This post is excerpted from the policy primer, Common Cent$: A Citizen’s Survival Guide, Appendix B.
What’s Wrong with Economics?
It’s a question worthy of a graduate seminar, but I’ll try to illustrate the key issues using intuitive shorthand. First I offer a quote from a popular science lecture course from The Teaching Company.
…there has been, ever since the 17th century, a predilection for linear modeling in science. That is to say, a linear assumption is that a small change in the input leads to a small change in the output. A big change in the input leads to a big change in the output. The equations that scientists use have this linear character, because linear equations are much easier to solve than non-linear equations. Linear behavior is much easier to model than non-linear behavior. And within a broad range, those models work.
Another predilection that we have seen again and again is the assumption that natural phenomena are always striving to get to equilibrium. They want to settle down once and for all. The combination of linearity and equilibrium is very distinctively characteristic of modern science from the 17th century into the 20th century.*
Nowhere is this more true than in the science of economics. One of the central assumptions in general equilibrium (GE) models that are widely used in neoclassical macroeconomics today is that people are pretty much alike, or homogeneous in their preferences. Consumers, workers, savers and investors, for example, all want to maximize utility and profits. A second assumption is that these preferences remain fairly fixed over time and do not vary or adapt under different circumstances. Lastly, economic models assume the availability of requisite information and the ability of individuals to process that information accurately.
Using these three basic assumptions, economic theory employs higher-order mathematics to build very complex and powerful models that enable us to study and understand the economic world we live in. Straight (or convex curve) functions usually yield clean solutions, known in economics as optimal equilibrium solutions. A good example is the graph of supply and demand curves that intersect at the equilibrium price. However, we have found that the assumptions cited above are frequently violated by real people acting in the real world, with significant implications for the results of the models.
As argued in a recent book on the limitations of economic and financial models, the behavior of individuals determines value—and people change their minds.** The basic problem, according to the author, is that “in physics you’re playing against God, and He doesn’t change His laws very often. In finance, you’re playing against God’s creatures.” And God’s creatures use “their ephemeral opinions” to value assets. Moreover, most financial models “fail to reflect the complex reality of the world around them.” The most serious weaknesses we have identified result from information failures and the ways in which we rationally adapt to these failures.
To conceptualize the difficulties of economic modeling, imagine a box of dried spaghetti pasta. Each strand of spaghetti represents an agent in our model. The individual strands can bend slightly without breaking, but they are mostly rigid and straight as an arrow. They cannot be folded back on themselves or twisted into a pretzel. Now, imagine building a model of a house or some other structure, such as a globe, with these sticks. The straight lines will intersect at many different points where you can glue them together. Depending on how big your model or how small the lengths of noodles, you can create a pretty good representation of curved surfaces using only straight lines. The model looks reasonably good, we can see what the model is meant to represent, and the structure holds together well.
Now, imagine trying to build the same model with wet, cooked spaghetti. Impossible. The strands wiggle around like worms, twisting and turning and refusing to retain a fixed shape. The model collapses into a tangled, gooey mess. This crude comparison illustrates the difference between an economic model that is built with simultaneous straight line equations that yield optimal solutions—those based on assumptions of homogeneous fixed preferences that don’t change over time and that reflect perfect information—and the often sad reality of a messy world. Economic behavioralists have discovered that our preferences vary, they change over time as we receive feedback and our circumstances change, and often they are not fully informed because we lack the requisite information. People are not like dried spaghetti noodles, they are like wiggly, cooked noodles – unpredictable. They are plagued by the uncertainty of their environment and are dominated by loss aversion. This creates serious limitations for modeling specific economic puzzles, especially those that relate to distributional dynamics characterized by variable preferences, feedback, and adaptability. [This just uses precise scientific language that refers to how the pie gets divvied up and how we all change our minds about what we want and what we’re going to do to get it.] Examples include wealth and income inequality, the business cycle, networked market solutions, and information cascades. So, what are we to do?
First, let’s not throw out the baby with the bathwater. Neo-classical GE models based on higher-order mathematics are very useful and powerful tools for explaining a wide range of market phenomena. Computer technology now offers us another powerful tool with simulation modeling. The exploding processing power of computers allows us to build market models from the bottom up using individual “agents” that can behave according to any number of simple rules that can be used in combination to create complexity. Individual agents can be unique in their preferences and adapt in an instant to changing circumstances, so our model doesn’t need to be constrained by rigid, representative agents. We can simulate how these programmed agents interact, as in a market, and observe the results. In effect, we can build ourselves interactive economic worlds that mimic the real world and then observe how these models behave as we change parameters.
I believe these techniques will allow us to fill in the gaps of conventional economic theory and improve our understanding of market dynamics.*** This promising new branch of economics is called agent-based computational economics. It’s not as elegant or intimidating as general equilibrium models, but it can do things GE models cannot. It is a sign of progress that hopefully in fifty years we won’t still be grappling with the same epistemological problems and policy failures.
* Great Scientific Ideas That Changed the World, lecture course by Steven L. Goldman, Lehigh University. 2007. Lecture 34: Systems, Chaos, and Self-Organization.
** See Emanuel Derman, Models. Behaving. Badly.