Hector Pollitt, Head of Modelling of Cambridge Econometrtics, explores: what is economic modelling anyway and why do we do it?
In the world of policy analysis models aim to represent the societies and economies that we live in, all held within a computer system. We do modelling to help understand what will happen to the economy under different policies, and what this means for things that matter, like jobs and inequality.
Different models have different representations of the economy and all macroeconomic models have their strengths and weaknesses – for sure some are better than others.
The importance of macroeconomic modelling
For better or worse, macroeconomic modelling is playing an ever-larger role in policy assessments. Just to be clear, we say macroeconomic modelling as opposed to ‘economic modelling’ in order to make the distinction between what we do with microeconomic analysis, which considers interactions at the level of the individual household or company.
The EU’s Better Regulation Guidelines make macroeconomic modelling an almost mandatory part of assessing the impacts of a new policy. In the rush to meet policy makers’ deadlines, it is all too easy, however, to lose sight of the question – why do we do it?
To answer that question, first let’s consider the sorts of questions that are modelled.
“At European level, any new policies may impact on 500 million citizens; it is only reasonable to ask what the effects might be in advance. But is modelling the only way to answer that question?
It is not – a comprehensive policy assessment will include modelling as just one of a combination of qualitative and quantitative techniques.
In an ideal world, these techniques would include laboratory experiments but, while such experiments may be sometimes possible at micro level, they cannot be conducted at macro level (and, even if they could, might be subject to substantial ethical issues). Modelling thus provides a substitute for this type of experiment.”
How can macroeconomic models be used?
Broadly speaking, modelling can be used for three purposes:
- To understand the past and present
- To predict (forecast) the future
- To test alternative futures
In all three cases the models aim to replicate a process of testing similar to that in a laboratory, in which a single input stimulus is changed at a time and the response to that input is tested. In the case of forecasting the future, however, the extremely large number of things that must be controlled for means the process becomes more like an artform, rather than a science-based approach.
In all cases, however, the validity of the experiment depends on how well the model can provide a representation of reality. To be clear, all models are simplifications of reality, otherwise they would be as complex as reality itself. In some cases a very simple model is desirable – if we strip out all the unnecessary stuff, then the bits that remain are a lot easier to understand.
Where models go wrong: missing the financial crisis and rational behaviour
So far, so good. But there are countless cases where modelling has gone badly wrong, and for lots of different reasons. It doesn’t take long to find strong critiques of macroeconomic modelling online, including several from highly respected economists. Two common reasons for problems…
First, oversimplification. All the forecasting models that excluded the financial sector and the build-up of private debts missed the financial crisis. That is, virtually all of them. The model builders decided that the financial sector was unnecessary and missed a crucial aspect of what they should have been analysing.
Second is the application of simplifying assumptions that run counter to reality. Textbook economics provides a long list of these (mostly including the words ‘perfect’ and ‘rational’). If they were explained individually to a non-economist, each would be rejected almost immediately. However, they all typically appear in standard Computable General Equilibrium (CGE) and Dynamic Stochastic General Equilibrium (DSGE) models, not least because the models often cannot be solved without them.
Trading off theory and empirical content
This is where economists disagree about what should and should not be included in a model. Neoclassical and New Keynesian economists accept the assumptions from textbook economics without question and incorporate them into their CGE and DSGE models.
In contrast, economists from other schools of thought strongly question these assumptions and have built models that adopt a more empirical approach to economic analysis. Cambridge Econometrics’ E3ME model is an example of such a tool.
So, to answer the questions in the title of this blog post, in the world of policy analysis macroeconomic models aim to represent the societies and economies that we live in within a computer system.
Most important is whether the model includes (in a reasonable way) all the necessary components to assess the policy that is being tested. If it does, then modelling can provide powerful insights to support a policy analysis. If it does not, then it could be a dangerous way of leading us in the wrong direction.
Cambridge Econometrics is a partner of the MONROE project. For the original blogpost of Hector Pollitt see: https://www.camecon.com/blog/what-is-macroeconomic-modelling-and-why-do-we-do-it/