Deflating the U.S. Decoupling Hypothesis: Fun With Selective Data

Fans of the U.S. economic decoupling hypothesis — the idea that current U.S. economic weakness won’t be as damaging worldwide because the rest of the world relies on the U.S. less — got backwards help today. U.S. markets fell only a little on Friday, but European markets tanked more than 5% today: Whoa, decoupling! More seriously, what we really had was evidence that U.S. troubles demonstrably worry market participants more than it apparently does pundits and Asian politicians.

Either way, it’s fun to look at the data. There are lots of angles, including cross-market correlations — but other markets tend to fall on U.S. crises, but not necessarily vice-versa, which doesn’t help decoupling devotees — and so on, but let’s take trade. Assuming all major markets can be usefully characterized as open, so trade matters, how has the percentage of global GDP accounted by the U.S. changed in recent years? A decreasing share would help the decoupling argument, and a flat (or increasing) share would refute it.

Here is a graph of the U.S. share of global GDP since 1999 (note: the data source for this figure and the following two is the USDA):


W00t! That seems to support the decoupling idea. After all, the percentage of global trade tied to the U.S. has seemingly been in free-fall since 1999, thus decoupling, non?

Well, there are at least two problems. First, the above figure is only since 1999, so it would help to take a longer view of the U.S.’s share of global GDP. Is the above decline a recent trend? Or something else? You asked, so here you go:


Now that changes things. While there has been a decline since 1999, that was preceded by a period from 1991 to 1999 during which the U.S.’s importance as a percentage of global GDP grew even more than it has since declined. Well, darn, that’s bad for decoupling. Then again, however, the current U.S. share of global GDP is on the lower end of things since 1969, so maybe it’s a teensy bit bad, right?

Maybe, but it would help to normalize the Y-axis a little. After all, the above figure has a minimum point which makes the percentage swings seem much larger than they are. If you take a proper Y-axis zero point of, well, zero, and still keep the time period back to 1969, then you get a more sane view of things with respect to U.S. global GDP share.


The result? Yes, there has been a slight recent decline in U.S. share of global GDP, but that was preceded by a slight increase, and all of that by … more wobbling around a fairly tight baseline. You could, in other words, make an awfully compelling argument that, when it comes to the importance of the U.S. to global GDP, darn little has changed — whatever the decoupling devotees might say.


  1. The X and Y axes are powerful instruments of obfuscation and they are often abused.
    Great work Paul!

  2. If I’m reading that second graphic correctly, US % of Global GDP is now just about exactly where it was when NAFTA was passed.
    Which just proves that correlation is not causation, but we can expect a WSJ editorial in the next few weeks about the need for “the next NAFTA” to return us to the pre-W heights.

  3. Where did you get this data from, Paul? I’m looking at the Global Vision fact sheet, and it seems to come to a different conclusion vis-a-vis US share of global GDP, no?

  4. While U.S. % of Global GDP is interesting data, I am not sure it can by itself provide much of an indication regarding decoupling, if any. Wouldn’t a decoupling thesis have to rely more on the % of non-US GDP that is tied to the US economy? For instance, what % of BRIC GDP is tied to US imports? I don’t have the answer, but I suspect it would be more informative as it would indicate whether the economic engines of these countries can continue steaming ahead if U.S. demand for their wares declines.

  5. Thomas Johnson says:

    That last graph is sort of ridiculous. Suppose we had a hypothetical economic indicator with a value of 5000. Over the past 35 years, it’s had a range of 4900-5100, and it’s been within 4950-5050 90% of the time. Now it hits 5500. On a graph with a Y-axis that ranges from 0 to 6000, we’d barely notice this. Does that mean that our new high is meaningless? Of course not–this is probably a 4 or 5-sigma event!
    Setting your Y-axis to have a minimum of 0 is not “normalizing” it.