In a conversation on Facebook, a friend recently raised an issue about the weaknesses of economic models. He was reacting to my Facebook post, in which I tried to explain why, at times, non-economists misunderstand language of economists. I used an example of the misunderstanding in the assumption that, ‘people are rational’, noting, “Non-economists think that rationality is acting in a reasonable way, for economists, rationality means behaving in the way a model would predict”. My friend responded that models were not reliable anymore, arguing, “the problem in rational choice theory models is that if you want to rely on them, they need to be incredibly complex. When models that are too simple fail to predict catastrophes as they do constantly, then you see the problem; another issue is that most models assumptions tend to be simply wrong outside a very specific set of circumstances”. He is not alone in arguing that models are not reliable. After the 2008 financial crisis, many have criticized the reliability of models since they failed to foresee the growing financial imbalances prior to the crisis. Others, argue that many macroeconomic models start from simplified, unrealistic assumptions over how households and businesses’ behavior depends on what will happen in the future, which often leads to some unusual and implausible results. One such unrealistic assumption that models make is that everyone is the same, or that, everyone can be characterized by a single, representative household or firm. This is not new. In the 1970’s, economic models quickly fell out of favor, because of the damaging methodological criticisms. In his 1976 critique, Lucas argued that the equations used in models were ill suited to evaluating changes in policy, since they were liable to change when the policies were altered. Another challenge came from the real-world development of stagflation, the 1970s occurrence of high inflation and unemployment. It was regarded as evidence against existing models, which almost exclusively featured negative Phillips curve relationships between inflation and unemployment. Is it now plausible, given the shortcomings of macroeconomic models to dismiss them as unnecessary and irrelevant? I would beg to differ! Models are literally abstractions. They simplify what is going on by removing unnecessary details. They help people to understand the potential effects of policies, to quantify and assess various mechanisms that might be at play and consider interactions beyond the direct or intended effects of the policy. For example, a central bank is tasked with ensuring price stability. In order to do this job well, models are needed to show how changes in policy affects inflation and the overall economy. Simply looking at what happened in the past when policy changes were made; to predict what will happen in future would not be helpful. For one thing, central banks typically increase interest rates when inflation is expected to rise above its target. Simply looking at the data would not disentangle the effect of the policy from the increase in inflation that was already expected. Often in macroeconomics there are many plausible stories as to why something has happened, or what might happen in future. Models therefore help to make sense of the data and give quantitative estimates of the size of different effects. The macro economy is a highly interconnected system where many aggregate variables depend on each other and models are required to help make sense of what is happening in the data, because relying on simple correlations can be misleading when trying to work out complicated causal logic. Models have failed on occasions, but we ought to look at models as tools that help us to understand the world around us rather than provide definitive answers. In this case, looking at models as maps would be helpful. Maps simplify our complex world to small-scale, flat figures. While travelling, a map can be used to show the route we need to follow, abstracting from loads of information that is not necessary to reach the destination. Although maps have become more accurate with improved data and technology, we are still surprised by unexpected delays, blocked roads or flooded paths. In the same way, economic models have improved with greater computing power, econometric techniques and data availability, but there is still significant uncertainty that cannot be eliminated, and this is why they cannot always provide definitive answers to the questions we ask. Despite the accuracy of maps, a person would not choose to follow a map if the terrain appeared different in reality: the dotted lines of a mountain footpath on an ordnance survey map could become a dangerous hazard following heavy snowfall. The same applies to our use of models. Policymaking should therefore be judgment based and models be used as an advisor to help us make decisions. Economics is less accurate and less predictable than cartography, but models remain important for decision-making. As the economic landscape changes, there will be need for models to change and adapt to the new realities. After failure of the Philips curve in the 1970’s, economists improved it to an augmented Philips curve we use today. Many macroeconomic models have been improved after the 2008 financial crisis, to incorporate the financial sector; economists should continue improving their models. Despite any improvements however, people should not expect models to be, tell all tools. We ought not to dismiss them as inaccurate, irrelevant and unnecessary tools, but instead look at them more as enablers of our decision-making. The writer is an Economist at the National Bank of Rwanda
