usual excuse

I noted that David gave Robert Barro's (!) [1] excuse for why macroeconomic models failed to see the financial crisis, and so I gave the one that is much more representative of the field. David replied:
The excuse you mention that the event was simply unpredictable is just a variation of “no one else predicted it either” as stated. There was enough data to warn of a storm, which is why a number of economists did exactly that. The problem was that most economists were looking in the wrong place – in large part because their models were misleading them.
First, despite what the Bezemer article claims, some people noted some increases in housing prices in some countries might be unsustainable and lead to recessions in those countries, not a global financial crisis. And we should be very, very careful about survivorship bias here. Several of these housing bubbles didn't pop, and one of the countries didn't even have a recession (Australia).

However reading Diane Coyle's, Mark Thoma's, and Noah Smith's version of the actual "usual" excuse as claiming it was "unpredictable" misunderstands them. No. Macro data is uninformative: it does not confirm or reject models, so it is impossible to know whether or not recessions are predictable or not (or what causes them). There was data that housing prices were off trend but today prices are above their 2007 peak in the US, and there were indicators of recession before that peak was reached. A decline in conceptions appears to lead the shock to the Case-Shiller index. Monetarists make claims that the housing bubble collapse and the recession were caused by monetary tightening starting in 2006.

Claiming it is somehow obvious the housing crisis caused the recession (when we don't know what a recession is or how they are caused) is more likely confirmation bias than empirically robust. We should try to lean over backwards more to think of any possible reason our theory might be wrong and present that evidence. I personally was once sympathetic to the monetary case, but have since changed my thinking [2]. I'm actually amenable to the Minsky view of overinvestment/collapse cycle for the past couple decades, but also think recessions are fundamentally social processes. Those are just ideas, and the evidence isn't conclusive.

FitzRoy
Obviously he started off in good shape because he was allowed to found the Met. Office in 1854! The point is that he lost that prestige.
That was David's reply to my comment that FitzRoy wasn't exactly outside of the scientific establishment. But I'll admit I now don't understand the purpose the FitzRoy analogy is at all. So science is about making inaccurate predictions until eventually you perservere and get better? How is this unlike macro making inaccurate predictions? Do we just wait until mainstream macro perserveres like FitzRoy?

Weather forecasts

In reponse to my comments about weather forecsts, David makes the point I am making in my blog post:
The standard test of forecasting is skill relative to naive forecasts (such as persistence or climatology). It also matters whether the phenomenon being predicted is important in itself, or is being chosen because it is easy to predict. Weather models are good at predicting things like the 500 mb heights quite far out, but are less good at things like low-level wind, temperature, and precipitation (i.e. what we call the weather), and storms are more difficult still.
As Chris Dillow says (and I quote him in my blog post): "what can people reasonably be expected to foresee and what not?"

...

Others

These are some other comments David made and my responses.
- “decried in Orrell's article via quotes” 
You mean statements in articles that I link to, which is not the same thing and you should make that clear.
I don't see how that is different. I say David is quoting sources that decry "neoclassical economics". I explicitly wrote it that way to say that "neoclassical economics" is being decried in quotes from others in the article written by David. I didn't say that "Orrell is decrying neoclassical economics" and it would be nonsensical to interpret it as Orrell quoting himself. However, David is choosing to show us quotes that decry neoclassical economics so the ambiguity as to whether David himself is decrying neoclassical economics using selective quotes or is just a conduit for the sentiments of the speakers is an open question that exists in the original. If I were to add that David is definitely not decrying neoclassical economics, I would be misrepresenting David's article.
- Papers are “based on reading Michio Kaku's popular science books” 
I have not read his books, though I did study quantum mechanics, and once taught it as part of a university level mathematics course.
I'll admit that was a bit mean (making friends), but in reading David's papers (and this Aeon article) there does not appear to be any substance to the metaphors using e.g. entanglement or other "quantum" behavior. As far as I can tell, David sees no difference between things that are "entangled", "strongly interacting", or forming a quasi-particle.

For example, from the Aeon article:
One of the more mysterious aspects of quantum physics is that particles can become entangled so that they become a unified system, and a measurement on one affects the other instantaneously [entanglement]. In economics, the information encoded in money is a kind of entanglement device, because its creation always has two sides, debt and credit [quasi-particle?, e.g. a Cooper pair] (for example modern fiat money represents government debt). And its use entangles people with each other and with the system as a whole [strongly interacting?], as anyone with a loan will know. If you go bankrupt, that immediately affects the state of your creditors, even if they don’t find out straight away.
These are different ideas in physics. Some more specifics about what this is supposed to mean here or in David's other papers would help.
- “There is no math in them”
That’s a deliberate choice on my part, because I am trying to make the topic accessible to an audience who may not be familiar with the quantum formalism (e.g. most people). In the book I do show worked examples where the math can be communicated graphically. And of course I cite many papers and books which do take a mathematical approach, if that is what you are after. One thing I point out in the book, though, is that mathematics is a double-edged sword, and can be used to obfuscate as well as communicate.
Wait, is David saying he might accidentally start trying to obfuscate with mathematics? I can only think of an old Space Ghost: Coast to Coast episode:

Tansut: Is it bad if a chicken bites you?

Space Ghost: Did a chicken bite you?

Tansut: Well, no. But he's gonna!

Space Ghost: Then go away from the chicken!

Then don't obfuscate with mathematics!

