Economics, theory, and data

This post was titled “Why data alone is not enough for economic inference”.  I was all prepared to write a post on the fact we need data and theory in order to do economic inference and create knowledge. I had links (*,*,*,*,*).  Then Noah Smith wrote this like really good post on the issue, so I’d suggest reading that.

On the other side there are those who are “too in love” with theory without any reference to data, or prior literature (which is a way of building a case for inference between theory and data).  A clear example of that comes from some of the comments that specific physicists moving into economics make – and Chris House has expressed that here.

I’d note that I want physicists to come in and add tools and debate to economics – things like agent-based modeling, which they are more familiar with, are growing exponentially in popularity in sociology and economics (due to recent data availability), so pro-tips from this would actually be superb!  Also keeping economics in a silo from other disciplines would be patently ridiculous – the questions we ask are too important not to have as many ideas as possible competing to create knowledge.

But many of these econphysicists (not all) that are most vocal they need to learn to accept the idea of “many models” and the lack of a general theory for economics (I am sure most physicists do in their own discipline, the people Chris mentions do not).  Even more importantly, these physicists do not understand the normative-positive distinction with economic modeling – their determination to say what we “should” do based on one perspective is incredibly unsound.  There are multiple ethical dimensions involved due to the existence of individual choice, and economic models are predicated not on giving a “solution” to policy questions, but informing them of tendencies and trade-offs that need to be considered.

Examples of concerns

The two worst examples of physicists being snide at economists recently were this article and this paper – although for very different reasons.

The article frustrates me a lot more than the paper, as it is someone talking about terms they don’t have a complete appreciation of!  They complain about the microfoundations endevour, and then follow it up with a call to heterogeneity.  Explanatory models that include more heterogeneity are indeed a form of microfounding a model – so this in itself is strange.  He may defend it by saying “I mean the representative agent microfounding you guys do” – ok but then he is ignoring all the work on introducing heterogeneity, accounting for habit formation, forms of bias, etc etc and attacking a strawman.  Note:  Attacking the political implementation of economic policies is so very different to attacking the academic discipline – academics got pretty frustrated during this crisis as well.

And he makes claims about journal articles being all about rationality and single equilibrium … does he actually read journal articles, as this is not what I’ve found.  Instead there is a LOT about data and stylized facts, and a bunch about trying to isolate causes from idealized models given a range of assumptions.  The call about rational choice is an overused archetype of a concern – these issues are complicated, economists (since Herbert Simon – but also a lot prior in a more ad hoc way) tend to view rational choice through a lens of limitations to cogitative efficiency, and try to work in a framework that could incorporate that – this makes us similar to psychologists, and makes neuroeconomics exciting.

I have also heard the author of that article say that economists need to study the behaviour of idealised systems/an artificial world, and then think about how that translates and can help us answer questions.  This solely tells me the author doesn’t know what economists do – and is selling books attacking a straw man.  Why?  As the behaviour he wants us to exhibit happens to be what the economics method is!

Update:  James had a very nice post earlier in the month about some of the specific claims made by Mark, it can be found here.

The papers concept was reasonable enough, exchange rates exhibit Brownian motion.  However, it was just weird to see scientists claiming it and sort of saying “hey it’s like fluid and stuff and this is a new idea”.  Exchange rates are asset prices, and within economics and finance asset prices are constantly modeled using Brownian motion.  It just disappoints me that these claims are made in a way that make economists sound stupid, even when it turns out there is already a well used current literature on the topic in economics that doesn’t look like it has been read 🙁 – Note:  It may be in the paper itself, I don’t have access right now – but this doesn’t change the tone of the abstract!

The key point of House’s piece was that there is an existing literature, and many of these econophysicists don’t feel like they need to reference and read it and that they can come up with things from scratch.  This is a neat way of doing research without being incumbered by others “priors” – but when it comes to actually publishing results, claiming things as “truth” without reference to whether the literature is already there, or whether there are good reasons already given for these things, is pretty naff.

I world usually ignore this sort of stuff – except I constantly hear comments like this:

FFS.  There are a bunch of crazy smart people who have spent their lives studying these issues, and are in competition so would jump at the chance for a new model that outperformed!  But the odd random person with their “alternative accounting” knows the real truth and is being suppressed – conspiracy!!!!!

