Complexity and decline
- Thread starter Voidwalkin
- Start date
Economics is an interesting example. The complexity of economics is that it is many different emergent systems interacting with each other. Economists have a very good understanding of different facets of this, such as oligopoly, monetary policy, taxation, externalities, asset pricing, and so on. Very robust models of each of these phenomena have been developed over the years. But every model focuses on a subset of these things to interact with each other and ignores the rest, because it would become intractable otherwise.
The real challenge of Economics research is knowing what frictions are important in your setting and what frictions aren't as important.
What does robust mean here, if not "valid in the real world"?
Voidwalkin
Emperor
- Joined
- Jun 12, 2024
- Messages
- 1,669
I agree with all of that and I fear its consequences.
1. Numerous bubbles as a result
2. Average person does not understand these models, leading to imbalance in evaluation. Or in other words: the better positioned wealthy and well educated are able to pursue their self interest without much if any meaningful pushback.
...which leads to political instability, apathy and nihilism is an age where climate change and nuclear war are potential existential threats. The structures can't comprehend the challenge, and there's so much room for bad faith actors to sabotage responses.
How interesting. I hadn't heard of that, but it does very well summarize the concerns I have. Stresses compound as different systems interact and lead to extremely unpredictable outcomes.
Monetary policy models are very valid for explaining inflation movements (and to a lesser extent aggregate employment) in the long run. They are very poor at explaining other knock on effects of monetary policy, ie distributional consequences, industrial consolidation and oligopoly, asset pricing implications.
At the same time, we have good models for income distribution dynamics (HANK models and their derivatives), asset pricing for options (Black-Scholes and many descendants using similar mathematical tools), and monopolistic competition (Dynamic Cournot competition with investment). But none of those can share assumptions with the monetary policy models without breaking. Yet each of them explains very well the phenomena they were designed to study.
Funny you mention nuclear war, the entire field of game theory was developed in large part for the government to be able to reason about nuclear deterrence in the 50s.
Which ones? If any had real predictive power in the long term I would like to see evidence of that, I did not think such a thing could be said of any economic model.
Rational Expectations models developed by Tom Sargent, Neil Wallace, and others in the 1970s have accurately predicted the long run effect of long-term monetary policies in western economies. You might be thinking of short term effects which are poorly understood due to all of the confounding factors I mentioned earlier. But the phrase "inflation is always and everywhere a monetary phenomenon" is completely true in the long run at least.
Before I jump too far in, can I confirm say this (from Rational Expectations and the Dynamics of Hyperinflation by Thomas J. Sargent and Neil Wallace) is what you are describing as "robust":
Rational Expectations is a concept best described in layman's terms as "Every person in this model behaves in a way to maximize their own wellbeing given what they know and the state of the world". It isn't tied intrinsically to that reduced form regression it looks like they are doing there. For context, this is looking at events that happened 50 years or more before the paper was written (and about a century ago today!) and where data quality is very poor.
This is important in the monetary policy context because it forecloses many explanations of inflation that go something like this "Inflation is caused by businesses realizing that they can get more profit by raising prices to screw consumers" because that would imply that the businesses were behaving stupidly yesterday, and simply not making a profit when they could have. The price raising is a symptom not a cause.
I will add that my research focus is not on monetary theory or macro finance so I may not be able to go into as much depth as you would like on the topic. I brought it up as an example where economics has a good understanding of a specific interaction between theory and practice.
This is important in the monetary policy context because it forecloses many explanations of inflation that go something like this "Inflation is caused by businesses realizing that they can get more profit by raising prices to screw consumers" because that would imply that the businesses were behaving stupidly yesterday, and simply not making a profit when they could have. The price raising is a symptom not a cause.
I will add that my research focus is not on monetary theory or macro finance so I may not be able to go into as much depth as you would like on the topic. I brought it up as an example where economics has a good understanding of a specific interaction between theory and practice.
Robust is a very specific term in maths, and I cannot see anything like that in that paper. I can see things I could question in that paper, for example I cannot see where the multiple testing correction is to get to those significant levels. Why they are quoting the F values rather than the converted P values I do not know, I am sure it is principally done as intentional obsvercation.
Do you have an example paper demonstrating robustness in the face of real data?
Look for papers about the "Taylor Rule". It is pretty old and known to be incomplete in some ways but broadly explains pre-ZLB monetary policy and how it drove inflation dynamics during the Great Moderation.
I did not ask for something that explained something, I asked for something that demonstrated robustness. You made the claim.
I'm not sure what mathematical definition of robustness you are demanding. I made the claim of robustness using that word in the common english meaning of working in many different times and different countries to conduct inflation targeting.
We may be talking about the same thing, it is usually defined something like (from wiki) "Robust statistics are statistics that maintain their properties even if the underlying distributional assumptions are incorrect".
"Working" in economics is a difficult question. If you are saying it is "robust" because most of the time when they increase the interest rate inflation goes down that is a very different thing to what I understood you to mean. If you can point to the paper you are referring to I would read it though, I have read a few economics papers recently and they all seem to have pretty bad maths. I would like to see one that was good.
It is more than that, it also has a rough idea of the slope of that relationship. Enough at least for central banks to calibrate their responses much quicker than in the 1960s and 1970s.
Linky link?
Narz
keeping it real
People think they want to have things & experiences but mostly people want to create things & experiences. But they can less & less so they watch netflix.
Narz
keeping it real
Pretty much, even the body works like that, you build it up over decades & then get cancer and are gone within a few years (or a few months), obviously violent death is faster still
PhroX
Deity
- Joined
- Jan 12, 2009
- Messages
- 2,801
This post and your others in the discussion have reminded me of something I've been wondering about for a while regarding economic models but never actually got round to finding out, so I'm intersted in your thoughts.
I'm a hydraulic modeller by trade (i.e. I model the behviour of water), and when I hear people talking about economic models they seem to be...well, not even really models how I would consider them. They're more like equations or rules. Is this right?
By comparison, I'll build a model of a river valley to investigate flooding. This will involve input data about the valley such as the topography, the bathymetry of the river, surveys or construction drawings of structures in the river, soil conditions, sometimes populations etc, as well as the impact of factors from outside the valley, such as the flow in any rivers that enter form outside my model area. All this will be input into dedicated computer software. The software then divides the valley up into hundreds of thousands if not millions of points, and solves fundamental hydraulic equations at each of them at regular time interavals for hours, days, sometimes even weeks of simulated time (this usually takes a few hours to solve, but some models can be several days).
Once we've done this, we then have to calibrate the model. This involves taking data about known storm events (e.g. rainfall), inputting them into the model, and seeing if the model predicts what actually happened (e.g. does the predicted flood areas match where flooding was reported?). If it doesn't, we then have to start adjusting out base model parameters - we have to make some assumptions and simplifications as we never have perfect data - for example, maybe we underestimated the friction in part of the river so water is flowing quicker there than in reality.
Only when the model has been calibrated against real events would we consider it complete. If the model can't give a pretty good recreation (it's not going to be exact as we never have perfect info) of what we know has happened, it's basically of no use as a model.
When we have such a model, we can use it to predict what will happen in various future scenarios, for example, what would increased rainfall from climate change do to flooding, or what wouldbe the impact of a new housing development, or what kind of flood defences should be installed.
Is this kind of detailed large scale computer simulation used in economics?
I find myself in a similar situation. I build models with stats, parameterise them with real data and demonstrate their predictive power numerically. I have not found a high profile economic paper that really seems to take the scientific and/or mathematical method seriously.
Similar threads