It is fine to make a deliberate choice to avoid math in a book for a general audience. I made exactly the same choice in my book. I'm not talking about that. I'm talking about the papers listed as "related research articles" published in Economic Thought. The journal appears to accept technical papers with math in them like this one [pdf]. My question was: since the ideas of quantum physics are mostly mathematical, where are the papers with the more explicit mathematical connection to the ideas of quantum physics? I may have mistaken David for a researcher in this field instead of him acting more like a journalist telling us about other people's work. In which case, I take it back: no math needed and I'll go look at the citations.
- “It's just the word quantum that sounds cool.” 
That actually has it the wrong way round. Using the word “quantum” outside its physics context is fraught with difficulty which is why most (not all) scientists still avoid it. One way to use the quantum approach is to say that it is just a more flexible way of describing probabilities. In the book I argue though that it makes sense to treat money as a quantum system in its own right. Just as physicists once adopted the Hilbert space as a way to model the behaviour of subatomic particles, social scientists can usefully adopt the same framework to describe things like the economy. But as I note in the book, it is interesting how much the typical objections to these ideas rest on the use of particular words.
I would love to see some papers. I can't find any that use Hilbert spaces or other quantum physics to describe empirical macroeconomic data. This one just makes an elaborate analogy, but no data. This one is just nonsense — PY = MV (!) — but also doesn't make any reference to data.

As I mention in my article, sure there are path integral approaches, but that isn't quite so "quantum" (and is probably more directly related to thermodynamics).
- “The galling thing is that Orrell starts with a quote about making predictions and saying economics is bad science, but then closes promoting a book that (based on the source material in his papers) won't make any predictions or engage with empirical data either.” 
That sounds like a prediction based on little data, given that you haven’t read the book.
My hunch is based on David's papers listed on a promo site for the book. If the book is wildly different from the way it is presented there, then maybe it will engage with data.

...

Footnotes:

[1] Barro is generally a fool, like most of the rest of the Harvard economics department that includes Alessina, Feldstein, Mankiw and Borjas. Whatever one thinks might help the less well off, they will find a way to show it doesn't work. But if you find a way to help the rich, they're totally on board.

[2] The apparent monetary tightening in the mid-2000s seems more likley to be the delayed response to the end of the demographic shift. The demographic shift is centered in 1977-8 and the response in the monetary base is centered in 1985 (about 8 years later). The fluctuations in women's labor force participation become correlated with men's in about 1998 (marking the end of the demographic shift), making a 2006 date for the end of the monetary response to the demographic shift completely plausible. I.e. the Fed wasn't tightening, and the apparent tightening was just the response to the ending demographic shift. Anyway, that's my best guess. It could be wrong.

2 comments:

  1. - “the quote is taken out of context (one in which Feynman himself said models could probably not predict the existence of thunderstorms without observational data from the Earth)”

    Feynman is dropping a general remark into another specific context. In my case, I am using it to sum up a commonly held understanding of how science works, with prediction forming a key part of the scientific method. As you say, the meaning of what counts as a prediction (e.g. it could arguably be something that is already known) is up for debate and I discuss that at some length in my book.

    - “Bezemer article appears to have fabricated its examples ... But no one seems to care.”

    Again, fabrication is a serious charge. Have you tried submitting it for publication, or contacted the author or editor or other people involved?

    - “it is impossible to know whether or not recessions are predictable or not”

    We are talking about the failure to warn of a particular financial catastrophe, or take steps to prevent it, or diagnose it properly. As Akerlof and Shiller wrote in Phishing for Phools, “It is truly remarkable that so few economists foresaw what would happen.”

    - “I now don't understand the purpose the FitzRoy analogy”

    There seems to be a pattern here where if I clear up something it only reveals some deeper problem (see below).

    - “neoclassical economics decried in Orrell's article via quotes”

    The way it reads is that I am making these statements in the article by quoting them. By that standard, you could have turned it around, and said that critics are decried in the article as “dangerous” etc. The right way to word it would be to say “decried in papers quoted in the article” though that doesn’t sound quite as snappy. I expect if I’d done it to Feynman you would have picked up on it.

    “Wait, is David saying he might accidentally start trying to obfuscate with mathematics?”

    Impressed that you can turn my statement around and use it to question my scientific integrity. This seems to be less about a serious response to my article or my forthcoming book, and more about your self-described irritation caused by “reference to physics in the context of economics.” One of the topics of the book is exactly the source of this irritation, which tends to afflict both physicists and economists, but which seems to apply especially to quantum approaches (not so much to e.g. classical thermodynamics).

    Anyway, I hope you do take a look at the book when it comes out, though I am sure you will find it irritating. But economics strikes me as an area where a little irritation might be constructive.

    ReplyDelete
    Replies
    1. "Again, fabrication is a serious charge. Have you tried submitting it for publication, or contacted the author or editor or other people involved?"

      Bezemer's paper wasn't peer reviewed or published in a journal. It's a pdf'd word document that's listed as "unpublished" in a citation from Bezemer in a voxeu blog post.

      Why should I have to go to greater lengths than Bezemer did to point out the problems with his quotes in order for you to take it seriously? Look up the citations yourself. I linked to the sources. I'm not the one citing it, you are.

      ...

      The bit about obfuscating math was a joke; immediately following the line is a quote from a (admittedly obscure) comedy TV show that was on the Cartoon Network in the US in the late 90s and early 2000s.

      Delete

Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.