From the Equation to the Algorithm: Is it time to update Economics?

One of the main sources of examples on the panel on Computational Social Science last week, with Edward Glaeser and David Lazer, was Economics. This was not entirely unexpected, and many of the examples that were discussed were genuinely interesting. But the novelty of most of the approaches relied almost exclusively on the ability to include more data into existing ways of conceptualizing economic problems. By focusing so much on the power of data, academics might be ignoring the full potential of a new era in computational power.

Economics has always had a complicated relationship with its being grouped together with ‘other’ Social Sciences. For much of the 20th century the discipline of Economics has tried to distance itself from its more qualitative cousins by developing an orthodoxy that relies heavily on rigorous mathematical and statistical foundations.

It is therefore not surprising that Economists feel that they are well placed to make the most of the recent explosion in available data and advanced analytical methods. But it appears that much of that focus has been on taking existing theory and attempting to fill in the missing pieces that were simply not technically feasible before.

One of the foundations of modern economics has been the equation-based model. Attempting to find the right balance of simplification and complexity to distill aggregated behavior in a useful way has been one of the primary challenges in both Micro and Macroeconomics. With the sudden increase in the quantity and quality of data available, Economists now find themselves in a position to build better, more complex models that can include more variables and be tested against more real word cases.

But this overlooks one fundamental area of potential disruption. In many ways, modern computational power has made the economic equation obsolete. Historical models were based on equations because of necessity. The equation was the only way to formally describe relationships that could then be tested against data. Without simplifying and aggregating individual human behavior there was no way to conceptualize it on the scale of entire economies.

This is no longer the case. A model no longer needs to be limited to a mathematical relationship. We are now capable of modelling problems and running simulations that can aggregate actions that span the full known spectrum of rationality and consumer preferences. One of the key breakthroughs in algorithms used in the technology industry, from Google’s PageRank to basic A/B testing, has been the idea that models that rely on chains of probabilistic outcomes can be even more powerful than formal, well-defined relationships.

We have already started to see the worlds of simulation and model disaggregation merge with Economics in other, related ways. Many major online role-playing games employ economists to manage their vast virtual economies. This has been a rich testing ground to experiment with and test long-standing economic principles from effects of monetary policy to the limits of rationality in markets approaching perfect information.

Maybe the time has come to rethink how we can define the fundamentals of Economics, based not on what has been possible in the past, but what will be possible in the future.