Is macroeconomics a science?


Scientific models must be falsifiable

Being able to falsify a result lies at the core of the scientific method. It must be possible to set up a test that could lead to a model being discarded. A good model should contain a single or simple set of highly explanatory equations. At a bare minimum the equations should be dimensionally self-consistent, i.e. they have the same dimensions or units on either side of the equals sign.

The physics-based model I have described here is based on a proposed constant relationship between energy consumption rates and a very general representation of total inflation-adjusted wealth. If this relationship does not hold to within observational uncertainty, then the model can be dismissed as being flat-out wrong and we can then go off to try to do something better. All we should really care about is figuring out how things work. No point in pursuing something false.

Of course, it does look like the relationship is constant, but the point is that the hypothesis could, in principle, be invalidated. Further evaluation of the model can be done by performing hindcasts, asking whether we predict the present with a deterministic model that is initialized at some point in the past. Again, in this case it appears we can: current global rates of energy consumption growth and GWP growth can be accurately predicted based on conditions observed in the 1950s, without appealing to any observations in the interim, with skill scores >90%.

Strangely, such rigorous evaluation appears alien to traditional macroeconomics. I can find no evidence of tests of whether the models can perform true hindcasts. This is probably because, in fact, no test could be devised that would definitively show whether or not these models actually describe anything fundamental about how the economy works. Even where they reproduce the past, they are tuned to the past with no obvious applicability to a future where the tuning constants might change.

At an even more basic level, traditional economic models use production functions that use totally nonsense units. Take for example the basic Cobb-Douglas production function used by economists as a starting point for relating economic production Y to labor L and capital K. The quantity A is a “total factor productivity” that has been thought to be related to innovation.

Here the parameters A and α are tuned to past data. There is nothing fundamental about the quantity α since it is just a number. It can have any value depending on the statistical fit, the country, or the period considered. Suppose α = 0.3. If A is just a number, labor has units of worker hours, and capital units of dollars, then Y has the absurd units of worker hours to the 0.3 power and dollars to the  0.7 power. Of course economic output should have units of dollars per time.

In economic studies, when the inelegant Cobb-Douglas function (or whatever is used as a replacement) doesn’t work well for whatever reason, the approach is not to ask whether or not something might be fundamentally wrong about the premise behind the fit, but rather to add ever more bells and whistles until once again a sufficient fit is obtained, totally independent of any consideration of dimensional self-consistency. For example, maybe a constant exponent α doesn’t provide a good fit unless A is allowed to change too according some equally complex function. Sometimes this function is attributed to government stimulus of R&D.

But to successfully achieve a fit, really A could be anything! With a sufficiently complex function one could fit the historical population of rodents under Wall Street to in such a manner that the Cobb-Douglas function once again reproduces timelines of Y.

Why such absurdity? Making things ever more mathematically complex does not make things more true, if anything less so. It feels akin to astrology, a highly complex, self-consistent model based on un-physical nonsense. Totally convincing to those who are looking to believe that the world has order and explanation, and that they alone have the years of training required to understand it, but completely lacking in any means for falsifiability.

Perhaps, this is too harsh -- everybody is trying their best -- but it looks like fluency in Latin in the Catholic Church, where established macro-economists need something sufficiently opaque in order to maintain their high-priesthood. More generously, economics is complicated and economists just don’t yet know yet how to describe it without such detailed fits; even in physics, similar fits are occasionally used to describe interactions of particles with turbulence, for example, simply because the underlying physics can be rather challenging.

But, in any situation, a useful first step to solving a seemingly complicated problem can be to step back to look at the larger whole. The physics model for the global economy that I have developed offers a simple, straightforward, and most importantly, falsifiable expression for economic production. Instead of the flexible complexity of the Cobb-Douglas production function or its cousins, the replacement is

where the link to the past is given by a growth rate, always adjusting economic production for inflation:

Here, a is the global rate of primary energy consumption (units Watts), η is a variable growth rate (units per time) and λ is a constant 7.1 Watts per $1000 inflation-adjusted to year 2005. The units work out, and there is nothing to tune. λ is either a constant or it isn’t. The rest of the parameters are measurable. Output has units of real dollars per time. There are no bells and whistles. It is a model that can be easily tested and discarded. Economic output is determined by the amount of energy consumed and a variable coefficient η that can be shown to be determined by the energy efficiency of converting energy consumption to work.

It might seem a bit dehumanizing to have an expression for the economy that doesn’t have a labor term that explicitly mentions people. It would be nice to think that human agency has a little more power at altering its collective future than an amoeba in a petrie dish. But maybe not. Hopefully, at least, there is some value to understanding the forces that control our future. It is only by understanding our current state that we can understand the future for us and our children. We will not escape this century’s ecological and economic dilemmas by furthering fairy tales as solutions.