A bad econ model is better than the alternative

Let us talk about this.


I will be using this model myself next week to talk about migration (opps post timing changed and that was already up), so lets have some fun talking about this.

Originally I was going to title this “why Twitter doesn’t understand the purpose of models” – but it might make it sound like I agree with this model when I don’t.  I do want to point out that I am not defending the tweeters complaining about LRAS being vertical – as they have betrayed a misunderstanding about what it represents.

It means “if prices were higher, we assume that in the long run the factor markets adjust so that all relative prices are unchanged” – in other words the long-run neutrality of money, or an assumption about the transition paths of shifts in the short-run curves.  The “cause” of the price change may stop that from happening, which is shown by shifting the curve – the appropriateness of this depends on the question asked but starting with a vertical LRAS is pretty innocuous by itself.  It is not saying that nothing changes LRAS, it is saying that changes in LRAS are exogenous (or that a change in the price level of everything by itself won’t change output) – if the price of all goods and services and the price paid to factors (eg wages) all doubled then why would the amount produced differ … starting with the assumption that it wouldn’t is not that wild.

Now, lets get to why this is bad modelling, but then why it is better than the alternative finger in the air.

The model

I can’t read this text book, so my assumptions are as follows.  The AD curve is a representation of a standard goods and money market equilibrium, the LRAS curve is all the points where the SRAS settles following a change in prices, where the SRAS curve is the choice of production and prices by firms for fixed factor input costs.

We can view the LRAS as some “aggregate production function” for an economy, and AD as aggregate expenditure which is linked to prices either due to the traditional reasons (liquidity preference, Pigou effects) or in the current treatment a monetary policy rule.  In the long run what we produce is determined by the countries ability to produce and this AD curve determines prices – but because factor prices (specifically wages) are slow to adjust there can be cycles.

The model is definitely not the greatest at explaining cycles – heterogeneity, economic frictions, expectations, multiple-equilibrium, dynamics of the model when including growth … these things are all sort of rammed in on the side.  But it is not that bad if we are trying to get an overarching idea of “what could have happened to explain what we observed”.

So they have a leftward shift in the LRAS and a leftward shift in AD.  Ignore the curves have 2008 put on it – that is quite bad, especially for LRAS as it does imply that the long-run path was determined following the shock, which is not true.  Pretend they say “2018”.  Here these shifts explain the observed facts – prices moved sideways and output (relative to trend) remained low.

In essence yes, these shifts MUST have happened to explain the data (as long as these concepts are useful in describing the long-run tendency of the economy).  It is a comparative static exercise – the data itself is only consistent with a situation where LRAS has shifted left and AD has shifted left.  However, this begs the questions “why” and “what does that actually mean”.

Why it is bad for the question at hand

Ok, so we want to explain the GFC.  What does this model tell us about the GFC.

It tells us that we need to explain why prices didn’t rise or fall but output relative to trend has stayed lower …

It achieves this by positing that the amount that can be produced by resources in the economy is lower.

Well why?  And how is policy associated with this?

The actual thing we are trying to explain, and where we would like some policy pointers, is exogenous in this model, and so other than describing what the data is, and the “types” of changes that must be consistent with it it will tell us nothing.

If we want to explain something, we then need to build a model of that … the AD-AS framework is not built for this.

What is the point of AD-AS

AD-AS comes from describing the circular flow, and trying to figure out what will happen to prices and output following a specific shock.  Given that, we can talk about the idea of stabilisation policy – which in itself is just another “AD shock”.  There are heaps of good criticisms of its effectiveness for this role, and whether it captures the true nature of such policies.  And it is good to debate those.  But that is what the model is intended for.

Now it is also useful for saying “here is another economic concept”.  For example, economic growth.  But would we really want a model where we just shift out the LRAS curve and say that is growth … because that is all we would do.  Here exogenous growth theory was an admission that we don’t understand what actually causes growth, but we can analyse the dynamics of economic indicators if growth was to occur.

Given that this doesn’t answer anything about how policy influences growth (which is a question people are interested in) the development of endogenous growth theory was useful.  Here the process of growth is determined by a process explained in the model.  Of course the key issue is that the endogenous theory doesn’t explain the data (on conditional convergence) as well as treating it as an external shock – indicating that growth theory needed more thought, which is what the entire field of growth economics does!

AD-AS is great for outlining what economic questions are, and giving a way of thinking about what a question refers to in terms of the SHOCKS it would involve to explain the data.  It is a teaching tool.  But it doesn’t actually tell us “this is the cause of the financial crisis” – and if the LRAS just shifts it doesn’t tell us anything about what this has to do with policy.

I teach AD-AS at 100 level, and I start the course by stating that we are trying to describe data from the Great Depression using the framework of the circular flow – I point out areas where the model doesn’t fit the data, the problems with the assumptions we make, and how introducing new assumptions (such as expectations) change our results.  This model acts as an introduction to thinking about economic aggregates and principles, not as a general theory of the economy.

The purpose of such a framework is to allow us to think critically about “why” something happened, and to in turn try to model that “why” and test it against data.  In that why there is nothing wrong with that diagram – instead the problem is with the conclusions they make when discussing it in later slides, which involve using conjecture that fits their prejudice to explain the shifts, rather than model and data which are necessary.

But the advantage of them using AD-AS to do this is that the bad assumptions they make, and the bad conclusions they get from them, are transparent.  Unlike a lot of discourse about the financial crisis, which involves people running themselves in verbal circles to try to fit whatever they believe, a model forces you to show your (assumptions) hand.

Hating on the entire model because their conclusions are poor is dumb – show you understand the basic model, and clearly articulate where their description falls down!

Why don’t you say where it falls down

Ok, well which bit do you have a problem with.  Take this:

Yup I can see why stating that the reduction in LRAS is just due to financial intermediation breaking down is inappropriate.  But the point is that they should be describing the process that led to this comparative static change in more detail, and with more potential causes – not that the graphical representation that starts that conversation was inherently wrong.

This is 100 level economics, people are just learning about the terms and issues and should be taught how to build a framework that allows for critical thinking – and which forces that thinking to be disciplined by data.  That is exactly what this graph CAN do, and how it should be taught … but just because our pet explanation for a single phenomenon isn’t the centrepiece of the conversation we don’t like it?  Is that it, or am I misreading people here?  😉