The inability to predict the financial crisis has raised a debate on an important toolkit of economists: economic models. How reliable and useful are they? To what extent can policy makers rely on model analyses in forming policies? And to what extent, can they be used, for example, for science policy to ensure most effective allocation of limited funding resources?
An economic model is a mathematical representation of the economy. There are many models differing in specific assumptions and the specific purpose of the model use. Different types of models are called for when, for example, making short-term forecasts or analysing the long-term consequences of ageing.
Economics deals with complex interdependencies and interactions between numerous decision makers. A model is a way of keeping track of these. Making these elements explicit has the advantage of ensuring consistency in the assumptions made, and it enforces discipline. This is important in its own right. But it also makes it possible for outside observers to assess the ingredients build into a given model.
A primary purpose of a model analysis is to gain insights and quantification of the likes of macroeconomic development but also the effects of more specific interventions like tax reforms or R&D funding. Insights are obtained by assessing the role of various assumptions made for the outcome. Model builders spend much time on such exercises to understand their tool-kit. Quantification is essential to assess the impacts and consequences of policy changes.
Empirical validation of models is essential. Are the specific structures and assumptions made consistent with available empirical evidence? This is an ongoing process within the profession. Evidence is accumulated, theories are tested, and models are reformulated.
Leaving aside the thorny question of statistical issues in model validation, one crucial caveat should be noted. Empirical validation is inevitably backward-looking depending on historical data. This is important information, but it misses new events like a financial crisis. This is why theoretical modelling is important to explore possible events which have not been observed historically. This is an ongoing process within the profession with progresses, but also shortcomings.
A case in point is the financial crisis. Mainstream models of the business cycle neglected financial factors, not that they were unimportant, but they were seen as an add-on not in itself a source problems and business cycles. The financial crisis has induced intensive research activity trying to resurrect the role of financial factors.
The outcomes of model analyses are inherently uncertain. Model builders and users are very well aware of such limitation. When the media report that model forecasts for, say , output growth next year are 2 %, the model analysis may say that with 95% certainty the growth rate will be between say 1.5% and 2.5 %. The best point estimate is 2%, but it is uncertain. This kind of uncertainty is difficult to communicate. And the media abstain from doing so, demanding clear-cut and simple messages. In this sense, model outcomes are often misused or over-interpreted.
Policy makers often find that model analyses of policy proposals are a straightjacket. But that is precisely their purpose. Policies should be based on careful assessments and evaluation. And not just beliefs. This is not implying that models are perfect – they are not. They must constantly be up-dated and reformulated to capture the ongoing changes in society. In that sense a good model is a moving target.
Torben Andersen, professor economics at the department of economics and business, Aarhus University, Denmark.
This is a post sponsored by ESOF 2014. The role of economic modelling will be discussed during a session entitled ‘Fiscal austerity and growth: what does science say?’ at the ESOF 2014 conference, due to be held in between 21st and 26th June in Copenhagen, Denmark.
Featured image credit: CC BY-SA 2.0 by Ten Keegardin
Go back to the Special Issue: Alternative research funding
Go back to the Special Issue: ESOF 2014
EuroScientist journal 
Torben highlights an important problem with models, which I’m persuaded need not necessarily be an inherent problem of the models themselves, when he writes: ‘[…] uncertainty is difficult to communicate. And the media abstain from doing so, demanding clear-cut and simple messages. In this sense, model outcomes are often misused or over-interpreted.’
The ‘media’ can be a distorting lens through which a clearer understanding and appreciation of life’s inherent uncertainties can be ‘airbrushed’ away by over-simplifying complex or grey-area issues. This has the unfortunate effect of creating unrealistic public expectation and subsequent disillusionment with our experts and institutions, depriving society of the chance to be intelligently informed and engaged with issues that may affect us all. Sadly, politicians are often complicit by demanding over-simplification in order to sell an idea to an uninformed public, perpetuating a poor comprehension of important issues. But academics too can be guilty of complacency, forgetting that their models are only a contingent representation – a description, not the described – and granting the models undue authority, not the required humility that Torben recognises.
However, another significant problem for models has been built upon the assumption in economics concerning rational choice. The rational mind’s need for certainty and predictability, its desire to make sense of the unknown and to feel in control is part of its nature but conceals its own potential for self-deception. Neither is it necessarily our default position. It is increasingly apparent that the central tenet of economic modelling revolving around ‘the choice of rational agents’ is a fallacy. It is worth noting that despite the academic persistence of this fallacy, businesses, when advertising, have almost exclusively focused on ’emotional’ appeal or persuasion to sell their products, not our capacity for rational choice – and so it is most often the motivation for our consumption, which models may then try to measure and make sense of.