Correcting The Wall Street Journal

A chart on the front page of the January 31 issue of The Wall Street Journal was informative, but not correct.*

There are a lot of ups and downs within individual business cycles. It makes sense to smooth them. That’s what The WSJ did, but it looks like they used Excel to make a bar chart. The problem with this is that it accentuated the length of recessions, both in absolute and relative terms.

Take a look at that. Do we really think the Great Recession was longer than the expansion that’s followed it? No, it was about a third the size. It gets worse: the 2 quarter recession in 1980 is shown to be longer than the only recessions that compete with the Great Recession for the title of worst since World War II (1973-5, and 1981-2). Or how about the halcyon days of the 1950’s; were they really  dominated by 3 recessions that were longer than the intervening expansions? In sum, the spirit that inspired this chart is on the right track, but the execution is poor.

The solution to this is to do an XY chart and connect the dots. For the chart below, I subdivided the data around NBER peaks and troughs. I then calculated the annualized geometric average† growth rate for each expansion and contraction.

The advantage of this presentation is it shows accurately 1) the length of expansions and contractions going horizontally, 2) the average strength of expansions and contractions going vertically, and further the 3) total impact of an entire expansion or contraction is proportional to the area between the blue line segments and the 0% gridline.

A disadvantage of this sort of presentation is that it does a bad job with the relative size of double-dip recessions (as in 2001) or triple-dip ones (as in 1981-2): the upward spikes within the recessions make their overall growth look OK.

Keep in mind that the threshold for what feels good is around 2% rather than at 0%. This is why Obama’s expansion is so poorly regarded: it’s barely beating the threshold. And there certainly seems to be a pattern of declining average growth in expansions over the last generation (although 3 expansions is not a big sample). Here's the same chart, with the contractions removed, and the threshold for "good times" set at 2% rather than 0%:

Many thanks to Jon Peltier of Peltier Tech Blog for help getting the charts to look right.
* Read the whole thing, entitled “US Economy Shows Signs of Gearing Up”.

† The geometric average is the rate which, when compounded, would produce the observed growth over the entire period.