Read of the Day: Beware Economists Peddling Elegant Models

Good thoughts here from theoretical physicist Mark Buchanan.  In a Bloomberg piece he warns us to beware of economists peddling elegant models:

“If economists jettisoned elegance and got to work developing more realistic models, we might gain a better understanding of how crises happen, and learn how to anticipate similarly unstable episodes in the future. The theories won’t be pretty, and probably won’t show off any clever mathematics. But we ought to prefer ugly realism to beautiful fantasy.”

Realism, realistic models and a better understanding of our problems using a purely descriptive version of the money system?  Sounds like Buchanan is referring to Monetary Realism.  Even better, none of its founders are economists.  :-)

Read the full thing here.

Cullen Roche

Mr. Roche is the Founder of Orcam Financial Group, LLC. Orcam is a financial services firm offering research, private advisory, institutional consulting and educational services.

More Posts - Website

Follow Me:


  1. Great article! Thanks for posting. I’ve often wondered about the existence of economic models which include a multitude of non-clairvoyant agents instead of one “rational agent” to represent the whole of one large sector of the economy. The author makes a lot of good points.

  2. This is the problem with Economics, it relies on what people SHOULD do when it comes to their economic interests. Sadly in the real world, people often do what they SHOULDN’T do in those situations. Economists need to adjust to get more behavioral in nature. Their are plenty of past events they can analyze and see where people do the opposite of what is in their best interest.

  3. What differntiates the science of economics from something like physics is that people are sentient agents, that is they are self-aware and make decisions based on a myriad of influences, besides “rational” thought processes. A human agent may act differently under the same exact circumstances from event to event. With that in mind, human behaviour is not necessarily “deterministic”….as an electron, or a gas particle would be….It’s this variable that makes mathematical modelling of economic/human behaviour so challenging.

    • I’m not sure that’s true. Certainly its very very difficult to model, and we certainly aren’t there yet. Plus there’s always the problem w/ economic models in general that any good model will create a feedback loop, thus changing the system its modeling (unlike models of other complex systems such as the weather).

      But regarding the human brain… I think we will be able to simulate it pretty well one day.

      • I’m skeptical of the notion that some day we will be able to accurately model irrationality.

        • I think human irrationality will ultimately be a lot easier to model than you think. Check out the work of Dan Ariely: a guy who does behavioral or experimental economics exploring irrational behavior. Dan has found the power of the “default” position is tremendous (for example, default entries on forms). Also he finds that humans don’t like to have too many choices. He did one experiment with physicians that was interesting. They seemed to act rationally with two possible choices of diagnosis, but adding in a third made most of them do the irrational (default) thing. It was pretty scary actually! I don’t recall the details, but Google him, and you’ll find it.

          The ant brain (and perhaps other insect brains) have already been modeled. Here’s an example of a bigger brain:

          Insect brains have been modeled. Here’s an article about larger animal brain models:

          I think the ant was modeled completely … all 100 neurons (or whatever it is they have).

          Personally I think human irrationality is more a product of a LACK of computing resources, and evolved ad-hoc ways of dealing with that.

          • So for example (on a micro level, suppose my neighbor and I are both in the market for a new car. My neighbor and I both do our homework (independently), and decide the Ford Focus is the best car in terms of value, mileage, resale value, maintenance, etc….a rational decision made independently by both of us.

            But, then he decides to buy his one week earlier than me. I (seeing my neighbor’s new Ford Focus), decide that I don’t want to copy my neighbor, and want to express my individuality by buying the more expensive Toyota Yaris…(a less rational decision, based on another’s decision).

            Are you saying this type of behavior can be mathematicaly expressed ? Suppose, my neighbor buys the Ford Focus, and I buy the Ford Focus (not because of any rational calculation on my part), but because I idolize my neighbor, and want to be just like him !

            Is this something that can be mathematically modelled ?… especially when extrapolated from the “individual” to the “aggreagate”.

            • That might be a tough one: I’ll absolutely agree with you there! However, my point is that ultimately we’re machines, just like everything else anybody has used the scientific method to look into. I’m a materialist, so that’s an easy one for me. If it’s made out of star dust, then it should follow the physical laws that everything else made out of stardust does! The brute force way to do it is to simulate a bunch of brains in a simulated world. We’re NOT EVEN CLOSE to doing that, but in principal I don’t see a problem… except perhaps ethically (because I think if we accurately simulate human brains… then those ARE humans we’d be toying with!).

              So the key is to make justifiable simplifying assumptions. Perhaps there’s a way to look at the example you give that could be captured by a simplifying assumption. That article Cullen links to talks about a simulation with millions of independent agents. A numerical example to be sure! But each one (I’d assume) perhaps “blessed” with a somewhat randomized initial condition and “internal behavior model.” I don’t know! I’m just imagining how that might be done.

