Economic systems are part of the physical universe, and should be amenable to a physical treatment. Prior work introduced a thermodynamically-based economic growth model that is fundamentally different from traditional approaches (Garrett, 2011). At its core is a result that the real fiscal value of global economic capital, where global capital subsumes labor and infrastructure into a whole, can be viewed as a representation of the total capacity of civilization to consume energy. Effectively, what sustains the purchasing power embodied in an inflation-adjusted 1990 dollar bill, and distinguishes it from a mere piece of paper, is a constant 9.7 +/- 0.3 milliwatts of primary energy consumption. Expressed on a global level, in 2008 the global economy was characterized by 1656 trillion 1990 dollars of economic worth that was supported by 16.4 terawatts of power.

The work here extends the Garrett (2011) model to provide a basis for making long-range economic forecasts with positive skill. Two key rates are introduced: the “expansion rate” represents the efficiency of converting energy consumption to an increase in the capacity to consume energy, or equivalently to convert wealth to GDP; the “innovation rate” represents an acceleration of the expansion rate.

Presently, global expansion rates are the highest they have ever been, at about 2.3% per year. But expansion rates have evolved dramatically over history. “Economic fronts” have passed when innovation has skyrocketed. Currently, innovation is at its lowest since at least since the 1930s. The all-time historical high for innovation was between 1950 and 1970 when rates ranged between two and four percent per year. This is important because past innovation plays the controlling role in current rates of global economic GDP growth. In fact, current innovation is very small, and the the largest contributor to today's energy consumption and GDP growth rates is innovation that took place during the 1950 to 1970 time period.

Three main forces control this evolution. One is a law of diminishing returns. A growing system is naturally inclined to grow at an ever slower rate because current consumption becomes progressively diluted in past consumption, and it adds ever less to an interface that drives flows from primary energy reservoirs to civilization. The second force is depletion of reserves, which dampens growth. Diminishing returns and depletion can be overcome only through the “discovery” of new available energy reserves. Expansion of energy reserves is fundamental to sustaining long-term growth.

This very simple theoretical model shows excellent consistency with multi-decadal observations of global innovation and expansion. A simple story is portrayed. High rates of energy reserve discovery that began just prior to 1950 catapulted civilization into an extremely innovative state that lasted for several decades. But, eventually innovation succumbed to a law of diminishing returns, to the point that innovation is currently near zero and GDP growth is currently sustained by past high innovation.

The model requires no reference to policy or people, yet it appears that positive forecast skill for the present global economic state might have been achieved as far back as the 1960s. While the model is fundamentally deterministic, at least insofar as a law of diminishing returns is concerned, rates of energy reserve expansion remain an unknown free parameter. Fundamentally discovery is a geological problem, and it is hard to rule out surprises.

So we are still left with forecast scenarios, even if they are thermodynamically constrained. For example, if civilization ever enters a long period where it ceases to uncover new reserves faster than they are consumed, then global GDP will start to decline rather rapidly. If this condition were to begin soon, the expected decline would be about 2.3% per year. On the other hand, long-term expansion of reserves can lead to reignited innovation and GDP growth. But reserve expansion will have to be quite fast, at a rate of at least 3.5% per year.