A Little Disruption, Big Economic Shocks
Sometimes a piece of macroeconomics research is so cool I just have to write about it. A new working paper by Daron Acemoglu, Ufuk Akcigit, and William Kerr is one of those papers.
Time and again, you have heard the critics of macroeconomics say that the economy is too complex to model with math -- especially the straightforward kind of math used in most of macro. Most macro models lump all consumers into one giant superconsumer, all companies into one giant supercompany and so forth. Even the models that differentiate people and companies tend to do so only in a limited way -- limiting the economy to two types of people, for example, or having people differ only in their level of wealth.
This frustrates a lot of people. The economy isn’t just some simple machine that takes in labor and capital and spits out dollars of gross domestic product, right? It’s a complex web of business relationships, inputs and outputs, expectations and interactions. Blogger Arnold Kling, for example, has criticized the idea of the economy as a “GDP factory,” and urged us to think about “patterns of sustainable specialization and trade,” or PSST.
There are a couple of reasons we would really like a PSST or network-based macro model. First of all, it would solve the problem of what causes recessions. Currently, we have very little idea of what tips economies from boom over to bust -- there is usually no big obvious change in productivity, technology or government policy at the beginning of a recession. If the economy is a fragile complex system, it might only take a small shock to send the whole thing into convulsions. A second reason is that modeling the complex guts of the economy would give us a lot more data to look at -- instead of just thinking about aggregates like national consumption and investment, we could look at the pattern of how recessions spread from company to company or region to region.
The problem -- as any macroeconomist would tell you -- is that actually modeling PSST with math is really, really hard. But without the concrete language of mathematical theory, you’re left just waving your hands, relying on intuition or anecdotes, or (at best) subjective interpretations of empirical studies.
People have wanted to model the economy as a system of moving parts for a while now. John Long and Charles Plosser took a stab at it back in the 1980s. Their basic idea was to model the economy as a system of inputs and outputs, with linkages connecting the pieces -- kind of like a bunch of balls connected by sticks. Acemoglu et al. use a somewhat updated version of this model, adding linkages between different regions.
What the authors then do is to look at the pattern of how economic disturbances propagate throughout the industrial and regional network. They examine several types of disturbances such as changes in Chinese imports, government spending and productivity. Some of these shocks propagate upstream through the value chain, from retailers to suppliers. They call these demand shocks. Others move in the opposite direction, and they call these supply shocks
Their main conclusion is the result we would want from a model like this -- small disturbances to any of these things create big effects.
This is very different from the way many macroeconomists would like to think about the economy. Robert Lucas, considered by many to be the father of modern macroeconomics, wrote in 1977 that the movements of the various pieces of a complex economy should average out. It was shown much later that this wasn't necessarily the case, but not until Acemoglu et al. have we really seen concrete examples and hard numbers.
The implication is that the rosy picture of the economy as a smoothly functioning machine isn't necessarily an accurate one. The tinker-toy web of suppliers and customers and regional economies in Acemoglu et al.’s paper is a fragile thing, easily disturbed by the winds of randomness.
The model also has policy implications. One of the biggest and longest-lasting economic debates is whether government spending can affect the real economy. Lucas and others have claimed that it can’t. But in Acemoglu et al.’s model, it absolutely can, since the government is part -- a very big, very important part -- of the network of buyers and sellers.
So thanks to the hard work and insight of Acemoglu and others, the old dream of a network model of the economy is a little closer to reality. Someday we may draw maps of economic linkages the way we now draw circuit diagrams, and use supercomputers to simulate economic disturbances as they make their way through the web. We may look back on the simple pen-and-paper models of yesteryear and laugh.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To contact the author on this story:
Noah Smith at firstname.lastname@example.org
To contact the editor on this story:
James Greiff at email@example.com