              So if the effect you describe is important to model, perhaps the essence of it could be captured with a FAR simpler model than the entire brain. I assume ahead of time that a pseudo-random number generator will be part of the simulation (as in all good monte-carlo simulations). Justifying the simplified model based on experiments with human subjects should be an integral part of the process.

              So, perhaps it boils down to an “individuality” desire parameter of the modeled agents (people). Perhaps some of the elements are randomly assigned a larger built in desire for individuality and it only comes into play when other actors are “in proximity” (i.e. neighbors). Then this parameter might express the chance that an irrational decision is made about a purchase based on the desire for individuality. The die still has to be cast to see if this irrational decision is ultimately made.

              Of course I’m just speculating about all this! But I can see some similarities w/ how particle filters work (in the Engineering world). Again, elegance is traded for a better fit to reality at the cost of having to use more computing power and ultimately making the problem intractable except through massive numerical simulation. But the reality is that particle filters are useful in some non-Gaussian circumstances because they do exactly that… the old standby, Kalman filter solution may be “elegant” and allow you to perform an elegant “covariance analysis” w/o having to do monte-carlo runs, but it’s assumptions may be so divergent from non-Gaussian, non-linear reality that the results it produces are ultimately useless.

  4. Interesting. I made a similar point recently, that we can best approximate the economy with a huge set of coupled differential equations (which can only be “solved” with numerical simulation). Compared with areas where this approach is extremely useful (the example used of modeling airflow over airplane wings is perfect) economics has a human, psychological, component.

    The recent housing related downturn is a classic example. I recently calculated an estimate of real per capita mortgage cost using FRED data. Taking the real mortgage debt outstanding times the 30 year mtg rate, divided by real GDP and population, at the peak of the housing boom the level was below what it had been from 1978 to 1994, and at the peak a bit more than half of the all time peak reached around 1982. There are lot’s of other factors, but by this measure (the mean real cost [or flow] of housing debt) was not the root cause.

    An important insight from MR and the coupled differential equation approach is that without regulatory oversight bubbles are inevitable because of positive feedback. As Cullen correctly points out, bank are not reserve constrained, they are capital constrained. As long as the collateral backing a loan remains stable or rises in price banks have an incentive to create money through loans. Given that more than half of household spending is on things other than food, shelter, and clothing (BLS data) this makes for an inherently unstable system. Government spending on real productivity provides stability (dampens the inherent instability in the private sector). It is almost certainly true that in many areas Government spending is inherently less productive than the private sector (take for example trash collection) but there are areas where this is not the case (take basic scientific research), but it is always stabilizing.

    • “As Cullen correctly points out, bank are not reserve constrained, they are capital constrained.”

      I’ev heard this notion “capital constrained” many times over the years but can’t remember it ever being fleshed out, at least not for the lay-people. EXACTLY what does it mean for a bank to be capital constrained?

      Let’s say there’s a small local bank in a small down, and multi billionaire Mr. Brown shows up with his billions in financial assets, and he attempts to obtain a loan from the small bank for $300 million. Can this small town bank just create that loan because he’s credit worthy? Is the small town bank “capital constrained” in some way?

      Cullen, any thoughts?

      • I can’t really address your example, but I can point you to a simple example of capital constraints that I put together:

        … variation on a theme:

        (Also example 3.1 is a more complicated version of 3.2 in you’re interested)

        This last one I’m least confident about, but here it is anyway:

        Hope that helps!

        • And BTW, I’m certainly a lay person myself! I recall having the same question as you. Joe in Accounting (commenter here on pragcap) was a big help to me in understanding what I do. I put together those balance sheet examples because that’s the best way for me to learn that stuff. What I’ve addressed here is “regulatory capital” (my last link above deals with “regulatory capital” vs “accounting capital” as I understood the concepts from Joe). Joe has brought up before another concept called CAMELS which is an acronym for a rating which can be applied to banks to judge their overall solvency health I think. I believe that’s what a regulator might be interested in.

      • So to address your example, if the bank’s CAR is > 10% they could do it. Since the guy taking the loan is a multibillionaire, it’s likely that the denominator of the CAR would be weighted by a relatively small weight as he could probably offer some very substantial collateral for the loan. The smaller the weight in the denominator of the CAR the easier it is for the CAR to exceed the 10% threshold.

        Other than that, they might already have “retainied earnings” or other capital, or they could raise some by charging Mr. Billionaire a loan origination fee (assuming it’s not so high that he’d take his business elsewhere).

  5. John….excellent points: “An important insight from MR and the coupled differential equation approach is that without regulatory oversight bubbles are inevitable because of positive feedback….”

    So maybe what you’re describing is like “forced resonance”…with the Tacoma Narrows Bridge failure as an example of that:

  6. Makes me think of how Volcker once said Ray Dalio models in a completely different way than most economists — and that though it is inaccurate, it provides a better yardstick for gauging the economy and the financial markets.