Can physicists please look at a basic textbook before releasing these things

FFS, this is probably the worst example of a physicist treating economists like idiots, and saying something both meaningless and already known, that I’ve seen for a while (via Marginal Revolution).

The basic inequality that plagues economies the world over may have a simple explanation—at least, according to physicists who’ve turned to economics.

Simple explanation aye, this sounds dodgy.  Let’s assume that they are talking solely about income inequality, as it looks like the article doesn’t understand the importance of the difference. Let’s see what the “simple explanation” is.

Pick a country, they claim, and you’ll find multitudes of people who earn next to nothing, a few who rake in plenty, and a distribution between the extremes that falls exponentially as income increases (see figure). That distribution applies to all but the very rich, they say, and it arises from an analogy to the concept of entropy, a measure of disorder in a physical system such as a gas. Just as a gas evolves to a state of maximum entropy, they argue, random churning in the economy ensures that the income distribution naturally tends to this inequitable form.

The reasoning is “not very close to the thinking of economists, but it’s pretty persuasive,” says Thomas Lux, an economist at the University of Kiel in Germany. But Frank Cowell of the London School of Economics and Political Science says “I’m extremely skeptical” that the argument provides any insight into the economy.

Two things that make me furious with Thomas Lux here are:

  1. This isn’t an explanation as there is no mechanism involving the people involved – it is a description.  These have value, but the two are very different, and descriptions are not policy relevant until we have a mechanism.  Update:  Good comment by Kiwi poll guy here regarding my first point being wrong – if described as inequality occurring due to luck it is an explanation.  However, it is A explanation not THE explanation.
  2. FFS READ THE LITERATURE.  Especially if you are going to say economists haven’t thought of this.

On the second point, the original (read first half of the 20th century) view of looking at income distributions was through a power law (Pareto distribution) and a log-normal distribution.  When it came to looking at inequality indices, one of the key indices used was the generalised entropy family, specifically the Theil index and log mean deviation.  The later two were the only distributions that met all the “axioms” economists put together for rating a distribution.

In modern times parametric forms have been studied to death, but with data increasingly bimodal (and these forms unimodal) there has been a strong push towards non-parametric analysis (without functional form, just using the microdata).  Teasing these things out is difficult, and any introductory book would have given these points – hell even read Cowell’s notes on measuring inequality here.  If you read Italian, just go back and read the Pareto and Gini papers from the late 19th and early 20th century and you will see a discipline that knows the points that are being raised (about the distribution being exponential, entropy measures came in the 1960s ;) ) and was trying to tackled with “value judgments” and “mechanisms”.

The “natural tendency” that can exist for the dispersion of income is something that is a centerpiece of economic analysis – instead of saying the dispersion happens though, we try to work out, and test, different mechanisms.  Hence this post by Shaz and Piketty’s text both talking about “natural” movements in income – pro-tip, what this “natural” is due to is an important element of whether leaning against it, or with it, is a good thing, and whether it is policy relevant.

If anyone thinks this is just “nit picking” and doesn’t indicate that the article is practically useless, then you don’t understand the basics of economics and policy making – that sounds strong, but it is so far away from being useful for policy or adding to knowledge that I’m comfortable saying it.  Note that I know practically nothing about physics.

That is cool, the article should be informing you of those things – but as it is written by someone with no understanding of these basic points they can’t inform you.  The article says virtually nothing and insults economists (directly) showing a complete lack of engagement with the literature that economists get to read in an undergraduate policy economics course.

If you want to look at economic data and say things, that is cool, do so physicists – I actually hold hope that you dudes and dudettes will add heaps to our understanding.  But before saying “look at this 100 level result, economists haven’t thought of it and are morons” actually read what economists have f’ing said in the past.  Arrogant sons of guns – I am utterly furious, and this was defacto published by the AAAS as it is their mag.  FFS!

  • Kiwi Poll Guy

    Agree on 2), disagree on 1).

    The lack of a mechanism doesn’t mean there is no explanatory power. There is no reason (in theory) why it isn’t possible to derive the shape of the distribution from first principles, so I think the method is legit. Perhaps there is some sort of nuanced difference between what physicists and economists think of as an explanation?

    Whether or not it works (it doesn’t really) and whether or not it’s already been done (it sounds like it has) are separate issues.

    • http://tvhe.co.nz/ Matt Nolan

      My thought process runs description -> explanation -> understanding.

      Description outlines the data. Explanation tells us whether something drives something else. Understanding tells us why something drives something else.

      Mechanism may have been too strong a term – I think you’ve got a good point there. If their argument is that inequality could occur purely due to chance, then that is a potential explanation – and I suppose my irritation has more to do with them acting like it is “the” explanation :)

      • Kiwi Poll Guy

        Fair enough.

        If the distribution predicted by entropy maximisation describes the true distribution well (it doesn’t) then I would argue the entropy maximisation “explains” the distribution. Entropy maximisation *is* the mechanism.

        But otherwise agree.

        • http://tvhe.co.nz/ Matt Nolan

          Interesting – even if it was shown that entropy maximisation describes the observed distribution, I still find it hard to ignore seemingly relevant covariates, or even more importantly the inherent heterogeniety in the characteristics of individuals in the sample.

          The kicker here is two-fold:

          1) Measurement problems and the such make it a lot more tenuous to compare even a gigantic sample to the “true” distribution. Something I think economists take into account implicitly when holding back on inference from data.
          2) Even if the true distribution can be seemingly described by entropy maximisation, I can’t make the leap to prediction due to the Lucas Critique – with agents making active choices, the underlying parameters of the model won’t be invariant to policy. This part of the “mechanism”, stemming from individual choice, also requires modeling. In this case, the process involved would be random chance – and the policy would be redistribution of income based on it solely being chance. So even if we were to accept this extreme assumption (which is patently too extreme), we are only part of the way towards policy design.

          This is how I understand it, but if you have any feedback where you think I’m confused I’d be happy to hear it! Methodology and language, especially when it starts crossing disciplines, is important to discuss.

  • Peter Cresswell

    Relevant Twitter comment this morning from https://twitter.com/FriedrichHayek :
    “You can math model almost anything in Econ … everything except the central causal explanatory elements of the science.”

    • http://tvhe.co.nz/ Matt Nolan

      Math is a useful language to help make our assumptions clear – but the assumptions need to be built on something. I read that as what you are pointing at here, and agree.

  • Pingback: Can physicists please look at a basic textbook before releasing these things

  • Blair
    • http://tvhe.co.nz/ Matt Nolan

      Always tempted to link that. However, this guy hasn’t just met the subject – he just loves applying basic physics and telling everyone that econs never have. A strange critique given how much I’m told by other disciplines that econ over-relies on classical physics techniques!

      Most of the time, both sides haven’t actually checked to see what econs actually do …

  • lalitha

    Hi Nice Post ,

    Love the ideas shared here. The list really huge and it will cover almost everything for any things. Thanks

    for the such a lovely post! Again thank you for sharing .

  • Pingback: Can physicists please look at a basic textbook before releasing these things | The Dismal Science