Trust me, if you have evidence of a forecasting model that outperforms Bayesian moving average combinations of time series and structural models then don’t tell me – get hold of your local central bank.  Given they have to “target their forecast” they are the guys who really really care about this stuff, and if its really improving policy they’ll be first on the line to try to get you your Nobel Prize.

Update:  I wrote this post as soon as House’s post came out.  It led to some arguments on Twitter – so I’m putting this up a day early, and also clearing myself up a bit.  John is right when he says this:

Yeah, that call was too far by House – that was part of the reason I’d written in this post about how much I want them to contribute, as I didn’t want that to get confused in my blog.  When House says “Neural-networks, agent based modeling, path-dependent equilibria” haven’t produced much I felt a bit uncomfortable – as they are areas that can offer significant marginal gain, and in the case of ABM where I see a lot of economists working (even I’m trying to build up capability in that field).

But my guess is it came from frustration, I read it as the sort of thing I feel after getting another call from a physicist/engineering telling me an economic variable is JUST X, and I’m stupid – and economists should feel bad.  I work in a consultancy, god knows why these people contact me, but anyway.  I say something like “interesting, what behaviour drives that, and what does that mean” and they’ll tend to say “it is X, are you listening!!”.  This is a philosophical gap pure and simple, and we need two way communication – instead of insults – to bridge it.

We solve that, and physicists will bring a lot to economics – not as much as psychology or biology – but still a bunch.

One extra point – often on the same day I get told economics isn’t mathematical enough (engineering, physics, drunk people in bars) and then get told economics is too mathematical (politics, sociology, drunk people in bars).  Could these people also argue with each other for a little bit, since they seem to have different definitions of what economics is, and what the “right” amount of math is (whatever the hell that means).  Is there a way to get them in the same room?

  • You might want to look at Micheal Keane’s Keynote Address ‘Structural vs. atheoretic approaches to econometrics’ published in the Journal of Econometrics (2010).

    Keane argued that it is not possible to learn anything of interest from data without theoretical assumptions. All statistical inference relies on some untestable assumptions


    In this paper I attempt to lay out the sources of conflict between the so-called “structural” and “experimentalist” camps in econometrics.

    Critics of the structural approach often assert that it produces results that rely on too many assumptions to be credible, and that the experimentalist approach provides an alternative that relies on fewer assumptions.

    Here, I argue that this is a false dichotomy. All econometric work relies heavily on a priori assumptions.

    The main difference between structural and experimental (or “atheoretic”) approaches is not in the number of assumptions but the extent to which they are made explicit.

    Also look at Keane, Michael P. 2010. “A Structural Perspective on the Experimentalist School.” Journal of Economic Perspectives, 24(2): 47–58 where he argues that we cannot begin a systematic assembly of facts and empirical regularities without a pre-existing theoretical framework that gives the facts meaning and tells us which facts we should establish.

    Koopman’s famous 1947 essay ‘measurement without theory’ argued that without theory, you do not know what to “look for” and what phenomena to observe and what measures to define and compute.

    • Sounds good, especially since it reinforces my priors 😉

      More seriously though I will have a look.

  • Kimble

    Fricken engineers! Amirightorright?

    • Hehe. Tbf, lots of economists started as engineers – but then came into the discipline, did a whole bunch of analysis, and helped build shiz up … rather than sniping from the sidelines and ignoring current lit. That’s the difference that matters right there.

  • Avishek Roy

    Show your outstanding engagement for your dearest mom, friends and family members through extravagant Hamper to everywhere
    in Hong Kong and make them special forever in your life and spread happiness and brightness in the lives of the people you love on Mother’s Day. Mother’s
    Day Flowers Delivery within Hong Kong
    for your lovest mother.

  • Nishan Roy

    The sweet surprises in the form of gifts create unforgettable moments that are restored in our hearts forever. Plan such a surprise gift for the people your mother whom you love
    and care. Send her amazing Flower arrangements, Cakes, Chocolates, Sweets, Apparels, Cosmetics, Jewelries, and lots more just with the help of the online shopping portals on the Mother’s Day. Cheap Online Flowers Delivery in Philippines Same Day for your lovest Mother.